摘要:壓力容器的設(shè)計(jì)具有冗雜性,設(shè)計(jì)人員需要嚴(yán)格按照相關(guān)法規(guī)及設(shè)計(jì)標(biāo)準(zhǔn)確保壓力容器的穩(wěn)定性和安全性。本文提出了一種雙層循環(huán)蟻群算法,將壓力容器的設(shè)計(jì)過程簡化為求解一個(gè)多目標(biāo)優(yōu)化問題,在滿足GB/T 150—2011《壓力容器》和GB/T 19905-2017《液化氣體汽車罐車》設(shè)計(jì)要求的前提下,高效地搜索全局最優(yōu)解。
關(guān)鍵詞:結(jié)構(gòu)可靠度優(yōu)化;壓力容器;雙層循環(huán)蟻群算法
中圖分類號:U469.7 收稿日期:2023-07-22
DOI:10.19999/j.cnki.1004-0226.2023.10.011
1 前言
結(jié)構(gòu)可靠度優(yōu)化算法是對目標(biāo)函數(shù)和約束條件從概率意義上進(jìn)行度量,用概率約束代替?zhèn)鹘y(tǒng)的確定性約束[1-3]。一般來說,優(yōu)化模型定義如下:
其中,[C(d,μ)x]和[d]為目標(biāo)函數(shù)和設(shè)計(jì)變量,[μx]為隨機(jī)變量x的均值。[Gi]為第i個(gè)功能函數(shù),[Pf]為功能函數(shù)的失效概率,[dL]和[du]分別為[d]的上下邊界。
在結(jié)構(gòu)可靠度優(yōu)化算法中,概率約束的處理直接影響著優(yōu)化方法的計(jì)算精度、計(jì)算效率和收斂性[4-6]。雙層循環(huán)算法[7-9]將可靠度的計(jì)算流程分為雙層優(yōu)化循環(huán),內(nèi)循環(huán)求解可靠度指標(biāo),外循環(huán)收斂目標(biāo)值。Nikolaidis和Burtissc[10]提出了可靠度指標(biāo)法,使用可靠度指標(biāo)約束代替概率約束,將求解失效概率轉(zhuǎn)化為求解最可能失效點(diǎn),降低了可靠度優(yōu)化的求解難度。解耦法[11-13]將優(yōu)化問題中的可靠度分析循環(huán)與確定性循環(huán)分離,將其轉(zhuǎn)化為一系列確定性子優(yōu)化問題,循環(huán)求解,直至找到最優(yōu)解。單層循環(huán)方法[14-16]采用Karush-Kuhn-Tucker條件代替原可靠度分析過程,由KKT條件推導(dǎo)出近似可靠度信息有效地減少了可靠度優(yōu)化的計(jì)算成本。
盡管這些經(jīng)典的優(yōu)化算法得到了廣泛的應(yīng)用,但它們并不適合解決一些復(fù)雜的優(yōu)化問題[17-19]。本文參考蟻群的覓食行為,對經(jīng)典算法進(jìn)行改進(jìn),提出了一種具有全域搜索能力的概率優(yōu)化算法,即雙層循環(huán)蟻群算法(Double-loop ant-colony algorithm,DAA)。DAA算法使用設(shè)計(jì)域中的樣本點(diǎn)來模擬蟻群覓食時(shí)的移動行為。其中,內(nèi)層循環(huán)使用改進(jìn)的均值方法求解功能度量約束更新每只螞蟻的位置,外層循環(huán)基于兩個(gè)目標(biāo)值的差值作為停止判據(jù),保證目標(biāo)值的準(zhǔn)確性。所提DAA算法可以有效地避免局部最優(yōu)解的干擾,在全域有效地搜尋最優(yōu)值。
2 DAA算法
本文所提出的DAA算法基于以下三個(gè)假設(shè):
a.螞蟻沒有性別區(qū)分,螞蟻只會被同伴所釋放的信息素所吸引。
b.螞蟻發(fā)現(xiàn)食物時(shí),會釋放信息素,螞蟻距離食物越近,所釋放的信息素越強(qiáng)。
c.螞蟻如果沒有接收到同伴所釋放的信息素,將會在域內(nèi)隨機(jī)移動。
其中,螞蟻由域中的樣本點(diǎn)xi=(x1,x2,…,xu)T,i=1,…,n表示,螞蟻所釋放信息素的強(qiáng)度由C(xi)計(jì)算,信息素對同伴的吸引力[θ]計(jì)算如下:
式中,[θ0]為r=0時(shí)的最大吸引力,螞蟻xi與螞蟻xj之間的距離為笛卡爾距離:
在第k次迭代時(shí),螞蟻[xi]的位置更新可以表示為:
式中,[α]是隨機(jī)參數(shù),[εki]是高斯分布中抽取的隨機(jī)向量。
3 DAA算法優(yōu)化模型
在DAA算法中,將螞蟻的位置[xk+1i]看作為最小性能目標(biāo)點(diǎn),可以得到該算法的可靠度優(yōu)化模型:
上述所提的DAA算法通過求解逆可靠度問題把可靠度指標(biāo)約束轉(zhuǎn)化為功能度量約束,將螞蟻在初始域中的位置[xi]轉(zhuǎn)換為標(biāo)準(zhǔn)域中的位置[μi],可以得到以下優(yōu)化列式:
式中,[n(uki)]和[?uiG(d,uki)]分別表示功能度量函數(shù)在第k次迭代時(shí)的方向向量和靈敏度向量。
該算法的可靠度優(yōu)化模型同樣使用罰函數(shù)法將非線性約束轉(zhuǎn)化為一個(gè)更容易求解的無約束優(yōu)化問題。對于不等式約束[Gi(x)]和非線性等式約束[Hj(x)],響應(yīng)函數(shù)[G(x)]可以定義為:
式中,[c(x)]為優(yōu)化問題的目標(biāo)值;[ri]和[tj]是懲罰項(xiàng)的系數(shù),它們的值取決于具體的優(yōu)化問題。
