摘要
利用一种称为传感器融合的新监测方法，集成来自各种传感器的信息，建立了一个基于神经网络的切屑类型监视的智能系统。当使用不同的切削用量时，系统的正确识别率达８４％，表明神经网络模型具有在较宽工作条件下识别复杂系统的优点。
Without satisfactory chip control,a machining operation cannot be successful.Becauseof the complexity,randomness and fuzziness of cutting processes,traditional monitoringmethods are unreliable,unrepeatable and with only a narrow range of application。The newstrategy of integrating information from a variety of sensors,termed sensor information fusion,is described An intelligent monitoring scheme based on neuraI networks for recogniz-ing chip typeshas been developed using a simple feedforward 15-6-l network.Fifteen phys-cal features including the total energy of the spectrum of cutting forces in x adn z directionsand the energy of the spectrum in several frequency bands and peak frequency are chosen asinputs of the neural network.The three chip types can be expressed by one output node inthe neural network,The discriminant F,or classification effectiveness of physical featuresis proposed.The most effective feature for classifying chip types is the energy of the spec-trum in all spectrum features of dynamic cutting fOrces。The correct recognizing rate is ashigh as 84 percent when different cutting conditions are used for evaluation。It is shown thatthe sensor fusion technique is capable of recognizing and controlling a complex process over awide range of conditions。
出处
《华中理工大学学报》
CSCD
北大核心
1994年第2期5-10,共6页
Journal of Huazhong University of Science and Technology
关键词
切屑类型
神经网络
金属切削
监视
chip types
physical features
neuraI network,sensor fusion
