唐和生+蘇瑜+薛松濤+鄧立新
文章編號(hào):16742974(2014)04003306
收稿日期:20130417
基金項(xiàng)目:國(guó)家自然科學(xué)基金資助項(xiàng)目(51178337,50708076);科技部國(guó)家重點(diǎn)實(shí)驗(yàn)室基礎(chǔ)研究項(xiàng)目(SLDRCE11B01);同濟(jì)大學(xué)土木工程學(xué)院光華基金資助項(xiàng)目
作者簡(jiǎn)介:唐和生(1973-),男,安徽安慶人,同濟(jì)大學(xué)副教授,博士
通訊聯(lián)系人,E-mail:thstj@#edu.cn
摘要:采用證據(jù)理論作為傳統(tǒng)概率的替代方法處理不精確的數(shù)據(jù)信息,提出了基于證據(jù)理論的可靠性優(yōu)化設(shè)計(jì)方法.該方法針對(duì)給定的失效概率許用值Pf,通過(guò)計(jì)算證據(jù)理論的不確定測(cè)度Pl(F),以Pl(F)<Pf作為可靠性約束條件的替代模型進(jìn)行可靠性優(yōu)化設(shè)計(jì).為了降低基于證據(jù)理論不確定量化分析的計(jì)算成本,引入微分演化優(yōu)化算法計(jì)算區(qū)間邊界值.以典型桁架結(jié)構(gòu)形狀優(yōu)化問(wèn)題為例,驗(yàn)證了所提出方法的準(zhǔn)確性和有效性.
關(guān)鍵詞:證據(jù)理論;微分演化;可靠性優(yōu)化設(shè)計(jì);形狀優(yōu)化
中圖分類號(hào):TU318;TU323.4 文獻(xiàn)標(biāo)識(shí)碼:A
AMethodofReliabilityDesignOptimizationUsingEvidence
TheoryandDifferentialEvolution
TANGHesheng1,2,SUYu1,XUESongtao1,DENGLixin1
(1.ResearchInstituteofStructuralEngineeringandDisasterReduction,TongjiUniv,Shanghai200092,China;
2.StateKeyLaboratoryofDisasterPreventioninCivilEngineering,TongjiUniv,Shanghai200092,China)
Abstract:Anewmethodofreliabilitydesignoptimizationusingtheevidencetheorywasproposed.Evidencetheorywaspresentedasanalternativetotheclassicalprobabilitytheorytohandletheimprecisedatasituation.TheplausibilitymeasurePl(F)basedonevidencetheory,withPl(F)
Keywords:evidencetheory;differentialevolution;reliabilitydesignoptimization;shapeoptimization
基于可靠性的優(yōu)化設(shè)計(jì)是不確定性結(jié)構(gòu)優(yōu)化設(shè)計(jì)的有效途徑.傳統(tǒng)的可靠性優(yōu)化設(shè)計(jì)常采用概率模型,但概率方法需要足夠的統(tǒng)計(jì)數(shù)據(jù)信息來(lái)擬合其概率特征,而實(shí)際工程中這些數(shù)據(jù)通常是無(wú)法準(zhǔn)確獲知的,所以傳統(tǒng)概率方法面臨巨大的挑戰(zhàn).
近年來(lái),國(guó)內(nèi)外很多學(xué)者致力于發(fā)展非概率的不確定建模手段,并在其基礎(chǔ)上提出非概率可靠性優(yōu)化設(shè)計(jì)方法,其中由Dempster[1]和Shafer[2]提出的證據(jù)理論具有較強(qiáng)的不確定處理能力,已經(jīng)成為不確定信息表達(dá)和量化的有力工具,在多目標(biāo)識(shí)別、信息融合、多屬性決策等領(lǐng)域獲得了廣泛應(yīng)用[3-6].而基于證據(jù)理論的工程可靠性優(yōu)化分析剛剛起步,并且主要應(yīng)用于機(jī)械與航空領(lǐng)域.Mourelatos等[7]將證據(jù)理論應(yīng)用于內(nèi)壓容器可靠性優(yōu)化設(shè)計(jì)中,研究了基于證據(jù)理論的失效概率或可靠度指標(biāo)的計(jì)算問(wèn)題.Bae等[8]運(yùn)用證據(jù)理論解決機(jī)械工程中的不確定問(wèn)題,實(shí)現(xiàn)了飛機(jī)機(jī)翼結(jié)構(gòu)的可靠性優(yōu)化設(shè)計(jì).郭慧昕等[9-11]提出了證據(jù)理論和區(qū)間分析相結(jié)合的可靠度優(yōu)化設(shè)計(jì)方法,將此應(yīng)用于內(nèi)壓容器和氣門彈簧的可靠性研究中.盡管已經(jīng)取得了一些進(jìn)展,但是證據(jù)理論仍然很少應(yīng)用于實(shí)際工程的可靠性優(yōu)化問(wèn)題中,計(jì)算成本是導(dǎo)致該問(wèn)題的主要原因.同時(shí),基于證據(jù)理論的可靠性優(yōu)化設(shè)計(jì)在土木工程中的應(yīng)用還是一個(gè)新課題.
