Hybrid Model of Molten Steel Temperature Prediction Based on Ladle Heat Status and Artificial Neural Network

Aiming at the characteristics of the practical steelmaking process, a hybrid model based on ladle heat sta- tus and artificial neural network has been proposed to predict molten steel temperature. The hybrid model could over- come the difficulty of accurate prediction using a single mathematical model, and solve the problem of lacking the consideration of the influence of ladle heat status on the steel temperature in an intelligent model. By using the hybrid model method, forward and backward prediction models for molten steel temperature in steelmaking process are es- tablished and are used in a steelmaking plant. The forward model, starting from the end-point of BOF, predicts the temperature in argon-blowing station, starting temperature in LF, end temperature in LF and tundish temperature forwards, with the production process evolving. The backward model, starting from the required tundish tempera- ture, calculates target end temperature in LF, target starting temperature in LF, target temperature in argon-blo- wiag station and target BOF end-point temperature backwards. Actual application results show that the models have better prediction accuracy and are satisfying for the process of practical production.

Artificial neural network modeling of mechanical properties of armor steel under complex loading conditions

An artificial neural network （ANN） model is established to predict plastic flow behaviors of the 603 armor steel, based on experiments over wide ranges of strain rates （0. 001 -4 500 s -1 ） and temperatures （288 -873 K）. The descriptive and predictive capabilities of the ANN model are com- pared with several phenomenological and physically based constitutive models. The ANN model has a much better applicability than the other models in characterization of the flow stress. The tempera- ture and the strain rate effects on the flow stress can be described successfully by the ANN model, with an average error of 1.78% for both quasi-static and dynamic loading conditions. Besides its high accuracy in prediction of the strain rate jump tests, the ANN model is more convenient in model es- tablishment and data processing. The ANN model developed in this study may serve as a valid and ef- fective tool to predict plastic behaviors of the 603 steel under complex loading conditions.

A Predictive Modeling Based on Regression and Artificial Neural Network Analysis of Laser Transformation Hardening for Cylindrical Steel Workpieces

Laser surface hardening is a very promising hardening process for ferrous alloys where transformations occur during cooling after laser heating in the solid state. The characteristics of the hardened surface depend on the physicochemical properties of the material as well as the heating system parameters. To exploit the benefits presented by the laser hardening process, it is necessary to develop an integrated strategy to control the process parameters in order to produce desired hardened surface attributes without being forced to use the traditional and fastidious trial and error procedures. This study presents a comprehensive modelling approach for predicting the hardened surface physical and geometrical attributes. The laser surface transformation hardening of cylindrical AISI 4340 steel workpieces is modeled using the conventional regression equation method as well as artificial neural network method. The process parameters included in the study are laser power, beam scanning speed, and the workpiece rotational speed. The upper and the lower limits for each parameter are chosen considering the start of the transformation hardening and the maximum hardened zone without surface melting. The resulting models are able to predict the depths representing the maximum hardness zone, the hardness drop zone, and the overheated zone without martensite transformation. Because of its ability to model highly nonlinear problems, the ANN based model presents the best modelling results and can predict the hardness profile with good accuracy.

Prediction of Al(OH)_3 fluidized roasting temperature based on wavelet neural network

The recycle fluidization roasting in alumina production was studied and a temperature forecast model was established based on wavelet neural network that had a momentum item and an adjustable learning rate. By analyzing the roasting process, coal gas flux, aluminium hydroxide feeding and oxygen content were ascertained as the main parameters for the forecast model. The order and delay time of each parameter in the model were deduced by F test method. With 400 groups of sample data (sampled with the period of 1.5 min) for its training, a wavelet neural network model was acquired that had a structure of {7 211}, i.e., seven nodes in the input layer, twenty-one nodes in the hidden layer and one node in the output layer. Testing on the prediction accuracy of the model shows that as the absolute error ±5.0 ℃ is adopted, the single-step prediction accuracy can achieve 90% and within 6 steps the multi-step forecast result of model for temperature is receivable.

