Advancing Process Safety Evaluation Through Dynamic Data Envelopment Analysis: Measuring Efficiency, Learning, and Risk Mitigation Outcomes

This paper applies Dynamic Data Envelopment Analysis (Dyn-DEA) to evaluate the intertemporal efficiency of Process Safety in large-scale Brazilian refineries, considering multiple inputs, outputs, and intertemporal linking variables (carry-overs). The objective is to identify which units effectively convert resources allocated to safety into safe operational performance, organizational learning, and risk mitigation, while accounting for the significance of accumulated assets and critical backlog. The model is output-oriented and employs linear programming, incorporating variable returns to scale, technological comparability, and normalization of structural factors. Inputs include maintenance, inspection, training, audits, and management-of-change efforts; desirable outputs encompass safe throughput, implementation of recommendations, and critical barrier integrity; undesirable outputs comprise loss-of-primary-containment events and unplanned downtime hours. Intertemporal carry-overs include barrier integrity (desirable) and critical action backlog (undesirable). The results reveal three patterns: refineries on the dynamic frontier, accumulating integrity and reducing backlog; transitional units, showing operational gains still constrained by backlog; and units exhibiting structural inefficiency, with slow recovery trajectories. The analysis demonstrates that efficiency is cumulative, dependent on intertemporal trajectories, organizational learning, and effective barrier management, and that investment-focused policies alone may be insufficient to enhance Process Safety performance.

Keywords: Process safety, data envelopment analysis, dynamic performance assessment, organizational learning.