Piezo device for micro-texturing surfaces with vision-based neural network quality assessment

With substantial progress in advanced manufacturing, industries now require effective means to detect, classify, and determine the topological parameters of micro-features imprinted on workpieces at the micro-scale. In this study, a novel piezo texture device was developed to generate micro-textures using vibrations. The tool comprises of an ultrasonic concentrator that amplifies vibrations, piezo-actuated sensors that convert electrical energy into microscale vibrations, and the corresponding electronic components needed to generate the required signal. The primary objective was to perform micro-texturing through turning with a device that creates vibrations using this setup. Micro-textures, often referred to as dimples, were successfully imprinted onto surfaces. Since it is impossible to check the micro-textures using traditional quality control techniques, artificial intelligence offers a powerful alternative by enabling real-time monitoring and defect detection based on pattern recognition. Indeed, Convolutional Neural Networks (CNNs), specifically GoogleNet and ResNet-50, were employed for dimple classification and further analysis, such as evaluating the similarity ratio of dimples on a workpiece.