首页> 外国专利> PREDICTION DEVICE FOR OPTIMUM PARAMETER COMBINATION

PREDICTION DEVICE FOR OPTIMUM PARAMETER COMBINATION

机译:最佳参数组合的预测装置

摘要

PROBLEM TO BE SOLVED: To predict an optimum parameter through a small number of operation trials by converting a parameter into an internal parameter (factor) of the same level number and performing simultaneously a factor analysis based on an orthogonal plan and the searching based on a genetic algorithm. SOLUTION: A parameter conversion part 1 stores the correspondence between an original parameter and an internal parameter and performs the conversion between them. The factor level is set at 2, and an orthogonal plan is produced at a level of 2 and with the intensity of 2. Then a factor combination is produced and evaluated at a combination evaluation part 4. Based on the evaluation value of the factor combination, the average and deviation of the evaluation value are calculated at every level. The presence or absence is decided for a factor that contributes to the improvement of the evaluation value based on the selection reference. If the relevant factor is searched, its level is decided. The processing is repeated at and after the production of the orthogonal plan for other factors. If the relevant factor is not searched, the factor contribution is defined ambiguous and a combination searching part 3 searches the factor. If no remaining factor is searched, the decided factor combination is outputted to a user interface part 5.
机译:解决的问题:通过将参数转换为相同级别编号的内部参数(因子)并同时执行基于正交计划的因子分析和基于对象的搜索来通过少量操作试验来预测最佳参数遗传算法。解决方案:参数转换部分1存储原始参数和内部参数之间的对应关系,并在它们之间进行转换。将因子级别设置为2,并以2的强度和2的强度生成正交计划。然后,在因子组合评估部分4中生成因子组合并进行评估。 ,在每个级别计算评估值的平均值和偏差。根据选择基准,决定是否有有助于评价值的提高的因素。如果搜索到相关因素,则确定其级别。对于其他因素,在正交图生成时和生成后重复该处理。如果没有搜索到相关因素,则将因素贡献定义为不明确,并且组合搜索部分3搜索该因素。如果未搜索到剩余因子,则将确定的因子组合输出到用户界面部分5。

著录项

  • 公开/公告号JPH10149345A

    专利类型

  • 公开/公告日1998-06-02

    原文格式PDF

  • 申请/专利权人 MITSUBISHI ELECTRIC CORP;

    申请/专利号JP19960309360

  • 发明设计人 TANAKA HIDETOSHI;

    申请日1996-11-20

  • 分类号G06F15/18;G06F17/00;

  • 国家 JP

  • 入库时间 2022-08-22 03:04:45

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号