A Data Driven Method for Quantitative Fault Diagnosability Evaluation

Knowledge of achievable diagnosability performance can provide theoretical guidance for developing diagnostic algorithms and optimizing sensor placement. A data driven approach for fault diagnosability quantitative evaluation without designing any diagnosis algorithm is proposed. Fault diagnosability is converted to similarity measure of output information which is denoted by fuzzy sets under different fault states. A quantitative evaluation measure and a specific diagnosability evaluation process are given. Finally, an example is presented to demonstrate the effectiveness of the proposed methodology.