Failure prediction of laser gyro based on neural network method

During the storage and using process, laser gyroscope zero bias will drift due to influence of temperature, vibration and other environmental factors. This paper uses FMMEA method to analyze the reason for the variation of the laser gyros parameters. Using learning mechanisms of BP neural network to train the model with zero bias data and establish a relationship between zero bias value and the time. According to the given zero bias threshold and the acquired neural network model, fault time can be predicted. This paper also uses radial basis network and time series analysis to establish the reasoning algorithm between zero bias and laser gyro navigation fault. The results show that, for laser gyroscope zero bias data, neural network method has higher fitting precision than time series analysis, and can achieve good reasoning model, also the prediction is more close to the real fault time.