"Scheduling tracing", a technique of knowledge elicitation for production scheduling

Complexity of production scheduling in real world applications arises due to a diversity of constraints. Actual schedules to date are usually generated by experienced human schedulers. It is thus important to obtain a feasible solution to meet a variety of scheduling environments rather than to obtain an optimal solution generated on many unrealistic assumptions. Each production schedule created by human schedulers can be regarded as a feasible solution. The focus of this study is on obtaining a scheduling procedure similar to the human scheduler's one as much as possible. This leads to an idea of knowledge elicitation by interviewing a human scheduler while he traces the scheduling process actually practiced by him. This reminds him to talk about his scheduling knowledge as explicitly as possible. This knowledge elicitation is called "scheduling tracing" in this study and is very appropriate for building a knowledge base of production scheduling. Some uncertain types of knowledge are, however, left to acquire, which the scheduler possesses implicitly in mind. Hence, the well known knowledge elicitation technique of "prototype" is combined with "scheduling tracing" to improve the knowledge base. This combination of knowledge elicitation of "scheduling tracing" with "prototype" is applied to an example of production scheduling in current industrial practice to demonstrate its effectiveness in building a knowledge base of production scheduling.