Statistical Methods Applied to Engine Testing

Abstract

In presenting the application of some statistical methods to problems in the field of research and development testing of internal-combustion engines, this paper attempts to stimulate interest in the use of modern statistical techniques in this important area of engineering practice. Owing to the wide range of possible applications the paper is selective and, besides saying little or nothing about many worth-while statistical methods, it also has to pass over the statistical approach to the control of quality and the inspection of the finished product.

After an introduction in which the need for the statistical approach is outlined, the paper considers an engineer faced with the problem of making an estimation from a few test observations. The outlying observation is dealt with followed by the desirability and value of ‘Confidence Limits’, ‘Tolerance Limits’ and the use of statistical inequalities. The third section moves one step further and considers two sets of observations. This deals with relationships, the scatter chart, and differences which must be tested for statistical significance: some attention is paid to the risks associated with drawing inferences from small samples of observations. The final section generalizes the position to many groups of data leading directly to the statistical technique of analysis of variance and the principles of statistical design of experiments. Some examples are given suitable for laboratory and bench tests as well as controlled trials and service tests. A brief glossary of statistical terms is given together with references to guide further reading.