Model-free approach and methodology for data anomaly detection for real time diagnostic solution

This paper describes an approach and methodology for data anomaly detection and fault isolation for a diagnostic solution of hot section aeroengine components. The essence of the methodology is the combination of qualitative data exceedence check and quantitative data analysis. Three data sets, namely reference, operational and test data and two significance test methods are proposed and tested using temperature data generated in the laboratory. Paired t-test and rank sum hypotheses test are found to be appropriate for different data set combinations. The algorithm and methodology work well with laboratory temperature data and seeded faults.