為此,本文采用證據(jù)理論處理不確定情況,引入微分演化算法降低證據(jù)理論在可靠性優(yōu)化中的計(jì)算成本,提出了證據(jù)理論和微分演化算法相結(jié)合的可靠性優(yōu)化設(shè)計(jì)方法,并將該方法應(yīng)用于桁架結(jié)構(gòu)的形狀優(yōu)化問(wèn)題中,來(lái)驗(yàn)證本文所提方法的有效性.
1證據(jù)理論的基本原理
證據(jù)理論是由Dempster和Shafer提出的,又稱為DS理論.它是建立在辨識(shí)框架上的一種不確定理論,設(shè)Θ為辨識(shí)框架,它表示關(guān)于命題互不相容的所有可能答案的有限集合,類似于概率論中有限的樣本空間,冪集2Θ定義為辨識(shí)框架Θ中所有子集的集合,證據(jù)理論是對(duì)冪集元素進(jìn)行基本概率賦值[12].定義函數(shù)m:2Θ→[0,1],AΘ.當(dāng)滿足:
m(Φ)=0,∑AΘm(A)=1,(1)
稱m為框架Θ上的基本信任分配函數(shù)(BBA),m(A)為A的基本信任度,表示證據(jù)對(duì)A的信任程度,若m(A)>0,則稱A為焦元.
若m為框架Θ上的基本信任分配函數(shù),則稱由
Bel(A)=∑BAm(B),(2)
Pl(A)=1-Bel()=∑B∩A≠m(B)(3)
所定義的函數(shù)Bel:2Θ→[0,1]為Θ上的信任函數(shù),函數(shù)Pl:2Θ→[0,1]為Θ上的似然函數(shù),Bel(A)表示對(duì)A為真的信任程度,Pl(A)表示對(duì)A為非假的信任程度,也稱為命題A的似然度,兩者之間的關(guān)系如圖1所示,Bel(A)和Pl(A)提供了概率P(A)的上限和下限.由此可見(jiàn),以概率論為基礎(chǔ)的傳統(tǒng)可靠性問(wèn)題只是證據(jù)理論的一個(gè)特例.
圖1對(duì)命題A的不確定描述
Fig.1UncertaintyrepresentationofpropositionA
對(duì)于認(rèn)識(shí)不夠透徹的不確定參數(shù),可能會(huì)有多個(gè)專家提出不同的理論或不同的數(shù)據(jù)來(lái)組成多源證據(jù),證據(jù)理論可以通過(guò)合成規(guī)則綜合考慮各種證據(jù)源的影響.經(jīng)典的DS合成規(guī)則為:假定m1和m2是同一辨識(shí)框架Θ上的2個(gè)基本信任分配函數(shù),焦元分別為Ai和Bj,則新的基本信任分配函數(shù)m為:
m(A)=∑Ai∩Bj=Am1(Ai)m2(Bj)1-K,A≠.(4)
式中:K=∑Ai∩Bj=m1(Ai)m(Bj),表示證據(jù)沖突性的大小.此即為2個(gè)證據(jù)合成的Dempster法則.當(dāng)證據(jù)沖突比較大時(shí),應(yīng)選用其他的合成方法[13].
