Robustness study and reliability growth based on exploratory design of experiments and statistical analysis: a case study using a train door test bench

This paper presents a methodology to improve the reliability/robustness of a complex mechatronic system for passenger access, which is greatly stressed over its life cycle. The proposed methodology aims to weed out design problems during the development phase and to test the robustness of the passenger access system under operating conditions as close to the real service conditions as possible. The empirical study process is based on an experimental design series in order to cover a complex experimental domain in a reasonable number of tests for each experimental design. The exploratory aspect that stands in the experimental stage also stands in the data analysis stage thanks to a descriptive multivariate approach that reduces the initial information as less as possible. An application concerning the passenger access system for a train is presented. The experimental process is mainly based on the D-optimal design and is used with a 1:1 scale test bench. An example with five input factors is presented; the output data consist in a multidimensional signal containing open/close cycles that may be perturbed. The statistical analysis process starts by considering a scale windowing while beeping both time and multivariate aspects. The frequency values are then investigated using multiple correspondence analysis. Advantages and disadvantages of the proposed design of experiments and data analysis are discussed.