基于奇异值分解极限学习机的维修等级决策
发布时间：2018-08-26 16:01
【摘要】：为降低航空发动机维修成本,增强维修等级决策的客观性,提出一种基于奇异值分解的极限学习机(SVD-ELM)算法,推导基于奇异值分解(SVD)的极限学习机(ELM)输出权重计算公式,从而有效地避免普通ELM在求解输出权重时因矩阵奇异而导致无法求逆的问题。将SVD-ELM应用于决策建模过程,提高决策模型的稳定性。研究结果表明:相比于SVM,SVD-ELM和ELM的决策准确率相同,且均比SVM的高,但SVD-ELM的模型稳定性高于ELM,且SVD-ELM和ELM的测试耗时相差不大,说明这2种方法的计算量相当。
[Abstract]:In order to reduce the maintenance cost of aero-engine and enhance the objectivity of maintenance grade decision, an algorithm of extreme learning machine (SVD-ELM) based on singular value decomposition (SVD) is proposed. The formula for calculating the output weight of (ELM) based on singular value decomposition (SVD) is derived. So it can avoid the problem that ordinary ELM can not solve the problem of matrix singularity in solving the output weight. SVD-ELM is applied to the decision-making modeling process to improve the stability of the decision model. The results show that compared with SVM,SVD-ELM and ELM, the accuracy of decision is the same and higher than that of SVM, but the model stability of SVD-ELM is higher than that of ELM, and the test time of SVD-ELM and ELM is not different, which indicates that the computation of the two methods is equal.
【作者单位】： 湖南大学信息科学与工程学院;湖南工学院数理科学与能源工程学院;
【基金】：国家自然科学基金资助项目(71501068)~~
【分类号】：TP181;V263.6
本文编号：2205418
[Abstract]:In order to reduce the maintenance cost of aero-engine and enhance the objectivity of maintenance grade decision, an algorithm of extreme learning machine (SVD-ELM) based on singular value decomposition (SVD) is proposed. The formula for calculating the output weight of (ELM) based on singular value decomposition (SVD) is derived. So it can avoid the problem that ordinary ELM can not solve the problem of matrix singularity in solving the output weight. SVD-ELM is applied to the decision-making modeling process to improve the stability of the decision model. The results show that compared with SVM,SVD-ELM and ELM, the accuracy of decision is the same and higher than that of SVM, but the model stability of SVD-ELM is higher than that of ELM, and the test time of SVD-ELM and ELM is not different, which indicates that the computation of the two methods is equal.
【作者单位】： 湖南大学信息科学与工程学院;湖南工学院数理科学与能源工程学院;
【基金】：国家自然科学基金资助项目(71501068)~~
【分类号】：TP181;V263.6
【相似文献】
相关期刊论文 前1条
1 张景瑞;陈立群;;基于奇异值分解的SGCMG操纵律分析[J];应用数学和力学;2008年08期
相关会议论文 前1条
1 张景瑞;;基于奇异值分解的SGCMGs输出误差分析及操纵律设计[A];第三届全国动力学与控制青年学者研讨会论文摘要集[C];2009年
，本文编号：2205418
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2205418.html
