《基于AdaBoost溷合模型的LF炉钢水终点温度软测量》.pdf

29 3 Vol29 No3 2008 3 Ch inese Journal of Scientific Instrum ent M ar. 2008 AdaBoost LF 毛志忠, 田慧欣, 王 琰 ( 110004) : LF, , , AdaBoost. RT BP , AdaBoost. RT BP , , ! 5 ∀ 85% : LF ; ; ; A daBoost; BP : TP206 : A : 510. 8040 M olten steel end tem perature soft sensing in LF based on hybrid model using AdaBoost M ao Zh izhong, T ian Huix in, W ang Yan (School of Inf ormation Science E ng ineering, N ortheast University, Shenyang 110004, Ch ina) Abstract: Th is paper pro ides a new hybrid modelingm ethod that comb ines traditionalm echan ism modelw ith artifi cial intelligence method to predict the temperature of lad le furnace. In th is new method, an mi pro ed AdaBoost. RT using BP netw ork as a w eak learning mach ine is presented to modify the parameters of themechanism mode.l Then the temperature pred iction resu lt is obtained by the mechanism model. The traditionalm echan ism method can not predict the temperature exactly and the b lack box m ethod depends on the data excessi ely. Th is new hybrid mod eling method could o ercom e abo e disad antages. A t the same tmi e, the new method o ercomes the shortcom ing of BP algorithm. Expermi ent results show that the new hybr id modeling ofAdaBoost. RT cou ld pred ict themolten steel temperature in LF more exactly. the number of furnace heatsw ith the predicti e error ofmolten steel end temperature in LF is not larger than ! 5 ∀ is greater than 85% . Key words: ladle furnace; hybridm odeling; soft sensing; A daBoost; BP network LF 1 , , LF LF , LF , , , LF , - 2 : 200701