Low Volatility Fuel Delivery Control during Cold Engine Starts
2005-01-0639 SAE TECHNICAL PAPER SERIES
Low Volatility Fuel Delivery Control during Cold Engine Starts Gerard W. Malaczynski, David B. Miller and Steven L. Melby Delphi Corporation
Reprinted From: Combustion and Flow Diagnostics, and Fundamental Advances in Thermal and Fluid Sciences (SP-1971)
2005 SAE World Congress Detroit, Michigan April 11-14, 200 400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org
The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE’s peer review process under the supervision of the session organizer. This process requires a minimum of three (3) reviews by industry experts. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. For permission and licensing requests contact: SAE Permissions 400 Commonwealth Drive Warrendale, PA 15096-0001-USA Email: [email protected] Tel: 724-772-4028 Fax: 724-772-4891
For multiple print copies contact: SAE Customer Service Tel: 877-606-7323 (inside USA and Canada) Tel: 724-776-4970 (outside USA) Fax: 724-776-1615 Email: [email protected] ISSN 0148-7191 Copyright © 2005 SAE International Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in SAE Transactions. Persons wishing to submit papers to be considered for presentation or publication by SAE should send the manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE. Printed in USA
2005-01-0639
Low Volatility Fuel Delivery Control during Cold Engine Starts Gerard W. Malaczynski, David B. Miller and Steven L. Melby Delphi Corporation
Copyright © 2005 SAE International
• • •
ABSTRACT The intensity of a combustion flame ionization current signal (ionsense) can be used to monitor and control combustion in individual cylinders during a cold engine start. The rapid detection of poor or absence of combustion can be used to determine fuel delivery corrections that may prevent engine stalls. With the ionsense cold start control active, no start failures were recorded even when the initially (prior to ionsense correction) commanded fueling had failed to produce a combustible mixture. This new dimension in fuel control allows for leaner cold start calibrations that would still be robust against the possible use of low volatility gasoline. Consequently, when California Phase 2 fuel is used, cold start hydrocarbon emissions could be lowered without the risk of an engine stall if the appropriate fuel is replaced with a less volatile one.
• •
All the above listed methods except RPM-based control algorithm either require engine/exhaust structural changes or the introduction of new sensors. Proposed in this paper are fuel delivery corrections based on combustion progress at the time of a cold engine start. This utilizes an ionsense virtual combustion sensor which is currently used for misfire and knock detection. Consequently, what is described here as a new application falls into the same category as RPM-based controller since it would require the development of a new software algorithm only. In addition, the authors are not aware of successful application of a RPM-based algorithm for engine starts with high DI fuel at temperatures well below the freezing point.
A cold start fuel calibration that produced lower hydrocarbon (HC) emissions but was too lean for reliable engine starts using low volatility (high driveability index) fuel was tested with high driveability index fuel (DI = 1266) down to temperatures as low as – 12 oC. Successful engine starts even at these low temperatures were achieved by activating the ionsense cold start control algorithm.
EMISSIONS AT COLD START If a cold start calibration does not have to protect against the possibility of a vehicle being fueled with a low volatility (high DI) fuel, emission levels would be substantially improved without any changes in the engine/exhaust system architecture. This is illustrated in Figure 1 that depicts the emissions results for two different fuel delivery schedules at cold engine start.
INTRODUCTION Today, increasing pressure to reduce emissions, is driving OEM’s to make cold starts as lean as possible, but this creates potential driveability problems when using less volatile fuel. Therefore, several methods to compensate for high DI fueling are known to be under investigation: •
Secondary Air Injection
Exhaust Gas Temperature Sensor Fuel DI Sensor Wide-range A/F sensor with Lost Fuel Adaptation Model Hydrocarbon Adsorber RPM Fluctuation Detection
A production-worthy calibration (performed on a high DI fuel and tested on Cal Ph. II fuel), secures reliable cold engine starts regardless of the type of fuel. However, if a more volatile fuel such as Cal. Ph. II is used, the fuel delivery represented by this 1
calibration becomes unnecessarily rich. As can be seen in Figure 1, when using the richer, production worthy calibration, the measured engine exhaust lambda was less than 0.9 while cranking the engine. On the other hand, the leaner calibration represented by the fuel delivery schedule calibrated and tested on Cal Ph. II fuel (1170 DI) would deliver a more optimized air-to-fuel ratio (lambda not less than 0.95 and cumulative HC better than 20% less in the first 50 seconds after crank) when a more volatile, low driveability index fuel is used, but would frequently fail to start smoothly when started on low volatility fuel (high DI). In short, unless there is a correction mechanism, the “leaner” calibration cannot be used without a risk of poor performance, since the choice of gasoline used in the vehicle cannot be controlled.
