Machine learning based fault detection approach to enhance quality control in smart manufacturing

In recent days the quickly changing manufacturing environment has pushed organizations to accomplish more consumer satisfaction by improving item quality, lessening production cost, and acknowledging maintainability. Anomaly recognition impacts the nature of items and it is typically directed through visual quality assessment. The visual quality review of an item can be performed either physically or naturally. This research proposes a novel technique in manufacturing industry-based fault detection and control management using machine learning technique. Here the input data has been collected as manufacturing fault historical data by IoT (internet of things) module. This data has been processed for noise removal, normalization, and smoothening. The processed data features have been extracted by using kernel principal vector component analysis. Gaussian quadratic Kernelized Generative Adversarial Network has been used to manage the control of extracted features. The experimental analysis has been done in terms of RMSE, MAP, AUC, F-1 score, recall, accuracy, and precision. Upon reviewing our procedures after our research and testing, we discovered that Auto-Encoder Neural Network is the best effective algorithm for identifying a production line failure. One-Class SVM gives the highest accurate findings in our machine-based investigations.