Soft Sensor Technology And Its Application In The Grinding

Grinding force and surface roughness are important physicsparameters showing the grinding process and grinding qualityrespectively. Due to the technology secret, there is no information aboutmeasuring the two parameters of spiral bevel and hypoid gears on line tobe vended, and the exceptional complicated teeth profile of the gearsmakes it difficult to measure the two physics parameters on line.The value of Ra is affected by a lot of facts and some of them areunknown, the surface roughness forecasting of spiral bevel and hypoidgears is a typical gray question. The newest information optimized graymodel GM (1, 1,α)is put forward in the paper. By changing the choiceof the initial condition and the white background series z⁽¹⁾(k)in graymodel GM (1, 1, D), the model GM(1, 1,α) is used to model toforecast the surface roughness of spiral bevel and hypoid gears. Theexperiment data proves that gray model GM (1, 1,α) can get a moreprecise forecasting result that the average relative error is less than 1%.The relationship among grinding force and all kinds of machiningconditions is complicated and nonlinear, besides some of unknown facts.Under the principles of geometry similitude and physics similitude,grinding force in grinding spiral bevel and hypoid gears is measured online in the paper by combining simulative experiment and soft-sensor.Namely when modeling, the correlation function in NN toolbox ofMATLAB is used to establish the RFBNN whose inputs are three facts ingrinding and outputs are three components of grinding forces. Based onthe experiment data, the RFBNN is trained and tested. Compared with theBPNN, the RFBNN can get a more precise measuring result of grindingforces in grinding spiral bevel and hypoid gears.By soft-sensor, the corresponding issues in grinding spiral bevel andhypoid gears are studied in the paper, it is evidential to design the toolmachine power, measuring the grinding quality on line becomes possible.