Design and Optimization of Neural Networks To Estimate the Chamber Pressure in Internal Combustion Engines by an Indirect Method

Abstract

A particular type of artificial neural network (ANN), with the aim of estimating the indicated pressure into cylinders from instantaneous angular speed measurement, has been developed: radial basis function (RBF). This is the main component of a methodology where input and output curves are parametrized. A modified RBF network from its general structure is designed to reduce the size of the network, approaching the hidden layer weights by means of polynomial functions. It makes it possible to use it in medium requirement PCs, such as in automobile applications. Results corresponding to different operating conditions of a single-cylinder diesel engine (DE), a three-cylinder spark ignition engine (SIE), and a 16-cylinder vee power plant DE (V-16 DE) are presented. Results show the accuracy and fast response of the network, making it feasible to use in online diagnostics and control systems in engines, with very low cost (removing the use of piezoelectric sensors).

Fast on-line identification of instantaneous mechanical losses in internal combustion engines

“…A first approach could be distributing the mean torque losses T fM between cylinders. On the other hand, for flexible crankshaft and driveline considerations (even for single-cylinder engines), there are several applications where an iterating procedure must be established, such as the indirect determination of in-cylinder pressure profile from the angular speed measurement [2,14,15]. Thus, the solution can be obtained by solving Eqs.…”

“…Second, for multi-cylinder engines, overlapping between torque profiles associated to the different cylinders makes it difficult to evaluate those coefficients. Finally, pressure acts as an unknown quantity when the inverse problem is applied; that is, pressure profile is determined from angular speed measurements by an indirect method [14,15]. In these conditions, a non-linear identification procedure is required.…”

Influence of a Combustion Parametric Model on the Cyclic Angular Speed of Internal Combustion Engines. Part II: Statistical Sensitivity Assessment Results

“…For this reason, in the present work results from the previously described response surface experimental design of a DI CIE have been used. Thus, conclusions may be taken as indicative of the procedure to follow in futures studies. , …”

“…To improve the exhaust emissions and performance of internal combustion engines, the influence of the design parameters over the engine behavior must be identified . Depending on the purpose (design, control, or diagnosis), pressure cycle measurements and modeling may be required. − Among the different procedures to model the combustion process into a combustion chamber, − the Wiebe’s parametric model that provides the mass burned fraction (MBF) corresponds to one of the most internationally extended . In this sense, many studies predict the correlations between some of the previous parameters. , …”

Influence of a Combustion Parametric Model on the Cyclic Angular Speed of Internal Combustion Engines. Part I: Setup for Sensitivity Analysis

“…Main parametric models are the Wiebe model and the Logistical models, besides general recursive procedures, providing a fuel mass-burned fraction (MBF) profile that depends on the crank angle degree (CAD) . Although there are other expressions such as the models from Chmela et al and Barba et al, , the widely used Wiebe model provides a good agreement between measured and calculated in-cylinder pressure. − Different analyses predict correlations between some of these parameters. , Engines are mechanical systems where pressures are considered to be the input forces, and the response of a given torque is the instantaneous angular speed. In this sense, in many failure diagnosis applications, the pressure cyclic curve is determined from the angular speed analysis. ,, Although the pressure profile is approached accurately, results from the analysis of combustion parameters through parametric models differ from real values, with the exception of two parameters (fuel equivalence ratio and injection timing), that can be fairly estimated, thus demonstrating the relation between input and output parameters is not unique. , …”