Prediction Of NC Machine Tool's Motion Precision Based On Phase Reconstruction Of Multivariate Time Series

Precision is an important performance index of NC machine tool.The precision of NC machine tool directly affects the precision and quality of machined parts,reflecting the technical level and competitive power of the manufacturing industry.China is a major consumer of NC machine tools,but it is not a major manufacturer of NC machine tools.The demand and manufacturing ability are in unbalanced state,especially the high-end NC machine tools.Compared with imported NC machine tools,the precision of NC machine tools are hard to maintain.Improving the precision retention of the domestic NC machine tool is a problem to be solved.The method of predicting NC motion precision which was based on phase reconstruction of multivariate time series was put forward.According to the forecast results,measures could be taken before the failure of NC machine tool precision to reduce or eliminate the sources of error,which prolonged the runtime of NC machine tools and indirectly improved the precision retention.In addition,it could be used to guide machine maintenance work to prevent the economic loss caused by lack of repairs.The low dimensional time sequence of NC machine tools' motion precision was mapped to high dimension space,in which the chaotic attractor and the ordered state could be restored.Principal component analysis method was introduced to remove redundant information of reconstructed phase space and reduce its dimensionality to simplify the structure of model.Then,the coordinate of the motion precision was input and the roundness error was output.The wavelet neural network model was constructed to achieve motion precision prediction.The main research work in this paper was as follows:(1)A platform was established to test the NC machine tool's motion precision.The time series data of motion precision characteristics was obtained from the test.And the data should be removed noise by the arithmetic mean method.chaotic characteristics of motion precision were analysed by the way of maximum lyapunov exponent.(2)The phase space of the NC machine tool's motion precision was reconstructed.The reconstruction parameters could be calculated by the C-C algorithm.Then the principal component analysis was used to reduce dimensions of high dimensional phase space and remove redundant information.The dimension of the motion precision phase space was taken as the number of neurons input in wavelet neural network.This way could ensure the completeness of input information.(3)The wavelet neural network prediction model of the NC machine tool's motion precision was constructed.Morlet wavelet was chosen as the activation function of the neuron hidden layer.After analyzing the structure of the wavelet neural network and algorithm of parameter correction,the prediction model was trained by reconstruction phase space data.Then according to the evaluation parameters included maximum relative error,relative mean square error and prediction accuracy,the results were evaluated.Finally,the prediction model established was compared with other models.The experimental results showed that the prediction accuracy of NC machine tool's motion precision based on multivariate phase space reconstruction was high.And the relative mean square error of the multivariate prediction model which was reduced dimensions was one order of magnitude lower than the other two models.It indicated that the prediction model proposed in this paper could effectively track the change rule of NC machine tool's motion precision.