论文部分内容阅读
为了解决有限长度且含有噪声时的单元精度时间序列相空间重构中的信息丢失问题,提出了基于多元混沌时间序列的数控机床运动精度预测方法。首先,引入多元相空间技术,将多个精度特征量时间序列映射到高维相空间,建立多元精度状态空间。然后采用主成分分析法,对高维相空间实现降维,去除冗余。最后,构建一种小波神经网络模型,将重构信息输入到预测模型中训练,实现对数控机床运动精度的预测。实验表明,该方法能够很好地分析数控机床运动精度变化规律,比单元混沌时间序列方法有更好的预测效果,且适应性和实用性更强。
In order to solve the problem of information loss in phase space reconstruction of unit-precision time series with finite length and with noise, a prediction method of motion accuracy of CNC machine tools based on multivariate chaotic time series is proposed. First of all, the introduction of multivariate phase space technology, the number of precision features mapped to the high-dimensional phase sequence, the establishment of multivariate state space. Then, using principal component analysis, dimensionality reduction and redundancy elimination are carried out on high-dimensional phase space. Finally, a wavelet neural network model is constructed, and the reconstruction information is input into the prediction model for training to predict the motion accuracy of the CNC machine tool. Experiments show that this method can well analyze the variation rule of NC machine tool motion accuracy, and has better prediction effect than chaotic time series method, which is more adaptable and practical.