基于熵理论的混凝土坝变形安全监控SVM模型

为解决利用混凝土坝安全监测全序列数据建立的支持向量机(SVM)模型存在结构复杂、计算工作量大等问题,提出利用熵理论选择具有代表性样本代替全序列样本进行建模,即通过建立外部档案,根据外部档案更新算法选择具有代表性的样本,然后将外部档案的样本用作支持向量机的训练样本。将该方法用于某蓄水初期的混凝土坝变形模型的构建中,结果表明,该组合算法在保证模型精度的同时有效降低了模型的复杂程度,减少了模型的训练时间,且使模型的泛化能力得到一定的提升。     

基于ABCA-LSSVM的复杂工程结构可靠度计算

针对复杂工程结构的可靠度计算中难以精确给出显式极限状态函数,提出了一种用于复杂工程结构可靠度计算的人工蜂群最小二乘支持向量机方法(ABCA-LSSVM),利用LSSVM建立映射关系替代有限元计算,使计算效率大幅提高;引入ABCA用于LSSVM模型参数γ和σ的优化,提高了预测精度.重力坝抗拉强度可靠度分析实例验证了本文方法的有效性.     

基于LSSVM-RSM的边坡可靠度研究

由于岩土体材料复杂多变,边坡稳定的功能函数往往具有高度非线性,常规的二次多项式响应面法(RSM)往往很难对其拟合。对此,利用最小二乘支持向量(LSSVM)的小样本优良的拟合能力,提出了基于LSSVM-RSM的可靠度分析方法对边坡稳定的功能函数进行拟合,并通过网格搜索法优化LSSVM参数,最后通过两个算例进行对比验证。计算结果表明,与传统二次多项式响应面法相比,该方法能更好地拟合抗剪强度参数与边坡稳定性之间的非线性关系,有效地扩展了响应面法在边坡稳定的可靠度分析中的应用。     

在役钢筋混凝土渡槽非概率时变可靠度分析

针对渡槽碳化概率可靠度分析需要大尺度的样本空间及统计变量特征难以获取等问题,基于区间模型非概率可靠度理论与混凝土碳化理论,提出了在役钢筋混凝土渡槽碳化的非概率时变可靠度计算方法,通过实例分析并与规范做比较,验证了计算模型的正确性。分别采用所提计算方法和Monte-Carlo概率可靠度法对既有渡槽结构进行可靠度计算。结果表明,非概率可靠度指标计算值较概率可靠度指标计算值小,该方法会使渡槽结构留有一定的冗余度;所提计算模型能简便有效地对渡槽结构做出碳化耐久性分析,弥补了概率可靠度分析的不足,研究结果对工程实践具有一定的指导意义。     

基于改进偏互信息的火电厂SCR脱硝系统建模

由于传统偏互信息(PMI)法利用条件期望计算回归值时计算误差较大,会对筛选正确率产生不利影响,将插值法应用于回归值的计算,提出了改进偏互信息(IPMI)法并利用Benchmark验证方法的有效性.将IPMI法应用于火电厂SCR脱硝系统,选取出最优变量集作为支持向量机(SVM)模型的输入,并利用现场历史数据建立了IPMI-SVM模型,将其与传统SVM方法进行了对比.结果表明:IPMI-SVM模型具有较高的预测精度,IPMI法可有效提高模型的泛化能力和鲁棒性,降低了模型复杂度.     

最小二乘支持向量机在煤着火特性研究中的应用

为避免煤热失重反应静态燃烧过程获取的燃烧参数与实际锅炉动态燃烧之间的误差,以及解决沉降炉实验工作量大、难以对所有煤种进行试验的问题,利用最小二乘支持向量机建立了煤粉着火特性模型,通过对热失重反应过程的机理分析确定最佳输入变量,并进行仿真实验.结果表明:利用热失重反应过程数据预测煤着火性能具有较高的预测精度,可以在实际中应用,指导大型电站锅炉及燃烧装置的设计和运行.     

