Sensitivity and uncertainty in variation simulation models for multi-stage manufacturing systems

Variation simulation models have been widely used in product design and manufacturing system development to reduce the dimensional variation of the final products. Variation reduction can be effectively achieved through a robust manufacturing system which is less sensitive to input variation. In order to evaluate and decrease the sensitivity of a system to its input variation, an appropriate set of metrics must be defined based on the variation simulation models. However, no such metrics have been systematically defined especially for multi-stage compliant assembly systems.;In addition, since no simulation models can completely capture all the characteristics of the simulated physical systems, the models always include uncertainty. The uncertainty strongly impacts the fidelity of the simulation output and consequently the applicability of simulation models. It is especially true for variation simulation models of multi-stage manufacturing systems in that uncertainty can propagate and accumulate from stage to stage. Unfortunately, no research has been conducted to analyze the uncertainty in multi-stage variation simulation models and to illustrate the effects of the uncertainty on the applications of the models.;To address these deficiencies, this dissertation studies the sensitivity and uncertainty in variation simulation models. First, the sensitivity indices for pattern, component and station are defined. These indices are critical for sensitivity analysis and robust design of multi-stage manufacturing processes. Additionally, the relationships among these sensitivity indices are established.;Second, an uncertainty model based on multi-stage variation simulation models is developed. From this uncertainty model, conclusions about uncertainty propagation and accumulation in the variation models are made, and guidelines for calibration of the models are also established. Moreover, the sources and the characteristics of the uncertainty in multi-stage variation simulation models are analyzed.;Third, the uncertainty model is applied to tolerance allocation to illustrate how uncertainty impacts the applications of variation simulation models. An original optimization formation is proposed and the impact is discussed by comparing the proposed formulation with traditional ones.;Fourth, an approach is developed for mitigating the uncertainty in compliant variation simulation models as induced by inaccurate part source shape representation. This approach includes an algorithm to generate basic shapes and implementation of genetic algorithm (GA) to identify a suitable set of nodes in the FEM model of a part as the inputs for the variation models.;Finally, the future directions of this research are suggested.