Abstract: Army materiel system analysis activity (AMSAA) model reliability growth prediction modeling method based on conditional distribution is proposed for the prediction of next failure time during reliability growth testing. According to Weibull process theory and reliability growth prediction theory, marginal conditional probability density function of the future failure time is established through obtaining joint probability density function of the failure time and joint conditional probability density function of the future failure time in the cases of failure censor and time censor, and then predictor and prediction intervals of the future failure time are given. By introducing the coefficient of variance, prediction accuracy of the future failure time is analyzed. Feasibility and effectiveness of this method are verified through a certain type of wind turbine pitch system reliability growth test data in the filed trial operation, not only overcame the limitations of traditional prediction method, but also improved the prediction accuracy. The research results is beneficial to dynamic analyzing system reliability growth trend, science adjusting reliability growth test time, reasonable determining the equipment delivery time, and developing maintenance strategies.

Key words: Conditional distribution, Prediction accuracy, Reliability growth prediction, Variance coefficient, Wind turbine