Tide Prediction Using Neural Networks

Prediction of tides at a subordinate station located in the interior of an estuary or a bay is normally done by applying an empirical correction factor to observations at some standard or reference station. This paper presents an objective way to do so with the help of the neural network technique. In complex field conditions this approach may look more attractive to apply. Prediction of high water and low water levels as well as that of continuous tidal curves is made at three different locations. The networks involved are trained using alternative training algorithms. Testing of the networks indicated satisfactory reproduction of actual observations. This was further confirmed by a high value of the accompanying correlation coefficient. Such a correlation was better than the one obtained through use of the statistical linear regression model. The training algorithm of cascade correlation involved the lowest training time and hence is found to be more suitable for adaptive training purpose.     

Real-Time Flood Forecasting Using Neural Networks

Real-time forecasting of stream flows during storms provides an essential input to operational flood management. This work is usually very complex owing to the uncertain and unpredictable nature of the underlying phenomena. The technique of neural networks therefore was applied to model it. Forecasting of flood values during storms with a lead time of one and more hours was made using a selected sequence of past flood values observed at a specific location. Training of the network was done with the help of three alternative methods, viz., error backpropagation, conjugate gradient, and cascade correlation. Resulting flood forecasts were found to be satisfactory—especially when warning time was the least.     

CPT-Based Liquefaction Evaluation Using Artificial Neural Networks

This article presents various artificial neural network (ANN) models for evaluating liquefaction resistance and potential of sandy soils. Various issues concerning ANN modeling such as data preprocessing, training algorithms, and implementation are discussed. The desired ANN is trained and tested with a large historical database of liquefaction performance at sites where cone penetration test (CPT) measurements are available. The ANN models are found to be effective in predicting liquefaction resistance and potential. The developed ANN models are ported to a spreadsheet for ease of use. A simple procedure for conducting uncertainty analysis to address the issue of parameter and model uncertainties is also presented using the ANN‐based spreadsheet model. This uncertainty analysis is carried out using @Risk, which is an add-in macro that works well with popular spreadsheet programs such as Microsoft Excel and Lotus 1-2-3. The results of the present study show that the developed ANN model has potential as a practical design tool for assessing liquefaction resistance of sandy soils.     

非线性局部剪力交互作用下的开裂组合梁长期变形神经网络研究

结合有限元分析和人工神经网络,提出一种新的思路,研究简支组合梁的短期和长期变形。本文建立两个神经网络模型,采用相关论文中有限元模型的结果进行样本训练。有限元模型考虑了抗剪连接件的非线性荷载-滑移关系,以及蠕变、收缩和混凝土板的裂缝。而对没开裂的混凝土只考虑了蠕变、收缩的影响。为训练及验证两个神经网络模型,建立了一个包括不同设计参数的大数据库。研究发现,两个神经网络模型均能预测组合梁的变形。因此,神经网络模型可用以评估非几何设计参数对简支组合梁的短、长期变形影响。最后,根据AISC规范和欧洲规范4方法计算简支组合梁的短、长期变形,并与有限元模型结果进行比较。结果表明,与有限元方法相比,AISC方法低估了短期变形而高估了长期变形。     

A Pavement Condition-Rating Model Using Backpropagation Neural Networks

Abstract: This paper presents an overview of the neural-network technique as a management tool for maintenance of flexible pavement. The paper discusses the development and implementation of a neural network for the condition rating of roadway sections. The condition-rating scheme developed by Oregon State Department of Transportation was used as the basis for the development of the network presented. A training set of 744 cases was used to train the network, and a set of 1736 cases was used to test the generalization ability of the system. The network adequately learned the training examples with an average training error of 0.019 and was able to determine the correct condition ratings with an average testing error of 0.023. The network's ability to deal with noisy data also was tested. Up to 60% noise was added to the data and introduced to the network. The results showed that the network presented could identify condition rating relationships at high levels of-noise. Finally, an expert determination was compared with that produced by the network. The network was able to mimic the expert's condition ratings with an average error of 0.0354.     

基于微人工鱼群算法优化RBF神经网络

在人工鱼群算法的基础上提出了一种新的优化算法--微人工鱼群算法,作为径向基神经网络(RBFNN)的训练算法.微人工鱼群算法利用两个鱼群(寻优鱼群和库存鱼群)来寻优,寻优鱼群使用人工鱼群算法来寻找全局最优解,库存鱼群保证了寻优鱼群的多样性,微人工鱼群算法使RBFNN的隐中心位置和相应的宽度值同时得以优化,提高了RBFNN的泛化能力.将微人工鱼群算法优化后的RBFNN应用于双螺旋和IRIS分类,试验结果表明,相对于K-means以及人工鱼群算法,本文方法在泛化能力上得到提高.     