IONSENSE FUNDAMENTALS In the event of a cold start, fuel volatility could be monitored directly by a “fuel driveability index” sensor, by sampling and evaluating relevant fuel properties, or indirectly by monitoring the progress of the combustion process. The latest is possible by sensing the combustion with the flame ionization signal. An ionization signal, or as it is frequently called, ionsense signal, represents the development of a flame’s conductivity [1, 2] over the combustion cycle which, in turn, correlates with combustion kinetics [3, 4]. In particular, poor combustion results in weak ionsense signals, and in the limit, the total absence of the signal. An arbitrarily defined combustion quality (CQ) can, therefore, be represented by an integral of the ionsense signal [5] over a selected window, measured in crank angle degrees (CAD), during which the combustion is expected. In the application described here [6], CQ is considered to be a key parameter for signal processing. It is the CQ value which is used for measuring combustion, and in particular, for monitoring and control of engine performance during cold starts, with a variety of fuels representing a full range of available volatilities. The method presented in this paper would be preferred to any arrangement that would include a driveability index (DI) sensor that samples fuel properties. Good cold starts are sometimes experienced with high DI fuels, while poor starts are possible even if low DI fuel is used. This happens because the combustion is governed by the air-tovapor phase fuel-ratio. Fuel delivered as liquid cannot participate in combustion. It must first be converted to vapor. As long as the intake system is cold, the quantity of vapor phase fuel cannot be precisely estimated. It may depend, for example, on the amount of residual fuel still present at the time of the previous shutdown. A combustion monitoring method, on the other hand, directly addresses engine performance. The proposed implementation of combustion quality monitoring allows adoption of a “one-type fits all” cold start air-to-fuel ratio calibration. It also addresses the issue of excessive emissions at cold starts that result from excessively rich fuel mixtures used to prevent engine stalls, misfires, and other type of cold-start induced malfunctions. In addition, it allows the monitoring
Figure 1. An example of hydrocarbon fuel emissions and exhaust air-to-fuel ratio for a 4-cylinder engine with production calibration (calibrated on a high DI fuel and tested on Cal Ph. II fuel), and a leaner calibration (calibrated and tested on Cal Ph. II fuel), which would not secure quality engine starts with high DI fuel. 2
and balance of the performance of individual cylinders. In other words, it may be utilized in correcting either acquired or inherited imbalance, and therefore, enhance emission control not only during cold starts but possibly also on a warmed up engine. EXPERIMENTAL SET-UP The cold start experiments reported in this paper were conducted on two different production vehicles from two different OEM’s. The first vehicle was powered by a 2.4L 4-cylinder engine, and the second by a 2.0 L 4 cylinder engine. Both vehicles used identical ignition coil design in which the ignition coil is integrated with circuitry for ion current biasing and measurement [7, 8]. Some of the ionsense control functions, which would normally be handled by the engine controller, were run externally, using a dSPACE/Simulink software-in-the-loop system for the sake of experimental flexibility (see Figure 2). Thus, the ioncurrent signals were delivered to the input of the fast A/D converter of the dSPACE interfacing system for further processing in Simulink. The intent of this hardware configuration was to emulate the system mechanization depicted in Figure 3. Subsequently, the ioncurrent signal integration and cylinder event synchronization were also emulated in Simulink. The Simulink processing segment ultimately outputs a calculated combustion quality index (CQ). The calculated combustion quality is then delivered to the engine controller for further processing. CQ value readings below a certain threshold level set a misfire flag, which activates the controller’s algorithm (see Figure 4).
Figure 2. Block diagram of the experimental set-up. The ioncurrent from individual cylinders are fed to DS2003 dSPACE board. Digitized ioncurrent is integrated over a time period representing individual combustion events and handled for further processing in the engine controller. If a misfire is sensed, the fuel delivery correction is calculated by the engine controller, and the modified fuel injector pulse width signal sent to the relevant fuel injector.
In addition, the Simulink/dSPACE setup was used to record relevant engine parameters, synchronization markers, and ionsense signals from individual cylinders. This enables post-test analysis of the data. The data recording block and its triggering line (once per combustion event) are removed from the schematic depicted in Figure 2, for clarity, since they do not contribute to understanding the system’s mechanization.
DESCRIPTION OF THE COLD START FUEL DELIVERY CORRECTION ALGORITHM The ionsense cold start feedback control algorithm utilizes a time integral of the raw ionsense current signal along with a variety of engine control parameters to rate combustion quality (CQ). This CQ value is then used to provide individual cylinder fueling corrections to prevent engine sag and stalling.
3
Upon conversion to engine run mode, the altered fuel pulse width can be reset to its initial value, and the control algorithm changed to monitor poor combustion performance of individual cylinders on a less reactive basis.
