Methodology for modeling and diagnosis of compliant structure assemblies.

摘要

All manufacturing processes have unavoidable variations due to the inherent randomness of the processes themselves. However, if assignable causes are present, excessive variations will result. Lack of knowledge and understanding regarding product/process variations often result in expensive quality problems. Especially, because of ineffective diagnostic approaches, root cause determination is the weakest step in the whole cycle of quality control efforts in manufacturing systems.;Since most current diagnostic approaches in compliant structure assembly processes, such as the automotive body assembly, are based on rigid body assumption, they are not able to model and predict effects of variations due to interferences and deformations of parts or subassemblies. Therefore, a modeling and diagnostic methodology for assembly variations considering part compliant characteristics is of great importance for developing more effective diagnostic systems in compliant assembly processes.;In this research, a methodology is proposed for dimensional fault detection and isolation in compliant structure assemblies. This methodology incorporates engineering information such as specifications of fixtures, parts joint functions, and process information such as sensor locations and in-process measurements. Diagnostic vectors are introduced based on the beam-based modeling technique which determine the nonlinear spatial responses of compliant assembly corresponding to each individual dimensional fault defined in the fault domain. More rigorous diagnostic performance can be accomplished based on these diagnostic vectors.;A diagnostic approach for multiple fault detection in compliant assemblies is developed for the first time. In this research, multiple fault diagnostics is classified into three diagnosability levels: fully diagnosable systems, conditionally diagnosable systems, and undiagnosable systems. The diagnostic capability for multiple faults in compliant assemblies is studied, the impacts of several key process variables, such as signal-to-noise ratio, sample size, number of sensor, and sensor placement are analyzed for the conditionally diagnosable systems.;Fault isolationability for compliant sheet metal assembly is studied for the first time. An adjusted least square approach is developed based on Singular Value Decomposition (SVD), which leads to the new method for analysis of diagnosability of ill-conditioned assembly systems. Diagnostic performance is significantly improved based on the results of OC curves. A new two-step diagnostic model is developed as well, which allows for precise isolation of collinear faults in ill-conditioned compliant assemblies.