Study On The Key Techniques Of Intelligent Cylindrical Traverse Grinding

The grinding result directly affects the final quality and performance ofproduct because the grinding is usually the ultimate machining procedure. But thegrinding process is non-linearity and randomcity with too many influence factors;the accuracy mathematical model cannot be built, and the traditional controlmethod cannot resolve the control problem during the grinding. Traverse grindingwhich is called universal grinding,is very important for shaft workpiece.Controlling the process of traverse grinding is more difficult than that of a plungegrinding because there are more parameters which affect the grinding process andthe nature of the process is non-liner and non-static. The deformation changes withthe wheel position because the system stiffness varies with the relative position ofthe wheel and the workpiece. In traverse grinding, the feed per-revolution duringcylindrical grinding is much smaller than the whole grinding wheel width.Consequently, each section along the width of the grinding wheel experiences adifferent effective depth of cut. Due to their complex nature, the grinding processesreveal stochastic responses because of uncontrollable and no-measurable processparameters such as changeable material and grinding wheel properties, coolingconditions.At present, adjusting the grinding process is strongly relied on theoperator's experience and skill. The grinding is one of the key technologies thatconstraints some elements of advanced technology development. Hence, intelligentgrinding is the development tendency.The paper is supported by the project of scientific and technical R&D of JilinProvince (20020632).The theoretical and experimental study were conducted forthe key techniques of intelligent cylindrical traverse grinding.The intelligentcontrol is preliminary implemented by introducing the advanced theory andmethods such as expert system,evolution algorithm,fuzzy logic and neuralnetwork.The project has carried out appraisal in July 2005,the result is internationaladvanced.been appraisaled. The accomplishment of this program is popularizedand applied by Chang Chun Bus factory etc. Corresponding conclusion and themajor innovation accomplishment of this paper are as follows:1. The Development of the Intelligent Experimental System for CylindricalTraverse GrindingIn order to verify the intelligent grinding theory, the experiments had beendone on the modified MMB1320 semiautomatic external grinder using computerand sensors.On hardware,the Industrial Computer is the core of the intelligentgrinding system,where as the sensors fuse technology was used and the stepmotor was used as servo motor,the precision can reach 0.5 micron;On sortware,theopen cylindrical traverse intelligent system is construct using Windows2000 as theplatform and the object-oriented software programming language VC++6.0 asdevelopment tool,which offer convenient interface for system develop twice byuser with friendly man-machine interface. The entire intelligent grinding systemcompose of expert system, intelligent forecast system and intelligent control systemetc.2. The Analysis Model of Cylindrical Traverse GrindingIt is proved that the ralationship presents nonlinear between grindingparameters (such as workpiece peripheral speed Vw, traverse feed rate of the tablefa and wheel depth of cut ap) and grinding process variables when the grinder'scondition is constant.In order to get the relationship between the grindingparameters and process variables,the analysis models of grinding force andgrinding power are built using angle just regression method combining withexperimental study.futhermore the size form mechanism is deeply studied;theconcept of the wheel contact deform is introduced;the mathematic model of wheelwear is built;the mathematic model of workpiece size is built, which provide thetheory base for size prediction and control.In addition,the analysis model of surfaceroughness is established,which can simplify the training process of Fuzzy BasedFuction Network and raise learning efficiency.3. The Expert System of Cylindrical Traverse GrindingThe choice of initial parameters is very important during grinding course. Atpresent,adjusting the grinding process is strongly relied on the operator'sexperience and skill.In this paper the task of the choice of initial parameters isresponded by an expert system.Knowledge get and regular generation always the bottleneck of expertsystem.In this paper,the optimized model was built combing the expert knowledgebased on heuristic rule with the grinding optimized model and the time-varyingwas considered. It is difficult to get the optimum solution using gradient-basedalgorithms or LS (least-squares) because the process variables involvedcontinuous variables(such as workpiece speed,table feed etc) and discrete( integer)variables (such as number of grinding passes) in the grinding process.The grindingintinal parameters' value can be obtained using the evolution strategies(ES),which can supply the optimum value for the intelligent control system.Case studies for minimization of grinding cycle time,minimization of grindingcost and process control are carried out according to different demand.4. The Prediction System for Cylindrical Traverse GrindingTo realize the intelligent control of grinding process , it is necessary to carryout detection for output variables and process variables of grinding and feedback tothe Industrial Computer, and then adjust the control variable's value to makegrinding process optimum.But it is difficult to detect some process parameters,such as surface roughness and workpiece size during grinding process,we shouldpredict the value of those parameters.This paper built the predictive model of surface roughness in cylindricaltraverse grinding process based on FBFN using ALS algorithm. The ALSalgorithm provides true hybrid structure-parameter learning without humanassistance by globally searching for an optimal fuzzy rule using GA. Simulationstudies showed that new algorithms generate superior results over conventionalalgorithms such as Backpropagation algorithms and conventional GA-basedalgorithm. The study on surface grinding process using a small amount ofVIIexperimental data demonstrated the potential of the ALS algorithm in constructingfuzzy models of complex manufacturing processes efficiently.According to the nature of highly non-liner and non-static state in traversegrinding, a size intelligent prediction control model based on the dynamic Elmanneural network is constructed. The first and the second derivative of the actualamount removed from the workpiece are added into the network input, which cangreatly improve the prediction accuracy, so the change tendency of the systemactual output is given to the network. The size prediction model can improve andraise prediction precision for cylindrical traverse grinding and remove theobstacles often appear during system identification, the build model time andcontrol respond speed, and result in compensation lag or uncontrolled.5. The Intelligent Control System for Cylindrical Traverse GrindingAccording to the nature of highly non-liner, non-static state and the variationin traverse grinding, it is desired to have an adaptive controller for the system.Inthis paper,a intelligent fuzzy controller was described ,which divide the grindingprocess into two stages:rough stage and finish stage.The function of the rough stageis removal most of the grinding allowance,and the object of optimization is often tohave higher grinding efficiency. The power control strategy was used in roughstage,which can maintain the G-ratio near its optimal value, which ensures a lowwheel wear rate and a high material removal rate. The motivation for using powercontrol is based on the possibility to control system deflection and avoid excessivewheel wear, which affects the surface finish and tolerances. Power control has theadvantage over force control because it is simpler to measure the power than theforces in cylindrical grinding.The table traverse feed is selected as the controlvariable,which can adjust the value of fa in real time according to the grindingcondition.The function of the finish stage is to accord with quality requirement of theworkpiece.In this stage,the geometrical control strategy was adopt , The tabletraverse feed fa is selected as the control variable,which can adjust the value offa in real time according to the grinding condition.The controller was implemented and tested on an open architecture cylindricalgrinding machine. The results show that the adaptively controlled system providesa better response than the non-adaptive control system over a wider range ofcutting conditions.The key techniques and methods of intelligent cylindrical traverse grindingsystem can apply to other grinding way and other mechanical manufacturingprocess by small modified.