分类挖掘方法在银行不良贷款信用风险评估中的应用

针对我国商业银行面临的不良贷款信用风险问题,提出了一种基于数据挖掘技术的决策树模型方法,对不良贷款信用风险问题进行预测分类。详细介绍了决策树模型的建立方法并且用实例结果表明该模型在预测银行不良贷款信用风险中的实用价值。     

决策树技术在电缆绝缘状态评估中的应用

电缆的绝缘状态通常可以分为良好、不好、差和故障等几种,以电缆的日常检修数据、试验数据和在线监测数据为基础,对电缆的状态进行判断是一个非常有意义的课题。采用决策树分类技术来对电缆的绝缘状态进行分类,分别对各种类型数据形成子树,然后通过子树合成技术形成最终的决策树,从而对电缆的绝缘状态进行判断。通过一个实际电缆的各种数据,采用SPSS软件进行实际应用,最终的仿真结果说明决策树技术是一种非常有效的电缆绝缘状态分类技术。     

SMC模型求解最优动态客户保持投入的一种方法

客户是资产的观念目前已被普遍接受,而企业如何对这一重要资产进行保持投入就成了亟待解决的问题.借由SMC模型的构建基础,求得个体客户在只发生一次购买行为的情况下未来每年的购买概率,再根据正向决策树构造客户保持率之公式表达,推导出未来各期客户的保持投入.在此推导过程中还可一并得到客户保持率的威布尔模拟的完整形式及各期平均客户收益.     

最小二乘支持向量机和小波电铲供电机组状态预测

针对电铲供电机组振动时间序列是个非线性、非平稳的复杂时间序列,难以用单一预测方法进行有效预测的问题,建立了一种基于小波分解和最小二乘支持向量机混合模型进行状态预测的方法.首先通过小波分解,将原始振动时间序列分解到不同层次,然后根据分解后各层次分量的特点选择不同的嵌入维数和LS-SVM参数分别进行预测,最后重构得到原始序列的预测值.对某电铲供电机组振动趋势的预测结果表明,该模型的预测性能好于单一的支持向量机预测方法.     

长延时网络控制系统预测控制方法研究

针对网络控制系统由于长时延导致性能下降的问题,提出了一种基于时间戳与快速隐式广义预测控制的时延补偿方法,首先利用时间戳获得网络时延,然后采用快速隐式广义预测控制求出未来时刻的控制量。仿真结果表明该方法具有很好的补偿效果,改善了系统的输出,保证了控制的稳定性。     

结构地震反应的预测控制方法研究

本文采用自回归（ＡＲ）模型对运动加速度进行短时在线预测,进而提出考虑时滞现象的离散状态方程形式的预测模型,对受控结构未来某时段内的状态和输出进行估计。通过邓预测性能指标函数的滚动优化求现实时控制律。仿真分析表明,本文方法在一需增加控制力作用的情况下,较瞬时最优控制和原有预测控制更好的抑制了结构的地震响应,且对控制系统的时滞问题解决良好。     

有关电力负荷的短期预测初探

所谓的电力负荷短期预测指对一年以内用户需求用电量的预测,其包括小时预测、日预测、周预测以及月预测。通常短期预测是预测电功率。在短期内用户需求电量呈现一种随机起伏的状态,其以过去负荷为基础,用户负荷变动、系统内部设备检修以及重大事件与气候变化等因素均会对其产生影响。所以对电力负荷进行短期预测可以为经济调度、发电机组的停启、错峰避峰用电等有着重要的现实意义。     

半主动预测控制系统的时滞与补偿

建立了一种基于磁流体阻尼器的半主动预测控制策略,该方法基于结构预测模型,根据当前结构状态预测其未来输出,通过优化反映受控对象性能的指标,实现最优控制,并在系统内部自动进行时滞补偿。以一装有2个磁流体阻尼器的5层框架结构为例,分析了不同地震作用下时滞与补偿对控制系统性能的影响,即使滞后时间达到1．0s～2．0s,该系统仍...     

