正在加载图片...
第5期 甘春雷等:基于ANFIS方法的连续定向凝固BFel01-1合金的压缩流变应力模型 ·605· 为了进一步分析模型计算流变应力值的偏差大 于金属流变应力模型的研究 小,将采用ANFIS模型和回归模型对预测的流变应 力值与试验值的偏差变化情况示于图9所示.从图 参考文献 中可以看出,采用回归模型计算的流变应力偏差范 [1]Yu Y Q,Zhao F,Li W G.Manufacturing of BFe3044 cupro- nickel tubes by heated mold continuous casting.Spec Cast Nonfer- 围为-63.85~42.58MPa,而ANFIS模型计算的偏 rous Alloys,2005,25(12):761 差范围为-8.64~7.94MPa,可知ANFIS模型预测 (余业球,赵锋,黎沃光.用热型连铸法制造BF30一1白铜 值的偏差范围非常小,远小于回归模型计算的偏差 管坯.特种铸造及有色合金,2005,25(12):761) 范围. [2] Yu Y Q,Zhao F,Li W G.Microstructure and performance of BFe30-4 cupronickel tubes.Hot Work Technol,2005(11)8 100 (余业球,赵锋,黎沃光.热型连铸BF30H白铜管的组织 ANFIS ·回归模型 与性能.热加工工艺,2005(11):8) B] Mandal S,Rakesh V,Sivaprasad P V,et al.Constitutive equa- tions to predict high temperature flow stress in a Ti-modified auste- 20 nitic stainless steel.Mater Sci Eng A,2009,500(1/2):114 4]Takuda H,Morishita T,Kinoshita T,et al.Modelling of formula 2 for flow stress of a magnesium alloy A731 sheet at elevated temper- atures.J Mater Process Technol,2005,164/165:1258 60 [5]Varol Y,Koca A,Oztop F,et al.Analysis of adaptive-network- -1006 based fuzzy inference system (ANFIS)to estimate buoyancy-in- 100200300400 500 duced flow field in partially heated triangular enclosures.Expert 样本数 Syst Appl,2008,35(4):1989 图9ANFS和回归模型预测值与试验值的偏差变化 [6]Hanbay D,Baylar A,Ozpolat E.Predicting flow conditions over Fig.9 Variation of the deviation between measured values and pre- stepped chutes based on ANFIS.Sof Comput,2009,13(7):701 dicted results for the ANFIS and regression model Sfetsos A.A comparison of various forecasting techniques applied to mean hourly wind speed time series.Renew Energy,2000,21 4结论 (1):23 [8]Chang F J,Chang Y T.Adaptive neuro-fuzzy inference system for (1)在变形温度为25~500℃,应变速率为 prediction of water level in reservoir.Ade Water Resour,2006,29 0.01~10s-的条件下,连续柱状晶组织BFe10-11 (1):1 合金压缩变形以应变硬化为主,未出现明显的再结 Wei M Z,Bai B J,Sung A H,et al.Predicting injection profiles using ANFIS.Inf Sci,2007,177(20)4445 晶现象 [io] Leidermark D,Moverare JJ,Simonsson K,et al.Room tempera- (2)采用自适应神经模糊推理系统建立了可反 ture yield behaviour of a single-erystal nickel-base superalloy with 映连续柱状晶组织BFe10-1-1合金在变形温度为 tension/compression asymmetry.Comput Mater Sci,2009,47 25~500℃,应变速率为0.01~10s-条件下的压缩 (2):366 [11]Jang J S R.ANFIS:Adaptive-network-based fuzzy inference sys- 变形全过程的流变应力预测模型,具有优良的预测 tem.IEEE Trans Syst Man Cybern,1993,23(3):665 精度:流变应力预测值的平均误差为0.75%,均方 [12]Shen J C.Fuzzy neural networks for tuning PID controllers for 根误差为2.13,相关系数为0.9996. plant with underdamped responses.IEEE Trans Fuzzy Syst, (3)在本文研究条件下,与传统回归模型预测 2001,9(2):333 结果相比,ANFIS模型预测连续柱状晶组织 [13]Yuan S C.MATLAB Language and Mechanical Engineering Ap- plication.Beijing:China Machine Press,2008 BFel01-]l合金流变应力的精度(平均误差)提高 (原思聪.MATLAB语言及机械工程应用.北京:机械工业 了5.53%,具有更高的拟合能力,表明该方法适合 出版社,2008)第 5 期 甘春雷等: 基于 ANFIS 方法的连续定向凝固 BFe10--1--1 合金的压缩流变应力模型 为了进一步分析模型计算流变应力值的偏差大 小,将采用 ANFIS 模型和回归模型对预测的流变应 力值与试验值的偏差变化情况示于图 9 所示. 