Study On Method Of The Synchronized Manager Based Flexible Workflow Modeling

Workflow management technology is vital to enhance the agility of the enterprises, especially the self-adaptability, flexibility to change and the structural flexibility within the enterprises. Therefore, how to enable the workflow management system to automatically adapt to the environment becomes key to workflow management technology research recently.As the foundation of the workflow management system, the extent of workflow process model’s flexibility would influence the range of application and adaptation of the whole management system. By taking into account the previous research of the flexible workflow modeling, this paper is carried out along three dimensions:process meta-model, process model and the new modeling technique-workflow mining. The content is as follows:1. Tackling the problem of flexible workflow process meta-model. The possible dynamic change of workflow management system is analyzed and the dynamic features and definition of process model by Workflow Management Conference are discussed along the lines of process and activity. Based on this, new elements like flexible workflow definition, synchronizer and converter are introduced to improve the expansion process definition meta-model. The improved meta-model has more flexibility and ability of re-structuring. Due to the introduction of converter, the foundation of the intelligence of process meta-model is laid down. Some improvements and expansions are also done to WPDL, which is the description language of process meta-model.2. Workflow pattern based on synchronized manager has been discussed. This paper offers the concept of synchronized manager on the basis of flexible workflow process meta-model, and further modeling various workflow patterns by using synchronized manager. The synchronized manager based modeling method distinguishes the executive behavior and management behavior of the process, thus making the executive behavior more independent and with it no control information is affiliated. We could encapsulate components individually for the sake of re-use. At the same time, this separation makes it certain that synchronized manager is the main part of the process change, and it makes sure of dynamic synchronized management operation and relative static model structure, which is very important to construct correctly_and analyze the adaptability of workflow model to the environment. Through the dynamic management behavior of synchronized manager, the semantic discourse of dependency relationship between activities in process is made more accurate. By introducing split synchronized manager, join synchronized manager and void transition, the majority of workflow pattern is described simply and clearly. The model is simple and valid, and has compatibility.3. The mining of non-certain choice and parallel, jump, loop and invisible activity is researched. Analyze logical relationship between activities among workflow logs and provide relevant features and verification between these relationship and process model. Based on this,γandγ+ process mining algorithm have been raised to discover synchronized manager based workflow model from system logs. These algorithms solve the current problems of process mining:not only mining the logical relationship between activities, but also mining the management operations of process. y algorithm specially tackles the problem of mining of non-certain choice and parallel models; andγ+algorithm could find out structure of mandatory loop and optional loop, of jumping forward and backward, of invisible activity etc.4. This paper studies on-line dynamic workflow mining and proposes incremental process mining algorithm based of relationship matrix. This algorithm utilizes relationship matrix to store successive relations of log, incrementally update the relationship matrix, and use directional loop cutting algorithm to remove redundancy in the matrix. To discover loop structure, this algorithm utilizes complement matrix to handle single and structural loop, integrates complement matrix to the relationship matrix, and transform the model. At last, difference degree and sliding window are added to make window self-adaptive, this improves the fast responding ability of the algorithm. Robustness is added to the algorithm by using the threshold value mechanism.