In the work reported in this paper, the on-board microprocessor utilizes a leaner than normal cold start calibration, sufficient to produce near flawless cold starts with California Phase II fuel, but producing inferior performance when a significantly lower volatility fuel is used. We have also selected an engine calibration which delivers sequential spark from the start. However, only slight modifications to the algorithm are necessary to adapt it to a waste spark system.
The fuel delivery correction is achieved by modifying the initially comanded fuel injector pulse width by multiplying it by a correction factor. The algorithm which calculates the multiplier considers the current engine state, and other parameters like the engine speed, load, etc., as indicated in the schematics presented in Figures 3 and 4. During the crank portion of the intervention algorithm, the
Spark Plug
B+ DSP
A/D Converter Bias Ion Current Signal
Igniter
Ion Current Buffer and Driver
Synchro
Ignition Coil
Buffered Ion Signal
Input Network
Electronic Spark Timing
Controller
Figure 4. Generic diagram of the fuel delivery correction algorithm based on low readings of the combustion quality CQ, which represents the intensity of the ionization during an individual combustion event.
Figure 3. System mechanization would consist of the ionsense-ready ignition coil and its communication lines with the engine controller performing the digital signal processing and its interpretation.
The combustion quality monitoring process starts as early as the first combustion event during the crank period. Based on the ongoing performance of individual cylinders, fuel pulse widths are adjusted, if necessary, to achieve the expected buildup of engine speed. Since the crank time period is very short and terminates either in a transition to the engine run state or in failure to start the engine, the pulse width modulation algorithm must be, at this phase, distinctly fast and decisive (at this stage there is not much room for a second correction).
misfire flag arriving at terminal # 1 (Fig.4) is delivered to the misfire counter through the eventsensitive switch enabled at terminal # 2. The reset signal coming from terminal # 4 is generated each time combustion occurs. Inputs to the algorithm, except the consecutive misfire count, are enginecondition dependent calibration parameters residing in the controller’s software as look-up tables and single value calibrations. The fuel correction logic, the activation switch signal (2), the correction function and its parametric entries, are all 4
first attempt, or engine stall after start. These cold engine starts were run with, and without the ionsense-based cold start controller active for a low volatility fuel. They were run using two different fuel delivery calibrations.
calibratible, to accommodate the varying needs of different engine architectures. The aggressiveness of the algorithm is different for the crank and run phases. The run mode does not require as aggressive correction as crank to prevent a stall, even when a load is suddenly applied, because the inertia of the engine allows adequate time for a stall preventing reaction at a more moderate rate.
An example of the commanded air/fuel ratio for the first 25 seconds, for the “normal” and “leaner” calibrations is depicted in Figure 5. Table I clearly indicates that with a standard, “normal” production calibration, good starts are always obtained, as might be expected. However, with a leaner calibration, which could be expected to reduce HC emissions on the FTP (Federal Test Procedure) test, cold starts on high DI fuel are usually poor. On the other hand, some good starts occur even with high DI fuel. This is usually caused by fuel remnants from the previous shutdown, which may be present depending on the operating conditions of the engine during the previous operating cycle.
The issue of fuel volatility becomes less important and eventually disappears with increasing temperature of the intake ports. Therefore, the correction algorithm, even if activated in response to initial poor combustion caused by low fuel volatility, is eventually terminated. The return path to normal fuel delivery, sufficient for the warmed-up engine, depends on a set of engine parameters. This set of parameters may include calculated intake port temperature, time elapsed from the start, current load, speed, etc. It also takes into account the potential for control system instability that might be initiated by removing a correction too soon or at too high of a rate.
Low volatility 1266 DI Fuel
To make the response to the proposed algorithm more efficient on subsequent cold starts, the previous cold start performance and required fuel pulse width modifications may be stored for use in the next cold start even before the first combustion event is evaluated, provided that re-fueling has not occurred in the meantime. This option, however, has not been explored yet.
Number of Number of poor starts acceptable starts 43 0
Ionsense-based control inactive, “normal” calibration Ionsense-based 7 control inactive, leaner fuel calibration 112 Ionsense-based control active, leaner fuel calibration
Finally, the combustion quality index, CQ, may be used to indicate that an excessive amount of fuel is being delivered to an individual cylinder at any time. This information could possibly be used to bring the individual fuel deliveries down to the desirable level. This would further assist in lowering HC emissions.
38
0
Table I. Cold starts performance for production-worthy (“normal”), and leaner fuel calibrations with, and without ionsense-based cold start control algorithm.
CONTROLLER’S PERFORMANCE
Fuel delivery corrections are, in general, different for each cylinder as the initial thermodynamic conditions may vary at each individual intake port. A key contributor is the partial fuel vapor pressure before the first injection (fuel remnants), but it might also be individual spark energy, statistical difference in fuel deliveries, firing sequence, temperature distribution, etc. In fact, in almost all cases, certain imbalance was observed, resulting in slightly different reaction of the controller for individual cylinders.
