Time Series Measurement of Force Distribution in Ice Hockey Helmets

Time Series Measurement of Force Distribution in Ice Hockey

Helmets during Varying Impact Conditions

Ryan Andrew Ouckama

Department of Kinesiology and Physical Education, Faculty of Education,

McGill University

Montreal, Quebec, Canada

March 2013

A thesis submitted to McGill University in partial fulfillment of the requirements of

the degree of Doctor of Philosophy in Kinesiology and Physical Education

ABSTRACT

Modern sport helmets have been effective in reducing catastrophic head injuries such as skull fracture and subdural hematoma; yet, the high prevalence of minor traumatic brain injuries (mTBI) is an unresolved public health concern.

Consequently, there is a need for greater scrutiny in a helmet’s ability to mitigate collision forces that may correspond to mTBI risk. Current safety standards primarily assess a helmet’s ability to minimize the whole head’s peak acceleration during blunt impacts. Absent are dynamic measures local to the impact site itself due to the technical challenge to spatial map high impact force magnitudes with high temporal resolution. Inclusion of the latter measures may enhance the functional assessment of helmets. Thus, the aim of this research was to develop a localized impact mapping system (LIMS) for placement between the helmet and head interface and then to utilize the LIMS to evaluate the mechanical behaviour of various padding foams and helmets during controlled headform drop and projectile collision tests. Interposed between the helmet shell/padding and head surface, this LIMS consists of an array of discrete, thin force sensors connected to a compact signal conditioner and high speed data acquisition digital recorder. A first study demonstrated the feasibility of the LIMS to accurately capture impact events in terms of both force magnitude and temporal response. The results of this initial study demonstrated that the system could capture impact forces with acceptable error (~5%) and high correlation (0.97) between measures of global force and the sensor array. Furthermore, the LIMS demonstrated the ability to capture impact “footprints” that functionally differentiated material properties of density and temperature. A second study incorporated the LIMS as part of a standard controlled surrogate headform drop test

iii

for blunt impacts. The LIMS performed equally well on the curved cranial surface

geometry of the headform and was able to differentiate unique impact contact

distribution patterns based on the ice hockey helmet model’s shell and padding

configurations, including identification of high focal force concentrations (>16 MPa)

during side impact. Of note, global head impact acceleration measures did not

correspond to the magnitude of localized contact forces (R2=0.22), but did

correspond to net global contact force (R2=0.98). A third study used the LIMS between a Hybrid III surrogate headform and an ice hockey helmet during controlled puck projectile collisions. The LIMS was effective at capturing local force distributions dynamics for short impact events lasting 2-4 ms, and again was able to

distinguish between varied helmet model’s padding materials and installed

configurations. Five helmet models were subject to highly localized puck impact at

two different velocities (V1=24.2 m/s, V2=33.3 m/s). At V2, peak contact pressures,

averaged across all helmet models, were nearly double (393 N/cm2) those recorded

at the same location during vertical drop testing (201 N/cm2). Again, linear

acceleration data did not discern these differences in localized pressures. In

summary, this novel testing approach provides an instrument for the assessment of

helmet design and material properties on local impact dynamics, and demonstrates

merit as an industrial and research tool to enhance head protection.

ABRÉGÉ

Les casques de sport modernes ont été efficaces pour réduire les

traumatismes crâniens sévères tels que les fractures du crâne et les hématomes sous-

duraux. Malgré tout, la prévalence élevée des lésions cérébrales traumatiques

mineures reste un problème de santé publique non résolu. Par conséquence, il existe

un besoin important pour un examen plus approfondi de la capacité des casques à

atténuer les forces de collision qui pourraient correspondre à un risque de

traumatisme cérébral mineur. Les normes actuelles évaluent principalement

l’efficacité des casques à minimiser les accélérations maximales de la tête lors

d’impacts contondants.

L’absence de mesures dynamiques locales, plus précisément au site d'impact,

est surtout dû au défi technique qui est d’insérer des matrices sensorielles avec une

haute résolution temporelle. Le développement de cette dernière technique de

mesure pourrait améliorer l'évaluation fonctionnelle des casques en général. Ainsi,

l'objectif principal de cette recherche était de développer un système de cartographie

d’impact local (CIL) tout en permettant l’insertion de ce système entre le casque et la

tête, et ainsi, utiliser le CIL afin d'évaluer les caractéristiques mécaniques de

differentes mousses de rembourrage et différents casques au cours de chute et de

collision contrôlée sur une fausse tête. Interposé entre la calotte/rembourrage et la

surface de la tête, ce CIL est constitué d'un réseau de capteurs de force discrets,

minces, connectés à une grande vitesse d'acquisition de données numériques. Une première étude a démontré la faisabilité d’utiliser le CIL pour capturer avec précision des événements d'impact en termes d’amplitude et de force ainsi que la réponse temporelle. Par ailleurs, le CIL a démontré la capacité de capturer les «empreintes»

d'impact et de différencier fonctionnellement divers matériaux en mousse et des densités.

Une deuxième étude a intégré le CIL dans le cadre d'une norme d'essai contrôlé de fausse tête de substitution lors de chute sur objets contondants. Le CIL s'est révélé tout aussi précis sur la géométrie de la surface crânienne courbe et a été en mesure de différencier les modèles uniques d'impact de contact de distribution basé sur le modèle de coque des casques de hockey et de configurations de remplissage, y compris l'identification de concentrations élevées de force de contact

(>16 MPa). Fait à noter, l'impact global des mesures d'accélération de la tête ne correspond pas nécessairement à l'ampleur des forces d'intervention (R2=0.22).

Une troisième étude a utilisé le CIL entre une fausse tête de substitution

Hybrid III et un casque de hockey sur glace lors de collisions de projectiles. Le CIL est efficace pour capturer des distributions locales de forces dynamiques lors d’événements de moins de 4 ms, et encore une fois a été en mesure de faire la distinction entre les matériaux de rembourrage des modèles de casques variés.

En résumé, cette approche de test innovatrice s'est avérée être un instrument précis pour l'évaluation de la conception du casque et des propriétés des matériaux sur la dynamique d’impact local, et démontre le mérite d'un outil industriel et de recherche visant à améliorer la protection de la tête.

TABLE OF CONTENTS

Table of Contents ...... vii List of Figures ...... x List of Tables ...... xiv List of Abbreviations ...... 1 Statement of Originality ...... 2 Contribution of Authors ...... 2 Acknowledgements ...... 3 Chapter 1: Introduction ...... 5 1.1 Rationale ...... 7 1.2 Objectives ...... 9 1.2.1 Principal objective...... 9 1.2.2 Objective: Chapter 3...... 9 1.2.3 Objective: Chapter 4...... 11 1.2.4 Objective- Chapter 5 ...... 12 Chapter 2: Review of Literature ...... 13 2.1 Anatomical and Physical Properties of the Human Head ...... 15 2.1.1 The Skull ...... 15 2.1.2 The Brain ...... 17 2.2 Head Injury Classification ...... 19 2.2.1 Skull Fracture ...... 19 2.2.2 Focal Brain Injury ...... 19 2.2.3 Diffuse Brain Injury ...... 22 2.3 Quantification and Prediction of Head Injury Risk ...... 24 2.3.1 Tolerance Curves ...... 26 2.3.2 Simulation of Brain Tissue ...... 28 2.4 Head Protection– The Helmet ...... 29 2.4.1 Helmet Construction ...... 30 2.4.2 Focal Injury Prevention: Impact Standardization ...... 32 2.4.3 mTBI Injury Rates in Helmeted Sport ...... 35 2.5 Improving the Helmet: Current Research ...... 37

vii

Chapter 3: Evaluation of a Flexible Force Sensor for Measurement of Helmet Foam Impact Performance ...... 40 3.1 Preface...... 42 3.2 Abstract ...... 44 3.3 Introduction ...... 44 3.4 Methodology ...... 47 3.4.1 System Design ...... 47 3.4.2 Calibration ...... 48 3.4.3 Foam Testing ...... 48 3.4.4 Statistics ...... 50 3.5 Results ...... 50 3.5.1 Nomenclature ...... 50 3.5.2 System Calibration ...... 51 3.5.3 Array Measurement Validation ...... 53 3.5.4 Foam Testing Results ...... 55 3.6 Discussion ...... 57 3.7 Acknowledgements ...... 59 Chapter 4: Impact Performance of Ice Hockey Helmets: Head Acceleration versus Focal Force Dispersion ...... 60 4.1 Preface...... 62 4.2 Abstract ...... 63 4.3 Introduction ...... 64 4.4 Methodology ...... 65 4.4.1 Instrumentation ...... 65 4.4.2 Data Processing & Statistics ...... 68 4.5 Results ...... 68 4.5.1 Sensor Calibration ...... 68 4.5.2 Helmet Testing ...... 69 4.6 Discussion ...... 75 4.6.1 Sensor Performance ...... 75 4.6.2 Peak acceleration and focal force ...... 75 4.6.3 Load Patterns ...... 76

viii

4.6.4 Implications ...... 77 4.7 Funding ...... 77 4.8 Declaration of Conflicting Interests ...... 77 Chapter 5: Projectile Impact Testing of Ice Hockey Helmets: Load Distribution measures ...... 78 5.1 Preface...... 80 5.2 Abstract ...... 81 5.3 Introduction ...... 81 5.4 Methods ...... 85 5.4.1 Statistics: ...... 88 5.5 Results: ...... 88 5.5.1 Event timing ...... 88 5.5.2 Headform Measures ...... 89 5.5.3 Linear Acceleration ...... 91 5.5.4 Angular Acceleration ...... 92 5.5.5 Contact Pressure ...... 93 5.5.6 Pressure Distribution ...... 94 5.6 Discussion ...... 96 5.6.1 Linear Accelerations ...... 97 5.6.2 Angular Accelerations ...... 98 5.6.3 Load distributions ...... 99 5.7 Summary and Conclusions...... 101 5.8 Limitations ...... 102 5.9 Funding ...... 103 Chapter 6: Summary and Conclusions ...... 104 6.1 Summary ...... 106 6.2 Conclusion ...... 109 6.3 Significance of this work to future studies ...... 110 6.4 Application of work to industry and applied fields: ...... 111 Chapter 7: References ...... 112

LIST OF FIGURES

Figure 2-1: The bones of the human skull. ©Patrick J. Lynch. Modified with permission through the Creative Commons Attribution Licence (CC BY 2.5)...... 16 Figure 2-2: Basic anatomy of the human brain. Sagittal section is shown. ©Patrick J. Lynch, modified with permission through the Creative Commons Attribution Licence (CC BY 2.5)...... 18 Figure 2-3: Mechanisms of brain injury arising from direct contact injury. Illustrations represent coronal section of the brain impacting a ridged surface in the direction of the impact vector. Injuries may occur at the site of contact (coup) or remote to the contract site (contrecoup) as a result of these mechanisms. Figure adapted from Schmitt 2011...... 21 Figure 2-4: Wayne State Tolerance Curve (WSTC) for the impact to the human head. Adapted from (Gurdjian, Roberts et al. 1966) ...... 26 Figure 2-5: The effect of differing material properties for absorption of equal energy impacts. Adapted from Newman (2002)...... 31 Figure 3-1: Foam testing apparatus using a guided monorail fitted with a 5 kg 73 mm radius hemispherical impactor. An array of 13 Flexiforce® sensors recorded load distribution...... 49 Figure 3-2: (A) Comparison of the Flexiforce® sensor calibrated up to 800 N with respect to the output from a calibrated load cell. Five repeated calibrations are superimposed for thirteen separate sensors. (B) Bland–Altman plot of sensor error shows a tendency toward increase in error with increasing load...... 52 Figure 3-3: (A) Average contact diameter of the hemispherical impactor with foam sample for each condition relative to the sensor array width and (B) mean force at the sensor array’s perimeter with respect to the contact diameter of hemispherical impactor...... 54

Figure 3-4: Bland–Altman error plot between the GFFF and GFLC variables. 95% of the error is contained within ± 164.9 N of the mean...... 54 Figure 3-5: Average (n = 9) peak load distribution for 3 densities of EPP foam during first impact by a 5 kg hemispherical impactor. Ambient (20°C) and cold (- 25°C) temperatures were tested for both initial and repeated impact. Peak focal force

(72.7 N) was observed for the cold sample of the highest density EPP foam during the initial impact...... 55 Figure 3-6: (A) Average (n = 9) peak acceleration of a 5 kg hemisphere on impact to three densities of EPP foam at ambient and cold temperatures for initial and repeated 5 J impact. (B) Average (n = 9) peak focal force measured using an array of 13 Flexiforce® sensors instrumented under the impacted foam. Error bars represent standard deviation. *Denotes significant difference...... 56 Figure 4-1: Impact calibration method for the Flexiforce® sensor. A vinyl cylinder adhered to a 4.7 kg headform was impacted directly onto the Flexiforce sensor to calibrate output voltage to force...... 67 Figure 4-2: Physical arrangement of the 25 Flexiforce sensors on a magnesium EN960 headform and their position relative to the helmet features for front, side and rear impact conditions...... 67 Figure 4-3: Sample calibration for a Flexiforce® sensor by direct impact of 4.7 kg mass. Ten unique drop impacts were recorded and a third-order polynomial was fit to the data...... 69 Figure 4-4: Peak linear acceleration (A) and peak focal force (B) results during front, rear and side impacts to a helmeted EN960 575 mm magnesium headform during linear drop impacts following the CSA z262.1-09 testing guidelines. Five models of ice hockey helmets were subjected to three repeated impacts and two temperatures (21 °C and –25 °C). A total of 50 helmets were tested (5 models × 2 temperatures × 5 samples)...... 71 Figure 4-5: Average (n = 5) peak focal force distribution maps during the first 4.5 m/s impact to ambient (21 °C) conditioned ice hockey helmets. Impacts to the front, rear and side of (a) Helmet 1, (b) Helmet 2, (c) Helmet 3, (d) Helmet 4 and (e) Helmet 5 are presented...... 73 Figure 4-6: Average (n = 5) peak focal force distribution maps during the first 4.5 m/s impact to cold (–25 °C) conditioned ice hockey helmets. Impacts to the front, rear and side of (a) Helmet 1, (b) Helmet 2, (c) Helmet 3, (d) Helmet 4 and (e) Helmet 5 are presented...... 74 Figure 5-1: Force sensor locations (25) were selected based on wireframe intersections of the Hybrid III finite element model (left). The virtual coordinates

were transferred to the physical headform (right). The point-to-point distance from reference point A (tip of nose) to each sensor row along the median plane is displayed. A red laser on the physical headform indicates the alignment of the projectile canon with the central sensor...... 86 Figure 5-2: Event sequence for the impact of a 160g hockey into a helmeted hybrid III headform. The period over which the material can dissipate force is very brief (<2ms). The rotational displacement of the headform peaked around 70ms...... 89 Figure 5-3: Peak resultant linear acceleration by helmet model for the repeated impact of an ice hockey puck at 24 and 33 m/s to the forehead of a hybrid III dummy headform. Asterisks within the bar indicate a significant difference between 1st and 2nd impacts within helmet models. Average head impact criterion values (HIC) and Severity Index (SI) are indicated for each helmet model. Error bars indicate ± 1 SE...... 91 Figure 5-4: Peak resultant angular acceleration across helmet model for two repeated impacts to the forehead of a hybrid III dummy headform by an ice hockey

puck travelling at 24 m/s (PI24) and 33 m/s (PI33). Average peak angular velocity (rad/s) about the y-axis is presented below each helmet model. Asterisks above the bars represent significant difference of the indicated model from all other helmet models. Asterisks within the bar represent significant differences between impacts. Error bars indicate ± 1 SE...... 93 Figure 5-5: Peak contact pressure (N/cm2) between the five different helmet models and a hybrid III dummy during repeated impacts of an ice hockey puck at 24 m/s and 33 m/s. Average pressure (indicated by shaded band of each bar) was calculated for the active contact area (indicated under each helmet model). Asterisks indicate a significant difference between impacts within helmet models. Error bars and shaded regions indicate ± 1 SE...... 94 Figure 5-6: Low-speed sample trials showing spatial loading profiles between 5 differing helmet models during impact by a 260g hockey puck at 24 m/s. Twenty- five discrete force signals (lower panel) were recorded at the helmet-head interface of a hybrid III headform. The dashed vertical line represents the instant of maximal net force. Intensity maps (upper panel) represent the spatial pressure distribution at this

xii

same instant. In some cases the greatest the point of peak net force did not correspond to the time of individual sensor maximums (eg. Helmet 5)...... 95 Figure 5-7: High-speed sample trials showing spatial loading profile between 5 differing helmet models during impact by a 260g hockey puck at 33 m/s. Twenty- five discrete force signals (lower panel) were recorded at the helmet-head interface of a hybrid III headform. Vertical dashed lines represent the instant of maximal net force. Colour maps (upper panel) represent spatial pressure distribution at this same instant...... 95

xiii

LIST OF TABLES

Table 2-1: Ice hockey helmet standards comparison. Simply comparing the peak g value between standards is not a sufficient comparison given the multiple variables of headform size, impact velocity and drop tower configuration...... 34 Table 3-1: Descriptive statistics for all 13 sensors used during foam impact testing. Each sensor was calibrated 5 times and the resulting force-to-voltage output was fit with a third order polynomial...... 51 Table 4-1: Physical characteristics of the five ice hockey helmet models evaluated in this study. Ten samples of each model were obtained. Manufacturer and model number data are excluded...... 68 Table 5-1: Physical characteristics of the five ice hockey helmet models evaluated in this study. Ten samples of each model were obtained. Manufacturer and model number data are excluded...... 87 Table 5-2: Total number of impact trials per helmet model...... 88 Table 5-3: Average (Least-square mean ± standard error) resultant linear acceleration, resultant angular acceleration, average pressure, maximal pressure and contact area during impact to various models of ice hockey helmets by an official- sized puck travelling at 24.2 m/s. Maximal values are indicated by bold type...... 90 Table 5-4: Average (Least-square mean ± standard error) resultant linear acceleration, resultant angular acceleration, average pressure, maximal pressure and contact area during impact to various models of ice hockey helmets by an official- sized puck travelling at 33 m/s. Maximal values are indicated by bold type...... 90 Table 5-5: Impact test results (linear acceleration, average and peak contact pressure) from a high-mass low-velocity vertical drop test (CSA z262.1-09) protocol for front impacts to ambient temperature ice hockey helmets. Helmet models matched those used in the current study. Data adapted from Ouckama & Pearsall (2012)...... 97

xiv

LIST OF ABBREVIATIONS

ABS Acrylonitrile-Butadiene-Styrene ASTM American Society for Testing and Materials AE Athlete-exposure ATD Anthropometric test devices CE Conformité Européene CIS Canadian Interuniversity Sport CDC Center for Disease Control and Prevention CNS Central nervous system COM Center of mass CSA Canadian Standards Association DAI Diffuse axonal injury DAQ Data acquisition EPE Expanded poly-etheleyne EPP Expanded poly-propelyne EPS Expanded poly-stryrene Fc Critical force to cause bone fracture FE Finite element FEA Finite element analysis SI Gadd Severity Index GFACC Global force (Newtons) estimated by F=ma, where m represents the mass of the impactor (4.977 kg) and a represents the calibrated accelerometer signal (g) * 9.81. GFFF Global force (Newtons) estimated by the approximate integral (sum of interpolated forces) of a 3D surface fit to the 13 focal forces GFLC Global force (Newtons) registered by the uniaxial load cell; used as a standard for reference and for calculation of RMS error HIC Head injury criterion HIT Head Impact Telemetry H3 Hybrid III Headform ISO International Organization for Standardization ISS Injury surveillance system KHL Kontinental Hockey League LOC Loss of consciousness LIMS Localized impact mapping system MEP Modular Elastomeric Polymer NCAA National College Athletic Association NHAMCS National Hospital Ambulatory Medical Care Survey NHL National Hockey League NOCSAE National Operating Committee on Standards for Athletic Equipment PMHS Post Mortem Human Subjects SDH Subdural hematoma TBI Traumatic brain injury mTBI Diffuse brain injuries TCL lésions cérébrales traumatiques VDT Vertical drop test VN Vinyl Nitrile VI Virtual Instrument (National Instruments Labview) WSTC Wayne State Tolerance Curve

STATEMENT OF ORIGINALITY

All materials presented in the thesis represent original work conducted by the authors and have not been published elsewhere except where specific references indicate.

The manuscripts presented in Chapters 3-5 represent original ideas and contribute to the advancement of the field of helmet research. The papers presented represent the first publication of research pertaining to dynamic measures of load distribution during impact to ice hockey helmets. All data presented in this thesis were collected at either the department of Kinesiology, McGill University, Montréal, Québec or the Advanced Research and

Development Center at Bauer Hockey Corp. in St-Jérôme, Québec.

CONTRIBUTION OF AUTHORS

The conception of experiments, data collection, analysis and preparation of the manuscripts presented in this thesis are primarily the work of Ryan Ouckama. The entire work was conducted under the direct guidance and supervision of Dr. David Pearsall, hence his inclusion as co-author on each paper presented.

ACKNOWLEDGEMENTS

First and foremost I would like to thank David Pearsall for his incredible patience

and understanding as a graduate supervisor. Over the course of my Master’s and Doctoral studies, Dr. Pearsall has been integral in my development as a graduate student, researcher, and as a person. His kind demeanour and commitment to academics, athletics and family have created an ideal learning environment for which I will strive to carry on in my future work.

From the biomechanics lab, I wish to acknowledge the many hours of work put in by our technicians Jonathan James ‘JJ’ Loh, Phillipe Dixon and Yannick Michaud-Paquette.

Their assistance in data processing through the development of various Matlab toolboxes was of great value to this project. Thank you to the multiple students whom assisted with data collection and additionally thank you to Yannick for his assistance with multiple French translations over the years.

