Application of Industrial Agents for Availability and OEE Improvement in CyberPhysical Production Systems

Abstract

Unplanned downtime is a key cause of availability and OEE losses in industrial systems. This article develops and implements a Cyber-Physical Production System with distributed control based on Multi-Agent Systems, where Resource Agents (RA) interact with physical process variables and their functionalities. The development of the CPPS is structured through model-based engineering using SysML for system description and analysis, facilitating conceptual transfer and reusability. The cyber layer is integrated non-intrusively with an industrial controller through the OPC UA protocol, preserving control determinism. In the presence of induced sensor faults (physical layer), the RA detects inconsistencies, classifies criticality and, when the fault is compensable, executes resilience through soft-sensors and override flags. This enables temporary substitution of the physical signal, maintaining operational continuity and logging maintenance alerts, improving the availability variable of OEE. The proposal is validated in an emulated Pick & Place cell and subjected to tests with 20 cycles and random perturbations, compared against centralized industrial control and with the multi-agent system activated. With ideal quality and performance (≈90%), preliminary results show an availability of at least 70%. The MAS-based CPPS can increase OEE in traditional industrial systems.