The invention discloses a fault diagnosis method for an unmanned aerial vehicle
gyroscope. The fault diagnosis method comprises the steps that the proximity of a sample and an expected value is calculated on the basis of a fuzzification technique, a sample membership degree factor is designed according to the proximity probability and configured into a
mathematical model representing a
support vector machine, and the influences generated by the sample distribution unbalance and the
noise characteristic are eliminated;
support vector machine training is conducted in an off-line mode,
decision function parameters of an optimal interval classification face for distinguishing a
positive sample and a
negative sample are acquired, when an unmanned aerial vehicle flies, attitude signals output by an airborne sensor
gyroscope are collected in real time, feature vectors of the signals are extracted by adopting a db4
wavelet basis
decomposition method, normalized and then input to a
decision function of the optimal interval classification face, a category
label of a to-be-decomposed
signal is calculated in real time, and the fault condition of the
gyroscope is recognized. According to the diagnosis method, the biasing degree of a fault classification hyperplane is significantly decreased, the fault diagnosis precision is improved, the calculation amount is low, and the on-line real-time performance of a fault diagnosis
algorithm is achieved.