2證據(jù)理論和微分演化算法相結(jié)合的可靠
性優(yōu)化設(shè)計(jì)方法
2.1基于證據(jù)理論的可靠性優(yōu)化設(shè)計(jì)模型
一般來(lái)說(shuō),基于概率理論的可靠性優(yōu)化問(wèn)題可表述為:
minF(xN,d),
s.t.P{gj(x,d)>g0}
P{gj(x,d)
j=1,2,…,k,
dL≤d≤dU,
xL≤xN≤xU.(5)
式中:x=[x1,x2,…,xn]T為n維不確定性量,分別服從一定的概率分布,[xL,xU]為不確定量x的名義值xN的取值區(qū)間;d=[d1,d2,…,dm]T為m維確定性量,[dL,dU]為d的取值區(qū)間;F和g分別為目標(biāo)函數(shù)和約束函數(shù);g(x,d)>g0表示結(jié)構(gòu)發(fā)生失效,g(x,d)<g0表示結(jié)構(gòu)安全,g0為許可的響應(yīng)值;k為約束條件的數(shù)目;P{}表示真實(shí)概率;Pf和R分別為結(jié)構(gòu)設(shè)計(jì)給定的失效概率和可靠度許用值.
當(dāng)問(wèn)題中不確定量的認(rèn)識(shí)信息較少或不完整時(shí),上述優(yōu)化模型中的約束條件不能采用概率理論來(lái)建立,此時(shí),可以利用證據(jù)理論的不確定建模手段解決這一問(wèn)題.如圖1所示,證據(jù)理論用似然函數(shù)和信任函數(shù)來(lái)進(jìn)行不確定性度量,可以證明[Bel,Pl]是真實(shí)概率的區(qū)間估計(jì),真實(shí)的失效概率或可靠度夾逼在該區(qū)間內(nèi):
Pl{gj(x,d)>g0}
P{gj(x,d)>g0}
Bel{gj(x,d)
P{gj(x,d)
由此,將Pl{gj(x,d)>g0}<pf或Bel{gj(x,d)<g0}>R作為傳統(tǒng)可靠性約束條件的替代模型,建立基于證據(jù)理論的廣義可靠度優(yōu)化設(shè)計(jì)數(shù)學(xué)模型:
minF(xN,d)
Pl{gj(x,d)>g0}
g0}>R,
j=1,2,…,k,
dL≤d≤dU,
xL≤xN≤xU.(7)
模型中Pl{gj(x,d)>g0}或Bel{gj(x,d)<g0}是進(jìn)行非概率可靠性分析的前提,由于結(jié)構(gòu)失效占整個(gè)設(shè)計(jì)空間的比例較小,故以Pl{gj(x,d)>g0}<pf作為約束條件可以提高優(yōu)化效率,下面就Pl{gj(x,d)>g0}的計(jì)算進(jìn)行詳細(xì)闡述.
2.2基于微分演化方法的失效似然度計(jì)算
對(duì)于結(jié)構(gòu)分析中出現(xiàn)的不確定量,證據(jù)理論將其表述為區(qū)間數(shù).在計(jì)算結(jié)構(gòu)失效似然度時(shí),首先根據(jù)不確定量的可能取值范圍,將其劃分為有限個(gè)互不相容的基本區(qū)間作為辨識(shí)框架.以圖2為例,不確定量x1的辨識(shí)框架X1={x11,x12,x13},冪集2X1={Ф,{x11},{x12},{x13},{x11,x12},{x11,x13},{x12,x13},X1},基于證據(jù)(專家意見(jiàn)或?qū)崪y(cè)數(shù)據(jù))分析,對(duì)冪集中的焦元進(jìn)行基本信任度賦值,得到x1的基本信任分配函數(shù)m.
圖2不確定參數(shù)基本區(qū)間
Fig.2Basicintervalsofuncertaintyparameter
然后,對(duì)每個(gè)不確定變量的焦元進(jìn)行笛卡爾運(yùn)算,得到聯(lián)合焦元區(qū)間,以二維不確定參數(shù)為例:
C=x1×x2={ck=[x1m,x2n]:x1m∈X1,
x2n∈X2}.(8)
式中:x1m,x2n和ck分別為X1,X2和C的焦元區(qū)間.考慮到x1和x2的獨(dú)立性,二維聯(lián)合焦元的基本信任度m(ck)=m(x1m)m(x2n).