More than 200 cold starts were performed to demonstrate the functionality of the ionsense feedback cold start control algorithm. Most of the starts were performed in wintertime ambient temperatures, which were as low as –12 oC. Table I shows a compilation of “good” and “bad” starts. An “acceptable” start was defined as a start on the first crank with no RPM sag more than 200 rpm below the idle desired rpm, following the flare exit. A “poor” start is defined as a failure to start on the 5
Figure 7. Engine speed trace associated with the start represented by fuel deliveries depicted in Fig.6.
Figure 5. An example of commanded air/fuel ratio traces for “normal” and “leaner” calibrations.
Figure 8. Examples of engine speed buildup without, and with the controller for cold starts with a high DI fuel (DI=1266). EST synchronized data acquisition system: if there is no EST (stall) then there is no data recorded; markers represent readings.
Figure 6. A typical reaction of the ionsense-based fuel delivery correction controller at a cold start with high DI fuel (DI=1266). Modified pulse widths are depicted for all 4 cylinders (cylinder #1 at the top through #4 at the bottom). 6
The ionsense method offers a very significant benefit over other methods of achieving robustness against the use of high DI fuel. It possesses the ability to react to poor combustion in individual cylinders. Consequently, fuel delivery corrections are cylinder specific, which clearly optimizes the emission control system. If applied in conjunction with misfire and knock detection, the method represents the modification of software and engine calibration, thus, becomes extremely cost- effective.
A typical example of individual fuel delivery corrections is illustrated in Figure 6. In the example presented in Figure 6, the fuel delivery on cylinder number 2 and 3 tapered back to the uncorrected amount practically upon conversion from crank to run marked by the step function, while cylinder 1 and 4 required more aggressive fueling for over 3 seconds. The corresponding engine speed trace shown in Figure 7 clearly signals the danger of engine stall that was prevented due to the prompt action of the ionsense feedback controller. The controller’s action, since it is activated in the response to combustion quality readings, is reactive, not proactive. Consequently, starts in lower temperature result in longer cranking time. This trend is illustrated in Figure 8b. However, at similar start temperatures, without the ionsense feedback controller, engine stalls were recorded (Figure 8a). Note that the engine speed readings were acquired with the recording system that contains an event trigger synchronous to the electronic spark timing signal (EST). Therefore, engine speeds approaching zero are not recorded, as the EST pulses are not produced in those instances.
ACKNOWLEDGMENTS We would like to acknowledge the helpful critical review provided by Carilee Moran as well as the hardware and testing support provided by Chris Zimmerman. REFERENCES 1. J. Lawton, F.J. Weinberg, Electrical Aspects of Combustion, Clarendon Press, 1969. 2. R. M. Clements, P. R. Smy, “The Variation of Ionization with Air/Fuel Ratio for a Spark Ignition Engine,” J. Appl.Phys., Vol. 47, No.2, 1976. 3. A. Saitzkoff, R. Reinmann, T. Berglind, M. Glavmo, “An Ionization Equilibrium Analysis of the Spark Plug as an Ionization Sensor,” SAE Paper # 960377, SAE Congress, 1996.
FUTURE WORK An obvious improvement in the ionsense feedback controller’s performance would be to enable proactive action. This could be done by storing a memory of the engine performance of the previous start at engine shutdown. This would substantially shorten the crank time, as the initial fuel correction would be delivered before the first combustion event. It would, however, require a separate algorithm employing, at least, knowledge of the time between starts, engine block and fuel temperatures, and above all, information on refueling.
4. A. Saitzkoff, R. Reinmann, F. Mauss, M. Glavmo, “In-Cylinder Pressure Measurements Using the Spark Plug as an Ionization Sensor,” SAE Paper # 970857, SAE Congress 1997. 5. A. Lee, J. S. Pyko, “Engine Misfire Detection by Ionization Current Monitoring,” SAE Paper # 950003, SAE Congress 1995.
CONCLUSIONS 6. U.S. patent pending: “Engine Control Method and Apparatus Using Ion Sense Combustion Monitoring”, Delphi Corporation.
Monitoring the flame ionization current provides a feedback signal that enables a form of closed loop fuel control starting as early as the onset of the first combustion. This closed loop control enables a vehicle to be started reliably on low volatility fuels. It also allows the use of leaner cold start calibrations resulting in lower HC emissions when using higher volatility fuels such as California Phase 2 fuel.
