基于SVR的回转窑烧成带温度软测量方法的研究

提出了一种新的回转窑烧成带温度的软测量方法.首先利用大津方法从烧结图像中分割出物料区和火焰区,并提取相应的特征值,然后采用最速下降法选择支持向量回归(SVR)模型的参数,基于这些参数建立了烧成带温度的软测量模型,最后利用前向滑动平均进行滤波提高拟合优度.将该方法应用于中国铝业公司山西铝厂的3#回转窑的氧化铝烧结过程,通过大量实验,验证了烧成带温度的软测量结果与实测值的拟合优度达到0.927,表明了该方法具有很强的实用性和准确性.

Study on Soft Sensor for Temperature of Burning Zone Based on SVR

A new soft sensor method for temperature of the burning zone of the rotary kiln was proposed. Firstly, the flame zone and material zone were segmented from the burning image using ostu method, and the characters were extracted. Then the steepest descents method was used to select the parameters of SVR model, and the SVR model for measuring the temperature of burning zone was set up based on selected parameters. At last, the prior moving average method was proposed as filter method. This method was applied to the 3# alumina rotary kiln of the Shanxi alumina production manufacture. Many experiments prove the final goodness of fit between the measured temperature and real temperature is 0.927, which shows the practicability and precision of this soft sensor method.