Title: Fault Failure Diagnosis for Compliant Structure Using Statistical Modal Analysis (SMA)

Sponsor: NSF, NIST-ATP

Collaborators: R. Kumar, J. Kong, Y. Zhou, S. Gogoneni (Dimensional Control Systems)

This paper develops a new diagnostics methodology for compliant assemblies with N-2-1 fixture layout using statistical modal analysis (SMA). The developed methodology is based on: (i) the CAD-based dimensional fault models; (ii) data-based dimensional variation fault models; and (iii) the fault mapping procedure isolating the unknown faults.  The CAD-based faults model are developed using piece-wise linear classification of rigid and non-rigid areas of the compliant part. The CAD-based fault models allow developing the fault library for the failures of the N-2-1 fixtures in a multi-station assembly system.  The proposed approach allow to model all piece-wise linear modes of part deformation using “linear bending modes”. Data-based dimensional variation fault models are based on the statistical model analysis (SMA) which allow to model part deformation with varying number of deformation modes. SMA uses sampling strategy of DCT (discrete cosine transformation) which decomposes error into series of variation fault/error modes. It was proved that these independent variation fault/error modes are equivalent to the CAD-based faults models obtained in (i). The fault mapping procedure allows to diagnose the unknown fault by comparing the unknown fault variation pattern obtained from the SMA model with one of the predetermined CAD-based fault patterns.  The proposed methodology considers part geometry, fixture layout, and sensors positions that help in detecting fault. A case study from an automotive roof framing station assembly has been conducted to validate the efficacy of proposed method.   

 Flow chart for Model-Based Diagnosis

Fixture locators failure models using various part deformation modes