Adjusted Least Squares Approach for Diagnosis of Ill-Conditioned Compliant Assemblies

Abstract

Least squares (LS) estimation has been extensively used for parameter identification and model-based diagnosis. However, if ill-conditioning is present, the LS estimation approach tends to generate imprecise results and thus impacts the diagnostic performance. In this paper, an adjusted least squares approach is proposed to deal with the ill-conditioning problem in the diagnosis of compliant sheet metal assembly process. The adjusted LS approach is able to overcome the ill-conditioning and give precise results for certain linear combinations of the faults. Simulations and industrial case study are conducted to compare the diagnostic performance of the adjusted and regular LS approach. In addition, a two-stage assembly model is developed for further fault isolation with inclusion of additional measurement information.