Frequency selection with oscillatory neurons for engine misfire detection

Detecting novelties over time series data is of practical interest in many signal processing applications. Especially engine misfire detection is one of the great issues in automobile systems to inform incomplete engine exhaustion to cause environmental problem and also to guarantee safe operation of vehicles. It requires continuous monitoring of the system in real-time to detect deviations from the normal signal patterns. This paper presents a special frequency selection method based on recurrent neural networks consisting of oscillatory neurons, and applies the method with genetic algorithm to engine misfire detection problem by observing engine speed in automobile system.