對于DAA來說,每只螞蟻的移動速度具有一定的隨機(jī)性。為保證目標(biāo)值的準(zhǔn)確性,每隔20次選取螢火蟲種群的最優(yōu)位置作為當(dāng)前最優(yōu)目標(biāo)值,并且根據(jù)連續(xù)兩次當(dāng)前最優(yōu)目標(biāo)值的差值作為結(jié)束循環(huán)的判定判據(jù):
一般來說,DAA算法使用50只螞蟻就可以解決大多數(shù)優(yōu)化問題。上述所提出方法的步驟,可以總結(jié)為圖1所示的偽代碼。
RBDO approach using DAA
Step 1 Set the parameters of DAA
Input the boundaries of the design variables
Determine the convergence precision [ε]
Iteration counters k=0 and l=0
Generate the initial locations of n ants
Step 2 While (k<20)
Evaluate the solution for all n ants using AMV
Determine the light intensity C for each ant by objective function C(x)
For i=1∶n all n ants
For j=1∶i all n ants
If (Cj>Ci)
Move ant xi towards xj
End if
Attractiveness varies with distance r via
[exp-γr2]
End for j
End for i
Rank the ants and find the current best
End while
Step 3 Output the current best objective value [Cl]
l=l+1
If l=1
Return to Step 2
Else
If [Cl-Cl-1/Cl≤ε]
Output the best objective value [C?=Cl]
Else
Return to Step 2
End
4 DAA算法在壓力容器設(shè)計(jì)中的應(yīng)用
先使用一個(gè)簡單的數(shù)值算例對所提DAA算法進(jìn)行說明。這個(gè)例子有兩個(gè)獨(dú)立的隨機(jī)變量和三個(gè)概率約束。隨機(jī)變量服從[σx=σy=10]的正態(tài)分布,目標(biāo)可靠度指標(biāo)[βt=4]。
根據(jù)公式(9),螞蟻的覓食范圍為[x∈(-400,300)],[y∈(-400,300)],蟻群數(shù)量為50只。每只螞蟻移動200次,所尋得的食物位置為(237.908,11.820)T。
現(xiàn)將壓力容器的設(shè)計(jì)過程簡化為求解一個(gè)多目標(biāo)優(yōu)化問題。假設(shè)壓力容器采用Q420R板材,抗拉強(qiáng)度Rm=655 MPa,屈服強(qiáng)度Rel=420 MPa,可以得到板材的許用應(yīng)力為:
根據(jù)客戶的使用需求,該罐體的設(shè)計(jì)壓力為0.7 MPa,焊接接頭系數(shù)[?=1],板材負(fù)偏差c1=0.3 mm,腐蝕余量c2=1 mm。根據(jù)GB 19905-2017耐壓試驗(yàn)要求,罐體的液壓試驗(yàn)需滿足公式(11):
此外,筒體的計(jì)算應(yīng)力[σT]還需滿足公式(12):
其中,圓筒的內(nèi)直徑[Di=2 400 mm],液壓試驗(yàn)壓力[pT]可以由公式(13)計(jì)算:
根據(jù)GB/T 150.3-2011,設(shè)計(jì)溫度下圓筒的最大允許工作壓力[pw]需滿足公式(14):
由上面公式可以得到壓力容器的優(yōu)化模型如下:
使用DAA算法對公式(15)進(jìn)行求解,蟻群數(shù)量為50只,得到筒體的有效厚度為4.70 mm,筒體的計(jì)算厚度4.21 mm,液壓試驗(yàn)壓力下圓筒應(yīng)力為232.80 MPa。將所求結(jié)果代入公式(11)、公式(12)和(14)中進(jìn)行校核,所求結(jié)果滿足相關(guān)法規(guī)設(shè)計(jì)標(biāo)準(zhǔn)及客戶的使用需求。
5 結(jié)語
本文通過對雙層循環(huán)蟻群算法的研究,及在相關(guān)壓力容器計(jì)算中的舉例應(yīng)用,論證了算法與標(biāo)準(zhǔn)規(guī)定公式的一致性,探索科學(xué)算法在罐車復(fù)雜計(jì)算中應(yīng)用的可能性,為技術(shù)研發(fā)與創(chuàng)新提供了一種全新的研究方案。
參考文獻(xiàn):
[1]L. Ungki , K. Namwoo and L. Ikjin, “Selection of Optimal Target Reliability in RBDO through Reliability-Based Design for Market Systems (RBDMS) and Application to Electric Vehicle Design,” Structural and Multidisciplinary Optimization, 60(3), pp. 949-963, March 2019.
