Light Duty Natural Gas Engine Characterization

Home , Automotive engine, Honda Civic GX, Manifold injection

THESIS

Presented in Partial Fulfillment of the Requirements for the Degree Master of Science in the Graduate School of The Ohio State University

David Roger Hillstrom

Graduate Program in Mechanical Engineering

The Ohio State University

2014

Master's Examination Committee:

Professor Giorgio Rizzoni, Advisor

Professor Shawn Midlam-Mohler

Dr. Fabio Chiara

David Roger Hillstrom

2014

Abstract

The purpose of this project was to characterize the baseline performance of a

2012 Honda Civic Natural Gas vehicle including: designing experiments to generate complete performance maps, executing the experiments, and analyzing the experimental data. In the end, the results yielded a deep understanding of the 1.8 L four cylinder CNG engine’s combustion and air flow performance, as well as a good understanding of steady state engine out emissions. This information is used to isolate inefficiencies in design and propose possible avenues for improvement. The data that was acquired was then used to inform an existing 1-D computational model of the same engine in order to determine if, and where, the model was inaccurate, and determine what steps were necessary to improve it.

The resulting test data provides a data based background to the well-understood issues regarding a CNG port-fuel injected vehicle. The volumetric efficiency at low engine speeds was typically around 70%, resulting in an IMEP loss of about 15% compared to the engines peak possible performance. A CNG direct injection system is one possible solution to this problem. Additionally, the engine efficiency and spark timing map demonstrate that, even with the high compression ratio, the vehicle is not currently limited by engine knock. This available pressure headroom could be used with

ii boosting to improve the overall performance of the vehicle to bring it more in line with consumer expectations.

The development of this natural gas vehicle technologies research platform will allow the Center for Automotive Research at The Ohio State University to more easily pursue CNG related research topics. Some particular thrust areas of interest regarding this platform are the reduction of hydrocarbons while operating with lean burn, CNG direct injection, turbocharging optimization, and possibly even CNG / gasoline concomitant operation. The benefits to be had from these technology improvements can be gleaned by examining the baseline performance covered herein.

iii

Acknowledgments

I would like to thank my advisor Dr. Giorgio Rizzoni for providing the opportunities I have received since I arrived at The Ohio State University. He placed me in the natural gas consortium project which allowed me to very quickly get my hands dirty with heavy experimental work. Without this, I would have struggled to get such an involved and independent project to use for my Master’s Thesis.

I would like to thank the Honda Partnership Program for their donation of a 2012

Honda Civic Natural Gas for our research. Without their support, there would have been no foundation for this work to begin.

I would also like to thank my co-advisor Dr. Shawn Midlam-Mohler, Eric Shacht, and Dr. Fabio Chiara for their continued guidance in my work and all the technical help they have given me throughout my graduate career.

Finally I would also like to thank Dr. Jim Durand for providing extra opportunities for me to get involved around the Center for Automotive Research to ensure that my education extended beyond just the academic and into actual industrial and business relations.

Vita

January 1989 Born – Tulsa, Oklahoma

December 2011 B.S. Mechanical and Aerospace Engineering, Oklahoma State

University

August, 2012 to Present Graduate Research Associate,

The Ohio State University,

Center for Automotive Research

Fields of Study

Major Field: Mechanical Engineering

Table of Contents

Abstract ...... ii

Acknowledgments...... iv

Vita ...... v

List of Tables ...... viii

List of Figures ...... ix

Chapter 1: Introduction ...... 1

Brief Overview of the State of Natural Gas in US Energy ...... 1

CNG vs Gasoline ...... 3

CNG Vehicle Market Overview ...... 6

Chapter 2: Literature Review ...... 10

Direct Injection Executive Summary ...... 11

Geometric Design Considerations Executive Summary ...... 16

Hydrogen Executive Summary ...... 18

Dual fuel and Bi-fuel Executive Summary ...... 20

Combustion Executive Summary ...... 24

Noise, Vibration, and Harshness Executive Summary ...... 27

Emissions Executive Summary ...... 28

Chapter 3: Experimental Setup ...... 31

Throttle Model: ...... 33

Combustion Model ...... 35

Emissions ...... 39

Sample Timing ...... 39

Chapter 4: Engine Characterization Results ...... 43

Experimental Plan ...... 43

Thermodynamic Method for Locating Top Dead Center...... 45

Calculating for the Throttle Model ...... 47

Fuel Burn Rate Analysis for the Combustion Model ...... 49

Emissions and Efficiency Analysis ...... 55

Volumetric Efficiency ...... 60

Chapter 5: Integration With GT Power ...... 62

Chatper 6: Conclusions and Future Work ...... 67

Appendix A: Instrumentation ...... 75

vii

List of Tables

Table 1. Motoring Tests and Resulting TDC ...... 46

viii

List of Figures

Figure 1. EIA Natural Gas Data ...... 2

Figure 2. EIA Fuel Price History and Projections ...... 3

Figure 3. Laminar Flame Speed S1 at High Pressure and High Temperature [6] ...... 5

Figure 4. Mercedes B200 NGT With a Well-Integrated Fuel System [9] ...... 7

Figure 5. Curves of Injector Needle Lift and Gas Mass Flow ...... 12

Figure 6. Scheme of the proposed SS simplification ...... 13

Figure 7. WOT Torque-Speed Curves for Three Engine Classes ...... 15

Figure 8. Pre-Chamber Design Example ...... 17

Figure 9. Brake Thermal Efficiency against EGR ...... 19

Figure 10. Normalized Bi-fuel BSFC ...... 20

Figure 11. BMEP at full load, nominal performance for each fuel ...... 22

Figure 12. Laminar Flame Speed at 10x atmospheric pressure ...... 24

Figure 13. Schematic setup of a catalyst coated heat exchanger with bypass valve ...... 29

Figure 14. Aged bi-fuel taxi emissions measurements ...... 30

Figure 15. 1-D Engine Model Block Diagram...... 32

Figure 16. Laminar Flow Element Setup ...... 34

Figure 17. Manifold Air Pressure Setup ...... 34

Figure 18. Sample Fuel Burn Rate with and ...... 35

Figure 19. Cylinder Head Cross Section...... 37

Figure 20. Cylinder Head Cross Section...... 37

Figure 21. IMEP Error as a Function of TDC Error ...... 41

Figure 22. Crank Speed Fluctuation ...... 41

Figure 23. Testing DAQ Schematic ...... 42

Figure 24. Steady State Point Density ...... 45

Figure 25. CdA as a Function of Throttle Position ...... 47

Figure 26. CdA as a Function of Engine Speed [RPM] ...... 48

Figure 27. Exhaust Pressure vs. Cylinder Pressure during Exh. Valve Open ...... 50

Figure 28. P-V Diagram with Gamma Values Indicated for Exp. and Comp...... 51

Figure 29. Heat Release Rates ...... 53

Figure 30. CA50 as a Function of RPM and MAP ...... 54

Figure 31. CA10-CA90 as a Function of RPM and MAP ...... 54

Figure 32. Total Hydrocarbon Emissions ...... 56

Figure 33.Steady State CO [% Vol.] ...... 57

Figure 34. Steady State NOx [ppm] ...... 57

Figure 35. Excess Air Ratio as a Function of RPM and Torque ...... 58

Figure 36. Total System Efficiency ...... 59

Figure 37. Spark Advance ...... 59

Figure 38. Manifold Vol. Efficiency ...... 61

Figure 39. IMEP...... 61

Figure 40. MAF Error Using Throttle Input ...... 63

Figure 41. MAF Error Using MAP Input ...... 63

Figure 42. Unmodified MAF Modeling Error [%] ...... 64

Figure 43. Stock Intake Valve Lift Profile...... 65

Figure 44. MAF Error After Implementation of Tighter Valve Timing ...... 66

Chapter 1: Introduction

Brief Overview of the State of Natural Gas in US Energy

Natural gas as a transportation fuel is not a new idea; however large finds of natural gas, and the technology to recover this fuel at reasonable costs, have spurred increased national interest in CNG. Although NG has been used as a fuel in IC engines for a number of years, much development and optimization are still possible, both for the case of dedicated and bi-fuel engines. World-wide emphasis on CO2 emissions reduction/fuel economy improvements suggests that it may be worthwhile to make a small investment to understand what can be achieved with a CNG (dedicated or bi-fuel) engine in passenger car applications.

Even with oil production growing domestically, the US consumption vs production ratio still hovers around 2:1 [1, 2]. This reliance on imported product is fueling the search for domestic reserves of energy that may help the US reach ‘energy independence’. The United State Energy Information Administration (EIA) natural gas reserve data demonstrates a growing supply of proven natural gas reserves within United

States territory topping 300 trillion cubic feet [3]. This is about 12 times as much natural gas as the country consumes annually providing a relatively large supply cushion. These statistics can be viewed in Figure 1.

350 Proven Total US Reserves 300 Total Annual US Consumption

250

200

150

TCFnaturalof gas 100

0 1980 1985 1990 1995 2000 2005 2010 Year

Figure 1. EIA Natural Gas Data

Historically, NG fuel has maintained a cost around half of that of competing fuels such as gasoline and diesel. As Figure 2 demonstrates, The United States Energy

Administration Short Term Energy Outlook (STEO) does not predict this trend changing within the near future as the price is protected by the abundant supply mentioned previously. It is important to note that this natural gas price is based on EIA residential pricing information and not on common fuel pump prices. The reason for using residential pricing is that the data is readily available from reputable sources, such as the

US government, and this price is representative of the price one can find at the pump. If this price differential persists, the amount of money that one could save by operating a natural gas vehicle instead of a gasoline vehicle is significant.

8 Residential NG Price Gasoline Diesel STEO Projections 6 Crude Oil Present Time

2 Fuel Prices in $/GGE in Prices Fuel

0 Jan-10 Jan-12 Jan-14 Jan-16

Figure 2. EIA Fuel Price History and Projections

CNG vs Gasoline

The engine design approach behind a CNG vehicle and a gasoline vehicle should be different due to some key differences in the thermodynamic properties of the fuel.

This section serves to give an overview of what some of the differences are and how they can have a drastic effect on performance, emissions, or reliability.

Methane, the primary component of natural gas, is composed of one carbon atom and four hydrogen atoms. This H/C ratio of 4:1 is advantageous to an engine’s emissions as compared to gasoline which has an H/C ratio of about 1.85 [4]. The reason being that during combustion, heat energy and oxygen mix with the methane to break the molecular bonds and re-combine them. This ideally turns carbon into and hydrogen into . Thus, if there is less carbon and more hydrogen in the reactants, there should

3 be less and more in the products. This is indeed the case for CNG as compared to gasoline as natural gas observes a emissions reduction of about 20% [5]. This is the reason that natural gas is typically regarded as a ‘greener’ fuel than its petroleum based brethren.

