Thermal comfort in offices:comfort values and optimization of indoor control variables

In order to search for reasonable air-conditioned indoor control variables and save energy consumption and meet tO need of personal thermal comfort,a method which is based on numerical simulation is employed to optimize indoor control variables.Computational fluid dynamics(CFD)is used to describe thermal state of office.An optimal method is proposed in this paper,dual neural network model is firstly used to acquire reliable information,data from CFD model are pre-processed,and the remaining data are used to train artificial neural networks(ANN),then CFD model is replaced by ANN model to reduce computational cost when is optimized,indoor control variables are optimized by genetic algorithm.Simulation results show that indoor thermal comfort is improved obviously,and the energy cost is decreased accordingly.     

Genetic adaptive search model of hot metal desulphurization

Genetic adaptive search (GAS) techniques, based on the mechanics of natural genetics, are used in this paper to find an optimal model of desulphurization process of iron-melt by powder injection. Industrial data on 400 t torpedo are used to evaluate contributions of transitory and permanent contact reactions to the overall desulphurization process during calcium carbide powder injection. The results obtained with GAS agree more closely with the plant data than the theoretical model. GAS solutions are also found to be better than regression models. These results suggest immediate application of GAS in similar metallurgical problems.     

Comparative Study of SVMs and ANNs in Aquifer Water Level Prediction

In this research, a data-driven modeling approach, support vector machines (SVMs), is compared to artificial neural networks (ANNs) for predicting transient groundwater levels in a complex groundwater system under variable pumping and weather conditions. Various prediction horizons were used, including daily, weekly, biweekly, monthly, and bimonthly prediction horizons. It was found that even though modeling performance (in terms of prediction accuracy and generalization) for both approaches was generally comparable, SVM outperformed ANN particularly for longer prediction horizons when fewer data events were available for model development. In other words, SVM has the potential to be a useful and practical tool for cases where less measured data are available for future prediction. The study also showed high consistency between the training and testing phases of modeling when using SVM compared to ANN. While for the proposed SVM model the relative error of mean square error increased by an average of 42% from the training phase to testing the phase, the corresponding testing error of the ANN model raised by approximately seven times the training error.     

Neural nets – types,configurations and pitfalls

Fuzzy logic, neural nets and genetic algorithms form the core of soft computing methods. They are useful when there is no possibility to compute an exact mathematical model (hard computing). Neural nets have the ability to learn by example. This advantage is exploited by a lot of applications and many software packages make it quite easy to use neural nets. A stage is reached, where some critical remarks should be made in order to avoid disappointments. Some frequently used net types (backpropagation, LVQ, SOM) are discussed together with configuration and training problems. Important topics are the avoidance of under- and overfit and the remark that neural nets produce correct outputs only if the inputs lie in the part of the feature space, the net was trained for. Therefore, a detailed analysis of the training data set should be made. In the context of safety relevant applications the missing interpretability of neural net outputs is often criticized. Fuzzy-neuro-systems try to improve this situation.     

改进BP网络在智能传感器设计中的应用

传感器大多数都对环境(比如温度)有一定的敏感度,这样就会使传感器的零点和线性发生偏移,从而造成输出值随环境温度的变化而变化,再加上气压、以及气体流量等因素,导致测量出现附加误差。本文将利用神经网络来处理各种环境因素而产生的误差,将低成本的微控制器与传感器结合起来,设计出了能够自动补偿环境影响的智能传感器。并针对硬件平台的局限性,根据学习网络的学习特性,做了相应的优化改进,实现了传感器高精度快速误差补偿。     

Neural Networks for Profiling Stress History of Clays from PCPT Data

This paper evaluates the feasibility of using artificial neural network (ANN) models for estimating the overconsolidation ratio (OCR) of clays from piezocone penetration tests (PCPT). Three feed-forward, back-propagation ANN models are developed, and trained using actual PCPT records from test sites around the world. The soil deposits range from soft, normally consolidated intact clays to very stiff, heavily overconsolidated fissured clays. ANN model 1 is a general model applicable for both intact and fissured clays. ANN model 2 is suited for intact clays, and ANN model 3 is applicable to fissured clays only. The models are validated using new PCPT data (not used for training), and by comparing model predictions with reference OCR values obtained from oedometer tests. For intact clays, ANN model 2 gives better OCR estimates compared to ANN model 1. For fissured clays, ANN model 3 gives better estimates compared to ANN model 1. Some of the existing interpretation methods are reviewed. Compared to the existing methods, ANN models 2 and 3 give very good estimates of OCR.     

