From Explanations to Actions: Leveraging SHAP, LIME, and Counterfactual Analysis for Operational Excellence in Maintenance Decisions

In the rapidly evolving landscape of Industry 4.0, the evolution from predictive Artificial Intelligence to explainable artificial intelligence (XAI) in predictive maintenance has emerged as a critical tool for achieving operational excellence. Because there are different methods for XAI, understanding the differences between the insights provided by various XAI methods would be highly beneficial for maintenance managers. By leveraging XAI maintenance managers can enhance operational efficiency, effectively address potential issues, and implement proactive maintenance strategies. And understanding which is the best method for their specific conditions would maximise the benefit they can get. This paper presents a comprehensive comparison of three prominent XAI methods: SHapley Additive exPlanations (SHAP), Local Interpretable Model-agnostic Explanations (LIME), and Counterfactual Explanations. By leveraging a dataset from the AI4I 2020 Predictive Maintenance Dataset, we demonstrate how each method provides unique insights into model predictions, supporting decision- making processes with different contributions. Our study addresses the existing gap in XAI applications by linking these methods to specific decision-making contexts in operational excellence. We present a decision flowchart to guide the selection of the most appropriate XAI method based on the user's needs, ensuring practical and actionable insights. This integrated approach not only enhances model transparency but also supports continuous improvement initiatives, driving sustainable growth and competitiveness in manufacturing operations.