Knowledge based approach to flexible workflow management systems

Abstract

Today's business environments are characterized by dynamic and uncertain environments. In order to effectively support business processes in such contexts, workflow management systems must be able to adapt themselves effectively. In this dissertation, the workflow is redefined in concept and represented with a set of business rules. Business rules play a central role in organizational workflows in context of cooperation among actors. To achieve business goals, they constrain the flow of works, use of resources, and responsibility mapping between tasks and actors using role concept. Business rules are explicitly modeled in the Knowledge-based Workflow Model (KWM) using frames. To increase the adaptability of workflow management system, KWM has several distinctive features. First, it increases expressiveness of workflow model so that exception handling rules and responsibility mapping rules between tasks and actors as well as task scheduling rules are explicitly modeled. Secondly, formal definition of KWM enables one to define and to analyze correctness of workflow schema. Knowledge-based approach enables more powerful analysis on workflow schema including checking consistency and compactness of routing rules as well as terminality of a workflow. Thirdly, providing change propagation mechanism which assures correctness of workflow after the modification of workflow schema increases adaptability. Change propagation rules for the modification primitives are provided to manage workflow evolution. On the other hand, metarules that control rules in KWM are used to handle exceptions that occur in a running workflow instance. Workflow participants can easily change workflow schema of a workflow instance with the support of extra rules and a metarule. Based on KWM, K-WFMS (Knowledge-based WorkFlow Management System) has been implemented in client/server architecture. Inference shell of knowledge-based systems is employed for enactment of business rules and integrated with database systems. From a real application based on the KWM architecture, it has been shown that system performance can increase notably by reducing the number of rules and facts that are used in the course of workflow enactment.