从图 中可以看出,采用回归模型计算的流变应力偏差范 围为 - 63. 85 ~ 42. 58 MPa,而 ANFIS 模型计算的偏 差范围为 - 8. 64 ~ 7. 94 MPa,可知 ANFIS 模型预测 值的偏差范围非常小,远小于回归模型计算的偏差 范围. 图 9 ANFIS 和回归模型预测值与试验值的偏差变化 Fig. 9 Variation of the deviation between measured values and pre￾dicted results for the ANFIS and regression model 4 结论 ( 1) 在变形温度为 25 ~ 500 ℃,应变速率为 0. 01 ~ 10 s - 1 的条件下,连续柱状晶组织 BFe10--1--1 合金压缩变形以应变硬化为主,未出现明显的再结 晶现象. ( 2) 采用自适应神经模糊推理系统建立了可反 映连续柱状晶组织 BFe10--1--1 合金在变形温度为 25 ~ 500 ℃,应变速率为 0. 01 ~ 10 s - 1 条件下的压缩 变形全过程的流变应力预测模型,具有优良的预测 精度: 流变应力预测值的平均误差为 0. 75% ,均方 根误差为 2. 13,相关系数为 0. 999 6. ( 3) 在本文研究条件下,与传统回归模型预测 结果 相 比,ANFIS 模型预测连续柱状晶组织 BFe10--1--1合金流变应力的精度( 平均误差) 提高 了 5. 53% ,具有更高的拟合能力,表明该方法适合 于金属流变应力模型的研究. 参 考 文 献 [1] Yu Y Q,Zhao F,Li W G. Manufacturing of BFe30-1-1 cupro￾nickel tubes by heated mold continuous casting. Spec Cast Nonfer￾rous Alloys,2005,25( 12) : 761 ( 余业球,赵锋,黎沃光. 用热型连铸法制造 BFe30--1--1 白铜 管坯. 特种铸造及有色合金,2005,25( 12) : 761) [2] Yu Y Q,Zhao F,Li W G. Microstructure and performance of BFe30-1-1 cupronickel tubes. Hot Work Technol,2005( 11) : 8 ( 余业球,赵锋,黎沃光. 热型连铸 BFe30--1--1 白铜管的组织 与性能. 热加工工艺,2005( 11) : 8) [3] Mandal S,Rakesh V,Sivaprasad P V,et al. Constitutive equa￾tions to predict high temperature flow stress in a Ti-modified auste￾nitic stainless steel. Mater Sci Eng A,2009,500( 1 /2) : 114 [4] Takuda H,Morishita T,Kinoshita T,et al. Modelling of formula for flow stress of a magnesium alloy AZ31 sheet at elevated temper￾atures. J Mater Process Technol,2005,164 /165: 1258 [5] Varol Y,Koca A,Oztop F,et al. Analysis of adaptive-network￾based fuzzy inference system ( ANFIS) to estimate buoyancy-in￾duced flow field in partially heated triangular enclosures. Expert Syst Appl,2008,35( 4) : 1989 [6] Hanbay D,Baylar A,Ozpolat E. Predicting flow conditions over stepped chutes based on ANFIS. Soft Comput,2009,13( 7) : 701 [7] Sfetsos A. A comparison of various forecasting techniques applied to mean hourly wind speed time series. Renew Energy,2000,21 ( 1) : 23 [8] Chang F J,Chang Y T. Adaptive neuro-fuzzy inference system for prediction of water level in reservoir. Adv Water Resour,2006,29 ( 1) : 1 [9] Wei M Z,Bai B J,Sung A H,et al. Predicting injection profiles using ANFIS. Inf Sci,2007,177( 20) : 4445 [10] Leidermark D,Moverare J J,Simonsson K,et al. Room tempera￾ture yield behaviour of a single-crystal nickel-base superalloy with tension /compression asymmetry. Comput Mater Sci,2009,47 ( 2) : 366 [11] Jang J S R. ANFIS: Adaptive-network-based fuzzy inference sys￾tem. IEEE Trans Syst Man Cybern,1993,23( 3) : 665 [12] Shen J C. Fuzzy neural networks for tuning PID controllers for plant with underdamped responses. IEEE Trans Fuzzy Syst, 2001,9( 2) : 333 [13] Yuan S C. MATLAB Language and Mechanical Engineering Ap￾plication. Beijing: China Machine Press,2008 ( 原思聪. MATLAB 语言及机械工程应用. 北京: 机械工业 出版社,2008) ·605·
<<向上翻页
©2008-现在 cucdc.com 高等教育资讯网 版权所有