(中南大學(xué) 信息科學(xué)與工程學(xué)院，長沙 410083)
摘 要: 針對數(shù)據(jù)量龐大引起模型參數(shù)更新時(shí)樣本選擇困難及訓(xùn)練速度慢的缺陷, 提出基于投影尋蹤回歸的銅閃速熔煉過程關(guān)鍵工藝指標(biāo)預(yù)測方法。首先采用機(jī)器學(xué)習(xí)方式提取用于建模所需的相似樣本集，借助投影尋蹤回歸思想, 建立銅閃速熔煉過程關(guān)鍵工藝指標(biāo)預(yù)測模型；然后利用基于實(shí)數(shù)編碼的加速遺傳算法進(jìn)行模型參數(shù)的實(shí)時(shí)更新。訓(xùn)練樣本的機(jī)器選擇可以避免人工選擇帶來的主觀性和盲目性缺陷, 模型參數(shù)的更新訓(xùn)練只在相似樣本集中進(jìn)行，可有效提高模型參數(shù)更新速度。實(shí)際生產(chǎn)數(shù)據(jù)仿真結(jié)果驗(yàn)證了所提方法的有效性和可行性。
關(guān)鍵字: 銅閃速熔煉過程；投影尋蹤回歸；相似性度量；加速遺傳算法
（School of Information Science and Engineering, Central South University, Changsha 410083, China）
Abstract:Aimed at the shortcoming of difficult in choosing training samples and the slow training speed, which are caused by enormous data, a method based on projection pursuit regression that can predict the key process indicators for copper flash smelting process was proposed. With the similar samples set retrieved from the data base, the model of predicting key process indictors was developed by using projection pursuit regression method, and the model parameters were updated with acceleration algorithm in time, which is useful in avoiding the defects of subjectivity and blindness caused by artificial factors to select training samples. As the model parameters update training is only carried out within the similar samples, the update speed of the model parameters is effectively improved. Simulation results of actual data from a copper flash smelting process are given to verify the effectiveness and feasibility of the proposed method.
Key words: copper flash smelting process; projection pursuit regression; similarity measure; acceleration genetic algorithm
