| The numerical control machine tool of numerical control processingfunctions in the machine testing detection with the quality of the three coordinatemeasuring device form an organic whole, which improve the efficiency and accuracy.In throughout the technical route of parts processing, high efficiency, high precisionto obtain measurement data is a foundation, and the data preprocessing is the key toensure the quality of late model reconstruction. In the contact measurement,measurement data is very huge and inevitably exsits noise, at the sametime,measured data collected the header heart point rather than the parts surfacecoordinates, the coordinates of probe radius compensation is required. Sopretreatment is very necessary. Current data preprocessing techniques are more orless have some deficiencies, for non-contact scanning probe measurements obtaineddense scattered point cloud data, based on the SOFM neural network is put forward akind of strong adaptability, high precision, high efficiency of data preprocessingmethods, to lay a good foundation for the follow-up of CAD model reconstruction.Using SOFM neural network to make measure the scattered point data obtainedform area set, the classification of the core promoters that neurons as the connectionweight vector of scattered point set engineering approximation and reconstructsurface sample points of inner topological relationship, also realize the surface dense3d scattered points of self-organization of data compression; At the same time itadjust Lateral inhibition adjacent regions according to the rectangular array, and thena certain measurement point set topology data smoothing ability rectangular grid hasgenerated compressed. According to the data information a certain point the methodof vector can be directly calculated, of edge points in the square at the same time, thelack of points can be found in field neurons, complete solution of the rectangulargrid method. Using the method of calculating vector, the probe radius compensationcan complete and validated to the rectangular grid data smoothing ability. In addition,due to the large amount of data, the training efficiency declines, so the measurement data are simple chunking training and integration to improve the efficiency of dataprocessing under the precision.Through computer simulation experiments of rectangular grid data smoothingability has been effectively verificated, and the massive data preprocessing has beenbetter achieved. Application of MATLAB software to construct spherical, generalcomplex curved surface, double three b-spline curved surface of three complexlevels from low to high surface model has been verified. After a lot of simulationcomparison and analysis, this data preprocessing accuracy up to micron level, andcan completely meet the engineering need, software module of algorithm isconstructed, in order to be applied to the numerical control machining in machinetest data processing software. |