Normal Tracking And Edge Detection On Grinding Of Complex Profile Parts

NC Profile Grinding is widely used in precision machining of partswith complex curve profiles，such as dies, forming tools and gauges. Thenormal tracking technology in curve profile grinding could improve forcecondition of abrasion wheel and grinding precision. But, the interferencebetween wheel system and workpiece is hard to avoid in profile grinding.As a result, the normal tracking technology should be advanced to avoidpotential interferences. The visual edge detection for complex curveprofiles is very essential to the error compensation in profile grinding, andmoreover, the detection precision and efficiency have a major influence onthe feasibility of the error compensation. Therefore, the normal trackingand edge detection technology on complex profile grinding should befurther researched.In order to detect potential interference between the wheel systemand the part, a mathematic model is established on the basis of thegeometric and kinematic relationship among the grinding wheel, the partand the grinding wheel support. Principles for determining interference areproposed. Measures that can be used to avoid interferences and theexpression of normal tracking angle are also proposed. Based on theproposed theories, the simulation and experiments of normal trackingconsidering interference avoidance is researched; The NC program iswritten with parameters obtained by simulation results. Experiment resultsshow that the proposed principles for judging interferences and methods for avoiding interferences are reasonable.The edge detection of complex curve profile parts is then studied.Theories and methods on edge detection are firstly discussed, and a newsub-pixel edge detection algorithm is then proposed in order to meet thedemand for precision and efficient edge detection. This new algorithmfirstly chooses threshold from the image histogram, and then segment theimage. Zernike moments are used in image edge detection after imagesegmentation. The experiment results show that the new algorithm canconsiderably reduce the detection time while keeping the sub-pixeldetection precision.