基于奇异值分解极限学习机的维修等级决策

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.%In order to reduce the cost of aviation engine maintenance and enhance the objectivity of maintenance level decision,singular value decomposition based extreme learning machine (SVD-ELM) algorithm was proposed.The output weight formula of extreme learning machine (ELM) was deduced based on singular value decomposition (SVD).Unlike conventional ELM,SVD-ELM effectively avoids the failure of calculating matrix inversion due to matrix singular,during the process of computing output weight.Then SVD-ELM was applied in decision modeling process,which increased decision model stability.The results show that compared with SVM,the decision accuracy of SVD-ELM is the same as ELM,which are both higher than that of SVM.But SVD-ELM stability is greater than ELM.Meanwhile,testing time of SVD-ELM and ELM is similar,indicating that these two methods have the same computing amount.