通过贝叶斯自适应实验设计方法实现工业注塑工艺的单目标与多目标实时优化
Single and Multi-Objective Real-Time Optimisation of an Industrial Injection Moulding Process via a Bayesian Adaptive Design of Experiment Approach
AI总结:
本文提出基于贝叶斯优化的 ADoE 实时优化注塑质量与循环时间,并结合模内传感器表征收缩,实验数量较传统方法最多减少近50%。
摘要:
在不引发质量缺陷的情况下最小化循环时间是注塑(IM)中的一项重大挑战。实验设计方法(DoE)已被广泛研究用于注塑优化,但现有方法存在局限,包括需要大量实验以及预先确定搜索空间。贝叶斯自适应实验设计(ADoE)是一个迭代过程,其中先前实验的结果被用于为下一个设计作出有依据的选择。在本研究中,首次在注塑成型中开发了一种基于贝叶斯优化的实验性 ADoE 方法,利用工艺和传感器数据实时优化质量与循环时间。研究还提出了一种用于实时表征生产后收缩的新方法,该方法利用制件冷却期间关于温差的模内传感器数据。该表征方法通过生产后计量结果得到验证。本文提出了对注塑制件循环时间和温差的单目标与多目标优化。多目标优化技术,即复合期望函数和使用响应面方法论(RSM)模型的非支配排序遗传算法 II(NSGA-II),与新型实时 ADoE 方法进行了比较。与复合期望函数和 NSGA-II 相比,ADoE 在温差单目标优化所需实验数量上实现了近 50% 的减少,在温差和循环时间联合优化上实现了近 30% 的减少。此外,ADoE 为多目标优化识别出的最优设置与 NSGA-II 找到的所选 Pareto 最优解相似。
英文摘要:
Minimising cycle time without inducing quality defects is a major challenge in the injection moulding (IM). Design of Experiment methods (DoE) have been widely studied for optimisation of the IM, however existing methods have limitations, including the need for a large number of experiments and a pre-determined search space. Bayesian adaptive design of experiment (ADoE) is an iterative process where the results of the previous experiments are used to make an informed selection for the next design. In this study, for the first time, an experimental ADoE approach, based on Bayesian optimisation, was developed in injection moulding using process and sensor data to optimise the quality and cycle time in real-time. A novel approach for the real-time characterisation of post-production shrinkage was introduced, utilising in-mould sensor data on temperature differential during part cooling. This characterisation approach was verified by post-production metrology results. A single and multi-objective optimisation of the cycle time and temperature differential in an injection moulded component is proposed. The multi-objective optimisation techniques, composite desirability function and Nondominated Sorting Genetic Algorithm (NSGA-II) using Response Surface Methodology (RSM) model, are compared with the real-time novel ADoE approach. ADoE achieved almost a 50% reduction in the number of experiments required for the single optimisation of temperature differential, and an almost 30% decrease for the optimisation of temperature differential and cycle time together compared to composite desirability function and NSGA-II. Also, the optimal settings identified by ADoE for multiobjective optimisation were similar to the selected Pareto optimal solution found by the NSGA-II.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。