摘 要
結(jié)構(gòu)可靠度的概念來(lái)源于工程實(shí)踐，在各類工程結(jié)構(gòu)中，存在著很多隨機(jī)因素的影響，如結(jié)構(gòu)的物理特性、幾何參數(shù)等結(jié)構(gòu)本身的特性及結(jié)構(gòu)所承受的荷載。因此，對(duì)結(jié)構(gòu)三維空間結(jié)構(gòu)進(jìn)行隨機(jī)分析并評(píng)價(jià)其可靠性就顯得非常有必要?；诿商乜逯苯映闃臃?，通過(guò)大量的抽樣，對(duì)結(jié)構(gòu)系統(tǒng)進(jìn)行可靠性分析，其計(jì)算精度相對(duì)較高，但對(duì)于大型復(fù)雜結(jié)構(gòu)系統(tǒng)，采用此方法費(fèi)時(shí)較多，實(shí)用性很差。因此，本文采用隨機(jī)有限元法對(duì)結(jié)構(gòu)可靠性進(jìn)行分析。
本文首先介紹了可靠性設(shè)計(jì)的國(guó)內(nèi)外研究現(xiàn)狀及發(fā)展趨勢(shì)，接著介紹了隨機(jī)有限元理論及確定性有限元分析的方法，然后闡述了神經(jīng)網(wǎng)絡(luò)及遺傳算法，并對(duì)其進(jìn)行簡(jiǎn)單運(yùn)用。在結(jié)構(gòu)可靠性分析中，對(duì)梁板結(jié)構(gòu)和船艙段結(jié)構(gòu)簡(jiǎn)例進(jìn)行分析，首先建立梁板結(jié)構(gòu)單元的安全余量，利用改進(jìn)的一次二階矩法計(jì)算各失效模式的安全余量，采用分枝限界法找出主要失效模式，利用概率網(wǎng)絡(luò)估算技術(shù)法計(jì)算結(jié)構(gòu)系統(tǒng)的失效概率并進(jìn)行敏度分析。然后將梁板結(jié)構(gòu)的可靠性分析與基于蒙特卡洛直接抽樣法的計(jì)算結(jié)果進(jìn)行比較，得出本文方法的可行性。在結(jié)構(gòu)優(yōu)化中，通過(guò)神經(jīng)網(wǎng)絡(luò)和遺傳算法，對(duì)船艙段結(jié)構(gòu)進(jìn)行優(yōu)化設(shè)計(jì)，每代遺傳操作中只需以上可靠性方法計(jì)算一次結(jié)構(gòu)系統(tǒng)可靠性指標(biāo)，將該代最優(yōu)解對(duì)應(yīng)的數(shù)據(jù)加入神經(jīng)網(wǎng)絡(luò)的訓(xùn)練樣本，從訓(xùn)練樣本中刪除最劣樣本，使樣本不斷處于更新?tīng)顟B(tài)。通過(guò)兩種算法的結(jié)合，利用隨機(jī)有限元法的基本理論，建立了一套完整的基于可靠性的隨機(jī)結(jié)構(gòu)系統(tǒng)優(yōu)化設(shè)計(jì)方法。
最后，本文開(kāi)發(fā)了基于隨機(jī)有限元的結(jié)構(gòu)可靠性分析及優(yōu)化原型系統(tǒng)，考慮材料的強(qiáng)度、梁板元的尺寸和外載荷等均為隨機(jī)變量的三維梁板空間結(jié)構(gòu)，通過(guò)幾個(gè)實(shí)例說(shuō)明了本文方法的可行性，并研究了優(yōu)化算法對(duì)優(yōu)化結(jié)果的影響，得出了一些有益的結(jié)論。
關(guān)鍵詞 隨機(jī)有限元法；分枝限界法；改進(jìn)的一次二階矩法；安全余量；可靠性優(yōu)化


Abstract
The concept of structure reliability comes from the engineering practice. In all kinds of engineering structure, there are many random factors, such as structural physical characteristics, geometric parameters which are structure characteristics of itself and the load. Therefore, it is very necessary to make reliability analysis and eva luation of the 3d spatial stochastic structure. Based on direct Monte-Carlo sampling method, the computing accuracy is relative taller for structural system reliability analysis through a lot of sampling. But for large complex structure system, using this method spends more time and the practicability is very poor. So this paper adopts stochastic finite element method to the structural reliability analysis.
This paper firstly introduces the domestic and foreign research situation and development trend of the reliability design. Then introduce the stochastic finite element theory and deterministic finite element analysis method. Expound the neural network and genetic algorithm, and carry on the simple example. In the structural reliability analysis, analyse the example of the beam and shell structure and the cabin paragraph structure. Firstly establish the unit safety margins of beam and shell structure. Calculate the various failure modes of the safety margins with using adapted first-order second-moment method. Use branch and bound method to identify major failure modes. Finally, use the failure mode correlation and PNET method to calculate the failure probability of structural systems and the sensitivity analysis. Compare the calculation results of structure reliability analysis with the results of the direct Monte-Carlo sampling method, conclude the feasibility method of this paper. In the structural optimization, use the neural network and genetic algorithm to have the optimization design of the structure of the cabin paragraph. Calculate the structural system reliability index one time with the above traditional method in each generation of genetic operations. Join the corresponding data of the generation optimal solution of in the training samples and delete the inferior samples from the training sample to update the training samples of neural network. Through the combination of two algorithms, using the basic theory of the stochastic finite element method, has established a complete method of the optimization design based on the reliability of the stochastic structural system.
Finally, this paper develops the prototype system of the reliability analysis and optimization of the structure based on stochastic finite element. Consider the random variable of the material strength, the dimensions of beam and shell units and outer loads of the 3d beam and shell spatial structure. Through several examples illustrate the feasibility of the method, and study the influence of the optimization algorithm, obtain some useful conclusions.
Keywords stochastic finite element method; branch and bound method; adapted first- order second-moment method; safety margin; reliability Optimization


目 錄
摘 要 I
Abstract II
第1章 緒論 1
1.1 選題背景、意義及實(shí)用價(jià)值 1
1.2 結(jié)構(gòu)可靠性設(shè)計(jì)國(guó)內(nèi)外研究現(xiàn)狀 1
1.3基于隨機(jī)有限元結(jié)構(gòu)優(yōu)化研究現(xiàn)狀及發(fā)展趨勢(shì) 2
1.3.1隨機(jī)有限元國(guó)內(nèi)外研究現(xiàn)狀 2
1.3.2基于可靠性優(yōu)化研究現(xiàn)狀及發(fā)展趨勢(shì) 3
1.4 論文研究?jī)?nèi)容及章節(jié)安排 5
1.4.1 論文研究?jī)?nèi)容 5
1.4.2 章節(jié)安排 5
1.5 本章小結(jié) 6
第2章 隨機(jī)有限元理論分析研究 ..