Residual generation and statistical pattern recognition for engine misfire diagnostics

Methods for diagnosing misfire in internal combustion engines are presented in this paper. Crank-angle domain digital filters are used to extract features from the measured engine speed signal that are characteristic of misfire. Features for intermittent and continuous misfires are developed separately, since the engine speed responses for intermittent and continuous misfires are distinctly different. Also, the influence of crankshaft torsional vibration and repeatable measurement errors must be addressed differently in each case. The outputs from the digital filters serve as inputs to a pattern recognition network based on linear parametric classifiers. Experimental results from implementation on a Ford 4.6L V-8 engine are provided.