| Dimensional quality plays an important role in the functionality and performance of many products in different industrial sectors. Dimensional variation, which negatively affects the dimensional quality, is inherent to any assembly process. Therefore, it is important to understand how it propagates through the process. Unfortunately, assembly processes are complex and in many cases highly non-linear, especially the widely occurred assembly of compliant components. Since parts and fixtures inherently have geometrical variation, understanding how these variations propagate through the system is of significant interest to the design and control of such systems. Although uncertainty propagation has been extensively investigated in many engineering fields, the lack of efficient non-linear modeling tools has limited current analysis of processes to simplified linear models. In order to extend the capabilities of current models, it is necessary to create new methods that predict geometrical variation propagation by taking into account the non-linear effects due to the contact between the components and tools in the physical assembly process.;The main objective of this research is to first analyze the phenomena for the dimensional variation in non-linear compliant sheet metal assembly systems, then achieve the goal of predicting how the dimensional variation propagates and accumulates through the assembly based on the results of the analysis. Finally, the prediction technique will be used and combined with other tools to control and reduce the dimensional variation propagation and accumulation through the assembly. In order to fulfill this objective, the industrial background that relates to the research domain is first introduced, and then the review of recent research developments has been presented. Two new analysis methodologies which could achieve the prediction for the variation propagation are proposed, with the consideration of contact and welding interactions respectively. Some new uncertainty methods have been introduced to serve the purpose of accurately and efficiently predicting the statistical response of the assembly to variation on the input parameters for these two analysis methodologies. Based on the achievement made above, a systematical approach that combines an off-line error control-learning module using virtual assembly models and an in-line control implementation using a feedforward control strategy is proposed to realize dimensional-related error compensation in compliant sheet metal assembly processes. Finally, conclusions are made for the contributions made by this dissertation and future works are suggested. |