Another benefit of natural gas is its RON octane number which is typically much higher than gasoline. This allows the fuel/air mixture to reach a much higher temperature, and therefore a much higher pressure, before auto-ignition starts to occur. This extra pressure headroom can be utilized during an engine’s design stage to achieve a more efficient engine by increasing the compression ratio, a more powerful engine by turbocharging inlet air, or some combination of these [5]. Additionally, the peak compressed flame speed of natural gas is nearer to stoichiometry than gasoline and there is no charge cooling effect from CNG which reduces the desire to run rich in high performance operating modes. For example, gasoline race engines typically operate around a lambda of 0.9 [6]. Figure 3 demonstrates these points.

Figure 3. Laminar Flame Speed S1 at High Pressure and High Temperature [6]

The caveat with the increased pressure headroom, lack of latent heat of vaporization, and low lubricating capability is higher mechanical and thermal stresses on the engine. If an engine is to be designed to run reliably on CNG and highly optimized for CNG it must be reinforced or modified as compared to a typical gasoline engine

(crank-shaft, connecting rod, valve seats, etc.) [7]. This and other drawbacks associated with natural gas result in some interesting design challenges.

As CNG is a gaseous fuel, it takes up a much larger volume than liquid fuels like gasoline. As a result of this, typical injectors take a relatively long time to inject all the fuel for a particular engine cycle. For example, some CNG race engines must incorporate two injectors in each manifold port to meet the fuel delivery demand and can’t consider direct injection presently as no available CNG fuel injector is able to deliver enough fuel in the shortened window of time [7]. This is a design concern for any engine attempting

5 to reach a higher RPM. Moreover, if the engine uses manifold injection, the gaseous fuel expansion post-injector displaces a lot of air, leading to a detrimental effect on volumetric efficiency [7].

Within the United States, natural gas vehicles are still rare in the light-duty market. Other countries however, such as Germany and Italy, have a much higher adoption rate as consumers are beginning to understand the significant cost savings that can be realized.

CNG Vehicle Market Overview

Within the United States, CNG vehicles have yet to catch major consumer attention with a light duty NG vehicle market share of about 0.1% [8]. In the light duty sedan segment there is only one LD car available which is the natural gas version of the

Honda Civic. Europe however, has had more success, with a larger variety of models at

19. Part of my initial work with this topic was to extensively analyze the market in

Europe. This helped us to gather insight into what vehicle technology is available which can direct our efforts when exploring how these vehicles can and should be improved.

Additionally, understanding the differences between what is available in the United

States, and what is available in Europe, might help to direct our attention to the possible reasons for the low domestic NGV adoption rate.

Some of the major takeaways from the study are the key differences between the technology present in the Honda Civic that we have experimented on, and the vehicles

6 available in Europe. The main idea behind a well-designed NGV is that the consumer should not realize it is an NGV, it should be indistinguishable from a gasoline or diesel counterpart unless it is better, else a consumer might not prefer the CNG version of a car.

Our study focused on a broad range of topics such as fuel system integration, vehicle performance, refueling infrastructure, and government incentives. I will only discuss our findings related to vehicle performance here.

Figure 4. Mercedes B200 NGT With a Well-Integrated Fuel System [9]

From a performance perspective, every natural gas vehicle (NGV) offered in

Europe is a bi-fuel vehicle [10] meaning that the vehicle can run on either CNG or gasoline fuel. This helps to alleviate range anxiety associated with being uncomfortable with the CNG refueling station density in one’s region. The downside of running a bi-fuel vehicle is that the engine must be capable of handling gasoline which, in all present cases, means it is not optimized for CNG. Additionally, every single vehicle is CNG port- 7 fuel injected, resulting in a substantial volumetric efficiency loss as compared to gasoline operation. For some of these vehicles, this results in lower performance than gasoline.

However, others that have a turbocharger can control the boost pressure such that output horsepower between both CNG and gasoline operation is virtually the same.

The bi-fuel engines present in these vehicles appear to be purpose built for gasoline and then slightly modified to accept CNG. This leads to sub-optimal performance as CNG engines should be designed around the much higher octane number.

An advanced CNG engine could utilize various control methods to control compression ratio, and boost, in order to maximize performance and/or efficiency at every given moment. The desire for such an optimized vehicle is what drives our efforts with the present research.

This study is meant to lay the ground work for future efforts of OSU-CAR in the arena of natural gas fuelled vehicles. The testing done herein will include information on volumetric efficiency, steady state emissions, in-cylinder pressure curves, and heat release analyses of a production 2012 Honda Civic NG. This study will help to provide insight into what specific design changes should be considered in order to harness more efficiency or performance where it is available, while simultaneously demonstrating the emissions and efficiency benefits that CNG is already providing.

This research is done alongside another study utilizing the GT-Power software to virtually explore the performance potentials of such an engine with modifications such as boosting and direct injection. The experimental data harvested herein can be compared with the results from the virtual test bench in order to validate the model. Afterwards,

8 design changes can be implemented on the model for an initial perspective on expected returns. The combination of these two theses will provide a compass for future experimental and computational efforts.

Chapter two of this thesis will cover the literature review that was performed in preparation of this thesis work, covering many aspects of CNG engine technologies from research publications released in the last five years. Chapter three will cover the motivations behind the experimental setup based on the intended use of the gathered information. Chapter four will cover the analysis performed on the experimental data and resulting key discussions. Chapter five will cover the integration of the information into a previously developed GT power model for validation purposes. Finally, chapter six will cover the main resulting conclusions of the work and recommendations for future studies to continue the work.

Chapter 2: Literature Review

The focus of the work with the 2012 Honda Civic is to develop a platform in which to pursue more focused topics of global interest. In preparation for the work with the 2012 Honda Civic, an extensive literature review was performed giving insight into what the major thrust areas are related to CNG. The results of the review will dictate the future directions of this research. It is therefore paramount to ensure that all modern developments are fully understood. As such, this literature review was focused on the analysis of academic publications within the last seven years (2006+) related to compressed natural gas automotive engine technologies. The study leveraged the

University’s access to publication networks such as SAE Digital Library, ScienceDirect,

OhioLink, SAGE Journals, and SpringerLink in order to review over 100 publications with around 80 being analyzed in depth. The papers that were deemed relevant to our interests could be lumped into the following categories: direct injection, geometric design, hydrogen mixtures, dual-fuel and bi-fuel technologies, combustion, NVH, and emissions [11].

Direct Injection Executive Summary

Direct injection is a topic that is garnering much interest in the pursuit of CNG engine optimization. The volumetric efficiency losses from manifold injection are widely known, and direct injection is the most obvious cure for this dilemma. However, since

CNG is a gaseous fuel it introduces several new dynamics to the injection system that must be considered in order to have a well-functioning engine.

Recently, much effort has been devoted to the creation of accurate injection models for CNG. The dynamics of the injector can arguably be very complex. Due to the gaseous nature of the fuel, pressure wave phenomena are present within the fueling system. In order to circumvent any negative effects from this, the fuel rail must be designed intentionally such that the opening and closing of each injector does not negatively affect subsequent injections through pressure wave troughs propagation.

Additionally, the act of opening and closing the injector needle itself is subject to fluctuations. In a gasoline injector, the liquid fuel acts as a sort of damper on the injector needle as it is opening and closing such that the needle does not have significant dynamic fluctuations. However a CNG needle will bounce when it is commanded open or closed causing pulsations of fuel to leak through the opening (Figure 5). At high engine speeds, these pulses of fuel can account for up to 1/3 of the total fuel injected rendering the understanding of this behavior significantly important [5].

Figure 5. Curves of Injector Needle Lift and Gas Mass Flow

In regards to 3-D modeling of the injection process, there are several levels of complexity that have been tested. The unified goal of each is to somehow prevent an asymptotically complex mesh near the injector tip typically required in order to accurately model the behavior of this critical region. Firstly, a method of simplification is to disregard phenomena upstream of the injector, accomplished by instead modeling a sudden increase of pressure just inside the injector. This pressure rise then propagates into the cylinder modeling the fuel injection. It has been observed, using the STAR-CD environment, that no matter how this pressure increase is modeled, eventually the mass flow rate coming into the cylinder will reach a steady state flow, typically within 1/3 the time of a typical injection [12]. This idea of steady state flow was than extrapolated to create a much simpler 3-D CFD code (Figure 6).

Figure 6. Scheme of the proposed SS simplification

This simpler injection method operates under the assumption that fuel mass flow creates a quasi-steady jet into the cylinder. This jet carves out a conical area that is predicted by a phenomenological model, whose boundaries impose the initial conditions of the 3-D CFD code. This technique yields a much lower computational time due to the fact that the injector tip, generally represented by the finest mesh, is now lumped into a steady flow model [13]. The drawbacks associated with this methodology are that the injector bounce is not accurately modeled, and any nuances present around the injector tip are ignored.

It is important to note that these previously mentioned techniques yield quicker results at the expense of accuracy as they do not properly model the injector behavior, which is the most important aspect of a CNG direct injection system. The fluctuations present at the injector, even if controls are held constant, can be significant. Conversely

13 to these simplification models, FKFS approached the problem in a slightly different way.

In order to maintain accuracy, a fine mesh is implemented near the injector tip, but a slight modification to how the code views the incoming fuel can improve the mesh performance without having to shrink to an unrealistically small size. The code views the incoming fuel as very small droplets, not gas, which initially pass through the mesh; than after some small distance from the injector, these fictive gas droplets evaporate without any latent heat. Additionally this model considers the mass flow rate injection fluctuations brought on by the injector tip bounce. All of these models have their advantages, but the modeling approach from FKFS yields very promising results in terms of its capability of predicting fuel jet development [5].

The ability to accurately model the fuel jet development is a significant step in accurately modeling the combustion chamber in an engine; moreover the results garnered from these simulations all yielded insight into important design considerations for flow development in a direct injection engine running on CNG fuel. Namely, the characteristic attributes of such a fuel injection. CNG direct injection, even at high injection velocities, has little impact on the flow field within the cylinder, such that, compared to its gasoline brethren, there is little impact on the level of turbulence in the fuel/air mixture. The charge motion must be controlled through some other design parameter such as the intake port layout or piston head. In order to represent this quantitatively, an injector was tested in different orientations and the ISFC was measured to determine if any improvements could be made. As the injector was angled from 50 degrees to 90 degrees, the ISFC changed by no more than +- 1% [14].

An additional important parameter to consider when discussing a CNG direct injector is the injection timing with respect to top dead center. In general, for a homogenous mixture, the emissions and combustion stability improve the more advanced the injection window, all the way up to intake valve close. The mixture needs as much time as possible to smooth out any rich and lean combustion zones [15].

Figure 7. WOT Torque-Speed Curves for Three Engine Classes

In conclusion, the performance improvements a CNG direct injection system has on engine can be substantial. In one case, a direct injection system was able to improve power and torque by around 20% across the majority of the RPM range as compared to a port injected system (Figure 7). Additionally, the BSFC of the engine was lowered by

16% yielding a significantly more efficient engine [16]. Another method for improving the efficiency of an engine is to consider a stratified charge. However, the successful

15 implementation of stratified charge combustion is reliant on the design of the cylinder geometry as discussed in the next section.

