RuleWorks

A Framework for Model-Based Adaptive Training

Scheduling Domain Analysis - Introduction

The domain of training application in this chapter is order scheduling. The training purpose is to make production control, material planners, supervisors and order administration people aware of scheduling rules. Many of these rules are actually embedded in an existing Knowledge Based System which is used to check order schedules, update material availability and update capacity commitments. Even if a successful expert system is not intentionally designed to address training needs, it can generate significant training benefits. For example, support and advisory KBS’s provide training benefits in test equipment setup, process diagnoses, manufacturing troubleshooting, distributed product repair and business modelling [van de Brug & Orciuch 1991]. The existing scheduling system has not been used in any form to provide training. However, its scheduling knowledge base is adapted as the basis for a Model-based Intelligent Training system. Aspect of research relevant for the training framework are called Scheduling-MOBAT. This research work is partly reported in [van de Brug 1995]. The objectives of Scheduling-MOBAT are:

  • to define distinct levels of reasoning occurring at the domain level, in the form of an expert system, within a generic training framework;
  • to verify the MOBAT methodology with a complex industrial problem.

The next Section 5.3 starts with an overview of the scheduling KBS module. This research is extending the training framework as shown in Figure 5-1. A discussion of training analysis methods is provided in Section 5.4 and design methods are considered in Section 5.5. Both these analysis and design sections contain summaries supporting the initial framework that resulted from Proto-MOBAT shown in Figure 4-1. The training system design needs to embody principles of learning.

Figure 5-1 Scheduling-MOBAT Framework Overview

Figure 5-1 Scheduling-MOBAT Framework Overview

A set of preferred trainee learning modes (rote, deductive, inductive) from ITSIE has been introduced in the previous chapter. In this chapter, the ability to embody transfer of learning methods within the domain expert are described. The SOAR categorisation of within-trial, across-task and within-task learning principles are discussed with training units in Section 5.5. Augmenting the existing knowledge base for the purpose of the Scheduling-MOBAT training domain is presented in Section 5.6. The MOBAT framework uses a multi-level knowledge representation to support the different ways that people learn (i.e., rote, deductive or inductive). By selecting the appropriate combination of learning mode and expected level of trainee behaviour (as per [Rasmussen 1986] classification), a generic trainer agent can make use of multi-level knowledge representations. In this chapter three different forms of domain knowledge are specified as procedural,associative and principled knowledge. This multilevel approach permits various trainer methods for a given training objective. Each of the distinct levels of reasoning are presented for procedural knowledge, associative knowledge and principled knowledge. Examples of the transfer of learning methods are discussed which support the classification of distinct knowledge models within the training framework.

*
Flower Show
  *  
© RuleWorks.co.uk | | Sitemap