(中南大學(xué) 資源與安全工程學(xué)院，長沙 410083)
摘 要: 建立全尾砂沉降速度GA-SVM優(yōu)化預(yù)測模型，利用遺傳學(xué)算法對全尾砂沉降速度進(jìn)行優(yōu)化預(yù)測。建立支持向量機(jī)(SVM)回歸預(yù)測模型，采用訓(xùn)練集對模型進(jìn)行訓(xùn)練，以驗證集預(yù)測值的均方誤差作為適應(yīng)度函數(shù)，通過遺傳算法(GA)對SVM模型參數(shù)進(jìn)行優(yōu)化選擇，應(yīng)用優(yōu)化得到的SVM模型對預(yù)測集進(jìn)行預(yù)測。以司家營鐵礦為例，在絮凝劑單耗8.6 g/t、尾砂濃度18%條件下，沉降速度即可達(dá)到1.31 m/h，滿足生產(chǎn)需要，比原生產(chǎn)所需絮凝劑單耗減少14%。應(yīng)用表明：該預(yù)測模型具有較高的實用性，為全尾砂沉降速度優(yōu)化預(yù)測提供一種全新思路。
關(guān)鍵字: 充填；沉降速度；支持向量機(jī)；遺傳算法
（School of Resources and Safety Engineering, Central South University, Changsha 410083, China）
Abstract:Based on the GA-SVM optimal prediction model of sedimentation velocity, the genetic algorithm was used to make an optimal predicition. Support vector machine (SVM) regression model was established and trained by the use of samples of training. The acquired mean error of the value was made as a fitness function. Then, the model parameters were optimized through the genetic algorithm (GA). At the end, the optimized SVM was applied to predict the prediction set. GA-SVM optimal prediction mode was used in Sijiaying Iron Mine, the results show that when the flocculating agent consumption and tailings concentration are 8.6 g/t and 18%, respectively, the sedimentation velocity reaches 1.31 m/h. which meet the production requirements. The optimal prediction mode has relatively high practical value, can provide a new method to optimize the sedimentation velocity of unclassified tailings.
Key words: backfill; sedimentation velocity; support vectormachine; genetic algorithm
