Diagnostic monitoring of internal combustion engines by use of independent component analysis and neural networks

An on-line diagnostic system to detect the deterioration in a critical state of an internal combustion engine has been developed by means of a multiplayer perceptron neural network and independent component analysis. The approach requires minimal operator input with regard to the acquisition of stationary data, selection of variables, model validation and the statistical interpretation of multivariate observations. The diagnostic method is demonstrated on a compression ignition engine under test conditions.