基于LSSVR-AR模型的电解电容状态预测方法

针对传统的状态预测方法预测精度不高的问题,提出了一种基于最小二乘支持向量回归机(LSSVR)和AR模型相结合的非平稳时间序列建模的方法(LSSVR-AR),并应用于Buck电路的电解电容等效串联的状态预测中.对非平稳时间序列进行最小二乘支持向量回归,得到非平稳时间序列的趋势项及剔除趋势项后的随机项;对随机项建立AR模型并与趋势项的LSSVR模型组合,得到非平稳时间序列模型;用组合模型对电解电容的等效串联电阻进行状态预测.用本文所提出的方法对其预测的平均绝对百分比误差为6.57%,低于单一的LSSVR模型.实例证明:本文所提出的模型能对电解电容的状态进行准确预测.     

加筋复合材料结构分层损伤的贝叶斯诊断及预测

针对复合材料结构疲劳损伤的在线监测和预测问题,提出了一种基于结构健康监测 (Structural health monitoring, SHM) 和贝叶斯理论的结构分层损伤诊断及结构剩余使用寿命预测方法。在贝叶斯概率理论框架下,采用指数模型描述复合材料结构疲劳分层损伤面积的先验演化规律,融合在线SHM数据对结构分层损伤状态,以及损伤面积演化模型的参数进行联合后验估计,即为损伤诊断结果。进一步通过后验估计得到的损伤状态和模型参数预测未来时刻结构分层损伤面积的演化,从而得到当前复合材料结构的剩余使用寿命预测结果。通过有限元仿真的加筋复合材料结构疲劳分层扩展对所提出的方法进行了验证。结果表明,方法可以在线准确地诊断结构分层损伤状态以及预测结构的剩余使用寿命。      

基于长短时记忆网络的旋转机械状态预测研究

作为深度学习算法的一种,长短时记忆网络越来越成为时间序列预测的重要手段,简要阐述长短时记忆网络的基本原理,并将其应用于旋转机械状态监控领域,以轴承数据为样本进行仿真,针对轴承数据的非平稳性,运用经验模态分解方法将其分解为平稳信号,并计算其本征模态分量能量熵作为状态特征,通过计算长短时记忆网络对旋转机械状态单步预测的结果,并与支持向量回归机模型的预测结果进行比较,证明长短时记忆网络在旋转机械状态预测方面可以取得比支持向量回归机更好的效果。     

基于混合高斯输出贝叶斯信念网络模型的设备退化状态识别与剩余使用寿命预测方法研究

提出了基于混合高斯输出贝叶斯信念网络模型的设备退化状态识别与剩余使用寿命预测新方法,将变量消元和期望最大化算法相结合对模型进行推理,应用聚类评价指标对状态数进行优化,通过计算待识别特征向量的概率值来确定设备当前的退化状态,在退化状态识别的基础上,提出了剩余使用寿命预测方法。最后,分别应用50组轴承全寿命仿真数据和3组轴承全寿命实验数据对模型进行验证。结果表明,该模型可有效地识别设备的退化状态并对剩余使用寿命进行预测。     

Modelling and simulation of automated manufacturing systems for evaluation of complex schedules

Automated manufacturing systems have been studied widely in terms of scheduling. As technology evolves, the behaviour of tools in automated manufacturing systems has become complicated. Therefore, mathematical approaches to the analysis of complex schedules no longer reflect reality. In this paper, we propose a systematic way of conducting simulation experiments to evaluate the complex operating schedules of automated manufacturing systems. A simulation model is based on a timed Petri net to take advantage of its mathematical strength. Since a Petri net cannot itself have token firing rules, we introduce additional states called operational states. Operational states are not directly related to a Petri net, and are only used for decision making. In addition, a decision function that is responsible for the conflict resolution of a Petri net model and an operational state transition function are introduced. The parallel simulation concept is also suggested by dividing a Petri net into several independent decision sub-nets. A multi-cluster tool system for semiconductor manufacturing is analysed as an application.     

