A unique technique for the on-line detection of engine misfire has been developed, which is based on multiple feature integration through fuzzy pattern recognition. The technique requires the measurement of the instantaneous angular velocity signals. As more as 10 dimensionless features for the engine misfire detection are extracted by different ways. With the help of fuzzy pattern recognition, all the features are integrated together as a fuzzy vector, which identifies uniformly whether the engine is healthy or faulty at first, then locates the position of a misfiring cylinder or cylinders if it is necessary. The experimental results show that such a strategy is able to use reasonably the redundant and complementary information of all the features, and thus leads to better diagnostic reliability and efficiency.
|