An improved method of reliability assessment for suspended tests

This paper reports on the development of a new method for suspended test reliability assessment and modeling. Reliability suspended tests are performed in order to accelerate testing duration and then to obtain information quickly on the life distribution of products. During suspended tests, products are randomly removed from initial samples before failure. The proposed method, which is an alternative to Johnson’s approximation, corrects the failure order number of failed products each time a product is suspended. The method is applied within a global approach carried out using the ADONIS software. The software developed for this purpose helps users to assess and model reliability for different test types (complete, censored, truncated, suspended, etc.). The proposed method is compared with the cumulated chance rate method and Johnson’s method. The three methods are then evaluated with reference to complete tests. Several simulations are carried out to validate this evaluation. The three main distributions (Weibull, exponential, normal) are used for reliability modeling. Reported simulations show that the proposed method provides better results than the two other methods in 80 per cent of the tests performed.