一种基于增量式神经网络的转炉静态控制模型

针对中小型转炉不宜增设副枪、难以对钢水成分和温度进行连续检测、难以建立动态模型的实际情况,本文将传统增量模型和神经网络模型有机结合,提出了一种基于增量式神经网络的转炉静态控制模型,对钢水终点进行控制。在该模型引入了RBF神经网络对钢水终点温度和碳含量进行实时预报,使得对增量式神经网络控制模型的训练以预报模型的输出值与所要求的钢水终点温度和碳含量之差为最小,克服了常规静态控制模型存在的不足,改善了控制效果,提高了炼钢一倒命中率。     

基于PSO-ICA 和RBF 神经网络的转炉炼钢终点预报模型

提出将微粒群优化算法和独立成分分析引入到径向基函数神经网络模型用于转炉炼钢终点预报．利用 微粒群优化算法的全局遍历特性和快速不动点算法的局部寻优能力,改进了传统的独立成分分析算法,解决了其目 标函数易陷入局部最优和独立特征排序不确定的问题,压缩冗余信息并降低输入维数．将提取出的独立特征输入径 向基函数神经网络,预报终点温度和碳含量．对转炉生产实测数据进行了仿真,结果表明该模型能有效提高预报精 度,保证预报的可靠性．     

基于GRBF神经网络的脱硫预报模型

脱硫过程是一个复杂的非线性系统,文章建立了一个基于RBF神经网络的脱硫预报模型,并提出使用GRBF算法,从而较好地解决了传统RBF神经网络中心难于确定,存在过拟合的缺点。实践证明,该算法应用在脱硫预报模型的建立中是合理的、可行的。     

模型输入对模糊神经网络预报模型的影响研究

为了探索模型输入对模糊神经网络预报模型预测性能的影响,研究了通过减少预报模型自变量组合的复共线性影响,并结合相似系数计算分析方法建立了一种新的模糊神经网络预报模型。以气象学科的逐日降水预报作为研究对象,利用这种新的模糊神经网络预报模型进行了实际预报试验,并与常规的模糊神经网络预报方法,中国气象局T213数值预报模式以及逐步回归预报方法的预报结果进行了对比分析。结果表明,这种基于条件数和相似系数计算的模糊神经网络预报新方法对49天降水的独立样本预报平均绝对误差为7.33 mm,预报误差比模糊神经网络预报模型下降了5.9%,比传统的逐步回归方法下降了14.9%,比中国气象局T213数值预报模式的预报结果下降了13.4%。显示了很好的应用前景。     

基于ANFIS的焦炉火道温度预报模型研究

针对焦炉生产过程中直接检测火道温度成本高、精度低等问题,提出运用自适应神经网络模糊推理系统理论(ANFIS)建立焦炉火道温度预报模型,模型采用模糊减法聚类方法选取模糊规则数目,大大减少规则冗余量;结合最小二乘和误差反向传播混合算法对神经网络参数进行优化,采用现场的热工数据作为输入,将获得的模型与传统的线性回归模型和BP神经网络模型进行了比较,数值仿真结果表明所建立的模型具有学习速度快、预报精度高、泛化能力强等优点.     

基于神经网络的转炉炼钢终点控制

转炉炼钢是一种非常重要的炼钢方法,终点控制是转炉炼钢后期的重要操作.由于冶炼过程温度极高,很难进行准确及时地测量,无法形成通常意义下的反馈控制.采用RBF神经网络预报转炉炼钢终点温度和碳含量,在此基础上提出了基于神经网络的动态终点控制方法来确定在补吹阶段需要的吹氧量和加入的冷却剂量,克服了传统控制方法中基于热平衡和氧平衡控制模型不准确的缺点,提高了终点命中率.     

模拟退火免疫算法及其在铁水含硅量预报中的应用

本文提出了一种模拟退火免疫算法，该算法借鉴生物免疫系统的免疫识别、多样性及学习功能，利用基于模拟退火的浓度调节抗体多样性保持机理克服遗传算法易早熟收敛的缺点．将此方法用于优化铁水含硅量神经网络预报模型中的连接权值和阈值，可避免陷入局部极小，从而得到最佳神经网络，提高铁水含硅量预报精度. 仿真结果证明了方法的有效性．     

一种特殊的神经网络结构及其应用

本文针对目前处理信息数据处理方法上的不完善之处，提出了一种特殊的神经网络结构。以预报对流层大气折射指数的水平不均匀和时变漂移为例，说明了该神经网络模型的建立方法。它的特点主要是能充分利用各种类型的信息，克服了目前网络只能用同类信息为样本的缺点。仿真结果表明该方法用于金融数据预报也优于传统方法。     

