论文部分内容阅读
以华智公司的日生产产品不合格率为时间序列,研究制造质量信息系统的重构相空间的延迟时间。在使用四种自相关法及互信息法计算时间序列的延迟时间后,发现普通自相关法无法得到延迟时间,而改进自相关法、去偏复自相关法、统计自相关法由于只能考虑线性相关性及部分非线性相关性,所以对于所研究的时间序列来说,对其中的非线性相关性无法体现,也就得不到可信的延迟时间。而互信息法则适用于非线性相关时间序列,所以使用互信息法可以得到可信的延迟时间。另外,在使用互信息法计算延迟时间时,网格的划分要使每一坐标方向上的网格数目与时间序列的长度在同一数量级上,并使网格数不要太多,以免造成计算机不能计算或计算时间过长,这将导致该方法不可用。
Take Huazhi’s daily product failure rate as time series to study the delay time of reconstructed phase space of manufacturing quality information system. After using four kinds of autocorrelation method and mutual information method to calculate the delay time of the time series, we find that the ordinary autocorrelation method can not get the delay time, and improve the autocorrelation method and the de-autocorrelation method. The statistical autocorrelation method can only consider Linear correlation and partial non-linear correlation, so for the study of the time series, which can not be reflected in the non-linear correlation, there is no credible delay time. The law of mutual information applies to non-linear related time series, so using the mutual information method can get credible delay time. In addition, when using the mutual information method to calculate the delay time, the division of the grid is such that the number of grids in each coordinate direction is on the same order of magnitude as the length of the time series, and the number of grids should not be too large to prevent the computer from failing Calculation or calculation takes too long, which will render this method unavailable.