Neural network modelling of flow stress and mechanical properties for hot strip rolling of TRIP steel using efficient learning algorithm

Abstract

Transformation induced plasticity (TRIP) steels exhibit excellent strength and ductility and can be engineered to provide excellent formability for manufacturing complex parts. In this study, a data driven multi-input multi-output multilayer perceptron based neural network model has been developed to predict the flow stress, yield strength, ultimate tensile strength and elongation as a function of composition and thermomechanical processing parameters for strip rolling of TRIP steels. The input parameters in this generalised regression artificial neural network (ANN) model are steel chemistry, cooling rate and finish roll temperature. The network training architecture has been optimised using the Broyden–Fletcher–Goldfarb–Shanno algorithm to minimise the network training error within few training cycles. The algorithm facilitates a faster convergence of network training and testing errors. There has been an excellent agreement between the ANN model predictions and the target (measured) values for flow stress and mechanical properties depicted by the respective regression fit between these values.     

H∞ Filtering in Neural Network Training and Pruning with Application to System Identification

An efficient training and pruning methodology based on the H∞ filtering algorithm is proposed for artificial neural networks (ANNs). ANNs are first trained by the H∞ filtering algorithm and then some unimportant weights are removed based on the training. The results presented in the paper show that the proposed method provides better pruning results of the network without losing its generalization capacity. It also provides a robust training algorithm for given arbitrary network structures. The usefulness and effectiveness of the proposed methodology are demonstrated in developing an ANN model of a hysteretic structural system.     

Feed forward neural networks for the analysis of censored survival data: a partial logistic regression approach

Flexible modelling in survival analysis can be useful both for exploratory and predictive purposes. Feed forward neural networks were recently considered for flexible non-linear modelling of censored survival data through the generalization of both discrete and continuous time models. We show that by treating the time interval as an input variable in a standard feed forward network with logistic activation and entropy error function, it is possible to estimate smoothed discrete hazards as conditional probabilities of failure. We considered an easily implementable approach with a fast selection criteria of the best configurations. Examples on data sets from two clinical trials are provided. The proposed artificial neural network (ANN) approach can be applied for the estimation of the functional relationships between covariates and time in survival data to improve model predictivity in the presence of complex prognostic relationships.     

Action potential classifiers: a functional comparison of template matching, principal components analysis and an artificial neural network

Multiunit neural activity occurs often in electrophysiological studies when utilizing extracellular electrodes. In order to estimate the activity of the individual neurons each action potential in the recording must be classified to its neuron of origin. This paper compares the accuracy of two traditional methods of action potential classification--template matching and principal components--against the performance of an artificial neural network (ANN). Both traditional methods use averages of action potential shapes to form their corresponding classifiers while the artificial neural network 'learns' a nonlinear relationship between a set of prototype action potentials and assigned classes. The set of prototypic action potentials and the assigned classes is termed the training set. The training set contained action potentials from each class which exhibited the full range of amplitude variability. The ANN provided better classification results and was more robust in analysis of across-animal data sets than either of the traditional action potential classification methods.     

Classification of observational data with artificial neural networks versus discriminant analysis in pharmacoepidemiological studies--can outcome of fluoxetine treatment be predicted?

For several years, there has been an ongoing discussion about appropriate methodological tools to be applied to observational data in pharmacoepidemiological studies. It is now suggested by our research group that artificial neural networks (ANN) might be advantageous in some cases for classification purposes when compared with discriminant analysis. This is due to their inherent capability to detect complex linear and nonlinear functions in multivariate data sets, the possibility of including data on different scales in the same model, as well as their relative resistance to "noisy" input. In this paper, a short introduction is given to the basics of neural networks and possible applications. For demonstration, a comparison between artificial neural networks and discriminant analysis was performed on a multivariate data set, consisting of observational data of 19738 patients treated with fluoxetine. It was tested, which of the two statistical tools outperforms the two other in regard to the therapeutic response prediction from the clinical input data. Essentially, it was found that neither discriminant analysis nor ANN are able to predict the clinical outcome on the basis of the employed clinical variables. Applying ANN, we were able to rule out the possibility of undetected suppressor effects to a greater extent than would have been possible by the exclusive application of discriminant analysis.     

逐步回归分析与人工神经网络用于多组分稀土元素的HPLC同时测定

本文首先用逐步回归分析法对三组分稀土元素的重叠色谱数据进行筛选，提高数据的显著性，减少数据量，然后用ＢＰ模型的人工神经网络进行训练与预测，对重叠色谱峰中的各组分含量进行了定量分析。结果表明：采用逐步回归分析法对数据进行处理后，不仅提高了训练速度，而且预测结果也得到了明显的改善。     

无数学模型的非线性约束单目标系统优化方法改进

为提高无法准确建立数学模型的非线性约束单目标系统优化问题的寻优精度,并考虑获取样本的代价,提出一种基于支持向量机和免疫粒子群算法的组合方法(support vector machine and immune particle swarm optimization,SVM-IPSO).首先,运用支持向量机构建非线性约束单目标系统预测模型,然后,采用引入了免疫系统自我调节机制的免疫粒子群算法在预测模型的基础上对系统寻优.与基于BP神经网络和粒子群算法的组合方法(BP and particle swarm optimization,BP-PSO)进行仿真实验对比,同时,通过减少训练样本,研究了在训练样本较少情况下两种方法的寻优效果.实验结果表明,在相同样本数量条件下,SVM-IPSO方法具有更高的优化能力,并且当样本数量减少时,相比BP-PSO方法,SVM-IPSO方法仍能获得更稳定且更准确的系统寻优值.因此,SVM-IPSO方法为实际中此类问题提供了一个新的更优的解决途径.      

