Abstract:

Due to the phenomenon of engine misfire in single cylinder and dual cylinders often occured in vehicles, PNN(Probabilistic Neural Network) was used to detect the engine misfire by analyzing the engine rotary velocity and crankshaft angular displacement. Under the circumstance of AMEsim software, we construct a kind of four cylinders in line engine and use the injecting approach to simulate engine misfire. Then, extract the engine velocity and crankshaft angular displacement, put these data in the Matlab software to process and classify them. A PNN for training and testing is established. The experiment results indicate that the engine rotary velocity and crankshaft displacement reflect the real operation condition of the engine, the tested PNN can diagnose and probe single cylinder or double cylinders misfire in the engine. This technique has series of advantages of simplicity, economy, efficiency and high accuracy.

Key words: AMEsim engine model, fault injection, misfire diagnosis, probabilistic neural network