Final Temperature Prediction Model of Molten Steel in RH-TOP Refining Process for IF Steel Production

In order to precisely control the final temperature of molten steel in RH （Ruhrstahl Heraeus）-TOP blowing refining, the final temperature prediction models of molten steel in RH-TOP blowing refining process for Interstitial Free （IF） steel production were established under the condition of oxygen blowing and non-oxygen blowing respec- tively. The results show that the beginning molten steel temperature of refining and the amount of added scrap were influential factors, the baking temperature in vacuum chamber was a factor that had small influence. When the model was operated, the hitting probability was above 95%（under the condition of both oxygen blowing and non-oxygen blo- wing） of prediction deviation ±10℃. The accuracy is analyzed.

Monthly Mean Temperature Prediction Based on a Multi－level Mapping Model of Neural Network BP Type

In terms of 34-year monthly mean temperature series in 1946-1979,the multi-level maPPing model of neural netWork BP type was applied to calculate the system's fractual dimension Do=2'8,leading tO a three-level model of this type with ixj=3x2,k=l,and the 1980 monthly mean temperture predichon on a long-t6rm basis were prepared by steadily modifying the weighting coefficient,making for the correlation coefficient of 97% with the measurements.Furthermore,the weighhng parameter was modified for each month of 1980 by means of observations,therefore constrcuhng monthly mean temperature forecasts from January to December of the year,reaching the correlation of 99.9% with the measurements.Likewise,the resulting 1981 monthly predictions on a long-range basis with 1946-1980 corresponding records yielded the correlahon of 98% and the month-tO month forecasts of 99.4%.

A Study on Prediction of Weld Geometry in Laser Overlap Welding of Low Carbon Galvanized Steel Using ANN-Based Models

Predictive modelling for quality analysis becomes one of the most critical requirements for a continuous improvement of reliability, efficiency and safety of laser welding process. Accurate and effective model to perform non-destructive quality estimation is an essential part of this assessment. This paper presents a structured approach developed to design an effective artificial neural network based model for predicting the weld bead dimensional characteristic in laser overlap welding of low carbon galvanized steel. The modelling approach is based on the analysis of direct and interaction effects of laser welding parameters such as laser power, welding speed, laser beam diameter and gap on weld bead dimensional characteristics such as depth of penetration, width at top surface and width at interface. The data used in this analysis was derived from structured experimental investigations according to Taguchi method and exhaustive FEM based 3D modelling and simulation efforts. Using a factorial design, different neural network based prediction models were developed, implemented and evaluated. The models were trained and tested using experimental data, supported with the data generated by the 3D simulation. Hold-out test and k-fold cross validation combined to various statistical tools were used to evaluate the influence of the laser welding parameters on the performances of the models. The results demonstrated that the proposed approach resulted successfully in a consistent model providing accurate and reliable predictions of weld bead dimensional characteristics under variable welding conditions. The best model presents prediction errors lower than 7% for the three weld quality characteristics.

Application of Nonlinear Predictive Control Based on RBF Network Predictive Model in MCFC Plant

This paper described a nonlinear model predictive controller for regulating a molten carbonate fuel cell (MCFC). A detailed mechanism model of output voltage of a MCFC was presented at first. However, this model was too complicated to be used in a control system. Consequently, an off line radial basis function (RBF) network was introduced to build a nonlinear predictive model. And then, the optimal control sequences were obtained by applying golden mean method. The models and controller have been realized in the MATLAB environment. Simulation results indicate the proposed algorithm exhibits satisfying control effect even when the current densities vary largely.