Geometric Design Considerations Executive Summary

The design of the cylinder geometry can be leveraged to enhance charge motion control and is required for fuel stratification in order to enhance engine efficiency. In regards to the application of a stratified charge, the design of the piston does not have to be overly complex to be effective at creating a stratified charge near the spark plug.

Additionally, exotic fuel injection systems have been tested in an attempt to maximize the efficiency of the engine, along with exotic crankshaft designs.

A simple and effective method for implementing stratified charge operation with

CNG fuel is a simple bowl in the center of the piston head with the fuel injector in a perpendicular orientation directly above. As the piston is rising to top dead center, the injector floods the bowl with fuel. The bowl can then hold the fuel in the middle of the cylinder reasonably well to be ignited by the spark plug at TDC [17, 18]. Several simulations have validated that this can successfully maintain ignitable mixtures near the spark plug. The narrower the bowl, the leaner the overall mixture in the cylinder can be.

Experimental results show that this mixture formation method is ignitable but are not conclusive as to the effect on the emissions and engine performance. Other injection methods have also been explored involving more unique methods of mixture ignition.

Figure 8. Pre-Chamber Design Example

One of these methods is the utilization of a pre-chamber, a small crevice volume, where ignition takes place as seen in Figure 8. Various approaches have been tested on how this should be best utilized however all of them function on the basis that the pre- chamber is ignited and then the flame propagates out to combust the mixture in the main chamber. The novelties lie in how the pre-chamber is ignited whether by spark plug or compression ignition, and the controls on how the two chambers interact [19, 20].

Thermal efficiencies have been observed as high as 44.1% using compression ignition, while spark ignition also leads to stable engine operation at mixtures leaned out to a lambda as high as 1.4. The compression ignition methodology is so sensitive that it was only successfully controlled at steady state operation which is not conducive to its implementation in a motor vehicle.

Other exotic designs have been tested as well such as the utilization of a z-shaped crankshaft which allows for different compression and expansion strokes [21]. The implementation of a higher expansion stroke allows for the engine to operate more

17 efficiently by extending the workable area on a P-V diagram reflecting in a 2.6% increase in thermal efficiency and a 7% increase in measured fuel economy as compared to a standard engine with a similar compression ratio; the only downside being the additional complexity in the crankshaft.

Hydrogen Executive Summary

Hydrogen has been proposed to solve a setback typically associated with CNG operation which is the low laminar flame speed as compared to gasoline. In practice, this flame speed difference results in longer 10-90 CAD burn durations for the CNG fuel.

This problem can be circumvented by diluting the fuel with hydrogen as hydrogen is known to burn very quickly. Many researchers have explored how different proportions of hydrogen can affect the combustion within the engine with the general consensus being that more hydrogen means a more efficient engine. However, the issue lies in making hydrogen readily available to the consumer. A method for hydrogen integration is custom tailored synthetic natural gas which is gaining popularity as an energy storage medium. Experimental work dictates that hydrogen dilution of 40% by volume can lead to about a 2% increase in thermal efficiency due to the higher laminar flame speed [22].

Hydrogen presence in the fuel also allows for a lower coefficient of variance due to the ease with which the fuel/air mixture ignites, however power output tends to decrease with increases in hydrogen dilution percent.

Figure 9. Brake Thermal Efficiency against EGR

When attempting to optimize the engine performance, there is a concern with regards to hydrogen enriched CNG fuel mixtures. EGR is oftentimes used in order to increase the efficiency of a CNG engine, however excessive amounts of hydrogen result in excessive amounts of water in the exhaust gas [23]. Hence, care must be taken when recycling too much water into the combustion chamber as this can lead to high cyclic variability even with fairly modest amounts of EGR. The limit depends on the amount of hydrogen in the fuel, but for a 25% by volume mixture, The EGR upper bound is about

8% less than if hydrogen was not present (Figure 9).

Dual fuel and Bi-fuel Executive Summary

CNG fuel has not yet integrated itself throughout the nation’s infrastructure. This lack of refueling stations can lead to a well-known ‘range anxiety’ issue among consumers. Therefore, CNG technology is presently only pushing hard into the heavy duty scene due to the low cost of the fuel and the tendency for this customer to perform a more complete financial analysis. On the light duty side of things, it is still necessary to relieve this anxiety issue through granting CNG vehicles the capability of running on alternative sources of power such as gasoline. The drawback with such an engine is that an optimized CNG engine generally would not work with gasoline, therefore design compromises must be made.

Figure 10. Normalized Bi-fuel BSFC

Nevertheless, CNG/gasoline bi-fuel engines may be necessary in courting the technology into the market. As such, the performance benefits and possibilities are a popular topic of study. Directly comparing the two fuels during operation yields the fact that the CNG fuel can readily be more efficient in terms of BSFC (Figure 10) [24, 25].

The fuel inherently burns leaner than gasoline due to its higher stoichiometric air to fuel ratio. Additionally, at high engine speeds, gasoline must sometimes inject extra fuel to cool the exhaust gas for fear of harming the catalyst. In these high speed operating regions, the CNG fuel can offer much better efficiency as it remains at stoichiometry throughout the RPM band; unfortunately most bi-fuel engines have port injected CNG resulting in a significantly detrimental impact to power . Moreover, having access to two fuels on a vehicle raises the question of what happens if both fuels are used simultaneously? Could one attempt to harvest the benefits of both, the knock resistance of

CNG, and the volumetric efficiency benefits of gasoline?

Research has been performed to see how concomitant injection may be optimized and what benefits could be extracted from such a system. If the system is capable of actively varying the proportion of gasoline and CNG entering the combustion chamber there are optimization algorithms that could be employed depending on the desired outcome (Figure 11) [26, 27, 28]. If maximum power is desired, the control scheme is dictated as follows: less CNG at low RPM, mixed gasoline and CNG at mid RPM,

Mostly CNG at high RPM. Typically at low engine speeds, CNG fuel has less power than gasoline due to the severe impact from volumetric efficiency. Therefore at low RPM it could be considered best to use as much gasoline as possible. In the midrange RPM, it is

21 best to have a concomitant injection of the two with a gasoline mass fraction around 40% which yields equivalent power to gasoline only operation. In the high RPM region it is best to taper off the gasoline fraction in order to take advantage of CNG’s strong knock resistance and to avoid having to enrich the gasoline injection. If maximum efficiency is desired, than a different algorithm could be employed. This aspect of optimization yields another level of control to bi-fuel vehicles.

Figure 11. BMEP at full load, nominal performance for each fuel

In the case of a CNG/gasoline bi-fuel engine, the CNG operational mode generates higher thermal stresses on the internal components. Hot spot temperatures within the cylinder can reach up to 20 degrees C higher for the CNG fuel. This is generally due to the lack of latent heat benefits the gasoline fuel enjoys when it evaporates. Hence, the cooling jacket must be developed with care [29].

There is another popular bi-fuel system, more commonly referenced as dual-fuel which is usually the combination of CNG and diesel. In most cases, these engines are converted diesel engines which operate with compression ignition. Compression ignition however is not preferable to the CNG fuel. Therefore, CNG is considered the primary fuel and enough diesel is injected such that its auto-ignition can serve as the combustion catalyst to propagate a flame through the CNG/air mixture. A typical ratio for such a mixture is 80-90% CNG with the rest being diesel pilot fuel [30, 31, 32]. For a vehicle operating with this fuel, it is beneficial to the BSFC of the vehicle to increase the intake air temperature. However, increased temperature within the cylinder can increase the risk of knock onset, which has been an issue among active vehicles in Thailand [33]. The increase of intake air temperature can also benefit CO emissions due to the fact the hotter mixture promotes a more complete combustion. Conversely, this increased temperature has a negative effect on NO emissions. The intake air temperature can be controlled via exhaust gas recirculation

Combustion Executive Summary

The most important facet of efficient CNG operation is to develop a thorough understanding of the combustion process. This section will attempt to bring together all of the research ideas and innovations that have been discovered in the last five years in order to shed light on the important considerations regarding the CNG combustion process.

Figure 12. Laminar Flame Speed at 10x atmospheric pressure

The first topic of importance is that of the laminar flame speed which gives insight into combustion quality. At ambient temperature and pressure, CNG has a higher flame speed than that of gasoline. However at 10x ambient pressure, which is brought

24 about by a typical engine compression stroke, the peak laminar flame speed of gasoline is

65% faster than that of CNG (Figure 12). Something else of interest is that the laminar flame speed of CNG is highest nearer to stoichiometry, therefore there is little incentive to en-richen the mixture, whereas the peak for gasoline resides at an excess air ratio of about 0.9. The rich gasoline mixture allows for a higher laminar flame speed and manages to generate cooler exhaust gases through the consumption of heat through the evaporation of the excess fuel [6].

The CNG fuel, being of a gaseous nature, does not absorb heat through vaporization like gasoline fuel. This is a very important consideration in the design of a race engine using a turbocharger as one of the design constraints is the temperature of the exhaust gas entering the turbocharger. The temperature of this exhaust gas should not exceed the tolerances of the materials used in the turbine. Typically, if CNG and gasoline are running on two separate identical engines at stoichiometry, the exhaust gas from the gasoline engine will be slightly higher than that of the CNG engine. However in a race engine, the gasoline fuel mixture is run at a lambda of 0.9, the region promoting the highest laminar flame speed. At this air/fuel ratio, the extra fuel in the cylinder absorbs some of the heat causing the exhaust gas temperature to decrease to levels lower than the stoichiometric CNG engine. The higher temperature in the exhaust gas for the CNG fuel can be detrimental to the turbine reliability, but beneficial in the sense that this increases the enthalpy of the fluid. This means a smaller turbine can be used while still maintaining the same level of boost [6].

Boosting is a concept that has gained momentum as of late in the advent of fuel economy relevance in engine design. The combination of engine downsizing and boosting means that the engine can bridge the compromise between fuel economy and performance. In this respect, CNG is extremely knock tolerant making it favorable to such an application. So much so, that a particular engine running on CNG is capable of more power through excessive boosting than that same engine running on gasoline, even if the CNG is indirectly injected into the manifold intake port. This is once again discussed in terms of its application to motorsports. Regulations restrict the ability of a motorsports CNG engine to perform due to regulated air restrictors and maximum allowable peak pressure, but if these rules are lifted, the engines could generate just as much power, or more than gasoline, while still maintaining less CO2 emissions [6].

In contrast to CNG’s involvement in motorsports is the pursuit of maximal efficiency. The fuel consumption of the engine can be reduced through leaning the combustible mixture. Within CNG engines, the lean limit typically falls around an excess air ratio of 1.2 to 1.3 for a homogenous mixture; however, experiments have shown that proper stratification techniques can bring the lean limit as high as 1.8 [34]. Another technique for enhancing efficiency is through exhaust gas recirculation. The probability for ignition of CNG fuel in the combustion chamber goes up with increased air temperature and the implementation of EGR allows for control of this temperature.

