tion results, as the simulations took longer than preparing and evaluating them. Today, computers wait for the opera- tor’s inputs, as simulations crank out results faster than the operator can evaluate them and set up modified runs. To gain the most advantage out of these computer
advances, casting process modeling now has been combined with autonomous optimization tools, which produce simula- tions based on a range of parameters, rather than specific design points. For example, instead of setting up 10 simula- tions and evaluating 10 sets of results, only one simulation setup is required to yield the optimum results. Te one-dimensional search is characterized by a limited number of trials and can result in a dead end. Proposed solutions usually are not tested in regard to their limita- tions. Te one-dimensional search faces another drawback: the ability to transfer gained knowledge to new projects is limited with increasing casting complexity (Fig. 1, left). DOE methods long have been used in the optimization of manufacturing processes, where they are configured and then evaluated using statistical methods. In a virtual DOE, real experiments are replaced by a number of simulation runs (Fig. 1, center). Te goal is to evaluate the impact of each process
parameter on the casting process in order to predict its behavior at any point of the process window. Te DOE then is developed based on statistical patterns, after which the simulations are started. After all the simulations have finished, the operator can evaluate the results according to his or her objectives. When autonomous optimization is utilized (Fig. 1, right),
the operator defines not only the degrees of freedom for the process parameters, but also the objectives, before the simulations are performed. Following the calculation of each simulation generation, the program automatically evaluates the results according to the objectives defined by the opera- tor. Ten, dependent on the simulation results and the cho- sen objectives, a genetic algorithm creates new variations of the casting gating and riser layout. Te procedure follows the rules of evolution: each layout variation is kept, eliminated, modified or combined with an already calculated or new design. Tis process is repeated until design modifications do not lead to additional improvements. Just as in the biological world, the evolutionary process
过去，铸造工作者不得不耐心等待模拟结果，模拟需要 的时间比准备和分析时间要长。今天，计算机要等操作 者的输入，模拟出结果比操作者分析结果和改进模拟方 案要快。
要利用计算速度提高带来的好处，铸造工艺模拟仿 真现在已经和自动优化工具结合在了一起，这样在一个 范围内变化的多工艺参数上进行仿真，而不是一个特 定工艺。例如，不用进行10次模拟然后评价10次的模 拟结果来进行优化，而通过一次模拟计算就得到了优化 结果。
一维搜索的特点是进行有限次的试验达到目的。考 虑到该方法的次数限制，不能试验提出的各种方案。一 维搜索还有另外一个缺点，即铸件复杂程度提高后（如 图1左图所示）将已有知识应用到新项目中的能力受到 限制。试验设计方法很早就应用在制造过程的优化中， 采用统计学的方法进行描述和分析。在虚拟的试验设计 中，数值模拟取代了试验（如图1中间图）。 优化的目标是评价各工艺参数对铸造工艺的影响， 以便预测在工艺窗口中任意点的行为。试验设计基于统 计学，其背后是基于数值模拟。当所有的模拟完成之 后，操作者就能根据自己的目的对结果进行分析评价。 当采用自动优化时（如图1，右图），模拟开始 前，操作者不仅定义了各工艺参数的自由度，也定义 了目标。接着进行每次的模拟计算，程序根据操作者 的目标自动对结果进行评价。然后，根据模拟结果和 设定目标，遗传算法设计出浇注系统和冒口布置的各 种方案。分析过程遵循的分析规律是：保持每次的工 艺方案、删除、修改、和已有计算的方案或新方案合
Fig. 2. Autonomous optimization uses the above generic optimiza- tion algorithm. Just as in the biological world, the evolutionary process of autonomous optimiza- tion occurs over several calculation generations.
圖。 2。自主優化使用 上述通用優化算法。正 如在世界上的生物，在 進化過程中發生了幾自 治優化計算代。
Summer 2011
FOUNDRY-PLANET.COM | MODERN CASTING | CHINA FOUNDRY ASSOCIATION | 59
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75