基于BP和GA的激光焊接熱源模型參數(shù)優(yōu)化
發(fā)布時間:2018-09-03 06:17
【摘要】:通過BP神經(jīng)網(wǎng)絡與遺傳算法GA對激光焊接有限元模擬中的熱源模型參數(shù)進行優(yōu)化,實現(xiàn)了對激光焊接溫度場的精確模擬。選取面體熱源模型對激光焊接溫度場進行了有限元模擬,將模擬中難以確定并且對結果影響較大的熱源有效功率系數(shù)、熱能分配系數(shù)和熱源作用半徑作為輸入量,以有限元模擬結果的誤差作為輸出量對BP神經(jīng)網(wǎng)絡進行訓練,得到具有一定預測能力的神經(jīng)網(wǎng)絡,并形成結合神經(jīng)網(wǎng)絡和遺傳算法的參數(shù)優(yōu)化方法。結果表明,經(jīng)過參數(shù)優(yōu)化后的激光焊接有限元模擬具有較高的精度。
[Abstract]:The heat source model parameters in laser welding finite element simulation are optimized by BP neural network and genetic algorithm GA, and the accurate simulation of laser welding temperature field is realized. The surface heat source model is selected to simulate the temperature field of laser welding by finite element method. The effective power coefficient, thermal energy distribution coefficient and the radius of action of heat source, which are difficult to determine in the simulation and have a great influence on the results, are taken as the input quantity. The BP neural network is trained with the error of the finite element simulation result as the output, and the neural network with certain predictive ability is obtained, and a parameter optimization method combining the neural network and genetic algorithm is formed. The results show that the finite element simulation of laser welding with optimized parameters has high accuracy.
【作者單位】: 武漢理工大學現(xiàn)代汽車零部件技術湖北省重點實驗室;武漢理工大學汽車零部件技術湖北省協(xié)同創(chuàng)新中心;武漢理工大學汽車工程學院;
【基金】:國家自然科學基金資助項目(51305317) 中國汽車產(chǎn)業(yè)創(chuàng)新發(fā)展聯(lián)合基金(U1564202) 湖北省自然科學基金重點項目(ZRS2014000009)
【分類號】:TG456.7
本文編號:2219136
[Abstract]:The heat source model parameters in laser welding finite element simulation are optimized by BP neural network and genetic algorithm GA, and the accurate simulation of laser welding temperature field is realized. The surface heat source model is selected to simulate the temperature field of laser welding by finite element method. The effective power coefficient, thermal energy distribution coefficient and the radius of action of heat source, which are difficult to determine in the simulation and have a great influence on the results, are taken as the input quantity. The BP neural network is trained with the error of the finite element simulation result as the output, and the neural network with certain predictive ability is obtained, and a parameter optimization method combining the neural network and genetic algorithm is formed. The results show that the finite element simulation of laser welding with optimized parameters has high accuracy.
【作者單位】: 武漢理工大學現(xiàn)代汽車零部件技術湖北省重點實驗室;武漢理工大學汽車零部件技術湖北省協(xié)同創(chuàng)新中心;武漢理工大學汽車工程學院;
【基金】:國家自然科學基金資助項目(51305317) 中國汽車產(chǎn)業(yè)創(chuàng)新發(fā)展聯(lián)合基金(U1564202) 湖北省自然科學基金重點項目(ZRS2014000009)
【分類號】:TG456.7
【相似文獻】
相關期刊論文 前3條
1 譚立云;劉海生;譚龍;;基于極限學習機的上證指數(shù)預測與分析[J];華北科技學院學報;2014年04期
2 曹晉宏;李國勇;;礦井瓦斯智能預警系統(tǒng)的設計[J];煤礦機械;2014年08期
3 ;[J];;年期
相關會議論文 前2條
1 王光蘆;徐明;;飛機飛行振動預計技術[A];第二屆民用飛機制造技術及裝備高層論壇資料匯編(論文集)[C];2010年
2 余建星;田佳;譚振東;;深水平臺施工過程自然風險定量評估方法研究[A];中國造船工程學會2007年優(yōu)秀學術論文集[C];2008年
,本文編號:2219136
本文鏈接:http://sikaile.net/kejilunwen/jiagonggongyi/2219136.html
最近更新
教材專著