Thank you to my wife Meghan for her years of support through life and academia. I am very thankful to my family for their financial support throughout all the years of education and encouraging me to pursue this unique field of research. I was fortunate to have received the Industrial Innovation Scholarship (IIS) through NSERC & FQRNT which funded the initial years of my work and provided ongoing industry contacts. Without the support of these agencies this project may not have been possible. Finally, I would like to acknowledge Marie-Claude Généreux and Jean-Francois Laperrier from Bauer Hockey Corp. for their extreme professionalism and openness to scientific research and the promotion of innovation.

Thank you all!

CHAPTER 1: INTRODUCTION

1.1 Rationale

Ice hockey is a winter sport enjoyed by millions of people around the world. The sport is played across all skill levels and age groups ranging from casual ‘shinny’ played on an outdoor frozen pond among friends and neighbours, to minor competitive leagues, college

leagues (CIS, NCAA), junior leagues, major professional leagues (NHL, KHL) and at the

Olympic games as a sport for both men and women. Today, it would be shocking to see a

competitive team play full contact hockey without helmets, yet this was common prior to the

1980’s. Initial adoption of protective helmets was low in part due to social stigma, but as

severe head injuries became more prevalent, the NHL took action and required mandatory

use of helmets for all new players in 1979. Those that were already in the league were given

an option to wear a helmet and, many continued to do so. In fact, the last NHL game to

have an unhelmeted player did not occur until 1996 (NHL 2009).

Modern hockey helmets provide added protection to the skull and brain thereby

reducing numerous injury risks (cranial fracture, focal brain injury, and laceration) that can

be attributed to direct impact to the head as a result of the high speed and physical nature of

the game (Biasca, Wirth et al. 2002). These impacts can include the head falling directly onto

the ice, skating or sliding into the boards or other arena feature (e.g., stanchion), contact with

another player (body check, illegal elbow to head, illegal slash to the head), or projectile

impact from puck. Given the multitude of impact scenarios and energies, helmets must be

able to withstand multiple modalities of assault. Helmet performance can be measured

experimentally by placing the helmet securely onto the surrogate head of a crash test dummy

(headform) and measuring the resulting acceleration during impact. Linear acceleration is the

primary metric used in these tests as it was most correlated to skull fracture and focal brain

injury during early studies of human tolerance to impact (Gurdjian, Roberts et al. 1966). All

hockey helmets must be certified for sale and use in Canada, meaning they meet specific sizing and performance criteria dictated by the Canadian Standards Association (CSA).

Hockey helmet impact standards have typically set performance criteria based on multiple 4.5 m/s vertical drops (CSA 2009); however, this only represents one type of impact and may not be representative of high speed, highly localized impacts, such as those delivered by a slash from a stick or from contact with the puck. These types of impacts have been measured to some degree using pressure sensitive films (Bishop and Arnold 1993) and measures of linear and angular acceleration (Coulson, Foreman et al. 2009). In Bishop’s study six helmet models were evaluated using three different puck velocities. Pucks were fired at the left temporal area of a Hodgson-WSU headform and contact parameters were recorded using pressure sensitive film. Peak pressures ranged from 8 to 28 MPa during puck impacts of 100-140 km/h (27.8-38.9 m/s). These pressures were significant as fracture of the temporoparietal area of the skill could occur at pressures as low as 3.1 MPa. Linear acceleration measures for impacts ranging from 100-120 km/h (27.8-33.3 m/s) were less than 250 G and the corresponding severity index showed little risk of injury. Despite these limitations of pressure sensitive film, Bishop’s work on puck impacts to the temporal region of ice hockey helmets identified a significant issue: the risks associated with global measures of acceleration did not correspond to the risk of skull fracture posed by excessive pressure at the contact site. Bishop found that the contact forces posed a significant risk to injury and contradicted the head injury risk calculated using traditional measures of linear acceleration.

This finding warranted further investigation of force distribution measures as a way to evaluate helmet performance and maximize player safety.

1.2 Objectives

1.2.1 Principal objective

The principal objective of this thesis was to explore measures of localized load

distribution in helmet impacts, identified as a significant injury risk in Bishop’s earlier work.

An innovative sensor solution capable of time-series measurement of impact load

distribution at the helmet-head interface was developed. The system was used in conjunction

with measures of acceleration during two fundamentally different impact types common to ice hockey: 1) high-mass. low-velocity impacts (e.g., fall on the ice) 2) low-mass, high-

velocity impacts (e.g., puck to the helmet). Following a review of literature (Chapter 2), the

body of this work (Chapters 3-5) is presented in manuscript format, separated into three district chapters, each with specific goals progressing toward the principal objective.

1.2.2 Objective: Chapter 3

The objective of this chapter was to develop and evaluate a localized impact mapping system (LIMS). Before application to the head-helmet interface, it was deemed necessary to assess the system performance on a flat surface to measure system performance using controlled variables of foam thickness and density. As discussed earlier (Section 1.1), prior efforts to quantify load distribution in helmets were achieved using pressure sensitive films. These films provided only a cumulative snapshot of the peak impact forces and, therefore, cannot be directly correlated to other time-series data such as linear and angular acceleration. It was thus determined that a time series measurement of load distribution was necessary in order to generate the most useful data for visualization of the force distribution and for identification of the exact point in time and space where localized forces may deviate significantly from global loading profiles. Several commercial systems already exist for the

measurement of load distribution, but they were not deemed suitable for helmet impact use due to limitations in sampling speed, flexibility, measurement range, durability and cost. This chapter, presented in manuscript format, evaluates the performance of the LIMS in the context of impact to flat foam samples. This work was published in the Journal of

Biomechanics in 2011 and appears in its original form with the exception of formatting changed to meet the criteria of the McGill e-thesis.

Ouckama, R. and D. J. Pearsall (2011). "Evaluation of a flexible force sensor for

measurement of helmet foam impact performance." J Biomech 44(5): 904-909.

1.2.3 Objective: Chapter 4

Following successful measurement of LIMS error and performance under controlled flat impact conditions, the objective of Chapter 4 was to instrument the curved helmet-head interface using the LIMS and to evaluate commercial hockey helmets for their load distribution performance. Given helmet testing standards are well established and provide the current benchmark for helmet performance, it was logical to use this testing regime (CSA z262.1-09) as a starting point for the analysis of load distribution performance. This first application of the LIMS to helmet testing examined an impact type that represented a blunt high-mass, low-velocity impact to the helmet. The goal of this paper was to report novel measures of load concentrations relative to linear acceleration. Differing load concentrations were evaluated through the use of multiple helmet models utilizing various load attenuating materials. The chapter is presented in manuscript format, modified only to match the McGill e-thesis template and was published in the Journal of Sports Engineering and Technology.

Ouckama, R. and D. Pearsall (2012). "Impact performance of ice hockey helmets: head

acceleration versus focal force dispersion." Proceedings of the Institution of

Mechanical Engineers, Part P: Journal of Sports Engineering and Technology

226(3/4): 185-192.

1.2.4 Objective- Chapter 5

The objective of this chapter was to finally revisit Bishop’s earlier work on

assessment of ice hockey helmet performance during highly localized puck impacts. Using

the LIMS (Chapter 3) in conjunction with measures of linear and angular acceleration, this

chapter provides a comprehensive assessment of helmet performance during highly localized

impact by a low-mass, high-velocity projectile (i.e., hockey puck). The goal of this work is to provide data for comparison to the common helmet test modality (Chapter 4) in order to assess helmet performance across a range of impact types. This work is presented in manuscript format and is in preparation for submission to the Journal of Biomechanics.

CHAPTER 2: REVIEW OF LITERATURE

This chapter provides an overview of the research related to head and helmet impact biomechanics. It begins with the identification of basic anatomical features and physical properties of the head, followed by a classification of head injury mechanisms.

Specific methods of quantifying the head impact dynamics will be presented in the context of known (or estimated) tissue tolerance limits. Helmet protection will be discussed including a detailed summary of impact absorbing materials, standardized testing methods, and effectiveness based on injury rates in contact sports. Finally, methods to improve helmet performance will be presented based on current research.

2.1 Anatomical and Physical Properties of the Human Head

The human head is formed by the rigid osseous structures forming the skull, the surrounding tissues of the scalp and face and the underlying cerebral tissues of the brain.

Together, the average human head weighs approximately 3.6-5.5 kg and represents

approximately 6.5-9% of the adult body mass (Yoganandan, Pintar et al. 2009).

2.1.1 The Skull

The skull (Figure 2-1) is a complex structure composed of 22 bones divided into two groups, the cranial bones (8) and facial bones (14). The eight bones of the cranium enclose

the soft brain tissue and provide attachment sites for the spinal column and the muscles of

the neck. The cranium consists of the frontal, occipital, sphenoid and ethmoid bones as well

as paired parietal and temporal bones. The majority of cranial bones are diploid flat bones

with stiff outer strata of cortical bone and an inner lightweight energy-absorbing core

(Motherway, Verschueren et al. 2009). The individual bones are fused to form a unified

convex structure by means of interlocking fused joints called sutures. The thickness of the

cranial bones varies significantly with the thinnest sections (2.6-5.0 mm) located on the temporal bone (Jarquin-Valdivia, McCartney et al. 2004), and the thickest sections (9.7-15.1 mm) located on the occipital protuberance situated at the posterior aspect of the head

(Ebraheim, Lu et al. 1976). The orbits, nasal cavity and jaw consist of fourteen facial bones: the mandible and vomer as well as the paired inferior nasal concha, lacrimal, maxilla, nasal, palatine and zygomatic bones.

Figure 2-1: The bones of the human skull. ©Patrick J. Lynch. Modified with permission through the Creative Commons Attribution Licence (CC BY 2.5).

Impact studies indicate that the mechanical properties of cranial bone vary largely due to differences in morphology, subject age, and condition (fresh, embalmed or dry).

Further, variations in ex vivo testing methods can also impact material properties. Ideally, testing of these materials should be conducted on freshly deceased subjects as they provide similar results to live tissue (Greenberg, Gonzalez et al. 1968). As bone ages and dries it increases in stiffness, though, the original material properties can be restored to some degree with rehydration of dry bone. The mechanical properties and morphology of bone vary between subjects and are rate dependant which further complicates reporting an average

value, for example, the elastic modulus of cranial bone was measured at 7.46 GPa at 0.5 m/s impact and 15.54 GPa at a 2.5 m/s impact (Motherway, Verschueren et al. 2009).

2.1.2 The Brain

The brain is divided into the four major regions: the cerebrum, cerebellum, diencephalon (thalamus, hypothalamus and epithalamus), and the brainstem (midbrain, pons and medulla omblongata) (Figure 2-2). The human brain sits suspended in cerebrospinal fluid, isolated from the body’s bloodstream by the blood-brain barrier and protected by the bones of the skull. Though surrounded in protection, the brain is fragile and susceptible to injury. To further protect these delicate structures, the central nervous system (CNS) made up of the brain and spinal cord is enveloped in a series of membranes called the meninges.

The meninges consist of three layers: the dura mater, the arachnoid mater and the pia mater.

The primary purpose of these membranes is to protect the CNS, both structurally and from foreign infection. In addition to the protective features, the meninges also provide a scaffolding to suspend the brain within the cerebrospinal fluid.

Histologically, the two primary components of the brain are the nerve cells and the glial cells. The nerve cells, or neurons, serve as electrical transmitters while the glial cells provide supporting functions for the neurons. The neurons consist of a cell body (soma), branching dendrites (structures for signal reception) and a long projection called axons

(nerve fibres). These parts of the neurons are distributed in an organized manner throughout

Figure 2-2: Basic anatomy of the human brain. Sagittal section is shown. ©Patrick J. Lynch, modified with permission through the Creative Commons Attribution Licence (CC BY 2.5).

the brain making up two differentiated tissues, white and grey matter. The grey matter is concentrated on the surface of the cerebral hemispheres and is made up of neuronal cell bodies, whereas the white matter contains mostly glial cells and myelin insulated tracts of the axons. These tissues differ both in histology and function. The grey matter is primarily associated with cognition and processing, while the white matter coordinates communications between different brain regions. Grey matter regions are not exclusive to the cerebellum and exist elsewhere in the brain including the basal ganglia, mid-brain and brain stem. While the skull provides protection to the brain, it can also result in injury to the underlying tissue through inward bending, fracture, or relative motion of the brain and skull causing contact with irregular boney structures. The following section will explore the classifications of head injuries to the skull and brain tissues.

2.2 Head Injury Classification

For the purposes of this thesis, classification of head injuries will be divided into those causing injury to the skull and those causing injury to the brain. These can be independent of each other as a skull fracture may or may not include injury to the cerebral tissue.

2.2.1 Skull Fracture

Fracture to the skull can be either depressed or linear in nature. A depressed fracture, occurs when a blow to the head results in some bone fragments being pushed inward toward the brain. This depression can result in damage to the underlying vasculature of the brain, and/or may cause laceration to the brain tissue if the cranial bone fragments. Linear fracture results from a break in the cranial bone resembling a thin line in the absence of depression or distortion of the bone. At the weakest part of the skull, the temporal bone, Aldman

(1984) found that a depressed fracture could occur if the contact area was less than 5 cm2 and the localized pressure exceeded 4 MPa. This weakness of the temporal bone is of particular concern in sports such as baseball, ice hockey, or lacrosse, where relatively small, high velocity projectiles are common. If the fracture breaks the skin, and tears the dura with a bone fragment exposing the isolated brain to the environment, it is termed an open fracture. Open fractures further complicate the injury by introducing the external environment and allowing possible infection of the brain tissue. If the dura and/or skin remain intact, the head injury is classified as closed.

2.2.2 Focal Brain Injury

A focal injury is caused by direct contact of an object to the head resulting in the transfer of mechanical forces. These forces, which may or may not fracture the skull, can

result in impulsive loading to the brain tissue. During an impact to the head the skull and brain decouple. The brain tends to lag the motion of the skull by ± 5mm before returning to its initial position (Hardy, Foster et al. 2001). This relative motion can result in rupture of delicate blood vessels passing through or near the dura. Bridging veins are delicate, thin- walled vessels that originate from the surface of the brain beneath the dura matter and run anteriorly toward the superior sinus of the skull. When relative brain motion occurs between the brain and skull, these bridging veins can be subject to high levels of strain and can rupture, leading to an accumulation of blood, creating a subdural hematoma (SDH)

(Hodgson 1991). Other types of hematoma can occur outside the dura (epidural hematoma) or deeper within the brain (intracranial hematoma) due to an arterial or venous rupture. The immediate effects of focal brain injuries are considered primary brain injuries. Primary injury occurs at the time of impact with immediate neurological effects based on the location of the damaged brain tissue. If not treated quickly a primary brain injury can lead to secondary injuries. Given the brain is enclosure by the skull, a significant increase in blood volume will result in a corresponding increase in intracranial pressure, beyond homeostatic management.

A cascading effect can then result, as the increased volume raises intercranial pressure, which in turn, leads to swelling and further increases in pressure. The increased intercranial pressure can result in compression of the underlying cortex causing localized ischemic brain damage and ultimately, neuronal death (Reilly and Bullock 2005). Brain contusion often occurs in the outer cortex and typically results from the interior collision of the brain and skull due to relative movement during impact. The contusion injury typically occurs at the impact site and is referred to as a “coup” injury. The coup injury is caused by increasing intracranial pressure at the impact site. This coup injury is often accompanied by a contusion remote to the impact site, or “contrecoup”. The contrecoup injury can be caused by several

mechanisms, including direct contact to the region by skull deformation or relative motion of the brain, or by tissue response to pressure gradients or waves (Figure 2-3).

Figure 2-3: Mechanisms of brain injury arising from direct contact injury. Illustrations represent coronal section of the brain impacting a ridged surface in the direction of the impact vector. Injuries may occur at the site of contact (coup) or remote to the contract site (contrecoup) as a result of these mechanisms. Figure adapted from Schmitt 2011.

The ‘pressure wave’ theory is based on the notion that the incompressibility of CSF leads to the impulse travelling through the compressible brain material until it reaches the unyielding skull bone at the opposite side of the head (Goggio 1941). Alternatively, the

‘pressure gradient’ theory suggests that development of negative pressures opposite to the impact site during rapid motion can cause the injury (Goggio 1941; Nusholtz, Wylie et al.

1995). The vacuum created on the contrecoup site can be injurious due to sudden cavitation of the brain tissue. Intracranial pressure has been measured experimentally in the brain through use of post-mortem human subjects (PMHS). During impact to helmeted PMHS heads intercranial pressure reached upwards of 20-153 kPa at the coup site and -163 to 54 kPa at the contrecoup site (Hardy, Mason et al. 2007), thus providing support for the theory of negative pressure generation as an injury mechanism.

2.2.3 Diffuse Brain Injury

The inertial effects (linear and angular acceleration) of a mechanical blow can lead to

a diffuse brain injury. A diffuse brain injury differs from a focal brain injury in that it

conveys widespread damage to the brain. The injury has been attributed to both direct (head

contact) or indirect (cases where the torso is stopped or accelerated rapidly) loading by

various researchers (King, Yang et al. 2003; McLean and Anderson 2005). However, despite

evidence of the indirect injury mechanism (non-contact whiplash) in animal models

(Ommaya and Hirsch 1971), the occurrence of a diffuse injury in the absence of head contact is rare, if at all (McLean 1995). Researchers have long debated the importance of linear versus angular acceleration in the production of diffuse brain injury. In a diffuse brain injury there is often a global disruption of neurologic function which could be caused by physiological or anatomical disruption of the brain. The injury is manifested by a multitude of symptoms ranging from loss of consciousness or amnesia, behaviour changes, cognitive impairment, emotional changes, light or motion sensitivity and sleep disturbances. Diffuse brain injuries include minor traumatic brain injury (mTBI), diffuse axonal injury (DAI) and brain swelling.

2.2.3.1 Mild Traumatic Brain Injury (mTBI) & Concussion

Mild traumatic brain injury (mTBI) and concussion are often used interchangeably in the literature; however, concussion is actually a form of mTBI. Within the scope of this thesis, the terms will be used interchangeably. mTBI is defined as a complex pathophysiological process affecting the brain (McCrory, Meeuwisse et al. 2009). The

condition is induced by biomechanical forces such as a direct blow to the head, or elsewhere

on the body, provided sufficient impulsive force is transmitted to the head. mTBI results in a

rapid onset of what are typically short-lived impairments of neurological functions. These impairments are due primarily to functional disturbances in the brain, rather than structural injury. This creates a challenge in effectively diagnosing concussion, as in the absence of structural injury, imaging techniques such as CT or MRI will not show any abnormality in the brain. While the majority of concussions resolve in a short time, roughly 7-10 days, acute cases can last much longer, affecting behavior, cognition and sleep. Management of mTBI injuries involves rest from all activities until symptoms resolve, followed by a graded return to regular activity, be it sport, or even just daily tasks such as reading or watching tv. It is important the symptoms do not return before moving to the next activity grade, as a second injury can further complicate recovery. In a population of football players it was found that those players that had suffered a concussion were at a higher risk of a second concussion injury (Guskiewicz et al. 2000), highlighting the importance of a gradual return to play. The issue of multiple concussions and the risks associated has been gaining increasing attention both in research and in the rules and conduct of various contact sports. Recently, the long term effect of multiple mTBI injuries has been realized. In a population of retried professionally football players sustaining three or more concussions, there were elevated rates of cognitive impairment, depression and possibly early onset of Alzheimers disease relative to the general population (Guskiewicz, Marshall et al. 2005; Guskiewicz, Marshall et al. 2007).

2.2.3.2 Diffuse Axonal Injury

A diffuse axonal injury (DAI) is a serious condition which involves mechanical disruption (excessive shearing from rotational forces, or rapid deceleration) of numerous axons throughout the cerebral hemispheres and subcortical white matter (Melvin and

Lighthall 2002). This mechanical disruption can lead to the formation of extensive lesions or

hemorrhages in the white matter tracts of the cortex and extending into the brainstem. The

lesions may be detectable through the use of a high resolution CT scan, though can often be

missed due to their small size. More recently the use of diffusion tensor imaging has been

successfully applied to determine the extent of damage to the white matter of the brain (Xu,

Rasmussen et al. 2007). DAI involves immediate loss of consciousness lasting days to weeks.

If the patient regains consciousness they will typically display decerebrate posturing, with

severe memory and motor deficits. Treatment of diffuse brain injury is supportive. There are

no surgeries or other treatments currently available. The goal is to allow the brain to heal

itself over time, which does happen in some cases, however with most severe injuries the

effects are permanent. The conditions of mTBI and DAI injuries may be amplified by

swelling of the brain tissues. The enclosed space causes an increase in intracranial pressure

(Gennarelli and Thibault 1982) which agitates the underlying injury further, causing a

cascade effect if not immediately resolved.

2.3 Quantification and Prediction of Head Injury Risk

As noted earlier, biological tissues tend to vary greatly in their mechanical properties

and pose significant experimental challenges both in terms of logistics and ethics approval.

While it is still necessary to conduct testing on PMHS’s to obtain ‘real’ impact responses, it is

far easier and repeatable to utilize anthropometric test devices (ATD), commonly referred to

as crash test dummies. ATDs are manufactured to mimic human characteristics of size,

shape, stiffness and energy absorption and dissipation qualities. The use of ATDs simplifies

measurement of kinematic and kinetic responses by the use of embedded sensors and solid,

precise mounting points. The most commonly used ATD is the Hybrid III (H3) dummy,

developed by General Motors for use in front impact collisions (Mertz 2002). While the H3 represents a full body ATD, many helmet testing standards utilize anthropometric headforms of various materials including magnesium, aluminum, epoxy or wood.