Moreover, the presence of the exhaust gas gives unburned hydrocarbons a second chance at combustion reducing these emissions from the vehicle. However the intake air temperature must be carefully controlled such that temperatures do not rise to knock

26 inducing levels. Of course, the CNG’s tendency to knock is dependent on its octane number, which varies depending on what CNG fuel is used.

CNG fuel can vary significantly from one fuel pump to another. Therefore, an engine programmed to operate on CNG fuel must be prepared to accept these variations.

Engine performance can be properly maintained as long as the vehicle implements some form of adaptive AFR control. The ECU needs to recognize when the methane number of the fuel has dropped and switch its control parameters accordingly. Typically, for CNG fuel, a lower methane number means a higher presence of the heavier hydrocarbons ethane and butane. These heavier hydrocarbons are beneficial to combustion as they increase the density of the fuel, which reduces volumetric efficiency losses, and increases the laminar flame speed. Conversely, lower methane number fuel has a higher tendency to knock [35].

Noise, Vibration, and Harshness Executive Summary

CNG fueled engines have been found to operate more quietly than an identical gasoline engine. The rate of pressure increase during combustion correlates with the noise emissions of the process, and due to the lower flame speed of CNG fuel, this results in quieter combustion. This means that a higher compression ratio can be utilized while still maintaining the same level of noise output from combustion, or the engine can just be more consumer friendly [36]. However, one point of concern for this fuel is the injectors themselves. CNG injectors can cause loud pulsation noises due to pressure

27 waves when the injector needle bounces open and closed. This is highly depended on the injector design, but is a concern nonetheless [37].

Emissions Executive Summary

The emissions of a CNG engine are heavily dependent on the design of said engine but nevertheless certain trends exist. The major players in emissions regulations are hydrocarbons, CO, NOx, and CO2. Thus, most of the papers which discuss emissions focus on these key players. Generally, a CNG engine can be expected to produce less

CO2 than a gasoline or diesel counterpart due to the inherent nature of the fuel: CNG has a much higher hydrogen/carbon ratio than gasoline or diesel. The NOx emissions are dependent on the peak pressure/temperature within the combustion cycle and show no clear trend for comparison with gasoline. Moreover, the CO and HC emissions can vary from one engine to the next and are heavily dependent on engine speed. CNG fuel has a tendency to hide in crevice volumes leading to incomplete combustion; additionally CH4 is a light hydrocarbon which more easily passes through 3-way catalysts leading to increased THC emissions.

Figure 13. Schematic setup of a catalyst coated heat exchanger with bypass valve

These THC emissions remain a topic of focus when discussing CNG vehicles. A catalyst can let slip large quantities of hydrocarbons before it is properly lit off, and typically the CH4 coatings are near the back of the catalyst so it is the last section to reach its operational temperature [38]. A solution to this problem is a bypass valve that ensures the CH4 coatings are heated promptly (Figure 13). The low exhaust temperature typical of CNG vehicles can additionally contribute to the delayed catalyst light off time.

Research has also been performed into a currently unregulated emission, ammonia.

Ammonia is expected to soon join the roster of unwanted emissions by government agencies around the world, and should it succeed in doing so, CNG engines have a tendency to produce about half as much ammonia as gasoline or diesel engines. This is due to the fact that much of the hydrogen required for NH3 slips through the catalyst as methane, deprived of the opportunity to recombine with nitrogen. However, as catalysts evolve, so too may this problem.

Figure 14. Aged bi-fuel taxi emissions measurements

According to a study performed on an aging taxi fleet, the emission benefits enjoyed by CNG vehicles as compared to gasoline should remain throughout the lifetime of the vehicle. The purpose of this study was to compare the emissions of heavily used bi-fuel engines operating in each mode (Figure 14). As expected the emissions were far worse due to the aging of the catalyst, but the trends present in new cars are still present in the used ones [39].

In conclusion, there are many opportunities for improvement and optimization of natural gas engine performance. The performance potential of the high octane fuel, the improvement of consumer acceptability due to lower combustion noise, and the control of excess hydrocarbon emissions are just a few of these topics. This study hopes to shed light on the relevance of these thrust topics to the 2012 Honda Civic Natural Gas presently under investigation.

Chapter 3: Experimental Setup

The goal of this project is to develop a 2012 Honda Civic Natural Gas as an experimental platform for exploring natural gas engine technologies. The first two tasks conducive to this goal are: 1. Validate an existing computational model of a CNG engine;

2. Generate emissions out maps of the engine. Once the model is validated, it can be used as a starting point for research into engine modifications such as direct injection, turbocharging, EGR, etc. The modeling software that will be used is the 1-D computational software GT-Power.

It is important to note that the typical convention for characterizing an engine or performing engine experimentation is to remove the engine from the vehicle and place it on an engine test bench [41]. For the entirety of our experimentation, testing will actually be performed in-vehicle on a light-duty chassis dynamometer. This allows us to get data from a more realistic operating environment, with the caveat that there are additional unknowns in our torque measurements, such as transmission dynamics and torque converter losses. For the experiments performed herein, steps will be taken to reduce the effects of the transmission on the data as much as possible and are detailed in the experimental plan section of chapter 4.

The design of the experiments takes focus on what data is needed to inform the computational model. Fortunately, a model has been previously developed at The Ohio

State University by a student projects team known as EcoCAR. The model is a 1-D engine computational model in the GT Power environment for a 2008 Honda Civic GX.

GT Power attempts to model the gas dynamics of the plumbing in an engine efficiently by only taking into consideration the forward and backward propagation of pressure waves [40]. Pipe bends, splits, etc. are taken into account through pressure loss coefficients. This method ignores the effect of full 3-D phenomena, but greatly improves the model run time as one can simulate a whole engine in a matter of minutes with reasonably accurate results. EcoCar’s particular model was calibrated to run on E85, but the engine geometry between the 2008 Civic and the 2012 Civic are nearly identical. For this reason, this model serves as an excellent starting point for our research. A flow chart of how the model works can be seen in Figure 15. The black arrows represent piping geometry, the red blocks represent models that have already been calibrated by the

EcoCAR team, and the green blocks represent the models that will be calibrated using this experimental data.

Throttle Fuel Injector Combustion Exhaust Ambient Ambient Model & Intake Port Model Port

Figure 15. 1-D Engine Model Block Diagram

Throttle Model:

The throttle model in GT Power uses compressibility equations for flow through an orifice (equation 3.1 & 3.2) where 𝑚 ̇ is the mass flow rate of air through the throttle,

𝑝 is ambient pressure, 𝑅 is the specific gas constant for air, 𝑇 is the ambient temperature, 𝑝 is the intake manifold pressure, 훾 is the specific heat ratio for air, and

is an effective discharge coefficient and area that changes as a function of throttle opening. In GT-Power, one needs to input the array for as a function of throttle opening angle. This can be easily calculated as long as all the other parameters are known, therefore pressure and temperature will be taken at the specified locations during experimentation.

𝑚 ̇ = ( ) { [ ( ) ]} if ≥ . 28 Equation (3.1) √

𝑚 ̇ = 훾 { } if < . 28 Equation (3.2) √

The mass flow rate of air will be measured by forcing all of the air the engine intakes through a laminar flow element. The laminar flow element (LFE) forces the air stream through extremely small flow channels effectively dropping the Reynolds number into the laminar regime by lowering the characteristic length of the flow. This is beneficial as it creates a linear relationship between the differential pressure of the flow and the volumetric flow rate. The density for air can be calculated from the ideal gas law

33 and used to turn the volumetric flow into mass flow. One caveat from using a device like this with an IC engine is the non-uniformity with which the engine breathes. The chaotic pulsation of pressure waves as each cylinder breathes has to be dampened. This is accomplished in our case by inserting a large oil drum for the pressure waves to disperse in between the engine air box and the LFE and sealing the entire system.

The other measurements of temperature and pressure are acquired using K-type thermocouples, a barometer, and piezo-resistive pressure sensors. Pressure is measured immediately post-throttle for the 𝑝 and with a barometer for the 𝑝 terms. Ambient temperature is measured at the LFE while 훾 and 𝑅 are considered constant at 1.4 and

287 [ ] respectively. Pictures of the instrumentation can be viewed in Figure 16 and

Figure 17.

Figure 16. Laminar Flow Element Setup Figure 17. Manifold Air Pressure Setup

Combustion Model

GT-Power has several options for modeling the combustion. The software can even interface will full 3-D software packages like Star-CD. For the purposes of this research, the wiebe curve fit will be used to model combustion which uses experimental data to simulate the heat release rate from the fuel. A sample normalized burn rate taken from the Honda Civic can be seen in Figure 18. The Wiebe function, seen in Equation

3.3, is not predictive in any sense, it just happens to be a function that matches up well with the desired shape [41].

Figure 18. Sample Fuel Burn Rate with and

Equation (3.3)

The coefficients in the Wiebe function are not what GT-Power takes as inputs.

For GT-Power, the inputs are the crank angle position after TDC corresponding to 50% fuel burned. Additionally, GT Power requires the 10% to 90% burn duration in crank angle degrees. Examples of these are on Figure 18. In order to calculate the fractional burn rate of the fuel, information is needed on the heat release rate within the combustion chamber during the combustion cycle. An equation for this can be derived from the first law of thermodynamics and is seen below in Equation 3.4.

Equation (3.4)

In order to model the heat release rate inside the cylinder we need access to the in- cylinder pressure and its derivative and the in-cylinder volume and its derivative. The volume vector and its derivative can be calculated from crank kinematics presuming one knows the geometry of the cylinder [41]. The in-cylinder pressure, however, must be measured directly during experimentation. One way to accomplish this is using an extremely fast response piezo-electric pressure transducer mounted directly into the engine. The initial plan for the project was to mount one sensor in each and every

36 cylinder. However, due to time constraints and the complexity of the cylinder head, it was decided to only measure pressure in one cylinder on the outside of the engine.

Section views of the cylinder head can be seen in Figure 19 and Figure 20. In order to get pressure in all cylinders, the sensors would have to be mounted on the top of the combustion chamber; however there is very limited room to drill a hole here due to the size of the intake/exhaust valves and the position of the spark plug. Conversely, the driver side of the engine had an exposed section of solid aluminum that was very easy to access and deemed sturdy enough to allow for a small hole. This location is indicated by the red square in Figure 20.