基于小波包与LSSVM的短期光伏输出功率预测研究

由于太阳辐照度及其他气象会随时发生变化,导致光伏电站输出功率具有可变性和不确定性,这将会对电网的安全运行造成重大影响。文章研究了影响光伏电站输出功率的几种气象因素,提出了一种基于小波包与最小二乘支持向量机(LSSVM)的短期光伏电站输出功率预测方法。首先,利用小波包将原始光伏电站输出功率,以及太阳辐照度、环境温度、环境湿度等气象因素进行分解,得到基频信号和多层高频信号;然后,利用最小二乘支持向量机所具有的处理小样本数据和解决非线性函数的能力,将得到的基频信号和多层高频信号作为最小二乘支持向量机的输入变量;最后,将不同尺度的输出结果进行叠加、合成,得到原始光伏电站输出功率的预测值。仿真结果表明,与传统的最小二乘支持向量机预测法、BP神经网络预测法,以及EMD与LSSVM相结合的预测方法相比,文章预测方法的预测精度较高,可以有效地预测光伏电站输出功率。     

积分随机有限元方法在汽轮机转子计算中的应用

针对结构的随机问题,将Laguerre积分法和有限元相结合,并考虑结构的物性参数和载荷参数等多变量的随机性及其相关性,推导出结构应力应变均值和方差的计算式。利用ANSYS的二次开发语言———APDL进行编程,实现结构全节点的应力应变响应均值和方差的计算,并将计算结果云图显示。最后,通过对汽轮机转子进行计算分析,并与Monte-Carlo法相对比,验证了该方法的正确性和快速性。     

汽轮机调节级叶片三维接触有限元强度可靠性分析

汽轮机运行转速通常存在波动,叶片材料特性也存在不确定性。为评估调节级叶片强度可靠度,假设汽轮机转速、叶片材料弹性模量及密度服从正态分布,建立三维有限元接触模型,自编程序完成了调节级叶片多个样本的静态应力分析,然后应用多项式响应面法拟合以及Monte Carlo模拟方法得到了叶片可靠度。建立的叶片强度可靠性分析方法具有计算效率和精度高的优点,分析结果对保证叶片安全性具有重要参考意义。     

基于随机森林的火电机组SCR脱硝反应器建模

火电机组选择性催化还原技术(SCR)脱硝反应器过程复杂多变,采用机理建模的SCR脱硝反应器出口NO_x质量浓度预测难以取得良好的效果。基于火电厂的历史运行数据,将主成分分析(PCA)和随机森林(RF)相结合建立了SCR脱硝反应器出口NO_x质量浓度预测模型。在建模过程中,采用主成分分析方法计算各个变量的贡献率来筛选变量,进而对随机森林模型进行试验验证,并与支持向量机(SVM)模型和BP神经网络模型的预测性能进行对比。结果表明:采用PCA变量选择方法确定SCR系统模型的输入变量是可行和有效的;与SVM和BP神经网络模型相比,RF算法得到的SCR系统模型具有更好的预测效果。     

Probability strength design of steam turbine blade and sensitivity analysis with respect to random parameters based on response surface method

Many stochastic parameters have an effect on the reliability of a steam turbine blade during practical operation. To improve the reliability of blade design, it is necessary to take these stochastic parameters into account. An equal cross-section blade is investigated and a finite element model is built parametrically. Geometrical parameters, material parameters and load parameters of the blade are considered as input random variables while the maximum deflection and maximum equivalent stress are output random variables. Analysis file of the blade is compiled by deterministic finite element method and applied to be loop file to create sample points. A quadratic polynomial with cross terms is chosen to regress these samples by step-forward regression method and employed as a surrogate of numerical solver to drastically reduce the number of solvers call. Then, Monte Carlo method is used to obtain the statistical characteristics and cumulative distribution function of the maximum deflection and maximum equivalent stress of the blade. Probability sensitivity analysis, which combines the slope of the gradient and the width of the scatter range of the random input variables, is applied to evaluate how much the output parameters are influenced by the random input parameters. The scatter plots of structural responses with respect to the random input variables are illustrated to analyze how to change the input random variables to improve the reliability of the blade. The results show that combination of the finite element method, the response surface method and Monte Carlo method is an ideal way for the reliability analysis and probability strength design of the blade.     