Alternatively, live humans can be instrumented with sensors embedded in mouth guards, earpieces, or placed around the head using adhesives or instrumented helmets. Once instrumented, one cannot ethically apply dangerous forces to live human subjects. One

solution to this problem is to instrument players in various sports that are at high risk of

concussion or head injury. The Head Impact Telemetry (HIT) system was developed for this

task and has lead to one of the largest impact databases to date spanning multiple sports

including football and ice hockey (Duma, Manoogian et al. 2005; Guskiewicz, Mihalik et al.

2007; Rowson, Duma et al. 2012). The HIT system consists of an array of six orthogonally

mounted accelerometers embedded into a band which replaces a section of padding within

the helmet. The system collects impact acceleration data and transmits it to a central PC unit

where it is further processed and visualized alerting team staff in the event of an impact of

high magnitude acceleration. The system, while novel, is not without limitations. In order to

minimize helmet/head decoupling the system employs a series of spring-like actuators on

each accelerometer to maintain contact with the head. Though there have been criticisms of

the system’s ability to separate helmet versus head acceleration, there is good

correspondence to head kinematics when tested in-lab (Manoogian, McNeely et al. 2006). In-

lab comparisons with ATD linear acceleration were strong (r2 = 0.903) however angular acceleration measures are not as reliable (r2 = 0.528) (Beckwith, Greenwald et al. 2012),.

Despite the acknowledged limitations of the system, it has provided a wealth of data

concerning athlete exposure to impact and has paved the way for future implementation of

‘smart’ helmets into multiple contact sports.

2.3.1 Tolerance Curves

Once an impact is quantified by instrumentation of an athlete, PMHS or ATD, it is

necessary to have some sort of threshold value to assess the outcome values to a risk of

injury. Tolerance curves provide an estimation of the range between fatal and non fatal

injury risk in terms of acceleration and time. Most criteria and tolerance data originated from

Dr. E. Stephen Gurdijian’s research, which first involved measures of acceleration and

changes in intracranial pressure of dogs. In later work, he augmented the dataset by

measuring the concussive effects of hammer blows to monkeys and human cadavers. This

dataset was transformed into what is now the well-known Wayne State Tolerance Curve

(WSTC) (Figure 2-4). The acceleration values from which the curve originates are referred to as effective acceleration, which is defined arbitrarily as a value somewhat greater than half the peak value of a modified triangular acceleration pulse (Versace 1971). This arbitrary measure lead to several criticisms of the WSTC and associated metrics based upon it, however, it still stands as a commonly referenced metric.

Figure 2-4: Wayne State Tolerance Curve (WSTC) for the impact to the human head. Adapted from (Gurdjian, Roberts et al. 1966)

2.3.1.1 Gadd Severity Index (GSI)

In 1966 Charles Gadd of General Motors, proposed a severity index based on the

WSTC. Gadd used the logic that the area under the acceleration/time curve could form the basis of an index. However, using that logic, the risk associated to a particular impulse value would be the same regardless of whether it was a short duration, high acceleration impulse or long duration and low acceleration impulse. To accommodate for this, knowing that low duration impacts are of less risk of head injury when subject to high acceleration, a weighting factor was applied to the acceleration value of 2.5. This factor happens to be the slope of the

WSTC when plotted on a logarithmic scale. The Gadd Severity Index is defined as:

GSI = a 2.5dt ∫T This severity index is currently utilized in several sports helmet standards including the

NOCSAE football and ASTM ice hockey helmet standards. While this metric is not perfect,

it does provide useful correlation to past injury data. The applicable time domain for this index is between 1-50ms, given the linear fit fails to match the data from the WSTC beyond these points (Versace 1971).

2.3.1.2 Head Impact Criterion (HIC)

In 1971, John Versace, of the Ford Motor Company, presented a modification to the

GSI. Based on criticism that the GSI still did not adequately account for long duration, low acceleration impacts, Versace proposed a severity index that was based only on the part of the impact that was relevant to brain injury risk. This was accomplished by integrating the acceleration/time curve only over the time interval that yielded the maximum value of HIC.

This essentially eliminates all secondary impacts and focuses solely on the primary impact pulse. To prevent the same issues of long duration, low acceleration impacts creating high

criterion values, the HIC utilizes a cut-off maximum time of 36ms. In some cases a 15 ms

clip (HIC15) is more appropriate, and is now often used in the automotive industry (Mertz,

Prasad et al. 1996). HIC, like the GSI is a measure of the integral of acceleration at the center of mass of the head. The integral is calculated across the acceleration pulse between two points in time that maximize the result. The formula for HIC is:

2.5  t   1 2   HIC =  a(t)dt (t − t ) t − t ∫ 2 1  2 1 t1    max The HIC is maximized by direct calculation of all possible time intervals up to the cut-off time at all possible positions across the acceleration/time plot. The HIC is non-specific to the type of injury or severity of injury, and it does not include variation in mass or direction of impact (Hutchinson, Kaiser et al. 1998). The threshold value for injury is HIC=1000 which represents approximately a 18% chance of life-threatening injury, or around a 50% chance of a serious head injury (Hutchinson, Kaiser et al. 1998). The HIC is a popular measure in automotive testing, however it is used less in helmet testing standards.

2.3.2 Simulation of Brain Tissue

Direct measurement of crash tests using instrumented PMHSs, can be restrictive due to financial, ethical, or physical limitations. Computerized simulation of human tissue can thus provide a valuable tool for assessment and understanding of injury. Finite Element

Analysis (FEA) is a simulation method used to analyze the behavior of structures subjected to various static or dynamic loads. FEA uses a complex system of points (nodes) that form a grid (mesh). The mesh is given material and structural properties, which define how it will respond to certain loading conditions. From each node the mesh element extends to each adjacent node creating a web of vectors that carry the material properties throughout the

object. The modeled object can have linear or non-linear properties in two or three

dimensions. Significant increases in computer processing power are required for non-linear,

three-dimensional solutions that are typical of biomechanics systems. The power of finite

element modeling is tremendous given the simulation is a realistic depiction of the material.

The challenge of FEA is that in order to model complex structures their geometry must be

simplified (to a finite number of elements) to satisfy the requirements of the software, and

reduce processing time. Thus, FEA only provides an estimation of the material behavior.

When modeling biological materials we face the challenges of accurately representing a living

tissue mathematically. Nevertheless, scientists have made significant gains in understanding

biological systems using this modeling technique. Several FEA models of the human head

have been developed and validated using data from PMHS impacts (Yang, Mao et al. 2011).

Currently, these models provide the only way to non-invasively estimate the magnitude and

propagation of intracranial stresses of various impacts.

2.4 Head Protection– The Helmet

The primary purpose of the helmet is to minimize the risk of catastrophic focal brain injuries (skull fracture, subdural hematoma) by management of forces during impact. An outer shell will act to distribute these forces over a larger area to avoid localized regions of high load concentration (Di Landro, Sala et al. 2002). In addition, the impact energy is absorbed through deformation of the liner materials, which in turn increase the impact duration thus reducing acceleration and contact forces acting on the head (Mills 2003).

Helmet performance can be evaluated by securing the device onto a surrogate ‘dummy’ head, and measuring linear acceleration of the headform during impact. The results can be scaled and then related to human injury tolerance curves (Section 2.3.1) to estimate the risk

of focal brain injury. Linear acceleration measures provide the basis of nearly all helmet impact standards, which are enforced to ensure products provide a basic level of safety. The use of certified helmets in sport has proven effective in reducing the frequency of focal head injuries such as skull fracture and subdural hematoma (Biasca, Wirth et al. 2002). However, mTBI rates remain high despite the use of helmets. The following sections will review helmet materials, testing standards based on prevention of focal injuries and finally, discuss helmet protection against diffuse mTBI in terms of injury rates of helmeted athletes and helmet related research being conducted to address this issue.

2.4.1 Helmet Construction

The typical helmet parts include the shell, liner and retention system. The shell serves to distribute forces over a larger area and resist puncture and abrasion depending on the thickness and type of materials used. Common shell materials include thermoplastics

(polycarbonate), fiber reinforced plastics, Acrylonitrile-Butadiene-Styrene (ABS), and high density polyethylene. The liner serves to absorb energy and typically consists of polyurethane, polyethylene or expanded polystyrene or polypropylene. The deformation properties of a material can be plastic, elastic or somewhere in between. Plastic deformation occurs when a material changes shape permanently due to the absorption of energy. An elastic deformation rebounds immediately after compression, which returns energy. Most helmet materials will fall somewhere between a plastic and elastic response.

The effectiveness of the foam at absorbing impact is highly dependent on the material properties matching the impact energy (Figure 2-5). For example a strong material such as a high density foam, will have a high energy absorbing capability but at the cost of higher accelerations and localized contact forces (Di Landro, Sala et al. 2002).

Figure 2-5: The effect of differing material properties for absorption of equal energy impacts. Adapted from Newman (2002). Likewise, if the material is of lower density, then excessive deformation will occur and

without sufficient thickness the material will reach its useful compressive limit, termed

‘bottoming out’. Once a material has bottomed out, the remaining energy will propagate

through the material to the tissues below (cranial bone). If critical force (Fc) is reached from

the transmitted forces, the bone may fail and fracture. An ideal material will balance both

deformation and force to manage the impact energy efficiently without excessive forces or

deformation: it would compress the foam to its fullest extent, minimize rebound, and

maintain a constant acceleration throughout the impact (Newman 2002). Materials other than foams have been used in helmets for impact absorption including thermoplastics, engineered compressible structures, and air bladders that attempt to use fluid flow properties to provide rate-dependent material responses (Lamb and Hoshizaki 2009).

Helmets can be divided broadly into two core groups: single impact crash helmets and multiple impact helmets. Crash helmets are designed to withstand a single catastrophic blow and then be replaced. These include bicycle helmets and ski helmets, for instance.

Closed cell expanded polystyrene (EPS) is the typical foam used for crash helmets for its

light weight, good compression qualities, and minimal rebound (Mills 2003). After such an

impact the helmet should be replaced regardless of its physical condition. Multiple impact

helmets are designed to sustain multiple low energy impacts, as well have provided

protection against catastrophic blows, but to a lesser degree than a pure crash helmet. These

include hockey helmets, football helmets, and any sport that typically involves repeated low

energy contacts to the head. The helmet liners used in multiple impact helmets are typically

made from materials that regain their shape. Multiple impact helmets utilize foams such as

expanded polypropelyne (EPP), expanded polyetheleyne (EPE), Vinyl Nitrile (VN) or a

mixture of foams to provide attenuation of both low and medium energy impacts. Although

liners such as EPP and VN recover their shape and are commonly used for repeated impact

helmets, it is important to change the liners occasionally as the material properties will

degrade with multiple impact (Hakim-Zadeh 2002; Hoshizaki, Robertson et al. 2006) and age

(Bishop, Norman et al. 1984; Ouckama and Pearsall 2012).

2.4.2 Focal Injury Prevention: Impact Standardization

The use of certified helmets in sport has nearly eliminated catastrophic brain injuries

such as skull fracture and subdural hematoma (Biasca, Wirth et al. 2002). In order to control

product safety, helmet testing standards have evolved to ensure that all helmets meet the

same test criteria. The standards are typically developed by a committee of experts from

various professional fields such as health, safety, engineering, and manufacturing. Several

committees and associations exist that have developed testing standards specifically for

sports helmets: the American Society for Testing and Materials (ASTM), the National

Operating Committee on Standards for Athletic Equipment (NOCSAE), the Canadian

Standards Association (CSA), International Standards Organization (ISO), and Conformité

Européene (CE) among others. The purpose of the committee is to provide educated

reasoning from various fields of industry and research for the particular requirements of the

protective device ranging from its dimensions, visual field, impact tolerance and mass. The

approval of a testing standard is valuable to the consumer as well; they can feel confident

purchasing a device that has passed national safety regulations. It is important to understand

that testing standards are not interchangeable. If a product complies with one particular

standard (i.e. NOCSAE), there is no guarantee that it will comply with another standard (i.e.

ASTM). The methods used to control the impact to the helmet and calculate risks can vary

significantly. In addition, the surrogate headform, to which the helmet is attached, can be dramatically different in its construction and material properties. Bishop (1993) recorded

impacts of the same energy utilizing various headform types. There were significant

differences in peak acceleration values, which suggested that a uniform pass/fail criteria

based on peak acceleration (G) was not appropriate. Thus, one cannot assume that a

particular standard’s values for impact tolerance are appropriate for another standard.

Another issue is that a headform does not behave as a human head during impact. Unless

the surrogate head has the same weight, mass distribution and deformation qualities as a

human head, one cannot relate the data directly to human tolerance (Hodgson 1991). Even

the method of dropping the headform can significantly change the impact dynamics for

example, when using a fixed monorail system, peak acceleration was 31% greater than the equivalent energy impact provided by a guide-wire system (Bishop 1993).

Due to global markets for equipment providers, a helmet will often have to meet

several standards to be sold in varying markets. For example, the sport of ice hockey has

mature helmet standards, including those from the CSA, ASTM, ISO and NOCSAE. The

standards differ primarily by the headform used and the impact conditions. The CSA, ISO

and ASTM utilize magnesium headforms, while the NOCSAE standard uses a biofidelic

headform with a glycerin filled cavity. Though the headform used by the CSA and ASTM

standard lacks these biofidelic properties they provide more repeatable data (Halstead 2001).

As an example a comparison of the various standards for ice hockey helmets and associated

metrics is presented in Table 2-1. Many standards use variable mass headforms for differing

helmet sizes, thus, for simplicity, only properties medium sized helmets will be presented.

Standard CSA z262.1-09 ASTM F1045-07 ISO 10256:2003 NOCSAE ND030-11m12 Headform Type EN960 Half Face F2220/EN960 Half EN960 Full Face NOCSAE Full Face (circumference) (575mm) Face (J-570mm) (570 mm) (570mm) Headform Mass 4.7 kg 4.7 kg 4.7 kg 4.8 kg Impact Type Guided Monorail Guided Monorail Guided Monorail Guided Wire (uniaxial) (uniaxial) (uniaxial) or (triaxial) Freefall (triaxial)

Impact Velocity 4.5 m/s 4.5 m/s 3.96 m/s 3.46, 4.46, 5.46 m/s (increasing by impact)

Impact Material MEP60 Shore A MEP60 Shore A MEP60 Shore A MEP Acceleration 275 g 300 g 275 g N/A Limit Severity Index N/A N/A 1500 300 (3.46 m/s) Limit 1200 (4.46-5.46 m/s)

Given each sport incurs varying levels of risk and unique mechanisms of injury (i.e.body

checking in hockey, asphalt abrasion in cycling, high speeds in downhill ski racing), each has

developed unique variances in helmet design for management of these specific risks. The helmets must provide adequate protection under all expected conditions, thus factors such as temperature, humidity, and impact location are often included in each standard to conform to the conditions unique to the sport.

2.4.3 mTBI Injury Rates in Helmeted Sport

Despite the effectiveness of helmets at reducing focal injuries an alarming number of minor traumatic injuries continue to occur. The US Center for Disease Control and

Prevention (CDC) estimated the occurrence of mTBI’s resulting from these activities at approximately 300,000 per year in the United States (Thurman, Branche et al. 1998). This regularly cited statistic, however, only includes TBI’s in which a loss of consciousness (LOC) was reported. More recent research shows that, taking into account the percentage of TBI’s that include a LOC, the total number of sport related TBI’s could be as high as 1.6-3.8 million per year (Langlois, Rutland-Brown et al. 2006). This estimate may still be low as many head injuries can go unnoticed or may be unreported by players in order to continue play. In a study of Canadian Interuniversity Sport (CIS) athletes, Delaney et al. (2002) found discrepancies in the reported rate of concussion and the number of athletes who had experiences symptoms of the injury. Of the 328 football and 201 soccer players that responded to a questionnaire, 70.4% of the football players and 62.7% of the soccer players experienced symptoms but only 23.4% and 19.8% of the respective players actually realized they had suffered the injury. With the calculation of rates of injury, particularly with concussion, there is a risk that the occurrence may be under reported. Athletes may not realize they are concussed or may not want to be pulled from game play or risk the repercussions of admitting to injury. This behaviour was reported recently in study of

Canadian Interuniversity Sport (CIS) hockey players (Echlin 2012). Ideally, these attitudes will be adjusted in the sporting world as more attention is brought upon the long term risk of multiple concussions and with further education to coaches and training staff to aid in detection of concussion injuries.

Tracking the rate of concussion in sport has been achieved through varying methods

such as the NCAA’s Injury Surveillance system (ISS), hospital records such as the National

Hospital Ambulatory Medical Care Survey (NHAMCS), Canadian National Health

Population Survey, or more recently the through the use of instrumented helmets (Gwin,

Chu et al. 2006). The NCAA utilized an Injury Surveillance System (ISS) throughout 1988-

2004 to track injury records of 15 collegiate sports. This database provides some invaluable data in terms of head injury statistics among this population. For example, over 9000 concussions were reported during the 16 years of surveillance. American football had both the greatest number of participants and total number of reported concussions. To account for variances in team sizes the authors calculate an injury risk per-participant or athlete- exposure (AE) rate. An athlete-exposure is defined as one athlete participating in one practice or contest where he or she is exposed to the possibility of injury. Of all the contact and non-contact sports monitored, ice hockey topped the list, with women in particular sustaining the highest rates at 0.91 concussions per 1000 AE. Men’s ice hockey had the next highest regular-season injury risk at 0.41 concussions per 1000 AE. (Hootman, Dick et al.

2007; Daneshvar, Nowinski et al. 2011). More recently, in a study of CIS hockey players, this rate was substantially higher, with women at a risk of 14.93/1000 AE and men, at 7.5 /1000

AE during regular season play (Echlin, Skopelja et al. 2012). Ice hockey is a high-speed sport played on a hard ice surface with a fast moving projectile puck. There are numerous head injury risks in ice hockey, including contact with another player, the ice, the surrounding boards or the puck. To further increase the risk of head injury, fighting is common in many amateur and professional ice hockey leagues.

2.5 Improving the Helmet: Current Research

The issue of mTBI rates in helmeted sports and the long-term effects of multiple mTBI injuries have motivated researchers to seek numerous approaches to reduce these injury types. At the highest level of importance should be prevention of the impact all together. In high risk contact sports (i.e., hockey and football), the prevention of dangerous hits has been addressed through the development of several intervention strategies including behavioural education of players coaches, parents and spectators toward the severity of mTBI injuries, and elimination of high risk body checking through awareness (e.g., Head’s

Up Hockey, STOP) and infractions (e.g., illegal check to the head, hit from behind, elbowing).

Manufacturers of protective equipment for sports have been challenged to improve

helmets to reduce concussion risks (McIntosh, Andersen et al. 2011). The helmet represents only one part of injury prevention, and is ultimately limited in its ability to restrict brain

motion due to relative movement of the brain inside the skull (Section 2.2.2). The magnitude of linear acceleration acting on the head is reduced during compression of padding materials;

however, rotational acceleration, related to diffuse brain injury (Section 2.2.3.2) may not be

mitigated as effectively without significant changes to the fixation of the helmet to the head.

Part of the challenge in incorporating these measures into helmet performance is the relative

unpredictability of mTBI injury based solely on kinematic responses of the head (Guskiewicz

and Mihalik 2011). Unlike detectable bone fractures and focal lesion injuries that current

helmet standards are based upon, mTBIs may not exhibit visible brain injury and are

challenging to assess and thus difficult to predict. The majority of sport-based mTBI

research has thus been based on post-hoc analysis of known cases of mTBI. These events

can be analyzed for the associated head kinematics using impact reconstruction (Pellman

2003; Viano, Pellman et al. 2006) or prior instrumentation of athletes using helmet sensors

(Gwin, Chu et al. 2006; Mihalik, Greenwald et al. 2010). From these large datasets, the

resulting injury corridors have been proposed stressing the importance of both linear and

angular acceleration (Guskiewicz and Mihalik 2011), though it is strongly cautioned that

these measures should not be used to predict mTBI symptom severity including

neurophysiological function.

In order to further understand neurovascular injury risk factors, more

comprehensive measures, including estimated internal strain of brain tissue, are needed.

Multiple researchers have applied FEA techniques to predict the brain’s inertial responses to helmeted impacts in an attempt to correlate these measures to injury occurrence (Viano,

Casson et al. 2005; Marjoux, Baumgartner et al. 2008; Forero Rueda, Cui et al. 2011;

McAllister, Ford et al. 2012; Post, Oeur et al. 2013). Although FEA provides a useful tool to

assess brain motion and internal tissue stress or strain, it is limited by the accuracy of the model to simulate biological tissue. One such limitation is the validation of contact forces at the interface of the helmet and head. In regards to modeling the helmet, highly compressible foams are difficult to simulate (Mills 2007). The model predictions would benefit from validation of the final product assembly in order to ensure the model is accurately representing the helmet materials. Obtaining these measures can be challenging as typical

force measurement systems fail to provide sufficient sampling speeds, thinness, flexibility

and durability for impact testing. In 2008, initial methods were developed for measurement

of contact forces in ice hockey helmets (Ouckama and Pearsall 2008; Ouckama and Pearsall

2010) using an array of OEM flexible force sensors integrated with high speed data

acquisition. Independently of this work, Rigby & Chan (2009) developed a similar method

using the same force sensors in an automotive context. Rigby’s work focused on the

application of average pressure response of many models of motorcycle helmets as a generalized measure to assess the biofidelity of skull fracture criterions. Rigby did not address differences in contact forces or load dispersion properties based on differing helmet materials, conditioning or impact types. These measures remain unknown, and in the context of ice hockey, they are valuable for helmet research both in terms of finite element model validation and for the identification of highly localized forces (Bishop and Arnold 1993). To this end, the objectives of the following three chapters in this thesis are: to develop and validate a new method to quantify localized impact load distribution between the helmet and head contact surface (Chapter 3), to demonstrate the use of this method in comparison to standard acceleration outcome measures in order to assess ice hockey helmet-head contact parameters during traditional vertical drop testing (Chapter 4), and during high-speed puck impacts (Chapter 5). This work will contribute to the future of helmet research by providing a measurement method to address the possibility of high concentrations of pressure at the contact site that may go undetected by current standardized measures of linear acceleration.

