[Paper Review] Artificial Intelligence Approaches for Predictive Maintenance in the Steel Industry: A Survey

Jakub Jakubowski, Natalia Wojak-Strzelecka+5 more|arXiv (Cornell University)|May 21, 2024

Quality and Safety in HealthcareHealth Professions3 citations

TL;DR

This survey synthesizes 219 studies on AI-driven predictive maintenance (PdM) in the steel industry, analyzing AI methods, data types, and implementation challenges. It reveals growing use of deep learning for PdM in blast furnaces and hot rolling mills, with limited real-world deployment and low practical impact despite strong technical progress.

ABSTRACT

Predictive Maintenance (PdM) emerged as one of the pillars of Industry 4.0, and became crucial for enhancing operational efficiency, allowing to minimize downtime, extend lifespan of equipment, and prevent failures. A wide range of PdM tasks can be performed using Artificial Intelligence (AI) methods, which often use data generated from industrial sensors. The steel industry, which is an important branch of the global economy, is one of the potential beneficiaries of this trend, given its large environmental footprint, the globalized nature of the market, and the demanding working conditions. This survey synthesizes the current state of knowledge in the field of AI-based PdM within the steel industry and is addressed to researchers and practitioners. We identified 219 articles related to this topic and formulated five research questions, allowing us to gain a global perspective on current trends and the main research gaps. We examined equipment and facilities subjected to PdM, determined common PdM approaches, and identified trends in the AI methods used to develop these solutions. We explored the characteristics of the data used in the surveyed articles and assessed the practical implications of the research presented there. Most of the research focuses on the blast furnace or hot rolling, using data from industrial sensors. Current trends show increasing interest in the domain, especially in the use of deep learning. The main challenges include implementing the proposed methods in a production environment, incorporating them into maintenance plans, and enhancing the accessibility and reproducibility of the research.

Motivation & Objective

• To provide a comprehensive synthesis of AI-based predictive maintenance (PdM) research in the steel industry, addressing a critical gap in the literature.
• To identify key equipment types, AI methods, data characteristics, and practical implementation challenges in industrial PdM applications.
• To assess the business impact and real-world applicability of proposed AI-based PdM solutions in steel manufacturing.
• To highlight research gaps, including low reproducibility, limited industrial deployment, and insufficient collaboration between researchers and practitioners.
• To guide future research by emphasizing the need for business-oriented evaluation and integration into maintenance strategies.

Proposed method

• Systematic literature review using PRISMA methodology to identify and select 219 relevant articles from over 2,000 initial results.
• Keyword-based search combining terms related to AI, predictive maintenance, and the steel industry to ensure domain specificity.
• Independent reviewer screening and data curation to ensure methodological rigor and reduce selection bias.
• Thematic and quantitative analysis of selected papers across dimensions: equipment types, AI techniques, data sources, and business impact.
• Inclusion of non-AI but relevant works (e.g., signal processing, statistical models) to provide a broader context for PdM in steel production.
• Manual analysis without generative AI tools, supported by custom automation scripts for data processing and organization.

Experimental results

Research questions

• RQ1 Which equipment and production processes in the steel industry are most frequently targeted by AI-based PdM research?
• RQ2 What types of AI and machine learning methods are predominantly used, and how have they evolved over time?
• RQ3 What data sources (e.g., sensor data, images, text) are most commonly used in these studies, and what are their characteristics?
• RQ4 To what extent do proposed PdM solutions demonstrate practical implementation and business impact in real industrial settings?
• RQ5 What are the main challenges and research gaps in the current body of work, particularly regarding reproducibility, collaboration, and industrial integration?

Key findings

• The majority of research focuses on two critical steel production stages: the blast furnace and hot rolling mill, indicating their high failure risk and maintenance importance.
• Deep learning is the dominant AI approach, reflecting a growing trend toward complex, data-driven models for failure prediction.
• Over 90% of studies rely on sensor data from industrial equipment, with limited use of images or text, despite their potential for multimodal PdM systems.
• Despite strong technical development, only a small fraction of studies report practical implementation or measurable business impact in real production environments.
• A significant gap exists between research outputs and industrial deployment, with most methods not integrated into maintenance planning or operational workflows.
• Research is largely conducted in isolated teams with limited collaboration, reducing reproducibility and slowing progress toward industrial adoption.