Towards On-Line Prediction of the In-Cylinder Pressure in Diesel Engines from Engine Vibration Using Artificial Neural Networks

This study aims at building efficient and robust artificial neural networks (ANN) able to reconstruct the in-cylinder pressure of Diesel engines and to identify engine conditions starting from the signal of a low-cost accelerometer placed on the engine block. The accelerometer is a perfect non-intrusive replacement for expensive probes and is prospectively suitable for production vehicles. In this view, the artificial neural network is meant to be efficient in terms of response time, i.e. fast enough for on-line use. In addition, robustness is sought in order to provide flexibility in terms of operation parameters. Here we consider a feed-forward neural network based on radial basis functions (RBF) for signal reconstruction, and a feed-forward multi-layer perceptron network with tan-sigmoid transfer function for signal classification. The networks are trained using measurements from a three-cylinder real engine for various operating conditions. The RBF neural network is trained with time series from in-cylinder pressure signals and vibration signals measured on a cylinder which is distant from the one in which the pressure signal is measured. The accuracy of the predicted pressure signals is analyzed in terms of mean square error and in terms of a number of pressure-derived parameters. The location of the accelerometer has little influence on the accuracy of the reconstruction. This is confirmed also by the fact that the perceptron network, constructed in the second part of the work, is able to distinguish, from the accelerometer signal, among motored and fired conditions for any of the cylinders. Here, training data are again composed of time series obtained from the accelerometer, plus the corresponding target classes (fired/non-fired). Despite of the noisy character of the vibration signal and the distance from the cylinders, the perceptron network classifies correctly almost 100% of the signals.