Additionally, the outcomes from this measurement system could be beneficial in providing empirical data of the helmet-head contact parameters for validation of finite element models incorporating this interface.

CHAPTER 3: EVALUATION OF A FLEXIBLE FORCE SENSOR FOR

MEASUREMENT OF HELMET FOAM IMPACT PERFORMANCE

Ryan A. Ouckama, David J. Pearsall

Reprinted with permission from Elsevier, Publisher of the Journal of Biomechanics:

Ouckama, R. and D.J. Pearsall. Evaluation of a flexible force sensor for measurement of

helmet foam impact performance. Journal of Biomechanics. 44:904-909, 2011.

3.1 Preface

The manuscripts of this thesis were motivated by a study conducted by Patrick

Bishop and James Arnold at the University of Waterloo in 1993. In the article Bishop pointed out that in the sport of ice hockey, focal head injuries such as skull fracture and subdural hematoma could be caused by high-mass low-velocity impacts such as striking the head on the ice, but also by means of low-mass high-velocity impacts such as being struck with the puck. Bishop pointed out that the helmet testing standards for ice hockey at that time evaluated the first condition of high-mass low-velocity by means of a vertical drop test, however, the latter condition is not addressed by the testing standards despite the inherent risk to high levels of localized loading, particularly at the temporal region of the skull.

Indeed, the findings of their study demonstrated that localized loading magnitudes of certified helmets could exceed the limits of bone fracture, whereas conflictingly, the severity index measures (calculated from linear acceleration) predicted low levels of injury risk.

Helmet standards have been modified since the publication of this work, however, the issues raised by Bishop and Arnold’s work remains unaddressed in the testing protocols.

Other authors have pointed out the importance of evaluating puck impacts to hockey helmets (Halstead, Alexander et al. 2000), yet no further research has investigated localized loading characteristics during such impacts in large part, due to the technical challenge of obtaining these data. Bishop utilized pressure indicating films (e.g. Fuji

Prescale®, Fujifilm, Mississauga, Ontario), a material that when compressed, causes the rupture of micro-bubbles and triggers a change in color intensity related to the application force. The materials require careful post processing to minimize error in interpretation and fails to provide a time-series measure for analysis of the pressure profile curves at each point.

A modern high-speed data-acquisition based solution is required to further interpret the

time-history of pressure generation at the contact site and to relate these values directly to corresponding metrics of acceleration. The following published manuscript will present a cost effective system developed to measure load distribution in helmet foams.

Ouckama, R. and D.J. Pearsall. Evaluation of a flexible force sensor for measurement of helmet foam impact performance. Journal of Biomechanics. 44:904-909,

2011.

This manuscript has been reprinted with permission from Elsevier, publisher of the Journal of Biomechanics. The paper is presented as it was published with the exception of formatting of figures and tables to comply with the McGill University e-thesis formatting guidelines.

3.2 Abstract

The association between translational head acceleration and concussion remains

unclear and provides a weak predictive measure for this type of injury, thus, alternative

methods of helmet evaluation are warranted. Recent finite element analysis studies suggest

that better estimates of concussion risk can be obtained when regional parameters of the

cranium, brain and surrounding tissues are included. Lacking however are empirical data at

the head-helmet interface with regards to contact area and force. Hence, the purpose of this study was to evaluate a system to capture the impact force distribution of helmet foams.

Thirteen Flexiforce® sensors were arranged in a 5cm x 5cm array, secured to a load cell.

Three densities of foam were repeatedly impacted with 5J of energy during ambient (20°C) and cold (-25°C) conditions. RMS error, calculated relative to the global force registered by the load cell, was <1.5% of the measurement range during individual calibration of the

Flexiforce® sensors. RMS error was 5% of the measured range for the global force estimated by the sensor array. Load distribution measurement revealed significant differences between repeated impacts of cold temperature foams for which acceleration results were non-significant. The sensor array, covering only 36% of the total area, possessed sufficient spatial and temporal resolution to capture dynamic load distribution patterns.

Implementation of this force mapping system is not limited to helmet testing. Indeed it may be adopted to assess other body regions vulnerable to contact injuries (e.g., chest, hip and shin protectors).

3.3 Introduction

Although the majority of head injuries are caused by falls and motor vehicle accidents (Faul, Xu et al. 2010), athletes involved in high speed, contact sports such as

football and ice hockey are exposed to a relatively high risk of concussion (0.37/1000 and

0.47-0.91/1000 injury rate per athlete exposure (Hootman, Dick et al. 2007) due to the

frequent collisions inherent to these sports. In the United States alone there are an estimated

1.6-3.8 million sport-related minor traumatic brain injuries every year (Langlois, Rutland-

Brown et al. 2006) despite the widespread use of modern protective headgear. The

evaluation of helmet performance is governed by several international standards committees

(ASTM, CSA, ISO, NOCSAE). The standards developed by these committees rely primarily

on minimizing the metric of global head acceleration (ISO 2003; ASTM 2007; CSA 2009)

during a free fall impact onto a hard surface. The pass/fail threshold for these standards

(275-300 g’s) originates from the work of Gurdjian et al. (1966) on the tolerance of the

human skull to blunt impacts. These criterion have been effective as supported by the near

complete elimination of skull fractures and related fatal brain injuries in American football

since the mandatory use of helmets certified to impact acceleration standards (Mueller

1998). However, there is no strong clinical evidence to support that these helmets can reduce

the incidence of concussion (McCrory, Meeuwisse et al. 2009). Indeed, concussion rates in

both amateur and professional levels of contact sports remain high despite the use of an

approved helmet (Langlois, Rutland-Brown et al. 2006; Tommasone and Valovich McLeod

2006).

Simulation studies using 3D finite element analysis (FEA) have repeatedly

demonstrated that global head acceleration criterion measures per se cannot project the

nature and severity of resulting neurocognitive injuries. Though the link between translational head acceleration and concussions has been scrutinized extensively, ultimately is has been found to be a poor predictor for injury risk (Viano and Pellman 2005; Greenwald,

Gwin et al. 2008; McCrory, Meeuwisse et al. 2009). Understanding the dynamics at the small

region of contact between the colliding object and helmeted head may be more relevant due

to the directly induced stresses. While these stresses are likely insufficient to cause bone fracture, they can induce transient focal cranial deformation that in turn propagates stress wave pulses to the underlying cortex, particularly during a lateral impact to the thin squamous temporal bone (Zhang, Yang et al. 2001). Hence, inclusion of loading force

distribution to the cranium could enhance estimation of local tissue distress (i.e., Von Mises

stress, principal and shear strain) related to injury (King 2000; Zhang, Yang et al. 2004;

Kleiven 2006; Hardy, Mason et al. 2007; Marjoux, Baumgartner et al. 2008).

One of the primary functions of the helmet is to distribute an impact load across a

sufficiently large contact area and duration to shield the underlying bone and neurovascular

structures from mechanical distress. Using FEA, (Forero Rueda, Cui et al. 2009)

demonstrated that foams offering greater contact area coincided with lower peak cranial

stresses during dynamic testing. Arguably an empirical measure of impact load distribution

is warranted so as to quantify the dynamic loads across the helmet – head localized impact

site. Prior study by Bishop & Arnold (1993) demonstrated the feasibility to capture peak

force dispersal during helmeted headform impacts. Using pressure sensitive films, these

authors observed a poor relationship between the risk associated with headform acceleration

and the corresponding risk due to high peak focal forces. Unknown however, due to

limitations of the films, were the impact durations and loading rates.

To capture both temporal and spatial loading responses, various existing force sensor

matrices may be considered. However, few fit the specific requirements for impact

measurement. Such a system requires high speed measurement, high loading range, sufficient

spatial resolution, flexibility to conform to irregular curved surfaces, and ideally a low cost.

The purpose of this study was to evaluate an array of discreet flexible force sensors in order

to achieve high speed capture of impact force distribution of helmet padding materials. At

this preliminary stage of study, in order to minimize the confounding factors of helmet

geometry and shell materials, only flat samples of EPP helmet foam were tested to validate the global force predicted by the array of individual sensors.

3.4 Methodology

3.4.1 System Design

A 16-channel amplifier was designed and constructed based on the recommended drive circuitry for Flexiforce® sensors (Tekscan 2009). The Flexiforce sensor is constructed

from two polyester film layers which surrounding a printed circuit consisting of two

conductive strips and the active sensor region. The active region (9.5 mm diameter circle) consists of a material which varies its electrical resistance inversely to the applied force.

Sixteen sensors (Flexiforce® A201-100 Tekscan, Boston) were wired to a single compact amplifier using low cost PC ribbon cable and IDE connectors. The amplifier was assembled using a custom printed circuit board and surface mount electronics to minimize size and weight. Amplifier gains were adjusted to provide the maximum range of 1000N (~14MPa based on sensor area). The amplified signal was output to a PC USB data acquisition device

(DAQ) (NI USB-6210, National Instruments, Austin, TX) where data were sampled at

15kHz per channel at 16-bit resolution. The sensor response time to an impact is within 5 microseconds (Tekscan 2009), thus it is capable of high (15kHz) sample speeds. Amplifier

power requirements were low enough to be provided from the USB connection, allowing for

system portability when used with a laptop computer. A Labview® virtual instrument (VI)

provided live display, calibration and file saving features.

3.4.2 Calibration

A material testing machine (Sun 1000, Galdabini, Italy) was used to apply a controlled force to each individual sensor. A precision milled aluminum cylinder

(d=9.53mm) acted as a load transfer device between the material testing machine and

Flexiforce® sensor to ensure contact with only the active sensing area. A 1.5 mm thick wax sheet was adhered to the cylinder surface to ensure even surface contact with the

Flexiforce® sensor. A 20kN range load cell (Kistler 9215M113, New York) was mounted at the base of the material testing machine to provide synchronized force output for calibration of the voltage signal recorded from the Flexiforce® sensors. To maximize sensor accuracy, an initial conditioning phase was conducted (Tekscan 2009). A static load of 110% of the maximum measurement range for the respective sensor was applied using the material testing machine and held for 60 seconds. The conditioning process was repeated an additional time. Each sensor was then subject to a ramp-increasing load from 0 to 1000N at the machine’s maximum speed (750N/s) to simulate the dynamic nature of the impact measurements. Five calibration cycles were recorded for each sensor. Using Matlab® (The

Mathworks, MA, USA) for post processing, a third order polynomial was calculated to best fit the plot of all five calibrations together, while forcing an intercept with [0,0]. Calibration trial force data were low pass filtered at 10 Hz using a 4th order digital Butterworth filter within Matlab®. Only measures during the loading phase were used for sensor calibration.

3.4.3 Foam Testing

During impact testing, signals were recorded concurrently from the array of force sensors, impact accelerometer and load cell by the 16-channel DAQ. The apparatus for foam testing is presented in Figure 3-1. Thirteen Flexiforce® sensors were arranged in a square

Figure 3-1: Foam testing apparatus using a guided monorail fitted with a 5 kg 73 mm radius hemispherical impactor. An array of 13 Flexiforce® sensors recorded load distribution.

array measuring 5x5cm. Sensors were arranged on an acetate template and secured using

double sided tape. The acetate template was then adhered directly to the surface of a 20kN

load cell. The load cell was mounted to the base plate of a guided monorail impact system

used for helmet testing. A steel hemispherical impactor (mass=4.977 kg, r=73 mm, Cadex

Inc, St-Jean-sur-Richelieu, Quebec) was dropped to provide 5J of impact energy to 10cm x

10cm x 1.5cm foam samples placed on the force array. Three different densities (54.8, 69.0 and 86.1) kg/m3 of expanded polypropylene (EPP; a commonly used multiple impact helmet

foam) were manufactured (Polymos Inc, Terrasse-Vaudreuil, Québec) and tested to both

demonstrate the system performance and to assess the differences in load distribution due

varying material properties. Nine samples of each foam density were evaluated at both

ambient (20°C) and cold (-25°C) conditions, with each sample exposed to two consecutive

repeated impacts within 90 seconds. A zinc cream was applied to the hemispherical impactor

prior to each trial to mark the contact area on the foam. Immediately following impact the

contact diameter was recorded using a digital caliper. Acceleration was recorded at the CoM

of the hemispherical impactor and global force was measured using a 20 kN load cell at the

base of the monorail system.

3.4.4 Statistics

Descriptive statistics (Mean, Standard Deviation, RMS error, and percent error) were calculated for each calibration trial. System error was assessed by calculation of RMS error

and Bland Altman plots (Bland and Altman 2010) between the Flexiforce® sensors and

force plate outputs for each compression trial. In order to validate the Flexiforce® array

against the global force measured at the 20 kN load cell, the Matlab® Surface Fitting

Toolbox was used to best-fit a interpolated 3-D surface to the force-position data collected

at each time interval (0.0006s). An approximate integration of this surface was calculated to

estimate the total force at that instant in time. Descriptive statistics were calculated for each

foam impact condition. Additionally a 3×2 repeated measures ANOVA was calculated to

determine if any significant differences existed between testing conditions using Statistica 9

(Statsoft, Inc. Tulsa, OK, USA).

3.5 Results

3.5.1 Nomenclature

Focal Force - Point force measurements ( Newtons ) obtained from an array of thirteen

individual Flexiforce® sensors.

GFLC - Global force (N) registered by the uniaxial load cell; used as a standard for reference

and for calculation of RMS error

GFFF - Global force (N) estimated by the approximate integral (sum of interpolated forces)

of a 3D surface fit to the 13 focal forces

GFACC - Global force (N) estimated by F=ma, where m represents the mass of the impactor

(4.977 kg) and a represents the calibrated accelerometer signal (g) * 9.81.

3.5.2 System Calibration

Repeated (n = 5) calibrations of the Flexiforce® sensors (n = 13) display high repeatability and linearity both within and between sensors (Figure 3-2A). Average R2 was

0.999±0.001. RMS error across all sensors was 6.4±2.7 N (Table 3-1).

Table 3-1: Descriptive statistics for all 13 sensors used during foam impact testing. Each sensor was calibrated 5 times and the resulting force-to-voltage output was fit with a third order polynomial.

Max Min Mean St dev (n = 13) (n = 13) Peak Calibration Force (N) 1017.0 828.0 982.0 68.7 RMS Error (N) 12.6 3.1 6.4 2.7 R2 0.999 0.997 0.998 0.001 % Error (RMS/Max Force) 1.2% 0.4% 0.7% 0.3%

All thirteen sensors had an average RMS error < 1.5% of the total measurement range.

Several specific load intervals across the measurement range were selected to compare the

load cell output to the Flexiforce® output directly (Figure 3-2B). From this data we can

conclude that 95% of the measurement error is contained within ±1.96 standard deviations

(±11.9 N) of the mean. This results in 95% of the error accounted within ±1.5% of the total

measurement range.

Figure 3-2: (A) Comparison of the Flexiforce® sensor calibrated up to 800 N with respect to the output from a calibrated load cell. Five repeated calibrations are superimposed for thirteen separate sensors. (B) Bland–Altman plot of sensor error shows a tendency toward increase in error with increasing load.

3.5.3 Array Measurement Validation

For each foam sample tested, the approximate integral (sum of all interpolated forces) was calculated under a 3D surface passing though the force-position data of the 13

Flexiforce® sensors for each point in time (10ms, ~150 frames). Across all EPP foam

2 samples the average correlation coefficients (R ) relative to GFLC were 0.997 ± 0.001 for

GFACC and 0.976 ± 0.017 for GFFF. Average RMS error for GFACC was 37.1 ± 9.6N. Percent

error (RMS / Max Force) was 1.9 ± 0.5% for GFACC with 95% of the error accounted for

within ±61.6N. GFACC tended to overestimate GFLC. This small difference may be attributed

to some force dissipation through the guided monorail impactor’s support armature which

would not be registered by the load cell. In many cases, the contact area went beyond the

sensing zone. Similarly, the load distribution area may well have expanded beyond both the

force array’s perimeter and the contact perimeter (Figure 3-3). Indeed, the high perimeter

force, particularly in the smaller diameter contact areas, suggests this may be the case and

hence the GFFF were more prone to underestimate GFLC. Extrapolation of the 3D surface

fit was not attempted. For the foams where the mean perimeter forces were low (<20 N e.g.,

low density, ambient foam), the RMS error of GFFF was 81.0 ± 24.2 N or 5.0±1.7%

(expressed as an average percent error = RMS/Max Force). When error is plotted across all

trials of the 5J low-density, ambient condition (Figure 3-4), 95% of the error was contained within ±164.9N. Over the average maximum range of 1642N, 95% of the error was contained within ±10% of the total range. A visual representation of load distribution is presented for the conditions of density and temperature (Figure 3-5).

Figure 3-3: (A) Average contact diameter of the hemispherical impactor with foam sample for each condition relative to the sensor array width and (B) mean force at the sensor array’s perimeter with respect to the contact diameter of hemispherical impactor.

Figure 3-4: Bland–Altman error plot between the GFFF and GFLC variables. 95% of the error is contained within ± 164.9 N of the mean.

Figure 3-5: Average (n = 9) peak load distribution for 3 densities of EPP foam during first impact by a 5 kg hemispherical impactor. Ambient (20°C) and cold (-25°C) temperatures were tested for both initial and repeated impact. Peak focal force (72.7 N) was observed for the cold sample of the highest density EPP foam during the initial impact.

3.5.4 Foam Testing Results

Peak acceleration for conditions of foam density, temperature and repeated impact are presented (Figure 3-6A). Using a 3×2 repeated measures ANOVA, significant between-

subject effects exist for density (F2,48=1004.5, p<0.01), temperature (F1,48=593.0, p<0.01) and the interaction of density*temperature (F2,48=39.2, p<0.01). Within-subject effects exist for the repeated measure of impact (F1,48=918.2, p<0.01), and the interaction of impact*density

(F2,48=6.0, p<0.01) and impact*temperature (F1,48=675.3, p<0.01). There was not a significant 3-way interaction of impact*density*temperature (F2,48=0.416, p=0.66).

Figure 3-6: (A) Average (n = 9) peak acceleration of a 5 kg hemisphere on impact to three densities of EPP foam at ambient and cold temperatures for initial and repeated 5 J impact. (B) Average (n = 9) peak focal force measured using an array of 13 Flexiforce® sensors instrumented under the impacted foam. Error bars represent standard deviation. *Denotes significant difference.

Peak focal force for conditions of foam density, temperature and repeated impact are presented (Figure 3-6B). Between-subject effects were significant for factors of density

(F2,48=195.0 p<0.01), temperature (F1,48=45.1 p<0.01) and the interaction of

density*temperature(F2,48=27.5 p<0.01). Within-subject contrasts were significant for

impact*density (F2,48=11.8, p<0.01), impact*temperature (F1,48=273.5, p<0.01) and impact*density*temperature (F2,48=4.7, p=0.014). There was, however, a non-significant within-subject contrast for impact alone (F1,48=0.05, p=0.821). Post-hoc analysis

(Impact*Density*Temperature, Tukey HSD) revealed significant differences for 1st to 2nd impact for all conditions but low density cold foam (p=0.41). Significant differences existed between mid and high density*temperature, however, not for low density*temperature as indicated in Figure 3-6B.

3.6 Discussion

The purpose of this study was to evaluate a measurement system that was capable of high speed capture of impact force distribution during the evaluation of helmet padding materials. The use of multiple individual sensors enabled for a custom configuration and responded with a high degree of linearity and low error. Individual sensors accounted for

95% of the error within ±1.5% of the total range, exceeding the manufacturers specifications of ±3% (Tekscan 2009) and providing acceptable performance characteristics for flexible force sensors which typically do not provide low levels of error. The improved experimental error relative to the manufacturer’s published error could be due to the dynamic calibration method employed using a high loading rate and synchronized load cell output.

Using multiple force sensors in combination, the 5 cm x 5 cm array sparsely populated with 13 sensors (covering 36% of the 25 cm2 array area) possessed sufficient spatial and temporal resolution to capture dynamic load distribution patterns. As evidence, for those impacts trials where force was contained solely within the boundary of the sensor array, the sensor RMS error was within 5% of the total measurement range and demonstrated high dynamic correlation (R2=0.976 to 0.997) to the reference load cell. These

findings indicate that this instrumentation technique accurately captured the impact force

characteristics of the foams tested. Limitations became evident when the impact distribution

exceeded the sensor array boundary. This was especially true with foams of increased

density and lower temperatures. Though force RMS errors increased, strong dynamic

correlations were preserved. To address the latter limitations, future designs of this system

need incorporate a higher sensor count to both increase the spatial resolution and the total

coverage area in order to minimize these errors.

The use of global acceleration measures to evaluate foam impact performance

provides a gross estimate of the attenuation behavior. However, the findings clearly

demonstrated that these estimates cannot predict focal contact forces. For example, during

cold conditions impacts (Figure 3-6) no differences in peak acceleration measures were

noted between impacts whereas significant peak focal force differences for both medium

and high density foams impacts were found. These results demonstrated that dynamic force

mapping can discern mechanical differences in impact properties that global acceleration

cannot. Some improvements to this system have been noted. Due to the physical location

of the sensors some aliasing may have resulted when fitting the 3D surface (Figure 3-5).

When the surface was quadratically interpolated between actual measurement points, it was

common to have a 2 cm2 ‘hotspot’ at the central locus. With an increased sensor count, and alignment to the same coordinates as the interpolated grid, we expect to see some improved surface fitting which may reduce overall error. Additionally, the system is limited to measurement of the normal force only.

Several potential industrial and clinical applications of this testing modality exist.

First, dynamic force distribution properties may be used to identify padding materials for

optimal helmet performance. In part this can be achieved by maximizing the foam contact

area. Cui et al. (2009) demonstrated through FEA that a functionally graded foam provided greater contact area, particularly at low energy impacts which reduced overall acceleration.

