CAJ | 학술논문

目前,国内许多钢铁生产企业为了提高无缝钢管的生产质量,都采用了加装狄赛尔导盘的斜轧穿孔机进行管坯穿孔.但足在斜轧穿孔过程中,由于客观条件限制,导盘转速无法通过直接测量的手段在线精确测得.通过软测量实现穿孔机导盘转速的在线精确测量,进而有效控制其转速,提高无缝钢管生产机组的生产效率,是解决问题的一个有效方法.针对导盘转速软测量建模中存在的问题,本文对极限学习机(ELM)方法提出了改进.一方面将极限学习机方法同主成分分析(PCA)方法相结合,通过对模型输入量进行主成分分析,去除了各变量之间的线性相关,最终提高了极限学习机算法的泛化性能;另一方面对极限学习机隐层节点进行分类,通过分类减少了极限学习机方法的训练时间和预报时间.最终得出了基于改进PCA-ELM方法的导盘转速软测量模型,并通过仿真实验证了其准确性和可应用性.
목전,국내허다강철생산기업위료제고무봉강관적생산질량,도채용료가장적새이도반적사알천공궤진행관배천공.단족재사알천공과정중,유우객관조건한제,도반전속무법통과직접측량적수단재선정학측득.통과연측량실현천공궤도반전속적재선정학측량,진이유효공제기전속,제고무봉강관생산궤조적생산효솔,시해결문제적일개유효방법.침대도반전속연측량건모중존재적문제,본문대겁한학습궤(ELM)방법제출료개진.일방면장겁한학습궤방법동주성분분석(PCA)방법상결합,통과대모형수입량진행주성분분석,거제료각변량지간적선성상관,최종제고료겁한학습궤산법적범화성능;령일방면대겁한학습궤은층절점진행분류,통과분류감소료겁한학습궤방법적훈련시간화예보시간.최종득출료기우개진PCA-ELM방법적도반전속연측량모형,병통과방진실험증료기준학성화가응용성.
Recently, to improve the quality of seamless steel tubes, many domestic billet enterprises use skew rolling piercer with Diescher guide-disc to pierce steel billets. However, in skew rolling piercing, the guide-disc speed can't be measured directly because of the limitations in practical conditions. Therefore, an online soft measurement of the guide-disc speed has to be developed for controlling the guide-disc speed to improve the productivity and quality of steel tubes. To cope with this problem, we propose a measurement model based on the extreme learning machine(ELM) method with improvements made by us in this paper. This improved ELM method is combined with the principal component analysis(PCA) to let the input values of ELM be analyzed by PCA for improving the generalization performance. On the other hand, the training and predicting time of ELM is reduced by classifying the hidden layer nodes. Thus, a model based on the improved PCA-ELM method is established. Simulation results verify that this model is effective and applicable.