nce and Modelling
Statistics and the Materials Chain
Christine Müller, TU Dortmund, Dortmund, Germany
Several areas of statistics are connected to the Materials Chain’s scope from atoms to technical components. One large area is the improvement of product quality. There are many statistical methods for finding factors that influence the quality of a product and methods for predicting the quality for special factor combinations, even if the number of factors is high. A special challenge is predicting the reliability and lifetime of products since classical methods cannot be used and investigations under realistic conditions require too much time. Usually, accelerated lifetime tests or degradation models are used. Our research focusses on the development of good designs and new statistical methods of lifetime experiments as in [1,2]. The combination of accelerated lifetime tests and degradation models is also researched and an example is presented in . Future challenges will be the development of statistical methods and experimental designs for a chain of product components since the existing methods are mainly concerned with one component only.
Another area of statistical applications is the control of the production process. Within the production process, the quality of a product does not only depend on the known influential factors which have been optimized before but also on some unknown factors such as aging of machines, service personal etc. Here, the challenge is to develop methods for a chain of product components.
A third area that connects statistics and materials is acceptance testing. The classical approach provides different rules for the producer and the consumer. However, in a production chain, other methods must be developed since the consumer is also the producer for the next step.
 Müller, Ch.H.: D-optimal designs for lifetime experiments with exponential distribution and censoring. In: mODa 10 - Advances in Model-Oriented Design and Analysis. Eds. D. Ucinski, A.C. Atkinson and M. Patan, Physica-Verlag, Heidelberg, 179-186, (2013).
 Müller, Ch.H., Szugat, S., Celik, N. and Clarke, B.R.: Influence functions of trimmed likelihood estimators for lifetime experiments. Statistics DOI: 10.1080/02331888.2015.1104313, (2015).
 Szugat, S., Schnurr, A., Maurer, R. and Müller, Ch.H.: Simulation free prediction intervals for a state dependent failure process using accellerated lifetime experiments. In preparation.