Anomaly Detection for Industry Product Quality Inspection based on Gaussian Restricted Boltzmann Machine

In Industry 4.0, anomaly detection plays an important role in the process of product quality inspection, where the product data with high dimensions and highly imbalanced distribution give rise to some challenges. To handle these challenges, a novel anomaly detection method based on Gaussian Restricted Boltzmann Machine (GRBM) is proposed. To make it more tractable for training the model, the method performs distinct gradient compensations through integrating the free-energy function into the objective function in two stages of product quality inspection. Extensive experimental studies are respectively carried out on two real-world cases, i.e. wine quality and cigarette product testing. Three state-of art anomaly detection methods and two conventional GRBM methods are used for comparison analysis, and the results demonstrate that our proposed method provides effectiveness and superiority in product quality inspection.