为了解决复杂工业生产过程多重时滞辨识难题,提出一种基于改进互相关函数的多重时滞辨识方法。对于工业过程中多个受控制信号作用的过程变量,确定一个参考变量,分别考虑其他各变量和参考变量之间的相关性,选择变量某个时间段内的一组数据作为辨识对象,通过计算两个变量的数据组在不同相对时间延迟对应的互相关矩阵,比较互相关矩阵的奇异值,其最大奇异值对应的滞后时间,即为所要辨识的时滞。将所提方法应用于某铝厂连续碳分过程多重时滞的辨识,基于工业现场采集的生产数据,分析变量间的关联关系,辨识出多重时滞,然后将辨识所得的时滞代入碳分过程模型。结果表明:计算值和实测值的最大平均相对误差仅为3.23%,验证了所提方法辨识多重时滞的有效性。
In order to resolve the problem of multi-delays identification for complex industrial process,an improved cross-correlation function was proposed.In all the process variables affected by control signals,the reference variable was selected by considering the correlation with the other variables respectively.For the considered variable,a set of data in a continuous time segment sampled was selected as identification object.The cross-correlation matrix for the data sets of the reference variable and the other variables was calculated.By comparing the singular values of cross-correlation matrix,the delay corresponding to the maximum singular value was the required delay.The proposed method was applied to identifying the multi-delays of alumina carbonation decomposition process using the field data.At last,the identified delays were substituted to alumina carbonation decomposition process model.The results show that the maximum average relative error between the calculated and tested results is only 3.23%,and the proposed multi-delays identification method based on improved cross-correlation function is effective.