An integer programming based approach for verification and diagnosis of workflows

Workflow analysis is indispensable to capture modeling errors in workflow designs. While several workflow analysis approaches have been defined previously, these approaches do not give precise feedback, thus making it hard for a designer to pinpoint the exact cause of modeling errors. In this paper we introduce a novel approach for analyzing and diagnosing workflows based on integer programming (IP). Each workflow model is translated into a set of IP constraints. Faulty control flow connectors can be easily detected using the approach by relaxing the corresponding constraints. We have implemented this diagnosis approach in a tool called DiagFlow which reads and diagnoses XPDL models using an existing open source IP solver as a backend. We show that the diagnosis approach is correct and illustrate it with realistic examples. Moreover, the approach is flexible and can be extended to handle a variety of new constraints, as well as to support new workflow patterns. Results of testing on large process models show that DiagFlow outperforms a state of the art tool like Woflan in terms of the solution time.