Analysis of Bridge Condition Rating Data Using Neural Networks

Currently bridges are evaluated using either a visual inspection process or a detailed structural analysis. When bridge evaluation is conducted by a visual inspection, a subjective rating is assigned to a bridge component. With analytical evaluation, the rating is computed based on the load applied and the resistance of the bridge component. There have been several attempts to correlate the subjective rating to the analytical rating. The conventional statistical analyses, as well as methods based on fuzzy logic, have not been very successful in providing a clear relationship between the two rating systems. This paper describes the application of neural network systems in developing the relation between subjective ratings and bridge parameters as well as that between subjective and analytical ratings. It is shown that neural networks can be trained and used successfully in estimating a rating based on bridge parameters. The specific application problem for railroad bridges in the commuter rail system in the Chicago metropolitan area is presented. The study showed that a successful training of a network can be achieved, especially if the input data set contains parameters with a diverse combination of intercorrelation coefficients. When the relationship between the bridge subjective rating and bridge parameters was investigated, the network had a prediction rating of about 73%. The study also investigated the relation between the subjective and analytical rating. In this case, the prediction rate was about 43%. Compared with conventional statistical methods and the fuzzy‐logic approach, the neural network system had a much better performance ratio in establishing the relation between the bridge rating and bridge parameters.     

Predicting the Outcome of Construction Litigation Using Neural Networks

In this study, neural networks were used to predict the outcome of construction litigation. Disagreements between the owner and the contractor can arise from such considerations as interpretation of the contract, changes made by the owner, differing site conditions, acceleration and suspension of work, and so forth. When there are disagreements between the contractor and the owner, the result is the inefficient use of resources and higher costs for both the owner and the contractor, as well as damage to the reputation of both sides. Neural networks may help to predict the outcome of construction claims that are normally affected by a large number of complex and interrelated factors. Data composed of characteristics of cases and circuit and appellate court decisions were extracted from cases filed in Illinois appellate courts in the last 12 years. A network was trained using these data, and a rate of prediction of 67% was obtained. If the parties to a dispute know with some certainty how the case would be resolved if it were taken to court, it is believed that the number of disputes could be reduced greatly.     

神经网络在系统辨识与控制中的应用

讨论了人工神经网络辨识的基本方法，分析了神经网络控制的几种基本结构：自校正自适应控制、模型参考自适应控制、自校正内模控制器等，并分别指出了这些控制策略的优、缺点及有待解决的一些问题．     

Iterative Equation Solver for Bridge Analysis Using Neural Networks

A technique for enhancing finite-element analysis equation solvers for particular problem domains, i.e., particular classes of structures such as highway bridges, is presented. The technique involves merging artificial neural networks, used as a domain knowledge-encoding mechanism, together with a preconditioned conjugate gradient iterative equation-solving algorithm. In the algorithm, neural networks are used to seed the initial solution vector and to precondition the matrix system using customizable and trainable neural networks. A case study is presented in which the technique is applied to the particular domain of flat-slab highway bridge analysis. In the case study, neural networks are trained to encode the load-displacement relationships for concrete flat-slab highway bridges. Analytical load-displacement data are generated using finite-element analyses and subsequently used to train neural networks. Acting collectively, the neural networks predict approximate displacement patterns for flat-slab bridges under arbitrary loading conditions.     

Failure Criterion of Concrete under Triaxial Stresses Using Neural Networks

A neural network approach to model the strength of concrete under triaxial stresses is presented in this paper. A radial basis function neural network (RBFNN) and a backpropagation neural network (BPNN) are used for training and testing the experimental data in order to acquire the failure criterion of concrete strength. Unlike the traditional regression analyses where the explicit forms of the equation must be defined first, the neural network approach provides a general form of strength envelope. The study shows that the RBFNN model provides better prediction than the BPNN model. Parametric studies on both models are carried out to find the best neural network structure. Finally, a comparison study between the neural network model and two regression models is made.     

电梯交通分析和选型配置系列专题（3）——神经网络在电梯系统NSF问题中的应用

包括电梯系统控制方面的调查及除了根据电梯的需求等待的PC控制系统外经典的电梯最优控制系统。阐述了电梯的NSF问题，即给定一个单独电梯轿厢的电梯系统，已知位置、方向和分散登记的上行或下行层站呼叫作为输入，找到NSF作为网络的输出。在这个研究中，人工神经网络被用来在一个电梯控制单元确定下一个停站层（NSF），确定用于电梯交通量控制的方法。根据仿真运行本文提出的隐含在控制算法中的神经网络，评价了标准系统的性能指标，且同其他控制算法诸如优先权时间控制、分区算法和动态分区算法进行了比较。     

Predicting Geotechnical Parameters of Sands from CPT Measurements Using Neural Networks

Predicting sand parameters such as D_r ,  K ₀, and OCR from CPT measurements is an important and challenging task for the geotechnical engineer. In the present study, a system of neural networks is developed for predicting these parameters based on CPT measurements. The proposed system uses backpropagation neural networks for function approximation and probabilistic neural networks for classification. By strategically combining both types of networks, the proposed system is able to predict accurately D_r ,  K ₀, and OCR of sands from CPT measurements and other soil parameters. Details on the development of the proposed system are presented, along with comparisons of the results obtained by this system with existing methods.     