The reported system in this study can directly estimate contact area; thus it can assist manufacturers in identifying padding foam materials best suited for their products. Second, given the temporal and spatial resolution of this system, higher fidelity information can be fed into finite element modeling to simulate brain injury mechanisms. Acceleration values are currently used in FEA, however, with the additional input of the force distribution over the contact area, the change in applied surface pressure, or stress, could be calculated thus improving the tissue simulation. Other potential applications of this system include mounting within a helmeted headform and subjecting it to high speed projectile impact as well and typical industry drop testing in order to build up a database of expected contact pressures based on material choice and overlying shell geometry. Lastly, implementation of this force mapping system is not limited to helmet testing. Indeed it may adopted to assess other body regions vulnerable to impact contact injuries (e.g., chest, hip and shin protectors).

3.7 Acknowledgements

This study was supported by the Industrial Innovation Scholarship from the Natural

Sciences and Engineering Research Council of Canada (NSERC), Le Fonds Québécois de la

Recherche sur la Nature et les Technologies (FQRNT) and the Bauer Hockey Corp. in St-

Jerome, Quebec.

CHAPTER 4: IMPACT PERFORMANCE OF ICE HOCKEY HELMETS: HEAD

ACCELERATION VERSUS FOCAL FORCE DISPERSION

Ryan Ouckama, David Pearsall

Reprinted with permission from Sage Publications:

Ouckama, R. and D.J. Pearsall. Impact Performance of Ice Hockey Helmets: Head

Acceleration versus Focal Force Dispersion. Proceedings of the Institution of Mechanical Engineers, Part P:

Journal of Sports Engineering and Technology. 226(3/4): 185-192.

4.1 Preface

Following the successful development of an accurate measurement system for the

capture of load distribution during foam impact (Chapter 3), this system was used to measure localized impact dynamics in conjunction with the standardized ice hockey helmet drop test. This second manuscript was motivated by the need to demonstrate that the measurement system could be applied successfully in the context of helmet testing without failure. The standardized impact test of ice hockey helmets is well established and utilizes a pass/fail criterion based on gross headform acceleration. This manuscript demonstrates not only the feasibility of the system to capture helmet-to-head impact dynamics during these drop tests, but also presents a novel dataset containing synchronized measures of global acceleration, global contact force and localize load distribution. This data will also be used for comparison to later evaluation of localized low-mass high-velocity impacts from the puck.

Ouckama, R. and D. Pearsall (2012). "Impact performance of ice hockey helmets: head

acceleration versus focal force dispersion." Proceedings of the Institution of

Mechanical Engineers, Part P: Journal of Sports Engineering and Technology

226(3/4): 185-192.

This manuscript has been reprinted with permission from Elsevier, publisher of the

Journal of Biomechanics. The paper is presented as it was published with the exception of

formatting of figures and tables to comply with the McGill University thesis formatting

guidelines.

4.2 Abstract

Modern sport helmets certified to various international safety standards have

virtually eliminated the incidence of cranial fracture and fatal brain injury in contact sports

(Mueller 1998); however, the occurrence of diffuse brain injuries (mTBI) are still prevalent

(Langlois, Rutland-Brown et al. 2006); (Tommasone and Valovich McLeod 2006). Local contact mechanics between the colliding surface (helmet/head) need to be considered as global measures of acceleration are insensitive to load distribution measures which are indicative of helmet performance. The purpose of this study was to demonstrate the ability to capture localized load distribution response between the helmet and headform and to examine factors that may influence these measures. Twenty-five flexible force sensors were arranged in a 5×5 array about 3 impact sites (front, side, rear) of a 575 mm EN960 headform. Test factors included helmet model (5), impact location (3) and temperature

(21°C, -25°C) as well as repeated impacts (3). Testing procedure followed the CSA Z262.1-

09 standard at the defined locations. Average error calculated during sensor calibration was

2.8±1% with an R2 value of 0.987 ± 0.009. As expected, peak global force correlated well to

peak acceleration (R2=0.98) but weakly corresponded to peak focal force (R2=0.22).

Furthermore, both load distribution magnitudes and patterns were found to vary substantially with mixed effects between helmet models, impact locations, and temperatures.

Given these findings, this novel approach may be used to quantify local contact mechanics between the colliding surface / helmet / head (Marjoux, Baumgartner et al. 2008; Forero

Rueda, Cui et al. 2009).

4.3 Introduction

International helmet impact safety standards (ASTM (ASTM 2007), CSA (CSA

2009), NOCSAE (NOCSAE 2004), ISO (ISO 2003)) primarily utilize linear acceleration

measures from surrogate headform/helmet drop tests as the criterion measure to assess a

helmet’s ability to reduce force transmission to the head. The commonly accepted linear

deceleration threshold of between 275–300g stems partially from work assessing human

cranial bone tolerance to fracture resulting from blunt impact (Gurdjian, Roberts et al. 1966).

Modern sport helmets certified to these impact safety standards have virtually eliminated the

incidence of cranial fracture and fatal brain injury in contact sports (Mueller 1998); however,

diffuse brain injuries (concussion or mTBI) are still prevalent (Langlois, Rutland-Brown et al.

2006); (Tommasone and Valovich McLeod 2006). For instance, in the United States there are an estimated 1.6–3.8 million sport related mTBIs each year (Langlois, Rutland-Brown et

al. 2006). The issue of diffuse brain injury in sport is a major public health concern that must

be addressed.

Linear acceleration criterion for helmet impact tolerance has proven effective in the

reduction of skull fracture and blunt force trauma; however, it is alone a poor predictor for

mTBI risk (Viano and Pellman 2005; Greenwald, Gwin et al. 2008; McCrory, Meeuwisse et

al. 2009; Rowson, Duma et al. 2012). To understand neurovascular injury risk factors, more

comprehensive measures are needed. In response, several researchers have used finite

element analysis (FEA) techniques to predict the brain’s inertial responses (Marjoux,

Baumgartner et al. 2008; McAllister, Ford et al. 2012) to better account for the

heterogeneous mechanical properties of the head and brain. In addition to more biofidelic

intrinsic estimates of the mechanical properties of the head and brain, enhanced accounting

for the extrinsic impact parameters are needed, such as details of the local contact mechanics

between the colliding surface / helmet / head (Marjoux, Baumgartner et al. 2008; Forero

Rueda, Cui et al. 2009). For instance, Forero Rueda and associates (Forero Rueda, Cui et al.

2009) found that foam contact area has been related directly to the reduction of peak head

impulse and thus suggested this as an important input variable for FEA models to optimize

helmet performance. While the contact area and force was computationally estimated in this

latter study, there has been limited work conducted to directly measure localized contact

surfaces empirically in a dynamic manner. Pressure sensitive films can provide a snapshot of

force magnitude distribution (Bishop and Arnold 1993) but lack any temporal history

response. Recent developments in combining multiple, flexible yet robust sensors with high-

speed data acquisition hardware have been validated in the testing of helmet padding

materials (Ouckama and Pearsall 2011). Hence, the purpose of this study was to demonstrate

the ability to record localized load distribution response between the helmet and headform

during traditional linear impact testing and identify how these distributions may be modified

by factors of material density and geometry (based on differing helmet models), temperature

and repeated impact.

4.4 Methodology

4.4.1 Instrumentation

Twenty-five flexible force sensors (Flexiforce® A201-100, Tekscan, Boston, MA, USA) were

individually calibrated by way of a guided linear drop of a 4.7 kg mass directly onto each

sensor (Figure 4-1). A vinyl cylinder (d=9.5 mm, h=13 mm) was adhered to the impactor in

order to isolate sole contact to the sensor’s active area (d=9.5mm). The impact attenuation properties of the vinyl cylinder were selected so that the contact time of the calibration loading curve matched the contact time (~12ms) and peak point force (1600N) of a typical

helmet acceleration profile at 4.5 m/s. A piezoelectric load cell (Type 9215M113, Kistler,

Winterthur, Wulfingen, Switzerland) was fixed at the base of the drop tower to record impact forces. Ten calibration trials were recorded for each of the 25 sensors. The

Flexiforce® sensor voltage was correlated to the load cell output and a best fit 3rd order polynomial was then calculated to create the calibration equations. Data were collected using a 32 channel acquisition device (NI-9205, National Instruments, Austin TX, USA) sampling at 9 kHz. The sample rate was limited by the maximum speed of the acquisition device (250 kS/s). A 2.5 kg EN960 575 mm magnesium half-headform (Cadex Inc, St-Jean-sur-

Richelieu, QC, Canada) was instrumented with a linear accelerometer at the center of mass.

The headform was positioned to three different impact locations (front, side, rear) as defined by Canadian ice hockey helmet testing standards (CSA z262.1-09). The 25 Flexiforce® sensors were then adhered to the headform using double-sided tape in a 5×5 array centered about each impact location (Figure 4-2). Sensor grid spacing was 2 cm with active region of the 25 sensors covering 28% of the 64 cm2 area. The bare instrumented head headform was impacted directly into a MEP (Modular Elastomeric Polymer, Shore 60A) pad three times

(collected pre and post-testing) to serve as reference condition and verify repeat performance of the net force calculated from the sensor array. Five different models of commercially available ice hockey helmets, representing various masses, material types, shell geometries and padding orientations were obtained for testing (Table 4-1). Test factors included helmet model (5) and temperature (21°C, -25°C) for the repeated measures of location (front, side, rear) and impact (3). Testing procedure followed the CSA Z262.1-09 standard for the defined factors of impact location and temperature. Impacts had a final velocity of 4.5 m/s just prior to contact with a MEP pad. Repeated impacts occurred within

60 ± 30s seconds of each other.

Figure 4-1: Impact calibration method for the Flexiforce® sensor. A vinyl cylinder adhered to a 4.7 kg headform was impacted directly onto the Flexiforce sensor to calibrate output voltage to force.

Figure 4-2: Physical arrangement of the 25 Flexiforce sensors on a magnesium EN960 headform and their position relative to the helmet features for front, side and rear impact conditions.

Table 4-1: Physical characteristics of the five ice hockey helmet models evaluated in this study. Ten samples of each model were obtained. Manufacturer and model number data are excluded.

Helmet model 1 2 3 4 5 Primary impact Vinyl nitrile EPP EPP EPP Plastic absorption material (multi-density) cylinder array Mass 557 ± 5 g 514 ± 2 g 578 ± 3 g 322 ± 5 g 609 ± 4 g (n = 10) Size Medium Medium Medium Medium Medium

4.4.2 Data Processing & Statistics

Post-processing was done using MATLAB 2010b routines (Mathworks, Natick, MA,

USA). Raw analog signals were processed using a low-pass, 1 kHz Butterworth digital filter.

Percent error for calibration trials was calculated by computing the RMSE between the 3rd

order polynomial best fit to the 10 calibration plots of load cell output (N) vs. Flexiforce®

sensor output (Volts) and dividing by the maximum measurement range (N) of the

respective sensor. Helmet impact data were transformed by application of calibration

functions, centering raw data files about peak acceleration within the impact sample time

window of 20ms and finally by quadratic interpolation of a 3D surface to fit the 25 point

force measurements in order to create a load distribution map for each trial. A 3×3 (location

x impact) repeated measures ANOVA (General linear model) was calculated using Statistica

9.0 (Statsoft, Tulsa, OK, USA) for the dependent variables of peak acceleration and peak

focal force. Categorical variables included helmet model and temperature.

4.5 Results

4.5.1 Sensor Calibration

Average (mean ± sd) peak impact force to the Flexiforce® sensors during calibration

trials was 1630±73.7 N (n=750, 3 calibrations x 25 sensors x 10 trials). Contact time for the

calibration load profile was 12.1±0.8 ms (n=750), compared to the average contact time for all helmet impacts of 12.4±1.4 ms (n=450). Due to differences in sensitivity between the

Flexiforce® sensors average measurement range was 1180 ± 312N. Average calibration error across all sensors was 2.8±1%. Average R2 values for the correlation of Flexiforce® sensor voltage with the force plate were 0.987 ± 0.009. A plot of the load cell (N) output versus Flexiforce® output (volts) is shown below (Figure 4-3).

Figure 4-3: Sample calibration for a Flexiforce® sensor by direct impact of 4.7 kg mass. Ten unique drop impacts were recorded and a third-order polynomial was fit to the data.

4.5.2 Helmet Testing

Repeated impact results are plotted for all five helmet models, at three impact sites, and two temperatures for both peak acceleration (Figure 4-4A) and corresponding peak focal force (Figure 4-4B). Average velocity measures for all helmet impacts were 4.51 ± 0.02 m/s falling within 2% of the 4.5 m/s requirement of the CSA standard. The 25 sensors, covering

28% of the impact zone, accounted for 22 ± 3% of the global force as recorded by the force

plate during all helmet impacts. Temperature did not have a significant effect on this

measure (F1,84=2.66, p=0.1), however, location did have a significant effect on the total load

capture (F2,84=14.1, p<0.01) with rear placement of the sensor array accounting for 24% of

the global force, followed by side (21%) and front (20%) locations. This was presumably due

to differences in helmet pad/shell configuration. R2 values were 0.99±0.01 for the

correlation of net force (summation of 25 sensors) and the global force as measured by the

load cell. As expected, peak global force correlated well to peak acceleration (R2=0.99 ±

0.01). In general, peak acceleration did not correlate well to the peak focal force

(R2=0.22).The maximum average peak acceleration (219.8±9.7 G, n=5) was observed during

the first front impact (cold) for Helmet 5; however, the maximum average peak focal force

(1205±149 N, n=4) was observed for the third side impact (ambient) for Helmet 1. A 3×3 repeated-measures ANOVA revealed significant differences (p<0.05) between peak

acceleration measures for the main effects of helmet model (F4,40=40, p<0.01), temperature

(F1,40=659, p<0.01), impact (F2,80=1684, p<0.01) and location (F2,80=218, p<0.01). All

interactions were significant (p<0.05) with the exception of location*impact*temperature

(p=0.276).

Post hoc-analysis (Tukey HSD) for the main effect of helmet model, revealed that

Helmet 2 had significantly lower acceleration (142 G) than Helmet 1(152G), Helmet 3

(152G), Helmet 4 (161G) and Helmet 5 (157G). Ambient temperature had significantly

lower acceleration (140G) values than cold (166G). For the main effect of impact, first

impact was significantly lower (141G) than second (157G) and third (161G) impacts. Finally

for the main effect of location, rear impacts had significantly lower acceleration (145G) than

front (150G) and side (163G) impacts.

Figure 4-4: Peak linear acceleration (A) and peak focal force (B) results during front, rear and side impacts to a helmeted EN960 575 mm magnesium headform during linear drop impacts following the CSA z262.1-09 testing guidelines. Five models of ice hockey helmets were subjected to three repeated impacts and two temperatures (21 °C and –25 °C). A total of 50 helmets were tested (5 models × 2 temperatures × 5 samples).

Repeated-measures ANOVA results for peak focal force measures revealed

significant main effects of helmet model (F4,39=48, p<0.01), temperature (F1,39=14, p<0.01),

impact (F2,78=29, p<0.01) and location (F2,78=130, p<0.01). All interactions were significant

(p<0.05). Post hoc-analysis (Tukey HSD) for the main effect of helmet model revealed that

Helmet 2 (201N) and 4 (209N) had significantly lower focal forces than the other helmet

models, however were not significantly different from each other. Helmet 1 (351N) and

Helmet 3 (322N) were not significantly different from each other, however were significantly

greater than Helmets 2 and 4. Helmet 5 had significantly greater focal force (423N) than all

other helmet models.

For the main effect of temperature, peak focal force was significantly less at ambient

temperatures (277N) than for cold temperatures (324N). For the main effect of impact, the

first impact had significantly lower focal force (278N) than the second (305N) and third

impacts (320N). Focal forces for second and third impacts were not significantly different

from each other. Average (n=5) load distribution maps interpolated from the 25 point force

measurements are presented for ambient (Figure 4-5) and cold (Figure 4-6) conditions.

Obvious visual differences in load distribution characteristics such as contact area, and force

magnitude are apparent between helmet types, impact locations and temperature. Pre- and

post-test results of the bare headform drop were compared using the maximum net force (of

the 25 Flexiforce® sensors), global force and acceleration. There was an average change of -

6.2±2.3% in total net force measured pre- and post-test. Global force measured during the

same trials registered a 1.3±1.4% change, while global acceleration remained unchanged at

0.0±1.3%.

Figure 4-5: Average (n = 5) peak focal force distribution maps during the first 4.5 m/s impact to ambient (21 °C) conditioned ice hockey helmets. Impacts to the front, rear and side of (a) Helmet 1, (b) Helmet 2, (c) Helmet 3, (d) Helmet 4 and (e) Helmet 5 are presented.

Figure 4-6: Average (n = 5) peak focal force distribution maps during the first 4.5 m/s impact to cold (–25 °C) conditioned ice hockey helmets. Impacts to the front, rear and side of (a) Helmet 1, (b) Helmet 2, (c) Helmet 3, (d) Helmet 4 and (e) Helmet 5 are presented.

4.6 Discussion

4.6.1 Sensor Performance

In the prior foam impact study (Ouckama and Pearsall 2011) a constant load rate was used to calibrate the force sensors on a flat surface, where as in the current study direct dynamic impact calibrations were performed to represent the load rate profiles to be experienced during drop tests. This latter approach agreed well with the concurrent base load cell measures. For instance, the modeled line of best fit between the 10 repeated calibrations and base load cell (Figure 4-3) well estimates the dynamic force magnitudes

(average R2 = 0.987) and an error level of 2.8%. In terms of the ability of the sensor array to record the total force of impact, 22 ± 3% of the total force was captured by the array.

However, using interpolation to estimate forces between sensors, the captured force was improved to 87 ± 12%. Utilization of a larger sensor array would improve this estimate, as there was likely some contact force extending beyond the current array’s boundary. These findings demonstrate the sensor array’s ability to measure impact force magnitude during helmet impact.

4.6.2 Peak acceleration and focal force

There was significant increase in both peak acceleration and peak focal force by impact number. This is a common phenomenon observed in helmet testing and can be attributed to compression and/or failure of the padding materials as a short amount of time

60±30s occurred between repeated impacts. Results indicate that local contact forces were not directly proportional to peak headform acceleration. The inter-relation between these two dependent variables was modified substantially by the mixed interactions of helmet model, impact location, repeated impact, and temperature. In one instance (Helmet 5, front,

ambient) the lowest set of acceleration values were observed (Figure 4-4A), despite showing

similar focal forces (Figure 4-4B). Another instance (Helmet 5, 1st impact, cold) presented an

average peak acceleration that was 50% higher than the other models, yet, peak focal force

was nearly 224% higher on average. As a final example (Helmet 1, side, ambient)

temperature had the inverse affect on acceleration trends than for focal force. Hence, in

summary the measures of focal force provide a different characterization of impact that

cannot be predicted from global acceleration.

4.6.3 Load Patterns

Load distribution patterns at equivalent locations and temperatures were similarly shaped between repeated impacts of the same helmet model; however, both the magnitude of the focal forces and the size of the contact regions generally increased with further

impact. The force distribution patterns (Figures 4-5 and 4-6) provided unique collision

“finger prints” characteristic of each helmet and helmet impact location. As a case in point,

at the front impact site all five helmets displayed distinct regional load distributions. About

the headform-helmet interface, we can describe quite varied load transfer contact patterns

such as an off-center, inferior focus (Helmet 1 front), bilateral contact foci (Helmet 2 front,

Helmet 5 rear), large diameter focus (Helmet 2 & 3 side) and small diameter (concentrated)

focal points (Helmet 1 side, Helmet 3 rear). These diverse patterns can be attributed to the

differing construction methods, impact attenuating materials and assorted manners of

juxtaposition of materials (e.g., layering, gaps between pads). The sensor array method used

can help identify high load concentration areas, distinctive for each helmet model. While

skull fracture at these high load concentration areas is unlikely, the peak pressure calculated

based on sensor area reached upwards of 16.8 MPa. This is of particular concern, for

example, during a lateral impact to the thinner, more vulnerable temporal squamous bone

such that the adjacent neurovascular structures may in turn sustain high focal tissue distress

(skull deformation, intracranial pressure and shear deformation) (Zhang, Yang et al. 2001).

4.6.4 Implications

The prior force sensor array system used for foam testing (Ouckama and Pearsall

2011) was successfully transferred to the head model. This technological approach could be adopted to assess other products in the sports and medical industry such as shin pads, hip protectors, and other impact absorbing devices, where the contact mechanics may be difficult to measure otherwise due to complex geometry, high magnitude forces, and very short contact times. The presented force sensor array method provides a means to describe, in high temporal resolution, the contact interaction between protective devices and the body,

thereby identifying critical load concentration points on the cranium. Furthermore, these

detailed contact interactions may be used to improve input parameters to enhance

optimization of helmet performance through finite element analyses.

4.7 Funding

This study was supported by the Industrial Innovation Scholarship from the Natural

Sciences and Engineering Research Council of Canada (NSERC), Le Fonds Quebecois de la

Recherche sur la Nature et les Technologies (FQRNT) and the Bauer Hockey Corp. in St-

Jerome, Quebec, Canada.

4.8 Declaration of Conflicting Interests

The author(s) declare that there is no conflict of interest.

