论文部分内容阅读
针对化工过程的强非线性问题,提出一种基于神经网络的非线性主元分析故障检测方法,结合主元曲线算法和2个径向基神经网络,实现非线性主元的识别,并采用统计方法进行故障检测.第一个网络建立输入数据到非线性主元的映射,第二个网络实现逆映射重构原数据.在某炼油厂常压蒸馏过程的常压炉装置中的应用结果表明,基于神经网络的非线性主元分析故障检测方法的效果大大优于线性主元分析(PCA)方法,可准确进行故障检测和分离,保证常压炉安全高效地运行.
Aiming at the strong nonlinear problem of chemical process, this paper proposes a nonlinear principal component analysis based on neural network fault detection method, combining principal component curve algorithm and two RBF neural network to realize the identification of nonlinear principal components, Method for fault detection.The first network establishes the mapping of input data to nonlinear principal components and the second realizes inverse mapping and reconstruction of the original data.The results of application in atmospheric furnace at atmospheric distillation in a refinery , The nonlinear principal component analysis based on neural network fault detection method is much better than the linear principal component analysis (PCA) method, which can accurately detect and separate faults to ensure the safe and efficient atmospheric pressure furnace operation.