Research On The On-Line Measurement Of Viscosity For C₅ Petroleum Resin Based On Soft-Sensing

Viscosity was an important property in the production process of C5 petroleum resin, if we could realize its continuous on-line detection, we could make a better understanding of the process of reaction according to the relationship between the viscosity and the molecular weight, thus make a better control of the product quality. However, most of the current detection means were off-line, which constrained the production efficiency and quality at this stage.Based on the study of on-line measurement of the viscosity of C5 petroleum resin, this paper introduced some important aspects of the soft-sensing technique in detail, focused on the soft-sensor modeling method such as multiple linear regression method, PLS, RBF neural networks, and hybrid modeling methods. Through a detailed comparative analysis of the actual performance of several soft-sensor models, chose a hybrid model made up of a simple mechanistic model and a data-driven model, the data-driven model was base on a weighted least squares regression in which the capacity of samples was fixed, and updated the model parameters with a on-line correction method.The results show that the hybrid model, which was fast response, had good fitting ability and prediction capability, could be used to guide the practical production. Besides, as a result of the universality of the soft-sensing model, this program could be extended to some other resin producing units, and could also improve product quality, enhance economic efficiency, and reduce energy consumption.