Abstract
This paper provides a methodology to assess the optimal multi-agent architecture for collaborative prognostics in modern fleets of assets. The use of multi-agent systems has been shown to improve the ability to predict equipment failures by enabling machines with communication and collaborative learning capabilities. Different architectures have been postulated for industrial multi-agent systems in general. A rigorous analysis of the implications of their implementation for collaborative prognostics is essential to guide industrial deployment. In this paper, we investigate the cost and reliability implications of using different multi-agent systems architectures for collaborative failure prediction and maintenance optimization in large fleets of industrial assets. Results show that purely distributed architectures are optimal for high-value assets, while hierarchical architectures optimize communication costs for low-value assets. This enables asset managers to design and implement multi-agent systems for predictive maintenance that significantly decrease the whole-life cost of their assets.
Highlights
The potential of using computational models to enable real-time machine failure prediction has been known since the 1980’s (Buchanan 1986)
We describe them within the context of collaborative prognostics, more general descriptions can be found in Brennan et al (2002), Marík and Lažansky (2007), Monostori et al (2006), Andreadis et al (2014) and Leitão and Karnouskos (2015)
A separated experiment is performed for each type of architecture described in this paper, and prognostics, clustering and maintenance recommendations are executed such as described in the multi-agent system architectures section
Summary
The potential of using computational models to enable real-time machine failure prediction (prognostics) has been known since the 1980’s (Buchanan 1986). We compare several canonical multi-agent architectures for collaborative prognostics on the basis of different cost balances between communication, maintenance and computation. The architectures reviewed in this paper are formed by four elements: Virtual Assets, Digital Twins, Mediator Agents, and a Social Platform. The main task of the Social Platform is to run algorithms leveraging information originating from the whole fleet These algorithms can be aimed at (1) forming clusters of collaborating assets, (2) retrieving and plotting enterpriselevel information, or (3) calculating prognostics and making maintenance decisions. All the architecture types described above were simulated using the same strategy: Netlogo simulated the behaviour of the agents (i.e. initiating the fleet of assets, connecting similar agents together, computing agent failures, etc), and prognostics/clustering algorithms were implemented using Python scripts. Select one random agent from the fleet; while number of centroids
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call