For any system to run seamlessly and without failure is very important in several aspects including overall cost, market value and competition as well as overall performance of the entire process. This is very essential for any manufacturing system that the faults can be predicted/detected and diagnosed, which may help for scheduling preventive maintenance as well as finding the right solution in right time. The chapter discusses the artificial intelligence and machine learning methods such as genetic algorithm, Bayesian classifier, decision tree, random forest, k-nearest neighbour, support vector machine as well as artificial neural networks, Long Short-Term Memory that have been used to detect, classify, and locate faults a variety of faults for the real-world cases of conditioning monitoring of the roller bearing systems, short-circuit fault detection, traction motor fault detection, etc. The chapter also reviews the self-supervised leaning method such as data driven Kernel Principal Component Analysis for the fault prognosis usable for the small scale as well as large scale manufacturing system. In addition, the chapter also describes generative adversarial network uses labelled and unlabelled images for classifying die cast products and helps in detecting the fault in the system.