Abstract
It is now around fifteen years since the design and analysis of computer experiment was initiated at the beginning of the quality revolution in the early 1980s. Since then there has been considered progress in the area which falls broadly into two bodies of work. The first consists of fitting statistical models (emulators) to simulators with regard particularly to speeding up optimisation. This continues to be successful in engineering design, for example. The second area consists of fitting statistical models, in more or less the same fashion, to computer code but more to help validate the code. Examples of where these have been used are environmental modelling and nuclear safety.
Both streams need to be used in the industrial setting given the rapid rise in the use of simulation models in design, manufacture, logistics, scheduling and so on. The principles stay the same: we need methods to speed up the understanding of simulator behaviour and to help validate the simulators against the reality they are supposed to model. This paper presents an overview of the use of experimental design and modelling, in particular the validation and optimisation of emulators, and their use in the engineering environment.