Design of validation experiments for life prediction models

This paper proposes a novel validation experiment design optimization (VEDO) method for the assurance of life prediction model, which is one of the key steps in guaranteeing the reliable design of products in meeting the target service life. Life testing data collected from experiments are important for the validation of time-dependent models. However, directly collecting life data for model validation at the operating stress level is usually time-consuming and expensive. In order to overcome this challenge, the accelerated life testing (ALT) method is employed in the proposed method to collect data for model validation. The connection between ALT and model validation is established first; then a VEDO model is developed using the prior information obtained from the computer simulation model. In the VEDO model, the information gain for model validation is maximized within the testing budget and available testing chamber constraints. The obtained optimal number of tests and testing stress levels are designed to maximize the confidence in the validation results. Various sources of uncertainty such as prediction uncertainty, uncertainty of prior information, and observation errors are included within the optimization process in order to improve the robustness of validation experiment design. A composite helicopter rotor hub component is used to demonstrate the effectiveness of the proposed VEDO method.