Concept drift is a critical phenomenon in business process management where enterprise process models undergo alterations over time . Traditional detection methodologies concentrate on control flow changes, with less emphasis on the temporal dimension of drift. This paper addresses this gap by calculating correlation coefficients of activity durations, leveraging the isolation forest algorithm to identify abnormal coefficients, and subsequently employing the Page-Hinkley algorithm to precisely detect points of temporal drift. The efficacy of our proposed strategy is rigorously tested through experiments incorporating both synthetic and real-world log datasets, with comparative assessments against three distinct methodologies. The results of these experiments demonstrate that the proposed method is effective and achieves higher precision.