由于x1m和x2n都是區(qū)間,焦元ck在集合上為一矩形,顯然,對(duì)于n維問(wèn)題,聯(lián)合辨識(shí)框架中的焦元為n維“超立方體”.令y(x,d)表示結(jié)構(gòu)極限狀態(tài)函數(shù),結(jié)構(gòu)失效域F為:
F={x:y=g0-g(x,d)<0,x=
[x1,…,xn]∈ck}.(9)
在聯(lián)合BBA和失效域F的基礎(chǔ)上,可根據(jù)式(2)和式(3)計(jì)算結(jié)構(gòu)失效測(cè)度Bel(F)和Pl(F)為:
Bel(F)=∑ckFm(Xc),
Pl(F)=∑ck∩F≠m(Xc).(10)
可見(jiàn),在計(jì)算失效測(cè)度時(shí)需要確定聯(lián)合焦元ck是否滿足ckF或ck∩F≠,圖3描述了ck對(duì)Bel(F)和Pl(F)的貢獻(xiàn).從圖3可以看出:1)若ymax>0,ymin>0,則ck∩F=,ck對(duì)Pl(F)和Bel(F)沒(méi)有貢獻(xiàn),即ck不參與Pl(F)和Bel(F)的計(jì)算.2)若ymax<0,ymin<0,則ck∩F=ck,ck對(duì)Pl(F)和Bel(F)的計(jì)算都有貢獻(xiàn).3)若ymax>0,ymin<0,ck僅對(duì)Pl(F)有貢獻(xiàn).
圖3焦元區(qū)間對(duì)失效似然度的貢獻(xiàn)
Fig.3Focalelementcontribution
toplausibilityofthefailureregion
因此,為準(zhǔn)確判斷,需要求解y(x,d)在ck對(duì)應(yīng)的“超立方體”域上的極值,即
[ymin,ymax]=[mincky(x,d),mincky(x,d)].(11)
求解式(11)中區(qū)間極值的主要方法有采樣法和優(yōu)化方法,采樣法的精度很大程度上取決于采樣點(diǎn)數(shù)目,計(jì)算代價(jià)巨大.優(yōu)化方法會(huì)極大降低計(jì)算量,但由于不確定量x的焦元區(qū)間數(shù)目多,而且結(jié)構(gòu)響應(yīng)并不是簡(jiǎn)單的顯式而是通過(guò)有限元分析得到的,故利用傳統(tǒng)的優(yōu)化算法求解復(fù)雜多維非凸的極限狀態(tài)函數(shù)y(x,d)在ck上的極值顯得非常困難.
近年來(lái)仿生智能優(yōu)化算法被廣泛引入到結(jié)構(gòu)優(yōu)化中,例如模擬退火法(SA)[14]、遺傳算法(GA)[15]、微分演化法(DE)[16]等,其中DE是一種新穎的啟發(fā)式智能算法,采用變異、交叉和選擇3項(xiàng)基本操作,通過(guò)若干代種群演化操作不斷舍棄劣質(zhì)解,保留優(yōu)質(zhì)解,最終獲取近似全局最優(yōu)解.研究表明,微分演化算法在求解非凸、多峰、非線性優(yōu)化問(wèn)題中表現(xiàn)出較強(qiáng)的穩(wěn)健性,同時(shí)具有收斂較快的優(yōu)點(diǎn)[17].因此,本文采用DE提高y(x,d)區(qū)間極值的求解速度,如圖4所示,從而減少優(yōu)化設(shè)計(jì)的計(jì)算成本.
圖4區(qū)間函數(shù)極值求解
Fig.4Intervaloptimizationforcomputingbounds
根據(jù)以上描述,基于證據(jù)理論的可靠性優(yōu)化設(shè)計(jì)是利用微分演化算法在滿足可靠性約束Pl{g(x,d)>g0}<pf的條件下,尋求結(jié)構(gòu)最優(yōu)設(shè)計(jì)變量和最優(yōu)目標(biāo)值.同時(shí),基于證據(jù)理論的結(jié)構(gòu)約束失效似然度又是通過(guò)微分演化算法來(lái)提高其計(jì)算效率的,可靠性約束條件的計(jì)算流程如圖5所示.
