Knowledge Representation And Modeling Methods For Disruption Management Problems In Distribution Decisions

Distribution system is complicated and has uncertainties. The distribution process is frequently disrupted by the events of customers changing their demands, changing their delivery times, vehicle breakdowns, etc. These events are named unexpected events, which may disable the distribution plan being executed and make the distribution system abnormal. Disruption management is a method which should promptly produce a new distribution plan deviating least from the original one after an unexpected event emerges. Hence, the key problem of disruption management is how the unexpected events can be handled in real time and in a scientific way. Based on the qualitative and quantitative analysis of the unexpected events, utilizing models and algorithms to produce a solution is one of the important strategies for improving the science of decision making, while utilizing the computer to efficiently realize the process of online modeling in real time is one of the important methods for improving the instantaneity of the decision process. However, due to the complexity of the distribution system, and the real-time and dynamic characteristics of disruption management problems, it's hard for existing theoretical methods to solve the real-time modeling of this kind of problem. This research aims to improve the science of the decision making and the instantaneity of the decision process in disruption management of the urban distribution. It studies the knowledge representation method for disruption management problems in urban distribution decisions. And based on the representation method, it studies the real-time and online computer-aided modeling and problem-solving method that can achieve effective plans for supporting the decision-making process of disruption management in distribution decisions. This research includes the following aspects.(1) The knowledge representation method for disruption management problems in distribution decisions is studied. The disruption management problem and the disruption management process are analyzed. The objects composing the distribution system are abstracted and their states are analyzed. The factors of policies used to handle unexpected events are analyzed. The elements of knowledge representation for disruption management problems are studied from three aspects, the representation of unexpected events, the representation of distribution states, and the representation of rules for analyzing and identifying disruptions. Based on the above research, the three-tuple-based knowledge representation method, ESR, for disruption management problems in urban distribution decisions is proposed. The online process of analyzing and idenfying disruptions is realized, which sets the foundation for the subsequent modeling process.(2) The knowledge-based modeling method for disruption management problems in distribution decisions is studied. The characteristics of disruption-handling policies, algorithms and models involved in the policy-optimizing processes are analyzed. The representation methods for the knowledge of modeling and the realization of the process of modeling are studied from the three aspects, the matching of the policies, the selecting and constructing of algorithms, and the constructing of models. Based on the above research, the knowledge-based modeling method for disruption management problems in distribution decisions is proposed.(3) The modeling support system for the decision making of disruption management is developed. Based on the above mentioned theoretical research of the knowledge representation method and the knowledge-based modeling method, a prototype of modeling support system for disruption management in distribution decisions is designed and developed. Based on the analysis of the functions and the structure of the system, six subsystems are designed and developed by using the tools of "Amzi!Prolog", "Visual C++", "Access", etc. The six subsystems are the subsystem of real-time monitoring the distribution state, the subsystem of man-machine interaction, the subsystem of knowledge base, the subsystem of database, the subsystem of model base, and the subsystem of real-time modeling and problem solving. Moreover, a computational experiment is conducted to prove the efficiency and effectiveness of the proposed modeling method. Finally, the prototype is tested by real-world cases from Dalian Marketing Subsidiary of PetroChina Corporation.The research is an intersection of the theories of Artificial Intelligence, Knowledge Engineering and Operations Research Optimization, which is a beneficial exploration for implementing the disruption management process in real time and on line. The research results combining with the techniques of collecting vehicle's data and monitoring vehicles in real time can provide decision support for real-time scheduling of the distribution process. The research is significant in the sense of improving the service quality of distribution companies. The methodology presented in the research provides a theoretical reference for solving multi-goal programming problems similar to disruption management problems in urban distribution decisions.