基于改进LLE-LSTM的化工过程故障诊断

Fault Diagnosis of Chemical Process Based on Improved LLE-LSTM

针对化工生产过程数据多样性、高维性以及相似性的特点,传统的局部线性嵌入难以发掘数据高维非线性、不均匀特征的问题,本文提出一种改进LLE-LSTM算法。首先,运用改进LLE算法求出样本集的协方差矩阵,计算权重系数矩阵,将样本集映射到低维空间。其次,将重构的低维样本集输入LSTM模型,进一步提取样本特征。最后,对故障类型进行诊断和分类。将该方法应用到田纳西-伊斯曼(TE)过程,实验结果表明该方法具有更高的准确性和优越性。

In view of the diversity,high dimensionality and similarity of chemical production process data,the traditional local linear embedding is difficult to explore the problem of high-dimensional nonlinear and inhomogeneous features of the data,and an improved LLE-LSTM algorithm is proposed.The improved LLE algorithm is first applied to find the covariance matrix of the sample set,calculate the weight coefficient matrix,map the sample set to a low-dimensional space,and then input the reconstructed low-dimensional sample set into the LSTM model to further extract the sample features,and finally diagnose and classify the fault types.Applying the method to the Tennessee-Eastman(TE)process,experimental results show that the method has higher accuracy and superiority.