[2]J. Romero and N. V. Queipo, “An ANOVA Approach for Accounting for Life-Cycle Uncertainty Reduction Measures in RBDO: The FSAE Brake Pedal Case Study,” Structural and Multidisciplinary Optimization, 57(6), pp. 2109-2125, June 2018.
[3]Z. Meng, D. Q. Zhang and G. Li, “An Importance Learning Method for Non-Probabilistic Reliability Analysis and Optimization,” Structural and Multidisciplinary Optimization, 59(4), pp. 1255-1271, April 2019.
[4]J. H. Chun, J. H. Song and G. H. Paulino, “Structural Topology Optimization under Constraints on Instantaneous Failure Probability,” Structural and Multidisciplinary Optimization, 53(4), pp.773-799, April 2016.
[5]Z. Meng and H.L. Zhou, “New Target Performance Approach for a Super Parametric Convex Model of Non-Probabilistic Reliability-Based Design Optimization,” Computer Methods in Applied Mechanics and Engineering, 339(9), pp. 644-662, September 2018.
[6]J. C. Barbosa, H. S. Bernardino and J. S. Angelo, “An Improved Differential Evolution Algorithm for Optimization Including Linear Equality Constraints,” Memetic Computing, 11(3), pp.317-329, March 2019.
[7]W. Y. Liu and L. N. Xing, “The Double Layer Optimization Problem to Express Logistics Systems snd Its Heuristic Algorithm,” Expert Systems with Applications, 41(1), pp. 237-245, January 2014.
[8]V. T. Dao, H. Ishii and Y. Hayashi, “Double-Layer Optimization of Home Energy Management Systems with Volt-Watt Functions,” IEEJ Transactions on Electrical and Electronic Engineering, 14(5), pp. 705-715, May 2019.
[9]Z. Rosario, R. W. Fenrich and G. Iaccarino, “Cutting the Double Loop: Theory and Algorithms for Reliability‐Based Design Optimization with Parametric Uncertainty,” International Journal for Numerical Methods in Engineering, 118(12), pp. 718-740, June 2019.
[10]Z. Meng, H. L. Zhou, G. Li and D. X. Yang, “A Decoupled Approach for Non-Probabilistic Reliability-Based Design Optimization,” Computers and Structures, 175(7), pp. 65-73, July 2016.
11]E. Nikolaidis, R. Curdisso, “Reliability-Based Optimization: A Safety Index Approach,” Computers and Structures, 28(6), pp. 781-788, June 1988.
12]R. G. Mohammad, V. C. Charles and D. Babak, “Novel Decoupled Framework for Reliability-Based Design Optimization of Structures Using a Robust Shifting Technique,” Frontiers of Structural and Civil Engineering, 13(4), pp. 800-820, April 2019.
[13]T. Chatterjee, R. Chowdhury and P. Ramu, “Decoupling Uncertainty Quantification from Robust Design Optimization,” Structural and Multidisciplinary Optimization, 59(6), pp. 1969-1990, June 2019.
[14]Z. Meng, D. X. Yang and H. L. Zhou, “Convergence control of single loop approach for reliability-based design optimization,” Structural and Multidisciplinary Optimization, 57(3), pp.1079-1091, March 2018.
[15]S. B. Jeong and G. J. Park, “Single Loop Single Vector Approach Using the Conjugate Gradient in Reliability Based Design Optimization,” Structural and Multidisciplinary Optimization, 55(4), pp. 1329-1344, April 2017.
[16]B. Keshtegar and H. Peng, “Enhanced Single-Loop Method For Efficient Reliability-Based Design Optimization With Complex Constraints,” Structural and Multidisciplinary Optimization, 57(4), pp. 1731-1747, April 2018.
[17]L. Chen, H. L. Liu, K.C. Tan and Y. M. Cheng, “Evolutionary Many-Objective Algorithm Using Decomposition-Based Dominance Relationship,” IEEE Transactions on Cybernetics, 49(12), pp. 4129-4139, December 2019.
[18]M. H. Ahmadi, M. A. Ahmadi and S. A. Sadatsakkak, “Thermodynamic Analysis and Performance Optimization of Irreversible Carnot Refrigerator by Using Multi-Objective Evolutionary Algorithms (Moeas),” Renewable and Sustainable Energy Reviews, 51(11), pp. 1055-1070, November 2015.
[19]A. L. Bustamante, J. M. Molina and M. A. Patricio, “MIJ2K Optimization Using Evolutionary Multiobjective Optimization Algorithms,” Expert Systems with Applications, 38(9), pp. 10999-11010, September 2011.
作者簡介:
王永彬,1980年生,男,工程師,研究方向?yàn)楣奘杰囕v結(jié)構(gòu)優(yōu)化研發(fā)。