Parameter Estimation Using EM Algorithm and Test Design Optimization of Constant Stress Accelerated Life Test with non-constant Parameters under Type-I Progressive Censoring

In this paper k- level constant stress accelerated life test under Type-I progressive censoring for Lomax distribution with non-constant shape and scale parameters is investigated. The purpose of this paper is to estimate the model parameters using the EM algorithm and optimize the test design.

Life testing often consumes a very long time for testing and this is a fundamental problem in test design. This problem is solved by accelerated life tests. There is a recommended method for reducing the time of failure, such that the stress level of the test units will increase and then they will fail earlier than normal operating conditions. Therefore, these approaches reduced the running time. In this paper, the k-level constant stress accelerated life test under progressive Type-I censoring used. The Expectation-Maximization (EM) algorithm is used to determine the maximum likelihood estimates of the unknown parameters. This algorithm is a very powerful tool in handling the incomplete data problem. Two different criteria used to optimize the test plan. And the effect of the sample size, number of stress levels and inspection and the intermediate censoring proportion are assessed on the design efficiency.

based on the simulation study and a real data set, it is demonstrated that the EM estimator is good. Also, under the optimization criterion II, a more efficient test was obtained than the optimization criterion I. In addition, the small sample size, the small number of stress levels, the small number of inspections and the large intermediate censoring proportion lead to a more efficient test.

In this paper, the periodic inspection is used to collect lifetime data. Although continuous is an ideal mode. But sometimes due to technical limitations and/or budgetary constraints, the continuous inspection is not possible in practice and the experimenter has to use the periodic inspection. In this case, the exact times of test units may not be available and only the failure counts are collected at certain time points during the test. Also, in this paper, we assumed that both scale and shape parameters to be log linear model by operating stress