深度稀疏最小二乘支持向量机故障诊断方法研究

提出一种结合多层结构和稀疏最小二乘支持向量机(Sparse Least Squares Support Vector Machine,SLSSVM)的机械故障诊断方法。该方法构建了多层支持向量机(Support Vector Machine,SVM)结构,首先在输入层利用支持向量机对信号进行训练,学习信号的浅层特征,利用"降维公式"生成样本新的表示,并作为隐藏层的输入,隐藏层支持向量机对新样本训练并提取信号的深层特征,逐层学习,最终在输出层输出诊断结果。针对因多层结构带来算法的复杂度以及运行时间增加的问题,采用最小二乘支持向量机(Least Squares Support Vector Machine,LSSVM)技术,并将稀疏化理论与最小二乘支持向量机结合,通过构造特征空间近似最大线性无关向量组对样本进行稀疏表示并依此获得分类判别函数,有效解决了最小二乘支持向量机稀疏性缺乏的问题。最后,通过滚动轴承故障诊断实验验证了该方法的有效性。     

钢筋混凝土柱地震破坏模式判别的两阶段支持向量机方法

研究提出了一种钢筋混凝土(RC)柱地震破坏模式判别的两阶段支持向量机(Support Vector Machine,简称SVM)方法.根据RC柱的三种地震破坏模式(弯曲破坏、弯剪破坏和剪切破坏),建立了 RC柱地震破坏模式判别的两阶段SVM模型;基于270组试验数据,利用十折交叉验证和网格寻优方法确定了两阶段SVM的关...     

基于支持向量机的机械系统状态组合预测模型研究

提出了一种新的支持向量机（Support Vector Machines，SVM）机械系统状态组合预测模型。应用FPE（Final Principle Error）准则优化样本的维数，采用时域内的振动烈度和频域内的特征频率分量作为预测机械系统状态的敏感因子，构建了预测模型。支持向量机采用新型的结构风险最优化准则，预测能力强、鲁棒性好。采用径向基函数和ε损失函数，将该模型应用于实验台和旋转注水机组的状态预测，取得了较好的效果。这表明利用支持向量机的组合预测模型，可以降低设备维修代价，提高设备的安全性和可靠性。     

最小二乘支持向量机在瞬变热环境实验研究中的应用

瞬变热环境下,热反应与环境参数是紧密联系的。本文基于最小二乘支持向量机LS-SVM(LeastSquares Support Vector Machine)理论,结合瞬变热环境下受试者的投票实验数据,试图将这种关系量化,以达到对瞬变热环境下整体热感觉预测的目的。通过样本测试对预测模型的验证结果表明,向冷环境过渡和向热环境过渡中误差﹤0.3的样本比例都达到了90%以上,预测结果较理想,并且预测精度优于BP神经网络所建立的模型。另外,考虑到热感觉的模糊性以及个体化差异造成的影响,还给出了测试样本集在置信水平为95%时的置信区间,能对测试样本的变化区间作出较为准确的判断。     

基于遗传-支持向量回归的车内稳态噪声声品质预测

为了稳定、精确地评价车内稳态噪声声品质,以车内稳态噪声为研究对象,进行主观评价试验,计算客观心理声学参数并完成了相关性分析。建立基于支持向量回归(Support Vector Regression,SVR)的车内稳态噪声声品质预测模型,并使用遗传算法(Genetic Algorithm,GA)对支持向量回归的超参数进行优化。其后建立基于反向传播神经网络(Back Propagation Artificial Neural Network,BPANN)的声品质预测模型。对比分析发现遗传-支持向量回归(GASVR)模型预测精度高于BP神经网络。结果表明,遗传-支持向量回归适用于车内稳态噪声声品质预测,能够较大提高车内稳态噪声声品质预测精度和工程效率。     