基于自适应小波回声神经网络的光纤陀螺测角仪温度误差补偿技术

基于光纤陀螺的测角仪可以实现对各项角运动参数的一体化动态精密测量,但在实际应用中,光纤陀螺测角仪受到温度变化的影响,导致测量精度下降。针对这一问题,本文提出了一种基于自适应小波回声神经网络的光纤陀螺测角仪温度误差补偿技术。为了提高温度误差建模的进度,提高传统神经网络的逼近能力,通过自适应前向线性预测滤波器对建模用测角仪温度漂移数据进行预处理,并采用自适应小波回声神经网络建立温度漂移模型,能够避免传统神经网络结构设计的盲目性和局部最优等问题,增强了网络学习能力和泛化能力,并利用自适应律代替神经网络梯度进行网络训练,提升神经网络的逼近精度和收敛速度。实验结果表明,该模型可以提高光纤陀螺测角仪的测量精度和环境适应性,为光纤陀螺测角仪的性能优化和实际应用提供了可靠的技术支撑。

弹药贮存微环境温度智能化预测

高效、准确获取弹药贮存微环境温度变化规律是开展弹药寿命评估与延寿的关键。结合神经网络算法以及传热学原理,研究了弹药贮存微环境温度预测模型,并基于此模型开发了一套弹药包装箱微环境预测软件,用以预测不同气候环境下包装箱内部的温度变化。在敦煌、漠河进行了相关试验,用以验证软件预测正确性以及参数优化。结果表明:应用基于神经网络算法开发的包装箱微环境预测软件,根据不同包装箱材料,调控贮存地点以及控制贮存起始温度,可以实现对箱内温度的控制。

钢拱塔斜拉桥的温度耦合效应和索力预测

钢拱塔斜拉桥的受力体系与传统斜拉桥有所不同,为研究环境温度变化对这种异形桥塔斜拉桥主要受力部件的影响,以某钢拱塔斜拉桥为工程背景,首先基于在线监测获取的环境和部件温度数据,分析斜拉索索力、拱塔倾角和主梁应变的温度时变效应;然后以斜拉索为研究对象,通过该桥的有限元模型升降温模拟,分析各部件温差引起的温度耦合效应对拉索索力的影响;最后以环境温度、主梁温度、桥塔温度为输入,索力为输出,利用长短期记忆神经网络对实测索力-温度数据进行映射,实现数据压缩和特征提取,建立温度-索力预测模型,再对网络模型输入新的温度监测数据,以预测索力。研究结果表明:主梁和钢拱塔温度变化具有周期性,且滞后于环境温度;主梁应变与环境温度的变化趋势基本一致但具有一定的滞后性,环境温度变化对拱塔倾角的影响很小且没有周期性规律;索力与环境温度呈线性负相关,且需要考虑斜拉桥各部件的温差所引起的温度耦合效应;长短期记忆神经网络对带有时序特性的数据训练效果好,建立的温度-索力关系模型准确度高,可用于该桥索力的实时预测。

基于LSTM-AT的温室空气温度预测模型构建

构建精确的温室空气温度预测模型是采用模型预测控制等控制算法实现温室空气温度精准控制的前提条件。长短记忆神经网络(long short-term memory,LSTM)以处理时间序列数据方面的优势而广泛应用于温室空气温度预测,然而其面对长时间序列数据存在由于数据遗忘而导致温室空气温度预测精度降低的问题。为解决以上问题,该研究将LSTM模型与注意力机制(attention mechanism,AT)结合构建LSTM-AT模型,根据LSTM模型隐藏层输出状态重要性程度为隐藏层输出分配权重,以提高温室空气温度长时间预测精度。该研究在不同预测时长(12、24和48 h)和不同天气状况两种情况下将LSTM-AT模型与递归神经网络、门控循环单元、双向长短记忆网络、LSTM模型进行对比。结果表明,LSTM-AT模型空气温度预测值与测量值变化趋势较为一致,模型计算值与空气温度测量值的决定系数最小为0.95,均方根误差最大为1.34℃,平均绝对误差最大为10.51%;LSTM-AT模型、LSTM模型、门控循环单元、递归神经网络、双向长短记忆网络5种模型温室空气温度预测均方根误差平均值分别为0.89、1.42、1.89、2.10、1.51℃,平均绝对百分比误差平均值分别为4.26%、8.96%、13.57%、17.70%、10.67%。由此可知,相较于其他4种模型,该研究提出的LSTM-AT模型具有更高的预测精度,能够精确预测温室空气温度。

