Design And Implementation Of Workflow Mining Algorithm For Railway Communication And Signal Process With Complex Structure

The extensive use of new technologies and new equipment in the railway system increases the pressure on the safe production of the railway communication and signaling department.At the same time,because of a series of problems,the effect of the information system that is put into use by the C&S department is not satisfactory.Facing the complex structure which cannot be accurately identified by the existing workflow mining algorithms in the C&S production activities,a new mining algorithm was proposed to generate a more accurate process model through extracting the objective control flow information of the business process from the event log recorded by the information system.This model was employed to drive the workflow engine to supervise and manage the execution of production activities,so as to improve the efficiency and quality of mission execution of all production activities,and ensure the safety of driving effectively.The main work of this paper is as follows:(1)The research status of workflow mining was analyzed.Four kinds of workflow mining algorithms,abstraction-based algorithms,heuristics-based algorithms,search-based algorithms and algorithms based on theory of regions,were summarized.The advantages and disadvantages of various algorithms were compared.The concept and application of workflow mining were described.The basic theory of Petri-net and WF-net was introduced.(2)Facing the problem that two types of complex structure,direct looping and indirect association,cannot be detected by existing algorithms,the basic definitions of two types of structure were given respectively based on Petri-net and WF-net.The behavior characteristics of two complex structure hidden in the event log were analyzed.Based on the basic ordering relations of the Alpha algorithm,the complex ordering relations were designed to detect any possible relationship between two tasks,such as single-OR-join,single-OR-split,delayed-execution,coexistence,interval-follow,repeated-execution,one-way-loop,direct-loop and indirect-relation,etc.Two kinds of detection theorems and theoretical proofs of complex structure were given respectively to solve the problems of existing algorithms and lay the foundation for designing the new algorithm.(3)In order to eliminate the influence on detection method when the two kinds of complex structure coexist,a detection sequence of two methods was determined,and a method to handle the event log was designed.Facing the redundant-relation caused by concatenation of indirect-relation structure,a method which eliminates the redundant-relation was designed to overcome the influence of redundant-relation on constructing the workflow-net.Based on the task relation identified by the detection method from the event log and the workflow-net constructed by the causality matrix,the Alpha-KD algorithm was designed to generate a process model with complex structure of direct-loop and indirect-relation.The algorithm was implemented in Java programming and was integrated into the ProM framework as a plug-in.(4)Based on the method of conformance checking,two types of datasets,real business and synthetic case,were used to evaluate and compare the Alpha-KD algorithm with the existing algorithms.The results show that the Alpha-KD algorithm can construct a process model with higher adaptability and accuracy.The process model constructed by Alpha-KD algorithm was deployed in the management information system for C&S production activities to realize the reminding of to-do tasks,the tracking of the business status and the viewing of the circulation history,and verify the effectiveness and feasibility of the algorithm.