Figure 19. Cylinder Head Cross Section Figure 20. Cylinder Head Cross Section

Piezo-electric pressure transducers are great at monitoring very fast changes in pressure, however the pressure measured is relative to some arbitrary zero point. This arbitrary zero point can drift over the course of a test, requiring another known pressure source in which to anchor too. For the purposes of these experiments the known pressure source that will be used is the exhaust pressure just after the exhaust ports. Inserted here will be a piezo-resistive pressure transducer which measures absolute pressure and does

37 not drift dramatically with time. As the exhaust ports are open, the two pressure sensors are temporarily very close to each other and attached to the same stream of air and should therefore report the same pressure. This information will be used in post processing to move the in-cylinder pressure measurement higher or lower with each cycle as necessary to ensure the two pressures are identical during this period of time.

The information calculated from the in-cylinder pressure is typically reported in crank angle degrees and not in time. It is therefore necessary to measure the crank position alongside the other measurements. In order to accomplish this, a 180 tooth encoder disc with photo-interrupters for 1/revolution and every two degrees are mounted to the passenger side of the crankshaft. It is worth noting that no production off the shelf encoder would fit in the engine bay, therefore this disc was custom manufactured at The

Ohio State University using a water-jet CNC machine.

When performing an engine characterization, it is typical to remove the engine from the vehicle and instrument it inside an engine test cell attached to an engine dynamometer. This removes the unknowns associated with the transmission and drivetrain. Unfortunately, in the case of our research facility, the engine test cells are not outfitted for natural gas operation, specifically in terms of the high pressure fueling infrastructure and fire safety codes. It was therefore necessary to design the instrumentation in order to accommodate installation on a chassis dynamometer. Ergo, all instrumentation had to fit within the engine bay. In this case, the testing could be performed taking advantage of the vehicle’s onboard fueling system. The detailed

38 experimental plan for performing the steady state tests in this environment is detailed at the start of chapter 4.

Emissions

In addition to the validation of the GT power model, this project also seeks to map the emissions output from the engine. In order to accomplish this, a Fourier transform infrared spectroscopy (FTIR) emissions analyzer was used to measure the engine-out emissions before the three way catalyst. The FTIR analyzer compares the absorption spectrum of the engine emissions sample gas against a known spectrum provided by a constant stream of pure nitrogen. Each molecule absorbs different wavelengths in the spectrum and the intensity of the absorption changes with increased quantity of the molecule [42]. The quantitative capabilities of the FTIR have been calibrated by the company that developed the instrument. Using this device we can track the volume percent of several gases within sample gas stream such as methane, carbon dioxide, etc.

Sample Timing

The data acquisition system used is made up of a high speed National Instruments data acquisition card and a low speed National Instruments data acquisition card inside a

Dell Workstation running Labview. Additionally, the CAN data is recorded using a separate laptop within the vehicle and the emissions data is measured using another

39 separate laptop near the man DAQ computer. The reason for the separate computer in the vehicle is due to the need to monitor CAN data in real time during testing, and the availability of funds to set the system up. The need for a separate computer for the emissions system is due to the fact that the FTIR was graciously loaned to us by the company Stoneridge as a total package including its own software, computer, and hardware. The two laptops record data at a standard low frequency rate, but as these tests are steady state, the samples will be averaged for each test. On the other hand, the Dell

Workstation responsible for in-cylinder pressure measurements, uses a more complex method for organizing the samples.

It is typical to trigger samples based on the digital signal from the encoder; this ensures that data is recorded in the crank angle domain and not in the time domain.

However, as the encoder disc only had 180 teeth, and due to manufacturing reasons, we could only use the leading edge of each tooth’s signal (the digital step up and not the digital step down). This would translate into a sample resolution of two degrees. A very small error in the location of top dead center, due to this relatively coarse resolution, could lead to incorrect values in our determination of the wiebe curve fit parameters for the GT Power model. If these values are off by even one degree, it will lessen the predictive capabilities of the model by more than 5% in terms of IMEP and hence, BMEP or torque / power prediction. This point is emphasized in Figure 21.

10 1200 IMEP Error [%] 8

4 1150 2

-2 IMEP Error [%] Error IMEP 1100 -4

-6 [RPM] Speed Engine

-8

-10 1050 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 100 200 300 TDC Error [CAD] data number

Figure 21. IMEP Error as a Function of Figure 22. Crank Speed Fluctuation TDC Error

In order to circumvent this issue, the data will instead be measured in time at very high frequencies (80 kHz). This allows us to interpolate between the encoder teeth with real data. This requires an assumption that the encoder rotational speed is constant in between each encoder tooth for a duration of two crank angle degrees, but is a reasonable assumption based on the small variability in speed seen in Figure 22. Additional details on the instrumentation can be viewed in Appendix A with the Labview details in

Appendix B. A diagram of the testing equipment can be seen in Figure 23.

The experimental setup with provide pressures and temperatures in key location in order to give insight into the performance of the engine in all its operational regions.

Additionally the logging of the data in time, rather than triggered by encoder pulse, will allow much finer crank angle resolutions leading to more accurate determinations of top dead center and indicated mean effective pressure. This will help minimize the error associated with the collection of in-cylinder pressure helping to give more accurate data.

Figure 23. Testing DAQ Schematic

Chapter 4: Engine Characterization Results

This chapter will serve to give a detailed breakdown of the fundamental objectives of the thesis: the design of the experiments, the execution of the experiments, and the analysis of the experimental data. The goal of the experimentation is to generate several maps of steady state performance for the vehicle. The performance maps will typically be presented with engine speed vs torque in the x-y plane, with a number of different parameters being exchanged for the z. This data will be useful for a number of tasks that require understanding the baseline performance of the engine, namely, the calibration of the GT Power model, and the understanding of engine-out emissions for the purpose of guiding further exhaust research.

Experimental Plan

For the purpose of steady state map generation, the experimentation should match the accuracy of a speed locked engine test cell as much as possible. In order to accomplish this, the chassis dyno will be locked to a specific vehicle speed, such as 10 mph. This will hold the wheels at a fixed speed, which if the transmission remains engaged, will indirectly hold the engine at a fixed speed. With the engine speed fixed, different levels of torque can be reached allowing one to populate a torque vs. speed map

43 with steady state data. While performing these tests, as the Civic has an automatic transmission, it must be ensured that the transmission does not shift. Even though The

Honda Civic has an automatic gearbox, the transmission controls are conducive to this type of testing as it allows the car to remain in second gear, never changing to 1st or 3rd for any reason. Of course, this does not remove the transmission and torque converter dynamics from our results. With these thoughts in mind, the following procedure has been developed:

Procedure:

1. Place Civic in 2nd gear 2. Bring dyno up to speed 3. Throttle to desired torque 4. Monitor MAP until deviations remain < 0.2 psi for 5 seconds (steady) 5. Collect Data 6. Repeat 3-5 for a total of six torque steps from low throttle to full throttle 7. Repeat 2-6 for speeds 5, 10, 15, … 55 mph.

In each operating point, it is recommended to take at least 100 engine cycles of data for averaging purposes as the engine is susceptible to cyclic variability and must be averaged for meaningful results [41]. This experimentation will record 250 engine cycles at each data point to ensure that the answers are as robust as possible. The entire operating space can be seen in Figure 24.

120 5 mph 15 mph 100 25 mph 35 mph 45 mph 80 55 mph

20 Torque After Out Dividing Torque [N-m] Ratio Final Drive

0 0 1000 2000 3000 4000 5000 6000 Engine Speed [RPM]

Figure 24. Steady State Point Density

The torque shown is that which is measured by the chassis dyno. Therefore this is post-transmission torque. It can be seen that the torque converter is playing a large role in the test at speeds less than 25 mph as it appears to be multiplying the torque reaching the wheels. The slight creep up in RPM with higher torque can be attributed to more wheel spin at higher speeds and a combination of wheel spin and torque converter slip at speeds less than 25 mph.

Thermodynamic Method for Locating Top Dead Center

The method used for locating top dead center is explicitly from reference 43, “An

Universally Applicable Thermodynamic Method for T.D.C. Determination” by Marek J.

Stas. This method allows the determination of top dead center from in-cylinder pressure

45 and volume alone, as it is based on the heat transfer equation seen previously as 3.4. For robustness, seven different motoring tests were performed for the application of this technique. Motoring was performed on the chassis dynamometer by disabling the fuel injector to our measured cylinder. The engine could than run using the other three cylinders remaining warmed up. Table 1details the experimental points explored and the final results of the TDC method. It is important to note that the ECU has safety measures in place during miss-fire conditions such as this, so the ability to explore higher throttle positions was hindered, nevertheless, engine speed effects on the algorithm were explored. The TDC location is normalized to Test 1. The final results show six of the seven tests within 0.1 degrees of one another, with one test 0.4 degrees off. These results will minimize the error in our characterization of the engine’s performance due to incorrectly phasing the volume vector with the data.

Table 1. Motoring Tests and Resulting TDC

Test # Engine Speed Peak Pres. Loc. TDC Loc. [deg. [RPM] [deg. ATDC] ATDC] 1 670 -0.7 0.0 2 670 -0.7 0.0 3 670 -0.7 0.0 4 1330 -0.5 0.1 5 2140 -0.3 0.4 6 3640 -0.2 0.0 7 3640 -0.2 0.1

Calculating for the Throttle Model

In order to inform the throttle model, it is necessary to calculate the discharge coefficient across the butterfly valve. As the effective area changes with increased throttle angle, it is convenient to lump the discharge coefficient and area term together.

The results of the CdA term as a function of throttle position can be seen in Figure 25.

-4 x 10

CdA

2 CdA

0 0 10 20 30 40 50 60 70 80 90 Throttle Position [degrees]

Figure 25. CdA as a Function of Throttle Position

The values seem to follow a trend until after around thirty degrees. From this point the throttle jumps up to 84% where the pedal is actually at full throttle. The torque however does not increase immensely from the last position around 30% to wide open throttle. If we examine the CdA values from another angle, it might shed light on this phenomenon.

-4 x 10

CdA

2 CdA

0 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 Engine Speed [RPM]

Figure 26. CdA as a Function of Engine Speed [RPM]

From the perspective of engine speed, it can be seen that the peak CdA value increases with engine speed. It is likely that above thirty degrees, the throttle is no longer responsible for chocking the flow. At the larger throttle angles, another flow restriction is the weakest link in permitting a higher mass flow rate for air. In this sense, the throttle is oversized for this engine. This makes sense as the throttle should not limit the engine’s full torque capability. Nevertheless, for the purpose of the GT Power model, a double sigmoid curve fit can be fitted to the CdA values beneath 30 degrees and the highest CdA value at full throttle as the throttle will not be responsible for choking the flow in the upper region.

Fuel Burn Rate Analysis for the Combustion Model

This section will detail the process of using the in-cylinder pressure to calculate the fuel burn rate. The in-cylinder pressure must first be used to calculate the heat release rate and heat transfer within the cylinder as given by equation 3.4. The starting point for this analysis is to first clean up the noise of the pressure signal and shift it using the exhaust pressure. The theory that when the exhaust valve is open, the two pressures should be completely identical, is not entirely true. This is due to the fact that much of the exhaust manifold is integrated into the cylinder block making it difficult to place the exhaust transducer as close as it needs to be to the exhaust valve. Therefore the exhaust transducer is actually far enough away that it does not see some of the cylinder pressure phenomena, demonstrated in Figure 27. In order to remedy this fact, a range of crank angle degrees that are safely within exhaust valve open and close were averaged before shifting. This range is indicated by the red bars in the figure. The entirety of exhaust valve open to close was not used as during the higher speed and power cycles, the cylinder pressure remained much higher than the exhaust pressure for several degrees after exhaust valve open, which severely skewed the average.

