Research Of Simulation-based Optimal Design For Accelerated Testing

Accelerated testing (AT) technology provides a feasible solution to life prediction for important equipment and complex systems, such as satellites in orbit, missile systems in storage and so on. Different AT plans may induce completely different test efficiency, test cost and prediction accuracy. It is still a main challenge in application how to design optimal plans for AT which result in the most accurate estimation within predetermined budget. Analytical approaches to optimal design for AT have been focused on so far. However, analytical solutions of optimal design are not available in many cases, and the procedure of its inference is very complicated, which makes it difficult to set up a framework for applications. Therefore, optimal design problems of complex AT plans for equipment are difficult to be solved by analytical approaches. Aimed at the application requirements for life prediction of equipment, this dissertation carries on a systemic research on optimal design for AT from a completely new perspective. A set of theories and methods of simulation-based optimal design for AT are established by incorporating Monte Carlo simulation, which provides technical support for optimal design for AT in life prediction of equipment. The main contributions of this dissertation are summarized as follows:1. From the analysis on analytical approaches to optimal design for AT, a theoretical framework of simulation-based optimal design for AT, including simulating, analyzing and optimizing, is established by incorporating statistical simulation method of Monte Carlo. To meet the requirements of the theoretical framework, approaches of generating simulated data in typical AT are presented; statistical methods to analyze the data are summarized; basic methods for simulation-based optimal design for AT are presented. The presented theoretical framework can convert a problem of optimal design for AT into a problem of data analyzing in AT, so it is applicable in the optimal design for complex plans of AT, to which analytical approaches are difficult to apply.2. On the basis of the theory of simulation-based optimal design for AT, methods of simulation-based optimal design for typical accelerated life tests (ALT) and accelerated degradation tests (ADT) are presented.(1) Methods of simulation-based optimal design for constant-stress accelerated life testing (CSALT) and step-down-stress accelerated life testing (SDSALT) are presented respectively from analyzing objective function, plan elements and test constraints. Examples are provided to compare the presented methods of simulation-based optimal design for ALT with analytical approaches to demonstrate their validity. The methods can provide key technical support for life prediction in equipment with 0-1 failure modes.(2) A method of simulation-based optimal design for degradation test (DT) is presented from analyzing objective function, plan elements and test constraints of optimal design for DT. Based on this, a method of simulation-based optimal design for constant stress accelerated degradation test (CSADT) is presented. Examples are also provided to compare the presented methods of simulation-based optimal design for DT / ADT with analytical approaches to demonstrate their validity. The methods can provide key technical support for life prediction in equipment with degradation failure modes.The aforementioned four methods are applicable in optimal design for complicate plans of AT and convenient to use in engineering in contrast to analytical approaches. The methods advance the theory of optimal design for AT, and provide directions for optimal design for other types of AT.3. The methods of simulation-based optimal design for AT are applied to life prediction of newly-developed electron multipliers (EM) in super-high-precision satellite-borne Cesium atom frequency standard system, which provides an engineering case for the study of optimal design for AT. Firstly, prior information of EM is obtained by a pilot experiment of multi-crossed step-stress ADT. Secondly, an optimal ADT plan is designed by the methods of simulation-based optimal design for AT. Thirdly, multi-constant stress ADT is conducted based on the optimal ADT plan. And finally, life prediction model of EM is established by which life of EM can be predicted at use condition.In summary, under the fund of related key projects of armament, this dissertation investigates theories, methods and applications of simulation-based optimal design for accelerated testing systematically. The research and further improvements will provide systematic accelerated testing technology for life prediction of equipment. So the research has important theory and engineering significance in life prediction of equipment.