Abstract
This paper presents an application of the Limited Memory Influence Diagrams (LIMID) for decision-making in Structural Health Monitoring (SHM). Essentially, information from sensors or algorithms provides no value unless a decision strategy is associated. The Bayes detector is the decision function that optimizes the expected utility, and the Value of Information (VoI) is the increase in expected utility, obtained by performing the sensing. Where the detector is a threshold for a continuous variable, it is a decision policy for discrete information. For SHM with multiple sequential decisions, the VoI is found for the set of policies that lead to the highest expected utility. The policies can be found up-front by solving Influence Diagrams, which are Bayesian nets with decision and utility nodes. However, the multiple decisions of SHM make the solution of influence diagrams intractable. In this paper, we implement LIMIDs modelled on Dynamic Bayesian Nets to model the decision problem of SHM damage detection. The approach is intuitive, as it is based on graphical models, and is advantageous in calculation efficiency, as the policies are solved using the Single Policy Updating algorithm. LIMIDs of various complexities are investigated and the results are compared against exhaustive optimization using Monte Carlo simulation. The calculation cost is found to depend on the number of parents of each decision node and on the inter-slice dependencies, and the most efficient models are found to be of the Hidden Markov Model (HMM) type. The solutions are observed to conservatively approximate the simulation results, due to the simplifications of the approach. The LIMIDs are found suitable for SHM decision problems, but current solution algorithms need improvement before large decision problems can be solved on normal PCs.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.