摘要
针对传统磨机负荷检测方法存在的测量精度低、性能不稳定等缺陷,建立一种基于筒体振动信号频谱特征提取的选择性极限学习机(ELM)集成方法.采用核主元分析(KPCA)提取振动信号频谱特征,避免输入信号维数过高引发维数灾难.在非线性频谱特征空间内选用学习速度快、泛化性好的ELM建立集成模型个体,有效克服了单一ELM个体模型存在的运行结果不稳定问题.基于遗传算法(GA)的子模型后续选择方法进一步排除部分劣势个体,构建泛化能力强的简约集成模型,降低计算复杂性.实验结果表明:该方法对于矿浆浓度、料球比、充填率磨机负荷参数具有较高的精度和稳定性.
Due to the low precision and unstable performance of the traditional measurements for the ball mill load, a selective extreme learning machine （ELM） ensemble model based on feature extraction of frequency spectrum from shell vibration signals was proposed. Kernel principal component analysis （KPCA） was used to extract the spectrum features of the shell vibration signals with high dimensions and colinearity in order to overcome the dimensional disaster. In the feature space of the frequency spectrum, ELM algorithm was inserted into the selective ensemble frame as a compoment model, since ELM runs much faster and provides better generalization performance than the other popular learning algorithm, which may overcome variations in different trials of simulation for a single ELM model. Selective ensemble based on GA algorithm luther excludes the bad ELM components from all the available ensembles. The concise resemble model has strong generalization capacity and low computation load. Experimental results show high stability and accuracy of the proposed method in terms of the mineral to ball volume ratio, pulp density and charge volume ratio in a ball mill.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2011年第12期2088-2092,共5页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(61020106003
60874057)
中国博士后科学基金资助项目(20100471464)
关键词
磨机负荷
特征提取
极限学习机
集成模型.
mill load
features extraction
extreme learning machine
ensemble model
