Scholarlink

<p>Machine Learning (ML) has become a central enabler of intelligent, data-driven decision-making, predictive analytics, and adaptive control in modern manufacturing. This systematic review synthesizes advancements in ML-based process optimization from 2015 to 2025, analyzing 185 peer-reviewed articles from Scopus, Web of Science, and IEEE Xplore. The dominant trends include the continued use of supervised learning for predictive tasks and the growing adoption of reinforcement and deep learning for real-time adaptive control. Emerging paradigms such as generative AI and diffusion models enhance design optimization and additive manufacturing, untrained neural networks and deep image priorities enable data-efficient anomaly detection, and liquid neural networks offer adaptive event-driven control suited for IoT-enabled factories. Integration with digital twins and the Industrial Internet of Things (IIoT) supports cyber-physical production systems, whereas multimodal deep learning fuses diverse data sources to provide context-aware intelligence. However, challenges remain, including data scarcity, interpretability, computational demands, scalability, and robust validation for safety-critical applications. Future research should focus on hybrid AI models, explainable AI (XAI) frameworks, standardized MLOps, data-efficient learning, and human-centric sustainable manufacturing aligned with Industry 5.0 principles. This review provides a comprehensive foundation for advancing ML applications in modern manufacturing systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning for process optimization in manufacturing systems: a systematic review of algorithms, applications, and performance evaluation

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Tesfa Guadie Ayaliew,
  • Tibebu Alene Asresa

摘要

Machine Learning (ML) has become a central enabler of intelligent, data-driven decision-making, predictive analytics, and adaptive control in modern manufacturing. This systematic review synthesizes advancements in ML-based process optimization from 2015 to 2025, analyzing 185 peer-reviewed articles from Scopus, Web of Science, and IEEE Xplore. The dominant trends include the continued use of supervised learning for predictive tasks and the growing adoption of reinforcement and deep learning for real-time adaptive control. Emerging paradigms such as generative AI and diffusion models enhance design optimization and additive manufacturing, untrained neural networks and deep image priorities enable data-efficient anomaly detection, and liquid neural networks offer adaptive event-driven control suited for IoT-enabled factories. Integration with digital twins and the Industrial Internet of Things (IIoT) supports cyber-physical production systems, whereas multimodal deep learning fuses diverse data sources to provide context-aware intelligence. However, challenges remain, including data scarcity, interpretability, computational demands, scalability, and robust validation for safety-critical applications. Future research should focus on hybrid AI models, explainable AI (XAI) frameworks, standardized MLOps, data-efficient learning, and human-centric sustainable manufacturing aligned with Industry 5.0 principles. This review provides a comprehensive foundation for advancing ML applications in modern manufacturing systems.