A Fuzzy Similarity-Based Method for Failure Detection and Recovery Time Estimation

In this work, an extension of a data-driven approach for estimation of the available Recovery Time (RT) is presented. The improvement is in avoiding the need of resorting to a fault detection module for the identification of the anomalous system behavior: the algorithm proposed jointly detects the onset of the accidental transient and performs the estimation of the available RT. This is achieved by fuzzy similarity analysis of the currently developing scenario and reference multidimensional trajectory patterns of failure scenarios; the RT remaining before the developing trajectory pattern hits a failure threshold is predicted by combining the times of failure of the reference patterns, weighed by their similarity with the developing pattern. For illustration purposes, failure scenarios of the Lead Bismuth Eutectic eXperimental Accelerator Driven System (LBE-XADS) are considered.