基于支持向量机的变压器碳排放预测模型

目的 解决变压器中主要设计参数影响下的碳排放量预测问题。方法 本文利用随机森林（Random Forest,RF）算法和支持向量机（Support Vector Machine,SVM）算法进行对比,构建一个变压器碳排放预测模型。结果 通过对变压器的全生命周期进行评价,确定铁芯的长宽比为影响碳排放量的主要因素,对给定参数下的碳排放量进行预测,并与实际值进行对比分析得出,3类预测模型中,SVM高斯核模型的平均绝对误差值约为5.37,与碳排放实际值最为接近,故采用高斯核函数的非线性支持向量机预测模型最优。结论 证明支持向量机高斯核函数预测模型更具有预测准确性和有效性,以期能为生产企业进行低碳设计提供参考依据,为电力行业生产设备的可持续设计研究提供一定的借鉴意义。     

基于不变性常识的支持向量分类机

不变性常识(Invariance)与支持向量机(Support Vector Machine,SVM)的融合技术是近年来支持向量机研究的重点之一,将不变性常识融合于学习模型,有助于提高模型的泛化能力。探讨不变性支持向量机(InvarianceSVM)的形式化描述、目前发展状况及不变性常识与SVM融合的主要技术,并分析各方法的优缺点。     

基于自适应相关向量机模型的轨姿控发动机试验推力矢量预测

为了实时、准确、可靠地预测轨姿控发动机试验中推力矢量的变化情况,本文在分析支持向量机(SVM)预测算法缺陷的基础之上,提出并建立了一种自适应能力较强的故障预测模型——ARVM(Adaptive Relevance Vector Machine),并将其应用于某型轨姿控发动机高模试验推力矢量参数预测中.研究结果表明,ARVM预测模型在稀疏性和算法精度方面均体现出较高的优越性,能够很好地预测轨姿控发动机试验推力矢量的变化趋势.     

基于支持向量回归机的中央空调负荷预测模型

由于中央空调系统的时滞性、时变性、非线性和大惰性等特性,使得当前采用的中央空调负荷预测算法精度并不高,本文在江阴某楼宇空调系统节能改造项目的基础上,从中央空调系统的组成和特性出发,提出了基于支持向量回归机(Support Vector Regression SVR)理论的中央空调负荷预测模型。对项目楼宇历史负荷数据进行分析,分别采用SVR负荷预测模型和BP神经网络负荷预测模型进行了训练和预测。预测结果表明:基于SVR负荷预测模型较BP神经网络负荷预测模型精度更高,具有较强的实用性和可行性。     

基于粒子群-向量机的汽车加速噪声评价

以乘用车由50 km/h加速到100 km/h时的噪声信号为评价对象,用成对比较法对车内加速噪声品质偏好性进行主观评价实验,获得每个样本的偏好性评价值。计算各噪声样本的主要心理声学客观参数并进行相关分析。鉴于评价者对非稳态噪声主观评价过程的复杂性,建立支持向量机(Support Vector Machine,SVM)的主客观评价模型,并利用粒子群优化算法(Particle Swarm Optimization,PSO)对模型参数进行优化。为对比优化后预测效果,建立BP神经网络回归模型。结果表明,优化后的粒子群-向量机回归模型用于噪声声品质评价能获得更好的预测效果,可较大程度提高声品质预测精度。     

Investigation of impact fracture behavior of automobile laminated glass by 3D discrete element method

A three-dimensional discrete element model of laminated glass plane is presented and a 3D numerical analysis code, which can simulate the impact fracture behavior of automobile laminated glass, is developed. The impact process of a single glass plane and a laminated glass plane are calculated in the elastic range by the code. Comparing its results with those calculated by the commercial FEM code LS-DYNA in the same condition, the validity of the 3D laminated glass model and the 3D discrete element method are proved. Furthermore, the impact fracture process of a single glass plane and a laminated glass plane are simulated respectively. The entire failure processes in detail are presented. By comparing the impact force and reduction of kinetic energy of impact body between those two models, the numerical method is applied to demonstrate the advantage of laminated glass in passenger’s safety.     