Probability strength design of steam turbine blade and sensitivity analysis with respect to random parameters based on response surface method

Many stochastic parameters have an effect on the reliability of a steam turbine blade during practical operation. To improve the reliability of blade design, it is necessary to take these stochastic parameters into account. An equal cross-section blade is investigated and a finite element model is built parametrically. Geometrical parameters, material parameters and load parameters of the blade are considered as input random variables while the maximum deflection and maximum equivalent stress are output random variables. Analysis file of the blade is compiled by deterministic finite element method and applied to be loop file to create sample points. A quadratic polynomial with cross terms is chosen to regress these samples by step-forward regression method and employed as a surrogate of numerical solver to drastically reduce the number of solvers call. Then, Monte Carlo method is used to obtain the statistical characteristics and cumulative distribution function of the maximum deflection and maximum equivalent stress of the blade. Probability sensitivity analysis, which combines the slope of the gradient and the width of the scatter range of the random input variables, is applied to evaluate how much the output parameters are influenced by the random input parameters. The scatter plots of structural responses with respect to the random input variables are illustrated to analyze how to change the input random variables to improve the reliability of the blade. The results show that combination of the finite element method, the response surface method and Monte Carlo method is an ideal way for the reliability analysis and probability strength design of the blade. __________ Translated from Proceedings of the CSEE, 2007, 27(20): 12–17 [译自: 中国电机工程学报]     

高温高压容器不连续区可靠性分析

采用ANSYS对压力容器不连续区在内压和温度共同作用下进行了应力评定以及采用蒙特卡罗法分析了输入随机变量对不连续区的可靠性的影响。结果表明由于不连续区的应力集中很大,一次加二次应力程度较高。得出了在温度影响下高温高压容器内外壁应力的变化情况以及工作压力和屈服强度对不连续区的可靠性影响较大,是高温高压容器不连续区结构失效的主要因素。     

A probabilistic structure life prediction system for container ship repair and inspection

A probabilistic fracture mechanics code has been developed to assess the impact of welding procedure, inspection frequency, and flaw detection/rejection criteria, on the reliability of butt weld joints in container ship deck doubler plates. A fatigue loading stress spectrum was developed based on statistical sea state data and a structural model of a ship. An initial flaw size distribution was developed from ultrasonic inspection results of deck doubler butt welds. Other random variables included fracture toughness and inspection detection of weld defects. Probability distributions of the input random variables were applied through Monte Carlo simulation to a deterministic fracture mechanics model for fatigue crack growth. This paper describes the development and application of the analysis program.     

蒙特卡罗法在水闸闸室可靠度分析中的应用

针对闸室可靠度分析采用蒙特卡罗法进行数值模拟,结果表明对于闸室抗滑稳定功能函数,不同变量的均值、变异系数、分布类型对其可靠度指标有不同的影响.     

A multi-point univariate decomposition method for structural reliability analysis

A new multi-point univariate decomposition method is presented for structural reliability analysis involving multiple most probable points (MPPs). The method involves a novel function decomposition at all MPPs that facilitates local univariate approximations of a performance function in the rotated Gaussian space, Lagrange interpolation for univariate component functions and return mapping to the standard Gaussian space, and Monte Carlo simulation. In addition to the effort in identifying all MPPs, the computational effort in the multi-point univariate method can be viewed as performing deterministic response analysis at user-selected input defined by sample points. Compared with the existing multi-point FORM/SORM, the multi-point univariate method developed provides a higher-order approximation of the boundary of the failure domain. Both the point-fitted SORM and the univariate method entail linearly varying cost with respect to the number of variables. However, the univariate method with less than nine sample points requires fewer calculations of the performance function than the point-fitted SORM. Numerical results indicate that the proposed method consistently generates an accurate and computationally efficient estimate of the probability of failure.     