圖5可靠性約束計(jì)算流程
Fig.5Flowchartofcalculationofreliabilityconstraint
3算例分析
為了便于比較,取文獻(xiàn)[18]中的25桿桁架形狀優(yōu)化進(jìn)行討論,結(jié)構(gòu)形式見(jiàn)圖6,彈性模量名義值E=68950MPa,作用于桁架上的荷載名義值列于表1,容許拉壓應(yīng)力[σ]=±275.6MPa,各節(jié)點(diǎn)三向允許的最大位移為8.89mm.其他參數(shù)見(jiàn)文獻(xiàn)[18].
圖625桿空間桁架結(jié)構(gòu)
Fig.625barspacetrussstructure
表125桿桁架節(jié)點(diǎn)荷載名義值
Tab.1Normalvalueofjointloadfor25bartruss
節(jié)點(diǎn)號(hào)
Fx/kN
Fy/kN
Fz/kN
1
4.448
-44.48
-44.48
2
0
-44.48
-44.48
3
2.224
0
0
6
2.668
0
0
文獻(xiàn)[18]對(duì)該桁架進(jìn)行了確定性優(yōu)化,本文在此基礎(chǔ)上考慮不確定情況,將外荷載和彈性模量視為不確定的,假定其不確定信息(焦元區(qū)間及基本信任度)如表2所示,在2種不確定因素存在的情況下,進(jìn)行既滿足可靠度約束條件又使結(jié)構(gòu)總質(zhì)量最小的最優(yōu)設(shè)計(jì),該不確定優(yōu)化問(wèn)題的數(shù)學(xué)模型為:
findd=[A1,A2,…,A8,X4,Y4,…,Y8],
minF(d)=∑8i=1ρAiLi+λM,
s.t.Pl{gi(x,d)<0}
g1(x,d)=278.6-maxσk(x,d),
g2(x,d)=8.89-maxujl(x,d),
x=[F1x,F1y,…,F6z,E].(12)
式中:d為尺寸和形狀設(shè)計(jì)變量;x為不確定參數(shù);Pl{gi(x,d)<0}
考慮結(jié)構(gòu)允許的失效概率Pf為0.05和0.1二種情況,采用本文所提方法對(duì)25桿桁架進(jìn)行可靠性形狀優(yōu)化.Pf=0.05情況的評(píng)價(jià)函數(shù)收斂曲線和最終形狀分別見(jiàn)圖7和圖8.圖9給出最優(yōu)設(shè)計(jì)時(shí)位移約束函數(shù)g2(x,d)的信任度和似然度累計(jì)分布曲線.為了與文獻(xiàn)[18]的確定性形狀優(yōu)化結(jié)果相比較,根據(jù)圖9,表3詳細(xì)列出了應(yīng)力和位移約束失效似然度.
表2外荷載和彈性模量的不確定信息
Tab.2Theuncertaininformationofloadandelasticmodulus
F1x/kN
F1y/kN
F1z/kN
F2y/kN
區(qū)間
BPA
區(qū)間
BPA
區(qū)間
BPA
區(qū)間
BPA
[3.54.6]
0.15
[-48.9-44.5]
0.2
[-48.9-44.5]
0.2
[-48.9-44.5]
0.2
[4.24.6]
0.65
[-44.5-40.0]
0.5
[-44.5-40.0]
0.5
[-44.5-40.0]
0.5
[4.65.3]
0.2
[-40.0-33.4]
0.3
[-40.0-33.4]
0.3
[-40.0-33.4]
0.3
F2z/kN
F3x/kN
F6x/kN
E/103MPa
[-48.9-44.5]
0.2
[1.82.3]
0.15
[2.12.8]
0.15
[6065]
0.1
[-44.5-40.0]
0.5
[2.12.3]
0.65
[2.52.8]
0.65
[6570]
0.5
[-40.0-33.4]
0.3
[2.32.7]
0.2
[2.83.2]
0.2
[7080]
0.4
迭代次數(shù)
圖725桿桁架形狀優(yōu)化的收斂曲線
