A nonlinear filtering technique is applied to a four-cylinder diesel engine for diagnosis of power faults inside the cylinders. The technique utilizes cepstral analysis to extract the robust transfer function information and then to predict the cylinder pressure distribution from the measured engine vibration signal. The measured vibration signal is A/D converted, using the trigger pulses from an encoder mounted at the crank shaft end, and then analyzed.
Although the inverse filtering based on cepstral analysis reduces the accuracy of the signal recovery, it can prevent the occurrence of large spurious signal components which could overwhelm the time domain structure of the estimated cylinder pressure distribution. Experimental results show that the technique developed can be successfully applied to the identification of cylinder pressure distribution from the measured engine vibration signal with fair accuracy.