Explainable AI for trustworthy intelligent process monitoring
Arne Johannssen, Peihua Qiu, Ali Yeganeh, Nataliya Chukhrova
Abstract
Statistical control charts are often based on assumptions that do not hold in complex, high-dimensional and dynamic environments. To counter these weaknesses, control charts based on artificial intelligence (AI) techniques have emerged as a powerful alternative in recent years. However, their black-box nature limits transparency, interpretability and trustworthiness that are essential to realize Industry 5.0. To address that issue, this Short Communication discusses the necessity of embedding explainable artificial intelligence (XAI) in AI-based control charts. Incorporating XAI provides a solution by enhancing the interpretability of AI-based control charts while maintaining their high predictive accuracy. This paper also identifies key challenges in embedding XAI and outlines future research directions for responsible and trustworthy AI-based process monitoring. • This paper addresses the black-box nature of artificial intelligence (AI) based control charts. • It highlights the necessity of embedding explainable AI (XAI) techniques into AI-based control charts. • It reviews recent AI-based control charts and XAI techniques. • Illustrative scenarios on how XAI can improve decision-making, compliance and trust in industrial settings. • Benefits, challenges and future directions of XAI-based control charts are discussed.