Damage Detection of Mechanical Components Using Null Subspace Analysis

This paper presents two original applications of the Null Subspace Analysis (NSA) method for fault diagnosis in mechanical components. The method is first applied to the case-study of electro-mechanical devices at the end of the assembly line with the aim of assessing their overall quality. The advantages of the proposed method rely in its rapidity of use and its reliability. At first, a set of five good (i.e. healthy) devices and four damaged devices was considered. The components were instrumented with one triaxial accelerometer on the flank and one monoaxial accelerometer on the top. Based on the NSA method, a mapping of the space [ active components, system order ] up to a system order of 100, was realized in order to select the appropriate order and number of active components. Eventually, thanks to this mapping, the method was able to successfully detect all the faulty components using the signal from only one accelerometer in one direction. The second application is related to the quality assessment of welded joints between stripes in a steel processing plan. Six welded joints with nominal welding parameters and twenty-seven welded joints with out-of-range parameters were realized. Again, the NSA method was able to diagnose successfully the welded joints using a single signal from one accelerometer.