STATISTICS AND DESIGN OF EXPERIMENTS
Design of Experiments (DoE) provides a systematic approach both for real and virtual experimentation of campaign designs. This approach has obsoleted the more traditional approach of changing one factor at a time when making experiments
Design of experiments techniques, descriptive statistics of a data set and analysis of variance are all basic techniques used in labs to organize daily work activities. These techniques are also relevant and important for simulation engineers who want to reduce the total number of runs and consequently the number of time consuming simulations.
The Numerical Aspects of Design of Experiments
A design of experiments technique involves the selection of a set of experimental points with an optimal distribution in the experimental domain. The optimality of the distribution is influenced by the type and number of controlling factors, the analyzed outcomes, and the information the engineer would like to obtain.
When the number of controlling factors is reduced, usually to two or three input variables, it may still be possible to manage a “manual” positioning of the experiments in the experimental domain or any combination may be evaluated using a full factorial approach.
Usually, engineering problems have more than three input parameters, and placing points in the space in an appropriate manner can be very hard without the proper expertise and dedicated tools.
The Added value of Design of Experiments for the Simulation Engineer
A successful design of experiments campaign gives the engineer the maximum knowledge of the problem at hand, while reducing costs and time for the design of the experiment itself.
The advantages of using Design of Experiments include:
- elimination of redundant experiments
- reduction in time and resources needed for a campaign
- maximize the knowledge gained from experimental data