Method and system for detecting a misfire of a reciprocating internal combustion engine

A method for detecting a misfire of a reciprocating internal combustion engine with a trainable pattern recognition system, the engine including a plurality of reciprocating components having an engine cycle frequency, at least one cylinder and a crankshaft, the method comprising the steps of:

• generating an electrical signal as a function of crankshaft angular velocity, the electrical signal containing data;
• sampling the data synchronously with crankshaft angular position at a sampling rate to obtain sampled data;
• transforming the sampled data to an equivalent frequency domain spectrum, including frequency components having complex amplitudes at the engine cycle frequency and harmonics thereof;
• and
• applying an algorithm to the frequency components, which algorithm distinguishes between a true misfire and normal cyclic variability which characterizes the combustion process in each cylinder of the engine to thereby obtain a misfire signal, the step of applying the algorithm including the step of inputting the frequency components as input vectors to the pattern recognition system to identify the misfire.