Application of kNN and artificial neural network classifiers for fault detection and diagnosis in a diesel engine cooling system

摘要

Diesel engines are used in a wide range of industrial applications, from power generators to mechanical engines. Detecting incipient faults is very important for safety and cost reduction in any of these applications. In this paper, the pressure and temperature of cooling water and the engine speed are used to detect faults in the water cooling system. Statistical features of these signals are used for training kNN and artificial neural networks classifiers to detect and diagnose faults. Data collected from a real engine operating in all possible speeds was used for training the classifiers using cross validation techniques. Some parameters of the classifiers were selected evaluating their effect on classifiers' performance. Data from normal operation and from two faults of very small magnitude was introduced and performance was compared using average mean square error and the classification errors in the confusion matrix. Both methods presented a high and similar performance. The online application shows that the method can be readily applied to a real engine with small detecting times.