Fig.7Shapeoptimizationconvergence
historyof25bartruss
圖825桿桁架的形狀優(yōu)化結(jié)果
Fig.8Optimumshapeof25bartruss
由圖7可知,該算法具有很高的計(jì)算效率,25桿桁架形狀的優(yōu)化計(jì)算在迭代大約100次后已經(jīng)收斂.從圖9和表3可以看出,在Pf=0.1和Pf=0.05兩種情況下,位移約束失效概率[P(g2<0)]的不確定區(qū)間
g2(x,d)(a)Pf=0.05
g2(x,d)(b)Pf=0.1
圖9位移約束的信任度和似然度累積分布
Fig.9Cumulativebeliefandplausibility
distributionfordisplacementconstraint
[Bel,Pl]分別為[0,0.066]和[0,0.036],失效似然度即概率上界均小于相應(yīng)的失效概率許用值,滿足設(shè)計(jì)可靠度的要求.由表3可知,由于考慮不確定性的存在,基于DS可靠性優(yōu)化結(jié)果總質(zhì)量要比確定性優(yōu)化結(jié)果有所增加,但是從失效似然度來(lái)看,前者的可靠性(93.4%,96.4%)要明顯高于后者(5%).由此可見(jiàn),對(duì)不確定量的認(rèn)識(shí)信息較少,無(wú)法采用概率理論時(shí),證據(jù)理論以區(qū)間測(cè)度[Bel,Pl]代替?zhèn)鹘y(tǒng)概率單值來(lái)描述這種認(rèn)知不確定顯得更為合理.由于基于DS的可靠性優(yōu)化將使結(jié)構(gòu)具有良好的魯棒性,有效避免由于錯(cuò)誤估計(jì)而造成優(yōu)化結(jié)果的偏差.
表325桿空間桁架形狀優(yōu)化結(jié)果比較
Tab.3Comparisonofoptimaldesignsforthe25bartruss
設(shè)計(jì)變量
/mm2
文獻(xiàn)[18]
結(jié)果/mm2
本文不確定分析
結(jié)果/mm2
Pf=0.05
Pf=0.1
A1
64.5
64.5
64.5
A2
64.5
64.5
64.5
A3
645
774.2
774.2
A4
64.5
64.5
64.5
A5
64.5
64.5
64.5
A6
64.5
64.5
64.5
A7
64.5
64.5
64.5
A8
580.6
838.7
709.6
X4
949.9
984.6
961.1
Y4
1406.6
1770.5
1375.6
Z4
3283.9
2599.1
3283.7
X8
1315.9
1322.5
1468.0
Y8
3544.8
3400.5
3533.2
總質(zhì)量/kg
Pl(g1<0)
Pl(g2<0)
53.1
0
0.95
66.5
0
0.036
62.2
0
0.066
4結(jié)論
可靠性優(yōu)化設(shè)計(jì)中,由于不確定信息較少無(wú)法構(gòu)造精確概率分布時(shí),證據(jù)理論代替?zhèn)鹘y(tǒng)的概率理論進(jìn)行不確定信息描述是一種理想的選擇.該方法用不確定區(qū)間測(cè)度[Bel,Pl]代替不可知的真實(shí)概率來(lái)處理不完備的數(shù)據(jù)信息,以Pl(F)<Pf作為概率可靠性約束條件的替代模型,本文同時(shí)引入微分演化算法提高求解聯(lián)合焦元內(nèi)反應(yīng)區(qū)間極值的計(jì)算效率以及可靠性優(yōu)化的尋優(yōu)速度,降低證據(jù)理論在可靠性優(yōu)化中的計(jì)算成本.
本文以25桿桁架形狀優(yōu)化為例,在考慮荷載和彈性模量均為不確定的情況下,基于DS進(jìn)行可靠性優(yōu)化設(shè)計(jì)得到了很好的結(jié)果.分析結(jié)果表明,本文所提方法在實(shí)際工程中具有一定的應(yīng)用前景.
