Search Results

With the current trend towards engine downsizing, the use of turbochargers to obtain extra engine power has become common. A great difficulty in the use of turbochargers is in the modelling of the compressor map. In general this is done by inserting the compressor map directly into the engine ECU (Engine Control Unit) as a table. This method uses a great deal of memory space and often requires on-line interpolation and thus a large amount of CPU time. In this paper a more compact, accurate and rapid method of dealing with the compressor modelling problem is presented. This method is physically based and is applicable to all turbochargers with radial compressors for either Spark Ignition (SI) or diesel engines.

Open-source software is growing in popularity and becoming a real alternative to proprietary software. This is not only happening in home and office applications, but in embedded, industrial and engineering applications as well. This paper shows how open-source software can be used to develop and test engine models, both in simulations and in laboratory tests with real engines. It begins by explaining what open-source software is, its advantages and disadvantages compared to proprietary software, and what relevant software for engine modelling is already available. The paper describes the experience of the Engine Control Group (ECG) at the Technical University of Denmark (DTU), using a full open-source Linux solution including some software developed by the group itself. Finally, the paper concludes with the pros and cons of using this approach, suggesting tools and features to be developed in the future.

Because there are no production-type sensors which are able to measure the flow directly at the intake port, it is becoming common practice to use models of varying complexity to infer the port air mass flow from other measurements. Given the tight requirements of modern air/fuel ratio (AFR) control strategies, the accuracy of these models needs to be better than ever, during steady-state of course (though λ feedback strategies are by design very robust), but mainly during transient operation. This paper describes why conventional models might be inaccurate during engine transients.

With the tightening of exhaust emission standards, wide bandwidth control of the air/fuel ratio (AFR) of spark ignition engines has attracted increased interest recently. Unfortunately, time delays associated with engine operation (mainly injection delays and transport delays from intake to exhaust) impose serious limitations to the achievable control bandwidth. With a proper choice of sensors and actuators, these limitations can be minimized provided the port air mass flow can be accurately predicted ahead in time. While the main objective of this work is to propose a complete AFR controller, the main focus is on the problems associated with port air mass flow prediction.

Many modern control strategies for engine control are based on event based sampling. Operating the control strategy in the event domain makes it possible to obtain samples at specific crank shaft angles in the engine cycle, which is often desirable for certain control strategies. One of the biggest disadvantages involved with event based strategies is signal aliasing at low engine speeds or a high computational burden at higher engine speeds. This paper presents an easy solution to the aliasing problem above. If the data between the event based samples is stored using a time based strategy, it is shown here that a subsequent treatment of the sampled data as a time series together with a suitable low pass filter structure can avoid aliasing.

Mean Value Engine Models (MVEMs) are simplified, dynamic engine models which are physically based. Such models are useful for control studies, for engine control system analysis and for model based engine control systems. Very few published MVEMs have included the effects of Exhaust Gas Recirculation (EGR). The purpose of this paper is to present a modified MVEM which includes EGR in a physical way. It has been tested using newly developed, very fast manifold pressure, manifold temperature, port and EGR mass flow sensors. Reasonable agreement has been obtained on an experiemental engine, mounted on a dynamometer.

Long term stoichiometric Air/Fuel Ratio (AFR) control of an SI engine is at the present mainly maintained by table mapping of the engine's fresh air intake as a function of the engine operating point. In order to reduce a stationary error in the AFR to zero the table based control normally works in conjunction with a PI feedback from a HEGO sensor. The effective bandwidth of this feedback loop is quite small and seldom exceeds 2 Hz. This is altogether too small for accurate transient AFR control. This paper presents a new λ (normalized Air/Fuel Ratio) control methodology (H∞ control) which has a somewhat larger bandwidth and can guarantee robustness with respect to selected engine variable and parameter variations.

An important paradigm for the modelling of naturally aspirated (NA) spark ignition (SI) engines for control purposes is the Mean Value Engine Model (MVEM). Such models have a time resolution which is just sufficient to capture the main details of the dynamic performance of NA SI engines but not the cycle-by-cycle behavior. In principle such models are also physically based, are very compact in a mathematical sense but nevertheless can have reasonable prediction accuracy. Presently no MVEMs have been constructed for intercooled turbocharged SI engines because their complexity confounds the simple physical understanding and description of such engines. This paper presents a newly constructed MVEM for a turbocharged SI engine which contains the details of the compressor and turbine characteristics in a compact way. The model has been tested against the responses of an experimental engine and has reasonable accuracy for realistic operating scenarios.

