Research on transition sequence partitioning based on dependency relationships

Programmable logic controllers (PLCs) are widely used in infrastructure systems such as energy supply, sewage treatment, and transportation. They continuously collect input signals and send output signals unless the production system fails. A transition sequence transformed from these signals would be continuous and unsuitable for further use, as it is very long. Specifically, the transition sequence lacks ‘Case id’, preventing direct application of process mining techniques. In this paper, we propose an approach to partition the transition sequence based on dependency relationships. The aim is to transform a transition sequence into a log that can be further utilized by process mining. The method classifies dependency relationships into three types: causal, concurrent and irrelevant, and develops partitioning rules based on these relationships, thus partitioning the transition sequence into traces. The accuracy of the partitioned result is further improved by two optimization strategies, namely partition correction and traces pruning. The experimental results demonstrate that our method, which partitions the transition sequence in polynomial time, is accurate, efficient, and effectively preserves original loops and concurrency structures. The method is applicable to any discrete event log analysis lacking clear traces by default, providing good technical support for automated process mining in industrial control systems.