Abstract:In the context of modern engineering, environmental, and societal concerns, there is an increasing demand for methods able to identify rational management strategies for civil engineering systems, minimizing structural failure risks while optimally planning inspection and maintenance (I&M) processes. Most available methods simplify the I&M decision problem to the component level, often assuming statistical, structural, or cost independence among components, due to the computational complexity associated with global optimization methodologies under joint system-level state descriptions. In this paper, we propose an efficient algorithmic framework for inference and decision-making under uncertainty for engineering systems exposed to deteriorating environments, providing optimal management strategies directly at the system level. In our approach, the decision problem is formulated as a factored partially observable Markov decision process, whose dynamics are encoded in Bayesian network conditional structures. The methodology can handle environments under equal or general, unequal deterioration correlations among components, through Gaussian hierarchical structures and dynamic Bayesian networks, decoupling the originally joint system state space to component networks conditional on shared random variables. In terms of policy optimization, we adopt a deep decentralized multi-agent actor-critic (DDMAC) reinforcement learning approach, in which the policies are approximated by actor neural networks guided by a critic network. By including deterioration dependence in the simulated environment, and by formulating the cost model at the system level, DDMAC policies intrinsically consider the underlying system-effects. This is demonstrated through numerical experiments conducted for both a 9-out-of-10 system and a steel frame under fatigue deterioration. Results demonstrate that DDMAC policies offer substantial benefits when compared to state-of-the-art heuristic approaches. The inherent consideration of system-effects by DDMAC strategies is also interpreted based on the learned policies.

An Inference Algorithm for Probabilistic Fault Management in Distributed Systems

Backward Inference in Bayesian Networks for Distributed Systems Management

Probabilistic Inference for Network Management

Predictive Fault Management in the Dynamic Environment of Ip Networks

Community-base Fault Diagnosis Using Incremental Belief Revision

Research on rapid diagnosis algorithm for complex system based on Bayesian theory

Model And Fault Inference With The Framework Of Probabilistic Sdg

Probabilistic Inference Strategy in Distributed Intrusion Detection Systems

THE BAYESIAN NETWORKS METHOD FOR NETWORK FAULT MANAGEMENT

Probabilistic SDG Model and Approach to Inference for Fault Analysis

Block-Sparse Bayesian Learning Method for Fault Location in Active Distribution Networks with Limited Synchronized Measurements

Probabilistic SDG model description and fault inference for large-scale complex systems

An adaptive decision method using structure feature analysis on dynamic fault propagation model

Probabilistic approaches to fault detection in networked discrete event systems

Construction of Simulation for Probabilistic Inference in Uncertain and Dynamic Networks Based on Bayesian Networks

Quantification of Highly Coupled Dynamic Fault Tree Using IRVPM and SBDD.

Inference and dynamic decision-making for deteriorating systems with probabilistic dependencies through Bayesian networks and deep reinforcement learning

Distributionally Robust Fault Detection Design and Assessment for Dynamical Systems.

A Novel Probabilistic Fault Detection Scheme with Adjustable Reliability Estimates

Distributed Fault Detection for Large-Scale Systems Based on Adaptive Approximation

Distributed Fault Estimation and Fault-Tolerant Control of Interconnected Systems with Plug-and-Play Features