CAJ | 학술논문

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。
근년래，류정공업사고빈발，저사득가강생산과정적안전보장박재미첩，대고장식별적준학성제출료경고적요구。본문제출료일충기우특정중구적엄의회귀신경망락고장식별방법。수선，인입“자전표”적공능구건“고장자전표”；기차，채용핵주원분석방법대“고장자전표”진행주원제취，실현수거강유이급강저계산복잡도；제삼，“고장자전표”색인정위，통과수거양본여“고장자전표”적비대，대수거양본진행특정중구；최후，운용엄의회귀신경망락산법대수거양본진행학습훈련，용이계산계통수출변량진행고장식별。통과대TE(Tennessee Eastman Process)과정진행고장식별방진실험，결과표명，해방법대비선성시서계통구유교고적고장식별능력，위복잡과정공업대형계통적고장식별제공료신적사로화방법。
Recently, industrial accidents occur frequently, which make the production safety work urgent. Now, more demand has been presented for the accuracy of fault identification. In the paper, a general regression neural network (GRNN) fault identification method based on character reconstruction is proposed. First, the Fault Dictionary Table (FDT) is constructed by introducing the dictionary function. Second, the principal component extraction of FDT is made by Kernel Principal Component Analysis (KPCA), which is used to reduce the data-dimension and calculation complexity. Third, make the index location for FDT. Through the comparison of data sample and FDT, the character reconstruction is made for the data sample. Finally, the data sample is trained by GRNN to calculate the system output so as to identify the fault. Through the simulation experiment of fault identification on Tennessee Eastman (TE) Process, the results demonstrate that the proposed method has higher fault identification ability for nonlinear system, which provides a new way for the fault identification of complex large-scale process industry system.