7. “Method of identifying engine cylinder combustion sequence based on combustion quality,” US Patent # 6,520,166 B1, February 18, 2003, Delphi Corporation. 8. “Ion sense ignition bias circuit,” US Patent # 6,498,490 April 30, 2001, Delphi Corporation. 7
Low Volatility Fuel Delivery Control during Cold Engine Starts Gerard W. Malaczynski, David B. Miller and Steven L. Melby Delphi Corporation
Reprinted From: Combustion and Flow Diagnostics, and Fundamental Advances in Thermal and Fluid Sciences (SP-1971)
2005 SAE World Congress Detroit, Michigan April 11-14, 200 400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org
The Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE’s peer review process under the supervision of the session organizer. This process requires a minimum of three (3) reviews by industry experts. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of SAE. For permission and licensing requests contact: SAE Permissions 400 Commonwealth Drive Warrendale, PA 15096-0001-USA Email: [email protected] Tel: 724-772-4028 Fax: 724-772-4891
For multiple print copies contact: SAE Customer Service Tel: 877-606-7323 (inside USA and Canada) Tel: 724-776-4970 (outside USA) Fax: 724-776-1615 Email: [email protected] ISSN 0148-7191 Copyright © 2005 SAE International Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper. A process is available by which discussions will be printed with the paper if it is published in SAE Transactions. Persons wishing to submit papers to be considered for presentation or publication by SAE should send the manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE. Printed in USA
2005-01-0639
Low Volatility Fuel Delivery Control during Cold Engine Starts Gerard W. Malaczynski, David B. Miller and Steven L. Melby Delphi Corporation
Copyright © 2005 SAE International
• • •
ABSTRACT The intensity of a combustion flame ionization current signal (ionsense) can be used to monitor and control combustion in individual cylinders during a cold engine start. The rapid detection of poor or absence of combustion can be used to determine fuel delivery corrections that may prevent engine stalls. With the ionsense cold start control active, no start failures were recorded even when the initially (prior to ionsense correction) commanded fueling had failed to produce a combustible mixture. This new dimension in fuel control allows for leaner cold start calibrations that would still be robust against the possible use of low volatility gasoline. Consequently, when California Phase 2 fuel is used, cold start hydrocarbon emissions could be lowered without the risk of an engine stall if the appropriate fuel is replaced with a less volatile one.
• •
All the above listed methods except RPM-based control algorithm either require engine/exhaust structural changes or the introduction of new sensors. Proposed in this paper are fuel delivery corrections based on combustion progress at the time of a cold engine start. This utilizes an ionsense virtual combustion sensor which is currently used for misfire and knock detection. Consequently, what is described here as a new application falls into the same category as RPM-based controller since it would require the development of a new software algorithm only. In addition, the authors are not aware of successful application of a RPM-based algorithm for engine starts with high DI fuel at temperatures well below the freezing point.
A cold start fuel calibration that produced lower hydrocarbon (HC) emissions but was too lean for reliable engine starts using low volatility (high driveability index) fuel was tested with high driveability index fuel (DI = 1266) down to temperatures as low as – 12 oC. Successful engine starts even at these low temperatures were achieved by activating the ionsense cold start control algorithm.
EMISSIONS AT COLD START If a cold start calibration does not have to protect against the possibility of a vehicle being fueled with a low volatility (high DI) fuel, emission levels would be substantially improved without any changes in the engine/exhaust system architecture. This is illustrated in Figure 1 that depicts the emissions results for two different fuel delivery schedules at cold engine start.
INTRODUCTION Today, increasing pressure to reduce emissions, is driving OEM’s to make cold starts as lean as possible, but this creates potential driveability problems when using less volatile fuel. Therefore, several methods to compensate for high DI fueling are known to be under investigation: •
Secondary Air Injection
Exhaust Gas Temperature Sensor Fuel DI Sensor Wide-range A/F sensor with Lost Fuel Adaptation Model Hydrocarbon Adsorber RPM Fluctuation Detection
A production-worthy calibration (performed on a high DI fuel and tested on Cal Ph. II fuel), secures reliable cold engine starts regardless of the type of fuel. However, if a more volatile fuel such as Cal. Ph. II is used, the fuel delivery represented by this 1
calibration becomes unnecessarily rich. As can be seen in Figure 1, when using the richer, production worthy calibration, the measured engine exhaust lambda was less than 0.9 while cranking the engine. On the other hand, the leaner calibration represented by the fuel delivery schedule calibrated and tested on Cal Ph. II fuel (1170 DI) would deliver a more optimized air-to-fuel ratio (lambda not less than 0.95 and cumulative HC better than 20% less in the first 50 seconds after crank) when a more volatile, low driveability index fuel is used, but would frequently fail to start smoothly when started on low volatility fuel (high DI). In short, unless there is a correction mechanism, the “leaner” calibration cannot be used without a risk of poor performance, since the choice of gasoline used in the vehicle cannot be controlled.
