Abstract:
Machine vision (MV) can help in achieving real-time data analysis in a manufacturing environment. This can be implemented in any industry to achieve real-time monitoring of workpieces for geometric defects and material irregularities. Identification of defects, sorting of workpieces based on their physical parameters, and analysis of process abnormalities can be achieved by using the real-time data from simple and cost-effective raspberry pi with camera and open source machine learning platform TensorFlow to run convolutional neural network (CNN) model. The proposed cyber-physical production system enables to develop a MV based system for data acquisition integrating physical entities of learning factory (LF) with the cyber world. Nowadays, LFs are widely used to train the workforce for developing competencies for emerging technologies and challenges faced due to technological advancements in Industry 4.0. This paper demonstrates the application of a cost-effective MV system in a learning factory environment to achieve real-time data acquisition and energy efficiency. The proposed low-cost machine vision is found to detect geometric irregularities, colours and surface defects. The simple cost effective MV system has enhanced the energy efficiency and reduced the total carbon footprint by 18.37 % and 78.83 % depending upon the location of MV system along the flow. The teaching-learning experience is also enhanced through action-based learning strategies. This not only ensures less rework, better control, unbiased decisions, 100% quality assurance but also the need of workers/operators can be reduced.