Industrial Process Model Mining And Parallel Retrieval Research

The information age enables production or enterprise management to be easily and quickly operated on the Internet.Especially for the complex and discretized assembly line of the production process in the industrial manufacturing field,of which production records are complicated and numerous,which can be conveniently stored in the information system.The goal of process mining is to extract valuable objective information from the records of available event logs in the information system to discover the real process.Process mining can help enterprise improve the production processes and optimize the management processes.Most of the existing model mining algorithms are performed on the complete event log data,and the mining on log data doesn't work well when there exists noise.Based on the research of process model mining algorithm,this article is focuses on the efficient retrieval technology of the generated process model library.It belongs to the scope of process model discovery and compliance detection.Most process mining algorithms cannot deal with the incomplete and infrequent trajectory problems in the event log at the same time.And whether the event log generated during the production process is consistent with the pre-set or already constructed process model.Based on the two questions above,this thesis has carried out the following research:(1)For incomplete event logs and infrequent behavior in logs,this paper proposes a heuristic process model mining algorithm based on FP-growth algorithm.The algorithm takes the task pairs appearing in the log as a single set and calculates its support.And then by calculating the confidence to determine the reliability between tasks.The dependencies between tasks are mined in the log.Then a heuristic mining algorithm is applied to mine the process model.In order to facilitate the model retrieval,the mined process model is transformed into a Petri net representation.(2)A process model library was established according to the constructed process model.A parallel retrieval algorithm,based on the consistency of the behavior trajectory in the event log and the execution sequence in the process model library,put forward to measure the quality of the mining model.According to the idea of “survival of the fittest” in genetic algorithm,the algorithm combines the behavior in the event log to be detected with the transition in the process model library to form a consistent behavior pair,then compares the consistency of the consistency behavior pairs,records the missing in it.according to the missing,calculate the fittness and accuracy of the model,and then judge whether the process model conforms to the actual situation.Due to the complexity and discretization of the industrial production process,the recorded production log is incomplete,making the constructed process model inaccurate and unable to reflect the actual production process.This thesis maximizes the use of the original event log to build a process model,and establishes a process model library to study the retrieval algorithm of the process model.Finally,the proposed algorithm is compared with the classical method —heuristic process model mining algorithm and heuristic process model mining algorithm based on FP-growth algorithm,Token-based trajectory replay detection algorithm and parallel retrieval algorithm based on behavior trajectory consistency,compare these algorithms in different event logs.Experimental result verifies the correctness and rationality of the process model mining algorithm and model retrieval algorithm proposed in this thesis.By continuously modifying the model,when the process model matures to a certain extent,the process model can guide the actual production,and solve the problem that the industrial manufacturing process has low production efficiency and the production process is opaque and difficult to monitor in real time.The purpose of real-time monitoring is to terminate those error or erroneous production processes in real time to save production costs and increase production efficiency.