Improvement of the Objects Sorting Process Using Machine Learning on the Example of a Created Prototype

Various industries such as waste management, food production, automotive, and aerospace industries require precise sorting and categorization of products and waste. This is particularly important in the era of Industry 4.0, where automation and robotics in manufacturing processes are crucial. Machine Learning is an important tool in this process, enabling machines to improve their performance based on past experiences. Products, semi-finished goods, and waste can be differentiated based on color, weight, density, consistency, shape, dimensions, and chemical composition. There are advanced but costly methods for distinguishing products with different dimensions and shapes using sensors, scanners, cameras, or sets of sieves that allow only desired products to pass through. On the other hand, products with uniform shape and/or dimensions, as well as unique products (e.g., PET bottles in different colors), are recommended to be optically sorted based on the desired color. Currently available solutions are effective but too expensive for smaller facilities that want to automate their production or processing processes. The author has developed a prototype sorting device that, with further refinement, can be easily and inexpensively implemented in various industrial sectors. The device uses a photo-optical object detection mechanism, which compares the objects with basic color patterns programmed based on user color interpretation. The uniqueness of the device lies in its utilization of Machine Learning by comparing the color of the sorted sample with the reference shade obtained from an online database. The prototype project has demonstrated that current technologies enable the easy implementation of a complex automatic sorting process based on machine learning and scalable electronic infrastructure without incurring significant costs. The development of this device would find applications in many industrial sectors, both in sorting and collecting objects. It would be particularly useful in industries where traditional vacuum sorting is not feasible due to the risk of product damage.