基于RBF神经网络的粮库温度预测

由于粮库温度是非线性的时间序列,文章提出了基于RBF神经网络的粮库温度预测模型。该模型优于传统的粮库温度分析方法,又避免了BP算法容易陷入局部极小点和收敛速度慢的缺点。根据实验的仿真结果显示,该模型对于粮库温度的预测效果较好。     

基于RBF神经网络的钢坯温度预报软测量模型研究

钢坯加热过程是钢铁企业热轧生产中非常重要的工艺环节。钢坯温度预报模型是实现加热炉优化控制的重要基础,用常规仪器很难直接测量出钢坯温度。给出了基于RBF神经网络的软测量模型结构,对钢坯温度进行预报的仿真结果。     

Multi-step Learning Rule for Recurrent Neural Models: An Application to Time Series Forecasting

Multi-step prediction is a difficult task that has attracted increasing interest in recent years. It tries to achieve predictions several steps ahead into the future starting from current information. The interest in this work is the development of nonlinear neural models for the purpose of building multi-step time series prediction schemes. In that context, the most popular neural models are based on the traditional feedforward neural networks. However, this kind of model may present some disadvantages when a long-term prediction problem is formulated because they are trained to predict only the next sampling time. In this paper, a neural model based on a partially recurrent neural network is proposed as a better alternative. For the recurrent model, a learning phase with the purpose of long-term prediction is imposed, which allows to obtain better predictions of time series in the future. In order to validate the performance of the recurrent neural model to predict the dynamic behaviour of the series in the future, three different data time series have been used as study cases. An artificial data time series, the logistic map, and two real time series, sunspots and laser data. Models based on feedforward neural networks have also been used and compared against the proposed model. The results suggest than the recurrent model can help in improving the prediction accuracy.     

Reliable optimisation control of a reactive polymer composite moulding process using ant colony optimisation and bootstrap aggregated neural networks

This paper presents a study on the optimisation control of a reactive polymer composite moulding process using ant colony optimisation and bootstrap aggregated neural networks. In order to overcome the difficulties in developing accurate mechanistic models for reactive polymer composite moulding processes, neural network models are developed from process operation data. Bootstrap aggregated neural networks are used to enhance model prediction accuracy and reliability. Ant colony optimisation is able to cope with optimisation problems with multiple local optima and is able to find the global optimum. Ant colony optimisation is used in this study to find the optimal curing temperature profile. In order to enhance the reliability of the optimisation control policy, model prediction confidence bound offered by bootstrap aggregated neural networks is incorporated in the optimisation objective function so that unreliable predictions are penalised. The proposed method is tested on a simulated reactive polymer composite moulding process.     

An intelligent neural-fuzzy model for an in-process surface roughness monitoring system in end milling operations

In this research, a new intelligent neural-fuzzy in-process surface roughness monitoring (INF-SRM) system for an end milling operation was developed. The success of the INF-SRM system depends on an accurate decision-making algorithm, which can analyze the input factors and then generate an accurate output. A new neural-fuzzy model was proposed and implemented as decision-making algorithm for the INF-SRM system. The objective of the new model is to achieve higher accuracy for surface roughness prediction and solve the disadvantages of both neural networks and fuzzy logic. The neural-assisted method was implemented to generate the fuzzy IF-THEN rules for the model. To evaluate the performance of the new neural-fuzzy model, a neural networks model was applied to develop another surface roughness monitoring system for comparison. A statistical method was finally employed to analyze the accuracy between these systems.     

蓄热式加热炉钢温预报与炉温优化设定研究

在分析新型蓄热式钢坯加热炉的炉型结构与炉温制度特点的基础上,基于钢坯在炉内受热辐射和钢坯内热传导的机理,设计了二维非稳态导热钢温预报模型,并给出了适用该模型的炉温优化策略.该模型能够对出炉钢坯进行网格化钢温预报,具有精度高,通用性强,能够实现炉内钢坯温度分布的在线预报的特点.仿真研究表明,所提出的炉温优化策略能较大地提高钢坯的加热质量.     

