李萬慶++左紅++孟文清++陳明欣

摘要： 為客觀、合理地進(jìn)行氣膜薄殼鋼筋混凝土穹頂儲倉的工期預(yù)測，提出了基于PSO-SVR的預(yù)測方法。采用粒子群算法（particle swarm optimization，PSO）對支持向量回歸機（support vector regression，SVR）的參數(shù)進(jìn)行優(yōu)化，并運用優(yōu)化后的支持向量回歸機對氣膜薄殼鋼筋混凝土穹頂儲倉的工期進(jìn)行預(yù)測。通過實例驗證表明：PSO-SVR模型的預(yù)測效果優(yōu)于遺傳算法（GA-SVR）和串聯(lián)型灰色神經(jīng)網(wǎng)絡(luò)（SGNN）。

Abstract： In order to forecast the duration on thin-shell concrete dome using inflated forms objectively and reasonably， the artical presents a prediction method named PSO-SVR. Using PSO to optimize the parameter of SVR， and forecasting the duration on thin-shell concrete dome using inflated forms by support vector regression which is optimized. The example show that the prediction effect of PSO-SVR model is better than genetic algorithm （GA-SVR） and series of grey neural network （SGNN）.

關(guān)鍵詞： 氣膜薄殼鋼筋混凝土穹頂儲倉；工期預(yù)測；PSO-SVR

Key words： thin-shell concrete dome using inflated forms；forecasting of the duration；PSO-SVR

中圖分類號：TU722 文獻(xiàn)標(biāo)識碼：A 文章編號：1006-4311（2017）05-0035-03