Products Reliability Prediction Model Based on Bayesian Approach

Predicting reliability of new products at their early life time is one of the important issues in the field of reliability. Lack of data in this period of life time causes prediction to be very hard and inaccurate. This paper proposes a model for predicting non repairable product's reliability early after its production and introduction to the market. It is assumed that time to failure of this product has a Weibull distribution with known shape parameter but the scale parameter is a random variable that could have different distributions like gamma, inverted gamma and truncated normal. Bayesian statistics is used to join prior information on past product failure and sparse few field data on current product's performance to overcome lack of data problem which is a major problem in the early reliability prediction of new products. The Bayesian model provides a more accurate and logical prediction compared to classical methods and indications are favorable regarding the model's practicality in industrial applications. This model has managerial usefulness because of giving more accurate predictions. In all previous studies, there is no comprehensive and precise model for reliability prediction. Different from other studies, we present a definite form for scale parameter of different prior distributions. We use a special form of Weibull distribution which leads us to this definite form. This model provides a suitable estimation value from uncertain environments of parameters because it uses more information for prediction.