A study on the reliability assessment methodology for pipelines with active corrosion defects

A study on the probabilistic methodology for the estimation of the remaining life of pressurized pipelines containing active corrosion defects is presented. This reliability assessment is carried out using several already published failure pressure models. A steady state corrosion rate is assumed to estimate the growth in the dimensions of corrosion defects. The first-order second-moment iterative reliability method, the Monte Carlo integration technique and the first order Taylor series expansion of the limit state function (LSF) are used in order to estimate the probability of failure associated with each corrosion defect over time. The uncertainty of the statistical variables on which the LSF depends are modeled using normal and lognormal distributions and the sensitivity of pipeline reliability to these variables is evaluated. This extended probabilistic analysis framework is applied to a sample operating pipeline which was inspected using a high resolution magnetic flux leakage inspection tool.     

Estimating design reliability of composite rotor blades under ultimate loading

A probabilistic stress analysis tool predicting reliability of composite wind turbine rotor blades was developed and validated by comparing with results from a three‐dimensional shell finite element model of a blade. Stress analysis was based on thin wall multicellular Euler–Bernoulli beam theory using as input section stress resultants directly from aeroelastic simulations; a finite strip method was implemented for elastic stability calculations. Reliability analysis was performed at the ply level of the multidirectional laminates implementing various methods such as the response surface method, β‐index and crude Monte Carlo simulation. Physical and statistical uncertainties of the basic variables was taken into account while several model uncertainties related to the material properties were further introduced and quantified in the light of appropriate test results. To prove the efficiency of the code as a design tool, the effect of various probabilistic assumptions concerning the material properties was directly investigated on the estimated reliability β‐index values for two rotor blade design cases typical of stall‐regulated and pitch‐regulated wind turbines. Copyright © 2014 John Wiley & Sons, Ltd.     

基于有向图和序贯蒙特卡洛法的含电动汽车配电网可靠性评估

大规模电动汽车接入后将对配电网可靠性带来极大影响,快速有效的可靠性评估方法是亟待解决的关键问题之一。针对含电动汽车的配电网,提出了一种基于有向图和序贯蒙特卡洛算法的可靠性评估方法。首先基于有向图矩阵,利用前推和回推搜索辨识系统故障后的局部网络,并构建系统故障模式影响分析表;分析提出了考虑电动汽车接入后故障模式影响分析表的修正方法,并基于序贯蒙特卡洛算法定量分析了电动汽车接入后对配电网可靠性带来的影响,并从电动汽车接入位置和接入时段等角度对电动汽车接入电网提出了相应的建议。算例分析的结果表明所提方法能够快速修正电动汽车接入后的系统故障模式影响分析表,并能够适应不同场景下的含电动汽车配电网可靠性评估,为电动汽车合理接入电网提供理论依据。     

Global moment‐independent sensitivity analysis of single‐stage thermoelectric refrigeration system

A mathematical model of single‐stage thermoelectric refrigeration system (SSTRS) was built considering the influence of external heat transfer. Based on the relationship between the input system design variables and the output cooling capacity, the coefficient of performance and the influence of the fluctuation of the design variables on the stability of the thermoelectric refrigeration system output performance parameters were studied using a moment‐independent sensitivity analysis. The Latin hypercube sampling method was used to simulate the fluctuation of the design variables, and their moment‐independent sensitivity indices were calculated by the Monte Carlo method. The design variables were then sorted according to their importance to the output parameters, and the key design variables affecting the coefficient of performance and the cooling capacity were obtained. These results can provide useful guidance for the design and optimisation of SSTRS.