An evolutionary self-learning methodology: Some preliminary results from a case study

Abstract

The self-learning methodology is a four-step technique; (1): data from a process is collected, (2): the collected data is used to infer a model of the process, (3): this model is then intelligently interrogated by the computer in order to ‘discover’ a process improvement opportunity, (4): as new data become available from the process Step 2 of the methodology may be repeated in order to build a process model which is more realistic.

This self-learning methodology is almost solely driven by information extracted from process data. Hence the requirement for process domain expertise is minimal and the self-learning methodology is consequently relatively generic.

This paper provides a description of the self-learning methodology. A four-fold improvement in the tensile tolerance of a popular steel product is shown to be possible in laboratory experiments.