带FIR滤波的非线性滑动平均动态软测量模型

非线性滑动平均(NMA)模型能有效描述工业过程的动态特性,是一种典型的动态软测量模型.而受限于模型复杂度, NMA模型的输入时序边界相对较窄,难以适应带有大滞后或强测量噪声的动态工业过程.针对该问题,本文将NMA模型与结构简单、输入时序边界宽的FIR滤波器相结合,构造一种非线性、强抗干扰的软测量建模策略.并设计层白化结构来避免二者间的参数耦合现象,采用Adam算法进行同步优化,提高模型的预测精度及训练效率.最后,利用数值仿真和硫回收过程建模实验,验证所提模型的预测精度以及模型设计的合理性.

Nonlinear moving average dynamic soft sensor model with FIR filter

The nonlinear moving average (NMA) model has been widely applied in the field of dynamic soft sensing. However, due to the high model complexity, the averaging horizon of the NMA model is relatively short. As a result, its accuracy could be compromised if the industrial process embodies long time delay or strong measurement noise. To solve this problem, a novel soft sensor modeling strategy, with relatively strong anti-interference ability and long averaging horizon, is proposed by combining the NMA model with the FIR filter. Besides, parameters in the NMA model and FIR filter are optimized synchronously based on the Adam algorithm. Meanwhile, the layer whitening strategy is designed to avoid the parameter coupling phenomenon between the two structures mentioned. Finally, numerical simulations and sulfur recovery process modeling experiments have been conducted to verify the prediction accuracy of the proposed model and the rationality of model design.