IONSENSE FUNDAMENTALS In the event of a cold start, fuel volatility could be monitored directly by a “fuel driveability index” sensor, by sampling and evaluating relevant fuel properties, or indirectly by monitoring the progress of the combustion process. The latest is possible by sensing the combustion with the flame ionization signal. An ionization signal, or as it is frequently called, ionsense signal, represents the development of a flame’s conductivity [1, 2] over the combustion cycle which, in turn, correlates with combustion kinetics [3, 4]. In particular, poor combustion results in weak ionsense signals, and in the limit, the total absence of the signal. An arbitrarily defined combustion quality (CQ) can, therefore, be represented by an integral of the ionsense signal [5] over a selected window, measured in crank angle degrees (CAD), during which the combustion is expected. In the application described here [6], CQ is considered to be a key parameter for signal processing. It is the CQ value which is used for measuring combustion, and in particular, for monitoring and control of engine performance during cold starts, with a variety of fuels representing a full range of available volatilities. The method presented in this paper would be preferred to any arrangement that would include a driveability index (DI) sensor that samples fuel properties. Good cold starts are sometimes experienced with high DI fuels, while poor starts are possible even if low DI fuel is used. This happens because the combustion is governed by the air-tovapor phase fuel-ratio. Fuel delivered as liquid cannot participate in combustion. It must first be converted to vapor. As long as the intake system is cold, the quantity of vapor phase fuel cannot be precisely estimated. It may depend, for example, on the amount of residual fuel still present at the time of the previous shutdown. A combustion monitoring method, on the other hand, directly addresses engine performance. The proposed implementation of combustion quality monitoring allows adoption of a “one-type fits all” cold start air-to-fuel ratio calibration. It also addresses the issue of excessive emissions at cold starts that result from excessively rich fuel mixtures used to prevent engine stalls, misfires, and other type of cold-start induced malfunctions. In addition, it allows the monitoring
Figure 1. An example of hydrocarbon fuel emissions and exhaust air-to-fuel ratio for a 4-cylinder engine with production calibration (calibrated on a high DI fuel and tested on Cal Ph. II fuel), and a leaner calibration (calibrated and tested on Cal Ph. II fuel), which would not secure quality engine starts with high DI fuel. 2
and balance of the performance of individual cylinders. In other words, it may be utilized in correcting either acquired or inherited imbalance, and therefore, enhance emission control not only during cold starts but possibly also on a warmed up engine. EXPERIMENTAL SET-UP The cold start experiments reported in this paper were conducted on two different production vehicles from two different OEM’s. The first vehicle was powered by a 2.4L 4-cylinder engine, and the second by a 2.0 L 4 cylinder engine. Both vehicles used identical ignition coil design in which the ignition coil is integrated with circuitry for ion current biasing and measurement [7, 8]. Some of the ionsense control functions, which would normally be handled by the engine controller, were run externally, using a dSPACE/Simulink software-in-the-loop system for the sake of experimental flexibility (see Figure 2). Thus, the ioncurrent signals were delivered to the input of the fast A/D converter of the dSPACE interfacing system for further processing in Simulink. The intent of this hardware configuration was to emulate the system mechanization depicted in Figure 3. Subsequently, the ioncurrent signal integration and cylinder event synchronization were also emulated in Simulink. The Simulink processing segment ultimately outputs a calculated combustion quality index (CQ). The calculated combustion quality is then delivered to the engine controller for further processing. CQ value readings below a certain threshold level set a misfire flag, which activates the controller’s algorithm (see Figure 4).
Figure 2. Block diagram of the experimental set-up. The ioncurrent from individual cylinders are fed to DS2003 dSPACE board. Digitized ioncurrent is integrated over a time period representing individual combustion events and handled for further processing in the engine controller. If a misfire is sensed, the fuel delivery correction is calculated by the engine controller, and the modified fuel injector pulse width signal sent to the relevant fuel injector.
In addition, the Simulink/dSPACE setup was used to record relevant engine parameters, synchronization markers, and ionsense signals from individual cylinders. This enables post-test analysis of the data. The data recording block and its triggering line (once per combustion event) are removed from the schematic depicted in Figure 2, for clarity, since they do not contribute to understanding the system’s mechanization.
DESCRIPTION OF THE COLD START FUEL DELIVERY CORRECTION ALGORITHM The ionsense cold start feedback control algorithm utilizes a time integral of the raw ionsense current signal along with a variety of engine control parameters to rate combustion quality (CQ). This CQ value is then used to provide individual cylinder fueling corrections to prevent engine sag and stalling.
3
Upon conversion to engine run mode, the altered fuel pulse width can be reset to its initial value, and the control algorithm changed to monitor poor combustion performance of individual cylinders on a less reactive basis.