CHAPTER 5: PROJECTILE IMPACT TESTING OF ICE HOCKEY HELMETS:

LOAD DISTRIBUTION MEASURES

Ryan Ouckama, David J. Pearsall

In preparation for submission to the Journal of Biomechanics

5.1 Preface

This final study will revisit the issue of ice hockey helmet performance during highly localized impacts, first presented nearly 20 years ago by Bishop and Arnold (1993). The impact testing standards for ice hockey helmets do not address this impact mechanism

despite researchers repeatedly calling for the need to evaluate puck-to-helmet impacts

(Bishop and Arnold 1993; Halstead, Alexander et al. 2000). While some work has assessed consequent head linear and angular kinematics during puck and elbow impact (McIntosh and

Janda 2003; Coulson, Foreman et al. 2009) the ensuing focal contact dynamics between the helmet and head remain fundamentally unknown. Using the system developed to measure load distribution (Chapter 3) we will assess the contact parameters of the helmet-head interface during low-mass high-velocity puck impacts and compare these findings to the previous study (Chapter 4) of high-mass low-velocity impact. This manuscript is currently in

preparation for submission to the Journal of Biomechanics.

5.2 Abstract

Ice hockey pucks are solid rubber projectiles that can strike with upward of 180J of

kinetic energy. Researchers have repeatedly called for the need to evaluate hockey helmets using puck impacts, as standardized vertical drop tests (VDT) fail to simulate the highly localized impact of the low-mass high-velocity projectile. Past research using static measures of load distribution reported that despite a low risk for focal injury predicted by linear acceleration, the pressures recorded were capable of bone fracture at the temporal region.

Surprisingly, despite this finding there have been no further investigations of localized pressures during puck impact. In this study, a selection of certified ice hockey helmets were

fit to a headform and subjected to projectile puck impact at 24.2 m/s (PI24) and 33 m/s

(PI33). The corresponding linear and angular kinematics of the headform and dynamic load distribution at the contact site were measured. A Hybrid III headform was instrumented with 25 flexible force sensors and a 3-2-2-2 array of linear accelerometers. Average linear

acceleration values were greater during the VDT (122 G), followed by PI33 (98 G) and PI24

2 (53 G). However, peak contact pressure was greatest during PI33 (393 N/cm ), followed by

2 2 VDT (201 N/cm ) and PI24 (180 N/cm ) Average angular acceleration measures increased

2 2 with puck velocity (PI24=3645rad/s , PI33=8629 rad/s ) and with impact number reaching

upward of 10000 rad/s2. The quantification of this impact type provides insight into current helmet effectiveness during these impacts which place players at risk of mTBI injury or scalp lacerations.

5.3 Introduction

Mandatory helmet use and the establishment of national safety standards were introduced to the sport of ice hockey following public outcry due to the high frequency of

severe head injuries in the sport during the 1960s and 1970s. These injuries were caused primarily by blunt force trauma to the player’s cranium (contact with ice, boards, puck, other players, etc.) resulting in skull fracture and cranial hematoma. Consequently, helmets were adopted in the sport to shield the head from mechanical distress and to reduce high magnitude localized loads on the skull (Reid and Reid 1981). Correspondingly, ice hockey helmet standards were established wherein the fundamental collision tests evaluate a helmet’s capacity to limit peak linear acceleration below 275 G for a vertical drop of a helmeted headform (ISO 2003; ASTM 2007; CSA 2009). Modern helmets designed to these criteria have largely eliminated the incidence of blunt force trauma (Mueller 1998; Biasca, Wirth et al. 2002); however, the high incidence of diffuse brain injuries such as concussions (herein referring to mild traumatic brain injuries or mTBI) remains an outstanding public health concern.

The inability to obtain direct mechanical measures of cerebral tissue stress (and distress) response due to impact has been a major obstacle to understanding the etiology of mTBI. A promising alternative has been to use finite element analysis (FEA) of the brain and its tissues to estimate the stress and strain wave propagation resulting from cranial impacts that, in turn, correspond to brain injury risks. Using this approach, researchers have determined that measures of peak linear and angular acceleration alone do not correlate well to mTBI injury parameters (Forero Rueda, Cui et al. 2011). Alternatively, the shape of loading curve inputs has been found to greatly influence the magnitude and distribution of principal strain and Von Mises stress values in FEA of the brain (Post, Hoshizaki et al. 2012) and are substantially affected by the helmet’s material and construction properties. Hence, the manner of force transmission within the local dynamic boundary of helmet / cranium contact site may greatly modulate the level of brain injury risk.

The potential to map local contact dynamics was demonstrated when Bishop and

Arnold (1993) investigated various ice hockey helmets’ ability to distribute force during puck

impacts to the temple region of a Hodgsen-WSU headform. This was accomplished by the

use of pressure sensitive contact films placed between the headform and helmet at the site of

puck impact. Their results showed that of the helmet models tested, none were capable of

managing the focal forces transmitted to the temporal region. Though headform global

accelerations were below 275g, substantial pressure magnitudes (>5MPa) were achieved.

Since the pressure films provided only a summative picture of the impact event, the

temporal history was lost. The authors noted that further exploration of load measurement

techniques and appropriate thresholds were needed. Current developments in flexible force

sensor arrays make possible accurate spatial and temporal mapping of foam impact events

(Ouckama and Pearsall 2011). Furthermore, these sensor arrays have been shown to function well in standardized linear helmet drop tests (Ouckama and Pearsall 2012). These sensor arrays provided gross estimates of headform acceleration comparable to that of the accelerometer, yet also were able to discriminate spatial contact differences between helmet models. This latter observation is most relevant, as it offers the potential to quantify the input characteristics needed for realistic finite element modeling of impact behavior. Further study using this testing technique is thus warranted.

In this vein, this study extends the use of force sensor arrays to examine puck projectile impact events to ice hockey helmets. The fundamental difference between the two test methods is in the relative mass and velocities of the impacting object. In the drop test, the headform falls at a low speed onto a static anvil (i.e. high-mass, low-velocity impactor) whereas in the projectile test, the puck moves at high speed on the static headform (i.e. low mass, high-velocity impactor). These impact types are typical of collision events in the game

of ice hockey (Halstead, Alexander et al. 2000), yet, few studies have examined projectile impacts to ice hockey helmets (McIntosh and Janda 2003; Coulson, Foreman et al. 2009).

The kinetic energy of a projectile puck impact can easily exceed the energy levels established in the helmet standards. For example, an official 160g hockey puck must travel at only 24.2 m/s to impact a helmet with 47J of kinetic energy, equal to the impact energy during drop testing of medium sized helmets in the CSA ice hockey helmet standard (CSA z262.1-09).

Professional hockey players can achieve puck velocities of up to 48 m/s during a slap shot

(NHL 2012), reaching kinetic energy (184J) levels nearly 4 times that of the linear drop test.

Considering the potential of puck-to-helmet impact for higher energy, shorter contact duration and a smaller contact area, there is potential for significantly greater levels of helmet material deformation and stress during this impact modality relative to traditional helmet testing methods. Given the lack of time series measures of load distribution during these highly focal impacts, the authors propose to quantify spatial load distribution across several commercial ice-hockey helmet models, representing various material types and geometries to assess their performance during puck impact.

5.4 Methods

A 50th percentile Hybrid III dummy headform (Model 78051-61X-1846, Humanetics

Innovative Solutions, Inc., Plymouth, MI) was fixed to a flexible neck (Model 78051-90).

Linear and angular acceleration variable were collected using a 3-2-2-2 orthogonal array

(Padgaonkar, Krieger et al. 1975; DiMasi 1995) of 9 linear accelerometers (Model 7264C-

500, Meggitt’s Endevco, San Juan Capistrano, CA). Load distribution was measured by instrumentation of the dummy forehead with 25 discrete flexible force sensors (Flexiforce® model A201, Tekscan, USA). A 5×5 grid, corresponding to wireframe intersections of the

Hybrid III finite element (FE) model (LS PrePost 3.1, LSTC, Livermore, CA) was transferred to the physical headform (Figure 5-1). The grid was symmetrical about the median plane with spacing of 2-3 wireframe intersections to provide room for the

Flexiforce® sensors on the physical headform. Two reference points on the FE model were established (tip of nose and 6.4 cm superior of nose along the median plane). Linear measures between the reference points and each grid location were recorded using LS-

PrePost software. A drafting compass was then used to transfer the grid coordinates to the physical Hybrid III head by intersecting arcs of fixed distances from the reference points.

This method of sensor placement was developed to allow future comparisons between the empirically measured forces and the contact stresses predicted by corresponding FE analysis.

The force sensors were adhered directly to the headform using double sided tape. Care was taken to minimize overlap of the sensors and cabling. A nylon stocking was placed over the instrumented headform to minimize shear forces, and to mimic the movement of human skin beneath the helmet (Pellman, Viano et al. 2006). Two 32-channel analog acquisition modules (NI-9205, National Instruments, Austin, TX) were installed in a CompactDAQ chassis (cDAQ-9174). One module, operating at 20 KHz, was dedicated to capture of the 9

head accelerometers, while the second, operating at 10 KHz captured the 25 analog force channels. Synchronization between the 34 channels was maintained by a common timing engine within the CompactDAQ chassis. Accelerometer channels were conditioned by three,

3-channel amplifiers (Model 136, Meggitt’s Endevco, San Juan, CA) each employing a

CFC1000 hardware anti-alias filter (SAE 1995). The Flexiforce® sensors were conditioned using custom hardware based on the recommended MCP-6000 series op-amp (Tekscan

2009). Accelerometer and force data were post-processed in MATLAB® (r2012 ,The

MathWorks, Natick, MA) using a digital Butterworth filter meeting CFC1000 specifications.

Force measures at each sensor were converted to pressure based on the active sensor area

(0.71 cm2) and then using the physical sensor coordinates, interpolated to a matrix of 50×50 pressure sensels using a MATLAB script.

Figure 5-1: Force sensor locations (25) were selected based on wireframe intersections of the Hybrid III finite element model (left). The virtual coordinates were transferred to the physical headform (right). The point-to-point distance from reference point A (tip of nose) to each sensor row along the median plane is displayed. A red laser on the physical headform indicates the alignment of the projectile canon with the central sensor.

The head/neck assembly was mounted in a pneumatic ice hockey puck cannon

fixture. The cannon fired a regulation ice hockey puck (mass=160g) at ambient room

temperature at a precise point on the headform. Ten samples of five different models of

hockey helmets, representing various material types and geometries were obtained for the

study (Table 5-1). For each model, 5 helmet samples were subject to two 24.2 m/s puck

impacts (PI24). This velocity was selected to provide approximately the same energy (47J) as

the linear impact test for medium size hockey helmets (CSA 2009). An additional 5 samples

for each model were then subject to two puck impacts at 33m/s (PI33) providing 87J of

kinetic energy. This velocity was selected as it is common to test the toughness of full-face

protectors (visors/cages) in various hockey helmet standards (ISO 2003; CSA 2009). Time

between impacts was between 60-90 seconds to allow for retrieval of the puck and

pressurizing of the air canon. Puck velocity was measured by two laser light traps at the

distal end of the cannon barrel. The puck was aimed such that the center of mass of the

puck was aimed at the central force sensor, 11.0 cm vertical from the tip of nose along the

median plane (Figure 5-1 above). The impact vector was in median plane and parallel to the x-axis (anterior-posterior) of the headform.

Table 5-1: Physical characteristics of the five ice hockey helmet models evaluated in this study. Ten samples of each model were obtained. Manufacturer and model number data are excluded.

Helmet 1 Helmet 2 Helmet 3 Helmet 4 Helmet 5 Primary Impact Vinyl Nitrile Plastic cylinder EPP EPP EPP Absorption Material (multi-density) array Two-piece Two-piece Two-piece One-piece One-piece Shell Type polyethylene polyethylene polyethylene polycarbonate polyethylene Mass (n=10) 557±5 g 514±1 g 578±4 g 322±5 g 609±4 g Size Medium Medium Medium Medium Medium

5.4.1 Statistics:

A repeated measures ANOVA (α=0.05) was calculated (Statistica v8.0, Statsoft, Inc,

Tulsa, OK) for each puck velocity with main effects of helmet model (5) and repeated measure of impact number (2). Dependent measures of peak linear acceleration, peak angular acceleration, peak pressure and average pressure were analyzed. Post–hoc analyses of significant factors were calculated using Bonferroni corrections. Data exclusions occurred if puck velocity was greater than ±5% from the goal velocity, or a visible material failure resulted in abnormal acceleration curves.

5.5 Results:

The total number of trials for each helmet model is presented in Table 5-2. Of the 50 samples tested, a total of 34 helmets with both successful 1st and 2nd impacts, were included in statistical analysis.

Table 5-2: Total number of impact trials per helmet model.

Puck Velocity 24 m/s 33 m/s Total Helmet 1 3 3 6 Helmet 2 3 5 8 Helmet 3 3 3 6 Helmet 4 3 3 6 Helmet 5 3 5 8 Total 15 19 34

5.5.1 Event timing

Average puck velocities were 24.2 ± 0.2 and 32.9 ± 0.6 m/s. Average time for the

headform to reach peak resultant acceleration (tacc) across all helmet models was 1.28 ± 0.28

ms for 24 m/s puck impacts (PI24) and 1.07 ± 0.10 ms at 33 m/s puck impacts (PI33). The average time to peak net force (tfnet) was 1.36 ±0.23 ms at PI24 and 1.07 ± 0.09 ms at PI33.

After initial compression of the helmet materials, the puck typically rebounded at an elevated

trajectory followed by rotation of the head by an average of 13.8 ± 1.1 degrees during PI24

and 18.9 ± 2.2 degrees during PI33 (Figure 5-2). Rotation angles were calculated by a double

integration of the rotational acceleration about the medio-lateral y-axis of the head. Average

time to maximal neck extension was 68.0 ± 2.4 ms for PI24 and 73.2 ± 2.6 ms for PI33.

Figure 5-2: Event sequence for the impact of a 160g hockey into a helmeted hybrid III headform. The period over which the material can dissipate force is very brief (<2ms). The rotational displacement of the headform peaked around 70ms.

5.5.2 Headform Measures

The dependent measures of linear acceleration, angular acceleration, average

pressure, peak pressure and contact area are presented for impacts to the five differing

helmet models by an ice hockey puck at 24 m/s (Table 5-3) and 33 m/s (Table 5-4). Bar

graphs of the same data are presented for visual comparison of linear acceleration (Figure 5-

3), angular acceleration (Figure 5-4), pressure (Figure 5-5) and contact area (Figure 5-6).

Table 5-3: Average (Least-square mean ± standard error) resultant linear acceleration, resultant angular acceleration, average pressure, maximal pressure and contact area during impact to various models of ice hockey helmets by an official-sized puck travelling at 24.2 m/s. Maximal values are indicated by bold type.

Impact Velocity = 24.2 m/s Helmet Impact Model # Linear (g) Angular Acc Angular Vel. Pavg Pmax Contact area (rad/s2) (rad/sec) (N/cm2) (N/cm2) (cm2) I 62.4 ±2.2 5423 ±154 6.4 ±0.2 76.7 ±4.6 210.8 ±16.8 47.7 ±1.5 1 II 67.9 ±2.6 5482 ±223 6.3 ±0.2 81.3 ±4.9 225.6 ±23.1 49.2 ±1.9 I 57.0 ±2.2 5043 ±154 6.7 ±0.2 66.3 ±4.6 186.8 ±16.8 63.0 ±1.5 2 II 65.4 ±2.6 5322 ±223 6.8 ±0.2 74.0 ±4.9 217.2 ±23.1 62.1 ±1.9 I 60.2 ±2.2 5185 ±154 7.2 ±0.2 70.8 ±4.6 186.6 ±16.8 48.9 ±1.5 3 II 77.2 ±2.6 5698 ±223 7.1 ±0.2 84.0 ±4.9 259.9 ±23.1 52.3 ±1.9 I 44.6 ±2.2 5434 ±154 7.7 ±0.2 59.8 ±4.6 132.7 ±16.8 44.5 ±1.5 4 II 67.5 ±2.6 5998 ±223 8.0 ±0.2 104.4 ±4.9 326.2 ±23.1 42.0 ±1.9 I 40.6 ±2.2 3645 ±154 6.8 ±0.2 64.7 ±4.6 181.1 ±16.8 48.9 ±1.5 5 II 38.4 ±2.6 3815 ±223 7.0 ±0.2 57.4 ±4.9 142.3 ±23.1 51.2 ±1.9 38.4 (H5) 3645 (H5) 6.3 (H1) 57.4 (H5) 132.7 (H4) 42.0 (H4) Range I & II 77.2 (H3) 5998 (H4) 8.0 (H4) 104.4 (H4) 326.2 (H4) 63.0 (H2) Average I 53.0 4946 7.0 67.7 179.6 50.6 Average II 63.3 5263 7.0 80.2 234.2 51.4

Table 5-4: Average (Least-square mean ± standard error) resultant linear acceleration, resultant angular acceleration, average pressure, maximal pressure and contact area during impact to various models of ice hockey helmets by an official-sized puck travelling at 33 m/s. Maximal values are indicated by bold type.

Impact Velocity = 33 m/s Helmet Puck Angular Acc Angular Vel. Pavg Pmax Contact area Model Impact Linear (g) (rad/s2) (rad/sec) (N/cm2) (N/cm2) (cm2) I 110.1 ±8.9 8538 ±835 7.6 ±0.2 120.4 ±8.8 374.1 ±46.6 50.7 ±1.4 1 II 123.1 ±8.5 8436 ±912 7.5 ±0.2 127.3 ±7.8 410.9 ±44.7 52.7 ±1.7 I 97.1 ±6.9 8742 ±647 8.8 ±0.2 100.9 ±6.8 353.0 ±36.1 67.3 ±1.1 2 II 121.9 ±6.6 9359 ±706 8.5 ±0.2 119.9 ±6.0 453.4 ±34.6 64.0 ±1.3 I 96.7 ±8.9 7981 ±835 8.1 ±0.2 97.9 ±8.8 344.6 ±46.6 56.8 ±1.4 3 II 154.5 ±8.5 11330 ±912 8.6 ±0.2 143.5 ±7.8 544.3 ±44.7 53.8 ±1.7 I 110.2 ±8.9 10281 ±835 10.7 ±0.2 127.5 ±8.8 521.4 ±46.6 55.3 ±1.4 4 II 115.2 ±8.5 11533 ±912 10.0 ±0.2 86.1 ±7.8 285.1 ±44.7 60.5 ±1.7 I 77.9 ±6.9 7604 ±647 8.6 ±0.2 96.5 ±6.8 373.0 ±36.1 55.6 ±1.1 5 II 96.4 ±6.6 9616 ±706 9.1 ±0.2 106.4 ±6.0 376.0 ±34.6 56.7 ±1.3 77.9 (H5) 7604 (H5) 7.5 (H2) 86.1 (H4) 285.1 (H4) 50.7 (H1) Range I & II 154.5 (H3) 11533 (H4) 10.7 (H4) 143.5 (H3) 544.3 (H3) 67.3 (H2) Average I 98.4 8629 8.8 108.6 393.2 57.1 Average II 122.2 10055 8.7 116.6 413.9 57.5

5.5.3 Linear Acceleration

All helmet models tended to have greater linear acceleration upon 2nd impact with the exception of model 5 during low speed puck impacts (Figure 5-3). There was a

significant interaction of impact*model for both the PI24 condition (F(4,10) = 7.0, p<0.01) and the PI33 condition (F(4,14) = 6.0, p<0.01). Helmet model 3 had a significant increase in linear acceleration between impacts at both puck velocities, whereas models 2 and 4 had significant differences at only one of the test velocities. Due to the short contact time (<2 ms) and the fact that both the head injury criterion (HIC) and Gadd Severity Index (SI) are related to duration of contact, the resulting criterion values were well below typical injury thresholds.

Average HIC and SI values for each helmet are presented for each helmet model.

Figure 5-3: Peak resultant linear acceleration by helmet model for the repeated impact of an ice hockey puck at 24 and 33 m/s to the forehead of a hybrid III dummy headform. Asterisks within the bar indicate a significant difference between 1st and 2nd impacts within helmet models. Average head impact criterion values (HIC) and Severity Index (SI) are indicated for each helmet model. Error bars indicate ± 1 SE.

5.5.4 Angular Acceleration

Angular acceleration results were similar to linear acceleration with some exceptions.

Though Helmet 3 had the highest linear acceleration at both puck velocities, it was Helmet 4 that had the greatest magnitudes of angular acceleration (Figure 5-4). There was a significant

main effect of both model (F(4,10)= 27.2, p=0.01) and impact (F(1,9)=9.1, p=0.01) for the PI24 conditions. Helmet 5 had significantly lower angular acceleration than all other models.

Angular acceleration increased with impact number. There was approximately a 6% average

increase in angular acceleration upon second impact at P24 compared to a 17% increase at

P33. There was a significant interaction between helmet model and impact number at PI33

(F(4,14)= 29.2, p=0.01). Helmet model 3 had comparably, a much larger increase in angular acceleration upon second impact, whereas all other models had no significant difference, resulting in the interaction. There was a significant main effect of helmet model for the

measure of angular velocity at PI24 (F(4,10)=12.8, p<0.01) and PI33 (F(4,14)=40.8, p<0.01).

Unlike angular acceleration, where helmet model 5 produced the lowest values, angular velocity was lowest for helmet model 1 for both impact conditions. Angular velocity was greatest for helmet model 4 just as it was for the angular acceleration values.

Figure 5-4: Peak resultant angular acceleration across helmet model for two repeated impacts to the forehead of a hybrid III dummy headform by an ice hockey puck travelling at

24 m/s (PI24) and 33 m/s (PI33). Average peak angular velocity (rad/s) about the y-axis is presented below each helmet model. Asterisks above the bars represent significant difference of the indicated model from all other helmet models. Asterisks within the bar represent significant differences between impacts. Error bars indicate ± 1 SE.

