CAJ | 학술논문

研究液压泵特征参数的在线预测问题。对表征液压泵工作状态的特征参数进行准确、快速的在线预测,对实时掌握液压泵健康状况的发展趋势具有重要意义。针对液压泵特征参数在线预测问题,提出一种在线核极限学习机方法( OL-KELM)。 OL-KELM采用Cholesky分解将核极限学习机( KELM)从离线模式扩展到在线模式,网络权值可在历史训练数据的基础上,随新样本的输入而递推求解更新,并以简单的四则运算替代复杂的矩阵求逆,从而提高网络的在线学习效率。仿真结果表明,在获得同样预测精度的条件下,OL-KELM比直接在线核极限学习机的计算速度更快,且预测误差小于贯序正则极限学习机,并具有更强的鲁棒性,故OL-KELM能够对液压泵特征参数进行快速准确的在线预测。
연구액압빙특정삼수적재선예측문제。대표정액압빙공작상태적특정삼수진행준학、쾌속적재선예측,대실시장악액압빙건강상황적발전추세구유중요의의。침대액압빙특정삼수재선예측문제,제출일충재선핵겁한학습궤방법( OL-KELM)。 OL-KELM채용Cholesky분해장핵겁한학습궤( KELM)종리선모식확전도재선모식,망락권치가재역사훈련수거적기출상,수신양본적수입이체추구해경신,병이간단적사칙운산체대복잡적구진구역,종이제고망락적재선학습효솔。방진결과표명,재획득동양예측정도적조건하,OL-KELM비직접재선핵겁한학습궤적계산속도경쾌,차예측오차소우관서정칙겁한학습궤,병구유경강적로봉성,고OL-KELM능구대액압빙특정삼수진행쾌속준학적재선예측。
Study hydraulic pump characteristic parameters online prediction. Conducting that fast and accurate on-line prediction of hydraulic pump characteristic parameters is very significant for controlling the state of health trends of hydraulic pump in real-time, to execute online prediction of hydraulic pump characteristic parameters, an online kernel extreme learning machine algorithm ( OL-KELM ) was proposed. The kernel extreme learning machine ( KELM) was extended from offline mode to online mode by Cholesky factorization in the OL-KELM. Compared with KELM, the OL-KELM can complete the training recursively on the basis of sequential input new samples. The com-putational efficiency was improved by replacing the matrix inverse operation with arithmetic in the OL-KELM. Simu-lation results show that the OL-KELM can achieve the same accuracy with KELM by saving lots of training time, and it has lower prediction error and stronger robustness than SRELM. So the OL-KELM can be used in the hydraulic pump characteristic parameters online prediction quickly and accurately.