Valve fault diagnosis in internal combustion engines by acoustic emission method using artificial neural network

The purpose of this paper is the valve failure detection due to different faults by the acoustic emission method. Three fault types (the clearance، the semi-crack and the notch) were investigated. These faults would be examined on different exhaust and intake valves. Tests were done on a cylinder head of a four-cylinder spark ignited engine. The test procedure in brief was that the bottom of the cylinder head was closed and the pressurized air was interred through the spark plug location into the combustion chamber. Produced signals due to the leakage and the damage in valves were recorded by four sensors. Measured results from experiments showed that the acoustic emission could be an effective method to detect the healthy and faulty valves in both time and frequency domains. An artificial neural network was used in an expert system for the valve fault diagnosis. This technique could recognize different faults properly; even for the same fault type in various valves، has a high accuracy.