5.5.5 Contact Pressure

There was a significant interaction effect (impact*model) for the measure peak

contact pressure during both the PI24 (F(4,10)=9.5, p<0.01) and PI33 (F(4,14)=8.9, p<0.01)

conditions. Post-hoc analysis revealed significant differences between impacts for helmet

model 4 at both puck velocities. Helmet 4 was unique in that it resulted in both the lowest

and highest values for maximal contact pressure during the PI24 condition. There was more

than a 2-fold increase in peak contact pressure between first and second impact. A

comparable increase in pressure between impacts was not realized during the PI33 condition

for model 4. In fact, peak pressure values were significantly lower upon second impact. Both linear and angular acceleration increased upon second impact of this model suggesting that greater magnitudes of force were transferred to the headform, that were not accounted for at

the instrumented impact site. While there was no extreme visible damage to this helmet

model, there was a palpable change to the padding at the impact site which was perhaps

related to this decreased force measure. Contact area, estimated by active area of the

interpolated pressure map, was significantly greater for helmet model 2 than all other models

at both impact velocities.

Figure 5-5: Peak contact pressure (N/cm2) between the five different helmet models and a hybrid III dummy during repeated impacts of an ice hockey puck at 24 m/s and 33 m/s. Average pressure (indicated by shaded band of each bar) was calculated for the active contact area (indicated under each helmet model). Asterisks indicate a significant difference between impacts within helmet models. Error bars and shaded regions indicate ± 1 SE.

5.5.6 Pressure Distribution

Visual plots of pressure distribution at peak-load and time-series data for each sensor

are presented for PI24 (Figure 5-6) and PI33 (Figure 5-7) conditions. The pressure

concentration was inferior to the central sensor, likely as a result of the puck trajectory not

being normal to the head’s curved surface. The natural curvature of the forehead caused the

lower portion of the puck to make first contact (see Figure 5-2 above) likely resulting in the

non-centric pressure concentration. Time-series plots of the individual load sensors averaged

across helmet replicates, is presented below the corresponding pressure distribution plots.

Figure 5-6: Low-speed sample trials showing spatial loading profiles between 5 differing helmet models during impact by a 260g hockey puck at 24 m/s. Twenty-five discrete force signals (lower panel) were recorded at the helmet-head interface of a hybrid III headform. The dashed vertical line represents the instant of maximal net force. Intensity maps (upper panel) represent the spatial pressure distribution at this same instant. In some cases the greatest the point of peak net force did not correspond to the time of individual sensor maximums (eg. Helmet 5).

Figure 5-7: High-speed sample trials showing spatial loading profile between 5 differing helmet models during impact by a 260g hockey puck at 33 m/s. Twenty-five discrete force signals (lower panel) were recorded at the helmet-head interface of a hybrid III headform. Vertical dashed lines represent the instant of maximal net force. Colour maps (upper panel) represent spatial pressure distribution at this same instant.

The foam-based helmets (Helmets 1-4) had similar loading profiles with a single rise to a

central maximum. This was not true in Helmet 5, which does not utilize traditional padding

materials. The resulting loading patterns were notably different from the other helmet types

and displayed a rapid rise of the central sensor (red-dashed) followed by delayed onset of the

remaining sensors which peaked at various different timings. Some load plots show small forces prior to the impact. These pre-load forces were due to the helmet fit resulting in

contact with the load sensors.

5.6 Discussion

Projectile puck impacts to a player’s head are a probable event during the game of ice

hockey and can result in injuries including laceration, concussion and brain contusion

(Biasca, Wirth et al. 2002). In the absence of a protective helmet, a puck impact to the head

may prove fatal (Winslow and Goldstein 2007). Thus, it is important that athletes

participating in ice hockey have adequate head protection from these highly focal impacts.

Yet, despite this injury risk, remarkably little research has focused on the measurement of

puck impacts to the helmeted head, particularly with regard to the forces generated; hence,

the purpose of this paper was to investigate the surface loading characteristics at the contact

interface of the helmet and head during low-mass, high-velocity puck impacts. In the

following text, the resulting measures of load distribution, and headform kinematics (linear

and angular acceleration) will be discussed and compared to an earlier study assessing load

distribution during a high-mass low-velocity impact protocol (Table 5-5).

Table 5-5: Impact test results (linear acceleration, average and peak contact pressure) from a high-mass low-velocity vertical drop test (CSA z262.1-09) protocol for front impacts to ambient temperature ice hockey helmets. Helmet models matched those used in the current study. Data adapted from Ouckama & Pearsall (2012).

Impact Velocity = 4.5 m/s Helmet Drop Model Impact Linear Accel. Pavg Pmax (g) (N/cm2) (N/cm2) I 134.2 99.8 220.9 1 II 150.6 106.0 256.8 I 121.8 80.0 138.2 2 II 142.3 94.6 202.3 I 114.9 77.4 164.0 3 II 146.0 103.0 253.8 I 130.8 84.6 199.9 4 II 147.2 94.7 209.1 I 107.4 120.8 284.0 5 II 110.2 104.0 247.2 107.4(H5) 77.4 (H3) 138.2 (H2) Range I & II 150.6 (H1) 120.8 (H5) 284.0 (H5) Average I 121.8 92.5 201.4 Average II 139.3 100.4 233.8

5.6.1 Linear Accelerations

The puck impacts at 24.2 m/s (PI24) resulted in headform peak accelerations ranging

from 38-77 G. This was substantially lower in comparison to the range measured in our

previous study (107-150 G) assessing the same helmet models according to the CSA Z262.1-

09 protocol (Ouckama and Pearsall 2012) (Table 5-5). The discrepancy in these acceleration

values were in part attributed to differing headform compliance (EN960 headform versus

Hybrid III dummy head and neck) and impact rebound conditions, affecting the total energy

transfer. At greater puck velocity (PI33), the resulting peak linear acceleration (78-155 G)

achieved a range closer to the previous study. In general, performance rankings based on

peak linear acceleration (1st impact) were consistent between blunt (drop test) and highly

localized (puck) modalities. Helmet model 1 (VN foam liner) produced the greatest linear

acceleration values in both cases, while helmet model 5 (plastic cylinder array) produced the

lowest. The remaining helmet models all utilized EPP foam liners. Helmet model 2 was

consistently middle ranked during both impact types; however models 3 and 4 produced

mixed rankings. Helmet model 4 differs from the other EPP based helmets in that it utilizes

a lightweight, polyethylene shell bonded directly to a thin foam layer. This helmet performed

particularly well during the first PI24 test, with an average linear acceleration (44.6 G) lower than all other EPP foam helmets; however, at higher puck velocity (PI33) this helmet model

was the worst performer (110.2 G) among the EPP based liners. This result may be

attributed to the softer liner utilized in this helmet. Using simulations, softer liners were

calculated to perform optimally during low speed impact, however, during higher speed

impact, a stiffer liner was optimal for reduction of linear acceleration (Forero Rueda and

Gilchrist 2012). Given the considerable variance in liner stiffness between helmet models,

material selection should be balanced for performance during both high mass low-velocity

and low-mass high-velocity type impacts (i.e. head drop and puck projectile tests,

respectively) to protect the athlete from these foreseeable risks of the game. Linear

acceleration during second impact resulted in similar ranking for the vertical drop test,

however, during puck impacts helmet 3 (EPP) produced the greatest acceleration values.

Two of these samples were excluded due to shell damage suggesting that material failures

during second impact were likely the cause of the increased linear acceleration.

5.6.2 Angular Accelerations

2 The peak rotational acceleration values (3.6-6.0 krad/s ) obtained during PI24 were

close in magnitude to a similar helmet population tested using a 7.5 m/s pneumatic linear

impactor (4.9-5.9krad/s2) (Walsh, Post et al. 2012). The average rotational acceleration

2 during PI33 (8629-10055 rad/s ) was in the same range of a punch (hook) from an Olympic

boxer (9300 ±4485 rad/s2) (Viano, Casson et al. 2005). The lowest angular acceleration

values were obtained for Helmet 5 constructed using a thermoplastic impact absorbing

material. Linear and angular acceleration results from puck impact were strongly correlated

(r2=0.85), however, there were some noteworthy findings. For example, helmet 4, resulted in

the greatest angular acceleration measures for both puck impact speeds despite mid-range

linear acceleration results. There were no significant differences in angular acceleration

between helmet models using VN (Helmet 1) and EPP (Helmets 2-4) materials. This result is

contrary to what was found when testing hockey helmets with a NOCSAE style linear impactor (Post, Oeur et al. 2011). The vinyl nitrile helmet used in this study incorporated multiple-density foams layers which may help explain the differing results.

5.6.3 Load distributions

The helmet models used in this study were evaluated for load distribution characteristics during standardized vertical drop tests in a prior study (Table 5-5). Under this

evaluation method, the peak localized pressure, averaged across all five helmet models, was

201 N/cm2 during first and 233 N/cm2 during second impact to the front of the helmet.

During projectile impact (PI24), the average peak pressure was similar to the vertical drop

test results for first (180 N/cm2) and second (234 N/cm2) puck impacts. However, at higher

velocity (PI33) nearly a 2-fold increase in pressure relative to the drop test protocol was

observed during first (393 N/cm2) and second (414 N/cm2) puck impacts. This difference in

pressure between impact types (drop and puck impact) is likely underestimated as the hard

magnesium headform used for drop testing is much less compliant that the surface of the

Hybrid III headform. Pressures ranging from 285-544 N/cm2 (2.85-5.44 MPa) were

recorded during PI33. These pressures were lower than Bishop’s (1993) side impact results (8-

20MPa); however, this could be expected due to the thicker padding and uniform shell

coverage at the front of the helmet in comparison to the temporal region. We assessed front

impacts due to alignment restrictions of the equipment in addition to the fact that the

Hybrid III neck is designed and validated specifically for use during frontal impacts

(Yoganandan, Pintar et al. 2011) and would thus provide the most accurate response.

Relating pressure to skull fracture tolerance is challenging, as literature typically report

fracture limits in net force despite the fact they often use different sizes and shapes of

impactor. Part of the challenge in reporting tolerance limits in pressure is inability to know the instantaneous contact region of the impactor. Frontal bone fractures were as low as

2670N using a 6.45cm2 cylindrical impactor (Nahum, Gatts et al. 1968), which, if fully

engaged with the bone surface would be 4.1 MPa. In a more recent study, a projectile impact

to the tempo-parietal bone was evaluated with an impactor of similar mass (103g) and speed

(33 m/s) to the ice hockey puck. A 50% risk of fracture was calculated at 4572N over an

area of 11.4 cm2 (4.0 MPa) (Raymond, Van Ee et al. 2009). Changes in contact area and

impact velocity create fundamentally different responses of the skull, with larger surface

areas corresponding to larger fracture forces (Allsop and Kennett 2002). The relation

between change in contact area and impact tolerance is unclear.

The average net force, calculated by the average pressure multiplied by total contact area,

was 3425N during PI24 and 6175N for PI33. While forces achieved ranges that caused

fracture in previous experiments, it is unlikely that a fracture would occur as the total contact

region was 51-61cm2 during helmeted impact. However, laceration can be caused over this

contact area with forces greater than 4000N (Sharkey, Cassidy et al. 2012).

100

The current study is unique from earlier work by the application of dynamic

measures of load distribution and inclusion of rotational kinematics. The synchronized

measures permitted visualization of the changes in load distribution across time. Functional

differences between helmets and impact tests could be identified. For example, the plots of

first impact at PI24 (Figure 5-6) and PI33 (Figure 5-7) speeds demonstrated the differences in

material responses undetected by global force or acceleration measures. Helmets with the

highest peak linear acceleration did not correspond to those the highest peak pressures, thus

further site by site analysis of puck impact induced pressure transmission is warranted,

particularly at more vulnerable locations of the head. The assumption that shell/liners will

uniformly transmit contact pressures was found to be untrue; for instance, during PI24 to helmet model 5, load concentration was initially focused on a single central sensor then yielded to multiple points away from this central sensor, perhaps as a function of the plastic engineered liner. Furthermore, this dynamic pressure mapping method identified unexpected

differences in loci of point of contact to headform. For example, the incident location of the

puck to the helmet’s outer shell did not correspond to the liner/headform surface maximum

location (Figures 5-6 and 5-7). Due to the curvature of the head, the load concentration

occurred at the inferior leading edge of the puck (Figure 5-2). Without this mapping

evidence, the difference between presumed and actual head impact location would not have been immediately apparent.

5.7 Summary and Conclusions

It is important that ice hockey helmets can function adequately across the spectrum

of impact types experienced during play. Our results show that despite the use of modern

helmet designs, 33 m/s puck impacts can generate substantial linear and angular

101

accelerations as well as localized contact forces sufficient for scalp injury. The majority of

PI24 impacts resulted in linear and angular acceleration values below the proposed 50% injury thresholds of 82 G and 5900 rad/s (Zhang, Yang et al. 2004) suggesting that they

could likely manage the risk of mTBI injury. However, at higher puck velocities (PI33), easily achieved by skilled players, linear and angular accelerations exceeded 106 G and 7900 rad/s2, representing an 80% risk of mTBI injury. These risk values presented should be interpreted cautiously as they originate from longer duration football head-to-head impacts. Further research is warranted to assess the risks posed by short duration projectile impacts experienced in sports such as ice hockey, baseball, cricket and lacrosse. The load distribution measurement system identified nearly a 2-fold difference in peak pressure between the drop test and puck impact methods that was not realized by global linear acceleration. The measurement of load distribution during projectile impact provided useful data for identification of focal injury risks (e.g., laceration), and understanding of material behaviours and load concentrations caused during various impact modalities. Additionally, with the increasing application of finite element modelling to helmet optimization and head injury prediction, these data may serve to validate the corresponding contact stresses estimated by these models at the helmet-head interface.

5.8 Limitations

The primary limitation of this study exists in the comparison of helmet impact methods using different headforms. The EN960 magnesium headform was necessary to meet the requirements of the CSA helmet standard used to certify all hockey helmets for sale and use in Canada. However, given the absence of a standard for puck impact we felt it necessary to provide an accurate representation of the event, which was not possible using

102

the EN960 headform. The Hybrid III head and neck provided rotational kinematics,

necessary for assessment of head injury risk, in addition to a biofidelic skin which we

assumed would create more realistic pressure measures than the magnesium headforms.

Although these data do not truly represent the response of a human head, they are useful to compare with the growing literature base using these headforms for sports-related injury reconstruction.

5.9 Funding

This research was supported by the Industrial Innovation Scholarship (IIS) in collaboration with the Fonds de Recherche du Québec – Nature et technologies (FQRNT),

the Natural Science and Engineering Research Council of Canada (NSERC) and the Bauer

Hockey Corp.

103

CHAPTER 6: SUMMARY AND CONCLUSIONS

104

105

6.1 Summary

The manuscripts presented in Chapters 3-5 of this thesis have presented a novel methodology to evaluate helmet performance based on parameters of localized contact forces. Traditionally, helmet performance has been assessed by a measurement of the corresponding headform motion (linear acceleration) during an impact. This global measure of acceleration represents the average kinematic response of a rigid body (the headform) to an applied net force vector. When mass is constant, as it is during helmet testing, net force and linear acceleration are highly correlated, thus it is redundant to collect both measures.

However, if we break down the net force vector into its smaller component forces, it is apparent that the same net force vector could be achieved by many combinations of individual forces. The same net force can be achieved from distributed smaller magnitude forces, or from concentrated high forces. Based on this notion, linear acceleration measures would not be entirely sensitive to highly localized pressures. This acceleration/ localized pressure discrepancy was initially demonstrated using static measures in ice hockey helmets

(Bishop and Arnold 1993), and was explored further through dynamic measurement methods presented in this thesis. The development of a measurement system for the assessment of dynamic helmet-head contact forces has been presented. Additionally, applications of this system were presented in the context of two differing impact types (blunt and highly localized) to ice hockey helmets. A brief summary of each chapter will follow.

In Chapter 3, an evaluation of a flexible force sensor array for testing of helmet foam impacts was presented. Typical flexible force sensing arrays use multiplexed signals to scan dense force sensor grids at the cost of sampling rate. The system presented used independently sampled channels instead in order to achieve the necessary high collection rates (10+ kHz) at the cost of reduced spatial resolution. Due to the curvature of the head, it

106

is difficult to place embedded matrix type sensors without crimping and damaging the

sensor. Loose hanging, independent sensors allowed for precise placement on the curvatures

of the headform. A custom amplifier was created and a data acquisition system was

incorporated in order to collect the resulting data. The use of a dynamic calibration method

instead of a static calibration of varying constant loads allowed for more accurate

measurement of forces and pressures. This system, based on flexible force sensors, was

successfully implemented into a foam measurement protocol in which, by use of load

interpolation, global forces were accounted for within 5% error with r-square values of 0.97.

The system was able to differentiate foam materials based on temperature and density

parameters and allowed for a visual representation of the loading profiles for each of these

foams. There were notable differences in localized maximal forces between differing foams

despite similar acceleration curves. This provides proof of concept that acceleration, being a global measure, is influenced by a summative total load on the object, but the localized

magnitudes and contact regions of that global load can vary greatly. It was apparent that the

contact area expanded beyond the sensing area in some cases, which warranted both expansion of the total number of sensors as well as increasing the sensor density to increase spatial resolution. Following evaluation of the system capabilities during a controlled impact onto a flat surface, it was necessary to then expand this to a curved headform and introduce commercial helmet samples.

In Chapter 4, the force array system was expanded to record 25 channels in a 5x5 array and was used in conjunction with a certification standard for ice hockey helmets (CSA

z262.1-09). The purpose of this study was to both evaluate the capabilities of the force

measurement system when used on a curved surface of a headform and to quantify the

contact forces generated in relation to linear acceleration values between several commercial

107

hockey helmets of varying material types. The CSA standard relies on a single measure of

vertical acceleration while providing a guided linear impact of a magnesium headform

wearing the helmet. Drop impact tests are well known industrial performance tests for

helmets, and were selected as a logical controlled environment to apply the sensor system as it minimized variables associated with free-motion impact and allowed for global force to be quantified through the use of a load cell. This experiment revealed differences in both force concentrations and magnitudes based on differing material types, and various helmet features (external geometry, adjustment hardware). This study provided unique data of considerable interest as global measures of load and acceleration did not predict local contact forces, but only the net effect of the total load applied. For example, high contact pressure

(>16 MPa), nearly four times those found in other models, were identified in one helmet model by force mapping despite no comparable differences in between models based on linear acceleration. The highly localized forces, possibly caused by a protrusion of a plastic part through the foam material, would go unnoticed in evaluation methods based only on linear acceleration, and thus further supports the use of force mapping the head surface. The vertical drop test employed in this study represented a relatively high-mass low velocity

(blunt) impact to the helmet. Ice hockey is a sport of multiple impact risks, and it has been expressed multiple times that the drop test does not represent the mechanics of a low-mass high-velocity projectile impact (i.e. puck) (Bishop and Arnold 1993; Halstead, Alexander et al. 2000). The final step was to evaluate a projectile impact methodology using matching helmet population in order to provide a comparison to the results obtained in this chapter.

In Chapter 5, helmet models matching those used during drop testing (Chapter 4), were subject to projectile impact of low-mass, high-velocity ice hockey pucks. The ability to measure both linear and rotational accelerations in combination with a biofidelic skin was

108

deemed necessary to provide the most representative and useful dataset to the research

community. This study provides for the first time an empirical time-series measurement of

load distribution during puck-to-helmet impacts. Our results demonstrated that during 33

m/s puck impacts, localized pressures reached 5.4 MPa, more than double the pressure

measured (2.3 MPa) during drop impact testing to the same site. Additionally, measures of

linear and angular acceleration of the Hybrid III headform reached the threshold for high

(80%) likelihood of concussion injury. The results of this final study provided valuable

information about the contact forces, and headform kinematics associated with puck impact

to the helmet and how these impacts differ fundamentally from typical drop testing

methods. Future developments in stretchable electronic circuitry technologies (e.g. organic

electronics/nanowire transistors) will likely aid in improvement of the quantification of

surface contact forces through better conformation to the surface, increased spatial

resolution and possible integration into headform skin. Additional studies are necessary to

further quantify injury rates and injury mechanisms, however, the data provided here will aid in reconstruction and simulation of these short duration impacts.

6.2 Conclusion

Based on the works presented in this thesis we can conclude that during helmet impact, global forces and global acceleration correlate well, however, localized load distribution measures are not accounted for by either linear or angular acceleration measures.

In instances where localized pressures reach extremes, laceration, contusion or localized skull deformation may occur at these high pressure regions. Further study into the loading rates and loading profiles between material types may aid in further improvement to computer

109

models of compressible foams and thus allow for better impact simulation and ultimately

better prediction of injury risks

6.3 Significance of this work to future studies

Many studies have been conducted that have measured head tolerance to impact by a

global measure of force or acceleration. This research led to the creation of the acceleration

tolerance limits currently in use to assess helmet performance. Helmet certification standards

for impact have greatly reduced the occurrence of catastrophic focal head injuries, however,

the occurrence of mTBI remains high. The future of helmet certification may include

calculation of internal stresses and strains in the brain obtained through finite element

modelling. Currently, the application of finite element modelling to injury prediction is an

expanding area of research and may help us gain a better understanding of mTBI. Several authors have used finite element modelling to create virtual helmets and foams in order to optimize mechanical properties (Mills 2007; Shuaeib, Hamouda et al. 2007; Cui, Rueda et al.

2009; Forero Rueda, Cui et al. 2011). In order to further validate these models, it is necessary to obtain empirical measure of the dynamically changing stresses at the contact surface. The instrumentation methods presented in this thesis can be used to assess these contact parameters and are applicable to many future studies. In fact, similar methods have already

been applied to validation of a skull fracture criterion (Rigby and Chan 2009; Rigby, Juhas et al. 2011). Pressure data from helmet impacts were collected using very similar instrumentation methods to those presented (Section 4.4.1 (Ouckama and Pearsall 2011)),

The independent presentation of a similar methodology and logic concerning the use of

contact pressure measurements in helmet tests justified the significance of this work and the

potential for its application to future research. Additionally, there is potential to integrate

110

similar load measurement devices in-helmet. There exists debate among researchers

regarding the use of in-helmet accelerometers and whether they are measuring the

appropriate head acceleration or helmet acceleration as often the two are not firmly coupled.

