Search Results

This paper presents a systematic approach for designing an experiment in situations where expensive and time consuming computer simulations are used to evaluate product characteristics. In the presence of many design parameters, the critical step is to find the best possible experimental set up with minimum number of simulations. Usually in such situations, designers use their intuition and experience to carry out a number of simulation runs and choose the design that gives better performance. This intuitive approach can be considerably improved by using statistical methods. “Classical experimental designs” were compared with “space filling designs” in terms of their results and requirements. A typical clutch booster bracket is used as an example to demonstrate the methodology.

In this paper a procedure has been established to estimate the field life of a component using the data of accelerated life tests on old and new designs of the component, and the field failure data on the old design of the component. For this, the correlation factors have been established between the accelerated life data and the field life data of the old design and these correlation factors have been applied to the accelerated life data of the new design to estimate the life distribution of the new design in the field. The formula for the correlation factor is based on the assumption that all the life data distributions under investigation have increasing failure rates (IFR). A case study on engine timing belt has been presented as an example at the end to validate this work. Such a procedure has been found to be useful to introduce the component in the field without any field trials.