Carlo Siviero

Modelling the volumetric efficiency of ic engines: Parametric, non-parametric and neural techniques

Abstract The volumetric efficiency (Th.) represents a measure of the effectiveness of an air pumping system, and is one of the most commonly used parameters in the characterization and control of four-stroke internal combustion engines. Physical models of r),. require the knowledge of some quantities usually not available in normal operating conditions. Hence, a purely black-box approach is often used to determine the dependence of 11,. upon the main engine variables, like the crankshaft speed and the intake manifold pressure. Various black-box approaches for the estimation of ην are reviewed, from parametric (polynomial-type) models, to non-parametric and neural techniques, like additive models, radial basis function neural networks and multi-layer perceptrons. The benefits and limitations of these approaches are examined and compared. The problem considered here can be viewed as a realistic benchmark for different estimation techniques.

Residual Gas Fraction Estimation: Application to a GDI Engine with Variable Valve Timing and EGR

The paper presents an original review and extension of existing mathematical models for on-line residual gas fraction estimation. The resulting model has first of all been extended to take into account also the presence of externally recirculated exhaust gas (external EGR), and then critically analyzed to highlight the importance of a correct Intake Valve Opening and Exhaust Valve Closing effective position identification. As shown in the paper, such quantities may be evaluated by using experimental data, either acquired in the test-cell or on a valve flow bench. The main objective is to obtain a simple and reliable model (that could be run in real time within the engine control unit) also in presence of Variable Valve Timing (VVT, both on intake and exhaust valves) and external Exhaust Gas Recirculation (EGR) systems. In fact, the two main contributions to residual gas fraction (backflow of the burned gas during the valve overlap period, and amount of gas trapped within the cylinder) are strongly affected by intake and exhaust valves timing, and EGR flow should be taken into account in order to determine the total exhaust gas mass within the cylinder at IVC. Therefore, real time estimation of residual gas mass and composition is crucial for designing VVT and EGR management strategies that allow an optimal control of the combustion process. The new model has been applied to experimental data acquired on a 3.2 liter V6 GDI engine, equipped with intake and exhaust Variable Valve Timing systems. Tests were performed throughout the engine operating range for different combinations of intake and exhaust valve timings, while varying EGR flow. Model results are in good agreement with other measured quantities (such as Spark Advance angle and NO x emissions), and the proposed approach therefore represents a powerful tool for on-board optimal combustion control.

A Phenomenological Combustion Model for Common Rail Multi-Jet Diesel Engine

The simplest way to describe the combustion process into the cylinder of an internal combustion engine and the associated heat release is to estimate at each crankshaft angular position the mass fraction of fuel burned using a proper function. There is a number of functions recorded in the literature that have been used for this purpose, the most relevant being likely the so-called Wiebe function. These functions have been developed both for spark ignition and diesel engines. The development of modern Common Rail injection systems makes the application of this kind of methodology particularly challenging. The trend seems to indicate, in fact, that in the near future Diesel engine injection systems will perform up to five injections per engine cycle. Therefore the way energy is released into the cylinder could become very complex to be described and the simple approaches developed up to now could be not sufficient anymore. This paper deals with the development of a single zone combustion model able to correctly describe the heat release rate for a common rail multi-jet diesel engine employing up to 4 injections per engine cycle. The model has been developed step-by-step from the simplest case of a single injection to the more complex one with 4 injections. It has been identified and validated using experimental data obtained employing from 1 to 4 different injections. Premixed and diffusive combustions have been taken into account, both modelled as “Wiebe functions”. Particular identification problems (such as modelling error with multiple injection or identification robustness procedure) are approached on the basis of real data. The main result is that increasing the number of injections actuated (and then the combustion phases) predictive properties of the model are still acceptable, and identification procedure is robust if initial values of unknown parameters are properly set. The obtained results allowed observing for example the way the combustion delays (i.e the time delays between each Start of Injection and the corresponding Start of Combustion) are modified as the number of injections increases, as well as other important combustion characteristics.Copyright

Modeling, Analysis and Simulation of a Gasoline Direct Injection System

Abstract In this paper a physical dynamic model of a Gasoline Direct Injection (GDI) Common Rail system is developed and validated with experimental data. The model is used to study different system configurations. Specifically, the placement of the pressure sensor is investigated in tenns of the observability of the pressure wave inside the rail, and the influence of the geometry of the pipe connecting the high pressure pump and the rail on the pressure variations is examined. The analysis is performed by resorting both to a simplified lumped linearized model and to the theory of distributed parameter systems applied to the original mass and momentum equations.