1.8 Cyl P 1.7 Exh. P 1.6 Averaging Range

1.5

1.4

1.3

1.2 Pressure [bar] Pressure 1.1

0.9

0.8 350 400 450 500 550 CAD

Figure 27. Exhaust Pressure vs. Cylinder Pressure during Exh. Valve Open

The next step is to smooth the pressure signal. A first order Butterworth filter was used within Matlab to this effect. The Butterworth filter cutoff frequency was altered until it could be seen that the filtered pressure signal did not lose any of the magnitude of the un-filtered signal and then applied across all cycles. The filtfilt command within

Matlab was used to ensure that the filtering did not shift the data. An ensemble average was then performed to blend the 250 engine cycles into one average curve for the test.

This average curve is the one that will be used for future calculations.

One of the variables of equation 3.4, is the specific heat ratio (gamma) for the mixture inside the combustion chamber. For air, this value is 1.4, however we cannot assume that the combustion mixture will have the same value. One method for calculating the specific heat ratio for the mixture during combustion is to look at the loglog plot of the pressure vs volume [41]. The slope of the compression and expansion curves is representative of the specific heat ratio for the mixture at those times. This value

50 is somewhat complex to measure exactly as it is also dependent on the heat transfer through the walls, therefore the compression and expansion strokes will result in different gammas. For the calculations performed herein, the compression and expansion stroke gammas will be averaged in order to find the final value that will be implemented into the equation. Figure 28 demonstrates the difference in gammas, and how they appear on a typical P-V diagram.

2 10 averaged pressure gamma exp = 1.286 gamma comp = 1.1424

1 10

Pressure [bar] Pressure 0 10

-1 10 TDC 1/4 1/2 3/4BDC Volume

Figure 28. P-V Diagram with Gamma Values Indicated for Exp. and Comp.

With the gamma value calculated, the net heat release rate can be easily calculated. The net heat release rate includes the chemical energy released by the fuel and the loss of energy dissipating through the cylinder walls. In order to calculate the burn rate of the fuel, these two parameters need to be separated. The method used on these results to calculate the heat loss through the walls is the Woschni correlation for the

51 convection heat transfer coefficient (equation 4.1 & 4.2) [41]. In these equations represents the gas side heat transfer coefficient, B represents the cylinder bore, p is the instantaneous cylinder pressure, T is the instantaneous cylinder temperature, w is the average cylinder gas velocity, is the mean piston speed, is the engine displaced volume, 𝑇 𝑝 are the reference temperature, pressure, and volume respectively,

𝑝 is the instantaneous motored pressure, and 𝑚 are model parameters.

. . = 𝑝 𝑇 Equation (4.1)

= [ ̅ 𝑝 𝑝 ] Equation (4.2)

In these equations T, Sp, 𝑝 , and the reference variables have not yet been discussed. T will be calculated using the ideal gas law, Sp can be calculated using engine geometry and engine speed, and the motored pressure trace will be calculated using the ideal gas behavior of = . For the motored pressure trace, the n value will be 1.3 instead of 1.4 for air as this value must include the heat loss through the cylinder walls [41]. After the gas side heat transfer coefficient has been calculated, the simple equation for convective heat transfer can be used. However an assumption has to be made as to the temperature of the wall, which will be kept constant at 370 K which is just slightly higher than the temperature of the engine coolant. The heat release rate for a particular test can be seen in Figure 29.

60 Q net 50 Q loss Q chem 40

10 dQ/dtheta [J/deg] dQ/dtheta

-10

-20 0 90 180 270 360 CAD

Figure 29. Heat Release Rates

The narrow peaks are attributed to noise induced in the pressure signal from the spark plug. These peaks were removed before further calculations were made on the fuel burn rate. The mass fraction burned can now be calculated from integrating the chemical heat release rate using equation 4.3 [41].

∫ = Equation (4.3)

∫

In order to clean the graph of the heat release rate, the result will be forced to zero when the spark happens, as no chemical heat should have been released up to this point.

Additionally, once the function reaches its maximum value, it will stop calculating in order to remove the effects of the heat loss later in the expansion stroke. Figure 18 53 demonstrates the mass fraction burned and associated burn angles necessary for GT

Power. The resulting maps that will be implemented in GT Power can be seen in Figure

30 and Figure 31. GT Power will then reconstruct the fuel burn rate based on this information and the operating point that is being simulated.

Figure 30. CA50 as a Function of RPM Figure 31. CA10-CA90 as a Function of and MAP RPM and MAP

From these figures, the effect of engine speed on burn duration can be readily viewed. Additionally, the variation in the 50% burn location is fairly small over most of the operating region. It appears that at high load and high speed the prolonged burn duration begins to have an impact.

Emissions and Efficiency Analysis

In this section, the engine out emissions will be presented and discussed. The

FTIR used is calibrated to measure a variety of hydrocarbons, the carbon monoxide, the carbon dioxide, and the nitrous oxides present in the exhaust gas stream. The exhaust gas was sampled prior to the three-way catalytic converter. These emissions are of particular concern because they are regulated by the environmental protection agency and must not exceed a certain value in order to allow the vehicle to go in to production. The values regulated by the EPA are a vehicle’s g/mile emissions over particular drive cycles. As the experiments performed were focused on steady state performance, discussions will be limited to the volume fractions of each exhaust gas. These volume fractions translate into higher or lower g/mile emissions depending on the air flow through the engine. The first emission to be discussed will be the total hydrocarbons as it is arguably the most significant CNG engine out emission.

The hydrocarbons emitted from a CNG engine are primarily composed of methane (CH4). Methane is much harder to ignite in a catalyst then the heavier hydrocarbons associated with gasoline and diesel as it requires a much higher temperature [45]. As demonstrated by Figure 32, the hydrocarbon emissions from this particular vehicle are somewhat consistent between 1000-1800 ppm across the entire engine operating space. There is however a spike in hydrocarbon emissions at low speed and low torque which will be explained later on after the discussion of air to fuel ratio.

Figure 32. Total Hydrocarbon Emissions

The other greenhouse gas emissions of interest are the nitrous oxides (NOx), and carbon monoxide (CO). These gases absorb more radiation than CO2 and are therefore of particular interest to regulatory committees such as the EPA. The Honda Civic steady state emission for these gases can be seen in Figure 33 and Figure 34. The area of intensified hydrocarbon emissions also lends to somewhat higher carbon monoxide emissions which come about as a result of incomplete combustion.

Figure 33.Steady State CO [% Vol.] Figure 34. Steady State NOx [ppm]

The emissions data can be used to determine the total system efficiency of the vehicle by determining the energy available from the fuel flow rate and comparing it against the measured power at the wheels. In order to do this, the emissions data must first be used to determine the air to fuel ratio using the carbon balance method. As we know the mass air flow rate into the engine accurately due to the LFE, we can use this air to fuel ratio to accurately determine the fuel flow rate [41].

The FTIR records the volume percent of several chemicals. The chemicals of particular interest to us are the total hydrocarbons, carbon monoxide, water, and carbon dioxide. The equation for converting these values into an air to fuel ratio is commonly known as the carbon balance method seen in equation 4.4.

= [ ] Equation (4.4)

In the carbon balance equation, y is the hydrogen to carbon ratio of the fuel and each molecule is its volume fraction. y is heavily fuel dependent and, as discussed previously, can vary from one sample of natural gas to another. For the purposes of this analysis, the value used by the EPA to demonstrate the calculation mechanisms for fuel economy with CNG fueled vehicles will be used, 3.97 [44]. The M values in the equation are the molecular weights of air and fuel respectively. Air is a known quantity and will be

28.96 while fuel can be calculated by the H/C ratio, 16.01. Finally, the map of A/F can be seen in Figure 35 presented as the excess air ratio with a stoichiometric A/F ratio of 17.2 as detailed by Heywood for methane [41].

Figure 35. Excess Air Ratio as a Function of RPM and Torque

Figure 35 shows that, for the most part, the engine runs ever so slightly lean without any large deviations until the vehicle is at full throttle. When the vehicle is at full throttle, the excess air ratio drops as low as 0.96 in order to enhance combustion. This is

58 still considerably less than most gasoline fueled vehicles may dare to go when pushed to full throttle [6]. As mentioned previously, the A/F ratio can be used to determine the fuel flow rate, which can be used to calculate total system efficiency (Figure 36).

Figure 36. Total System Efficiency Figure 37. Spark Advance

The total system efficiency falls within expectations for a high compression ratio internal combustion engine such as this, with a maximum value around 31% [41]. It is also interesting to see that the efficiency remains very near 30% for most of the high torque region. Comparing this information with the spark timing Figure 37 is evidence that the vehicle is not retarding the spark out of fear for engine knock. Ergo there is likely pressure headroom available for extra performance either through turbocharging or potentially further increasing the compression ratio. Also of note, below 15 mph, it would seem that the torque converter has a massively detrimental impact on total system efficiency so these values are not necessarily representative of real world performance and have been removed from the plot to avoid confusion. The efficiency in the low rpm region depends heavily on torque converter lock-up controls at these engine speeds in 59 higher gears. In the future it would be interesting to re-perform these tests with a torque converter lock-up override in order to get more consistent results in this region.

Nevertheless, the majority of the engine operating space is well populated.

Volumetric Efficiency

Natural gas is, as the name suggests, a gaseous fuel. Therefore its physical density is far less than that of a liquid fuel such as gasoline and diesel. This is detrimental to engine performance if the fuel injectors lie outside the combustion chamber as it does with port injection. When the intake valves are open, the natural gas pushes a lot of the air out of the way as they surge together into the cylinder leading to less air in the combustion chamber, which in turn yields to less power from the combustion process [6].

Volumetric efficiency helps to quantify this effect representing the ratio of the actual air in the cylinder against the amount of air that could get in the cylinder based on its total volume. The map of the volumetric efficiency for the Honda Civic can be viewed in

Figure 38 along with the indicated mean effective pressure in Figure 39.

Figure 38. Manifold Vol. Efficiency Figure 39. IMEP

The full load torque is the most important with regards to volumetric efficiency as part load, by definition, is already restricted performance. At full load, the volumetric efficiency ranges from between 72-81%. This results in an IMEP that ranges from about

9.5-11. The volumetric efficiency is directly responsible for the ~15% drop in IMEP at the lower RPM values. These results verify the potential for either direct injection or turbocharging to overcome this performance drawback.