A comprehensive approach for modeling fatigue and fracture of rails

A comprehensive approach is developed for studying the fatigue phenomena (crack initiation and propagation) induced by repeated rolling or rolling-sliding contacts between wheel and rail. Cracks initiate and propagate in the rail head in a complex varying multiaxial stress regime due to Hertzian or non-Hertzian contacts generating 3D residual stress pattern. This paper presents the main steps of such an approach devoted to the modeling of defects induced in the rails by the traffic. Special attention is paid to some of the principal difficulties met as well as to the proposed solutions. Examples of applications for the prediction of initiation as well as propagation of some defects are presented. It is shown that numerical simulations predict very well the locus of crack initiation as well as its propagation in the rail. Our approach presents at least three main originalities: first, it is a global approach starting from the evaluation of the initial state of the rail to the simulation of the crack propagation under complex loading including multiaxial residual stresses. Second, special and original numerical methods for the evaluation of the initial states, the overloads and the elastoplastic state under service loading have been developed. Third, a new concept based on a “structural Paris law” has been developed and used in the crack propagation simulations.     

Well-being analysis for composite generation and transmission systems based on pattern recognition techniques

A new methodology to evaluate well-being indices for a composite generation and transmission system, based on non-sequential Monte Carlo simulation and pattern recognition techniques, is presented. To classify the success operating states into healthy and marginal, an artificial neural network based on group method data handling techniques is used to capture the patterns of these state classes, during the beginning of the simulation process. The idea is to provide the simulation process with an intelligent memory, based on polynomial parameters, to speed up the evaluation of the operating states. The proposed methodology is applied to the IEEE reliability test system (IEEE-RTS), to the IEEE-RTS-96 and to a configuration of the Brazilian South-Southeastern system.     

Condition Based Structural Health Monitoring and Prognosis of Composite Structures under Uniaxial and Biaxial Loading

This paper presents a condition based structural health monitoring (SHM) and prognosis approach to estimate the residual useful life (RUL) of composite specimens in real time. On-line damage states, which are estimated using real time sensing information, are fed to an off-line predictive model to update future damage states and RUL. The on-line damage index or damage state at any given fatigue cycle is estimated using correlation analysis. Based on the on-line information of the previous and current damage states, an off-line model is developed to predict the future damage state and estimate the RUL. The off-line model is a stochastic model which is developed based on the Gaussian process approach. In this paper, the condition based prognosis model is used to estimate the cumulative fatigue damage in composite test structures under constant amplitude fatigue loading. The proposed procedure is validated under uniaxial fatigue loading as well as biaxial fatigue loading. Experimental validations demonstrate that the prediction capability of the prognosis algorithm is effective in predicting the RUL under complex stress states.     

Formation of water-in-oil emulsions and application to oil spill modelling

Water-in-oil mixtures were grouped into four states or classes: stable, mesostable, unstable, and entrained water. Of these, only stable and mesostable states can be characterized as emulsions. These states were established according to lifetime, visual appearance, complex modulus, and differences in viscosity. Water content at formation was not an important factor. Water-in-oil emulsions made from crude oils have different classes of stability as a result of the asphaltene and resin contents, as well as differences in the viscosity of the starting oil. The different types of water-in-oil classes are readily distinguished simply by appearance, as well as by rheological properties. A review of past modelling efforts to predict emulsion formation showed that these older schemes were based on first-order rate equations that were developed before extensive work on emulsion physics took place. These results do not correspond to either laboratory or field results. The present authors suggest that both the formation and characteristics of emulsions could be predicted using empirical data. If the same oil type as already studied is to be modelled, the laboratory data on the state and properties can be used directly. In this paper, a new numerical modelling scheme is proposed and is based on empirical data and the corresponding physical knowledge of emulsion formation. The density, viscosity, saturate, asphaltene and resin contents are used to compute a class index which yields either an unstable or entrained water-in-oil state or a mesostable or stable emulsion. A prediction scheme is given to estimate the water content and viscosity of the resulting water-in-oil state and the time to formation with input of wave height.     