Contact force measurements would be insensitive to relative motion between the helmet and head and could provide a worthy alternative to linear acceleration given net force is directly

proportional.

6.4 Application of work to industry and applied fields:

The work presented in this thesis can be readily applied to various tasks within sporting, automotive and medical industries. At the development level, this tool can help aid engineers in selection of materials by providing quick feedback about the loading

characteristics and localized pressures. Additionally, if engineers want to simulate material

testing, the system developed in this thesis would verify the empirical assessment of model behavior. The real-world interaction of equipment geometry, material variances, and fit quality is important to quantify empirically and will provide a useful evaluation tool to various sporting industries. Expanding beyond helmets, this system has potential in any environment testing short duration impact, or where high speed measurement is desired.

The flexibility of sensor placement and low cost are beneficial for evaluation of medical protective devices (e.g. hip protectors) or in automotive testing using the Hybrid III

headform, for example the measurement of load distribution on the head during deployment

of an airbag.

111

CHAPTER 7: REFERENCES

112

113

Aldman, A. (1984). A method for the assessment of the load distrbuting capacity of

protective helmets. Goteborg, Chalmers University of Technology.

Allsop, D. and K. Kennett (2002). Skull and Facial Bone Trauma. Accidental injury :

biomechanics and prevention. A. M. Nahum and J. Melvin. New York, Springer:

254-276.

ASTM (2007). F1045-07 Standard Performance Specification for Ice Hockey Helmets. West

Conshohocken, PA, USA, ASTM International 10.1520/F1045-07

Beckwith, J., R. Greenwald, et al. (2012). "Measuring Head Kinematics in Football:

Correlation Between the Head Impact Telemetry System and Hybrid III

Headform." Annals of Biomedical Engineering 40(1): 237-248.

Biasca, N., S. Wirth, et al. (2002). "The avoidability of head and neck injuries in ice hockey:

an historical review." Br J Sports Med 36(6): 410-427.

Bishop, P. J. (1993). A comparison of the epoxy alloy and magnesium alloy

headforms. Safety in Ice Hockey. A. A. Ashare. Pennsylvania, ASTM. 3: 118-123.

Bishop, P. J. (1993). Protective equipment: biomechanical evaluation. Sports Injuries: Basic

Principles of Prevention and Care. P. A. Renstroem. Boston, Blackwell Scientific

Publications: 355-373.

Bishop, P. J. and J. Arnold (1993). The effectiveness of hockey helmets in limiting localized

loading on the head. Safety in Ice Hockey. P. J. Castaldi, P. J. Bishop and H. E.F.

Philadelphia, American Society for Testing Materials: 175-182.

Bishop, P. J., R. W. Norman, et al. (1984). "An evaluation of football helmets under impact

conditions." Am J Sports Med 12(3): 233-236.

114

Bland, J. M. and D. G. Altman (2010). "Statistical methods for assessing agreement between

two methods of clinical measurement." International Journal of Nursing Studies

47(8): 931-936.

Coulson, N. R., S. G. Foreman, et al. (2009). "Translational and Rotational Accelerations

Generated During Reconstructed Ice Hockey Impacts on a Hybrid III Head

Form." Journal of ASTM International 6(2).

CSA (2009). Z262.1-09 Ice Hockey Helmets. Mississauga, Ontario, Canada, Canadian

Standards Association.

CSA (2009). Z262.2-09 Face protectors for use in ice hockey. Mississauga, Ontario, Canada,

Canadian Standards Association.

Cui, L., M. A. F. Rueda, et al. (2009). "Optimisation of energy absorbing liner for equestrian

helmets. Part II: Functionally graded foam liner." Materials & Design 30(9): 3414-

3419.

Daneshvar, D. H., C. J. Nowinski, et al. (2011). "The epidemiology of sport-related

concussion." Clin Sports Med 30(1): 1-17.

Delaney, J. S., V. J. Lacroix, et al. (2002). "Concussions Among University Football and

Soccer Players." Clinical Journal of Sports Medicine 12(6): 331-338.

Di Landro, L., G. Sala, et al. (2002). "Deformation mechanisms and energy absorption of

polystyrene foams for protective helmets." Polymer Testing 21(2): 217-228.

DiMasi, F. (1995). Transformation of Nine-Accelerometer-Package (NAP) Data for

Replicating Headpart Kinematics and Dynamic Loading. John A. Volpe National

Transportation Systems Center, U.S. Department of Transportation, National

Highway Traffic Safety Administration.

115

Duma, S. M., S. J. Manoogian, et al. (2005). "Analysis of real-time head accelerations in

collegiate football players." Clin J Sport Med 15(1): 3-8.

Ebraheim, N. A., J. Lu, et al. (1976). "An anatomic study of the thickness of the occipital

bone. Implications for occipitocervical instrumentation." Spine 21(15): 1725-1729.

Echlin, P. S. (2012). "A prospective study of physician-observed concussion during a varsity

university ice hockey season. Part 1 of 4." Neurosurg Focus 33(6): E1: 1-7.

Echlin, P. S., E. N. Skopelja, et al. (2012). "A prospective study of physician-observed

concussion during a varsity university ice hockey season: incidence and

neuropsychological changes. Part 2 of 4." Neurosurg Focus 33(6): 1-11.

Faul, M., L. Xu, et al. (2010). Traumatic Brain Injury in the United States: Emergency

Department Visits, Hospitalizations and Deaths 2002–2006. Atlanta (GA), Center

for Disease Control and Prevention, National Center for Injury Prevention and

Control.

Forero Rueda, M. A., L. Cui, et al. (2009). "Optimisation of energy absorbing liner for

equestrian helmets. Part I: Layered foam liner." Materials and Design 30: 3405-3414.

Forero Rueda, M. A., L. Cui, et al. (2011). "Finite element modelling of equestrian helmet

impacts exposes the need to address rotational kinematics in future helmet

designs." Comput Methods Biomech Biomed Engin 14(12): 1021-1031.

Forero Rueda, M. A. and M. D. Gilchrist (2012). "Computational analysis and design of

components of protective helmets." Proceedings of the Institution of Mechanical

Engineers, Part P: Journal of Sports Engineering and Technology 226(3-4): 208-219.

Gennarelli, T. A. and L. E. Thibault (1982). "Biomechanics of acute subdural hematoma." J

Trauma 22(8): 680-686.

116

Goggio, A. F. (1941). "The Mechanism of Contre-Coup Injury." J Neurol Psychiatry 4(1):

11-22.

Greenberg, S., D. Gonzalez, et al. (1968). "Changes in Physical Properties of Bone Between

the In Vivo, Freshly Dead, and Embalmed Conditions." SAE Technical Paper

680783 doi:10.4271/680783.

Greenwald, R. M., J. T. Gwin, et al. (2008). "Head impact severity measures for evaluating

mild traumatic brain injury risk exposure." Neurosurgery 62(4): 789-798.

Gurdjian, E. S., V. L. Roberts, et al. (1966). "Tolerance curves of acceleration and

intracranial pressure and protective index in experimental head injury." J Trauma

6(5): 600-604.

Guskiewicz, K. M., S. W. Marshall, et al. (2005). "Association between recurrent concussion

and late-life cognitive impairment in retired professional football

players." Neurosurgery 57(4): 719-726.

Guskiewicz, K. M., S. W. Marshall, et al. (2007). "Recurrent concussion and risk of

depression in retired professional football players." Med Sci Sports Exerc 39(6): 903-

909.

Guskiewicz, K. M. and J. P. Mihalik (2011). "Biomechanics of Sport Concussion: Quest for

the Elusive Injury Threshold." Exercise and sport sciences reviews 39(1): 4-11.

Guskiewicz, K. M., J. P. Mihalik, et al. (2007). "Measurement of head impacts in collegiate

football players: relationship between head impact biomechanics and acute clinical

outcome after concussion." Neurosurgery 61(6): 1244-1252; discussion 1252-1243.

Gwin, J. T., J. J. Chu, et al. (2006). "Head impact telemetry system TM for measurement of

head acceleration in ice hockey." Journal of Biomechanics 39(1): S153.

117

Hakim-Zadeh, R. (2002). Durability of Ice Hockey Helmets to Repeated Impacts. Master of

Arts, McGill University.

Halstead, D. P. (2001). "Performance Testing Updates in Head, Face, and Eye Protection." J

Athl Train 36(3): 322-327.

Halstead, P. D., C. F. Alexander, et al. (2000). "Hockey headgear and the adequacy of

current designs and standards." Safety in Ice Hockey: Third Volume 1341: 93-100.

Hardy, W. N., C. D. Foster, et al. (2001). "Investigation of Head Injury Mechanisms Using

Neutral Density Technology and High-Speed Biplanar X-ray." Stapp Car Crash J 45:

337-368.

Hardy, W. N., M. J. Mason, et al. (2007). "A study of the response of the human cadaver

head to impact." Stapp Car Crash J 51: 17-80.

Hodgson, V. R. (1991). Impact Standards for Protective Equipment. Athletic Injuries to the

Head, Neck and Face. J. S. Torg. St.Louis, Mosby-Year Book: 28-43.

Hootman, J. M., R. Dick, et al. (2007). "Epidemiology of collegiate injuries for 15 sports:

summary and recommendations for injury prevention initiatives." J Athl Train 42(2):

311-319.

Hoshizaki, T. B., T. B. Robertson, et al. (2006). Impact properties of materials during drop

testing. Canadian Society for Biomechanics XIV. Waterloo, Canada.

Hutchinson, J., M. Kaiser, et al. (1998). "The Head Injury Criterion (HIC)

functional." Applied Mathematics and Computation 96: 1-16.

ISO (2003). 10256:2003 Head and face protection for use in ice hockey. Geneva,

Switzerland, International Standards Organization.

118

Jarquin-Valdivia, A. A., J. McCartney, et al. (2004). "The Thickness of the Temporal Squama

and Its Implication for Transcranial Sonography." Journal of Neuroimaging 14(2):

139-142.

King, A. I. (2000). "Fundamentals of impact biomechanics: Part I - Biomechanics of the

head, neck, and thorax." Annual Review of Biomedical Engineering 2: 55-81.

King, A. I., K. H. Yang, et al. (2003). Is Head Injury Caused by Linear or Angular

Acceleration. IRCOBI. Lisbon, Portugal, International Research Council on

Biomechanics of Injury.

Kleiven, S. (2006). "Evaluation of head injury criteria using a finite element model validated

against experiments on localized brain motion, intracerebral acceleration, and

intracranial pressure." International Journal of Crashworthiness 11(1): 65-79.

Lamb, L. and T. B. Hoshizaki (2009). "Deformation mechanisms and impact attenuation

characteristics of thin-walled collapsible air chambers used in head

protection." Proceedings of the Institution of Mechanical Engineers. Part H, Journal

of engineering in medicine 223(8): 1021-1031.

Langlois, J. A., W. Rutland-Brown, et al. (2006). "The epidemiology and impact of traumatic

brain injury: a brief overview." J Head Trauma Rehabil 21(5): 375-378.

Manoogian, S., D. McNeely, et al. (2006). "Head acceleration is less than 10 percent of

helmet acceleration in football impacts." Biomed Sci Instrum 42: 383-388.

Marjoux, D., D. Baumgartner, et al. (2008). "Head injury prediction capability of the HIC,

HIP, SIMon and ULP criteria." Accid Anal Prev 40(3): 1135-1148.

McAllister, T. W., J. C. Ford, et al. (2012). "Maximum principal strain and strain rate

associated with concussion diagnosis correlates with changes in corpus callosum

white matter indices." Ann Biomed Eng 40(1): 127-140.

119

McCrory, P., W. Meeuwisse, et al. (2009). "Consensus statement on concussion in sport - the

Third International Conference on Concussion in Sport held in Zurich, November

2008." Phys Sportsmed 37(2): 141-159.

McIntosh, A. S., T. E. Andersen, et al. (2011). "Sports helmets now and in the

future." British Journal of Sports Medicine 45(16): 1258-1265.

McIntosh, A. S. and D. Janda (2003). "Evaluation of cricket helmet performance and

comparison with baseball and ice hockey helmets." Br J Sports Med 37(4): 325-330.

McLean, A. J. (1995). "Brain injury without head impact?" J Neurotrauma 12(4): 621-625.

McLean, A. J. and R. W. G. Anderson (2005). Biomechanics of Closed Head Injury. Head

injury : pathophysiology and management

P. Reilly and R. Bullock. London; New York, Hodder Arnold ; distributed in the U.S.A. by

Oxford University Press.

Melvin, J. W. and J. W. Lighthall (2002). Brain Injury Biomechanics. Accidental Injury:

biomechanics and prevention. A. Nuauham and J. Melvin. New York, Springer-

Verlag: 277-302.

Mertz, H. J. (2002). Anthropometric Test Devices. Accidental Injury: biomechanics and

prevention. A. Nuauham and J. Melvin. New York, Springer-Verlag: 72-88.

Mertz, H. J., P. Prasad, et al. (1996). "Head Injury Risk Assessment for Forehead

Impacts." SAE Technical Paper 960099 10.4271/960099.

Mihalik, J. P., R. M. Greenwald, et al. (2010). "Effect of infraction type on head impact

severity in youth ice hockey." Med Sci Sports Exerc 42(8): 1431-1438.

Mills, N. J. (2003). Foam protection in sport. Materials in sports equipment. M. Jenkins.

Cambridge, England, Woodhead prublishing Ltd.: 9-44.

120

Mills, N. J. (2007). Finite element modelling of foam deformation. Polymer foams

handbook.

Motherway, J. A., P. Verschueren, et al. (2009). "The mechanical properties of cranial bone:

the effect of loading rate and cranial sampling position." J Biomech 42(13): 2129-

2135.

Mueller, F. O. (1998). "Fatalities from head and cervical spine injuries occurring in tackle

football: 50 years' experience." Clin Sports Med 17(1): 169-182.

Nahum, A. M., J. D. Gatts, et al. (1968). "Impact Tolerance of Skull and Face." Sae

Transactions 77: 178-&.

Newman, J. A. (2002). Biomechanics of head trauma: Head protection. Accidental Injury:

biomechanics and prevention. A. Nuauham and J. Melvin. New York, Springer-

Verlag: 303-323.

NHL. (2009). "Last Helmetless Player in the NHL." Retrieved January 2013, from

http://video.nhl.com/videocenter/console?id=38664.

NHL. (2012). "NHL Skills Competition All-Time Results, 1990-2011." Retrieved 9-Aug-

2012, from http://www.nhl.com/ice/page.htm?id=67157.

NOCSAE (2004). ND 030- 04m04a - Standard Performance Specifications for Newly

Manufactured Hockey Helmets. Overland Park, Kansas, USA, National Operating

Committee on Standards for Athletic Equipment.

Nusholtz, G. S., B. Wylie, et al. (1995). "Cavitation/boundary effects in a simple head impact

model." Aviat Space Environ Med 66(7): 661-667.

Ommaya, A. K. and A. E. Hirsch (1971). "Tolerances for cerebral concussion from head

impact and whiplash in primates." Journal of Biomechanics 4(1): 13-21.

121

Ouckama, R. and D. Pearsall (2008). Dynamic pressure mapping of the head-helmet

interface. North American Congress of Biomechanics (NACOB), Ann Arbor,

Michigan.

Ouckama, R. and D. Pearsall (2010). High Speed Force Measurement System for Evaluation

of Helmet Impact Load Distribution. Injury Biomechanics Symposium, Columbus,

OH.

Ouckama, R. and D. Pearsall (2012). "Impact performance of ice hockey helmets: head

acceleration versus focal force dispersion." Proceedings of the Institution of

Mechanical Engineers, Part P: Journal of Sports Engineering and Technology

226(3/4): 185-192.

Ouckama, R. and D. Pearsall (2012). Performance of Ice Hockey Helmets Following 10+

Years of Storage. 17th Biennial Meeting of the Canadian Society of Biomechanics,

Simon Fraser University, Burnaby, BC.

Ouckama, R. and D. J. Pearsall (2011). "Evaluation of a flexible force sensor for

measurement of helmet foam impact performance." J Biomech 44(5): 904-909.

Padgaonkar, A. J., K. W. Krieger, et al. (1975). "Measurement of Angular Acceleration of a

Rigid Body Using Linear Accelerometers." Journal of Applied Mechanics 42(3): 552-

556.

Pellman, E. J. (2003). "Background on the National Football League's research on

concussion in professional football." Neurosurgery 53(4): 797-798.

Pellman, E. J., D. C. Viano, et al. (2006). "Concussion in professional football: helmet testing

to assess impact performance--part 11." Neurosurgery 58(1): 78-96; discussion 78-96.

Post, A., B. Hoshizaki, et al. (2012). "Finite element analysis of the effect of loading curve

shape on brain injury predictors." Journal of Biomechanics 45(4): 679-683.

122

Post, A., A. Oeur, et al. (2011). "Examination of the relationship between peak linear and

angular accelerations to brain deformation metrics in hockey helmet

impacts." Comput Methods Biomech Biomed Engin

10.1080/10255842.2011.627559: 1-9.

Post, A., A. Oeur, et al. (2013). "An examination of American football helmets using brain

deformation metrics associated with concussion." Materials & Design 45(0):

653-662.

Raymond, D., C. Van Ee, et al. (2009). "Tolerance of the skull to blunt ballistic temporo-

parietal impact." J Biomech 42(15): 2479-2485.

Reid, S. E. and S. E. Reid, Jr. (1981). "Advances in sports medicine. Prevention of head and

neck injuries in football." Surg Annu 13: 251-270.

Reilly, P. L. and R. Bullock (2005). Head Injury: Pathophysiology and Management, Hodder

Arnold.

Rigby, P. and P. Chan (2009). "Evaluation of the Biofidelity of FMVSS No. 218 Injury

Criteria." Traffic Injury Prevention 10(2): 170-177.

Rigby, P., B. Juhas, et al. (2011). Evaluation of Biofidelity of ECE Regulation No. 22 Injury

Criteria, Paper No. 11-0366. International Technical Conference on the Enhanced

Safety of Vehicles. Washington, DC.

Rowson, S., S. M. Duma, et al. (2012). "Rotational head kinematics in football impacts: an

injury risk function for concussion." Ann Biomed Eng 40(1): 1-13.

SAE (1995). Instrumentation for Impact Test-Part 1: Electronic Instrumentation -SAEJ211-

1. Warrendale, PA, Society of Automotive Engineers.

123

Sharkey, E. J., M. Cassidy, et al. (2012). "Investigation of the force associated with the

formation of lacerations and skull fractures." International Journal of Legal Medicine

126(6): 835-844.

Shuaeib, F. M., A. M. S. Hamouda, et al. (2007). "A new motorcycle helmet liner material:

The finite element simulation and design of experiment optimization." Materials and

Design 28: 182-195.

Tekscan, I. (2009). "Flexiforce Sensor User Manual (rev G)." Retrieved May 6th, 2010,

from http://www.tekscan.com/pdfs/FlexiforceUserManual.pdf.

Thurman, D. J., C. M. Branche, et al. (1998). "The epidemiology of sports-related traumatic

brain injuries in the United States: recent developments." J Head Trauma Rehabil

13(2): 1-8.

Tommasone, B. A. and T. C. Valovich McLeod (2006). "Contact sport concussion

incidence." J Athl Train 41(4): 470-472.

Versace, J. (1971). "A Review of the Severity Index." SAE Technical Paper 710881

10.4271/710881.

Viano, D. C., I. R. Casson, et al. (2005). "Concussion in professional football: comparison

with boxing head impacts--part 10." Neurosurgery 57(6): 1154-1172; discussion

1154-1172.

Viano, D. C., I. R. Casson, et al. (2005). "Concussion in professional football: brain

responses by finite element analysis: part 9." Neurosurgery 57(5): 891-916; discussion

891-916.

Viano, D. C. and E. J. Pellman (2005). "Concussion in professional football: biomechanics

of the striking player--part 8." Neurosurgery 56(2): 266-280.

124

Viano, D. C., E. J. Pellman, et al. (2006). "Concussion in professional football: performance

of newer helmets in reconstructed game impacts--Part 13." Neurosurgery 59(3): 591-

606; discussion 591-606.

Walsh, E. S., A. Post, et al. (2012). "Dynamic impact response characteristics of a helmeted

Hybrid III headform using a centric and non-centric impact protocol." Proceedings

of the Institution of Mechanical Engineers Part P-Journal of Sports Engineering and

Technology 226(P3-4): 220-225.

Winslow, J. E. and A. O. Goldstein (2007). "Spectator Risks at Sporting Events." The

internet Journal of Law, Healthcare and Ethics 4(2).

Xu, J., I. A. Rasmussen, et al. (2007). "Diffuse axonal injury in severe traumatic brain injury

visualized using high-resolution diffusion tensor imaging." J Neurotrauma 24(5): 753-

765.

Yang, H., H. Mao, et al. (2011). Modeling of the Brain for Injury Prevention. Neural Tissue

Biomechanics. A. Gefen. New York, Springer-Verlag 10.1007/978-3-642-13890-4.

Yoganandan, N., F. A. Pintar, et al. (2011). "Comparison of head-neck responses in frontal

impacts using restrained human surrogates." Ann Adv Automot Med 55: 181-191.

Yoganandan, N., F. A. Pintar, et al. (2009). "Physical properties of the human head: mass,

center of gravity and moment of inertia." J Biomech 42(9): 1177-1192.

Zhang, L., K. H. Yang, et al. (2001). "Comparison of brain responses between frontal and

lateral impacts by finite element modeling." J Neurotrauma 18(1): 21-30.

Zhang, L., K. H. Yang, et al. (2004). "A proposed injury threshold for mild traumatic brain

injury." J Biomech Eng 126(2): 226-236.

125

126