Chapter 5: Integration With GT Power

This experimental work was done in parallel with a GT power model investigation of the same engine. The model was previously developed by EcoCAR and validated using their experimental data obtained on an engine dynamometer with the

2008 Honda Civic GX engine. In the validation of the model, one of the primary modeling components of interest is the mass air flow through the engine. If the model can accurately predict airflow, it will more accurately predict fuel flow providing the foundation for a good engine model [46].

Initially, the engine was simulated by inputting engine speed and throttle position and observing the resulting mass air flow values. As can be seen in Figure 40, this yields fairly inaccurate results especially at lower mass air flows. This is due to the somewhat complicated nature around modeling the throttle opening angle’s effect on air flow through the intake system. In fact, the GT Power user document recommends not attempting to model the throttle exactly, but rather forcing the throttle to achieve the desired end result: its ability to match the desired manifold air pressure. Therefore, for the immediate purposes of validating the model, the throttle calibration data that was collected will not be implemented into the GT Power model until the air flow from imposing forced manifold air pressure values is correct. It is therefore desirable to remove the throttle variables from the equation through the use of a PID controller which

62 seeks out the required throttle angle to match the experimentally acquired manifold air pressure. If the manifold air pressure is forced to the desired value in the GT Power model, this sets up the air path to more accurately simulate flow through the rest of the system including the intake/exhaust valves [46].

60 60

50 50

40 40

30 30

20 20

10 10

MAF-Simulation Data[g/s] MAF-Simulation Data[g/s] MAF-Simulation

0 0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 MAF-Experimental Data[g/s] MAF-Experimental Data[g/s]

Figure 40. MAF Error Using Throttle Figure 41. MAF Error Using MAP Input Input

As can be witnessed by Figure 40 and Figure 41, the experimental data collected from the EcoCAR team has excellent agreement with simulation results. However in this case, if the model is taken as is and run to simulate the new experimental operating points, the model consistently under predicts total air flow. This trend can be seen in

Figure 42. This could either mean that the model is not correctly simulating the engine, or there is some error associated with the imposed manifold air pressure. In other words, the measured manifold air pressure may not represent the actual values. Due to time considerations, the validation of this hypothesis will be included as future work for the project.

20 20 mph 25 mph 10 30 mph 0 35 mph 40 mph 45 mph -10 50 mph 55 mph -20 MAF Error [%] Error MAF 5 mph -30

-40 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Manifold Air Pressure [bar]

Figure 42. Unmodified MAF Modeling Error [%]

If the model is not accurately predicting air flow through the engine, even when the manifold air pressure is forced to match the experimental data, the combustion model has little chance of accurately predicting torque. It was therefore pertinent to investigate the cause for the consistent air flow offset and move the combustion model validation to future work for the project. As the manifold air pressure is matching the experimental data by design of the model, the air flow restriction would seem to be between the manifold and the cylinder. This leads one to believe the valves from the 2012 Honda

Civic Natural Gas must not be correctly represented in the model. The stock valve timing implemented in the model can be viewed in Figure 43.

Figure 43. Stock Intake Valve Lift Profile

The suspicion is that the somewhat delayed closure of the implemented valve timing is allowing a lot of air to flow back through the intake ports during compression.

In order to test this, the valve timing was modified to advance the closing crank angle position. This was accomplished by scaling the valve lift duration by a factor of 0.95, effectively making the valve lift profile 5% shorter, and also shifting it such that the intake valves continue to open at the same point. The simulations were then re-ran with the new valve timing to see if this had a positive effect on the mass air flow error.

20 20 mph 25 mph 10 30 mph 0 35 mph 40 mph 45 mph -10 50 mph 55 mph -20 MAF Error [%] Error MAF 5 mph -30

-40 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Manifold Air Pressure [bar]

Figure 44. MAF Error After Implementation of Tighter Valve Timing

From Figure 44, it can be seen that this is extremely beneficial to matching the predicted air flow against the experimental airflow. In conclusion, the cylinder head needs to be re-flowed in a flow lab and the valve lift profiles need to be re-calculated using information from the new engine. The assumption that the intake and exhaust valves were the same here is apparently invalid.

The flowing of the new head and experimental characterization of the new valve lift profile will not be covered herein, but will be included in the desired future work. As the air flow is not very well predicted by the model, the implementation of the combustion model will be put on hold until the issues regarding valve timing have been resolved.

In conclusion, the air flow of the 2012 engine seems to have been improved over the 2008 engine of which this model was derived. Nevertheless, the model still provides a good starting point for the work, as the intake manifold, and air path are very similar, however the cylinder head geometry changes must be taken into account.

Chatper 6: Conclusions and Future Work

The primary goal of this research was to design and execute the experiments necessary to characterize the performance of a 2012 Honda Civic Natural Gas, analyze the data from the experiments, and prepare the results in such a way that they can be used to inform a computational model. Additionally, this work was designed to build up the

2012 Honda Civic Natural Gas into a new experimental platform at The Ohio State

University for further natural gas research.

The experimental procedure has been developed for characterizing an engine on a chassis dynamometer through the use of the dyno’s ability to lock the roll speed. Locking the roll speed is an effective way of locking the engine speed through the transmission allowing the operator to step through different values of torque in order to populate a steady state performance map. The instrumentation and data acquisition systems necessary to facilitate such experimentation were also developed. The data acquisition system is mobile and can be moved to any experimentation room necessary, and all instrumentation was designed and installed in-vehicle such that the vehicle can still operate like normal, allowing it to be moved where necessary.

The testing results reinforce what is commonly known to be design challenges on natural gas vehicles. The volumetric efficiency is a performance limiting factor as the natural gas displaces air during the injection process. At full throttle the volumetric

67 efficiency resides around 80% at high RPM but drops as low as 70% at lower engine speeds. A potential avenue to overcome this issue is to move the fuel injectors into the combustion chamber operating as a direct injection vehicle. Additionally, the spark timing and efficiency maps demonstrate that the vehicle is operating very near maximum brake torque at all times. The total system efficiency hovers around 30% for the majority of the high torque map and the spark timing does not significantly retard beyond 20 degrees spark advance meaning the vehicle is not presently concerned with engine knock.

This may potentially point to the possibility of enhanced performance through turbocharging without having a significant impact on efficiency as there may be a significant amount of peak pressure headroom for most of the engine operating space.

The data gathered has been used to validate a GT Power model. The GT Power model was developed to simulate a 2008 Honda Civic GX engine by a student projects team several years ago and was used as the starting point for the model. After initial investigations of the differences between the two engines, it was concluded that they were similar enough that no serious geometry modifications had to be made. The data from the 2012 Honda Civic Natural Gas proved that the cylinder head had in fact changed and needs to be re-calibrated using flow bench data for the new components.

This work is outside the scope of the present project, but would nevertheless need to be performed in order to utilize the GT Power model for further design studies on the engine if the results are expected to have meaningful relationships with the experimental platform.

In order to further investigate the engine performance it would also be necessary to acquire control of the vehicle ECU. This would enable direct control over things like spark timing, fuel injection timing, and torque converter lock-up. With these parameters controlled, it would allow the systematic removal of variables allowing a more precise identification of the different control parameter’s impact on vehicle performance.

Nevertheless, the baseline performance and control of the vehicle is now well understood. This new experimental platform can now be utilized for further design studies involving advanced natural gas vehicle technologies.

References

1. U.S. Energy Information Administration. (2014). U.S. crude oil supply & disposition summary. Retrieved from http://www.eia.gov/dnav/pet/pet_sum_crdsnd_k_a.htm

2. U.S. Energy Information Administration. (2014). U.S. product supplied. Retrieved from http://www.eia.gov/dnav/pet/pet_cons_psup_dc_nus_mbbl_a.htm

3. U.S. Energy Information Administration. (2014). Natural gas reserves summary. Retrieved from http://www.eia.gov/dnav/ng/ng_enr_sum_dcu_NUS_a.htm

4. U.S. Environmental Protection Agency. (n.d.). 40 CFR 86.1342-94 - Calculations; exhaust emissions.. Retrieved From http://www.law.cornell.edu/cfr/text/40/86.1342-94

5. Chiodi, M., Berner, H., and Bargende, M., "Investigation on different Injection Strategies in a Direct-Injected Turbocharged CNG-Engine," SAE Technical Paper 2006-01-3000, 2006, doi:10.4271/2006-01-3000.

6. Chiodi, M., Ferrari, A., Mack, O., Bargende, M., and Wichelhaus, D., “A Way Towards Remarkable Reduction of Co2 Emissions in Motorsports: The CNG- Engine,” SAE Technical Paper 2011-37-0006, 2011, doi: 10.4271/2011-37-0006.

7. Chiodi, M., Ferrari, A., Mack, O., Bargende, M., Wichelhaus, D., “Improvement of a High-Performance CNG-Engine based on an innovative Virtual Development Process,” SAE Technical Paper 2011-24-0140, 2011, doi: 10.4271/2011-24-0140.

8. Natural and bio Gas Vehicle Association Europe, (2013). Worldwide ngv shares in total vehicle market (June 2013). Retrieved from The GVR website: http://www.ngvaeurope.eu/worldwide-ngv-statistics

9. Mercedes-Benz introducing B200 Natural Gas Drive. (2012, September 6). Greencarcongress. Retrieved from: http://www.greencarcongress.com/2012/09/b200-20120906.html#more.

10. Cars. (n.d.). NGVA Europe. Retrieved , from http://www.ngvaeurope.eu/cars

11. Hillstrom, D., Chiara F., and Durand, J., “CNG Literature Review.” Center for Automotive Research, The Ohio State University.

12. Baratta, M., Catania, A., Spessa, E., et al., “Multi-Dimensional Modeling of Direct Natural-Gas Injection and Mixture Formation in a Stratified-Charge SI Engine with Centrally Mounted Injector,” SAE Technical Paper 2008-01-0975, 2008, doi: 10.4271/2008-01-0975.

13. Andreassi, L., Facci, A., Ubertini, S., “Multidimensional Modelling of Gaseous Injection in Modern Direct Injection Internal Combustion Engines: Analisys of Different Fuel Injection Strategies,” SAE Technical Paper 2009-24-0030, 2009, doi: 10.4271/2009-24-0030.

14. Douailler, B., Ravet, F., Delpech, V., et al., “Direct Injection of CNG on High Compression Ratio Spark Ignition Engine: Numerical and Experimental Investigation,” SAE Technical Paper 2011-01-0923, 2011, doi: 10.4271/2011-01- 0923.

15. Hu, C., Hou, S., “Investigations on Combustion Process of Low-pressure CNG Compound Direct Injection Spark-ignited Engines,” SAE Technical Paper 2010- 32-0052, 2010, doi: 10.4271/2010-32-0052.

16. Kalam, M., Masjuki, H., Mahlia, T., et al., “Experimental Test of a New Compressed Natural Gas Engine with Direct Injection,” SAE Technical Paper 2009-01-1967, 2009, doi: 10.4271/2009-01-1967.

