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.
A population-level analysis is proposed to address data sparsity when building predictive models for engineering infrastructure. Utilizing an interpretable hierarchical Bayesian approach and operational fleet data, domain expertise is naturally encoded (and appropriately shared) between different subgroups, representing (1) use-type, (2) component, or (3) operating condition. Specifically, domain expertise is exploited to constrain the model via assumptions (and prior distributions) allowing the methodology to automatically share information between similar assets, improving the survival analysis of a truck fleet (15% and 13% increases in predictive log-likelihood of hazard) and power prediction in a wind farm (up to 82% reduction in the standard deviation of maximum output prediction). In each asset management example, a set of correlated functions is learnt over the fleet, in a combined inference, to learn a population model. Parameter estimation is improved when subfleets are allowed to share correlated information at different levels in the hierarchy; the (averaged) reduction in standard deviation for interpretable parameters in the survival analysis is 70%, alongside 32% in wind farm power models. In turn, groups with incomplete data automatically borrow statistical strength from those that are data-rich. The statistical correlations enable knowledge transfer via Bayesian transfer learning, and the correlations can be inspected to inform which assets share information for which effect (i.e., parameter). Successes in both case studies demonstrate the wide applicability in practical infrastructure monitoring, since the approach is naturally adapted between interpretable fleet models of different in situ examples.
We present the first steps towards real-time distributed collaborative prognostics enabled by an implementation of the Weibull Time To Event-Recurrent Neural Network (WTTE-RNN) algorithm. In our system, assets determine their time to failure (TTF) in real-time according to an asset-specific model that is obtained in collaboration with other similar assets in the asset fleet. The presented approach builds on the emergent field of similarity analysis in asset management, and extends it to distributed collaborative prognostics. We show how through collaboration between assets and distributed prognostics, competitive time to failure estimates can be obtained. 1
Despite increasing interest, real-time prognostics (failure prediction) is still not widespread in industry due to the difficulties of existing systems to adapt to the dynamic and heterogeneous properties of real asset fleets. In order to address this, we present an Industrial Multi Agent System for real-time distributed collaborative prognostics. Our system fulfils all six core properties of Advanced Multi Agent Systems: Distribution, Flexibility, Adaptability, Scalability, Leanness, and Resilience. Experimental examples of each are provided for the case of prognostics using the C-MAPPS engine degradation data set, and data from a fleet of industrial gas turbines. Prognostics are performed using the Weibull Time To Event -Recurrent Neural Network algorithm. Collaboration is achieved by sharing information between agents in the system. We conclude that distributed collaborative prognostics is especially pertinent for systems with presence of sensor faults, limited computing capabilities or significant fleet heterogeneity.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.