<p>Input and output signals in an industrial automation system are continuous, and when these signals are converted into transition sequences, the transition sequences become infinite. Effective partitioning helps to build accurate models when analyzing these sequences. A Petri net is a graphical mathematical tool that is employed for the description and analysis of discrete event systems. The utilization of places and transitions enables the modeling of system behaviors, thereby facilitating the comprehension of the system and providing a foundation for the optimization of industrial processes. To facilitate the construction of a Petri net, the infinite transition sequences must be partitioned into multiple traces. The challenge in the partitioning process is that the breakpoints for partitioning the sequences are unknown, which can lead to inefficiencies in the analysis if a simple exhaustive approach is used. This study aims to ensure that the sequence is partitioned accurately and efficiently. A framework traces recognition based on frequent itemset is proposed in this paper. The framework first generates frequent items based on the frequency of occurrence of transitions, then groups the frequent items based on their prefix relationships, and finally generates candidate itemsets by identifying repeatable segments of frequent items. The optimal candidate set generated is the traces in event log. Experimental results show that the method can generate traces quickly. The exponential time complexity of the original partitioning algorithm is compressed to polynomial time.</p>