17. Yadollahi, B., Boroomand, M., “The Effect of Piston Head Geometry on Natural Gas Direct Injection and Mixture Formation in a SI Engine with Centrally Mounted Single-Hole Injector,” SAE Technical Paper 2011-01-2448, 2011, doi: 10.4271/2011-01-2448.

18. Liu, Y., Yeom, J. K., & Chung, S. S. “An experimental study on the effects of impingement-walls on the spray and combustion characteristics of SIDI CNG.” Journal of Mechanical Science and Technology 26, no. 11 (2012): 2239-2246.

19. Sasaki, H., Sekiyama, S., Hashimoto, M., Nakashima, K., “Low-Emission Combustion of a Pre-Chamber-Type Compression Ignition Natural Gas Engine,” International Journal of Engine Research 8, no. 6 (2007): 465-476.

20. Liu, Y., Dong, Y., Yeom, J. K., & Chung, S. S. “An experimental investigation of the engine operating limit and combustion characteristics of the RI-CNG engine.” Journal of Mechanical Science and Technology 26, no. 11 (2012): 3673-3679.

21. Kono, S., Koga, H., Watanabe, S., “Research on Extended Expansion General- Purpose Engine-Efficiency Enhancement by Natural Gas Operation,” SAE Technical Paper 2010-32-0007 (2010), doi: 10.4271/2010-32-0007.

22. Zheng, J., Hu, E., Huang, Z., Ning, D., Wang, J., "Combustion and Emission Characteristics of a Spray Guided Direct-injection Spark-ignition Engine Fueled with Natural Gas-hydrogen Blends."International Journal of Hydrogen Energy 36, no. 17 (2011): 11155-1163.

23. Saanum, I., Bysveen, M., Tunestal, P., Johansson, B., “Lean Burn Versus Stoichiometric Operation with EGR and 3-Way Catalyst of an Engine Fueled with Natural Gas and Hydrogen Enriched Natural Gas.” SAE Technical Paper 2007- 01-0015 (2007), doi: 10.4271/2007-01-0015.

24. Kalam, M., Masjuki, H. “An Experimental Investigation of High Performance Natural Gas Engine With Direct Injection,” Energy 36, (2011): 3563-3571.

25. Abianeh, O., Mirsalim, M., Ommi, F., “Combustion Development of a Bi-fuel Engine,” International Journal of Automotive Technology 10, no. 1 (2009): 17-25.

26. Delpech, V., Obiols, J., Soleri, D., et al., “Towards an Innovative Combination of Natural Gas and Liquid Fuel Injection in Spark Ignition Engines,” SAE Technical Paper 2010-01-1513 (2010), doi: 10.4271/2010-01-1513.

27. Pipitone, E., Beccari, S., “Performances Improvement of a S.I. CNG Bi-fuel Engine by Means of Double-fuel Injection,” SAE Technical Paper 2009-24-0058 (2009), doi: 10.4271/2009-24-0058.

28. Obiols, J., Soleri, D., Dioc, N., Moreau, M., “Potential of Concomitant Injection of CNG and Gasoline on a 1.6L Gasoline Direct Injection Turbocharged Engine,” SAE Technical Paper 2011-01-1995 (2011), doi: 10.4271/2011-01-1995.

29. Ghorbanian, J., Ahmadi, M., “Experimental Thermal Analysis of Cylinder Block and Head of a Bi-fuel Turbocharged Engine,” Meccanica 47, (2012): 1987-2004.

30. Papagiannakis, R.G., Kotsiopoulos, P.N., Zannis, T.C., Yfantis, E.A., et al., “Theoretical Study of the Effects of Engine Parameters on Performance and Emissions of a Pilot Ignited Natural Gas Diesel Engine,” Energy 35, (2010): 1129-1138.

31. Lim, O., Iida, N., Cho, G., Narankhuu, J., “The Research About Engine Optimization and Emission Characteristic of Dual Fuel Engine Fueled with Natural Gas and Diesel,” SAE Technical Paper 2012-32-0008 (2012), doi: 10.4271/2012-32-0008.

32. Qi, Y., Srinivasan, K., Krishnan, S., Yang, H., Midkiff, K., “Effect of Hot Exhaust Gas Recirculation on the Performance and Emissions of an Advanced Injection Low Pilot-ignited Natural Gas Engine,” Inetrnational Journal of Engine Research 8, no. 3 (2007): 289. 72

33. Wannatong, K., Akarapanyavit, N., Siengsanorh, S., Chanchaona, S., “Combustion and Knock Characteristics of Natural Gas Diesel Dual Fuel Engine,” SAE Technical Paper 2007-01-2047 (2007), doi: 10.4271/2007-01-2047.

34. Ohm, I., Park, C., “Effect of Fuel Stratification on Initial Flame Development: Part 3 – High Swirl Condition,” International Journal of Automotive Technology 12, no. 5 (2011): 653-660.

35. Feist, M., Landau, M., Harte, E., “The Effect of Fuel Composition on Performance and Emissions of a Variety of Natural Gas Engines,” SAE Technical Paper 2010-01-1476 (2010), doi: 10.4271/2010-01-2047.

36. Bozza, F., Fontanesi, S., Gimelli, A., Severi, E., Siano, D., “Numerical and Experimental Investigation of Fuel Effects on Knock Occurrence and Combustion Noise in a 2-Stroke Engine,” SAE Technical Paper 2012-01-0827 (2012), doi: 10.4271/2012-01-0827.

37. Sachdeva, A., Mansuri, S., “Reduction of Pulsation Noise in CNG Vehicles,” SAE Technical Paper 2013-26-0098 (2013), doi: 10.4271/2013-26-0098.

38. Rink, M., Eigenberger, G., Nieken, U., Tuttlies, U., “Optimization of a Heat- integrated Exhaust Catalyst for CNG Engines,” Catalysis Today 188, (2012): 113- 120.

39. Zhang, C., Xie, Y., Wang, F., Ma, Z., Qi, D., Qiu, Z., “Emission Comparison of Light-duty In-use Flexible-fuel Vehicles Fuelled with Gasoline and Compressed Natural Gas Based on the ECE 15 Driving Cycle,” Journal of Automobile Engineering 225, (2011): 90.

40. Gamma Technologies. "GT Power Engine Simulation Software." . https://www.gtisoft.com/applications/a_Engine_Performance.php (accessed July 1, 2014).

41. Heywood, John B.. Internal combustion engine fundamentals. New York: McGraw-Hill, 1988.

42. Amand, Lars-Erik. "The Theory Behind FTIR analysis." Lecture, from Department of Energy Conversion Chalmers University of Technology, Göteborg, Sweden.

43. Stas, M., “An Universally Applicable Thermodynamic Method for T.D.C. Determination,” SAE Technical Paper 2000-01-0561 (2000), doi: 10.4271/2000- 01-0561.

44. "40 CFR 600.113-93 - Fuel Economy Calculations." LII / Legal Information Institute.

45. Rink, M., Eigenberger, G., Nieken, U., Tuttlies, U., “Optimization of a Heat- Integrated Exhaust Catalyst for CNG Engines,” Catalysis Today 188, (2012): 113- 120.

46. Mukherjee, T. (2014), “One-Dimensional Air System Modeling of Advanced Technology Compressed Natural Gas Engine,” MS Thesis, The Ohio State University.

Appendix A: Instrumentation

Temperature Sensors:  All: K-type thermocouples from Temprel Inc. Pressure Sensors:  Intake Pressure: Omega PX209 Series 030A5V o 0-30 PSI absolute pressure sensor  Exhaust Pressure: Kistler 4045A5v200s Piezo-resistive pressure transducer o Water cooled 0-5 bar absolute pressure sensor  In-cylinder pressure: AVL GH13Z-24 Piezo-electric pressure transducer Wideband O2 Sensor:  Bosch Wideband O2 Sensor : 0 256 007 151 Intake Mass Air Flow  Laminar Flow Element: Meriam Instruments Model: 50MH10-5 o Flow: 270.33 CFM at 8 in H2O @ 70 F and 29.92” Hg. Abs. Emissions:  MKS Instruments FTIR: Model #: 2030D-28229 Encoder Disc:  Custom manufactured disc with 180 teeth, and one extra deep groove for 1/rev Encoder Photo-interruptors:  TT Electronics Photologic Slotted Optical Switch “Wide Gap” Series  TT Electronics Photologic Slotted Optical Switch OPB916 Series

Appendix B: Labview Code Overview

Efficiency was paramount when designing the labview data acquisition code. One of the drawbacks of the system was the limitation of the high speed data acquisition card’s recording rate of 250 kHz. This 250 kHz was the translation rate of the one analog to digital converter on the card. The allocation of only one ADC meant that all signals coming in would be multiplexed, so the recording rate of all signals added together must be less than 250 kHz. Three signals were being recorded so the recording rate of each signal was set at 80 kHz which totals to 240 kHz when multiplexed, which is just beneath the maximum capacity of the card. Multiplexing was not a concern for data accuracy because the three signals were: in-cylinder pressure, exhaust pressure, and spark voltage.

The phase shift that might occur due to the sampling of one signal after another was not a major concern as there was only one pressure signal. The accuracy of the exhaust pressure and spark voltage was acceptable to be phase shifted by up to ~1/3 of a degree as they were not imperative for placing top dead center, or for other calculations that required the crank angle position to be very finely resolved.

The Labview program functions by breaking up the experimental setup into data acquisition tasks. For this program, the tasks were as follows: high speed data analog input task, 1/rev encoder pulse counter task, two degree encoder pulse counter task, low speed analog input task (thermocouples), and low speed USB analog input task. In order

76 to synchronize the high speed task and the counter input tasks, these three channels were start triggered by the passing of a 1/rev pulse. The low speed tasks were synchronized utilizing Labview’s sequence blocks such that they did not start until right after the high speed tasks were triggered, pending the CPU getting around to it.

The Labview code was kept rather simple to lower CPU usage. After all of the signals were triggered, the high speed tasks would cycle through a while loop for five engine cycles buffering the data in binary form on the computer RAM. After five engine cycles, the buffered data would be dumped to the hard drive through a TDMS save file.

The TDMS streaming ability with Labview is a very efficient method of storing data at high speed as it minimizes the manipulations that must be made before the data is written.

The low speed data is buffered throughout the experiment and dumped to the hard drive after the experiment had completed. No calculations were performed on the data in the

Labview setup in order to improve CPU efficiency.

Some caveats associated with the system were the allocation of direct memory access (DMA) channels from the National Instruments PCI cards in order to not over-run the CPU. The high speed signals had DMA, while the low speed signals (intake pressure,

LFE differential pressure, and wideband O2 sensor) were funneled through a USB DAQ using USB polling at 10 Hz and the low speed thermocouple signals were funneled through a PCI card with CPU polling at 10 Hz. With the system set up in this manner, the

CPU usage never went over 40-50%, allowing the experiments to be performed as intended.