Southampton researchers are inventing new methods for the statistical design of experiments, helping industries to understand and optimise complex processes in an efficient and effective way, with the ultimate aim of increasing their performance and reliability.
Experimentation in labs or pilot plants is essential for the development of new products or systems, but is becoming increasingly costly and time-consuming. Working closely with scientists and engineers, our researchers are overcoming these barriers through scientifically designed experiments and careful selection of statistical models to describe the data obtained.
"These experiments are reducing costs and time-to-market in the pharmaceutical, automotive, food and chemical industries and have wide applications in other areas," says Dave Woods, Reader in Statistics and Engineering and Physical Sciences Research Council Fellow. "In addition, the companies we work with are able to gain insight into their processes and a better understanding of the reliability of their conclusions."
For example, there can be many factors that contribute to the process of manufacturing a new drug or the performance of a car engine. However, statistically designed experiments can reveal that only a very small proportion of these factors actually have a real impact. "By detecting the key factors, and how they affect performance, you can set their values to improve the system, for example, by choosing features of a new car engine to optimise starting capability," explains Sue Lewis, Professor of Statistics.
This work is being undertaken by the Design of Experiments research group, which is part of Southampton Statistical Sciences Research Institute (S3RI) and Mathematics. "Our team is one of the largest and most active groups of its kind in the world, with a list of industry partners including GlaxoSmithKline, Jaguar and Unilever," says Steve Gilmour, Professor of Statistics.