參考文獻(xiàn)1] DEMPSTER A P. Upper and lower probabilities induced by a multiplicand mapping[J]. Annals of Mathematical Statistics, 1967,38:325-339.[2] SHAFER G. A mathematical theory of evidence [M]. Princeton: Princeton University Press, 1976:9-50.[3] MCCLEAN S, SCOTNEY B. Using evidence theory for the integration of distributed database[J]. International Journal of Intelligent Systems, 1997,12(10):763-776.[4] FRANCOIS J, GRANDVALET Y, DENOEUX T, et al. Addendum to resample and combine: an approach to improving uncertainty representation in evidential pattern classification[J]. Information Fusion, 2003(4):75-85.[5] 張曉明,王航宇,黃達(dá).基于DS證據(jù)理論的多平臺(tái)協(xié)同數(shù)據(jù)融合[J].計(jì)算機(jī)工程,2007,33(11):242-243.ZHANG Xiaoming, WANG Hangyu, HUANG Da. Cooperative multiplatform data fusion based on DS evidence theory[J]. Computer Engineering, 2007,33(11):242-243.(In Chinese)[6] 郝軍,吳煒,楊棧.改進(jìn)DS算法在船舶汽輪機(jī)故障診斷中的應(yīng)用研究[J].船海工程,2007,36(3):49-51.HAO Jun, WU Wei, YANG Zhan. Application of the improved DS algorithm in fault diagnosis of ships turbinemachinery[J]. Ship & Ocean Engineering, 2007,36(3):49-51.(In Chinese)[7] MOURELATOS Z P, ZHOU J. A design optimization method using evidence theory[J]. Journal of Mechanical Design, ASME, 2006,128(7):901-908.[8] BAE H R, GRANDHI R V, CANFIELD R A. Reliabilitybased design optimization under imprecise uncertainty[C]//46th Structures, Structural Dynamics and Materials Conference, AIAA/ASME/ASCE/AHS/ASC, 2005:3301-3313.[9] 郭慧昕,劉德順,胡冠昱,等.證據(jù)理論和區(qū)間分析相結(jié)合的可靠性優(yōu)化設(shè)計(jì)方法[J].機(jī)械工程學(xué)報(bào),2008,44(12):36-41.GUO Huixin, LIU Deshun, HU Guanyu,et al. Method of reliability design optimization using evidence theory and interval analysis[J]. Chinese Journal of Mechanical Engineering, 2008,44(12):36-41.(In Chinese)[10]郭慧昕,夏立龍,戴娟.基于證據(jù)理論的結(jié)構(gòu)失效概率計(jì)算方法[J].應(yīng)用基礎(chǔ)與工程科學(xué)學(xué)報(bào),2008, 16(3):457-464.GUO Huixi, XIA Lilong, DAI Juan. Method for calculating the failure probability of engineering structure using evidence theory[J]. Journal of Basic Science and Engineering, 2008,16(3):457-464.(In Chinese)
[11]任丕順.基于證據(jù)理論的氣門彈簧可靠性設(shè)計(jì)優(yōu)化[J].裝備制造技術(shù),2012(5):49-52.REN Pishun. The reliability optimal design of the valve spring based on evidence theory[J]. Equipment Manufacturing Technology, 2010(5):49-51.(In Chinese)[12]段新生.證據(jù)理論與決策、人工智能[M].北京:中國(guó)人民大學(xué)出版社,1993:13-32.DUAN Xinsheng. Evidence theory and decision & artificial intelligence [M]. Beijing: China Renmin University Press, 1993:13-32. (In Chinese)[13]SENTZ K, FERSON S. Combination of evidence in dempstershafer theory, SAND20020835 report[R]. Sandia: Sandia National Laboratories, 2002.[14]BALLING R. Optimal steel frame design by simulated annealing[J]. Journal of Structural Engineering, 1991,117(6):1780-1795.[15]GOLDBERG D, KUO C. Genetic algorithms in pipeline optimization[J]. Journal of Computing in Civil Engineering, 1987,1(2):128-141.[16]唐和生,范德偉,王兆亮,等.桁架尺寸優(yōu)化微分演化算法[J].湖南大學(xué)學(xué)報(bào):自然科學(xué)版,2011,38(11):13-18.TANG Hesheng, FAN Dewei, WANG Zhaoliang, et al. Differential evolution algorithm to size the optimization of truss structures[J]. Journal of Hunan University: Natural Sciences, 2011,38(11):13-18.(In Chinese)[17]STORN R, PRICE K. Differential evolution—a simple and efficient adaptive scheme for global optimization over continuous spaces[J]. Journal of Global Optimization, 1997,11(4):341-359.[18]GHOLIZADEH S, BARATI H. A comparative study of three metaheuristics for optimum design of trusses[J]. International Journal of Optimization in Civil Engineering, 2012,3:423-441.endprint