改进人工蜂群算法优化RBF神经网络的短时交通流预测

为了提高径向基函数RBF神经网络预测模型对短时交通流的预测准确性,提出了一种基于改进人工蜂群算法优化RBF神经网络的短时交通流预测模型。利用改进人工蜂群算法确定RBF网络隐含层的中心值以及隐含层单元数,然后训练改进的人工蜂群算法RBF神经网络预测模型,并将其应用到某城市4天的短时交通流量数据的验证。将实验结果与传统RBF神经网络预测模型、BP神经网络预测模型和小波神经网络预测模型进行了比较。对比结果表明,该方法对短时交通流具有更高的预测准确性。     

Prediction of indoor temperature and relative humidity using neural network models: model comparison

The use of neural networks grows great popularity in various building applications such as prediction of indoor temperature, heating load and ventilation rate. But few papers detail indoor relative humidity prediction which is an important indicator of indoor air quality, service life and energy efficiency of buildings. In this paper, the design of indoor temperature and relative humidity predictive neural networks in our test house was developed. The test house presented complicated physical features which are difficult to simulate with physical models. The work presented in this paper aimed to show the suitability of neural networks to perform predictions. Nonlinear AutoRegressive with eXternal input (NNARX) model and genetic algorithm were employed to construct networks and were detailed. The comparison between the two methods was also made. Applicability of some important mathematical validation criteria to practical reality was examined. Satisfactory results with correlation coefficients 0.998 and 0.997 for indoor temperature and relative humidity were obtained in the testing stage.     

Artificial neural networks to optimize the extrusion of an aluminium alloy

Extrusion of aluminium alloys is a complex process which depends on the characteristics of the material and on the process parameters (initial billet temperature, extrusion ratio, friction at the interfaces, die geometry etc.). The temperature profile at the die exit, largely influences microstructure, mechanical properties, and surface quality of an extruded product, consequently it is the most important parameter for controlling the process. In turn the temperature profile depends on other process variables whose right choice is fundamental to avoid surface damage of the extruded product. In the present work, two neural networks were implemented to optimize the aluminium extrusion process determining the temperature profile of an Al 6060 alloy (UNI 9006/1) at the exit of induction heater (ANN1) and at the exit of the die (ANN2). The three-layer neural networks with Levemberg Marquardt algorithm were trained with the experimental data from the industrial process. The temperature profiles, predicted by the neural network, closely agree with experimental values.     

基于神经网络的时序预测模型研究

古人云“以史为鉴”,说的是吸取历史的经验教训,对未来的情况做出预判或者改变。生活中,亦是存在相似的利用历史数据对未来变化趋势进行预测分析的时间序列问题。本文就时间序列一类的问题进行研究,探讨如何更好地根据历史统计数据,对未来的变化趋势进行预测分析。本文基于神经网络,以气象观测历史数据作为研究的对象,建立了气温变化时序预测模型。本模型利用大数据相关技术对数据进行特征处理,通过深度神经网络,学习特征数据和标签数据之间复杂的非线性关系,从而实现对气温变化的趋势预测。实验结果表明,相较其他模型,本文的模型能够更好地进行时序预测,同时也证明了神经网络用于气象预测的可行性。     

A reliable multi-objective control strategy for batch processes based on bootstrap aggregated neural network models

This paper presents a reliable multi-objective optimal control method for batch processes based on bootstrap aggregated neural networks. In order to overcome the difficulty in developing detailed mechanistic models, bootstrap aggregated neural networks are used to model batch processes. Apart from being able to offer enhanced model prediction accuracy, bootstrap aggregated neural networks can also provide prediction confidence bounds indicating the reliability of the corresponding model predictions. In addition to the process operation objectives, the reliability of model prediction is incorporated in multi-objective optimisation in order to improve the reliability of the obtained optimal control policy. The standard error of the individual neural network predictions is taken as the indication of model prediction reliability. The additional objective of enhancing model prediction reliability forces the calculated optimal control policies to be within the regions where the model predictions are reliable. By such a means, the resulting control policies are reliable. The proposed method is demonstrated on a simulated fed-batch reactor and a simulated batch polymerisation process. It is shown that by incorporating model prediction reliability in the optimisation criteria, reliable control policy is obtained.     

Dynamic neural networks for gas turbine engine degradation prediction,health monitoring and prognosis

In this paper, the problem of health monitoring and prognosis of aircraft gas turbine engines is considered by using computationally intelligent methodologies. Two different dynamic neural networks, namely the nonlinear autoregressive with exogenous input neural networks and the Elman neural networks, are developed and designed for this purpose. The proposed dynamic neural networks are designed to capture the dynamics of two main degradations in the gas turbine engine, namely the compressor fouling and the turbine erosion. The health status and condition of the engine in terms of the turbine output temperature (TT) are then predicted subject to occurrence of these deteriorations. Various scenarios consisting of fouling and erosion separately as well as combined are considered. For each scenario, several neural networks are trained and their performance in predicting multiple flights ahead TTs is evaluated. Finally, the most suitable neural networks for achieving the best prediction are selected by using the normalized Bayesian information criterion model selection. Simulation results presented demonstrate and illustrate the effective performance of our proposed neural network-based prediction and prognosis strategies.