Reliability Growth Prediction Method Based on GA-Elman Neural Network

Abstract

Reliability growth techniques is an effective means to track and predict reliability growth by planning growth paths in advance, and it can achieve quantitative improvement in the reliability of a product over a period of time. To evaluate such process, some reliability growth models which have extensively utilizations, such as the Duane model, the AMSAA model and other models, are proposed by researchers. However, some of these models still have some limitations, such as limited application scope, complicated model parameters calculation, and delayed tracking process. Applying these models to reliability growth may affect the prediction accuracy and tracking efficiency. In this paper, a novel reliability growth model is proposed to model the reliability growth and tracking process to solve the foregoing problems. First, GA-Elman neural network is chosen for short-term prediction of reliability growth. Second, based on this short-term prediction method, the reliability growth prediction and tracking model are established to achieve real-time of reliability growth. Finally, the proposed predictive model is verified by using simulated data and real engine reliability growth data from U.S.S. Grampus Diesel. The results illustrate that the proposed method is more accurate and effective than traditional models.