Construction Method Of Curve Surface Model And Its Applications

Modeling of curve surface refers to construct a model with date point in a givenprecision and thus gaining a smooth surface. This is an ordinary problem encounteredin manufacturing industry and also a key problem in designing and producing ofpropeller.This paper mainly includes three parts: the first part describes a constructionmethod of propeller surface model by using cubic spline interpolation.This methodcan be divided into two steps .The first step aims at calculating data point coordinatesof any section and the second step is to calculate any point's coordinate on the sectionby the result of the first step.The second part provides another two constructionmethod by using BP network or RBF network for the first time.These two methodsinclude definition of neural network's structure and parameter,and training of neuralnetwork with the given data point,which will not stop until deviation meets therequirement.In the third part, to satisfy some manufacturing factory ,a set of specialsoftware is put forward, which includes data input, data dealing ,data store anddrawing generation ,based on the construction method of propeller surface model.By comparing and analyzing,the three methods above are proved feasible insolving practical problem. And at the same time,they have their own advantages andshortcomings.Firstly,comparing to neural network, cubic spline interpolation featuressimple operation and program development, but a smaller fit range. Neural networkcan theoretically realize any mapping and approximation of function,but there needshuman's intervene in its modeling and neural network is relatively complex to realizethe program.Secondly,comparing to BP,RBF has a higher learning rate and has not theproblem of tending to converge to a local minimum,but its learning algorithm is morecomplex.