Systematic review of predictive maintenance and digital twin technologies challenges, opportunities, and best practices

Nur Haninie Abd Wahab, Khairunnisa Hasikin, Khin Wee Lai, Kaijian Xia, Lulu Bei, Kai Huang, Xiang WuPeerJ Comput. Sci 2024 · DOI

TL;DR

This paper presents a comprehensive systematic review of predictive maintenance (PdM) and digital twin (DT) technologies, analyzing 34 studies from 2018-2023. It identifies key challenges, opportunities, and best practices for integrating these technologies, particularly highlighting their potential in under-explored sectors like healthcare, utilities, and agriculture. The review underscores the growing importance of DT-driven PdM for enhancing equipment reliability, optimizing maintenance, and driving operational efficiency across diverse industries.

What Problem Does This Paper Address?

Modern industrial machinery is increasingly complex, making it challenging to predict and manage equipment failures. Traditional reactive or preventive maintenance strategies are often insufficient, leading to unexpected downtime, reduced efficiency, and potential safety hazards. Predictive Maintenance (PdM) aims to address this by monitoring equipment health in real-time to anticipate failures. Digital Twins (DTs), virtual replicas of physical assets, offer a powerful platform for enhancing PdM by enabling simulations, real-time monitoring, and data-driven insights. However, the integration of PdM and DT, especially in critical but less-explored sectors like healthcare, utilities (smart water management), and agriculture (smart farm), faces challenges related to scalability, interoperability, data integration, and human-centric adoption. This paper aims to systematically review existing research to identify these challenges, explore opportunities, and outline best practices for effective PdM and DT implementation.

Methods

The authors conducted a systematic literature review following the PRISMA guidelines. They searched eight academic databases (Web of Science, Scopus, IEEE Xplore, Science Direct, Medline Complete, Emerald, Springer Link, and Dimensions) for relevant articles published between 2018 and 2023. Inclusion criteria focused on research articles and conference proceedings in English related to PdM and DT, while excluding review articles, book chapters, and non-English publications. A total of 34 articles were selected after rigorous screening, eligibility assessment, and quality evaluation. Thematic analysis was employed to categorize the findings into four main themes: field of interest and model types, approaches, predictive outcomes, and maintenance management implementation. Visualization tools like VOSviewer were used to map keyword co-occurrence and identify research clusters.

Datasets

Not discussed in the paper. The paper focuses on a systematic review of existing literature and does not introduce or utilize specific datasets for training or evaluation.

Key Insights and Ideas

  • Synergistic Power: The combination of PdM and DT is not just additive; it creates a synergistic effect that significantly enhances equipment maintenance, moving from reactive to proactive strategies.
  • Untapped Potential in Critical Sectors: While PdM and DT have seen significant adoption in manufacturing, their application in healthcare, utilities (smart water management), and agriculture (smart farm) remains largely underexplored, representing a critical research gap with immense potential for societal benefit.
  • Model Diversity: A variety of models are employed, including physical, behavioral, decision-making, and hybrid models, each with its strengths and limitations, indicating that the "best" model is context-dependent.
  • AI as the Engine: Deep learning (DL) and machine learning (ML) are the core algorithmic approaches driving PdM and DT, with ML generally showing higher accuracy but DL offering faster training times.
  • Beyond Technology: Human Factor is Crucial: Successful implementation hinges not only on technological integration but also on organizational culture shifts, comprehensive training, and effective collaboration between diverse stakeholders.

Key Results

  • Publication Trend: The study identified a consistent upward trend in publications related to PdM and DT from 2018 to 2023, indicating growing research interest.
  • Dominant Disciplines: Engineering, computer science, and energy are the leading disciplines in PdM research using DT.
  • Key Themes: The review identified four main themes: field of interest and model types (41% physical models, 12% behavioral, 6% decision-making, 41% hybrid), approaches (deep learning and machine learning), predictive outcomes (prognostics and health management, fault diagnosis, anomaly detection, RUL estimation), and maintenance management implementation (CBM, FMEA, CMMS, RCM).
  • Industry Adoption: Successful implementations were observed in automotive, civil/architectural, aerospace, mechanical, marine, railway, and electric/electronic engineering sectors.
  • Challenges: Key challenges identified include data collection and management, data integration, model development, compliance with regulations, cost of implementation, and organizational culture changes.

Limitations and Future Work

Limitations: * The study is a literature review, and its findings are based on existing research. * A limitation in implementing PdM with DT is that every equipment or system must be monitored in real-time with sensors, which requires internet coverage. * Implementing PdM with DT comprehensively can incur significant costs for facility management.

Future Work: * Further research is needed to address the knowledge deficit in critical sectors like healthcare, utilities, and agriculture. * Developing practical solutions for improved equipment dependability and optimized maintenance procedures is crucial. * Exploring innovative approaches and frameworks to optimize the computational efficiency of real-time DT simulations, especially in large-scale and complex industrial environments. * Developing more comprehensive PdM decision-making models based on DT for better simulation of real mechanisms across organizations. * Investigating alternative machine learning techniques and integrating additional variables into predictive models. * Addressing the challenges of data quality, availability, and security, particularly for vital infrastructure.

Summary updated

BibTeX
@article{wahab2024systematic,
  author = {Nur Haninie Abd Wahab and Khairunnisa Hasikin and Khin Wee Lai and Kaijian Xia and Lulu Bei and Kai Huang and Xiang Wu},
  title = {{Systematic review of predictive maintenance and digital twin technologies challenges, opportunities, and best practices}},
  journal = {PeerJ Comput. Sci.},
  year = {2024},
  doi = {10.7717/peerj-cs.1943},
}
Figure