Derailment Prediction Models for Canada’s Rail Network TAVIA WING TUNG CHOW A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE IN CIVIL ENGINEERING GRADUATE PROGRAM IN CIVIL ENGINEERING YORK UNIVERSITY TORONTO, ONTARIO March 2020 © Tavia Wing Tung Chow, 2020 Abstract Train related accidents, particularly derailments, can lead to severe consequences especially when they involve injuries or fatalities or involve dangerous goods that might result in environmental impacts. A literature review found that rail safety assessment and derailment prediction models have often been constrained by aggregated data which may yield inaccurate assessments of the safety performance of a rail network by, for example, failing to consider segment-level characteristics. This study focused on the development of segment-level derailment prediction models for Canada’s rail network using negative binomial and logistic regression modelling methods. The study used a network screening process to identify key segments of derailment concern. A thorough quantitative review of the models’ results and performance was conducted to understand the predictive capabilities and applications of the models in derailment prediction. The analytical approach and findings in this study have strong implications for advancing research on rail safety in North America. Keywords: rail safety, track segment, network screening, safety performance function, derailments, risk prediction, negative binomial regression, logistic regression. ii Dedication I want to dedicate my thesis work to my husband and new baby girl for giving me the motivation to pursue and finish my research. Also, to my family who provided tremendous support in every way possible in this journey. I would never be able to make this accomplishment without your love and understanding. iii Acknowledgements I would like to thank my supervisor, Dr. Peter Park for his continuous guidance, understanding and support. His professional knowledge and experience in industry were extremely helpful for my research. I would also like to express my appreciation to Dr. Kevin Gingerich and Dr. Mehdi Nourinejad for their time reviewing this thesis. I am thankful for Tayyab Shah and Seun Oluwajana’s contribution towards the model development. Additionally, I would like to thank Transport Canada for partially funding my research as part of the Collaborative Research and Development for Improved Rail Safety in Canada project. iv Table of Contents Abstract ............................................................................................................................... ii Dedication .............................................................................................................................. iii Acknowledgements .............................................................................................................................. iv List of Tables ............................................................................................................................ viii List of Figures ............................................................................................................................... x CHAPTER 1. INTRODUCTION ............................................................................................... 1 Research Overview ........................................................................................................... 1 Safety Network Screening ................................................................................................ 6 Research Goal and Objectives .......................................................................................... 6 Scope and Structure of Thesis ........................................................................................... 7 Chapter Summary ............................................................................................................. 7 CHAPTER 2. LITERATURE REVIEW ................................................................................... 9 Railway Safety .................................................................................................................. 9 Derailment Prediction Models and Factors ..................................................................... 14 Modelling Methods ......................................................................................................... 21 Chapter Summary ........................................................................................................... 25 CHAPTER 3. STUDY DATA ................................................................................................... 27 Data Sources ................................................................................................................... 27 Data Preparation and Validation ..................................................................................... 28 Descriptive Data Analysis ............................................................................................... 40 Chapter Summary ........................................................................................................... 48 CHAPTER 4. MODELLING METHODOLOGY .................................................................. 49 Input Data Preparation .................................................................................................... 49 Model Descriptions ......................................................................................................... 51 Model Form .................................................................................................................... 54 Empirical Bayes (EB) Approach ..................................................................................... 55 v Model Calibration and Validation ................................................................................... 56 Goodness-of-fit Tests ...................................................................................................... 56 Variable Selection ........................................................................................................... 61 Chapter Summary ........................................................................................................... 62 CHAPTER 5. DEVELOPMENT AND ANALYSIS OF PREDICTION MODELS ............ 64 Derailment Prediction Model for Canada ....................................................................... 65 Derailment Prediction Model for Eastern Canada .......................................................... 68 Derailment Prediction Model for Western Canada ......................................................... 71 Derailment Prediction Model for the Canadian National Railway ................................. 75 Derailment Prediction Model for the Canadian Pacific Railway .................................... 78 Chapter Summary ........................................................................................................... 81 CHAPTER 6. SAFETY NETWORK SCREENING ............................................................... 82 Hotspot Analysis for Canada .......................................................................................... 82 Hotspot Analysis for Eastern Canada ............................................................................. 84 Hotspot Analysis for Western Canada ............................................................................ 86 Hotspot Analysis for Canadian National Railway .......................................................... 88 Hotspot Analysis for Canadian Pacific Railway ............................................................. 89 Chapter Summary ........................................................................................................... 91 CHAPTER 7. DEVELOPMENT OF BINOMIAL LOGISTIC MODELS .......................... 92 Logit Model Development .............................................................................................. 92 Model Calibration and Validation ................................................................................... 93 Goodness-of-fit Tests for Logit Model ........................................................................... 93 Logit Model for Canada .................................................................................................. 97 Logit Model for Eastern Canada ................................................................................... 100 Logit Model for Western Canada .................................................................................. 104 Logit Model for Canadian National Railway ................................................................ 107 Logit Model for Canadian Pacific Railway .................................................................. 111 vi Chapter Summary ......................................................................................................... 114 CHAPTER 8. COMPARISON OF PREDICTION RESULTS ........................................... 116 Tetrachoric Correlation Analysis .................................................................................. 116 Comparison of Network Screening Results .................................................................. 117 Segment Identification Methods ................................................................................... 118 Discussion of Model Performances .............................................................................