Sotf Sensor Modeling For On-line Estimation Of Gelatin Concentration

Concentration of gelatin is one of the important parameters of the gelatinproduction process, it can be reflects the level of quality of gelatin, so the gelatinconcentration detection speed affect the gelatin production process. Domestic gelatinmanufacturers mostly use artificial timing sampling method to measure liquidconcentrations, but this is not easily contaminated sample, but also measured with alag. In order to change the existing glue concentration measurement, improve thelevel of automation of the gelatin production, we deduced dominant variable difficultto measure by measuring auxiliary variable which are easily accessible with theadvantage of the soft sensor modeling to achieve the online measurement of theconcentration of gelatin.This paper takes the Qinghai Gelatin company refining processes technology asthe background and takes the concentration of gelatin as study object, after carefulanalysis of the gelatin production process and analyze data collected from anindustrial site, we select time, temperature and specific gravity which are easiermeasurement closely related to the glue concentration as an auxiliary variable.Toreduce the impact on the measurement results by the data due to differences inmagnitude, the data were normalized, and then create a series of soft measurementmodel to achieve the online measurement of the concentration of gelatin.In this paper, multi-models based on BP_Adaboost and the fuzzy C-meansclustering (Fuzzy C-means FCM) algorithm and support vector machine and GKclustering (Gustafson_Kessel, GK) algorithm of radial basis function is establishedrespectively and the multi-models used to measure the concentration of gelatin.werecompared with the single model use the same algorithm, measurement results showthat the multi-model has a higher measurement accuracy.To make the soft instrument really play the practical utility, we use the MATLABand VC for mixing programme. Combine the MATLAB engine and the VCprogramming methods and match the powerful computing features in MATLAB andVC operating environment perfectly, and create interactive interface, a true sense ofthe soft sensor field applications is realized at last.