Automated diagnosis of piston slap in engines

Abstract

Excessive impact between the piston and cylinder inner wall is a common mechanical fault of IC engines. This impact is mostly caused by wear during the engine operation. Even though many researchers have studied the dynamic process of piston slap for the piston design, only a limited number of researchers used the vibration signals measured on the surface of the block to detect and identify piston slap faults. Moreover, when these vibration-based diagnostic techniques are implemented in a real situation, it usually requires an expert to interpret the analysis results from measured vibration signals. This paper proposes an Artificial Neural Network (ANN)-based automated system for piston slap diagnosis. In order to provide sufficient input data to train the networks, simulation models were built in an advanced dynamic simulation software (LMS Virtual.Lab). The simulation models were validated and updated by a series of experiments. The experimental vibration signals for both normal condition and with a range of faults were processed by a number of signal processing techniques, such as envelope analysis and the "kurtogram". The transfer characteristic of the engine block was also measured. The best features were selected from the processed experimental and simulation results as the inputs to the ANNs. The automated diagnostic system consists of three main stages: fault detection, fault localization and severity identification. The simulated data was used to train the networks, and the measurement data was used for test the networks. The final results have shown that the developed system can efficiently diagnose different piston slap faults, including location and severity.