Intelligent Uncertainty Handling Using Artificial Neural Networks in a Programmatic Logic Controller-Based Automation System

Abstract

Abstract The manufacturing industry is relying more and more on artificial intelligence (AI). Despite this, Programmable Logic Controllers (PLCs), one of the most common industrial tools, still lack AI integration. This means that PLCs are still very rigid and struggle with handling uncertainties. Uncertainties are input data that the controller has not been programmed to address. Uncertainties can be caused by variation in sensor data or other factors. Because PLCs struggle to handle uncertainties, this makes them very tedious to program. Implementing AI into the PLC may eliminate this problem. This study answers the question of whether AI can be implemented into a PLC (Allen-Bradley). To do this, a Levenberg-Marquardt backpropagation artificial neural network (ANN) was trained in MATLAB to approximate a mathematical function. A representation of the ANN was then constructed using the PLC ladder logic diagram by implementing the weight and bias values obtained from the trained ANN. The PLC-implemented ANN’s results were validated by comparing its output values to those of MATLAB trained ANN. Negligible differences between the outputs indicate that implementing an ANN framework into the PLC ladder logic code is viable. The ability to implement an ANN into the PLC code may help PLCs handle uncertainties better. This implementation may help the PLC become more versatile for more jobs and may also make it easier to program. Future studies should test out this method of implementation and apply it in real world applications.