中国机械工程
中國機械工程
중국궤계공정
CHINA MECHANICAl ENGINEERING
2014年
16期
2225-2230,2231
,共7页
余治民%刘子建%艾彦迪%熊敏
餘治民%劉子建%艾彥迪%熊敏
여치민%류자건%애언적%웅민
Takagi-Sugeno型模糊推理%隶属度函数%模糊神经网络%鲁棒性%泛化能力
Takagi-Sugeno型模糊推理%隸屬度函數%模糊神經網絡%魯棒性%汎化能力
Takagi-Sugeno형모호추리%대속도함수%모호신경망락%로봉성%범화능력
Takagi-Sugeno type fuzzy inference%membership function%fuzzy neural network%ro-bustness%generalization ability
将基于神经模糊控制理论的建模方法---模糊神经网络建模法应用到数控机床热误差建模当中,讨论了热误差模糊神经网络的结构及建模原理;对大型数控龙门导轨磨床主轴箱系统进行建模试验,采用非接触式红外温度测量仪和千分表分别测量主轴箱系统温度值与主轴热误差,得到两组独立的试验数据,一组用来建立主轴箱系统热误差模糊神经网络预报模型,另一组用来对模型进行验证。试验结果表明,模糊神经网络模型预测精度高,泛化能力强;将模糊神经网络建模方法与径向基函数神经网络建模方法进行综合对比,分析结果表明,模糊神经网络建模方法具有更好的建模效率、建模鲁棒性及预测性能。
將基于神經模糊控製理論的建模方法---模糊神經網絡建模法應用到數控機床熱誤差建模噹中,討論瞭熱誤差模糊神經網絡的結構及建模原理;對大型數控龍門導軌磨床主軸箱繫統進行建模試驗,採用非接觸式紅外溫度測量儀和韆分錶分彆測量主軸箱繫統溫度值與主軸熱誤差,得到兩組獨立的試驗數據,一組用來建立主軸箱繫統熱誤差模糊神經網絡預報模型,另一組用來對模型進行驗證。試驗結果錶明,模糊神經網絡模型預測精度高,汎化能力彊;將模糊神經網絡建模方法與徑嚮基函數神經網絡建模方法進行綜閤對比,分析結果錶明,模糊神經網絡建模方法具有更好的建模效率、建模魯棒性及預測性能。
장기우신경모호공제이론적건모방법---모호신경망락건모법응용도수공궤상열오차건모당중,토론료열오차모호신경망락적결구급건모원리;대대형수공룡문도궤마상주축상계통진행건모시험,채용비접촉식홍외온도측량의화천분표분별측량주축상계통온도치여주축열오차,득도량조독립적시험수거,일조용래건립주축상계통열오차모호신경망락예보모형,령일조용래대모형진행험증。시험결과표명,모호신경망락모형예측정도고,범화능력강;장모호신경망락건모방법여경향기함수신경망락건모방법진행종합대비,분석결과표명,모호신경망락건모방법구유경호적건모효솔、건모로봉성급예측성능。
A modeling method of fuzzy neural network based on neural fuzzy control theory was applied in thermal error modeling of NC machine tool and the structure and modeling principle of fuzz-y neural network on thermal errors was discussed.The experiment on spindle head of large NC gantry guide grinder was conducted and two independent sets of experimental data were obtained by measur-ing the temperature of spindle head and the thermal error of spindle with non-contact infrared ther-mometer and dial gauge,one was used to establish thermal error fuzzy neural network prediction model of spindle head and the other was used to validate the model.The test results show that fuzzy neural network model has high prediction accuracy and good generalization.Compared with radial basis gunction(RBF)neural network modeling method,fuzzy neural network modeling method has better modeling efficiency,robustness and predict performance.