Micromechanical Modeling of Impact Damage Mechanisms in Unidirectional Composite Laminates

Composite laminates are susceptible to the transverse impact loads resulting in significant damage such as matrix cracking, fiber breakage and delamination. In this paper, a micromechanical model is developed to predict the impact damage of composite laminates based on microstructure and various failure models of laminates. The fiber and matrix are represented by the isotropic and elastic-plastic solid, and their impact failure behaviors are modeled based on shear damage model. The delaminaton failure is modeling by the interface element controlled by cohesive damage model. Impact damage mechanisms of laminate are analyzed by using the micromechanical model proposed. In addition, the effects of impact energy and laminated type on impact damage behavior of laminates are investigated. Due to the damage of the surrounding matrix near the impact point caused by the fiber deformation, the surface damage area of laminate is larger than the area of ??impact projectile. The shape of the damage area is roughly rectangle or elliptical with the major axis extending parallel to the fiber direction in the surface layer of laminate. The alternating laminated type with two fiber directions is more propitious to improve the impact resistance of laminates.     

Dynamic strains in architectural laminated glass subjected to low velocity impacts from small projectiles

An experimental validation of a mechanics-based finite element model for architectural laminated glass units subjected to low velocity, two gram projectile impacts is described. The impact situation models a scenario commonly observed during severe windstorms, in which small, hard projectiles, such as roof gravel, impact windows. Controlled experiments were conducted using a calibrated air gun to propel a steel ball against simply supported rectangular laminated glass specimens. Dynamic strains on the inner glass ply were measured using foil strain gages and a high speed data acquisition system. Impact speed, interlayer thickness, glass ply thickness, and glass heat treatment conditions were varied. Dynamic strains predicted by the finite element model were in close agreement with those measured in the laboratory.     

An Optimized Ensemble Model for Prediction the Bandwidth of Metamaterial Antenna

Metamaterial Antenna is a special class of antennas that uses metamaterial to enhance their performance. Antenna size affects the quality factor and the radiation loss of the antenna. Metamaterial antennas can overcome the limitation of bandwidth for small antennas. Machine learning (ML) model is recently applied to predict antenna parameters. ML can be used as an alternative approach to the trial-and-error process of finding proper parameters of the simulated antenna. The accuracy of the prediction depends mainly on the selected model. Ensemble models combine two or more base models to produce a better-enhanced model. In this paper, a weighted average ensemble model is proposed to predict the bandwidth of the Metamaterial Antenna. Two base models are used namely: Multilayer Perceptron (MLP) and Support Vector Machines (SVM). To calculate the weights for each model, an optimization algorithm is used to find the optimal weights of the ensemble. Dynamic Group-Based Cooperative Optimizer (DGCO) is employed to search for optimal weight for the base models. The proposed model is compared with three based models and the average ensemble model. The results show that the proposed model is better than other models and can predict antenna bandwidth efficiently.     

PVB夹层玻璃的失效扩展

本文为探讨聚乙烯醇缩丁醛(PVB)夹层玻璃断裂失效的扩展规律,运用ANSYS/LSDYNA软件对夹层玻璃在聚能射流冲击作用下的变化进行了仿真研究,分析了玻璃厚度对断裂破坏效果及PVB薄膜厚度对失效扩展特性的影响。结果表明:冲击作用下玻璃厚度对夹层玻璃断裂破坏的影响较平缓。随着PVB薄膜厚度的增加,玻璃试件的失效扩展特性(速度、起始加速度、后期减速度和稳定加速度)减小,其吸能特性增强,断裂失效的扩展过程趋于平缓。并从理论方面判定夹层玻璃在冲击过程中的失效扩展速度主要取决于应力分量,且PVB粘弹性层具有较强的吸能作用。     

复合材料层合板冲击后压-压疲劳寿命预测方法

针对冲击后复合材料层合板, 发展了含冲击初始损伤层合板的压-压疲劳寿命预测方法。该方法基于无损单向板的力学性能和疲劳特性, 对不同铺层参数、 不同几何尺寸以及不同冲击条件下层合板的疲劳寿命进行预测。为消除人为假设冲击损伤造成的误差, 对层合板在冲击载荷及冲击后疲劳载荷作用下的破坏进行全程分析, 即把冲击后层合板的实际损伤状态直接作为疲劳分析的初始状态。同时基于逐渐损伤思想, 推导了含冲击初始损伤层合板的应力分析过程, 建立了相应的三维逐渐累积损伤模型, 开发了参数化的复合材料层合结构冲击及冲击后疲劳破坏模拟程序, 为复合材料层合结构的抗冲击设计及其疲劳损伤扩展行为研究提供了较好的技术平台。      

