Research Of Soft Measurement Method Of Raw Material’s Grinding Fineness Based On Support Vector Machine

In cement production process, raw materials required by grinding formation with certain fineness of powder had a huge impact on quality of cement, raw meal fineness of detection is an important part of the system optimization control of raw material grinding. Domestic usually adopt the way of artificial offline test for raw meal fineness screen residue percentage, but the test cycle generally for two to four hours; Usually adopted abroad online analyzer of raw meal fineness measurement, but it is expensive and difficult to maintain. For raw meal fineness of artificial test cycle time is long, low working efficiency, online analyzer is expensive and difficult to maintain cause raw meal fineness could not have qualified rate is low and the testing efficiency and economic benefit. This paper proposes a fast effective raw meal fineness soft measurement method, raw meal fineness was designed and implemented online soft measurement system, effectively solve the artificial low detection efficiency and on-line analyzer is expensive. Specific studies are as follows:First, this paper summarizes the existing soft measurement modeling method, and further study of the working principl e of the system of raw material grinding process and raw meal grinding. Through the analysis of the influence factors of raw meal fineness of raw meal fineness soft measurement system auxiliary variables, is proposed based on support vector machine(SVM) a nd data-driven modeling method, realize the soft measurement of raw meal fineness.Secondly, in view of the raw meal fineness soft measurement modeling due to there is delay in the process of modeling, solve the problem of building the system time delay(T) and least squares support vector machine(LSSVR) with the combination of soft measurement model: the system time delay is divided into "static response delay" and "dynamic response delay"; By using normalized cross-correlation and first-order generalized difference method, get "static response delay" and "dynamic response latency" two parameters of the mathematical model of time delay estimation; And by using the least squares support vector machine(SVM) samples treated by time delay system prediction mo del is established. Based on this T- LSSVR model for soft measurement modeling experiment, raw meal fineness of raw meal fineness in this paper, the proposed method was verified the validity and reliability of soft measurement modeling.Finally, the fineness of raw meal was designed and implemented online soft measurement system. Through the soft measurement system on the client and the server database, the data interaction between complete online soft sensor for the raw meal fineness, thus achieved in raw meal grinding of raw meal fineness of forecasting. The system meets the requirements of basic industry, save the human and economic cost, improve work efficiency, create conditions for optimal control of raw meal fineness.