基于机器学习的多气体指标煤自燃温度预测

采空区煤自燃是诱发矿井火灾的主要因素之一,在矿井火灾中的占比高达90%。为实现采空区自燃的精准防治,需要准确获得采空区内部的高温点温度,以砚北煤矿为工程背景,分析其煤样氧化升温过程中产生的指标气体,建立煤自燃温度预测的指标体系,进而开展基于深度学习的多指标气体煤自燃温度预测研究。首先对砚北煤矿采集的煤样进行煤氧化升温实验,根据实验结果划分为单一气体指标与复合气体指标,分析各指标随温度上升的变化规律,进而确定合适的指标作为煤自燃温度预测指标;使用多源数据处理方法对煤自燃温度预测指标进行处理,应用库克距离法和多重插补法对数据进行清洗,并结合灰色关联度分析法建立煤自燃温度预测指标体系;使用Elman神经网络构建预测模型,确定模型结构与超参数后,进而建立煤自燃温度预测模型,获得煤氧化升温过程中温度的准确预测。

气象数据驱动的拉林铁路简支T梁温致变形预测

为快速、准确预测日照环境下高原铁路桥梁的温度变形效应,基于长短期记忆神经网络(LSTM)提出了一种气象数据驱动的桥梁温度变形智能预测模型。以拉林铁路简支T梁为背景,采用热力耦合有限元仿真分析,构建了“气象数据—温度变形”映射的样本数据库,以此训练预测模型并对桥梁温致变形进行预测。结果表明,LSTM模型表现出了较高的精度和优势,其梁体竖向挠度预测的决定系数(R2)超过0.97,平均绝对误差(MAE)和均方根误差(RMSE)较反向传播神经网络(BP)模型提升超过70%,较随机森林(RF)模型分别提升了24%和27%。预测挠度与真实值在趋势和数值方面均基本一致,表明所提出的预测方法性能优异,为高原铁路轨道平顺性变化规律研究及动态检测数据评价提供参考。

一种基于GRU的氢燃料重卡汽车工况下锂离子电池温度预测模型

针对目前氢燃料重卡在行驶过程中,动力电池工况复杂、外表面温度变化难以预测、滞后时间长等问题,以氢燃料重卡锂离子动力电池外表面温度为研究对象,提出一种类交叉熵损失函数和自适应矩估计(adaptive moment estimation,Adam)优化的改进型门控循环单元神经网络(gate recurrent unit,GRU),建立锂离子动力电池表面温度预测模型。该模型利用GRU神经网络的特殊门机制和全局处理能力,得到锂离子电池表面温度和电池充放电电流、电压、充放电时间、历史温度、当前温度以及环境温度之间的非线性关系。采用4个精度评价函数对预测模型进行评价:经过5种环境温度下的模拟工况实验,验证该模型的准确性。结果表明,基于GRU的电池温度预测模型的误差相对于反向传播(back propagation,BP)神经网络模型和循环神经网络模型(recurrent neural network,RNN)来说较小,说明GRU的锂离子电池温度预测模型具有更高的精度。该文为磷酸铁锂电池表面温度的精准预测提出了一种新的方法。