Residual Gas Fraction Estimation for Model-Based Variable Valve Timing and Spark Advance Control

As it is well known, the combustion process in Spark Ignition (SI) engines is strongly affected by the quality and quantity of the fluid within the cylinder at Intake Valve Closing (IVC). Residual gas affects the engine combustion processes (and therefore emissions and performance) through its influence on charge mass, temperature and dilution. Moreover, in Gasoline Direct Injection (GDI) engines, the amount of oxygen in the residual gas may be significant if the engine is operated in stratified charge mode (low loads and speeds), while almost no oxygen may be found in the residual gas during homogeneous-charge operation. In this paper, different approaches to residual gas fraction estimation are analyzed and compared. The main objective is to obtain a simple and reliable model also in presence of Variable Valve Timing (VVT, both on intake and exhaust valves) and External Gas Recirculation (EGR) systems, that could be used to control combustion duration and position. In fact, the two main contributions to residual gas fraction (backflow of the burned gas during the valve overlap period, and amount of gas trapped within the cylinder) are strongly affected by intake and exhaust valves timing, and EGR flow should be taken into account in order to determine the total exhaust gas mass within the cylinder at IVC. Therefore, estimation of residual gas mass and composition is crucial for designing VVT and EGR management strategies, integrated with optimal control of Spark Advance (and therefore of the combustion process). Experimental data have been acquired on a 3.2 liter V6 GDI engine, equipped with intake and exhaust VVT systems. Tests were performed throughout the engine operating range for different combinations of intake and exhaust valve timings, while varying EGR flow.Copyright

Estimation of Engine Maps: A Regularized Basis-Function Networks Approach

In this brief, a new methodology for the identification of engine maps from static data is presented. In order to enhance the flexibility of the model and exploit prior knowledge on the boundary conditions of the maps, a basis function neural network with a large number of neurons is used. To ensure smoothness of the estimated map as well as guarantee reliable extrapolation properties, the weights are estimated via a regularization strategy. Dynamic data are used to validate the new methodology. For this purpose, the estimated maps are included in a mean value model whose simulated manifold pressure and crankshaft speed are compared with the experimental ones. The results show a clear improvement with respect to the performances obtained resorting to standard radial basis function networks.

Combustion Development Investigation in a Common Rail Multi-Jet Diesel Engine With Up to 4 Injections per Engine Cycle

Newly developed technologies for modern diesel engines allow designing injection patterns with many degrees of freedom. Multi-jet engines, for example, can perform up to 5 injections within the same engine cycle: Position and duration of each injection, together with rail pressure and EGR rate can be chosen in order to properly design the desired in-cylinder combustion process. This means that during the injection system setup process all the free parameters have to be set to the desired value. If all the injection parameters variations have to be investigated in order to properly set their values, a huge amount of experimental tests should be needed. From this point of view, in order to reduce the need for test bench experimental work, the development of a combustion model can be very useful, to help determining the best injection configuration, and therefore the desired combustion into the cylinder. Single zone combustion models seem to be suitable for this task, thanks to the quick response they can give, and the possibility of using them for control purposes. In the paper a model developed for injection patterns with up to 4 injections is used in order to describe the combustion behavior as a function of the injection parameters. A properly designed set of tests has been performed in order to identify the combustion model. The obtained results give information on the way the combustion parameters, for example the combustion delays (i.e. the time delays between each Start Of Injection SOI, and the corresponding Start Of Combustion SOC), or the amount of fuel burnt for each injection are modified as the combustion process proceeds into the cylinder or as the injection parameters change. The information obtained can be in the following used in order to design the desired injection pattern, using the identified model as a virtual experimental tests generator.Copyright

Confidence Level Analysis for On-Board Estimation of SI Engine Catalytic Converter Efficiency

Abstract For the on-board diagnosis of the catalytic converter for SI engine, the comparison between upstream and downstream catalyst lambda signals is used to derive information about the actual conversion efficiency. In the present work three indices have been computed by means of both deterministic and statistical approaches, allowing to perform confidence analysis concerning the estimation of catalyst efficiency. The proposed approaches have been applied to process experimental data measured during New European Driving Cycle (ECE+EUDC) and to estimate the probability of fault occurrence during intermediate aged catalyst status. Furthermore, the three indices have been compared with respect to the risk levels associated with both undetected catalyst faults and false alarms.

Nonparametric Estimation of Engine Maps Using regularized Basis-Function Networks

Abstract In this paper a new methodology for the identification of engine maps from static data is presented. In order to reduce the bias error, a basis function neural network with a large number of neurons is used. Then, to avoid overtraining the weights are estimated via a regularization strategy. Dynamic data are used to validate the new methodology. For this purpose, the estimated maps are included in a Mean Value Model whose simulated manifold pressure and crankshaft speed are compared with the experimental ones. The results improve on a previous approach based on interpolating Radial Basis Function Networks.

Misfiring Detection in 8 Cylinders Turbo Engine

Abstract The paper describes a Heuristic Misfire detection strategy, called MEDOC, based on flywheel speed analysis. Tests carried out by means of numerical simulation show both control algorithm robustness and “easy tuning”. MEDOC low CPU load request, low memory occupation and good results in the actual case study prove that its application in automotive industry can be cost effective and market competitive.