基于误差修正 NLSTM 神经网络的无人机航迹预测

无人机产业近年来发展迅猛,在军用和民用方面都拥有广泛的应用前景。无人机的航迹记录在其航行过程中发挥着重要作用,无人机的航迹预测也成为当前世界研究的热点,使用神经网络进行航迹预测更可以充分发挥其优势。首先对国内外学者关于航迹预测的文献进行了梳理,根据航迹预测的原理对目前飞行器航迹预测算法进行了总结和分类,针对利用神经网络模型预测无人机航迹并逐步改进模型以提高预测精度的问题进行了研究。接着对于传统神经网络模型预测精度不够高的问题,提出一种带误差修正的嵌套长短期记忆 (ENLSTM) 神经网络预测模型。ENLSTM 在嵌套长短期记忆网络模型的基础上引入了误差修正项,从而使得预测精度更高。最后使用 BP、RNN、LSTM 和 ENLSTM 四种神经网络模型分别对无人机的真实航迹数据和模拟航迹数据进行仿真实验,得出结论：循环神经网络相对 BP 神经网络在无人机航迹的预测上更具有优势,基于基础循环神经网络的逐步改进提升了模型的预测能力,ENLSTM 模型对于无人机的航迹预测具有更好的效果。     

变速直膨式空调系统运行特性的稳态ANN模型

在假设系统输出显、潜冷量的相对值在不同的蒸发器入口空气状态下不发生明显变化的前提下,本文针对实验用变速直膨式空调系统建立了稳态人工神经网络(ANN)模型,预测其在不同压缩机、风机转速组合下的系统输出,利用输出显、潜冷量的相对值可以消除室内空气状态对系统输出的影响。通过稳态实验获得数据训练、检测并验证ANN模型预测变速直膨式系统运行特性的准确性,并通过非训练状态点下的稳态实验验证所提出假设与ANN模型的适用性。ANN模型的训练、检测以及验证实验结果的最大误差均小于5%,平均误差均小于3%,表明该稳态ANN模型可以在训练状态点以及非训练状态点较为准确地预测变速直膨式系统的运行特性。     

Dynamic behavioural model for assessing impact of regeneration actions on system availability: Application to weapon systems

Mastering system availability all along the system life cycle is now a critical issue with regards to systems engineering. It is more true for military systems which operate in a battle context. Indeed as they must act in a hostile environment, they can become unavailable due to failures of or damage to the system. In both cases, system regeneration is required to restore its availability. Many approaches based on system modelling have been developed to assess availability. However, very few of them take battlefield damage into account and relevant methods for the model development are missing. In this paper, a modelling method for architecture of weapon system of systems that supports regeneration engineering is proposed. On the one hand, this method relies on a unified failure/damage approach to extend acknowledged availability models. It allows to integrate failures, damages, as well as the possibility of regeneration, into operational availability assessment. Architectures are modelled as a set of operational functions, supported by components that belong to platform (system). Modelling atoms (i.e. elementary units of modelling) for both the architecture components and functions are defined, based on state-space formalism. Monte Carlo method is used to estimate availability through simulation. Availability of the architecture is defined on the basis of the possible states of the required functions for a mission. The states of a function directly depend on the state of the corresponding components (i.e. the components that support the function). Aggregation rules define the state of the function knowing the states of each component. Aggregation is defined by means of combinatorial equations of the component states. The modelling approach is supported by means of stochastic activity network for the models simulation. Results are analysed in terms of graphs of availability for mission's days. Thus, given the simulation results, it is possible to plan combat missions based on criteria such as the number of platforms to be involved given functions required for the mission or the mean of regeneration to be deployed given the possible threats. Further, the simulation will help towards the design of improved architecture of system of systems which could focus on the factors affecting the availability.