Finite–Element Mesh Generation Using Self–Organizing Neural Networks

Neural networks are applied to the problem of mesh placement for the finite–element method. When the finite–element method is used to numerically solve a partial differential equation with boundary conditions over a domain, the domain must be divided into "elements." The precise placement of the nodes of the elements has a major affect on the accuracy of the numeric method. In this paper the self–organizing algorithm of Kohonen is adapted to solve the problem of automatically assigning (in a near–optimal way) coordinates from a two–dimensional domain to a given topologic grid (or mesh) of nodes in order to apply the finite–element method effectively when solving a partial differential equation with boundary conditions over that domain.
One novelty of the method is the interweaving of versions of the Kohonen algorithm in different dimensions simultaneously in order to handle the boundary of the domain properly.
Our method allows for the use of arbitrary types of two–dimensional elements (in particular, quadrilaterals or mixed shapes as opposed to just triangles) and for varying desired densities over the domain. (Thus more elements can be placed automatically near "areas of interest.")
The methods and experiments developed here are for two–dimensional domains but seem naturally extendible to higher–dimensional problems. The method uses a mixture of both one– and two–dimensional versions of the Kohonen algorithm, with an improvement suggested by Tabakman and Exman, and further adapted to the particular problem here. Experimental results comparing this algorithm with a well–known two–dimensional grid–generating system (PLTMG) are presented.     

利用人工神经网络预测火灾下管桁架的温度

3层反向传播(BP)神经网络已用于预测火灾下平面管桁架钢的极限温度。网络模型的输入参数有直径比(β)、墙宽厚比(τ)、径厚比(γ)和荷载比,输出参数有极限温度。利用有限元软件ABAQUS建立神经网络模型。105组数据用于建立BP神经网络,15组数据用于测试和验证BP网络。建立BP网络的过程中,选用Levenberg-Marquardt反向传播算法。隐藏层选用tansig函数,输出层选用purelin函数。分析结果表明,使用BP网络模型预测的极限温度是准确有效的。     

A Comparative Analysis of Two Artificial Neural Networks Using Pavement Performance Prediction

The management of pavements requires the ongoing allocation of substantial manpower and capital resources by the responsible agencies. These agencies ultimately report to the executive and legislative branches of government, which require justification and proof of the efficacy of these expenditures. This and the need for improved engineering technical feedback have encouraged the development of pavement management systems (PMS). One goal of a PMS is to provide decision makers at all levels with optimal resource-allocation strategies. This requires evaluation of alternatives over an analysis period based on predicted values of pavement performance. This necessitates more reliable pavement performance prediction models. Traditional modeling uses multiple regression techniques to predict pavement performance from traffic, time, and pavement distress or various combinations of these factors. Within the last 10 years, new modeling techniques, including artificial neural networks (ANNs), have been applied to transportation problems. The ANNs examined usually have been of a single type called a dot product ANN. This paper examines a different type called the quadratic function ANN and compares the results to the dot product ANN. The quadratic function ANN is a generalized adaptive, feedforward neural network that combines supervised and self-organizing learning. Models were developed to predict roughness using both types of ANN on the same data samples and the results compared. The data samples were drawn from the Kansas Department of Transportation's PMS database. The results indicate a significant improvement in the use of the self-organizing quadratic function ANNs and lead to recommendations for specific areas of additional research.     

The Effects of Detector Spacing on Traffic Forecasting Performance Using Neural Networks

An investigation was made as to how short-term traffic forecasting on motorways and other trunk roads is related to the density of detectors. Forecasting performances with respect to different detector spaces have been investigated with both simulated data and real data. Pruning techniques to the input variables used for neural networks were applied to the simulated data. The real data were collected from the M25 motorway and included flow, speed, and occupancy. With the data used in our study, the forecasting performances decrease with the increase of detector spaces. However, by taking the assumed costs of detector infrastructure into account, it may be concluded from this study that increasing coverage to a spacing of 500 m gives little extra benefit and may actually be counter productive in certain circumstances. It was concluded that, on the basis of current evidence, a detector spacing of between 1 and 1.5 km might be optimal.     

人工神经网络在强夯工程中的运用分析与讨论

鉴于当前人工神经网络在岩土工程中的应用越来越广泛的情况下 ,本文分析和比较了多种人工神经网络模型对强夯问题的适用性和可靠性 ,并提出了几个人工神经网络在应用过程中应注意的问题 ,使之能够更好地指导强夯工程实践