In the work reported in this paper, the on-board microprocessor utilizes a leaner than normal cold start calibration, sufficient to produce near flawless cold starts with California Phase II fuel, but producing inferior performance when a significantly lower volatility fuel is used. We have also selected an engine calibration which delivers sequential spark from the start. However, only slight modifications to the algorithm are necessary to adapt it to a waste spark system.
The fuel delivery correction is achieved by modifying the initially comanded fuel injector pulse width by multiplying it by a correction factor. The algorithm which calculates the multiplier considers the current engine state, and other parameters like the engine speed, load, etc., as indicated in the schematics presented in Figures 3 and 4. During the crank portion of the intervention algorithm, the
Spark Plug
B+ DSP
A/D Converter Bias Ion Current Signal
Igniter
Ion Current Buffer and Driver
Synchro
Ignition Coil
Buffered Ion Signal
Input Network
Electronic Spark Timing
Controller
Figure 4. Generic diagram of the fuel delivery correction algorithm based on low readings of the combustion quality CQ, which represents the intensity of the ionization during an individual combustion event.
Figure 3. System mechanization would consist of the ionsense-ready ignition coil and its communication lines with the engine controller performing the digital signal processing and its interpretation.
The combustion quality monitoring process starts as early as the first combustion event during the crank period. Based on the ongoing performance of individual cylinders, fuel pulse widths are adjusted, if necessary, to achieve the expected buildup of engine speed. Since the crank time period is very short and terminates either in a transition to the engine run state or in failure to start the engine, the pulse width modulation algorithm must be, at this phase, distinctly fast and decisive (at this stage there is not much room for a second correction).
misfire flag arriving at terminal # 1 (Fig.4) is delivered to the misfire counter through the eventsensitive switch enabled at terminal # 2. The reset signal coming from terminal # 4 is generated each time combustion occurs. Inputs to the algorithm, except the consecutive misfire count, are enginecondition dependent calibration parameters residing in the controller’s software as look-up tables and single value calibrations. The fuel correction logic, the activation switch signal (2), the correction function and its parametric entries, are all 4
first attempt, or engine stall after start. These cold engine starts were run with, and without the ionsense-based cold start controller active for a low volatility fuel. They were run using two different fuel delivery calibrations.
calibratible, to accommodate the varying needs of different engine architectures. The aggressiveness of the algorithm is different for the crank and run phases. The run mode does not require as aggressive correction as crank to prevent a stall, even when a load is suddenly applied, because the inertia of the engine allows adequate time for a stall preventing reaction at a more moderate rate.
An example of the commanded air/fuel ratio for the first 25 seconds, for the “normal” and “leaner” calibrations is depicted in Figure 5. Table I clearly indicates that with a standard, “normal” production calibration, good starts are always obtained, as might be expected. However, with a leaner calibration, which could be expected to reduce HC emissions on the FTP (Federal Test Procedure) test, cold starts on high DI fuel are usually poor. On the other hand, some good starts occur even with high DI fuel. This is usually caused by fuel remnants from the previous shutdown, which may be present depending on the operating conditions of the engine during the previous operating cycle.
The issue of fuel volatility becomes less important and eventually disappears with increasing temperature of the intake ports. Therefore, the correction algorithm, even if activated in response to initial poor combustion caused by low fuel volatility, is eventually terminated. The return path to normal fuel delivery, sufficient for the warmed-up engine, depends on a set of engine parameters. This set of parameters may include calculated intake port temperature, time elapsed from the start, current load, speed, etc. It also takes into account the potential for control system instability that might be initiated by removing a correction too soon or at too high of a rate.
Low volatility 1266 DI Fuel
To make the response to the proposed algorithm more efficient on subsequent cold starts, the previous cold start performance and required fuel pulse width modifications may be stored for use in the next cold start even before the first combustion event is evaluated, provided that re-fueling has not occurred in the meantime. This option, however, has not been explored yet.
Number of Number of poor starts acceptable starts 43 0
Ionsense-based control inactive, “normal” calibration Ionsense-based 7 control inactive, leaner fuel calibration 112 Ionsense-based control active, leaner fuel calibration
Finally, the combustion quality index, CQ, may be used to indicate that an excessive amount of fuel is being delivered to an individual cylinder at any time. This information could possibly be used to bring the individual fuel deliveries down to the desirable level. This would further assist in lowering HC emissions.
38
0
Table I. Cold starts performance for production-worthy (“normal”), and leaner fuel calibrations with, and without ionsense-based cold start control algorithm.
CONTROLLER’S PERFORMANCE
Fuel delivery corrections are, in general, different for each cylinder as the initial thermodynamic conditions may vary at each individual intake port. A key contributor is the partial fuel vapor pressure before the first injection (fuel remnants), but it might also be individual spark energy, statistical difference in fuel deliveries, firing sequence, temperature distribution, etc. In fact, in almost all cases, certain imbalance was observed, resulting in slightly different reaction of the controller for individual cylinders.
