Abstract
Diesel engines have been widely used as electrical power sources for small islands and also as mechanical power sources for ships under the isolated circumstances. Therefore, failure diagnosis is of great importance. The paper pays a great attention to the abnormal sealing of exhaust valves, because it occurs frequently and tends to grow into a serious failure.
Based on the data obtained from artificial failure tests using an actual engine, pattern recognition theory was applied to the time series of the longitudinal vibration signal measured by accelerometers on cylinder heads. The paper proposed a method to discriminate between failures in sealing of exhaust valves and in fuel injection timing. Effectiveness of the method was finally verified with experiments.
