Determination of physical characteristics of mechanical elements using machine learning

Abstract

Introduction.  One of people's daily problems is verifying the status of various auto parts that are distributed in large quantities, because several spare parts warehouses that distribute mechanical parts depend on it to satisfy customers and their environment in general. To cover such needs, artificial neural networks have been developed that will classify these elements according to their physical characteristics. The acquisition of mechanical parts in the automobile industry is repeated countless times, so they can purchase defective mechanical parts. Objective.  Classification of mechanical elements using artificial neural networks for use in quality control of their physical characteristics. Methodology. A data collection methodology is applied that will help train the artificial neural network. The artificial neural network will be able to determine the quality status of the mechanical element based on the collected image data and will act as training of the neural network in the following process. Results. In the final test, 200 of these classified metal joints were used and it was observed that 10 had physical defects. Conclusion. Convolutional neural networks can be used to classify mechanical parts, extract their features from images, and then use them as a neural network database.