Long term control of the AFR (Air/Fuel Ratio) of spark ignition engines is currently accomplished with a selvoscillating PI control loop. Because of the intake/exhaust time delay, the oscillation frequency and hence bandwidth of this loop is small. This paper describes a new approach to the design of this control loop using a novel observer system. In this way the bandwidth of this important loop is increased by a factor of 2 - 6 times, leading to more accurate overall AFR control. Moreover the observer approach is so robust and allows such feedback levels that it reduces significantly the accuracy required in the calibration of the base fuel control system with which it is be used. It can be used with either conventional- or advanced observer based- base fuel strategies.

In general most engine models for control applications have been constructed using regressions fitting and measured engine data. Such techniques have also been used to model the dynamic performance of engines. Unfortunately regression equation models are very complex and do not show directly the physical reality from which they emerge. This has for example made it impossible to write down explicitly the dymanic equations for, for example, the air exchange process in an SI engine in any form other than as the manifold pressure state equation. In recent a publication a Mean Value Engine Model (MVEM) has been constructed for an SI engine which is physically based and which has a simple physical form which can be immediately understood and manipulated.

Mean Value Engine Models (MVEMs) are dynamic models which describe dynamic engine variable (or state) responses as mean rather than instantaneous values on time scales slightly longer than an engine event. Such engine variables are the independent variables in nonlinear differential (or state) equations which can be quite compact but nevertheless quite accurate. One of the most important of the differential equations for a spark ignition (SI) engine is the intake manifold filling (often manifold pressure) state equation. This equation is commonly used to estimate the air mass flow to an SI engine during fast throttle angle transients to insure proper engine fueling. The purpose of this paper is to derive a modified manifold pressure state equation which is simpler and more physical than those currently found in the literature. This new formulation makes it easier to calibrate a MVEM for different engines and provides new insights into dynamic SI engine operation.

In an earlier paper, some of the authors of this paper pointed out some of the difficulties involved in event based engine control. In particular it was shown that event based (or constant crank angle) sampling is very difficult to carry out without running into aliasing and sensor signal averaging problems. This leads to errors in reading the air mass flow related sensors and hence inaccurate air/fuel ratio control. The purpose of this paper is first to demonstrate that the conjectures about the operator input spectrum in a vehicle do actually obtain during vehicle operation in realistic road situations. A second purpose is to extend earlier modelling work and to present an approximate physical method of predicting the level of engine pumping fluctuations at any given operating point. The physical method given is based on a modification of the Mean Value Engine Model (MVEM) of a Spark Ignition (SI) engine presented previously.

Many existing production engine controllers use event (or constant crank angle increment) based sampling and computation systems. Because the engine events are synchronized to the internal physical processes of an engine, it is widely accepted that this is the most logical approach to engine control. It is the purpose of this paper to deal with this assumption in detail and to illuminate various failures of it in practical systems. The approach of the paper is in terms of overall general control system design. That is to say that the problem of event based engine control is considered as a general control problem with its standard components: 1. modelling (engine plus actuator/sensor), 2. specification of desired performance goals, 3. control system design method selection and 4. experimental testing.

One of the most important operating modes for SI engines is in the idle speed region. This is because SI engines spend a large part of their time operating in this mode. Moreover, a large measure of operator satisfaction is dependent on an engine operating smoothly and reliably in and around idle. In particular the operator expects that the idle speed will remain constant in spite of the engine loads due to power steering pumps and air conditioning compressors. In the idle speed region an SI engine is thought to be quite nonlinear because the engine loading can be quite significant, thus forcing the engine to be driven through a reasonably large portion of its lower operating range. Many of the earlier studies of idle speed control systems have dealt with linearized models which in principle have limited validity for the problem at hand. In order to improve this situation, it is necessary to deal with the more general nonlinear control problem.

