Time Series Prediction of Complex Equipment Accident Rate Based on Wavelet Packet and Elman Neural Network

Accident rate prediction is an important basis for safety analysis of complex equipment and allocating safety management resources. For better description the non-stationary characteristics of the accident rate time series of complex equipment and overcome the poor accident rate data, the wavelet packet denoising (WPD) and Elman neural network combined prediction algorithm are constructed, and the specific steps are proposed. Finally, taking the class A flight accidents rate of the US Air Force from 1962 to 2001 as an example, the combined WPD–Elman neural network has higher prediction accuracy, and the mean error sum of squares is 6.11% lower than Elman neural network alone. The prediction results of 100,000-h accident rate test data from 1998 to 2001 are 1.4481, 1.3094, 1.0582, and 1.0963, and the mean error sum of squares is 2.67%, which can better solve the problem of small-scale non-stationary time series prediction.