论文部分内容阅读
近年来,流程工业事故频发,这使得加强生产过程的安全保障迫在眉睫,对故障识别的准确性提出了更高的要求。本文提出了一种基于特征重构的广义回归神经网络故障识别方法。首先,引入“字典表”的功能构建“故障字典表”:其次,采用核主元分析方法对“故障字典表”进行主元提取,实现数据降维以及降低计算复杂度;第三,“故障字典表”索引定位,通过数据样本与“故障字典表”的比对,对数据样本进行特征重构;最后,运用广义回归神经网络算法对数据样本进行学习训练,用以计算系统输出变量进行故障识别。通过对TE(Tennessee Eastman Process)过程进行故障识别仿真实验,结果表明,该方法对非线性时序系统具有较高的故障识别能力,为复杂过程工业大型系统的故障识别提供了新的思路和方法。
In recent years, frequent industrial accidents in the process make it very urgent to enhance the safety of the production process and put forward higher requirements on the accuracy of the fault identification. This paper presents a generalized regression neural network fault recognition method based on feature reconstruction. First of all, the introduction of “dictionary table ” function to build “fault dictionary table ”: Second, the use of nuclear principal component analysis of the “fault dictionary table ” principal component extraction to achieve data dimension reduction and reduce computational complexity Thirdly, the index of “fault dictionary table” is indexed, and the data samples are reconstructed by comparing the data samples with the “fault dictionary table”. Finally, the data samples are learned by using the generalized regression neural network algorithm Training, used to calculate the system output variables for fault identification. The simulation experiment of fault recognition in TE (Tennessee Eastman Process) shows that the proposed method has high fault identification ability for nonlinear time series systems, and provides a new way of identifying faults in complex process industrial large systems.