Study on Prediction Model Based on Grey Residual Information Neural Network

The purpose of the investigation is to ensure equipment dependable operation. A novel prediction model based on grey residual neural network is proposed in this paper, which is composed of grey residual model (GRM) and information neural network (INN). Firstly, it predicts key feature information by GRM which is expressed as quantitative values. In this process, the prediction values of GRM are compensated by INN1, so the final prediction values equal to the output values of GRM plus the compensated error values. Then, the final prediction values are inputted INN2 to predict fault or life. And the influence of environment is an input node for network. In the prediction model, INN1 is an error compensator and INN2 is a prediction network. The investigation shows that the model can carry out the prediction successfully, because it uses well the ability of processing the uncertain information for grey system and the function of integrating decision for INN. In this paper, the model is validated with crack data which are obtained from the aerospace launcher. The result shows that the model is tried and has relatively good prediction accuracy and flexibility.