On-line Monitoring Of Micro-hole Drilling Force Based On Fuzzy Neural Network

Withthedevelopmentofmechanical-electronicproductsforaccurate,integraland miniature applications, micro-hole is widelyused. At present there are manymethods in machining micro-holes, but machining micro-holes by using drillstends to be the most economical and useful in practice. However, when drillingmicro-holes are used to machine micro-holes, there is a serious shortcoming thatintensityof micro-drills is low and break easily when they are wore away or stopup by scrap then draw back. Besides these drawbacks, the statistical date ofmicro-drilllifeisdisperse andlargeso itisverydifficultytopredictjusthowlongthe machining life of a micro-drill might be. Once drills break, it is difficult toextract trappedpart from workpiece,therebyusuallymakingwork piece wastage.Experiment shows that increment of drilling forces when the drill is being wornaway, including thrust and torque, is main reason. To prevent breakage of microdrills, asuit oftheon-lineandreal-timemonitoringsystemofmicro-drillingforcebasedonFuzzyNeuralNetworktechnologyisbuiltupinthisdissertation.Thispaper takes theLabVIEW7.0softwareas aplatform,establishesasuit ofthe on-line and real-time monitoring system of micro-drilling force based onFuzzy Neural Network technology. This system has functions of the dataacquisition, the information processing, recognizing and forecasting, and earlywarning then drawing back. Gets samples through this system, and uses thesesamples that are gathered to train the Fuzzy Neural Network. The obtainedtraining parameters establish Fuzzy Neural Network. Then,completes thereal-timemonitorsystem.This paper introduces the demarcating method of the dynamometers of themicro-drilling force, the basic wavelet theory. Selects the appropriate waveletfunction in view of micro-drilling force, and carries on the analysis of highfrequency components ofmicro-drilling force. Withdraws wavelet decompositionsignals asthe characteristicvectorsofthemonitoringmodels.This paper introduces the Fuzzy Neural Network basic structure, the workprinciple and the algorithm. Based on theory of BPalgorithm, practical problemssuch as design of network structure, choosing of training samples, definition oftraining epochs and pre-process of input data are discussed in this paper. Theresults are applied to the on-line and real-time system. Experiment demonstratesthe real-time monitoring of micro-drilling force, taking advantage of FuzzyNeuralNetworktechnology,canpredictdrillingstates accurately.