Artificial Neural Network Applied to Virtual Commissioning and Control of a Robotic Cell

Abstract

The world is undergoing significant changes in information technologies and industrial processes. The rise of Industry 4.0 and the advancement of artificial intelligence are creating new opportunities and challenges for industries. This study investigates the feasibility of substituting a traditional programmable logic controller (PLC) with a neural network-based control system for discrete event management within a robotic cell. The research assesses the feasibility of this replacement, analyzing the associated challenges, limitations, and advantages compared to traditional methods. Digital manufacturing software is employed for simulating and validating the proposed model through a Virtual Comissioning (VC). The control system of the proposed model utilizes artificial neural networks, trained using data derived from a Boolean logic model. The results indicate that it is possible to swiftly train an artificial neural network (ANN) to take over cell control. This approach opens up the possibility of implementing low-cost hardware, aligning the system with the concepts of Industry 4.0. Additionally, the virtual modeling conducted using digital manufacturing software paves the way for a future implementation of a digital twin. Findings indicate that the neural network control approach is feasible and offers operational advantages over traditional programming methods.