基于MLP和Transformer模型的大气温度预测

文章以运城市2015年1月1日至2020年12月21日期间监测的大气温度数据作为研究的基础资料,运用MLP模型和Transformer模型,预测了运城市大气温度。由于温度数据具有很强的时序性,对MLP模型与Transformer模型,各选取了两层、四层(MLP-2、MLP-4、Transformer-2、Transformer-4),进行了3天、5天、7天多组试验对比。结果显示:MLP-4模型7天的均方误差为3.2649,Transformer-4模型3天的均方误差为5.3767,预测精度都比较高,且MLP模型预测温度的精度高于Transformer模型预测温度的精度;MLP-2模型的均方误差分别为3.2662、3.2996、3.3579,MLP-4模型的均方误差分别为3.2674、3.2996、3.2649,均方误差有变化,但比较平稳;Transformer-2模型的均方误差分别为5.6225、5.9491、5.3892,Transformer-4模型的均方误差分别为5.3767、6.3787、6.1108,增加模型层数和参数量,均方误差增大,存在过拟合现象。运用Transformer模型进行预测,出现过拟合现象,原因是Transformer模型太过庞大(接近四百万个参数),而研究数据只有1531组,即使使用Weight decay和Dropout正则化的方法,仍然过拟合文章中提供的1531组研究数据,使其预测精度出现一定程度的下降。

马钢120 t LF精炼炉钢液温度预测模型的应用与优化实践

针对LF精炼炉钢液温度控制过度依赖人工经验的问题,马钢长材事业部以120 t LF精炼炉为研究对象,基于能量平衡原理,计算分析LF精炼过程中输入电能、合金化、炉渣热效应、钢包内衬散热、渣面辐射、吹氩搅拌和烟气热损失等热量对钢液温度的影响,建立LF精炼钢液温度的预测模型。经过跟踪实际生产试验、测温校正并优化模型,使模型取得了良好的应用效果。模型预测温度与实际测量值偏差绝对值≤5℃的比例为97.73%,偏差绝对值≤6℃的比例为100%。

基于循环神经网络的动车组温度数据预测研究

采用循环神经网络建立了基于CRH5A型动车组温度类数据的预测模型,对影响预测结果的影响因子、模型层数及神经元个数进行了明确的界定,对CRH5A型动车组实车开展持续性追踪分析,采集动车组运行真实数据,进行积累和培养。在利用神经网络预测模型对数据进行训练后,CRH5A型动车组变压器温度峰值预测模型精度可达94.2%,牵引电机温度峰值预测模型精度可达93.8%,齿轮箱温度峰值预测模型精度可达95.3%,轴箱温度峰值预测模型精度可达92.7%。动车组温度数据预测结果的精确度可满足实际应用需求,预测模型在提高列车检修效率、节支降耗方面有着重要的作用。

基于县域尺度神经网络模型的中国钢铁工业污染物排放预测研究

实现精细尺度的钢铁工业排放预测对我国的区域污染控制和产业政策调整具有重要意义。该研究通过融合多源数据构建神经网络模型并实现2025—2060年的县域尺度活动水平预测,建立中国钢铁工业的大气污染物排放预测清单。结果表明,粗钢产量将在2025年达到最高值,且随后呈现缓慢下降趋势。同时,未来中西部地区将形成大型县域钢铁生产基地。2060年,我国粗钢产量预计为7.6亿t,较2015年下降4%,共排放SO_(2)、NO_(X)、PM_(2.5)及CO_(2)分别为10.6万t、8.6万t、25.5万t和5.9亿t,较2015年分别下降88%、89%、76%和52%。

基于改进SSA-BP神经网络的钠硫电池拆解刀具温度预测模型研究

钠硫电池中含有大量的高纯度钠,在自动化拆解过程中存在燃烧、爆炸等安全风险。针对钠硫电池在车削拆解时存在的安全性问题,提出一种改进SSA-BP神经网络算法来预测刀具加工的最高温度。利用ABAQUS软件计算出刀具加工的实时温度,通过电池拆解实验验证仿真数据的可靠性;然后以仿真温度数据建立样本,利用Tent混沌映射对SSA-BP神经网络算法进行优化,建立刀具温度仿真预测模型。实验结果表明:该仿真预测模型收敛速度快,鲁棒性强,模型误差小。