Industry 4.0 has transformed the way modern manufacturing is being approached with the interplay of cyber physical systems, industrial internet of things (IIoT), cloud edge computing, and data driven intelligence. Machine learning (ML) has become one of the core technologies to enable the derivation of actionable insights out of heterogeneous manufacturing data. This paper provides a longer and comparative analysis of ML and deep learning (DL) algorithms with supervised, unsupervised, semi supervised, reinforcement learning and hybrid algorithms used in basic manufacturing areas like predictive maintenance, quality inspection and defect detection, process optimization, production planning and supply chain management. The review compares the performance of algorithms, the computational complexity, interpretability, and deployability based on a systematic review of the literature published since 2015. To increase methodological clarity, mathematical expressions of popular modeling, such as regression, support vector machines, convolutional neural networks (CNNs), and long short-term memory (LSTM) networks, are provided. The new areas of transfer learning, federated learning, edge AI, and explainable artificial intelligence (XAI) are presented in the framework of industrial scalability and reliability. This paper concludes that the key to closing the gap between laboratory scale ML models and real-world smart manufacturing systems lies in context aware model selection, hybrid, and explainable frameworks.
Paswan et al. (Fri,) studied this question.