In earlier work it has been shown that a nearly ideal solution to the problem of accurate estimation of the air mass flow to a central fuel injection (CFI) (or throttle body (TBI)) or EFI (or multi-point (MPI)) equipped engine is provided by using a closed loop nonlinear observer for the engine. With proper design this observer was shown to be both accurate and robust with respect to modelling end measurement errors. It is based on a Constant Gain Extended Kalman Filter (CGEKF). Since the publication of this work, another type of observer has emerged in the literature for which claims of great robustness have been made. This observer is based on new developments in the area of nonlinear control theory and is called a Sliding Mode Observer (SMO). In this paper these two types of observers are compared theoretically and experimentally on an engine mounted on a dynamometer. A very aggressive driving scenario is assumed for these tests.

A very important component of an accurate steady state and transient air/fuel (A/F) ratio control strategy is the transient fuel compensation (TFC) substrategy. This is the part of an engine control algorithm which cancels the fuel film dynamics and makes it possible to place injected fuel into the intake manifold (or close to the intake ports or valves) of a spark ignition (SI) engine at the correct time and location. This paper presents the results of a very large series of experiments conducted with the same engine with either a throttle body (TBI) (or central fuel injection (CFI)) manifold or with a multi-point port injection (MPI) (or electronic fuel injection (EFI)) manifold. These experiments have shown that in some practical applications it may be necessary to model the intake manifold as a two time constant dynamic system rather than as a single differential equation system.

In an earlier paper one of the authors of this paper (E. Hendricks and co-authors) treated the question of obtaining correct steady state and transient control of the air/fuel (A/F) ratio of an SI engine. This study was based in part on simulations conducted with a dynamic engine model developed earlier and in part on experimental results. The main conclusions were that conventional control strategies (Speed-Throttle, Speed-Density and Mass Air Flow (MAF)) cannot give proper A/F control because of 1. sensor and anti-aliasing filter time constants and 2. improper or lacking compensation for manifold fuel film and (air) filling dynamics. In this paper, the results of a long series of experiments conducted with the control systems above are to be presented. Both central fuel injection (CFI) (or throttle body (TBI)) and electronic fuel injection (EFI) (or multipoint (MPI)) manifolds have been investigated.

An important element in nearly all engine control systems is the lambda control feedback system and its associated switching exhaust gas oxygen sensor (EGO). This feedback loop is necessary to keep the mean value of the normalized air/fuel ratio close to one. This is a necessary condition for proper operation of the three-way catalyst systems which are a part of nearly all production emissions control systems. Currently many systems are based on using classical proportional-integral (PI) controllers in lambda control feedback loops which are self-oscillating. Proper design of such systems is dependent on knowing the time delay between the injection time and the time when a corresponding signal appears at the engine exhaust EGO sensor. Recently a new method of designing the vital larnbda control loop has emerged which is claimed to be very robust with respect to the injection/exhaust time delay.

Mean value engine models (MVEMs) seek to predict dynamically the mean values of important SI engine variables such as the crank shaft speed, the manifold pressure and the theoretical air/fuel ratio (lambda). Previous work also shows that such models can be made quite accurate, both for stationary and transient operating modes. Because these models can be made mathematically simple and compact, they are also tractable for direct mathematical and physical analysis. In this paper an analysis of a mean value engine model is carried out which reveals the underlying structure of the problems which face engine control system designers. In particular it is shown that an SI engine is extremely nonlinear and time dependent. Because of this, conventional control strategies using conventional sensors cannot be made to operate correctly in the transient regime. An “ideal” nonlinear compensator is also described for the fueling dynamics which works over a wide operating range.

Conventional electronic engine control systems suffer from poor transient air/fuel ratio control accuracy. This is true of speed-throttle, speed-density, and mass air flow (MAF) control systems with either single point (or central) or port fuel injection. The reason for this is that they fail to 1. compensate for the nonlinear dynamics of the fuel film in the intake manifold or in the vicinity of the intake valves. 2. estimate correctly the air mass flow at the location of the injector(s). This paper presents a nonlinear fuel film compensation network and a nonlinear closed loop observer. The nonlinear fuel film compensator gives improved global cancellation of the fuel film dynamics, while the closed loop observer has improved robustness with respect to modelling error and measurement noise. The closed loop observer is based on a modified constant gain extended Kalman filter.