More than 200 cold starts were performed to demonstrate the functionality of the ionsense feedback cold start control algorithm. Most of the starts were performed in wintertime ambient temperatures, which were as low as –12 oC. Table I shows a compilation of “good” and “bad” starts. An “acceptable” start was defined as a start on the first crank with no RPM sag more than 200 rpm below the idle desired rpm, following the flare exit. A “poor” start is defined as a failure to start on the 5
Figure 7. Engine speed trace associated with the start represented by fuel deliveries depicted in Fig.6.
Figure 5. An example of commanded air/fuel ratio traces for “normal” and “leaner” calibrations.
Figure 8. Examples of engine speed buildup without, and with the controller for cold starts with a high DI fuel (DI=1266). EST synchronized data acquisition system: if there is no EST (stall) then there is no data recorded; markers represent readings.
Figure 6. A typical reaction of the ionsense-based fuel delivery correction controller at a cold start with high DI fuel (DI=1266). Modified pulse widths are depicted for all 4 cylinders (cylinder #1 at the top through #4 at the bottom). 6
The ionsense method offers a very significant benefit over other methods of achieving robustness against the use of high DI fuel. It possesses the ability to react to poor combustion in individual cylinders. Consequently, fuel delivery corrections are cylinder specific, which clearly optimizes the emission control system. If applied in conjunction with misfire and knock detection, the method represents the modification of software and engine calibration, thus, becomes extremely cost- effective.
A typical example of individual fuel delivery corrections is illustrated in Figure 6. In the example presented in Figure 6, the fuel delivery on cylinder number 2 and 3 tapered back to the uncorrected amount practically upon conversion from crank to run marked by the step function, while cylinder 1 and 4 required more aggressive fueling for over 3 seconds. The corresponding engine speed trace shown in Figure 7 clearly signals the danger of engine stall that was prevented due to the prompt action of the ionsense feedback controller. The controller’s action, since it is activated in the response to combustion quality readings, is reactive, not proactive. Consequently, starts in lower temperature result in longer cranking time. This trend is illustrated in Figure 8b. However, at similar start temperatures, without the ionsense feedback controller, engine stalls were recorded (Figure 8a). Note that the engine speed readings were acquired with the recording system that contains an event trigger synchronous to the electronic spark timing signal (EST). Therefore, engine speeds approaching zero are not recorded, as the EST pulses are not produced in those instances.
ACKNOWLEDGMENTS We would like to acknowledge the helpful critical review provided by Carilee Moran as well as the hardware and testing support provided by Chris Zimmerman. REFERENCES 1. J. Lawton, F.J. Weinberg, Electrical Aspects of Combustion, Clarendon Press, 1969. 2. R. M. Clements, P. R. Smy, “The Variation of Ionization with Air/Fuel Ratio for a Spark Ignition Engine,” J. Appl.Phys., Vol. 47, No.2, 1976. 3. A. Saitzkoff, R. Reinmann, T. Berglind, M. Glavmo, “An Ionization Equilibrium Analysis of the Spark Plug as an Ionization Sensor,” SAE Paper # 960377, SAE Congress, 1996.
FUTURE WORK An obvious improvement in the ionsense feedback controller’s performance would be to enable proactive action. This could be done by storing a memory of the engine performance of the previous start at engine shutdown. This would substantially shorten the crank time, as the initial fuel correction would be delivered before the first combustion event. It would, however, require a separate algorithm employing, at least, knowledge of the time between starts, engine block and fuel temperatures, and above all, information on refueling.
4. A. Saitzkoff, R. Reinmann, F. Mauss, M. Glavmo, “In-Cylinder Pressure Measurements Using the Spark Plug as an Ionization Sensor,” SAE Paper # 970857, SAE Congress 1997. 5. A. Lee, J. S. Pyko, “Engine Misfire Detection by Ionization Current Monitoring,” SAE Paper # 950003, SAE Congress 1995.
CONCLUSIONS 6. U.S. patent pending: “Engine Control Method and Apparatus Using Ion Sense Combustion Monitoring”, Delphi Corporation.
Monitoring the flame ionization current provides a feedback signal that enables a form of closed loop fuel control starting as early as the onset of the first combustion. This closed loop control enables a vehicle to be started reliably on low volatility fuels. It also allows the use of leaner cold start calibrations resulting in lower HC emissions when using higher volatility fuels such as California Phase 2 fuel.
7. “Method of identifying engine cylinder combustion sequence based on combustion quality,” US Patent # 6,520,166 B1, February 18, 2003, Delphi Corporation. 8. “Ion sense ignition bias circuit,” US Patent # 6,498,490 April 30, 2001, Delphi Corporation. 7
