Use of Neural Networks for Modelling and Fault Detection for the Intake Manifold of a SI Engine
A Jaguar Car engine is used to provide data for modelling the throttle body, engine pumping and manifold body. Based on the gas law of the intake dynamics, input/output variables are identified and used to train a neural network. Various structures are co
- PDF / 911,707 Bytes
- 6 Pages / 595.276 x 793.687 pts Page_size
- 30 Downloads / 162 Views
Use of Neural Networks for Modelling and Fault Detection for the Intake Manifold of a SI Engine Jocelyn A. F. Vinsonneau 1, Derek N. Shields 1 , Paul King 2, Keith J. Burnham 1 lControl Theory and Applications Centre, School of Mathematical and Information Sciences, Coventry University, Coventry CVl 5FB. U.K. Tel: +442476888972 Email: j. [email protected]. uk 2Jaguar Cars Ltd, Coventry, U.K.
Abstract A Jaguar Car engine is used to provide data for modelling the throttle body, engine pumping and manifold body. Based on the gas law of the intake dynamics, input/output variables are identified and used to train a neural network. Various structures are compared and assessed. The best structure is then used for fault detection. A neural network observer is developed and error stability is assessed. Two fault scenarios are considered.
1 Introduction On-board fault diagnosis in most modem cars is introduced to aid control and diagnose engine malfunctions. More recently, it has been used to meet increasingly stringent emissions standards set by the Environmental Protection Agency (EPA), in the context of a low emission vehicle (LEV) programme. The first LEV standards were adopted by the California's Air Resources Board (CARB) and operate from 1994 through to 2003. LEV II regulations, running from 2004 until 20lO, require on-board diagnostics to be more sophisticated, detecting emission problems relating to any sensor or component of the engine [1]. A mean of detecting those malfunctions is to make use of fault detection algorithm, to diagnostic the system (ref. [2], [3] and [4]). Over the great variety of methods used in literature based on mathematical models of the monitored system, the most common is the modelbased approach, where residuals are generated [5]. In the framework of Fault Diagnosis and Isolation (FDI), fault are detected by setting threshold on each residual signal. FDI was recently introduced in most modem cars to control and diagnose engine malfunctions (ref. [6], [7] and [8]). Also, neural networks have been proven to be useful for nonlinear system modelling in the automotive field (ref. [9], [lO] and [11]). This paper is organized as follows: A model description will be given. In section 3, neural network modelling for the intake manifold will be developed. Section 4 will propose a fault detection observer using neural network. Section 5 is an application section.
D. W. Pearson et al. (eds.), Artificial Neural Nets and Genetic Algorithms © Springer-Verlag/Wien, 2003
2 Model Description The purpose of the modelling phase is to predict the mass airflow and the manifold pressure through the throttle and the intake manifold of a spark ignition (SI) engine, respectively. This is required to be done for various atmospheric pressures and air temperatures, using mathematical expressions. The airflow rate through the intake manifold is controlled by a throttle plate. Fundamental work (ref [12] and [13]) give equations to describe the throttle and intake manifold dynamics, which form part of a complete e
Data Loading...
