Using neural networks in programmable logic controllers as an opportunity to abandon classical programming
Abstract
Using neural networks in programmable logic controllers as an opportunity to abandon classical programming
Incoming article date: 17.08.2025The article describes an experiment on designing a neural network for a programmable logic controller in order to eliminate the need to involve programmers in the development of automated control systems. The main task of programmable logic controllers is to simplify the automation of technological processes; they practically eliminate the tasks of developing printed circuit boards and soldering operations of elements. Obviously, the fewer different tasks that have to be solved and the simpler these tasks are, the faster the development and launch of a new system will be, and its cost will be lower. For the same purpose, for programming controllers, fairly simple and visual languages are used, this greatly simplifies the work of programmers. At the current level of microelectronics development, the computing resources of controllers significantly exceed the level necessary for most automation tasks. The question certainly arises whether it is possible, using excess computing power, to develop a single universal program capable of adapting to any technological process. Surely, such a program will work slower and take up more memory, but in this case, the programming task should degenerate into the task of setting up ready-made software. The article is devoted to the development of a prototype of such a program based on the single-layer perceptron model. The structure and parameters of the developed neural network are described taking into account the characteristics of the target platform. The training process of the designed neural network is analyzed. The limitations imposed on the development are listed and substantiated. The advantages and disadvantages, as well as development options, are outlined.
Keywords: programmable logic controller, artificial neural network, single-layer perceptron, relay logic language, automated process control system