Process Optimization of Solid Freeform Fabrication Using Experimentation, Simulation, and Adaptive Statistical Models

这项研究的目标是创建一种使用统计模型优化制造过程的方法。这些模型将从不同来源获得的知识创建，包括启发式、专家知识、物理实验、第一性原理关系和物理模型的大规模计算机模拟。优化制造工艺需要了解设计和工艺变量对所制造人工制品的最终质量的影响和相互作用-变量是所选材料(或材料组合)的特性、工艺的物理特性、零件的几何形状、设备设置和制造环境条件。流程建模框架必须是灵活和可适应的，以便子流程的模型可以随着流程的发展和新知识的获取而更新和调整。这项研究的重点是一种新的制造工艺，称为实体自由形状制造(SFF)或分层制造，它正在改变传统的机械零件设计和制造方法。SFF制造过程的更好的过程模型将导致快速制造过程的更快发展，并将加速它们的使用和被行业接受。好的模型还可以更好地识别设计可能性，以便将制造方面的考虑纳入设计周期的早期阶段。为设计周期的不同阶段创建和优化具有不同详细程度的工艺模型的方法将消除设计和制造问题整合的障碍。这项工作还将有助于统计模型的建立，使用户能够纳入对设计变量的主动和关键约束，有效地处理高维和不规则的参数空间，并评估模型的可靠性、预测能力和参数敏感性。

The objective of this research is to create a methodology for optimizing manufacturing processes using statistical models. These models will be created from knowledge acquired from different sources, including heuristics, expert knowledge, physical experimentation, first principle relationships and large-scale computer simulations of physical models. Optimizing a manufacturing process requires an understanding of the influence and interactions of design and process variables on the final quality of the manufactured artifact - Variables are properties of the material (or combinations of materials) selected, of the physics of the process, of the geometry of the part, of the equipment settings, and of the manufacturing environmental conditions. The process modeling framework must be flexible and adaptable, so that models for subprocesses can be updated and adapted as processes evolve and new knowledge is acquired. This research focuses on a new class of manufacturing processes called Solid Freeform Fabrication (SFF) or layered manufacturing, which are changing traditional approaches to design and manufacture of mechanical parts. Better process models of SFF manufacturing processes will lead to faster development of rapid manufacturing processes and will accelerate their use and acceptance by industry. Good models will also allow better identification of design possibilities so that manufacturing considerations can be brought into the early phases of the design cycle. A methodology for creating and optimizing process models with differing levels of detail for different phases of the design cycle will remove impediments to the integration of design and manufacturing concerns. This work will also contribute to statistical model building by enabling users to incorporate the active and critical constraints on design variables, to handle high-dimensional and irregular parameter spaces efficiently, and to assess model reliability, predictive capability and parameter sensitivity.