基于粒子群算法优化BP神经网络的溶浸开采浸出率预测

为了研究溶浸开采过程中浸出率的预测问题,以含锑硫化矿的浸出过程为例,采用经粒子群算法优化的BP神经网络模型预测浸出率。首先分析得出影响矿物浸出率的主要因素,并将已有样本数据进行变量训练,建立BP神经网络预测模型;其次利用粒子群算法优化该模型;最后分别利用BP神经网络模型和PSO-BP神经网络模型预测浸出率,并对比2种模型预测值与实际值的误差精度。研究结果表明:影响含锑硫化矿浸出率的主要因素有温度、时间、液固比、搅拌速度和HCl浓度,且这些因素相互影响,其与浸出率呈现高度非线性关系,采用粒子群算法优化的BP神经网络模型训练精度较高,对浸出率的预测更精确,相比BP神经网络,该模型得出的预测结果与实际值的相对误差以及方差都有明显下降。由此可见,该预测模型对当前矿区溶浸开采的浸出率优化有一定的参考价值。     

基于IPSO-BP神经网络的富氧底吹铜熔炼炉喷枪寿命预测模型

富氧底吹铜熔炼炉喷枪是整个熔炼炉中最重要的部件,并且造价高,易损坏,工作环境恶劣复杂,对其进行准确的寿命预测比较困难。提出了一种基于IPSO-BP神经网络的寿命预测模型,粒子群优化算法解决了BP神经网络容易陷入局部极小值和训练速度慢的问题,优化的粒子群算法优化了惯性权重和学习因子,进一步加快了训练速度和搜索速度,提高了BP神经网络跳出局部极小值的能力。以工作环境中容易对喷枪寿命造成影响的因素作为输入,喷枪寿命作为输出,通过实际生产采集的数据做验证,并与BP神经网络和PSO-BP神经网络预测模型作对比。结果表明,本文构建的寿命预测模型预测效果比BP神经网络和PSO-BP神经网络的预测更加准确,精度更高,该预测模型为富氧底吹铜熔炼的喷枪寿命预测提供了一种方法借鉴。     

人工神经网络解析分光光度法同时测定钨和钼

用单纯形法优化了钨、钼－二溴羟基苯基荧光酮－ＣＴＭＡＢ显色体系的实验条件。应用三层ＡＮＮ－ＢＰ网络解析钨和钼的吸收光谱,分光光度法同时测定了钨和钼并与偏最小二乘法、因子分析、Ｐ－矩阵法、卡尔曼滤波、主成分回归等化学计量学方法的解析结果进行了比较,表明神经网络方法优于其它方法。使用改进的ＢＰ算法,避免了神经网络学习过程中可能产生的麻痹现象。提出了目标向量的简单变换方法及便于网络参数选择的收敛评价函数。     

The target switch algorithm: a constructive learning procedure for feed-forward neural networks

We propose an efficient procedure for constructing and training a feed-forward neural network. The network can perform binary classification for binary or analogue input data. We show that the procedure can also be used to construct feedforward neural networks with binary-valued weights. Neural networks with binary-valued weights are potentially straightforward to implement using microelectronic or optical devices and they can also exhibit good generalization.     

热镀锌过渡卷的力学性能预测

在收集大量现场数据的基础上,运用BP神经网络算法,建立了退火处理各工艺参数对热镀锌过渡卷力学性能影响的数学模型,并与线性回归模型相比较.结果表明,BP算法预测误差较线性回归预测误差小;神经网络用于热镀锌过渡卷力学性能预测是可行的.     

ANN and Fuzzy Logic Models for Simulating Event-Based Rainfall-Runoff

This study presents the development of artificial neural network (ANN) and fuzzy logic (FL) models for predicting event-based rainfall runoff and tests these models against the kinematic wave approximation (KWA). A three-layer feed-forward ANN was developed using the sigmoid function and the backpropagation algorithm. The FL model was developed employing the triangular fuzzy membership functions for the input and output variables. The fuzzy rules were inferred from the measured data. The measured event based rainfall-runoff peak discharge data from laboratory flume and experimental plots were satisfactorily predicted by the ANN, FL, and KWA models. Similarly, all the three models satisfactorily simulated event-based rainfall-runoff hydrographs from experimental plots with comparable error measures. ANN and FL models also satisfactorily simulated a measured hydrograph from a small watershed 8.44?km2 in area. The results provide insights into the adequacy of ANN and FL methods as well as their competitiveness against the KWA for simulating event-based rainfall-runoff processes.