OSC—PLS方法用于渣油裂解装置的软测量建模

渣油裂解反应中，影响沥青产率的因素多，反应机理十分复杂，难以建立准确的机理模型。采用基于正交投影的正交信号校正（OSC）算法对输入变量测量数据进行预处理，剔除数据中所含的与待测变量如浓度、收率等无关的噪声信息；再实施OSC与偏最小二乘（PLS）回归相结合的OSC-PLS方法，建立渣油裂解装置沥青产率的软测量校正模型。结果显示：模型精度和稳定性较非线性方法均有显著提高，而且模型所需PLS成分数减少，模型更简洁。

Soft Sensor Modeling for Cracker of Residue Oil by OSC-PLS

It is hard to get the satisfied mechanism model in the deasphalting process of residue oil for complicated influential factors and intricate reactive mechanism. A novel partial least squares algorithm that embedded the orthogonal signal correction （OSC） into the regression framework of the partial least squares（PLS）method, termed as OSC-PLS method, is implemented. OSC technique is used the to delete the noise signal that is independent of the property variable, and the PLS algorithm is applied to build the cracker soft sensor calibration model. The result indicates that the OSC-PLS approach can not only improve the model accuracy and stability with comparison to some nonlinear modeling methods, but also decreases the PLS factors and the model becomes more concise.