[11]任丕順.基于證據(jù)理論的氣門彈簧可靠性設(shè)計(jì)優(yōu)化[J].裝備制造技術(shù),2012(5):49-52.REN Pishun. The reliability optimal design of the valve spring based on evidence theory[J]. Equipment Manufacturing Technology, 2010(5):49-51.(In Chinese)[12]段新生.證據(jù)理論與決策、人工智能[M].北京:中國(guó)人民大學(xué)出版社,1993:13-32.DUAN Xinsheng. Evidence theory and decision & artificial intelligence [M]. Beijing: China Renmin University Press, 1993:13-32. (In Chinese)[13]SENTZ K, FERSON S. Combination of evidence in dempstershafer theory, SAND20020835 report[R]. Sandia: Sandia National Laboratories, 2002.[14]BALLING R. Optimal steel frame design by simulated annealing[J]. Journal of Structural Engineering, 1991,117(6):1780-1795.[15]GOLDBERG D, KUO C. Genetic algorithms in pipeline optimization[J]. Journal of Computing in Civil Engineering, 1987,1(2):128-141.[16]唐和生,范德偉,王兆亮,等.桁架尺寸優(yōu)化微分演化算法[J].湖南大學(xué)學(xué)報(bào):自然科學(xué)版,2011,38(11):13-18.TANG Hesheng, FAN Dewei, WANG Zhaoliang, et al. Differential evolution algorithm to size the optimization of truss structures[J]. Journal of Hunan University: Natural Sciences, 2011,38(11):13-18.(In Chinese)[17]STORN R, PRICE K. Differential evolution—a simple and efficient adaptive scheme for global optimization over continuous spaces[J]. Journal of Global Optimization, 1997,11(4):341-359.[18]GHOLIZADEH S, BARATI H. A comparative study of three metaheuristics for optimum design of trusses[J]. International Journal of Optimization in Civil Engineering, 2012,3:423-441.endprint
[11]任丕順.基于證據(jù)理論的氣門彈簧可靠性設(shè)計(jì)優(yōu)化[J].裝備制造技術(shù),2012(5):49-52.REN Pishun. The reliability optimal design of the valve spring based on evidence theory[J]. Equipment Manufacturing Technology, 2010(5):49-51.(In Chinese)[12]段新生.證據(jù)理論與決策、人工智能[M].北京:中國(guó)人民大學(xué)出版社,1993:13-32.DUAN Xinsheng. Evidence theory and decision & artificial intelligence [M]. Beijing: China Renmin University Press, 1993:13-32. (In Chinese)[13]SENTZ K, FERSON S. Combination of evidence in dempstershafer theory, SAND20020835 report[R]. Sandia: Sandia National Laboratories, 2002.[14]BALLING R. Optimal steel frame design by simulated annealing[J]. Journal of Structural Engineering, 1991,117(6):1780-1795.[15]GOLDBERG D, KUO C. Genetic algorithms in pipeline optimization[J]. Journal of Computing in Civil Engineering, 1987,1(2):128-141.[16]唐和生,范德偉,王兆亮,等.桁架尺寸優(yōu)化微分演化算法[J].湖南大學(xué)學(xué)報(bào):自然科學(xué)版,2011,38(11):13-18.TANG Hesheng, FAN Dewei, WANG Zhaoliang, et al. Differential evolution algorithm to size the optimization of truss structures[J]. Journal of Hunan University: Natural Sciences, 2011,38(11):13-18.(In Chinese)[17]STORN R, PRICE K. Differential evolution—a simple and efficient adaptive scheme for global optimization over continuous spaces[J]. Journal of Global Optimization, 1997,11(4):341-359.[18]GHOLIZADEH S, BARATI H. A comparative study of three metaheuristics for optimum design of trusses[J]. International Journal of Optimization in Civil Engineering, 2012,3:423-441.endprint