The study of the method about detection on throttle failure using the bispectrum

The detection of non-Gaussian signal in additive Gaussian noise using bispectrum is a matter of practical interest in many areas, such as throttle failure detection and so on. An ARMA model of a nonload gasoline engine which works on little throttle valve opening is given. The detection approach of engine speed sensor and throttle sensor failure based on the model is elaborated which is employed by the electric control system in the period of shifting. When there is no failure, the residual between the series of ARMA model of throttle (nonload engine) and the measuring data series of engine speeding and throttle should be Gaussian, the mean of it is zero, otherwise it should be non-Gaussian. Then we can use the bispectrum to detect the failure of the engine speed sensor and throttle sensor.