Collaborative Research: Applying Bayesian Predictive Modeling and Decision Theory to Milling Profit Optimization under Uncertainty

本合作研究项目的研究目标是为不确定条件下最优铣削参数的选择建立一种新的范式。研究计划包括两个基本组成部分：1)建立贝叶斯预测模型；2)实现利润最大化的决策框架。它最终将产生两个重大结果。首先，将进行验证测试，将刀具制造商推荐的铣削参数与新的优化结果进行生产成本比较。其次，将开发一个软件平台，指导用户通过新方法不仅确定利润最大化的参数，而且还为新数据收集选择最佳实验。如果成功，这种将贝叶斯预测建模和决策理论与加工建模能力相结合的新方法，将从确定性的、基于模型的铣削参数选择转向一种更现实的方法，该方法将模型预测中固有的不确定性纳入其中。这将通过以下方式为最佳铣削参数选择提供新的见解：1)将铣削表述为不确定条件下的决策问题；2)为铣削试验值的计算提供规范依据；3)基于实验结果，提供铣削性能不确定性实时更新算法；4)在软件中实现所导出的算法；5)进行铣削试验，验证规范贝叶斯更新方法与实验结果的一致性。通过在不确定的情况下实现离散零件铣削生产的利润最大化，本研究将对国家产生积极影响。美国的经济和国防能力。

The research objective of this collaborative research project is to establish a new paradigm for the selection of optimal milling parameters under uncertainty. The research plan includes two fundamental components: 1) develop a Bayesian predictive model; and 2) implement a decision making framework for maximizing profit. It will culminate in two significant outcomes. First, validation tests will be performed that compare production costs using cutting tool manufacturer-based recommendations for milling parameters to the new optimized result. Second, a software platform will be developed that guides users through the new approach to not only determine parameters for maximized profit, but also the optimal selection of experiments for new data collection. If successful, the new approach, which combines Bayesian predictive modeling and decision theory with machining modeling capabilities, will provide a fundamental departure from deterministic, model-based selection of milling parameters to a more realistic approach that incorporates the inherent uncertainty in model predictions. This will lead to new insights into optimal milling parameter selection by: 1) formulating milling as a decision problem under uncertainty; 2) providing the normative bases for calculation of value of experimentation in milling maximize expected utility; 3) providing algorithms for the real-time updating of milling performance uncertainties based on experimental results; 4) implementing the derived algorithms in software; and 5) conducting milling tests to verify the consistency of the normative Bayesian updating approach with experimental results. By enabling maximized profit under uncertainty for discrete part production by milling, this research will positively influence the nation?s economy and defense capabilities.