Spontaneous failure of a passable laminated glass floor element

It is well-known that thermally toughened safety glass is subjected to a certain risk of spontaneous failure due to nickel sulphide particles included in the material. However, the present contribution focuses on a very uncommon case in which two out of three glass layers of a thermally toughened laminated slab of a passable floor element failed spontaneously.After initial on-site observations, witness interviews and reconstruction of the exact circumstances of the failure, no direct external cause such as hard or soft body impact, and local heating could be found. Consequently, the laminated glass slab was further investigated in the laboratory. During the laboratory analysis, several techniques have been used, such as crack pattern analysis, optical microscopy, and FEG scanning electron microscopy.As a result of the failure analysis, a nickel sulphide particle could be clearly detected. Based on striking similarities, a.o. in the crack pattern, breakage of the underlying glass layer of the laminate could be attributed to nickel sulphide as well.A method to assess the probability of failure of a double NiS failure is proposed. In spite of the design philosophy followed by the designer, in which the probability of this failure mode was not considered to be significant, this failure and risk analysis demonstrated clearly its importance for building practice.     

自适应人工蜂群优化极限学习机在拉曼光谱血液定量分析中的应用

提出了一种基于自适应差分进化人工蜂群优化极限学习机预测血液各组分浓度的方法。首先应用人工蜂群算法对输入权值和隐含层阈值迭代寻优;其次结合差分进化进一步提高模型精度且避免后期易陷入局部最优等问题;由于差分进化算法交叉率和变异率存在凭经验给定的不确定性,最后引入了自适应调整的思想提出自适应差分进化人工蜂群算法优化极限学习机算法的模型,将其应用于血液成分定量分析中。实验表明,自适应差分进化人工蜂群算法优化的极限学习机模型具有较高的预测精度,模型具有较强的稳健性。     

融合材料领域知识的数据准确性检测方法

材料数据由于小样本、高维度、噪音大等特性, 用于机器学习建模时常常会产生与领域专家认知不一致的结果。面向机器学习全流程, 开发材料领域知识嵌入的机器学习模型是解决这一问题的有效途径。材料数据的准确性直接影响了数据驱动的材料性能预测的可靠性。本研究针对机器学习应用过程中的数据预处理阶段, 提出了融合材料领域知识的数据准确性检测方法。该方法首先结合材料专家认知构建了材料领域知识库。然后, 将其与数据驱动的数据准确性检测方法结合, 从数据和领域知识两个角度对材料数据集进行基于描述符取值规则的单维度数据正确性检测、基于描述符相关性规则的多维度数据相关性检测以及基于多维相似样本识别策略的全维度数据可靠性检测。对于每一阶段识别出的异常数据, 结合材料领域知识进行修正, 并将领域知识融入到数据准确性检测方法的全过程以确保数据集从初始阶段就具有较高准确性。最后该方法在NASICON型固态电解质激活能预测数据集上的实验结果表明: 本研究提出的方法可以有效识别异常数据并进行合理修正。与原始数据集相比, 基于修正数据集的6种机器学习模型的预测精度都有不同程度的提升。其中, 在最优模型上R²提升了33%。     

基于机器学习的电喷印精度预测方法研究

目的 节省电流体喷射打印精度预测的时间和解决电流体工艺参数的选择问题,达到提高电流体打印的质量和效率的目的。方法 为了对电流体喷射打印精度进行预测,提出有限元模型与机器学习相结合的方法。基于线性回归、支持向量回归和神经网络等机器学习算法建立4种参数与射流直径的关系模型。结果 算法结果表明:支持向量回归和神经网络预测模型的决定系数R²能达到0.9以上,表示模型可信度高;支持向量回归和神经网络预测模型指标都比线性回归预测模型的小。结论 机器学习算法可对电喷印打印精度进行有效预测,预测效率提高了十几倍,节省了精度预测的时间。