2022亚太建模A题Feature Extraction of Sequence Images and Modeling Analysis of Mold Flux Melting and Crystallization思路分析

简介: 2022 亚太建模A题序列图像的特征提取与建模分析 模具流量的熔融和结晶Feature Extraction of Sequence Images and Modeling Analysis ofMold Flux Melting and Crystallization

2022亚太建模A题 序列图像的特征提取与建模分析 模具流量的熔融和结晶


image.png


点击链接加入聊【2022亚太建模竞赛交流】:https://jq.qq.com/?_wv=1027&k=QSH410XM


点击链接加入聊【2022亚太建模竞赛交流】:https://jq.qq.com/?_wv=1027&k=QSH410XM

点击链接加入聊【2022亚太建模竞赛交流】:https://jq.qq.com/?_wv=1027&k=QSH410XM


2022亚太建模思路免费获取:297310168

2022年亚太地区数学建模大赛

问题A

序列图像的特征提取与建模分析

模具流量的熔融和结晶

连铸过程中的模具通量对钢半月板进行热绝缘,防止液态钢连铸过程中液态钢再氧

化,控制传热,提供链润滑,吸收非金属夹杂物。模具通量的冶金功能主要由温度

控制曲线下的熔化速率和结晶速率决定。因此,研究模具壁与股壳间隙中模具通量

的相分布具有重要意义。

在模具的顶部加入连铸模具通剂。这些固体矿渣以粉末层的形式堆积在液态钢的表

面,可以防止由于液态钢的温度下降过高而导致的液态钢水平结皮。模具通量的温

度随后逐渐上升到熔点,模具通量被熔化形成烧结层。模具通量的原料通过化学反

应形成低熔点物质,然后形成液渣,模具通量的组成会发生一定程度的变化。这是

熔化过程。

由于模具焊剂完全熔合,将形成液渣层并覆盖在液态钢表面。当液渣从钢液表面的

渣池渗入壳体与铜模壁的间隙时,形成渣膜。由于链表面的高温,对链的矿渣仍然

保持液相。但随着液渣的温度随模具纵向表面的降低,渣膜相对于铜模壁,淬火固

化形成玻璃固体渣膜(渣膜的固化行为),经过模具的强制冷却,而渣膜在一定区

域结晶形成结晶层(渣膜结晶行为),最终形成典型的三层渣膜结构:玻璃层、结

晶层、液渣层。这个过程是结晶。

由于高温、瞬态流体流动、复杂的相变和化学反应以及模具壁的不透明度,很难直

接观察模具通量的相位变化。SHTT II的熔化和结晶温度测试仪

12

目前已广泛应用于观察模具通量的结晶行为。实验结束后,实验人员逐一对图像进

行演示,记录在图像左上角,用肉眼和经验识别关键节点图像(见图1),指导模

具通量设计,满足钢级的凝固要求。这一过程浪费了人力,阻碍了实验过程信息的

开发。发展序列图像的自动特征提取和数学建模技术是当务之急。

附件1有562张模具通量的熔化和结晶的序列图像。这些序列图像是从实验开始时的

第110秒到第671秒之间收集起来的。文件序列号遵循采集时间序列,每隔1秒采集

一次图像。该信息由附件1中的数字图像呈现(见图1)。每幅图像的左上角都标记

有图像对应的时间和No. 1热电偶和No. 2热电偶的温度值。

(a)样本

(b)样品开始融化

在110秒

(c)这个样品是完全的

在141秒熔化

(d)样品可以看到钢

在149秒熔化

(e)结晶开始于

150s

(f)结晶结束于

671s

图1模具通量熔化结晶

为了实现模通量熔化和结晶序列图像的自动特征提取和数学建模,请回答以下三个

问题。

问题1:通过图像分割和识别等技术,请自动提取每个图像左上角的No. 1热电偶和

No. 2热电偶的温度,并自动导入到相应的表中

2022亚太建模思路免费获取:2973101683

附件2(请逐编写技术操作文件),制作温度时间曲线图(

1线温度2线温度时间图

;1线平均温度2线平均温度时间图)。另外,1#线或2#线的测试结果不准确。请把

它指出来并解释一下。

问题2:根据图1中的6个节点图像,应用数字图像处理技术,研究并量化模通量熔

融和结晶过程中相邻序列图像之间的动态差异。在此基础上,请对量化的不同特性

进行时间序列建模,并根据数学模型的仿真结果,讨论模具通量的熔化和结晶过程

曲线。

问题3:鉴于温度和时间的变化,以及问题2的研究结果,请建立数学模型,讨论温

度与时间变化的函数关系以及模具通量的熔融结晶过程,并根据数值模拟结果讨论

模具通量的熔化结晶动力学(温度、熔化速率和结晶速率的关系)。

您的PDF解决方案总共不超过25页,应该包括:

l 一页汇总表。目录。您

l 的完整解决方案。

l

注:APMCM竞赛的页面限制版数为25页。您提交的所有方面都达到了25页的限制(

摘要表,目录,您的完整解决方案)。但是,参考文献列表和附录的页面并不受限

制。

附件:

附件1.zip,请在website:https://share.weiyun.com/ubtXPGz0上下载

附件2.xlsx

2022亚太建模思路免费获取:297310168





1

Problem A

Feature Extraction of Sequence Images and Modeling Analysis of

Mold Flux Melting and Crystallization

Mold fluxes in continuous casting process thermally insulate the molten steel meniscus,

prevent reoxidation of liquid steel during continuous casting of liquid steels, control heat

transfer, provide lubrication of strand, and absorb nonmetallic inclusions. Metallurgical

functions of mold flux is mainly determined by its melting rate and crystallization rate under

the temperature control curve. It is therefore important to study the phase distribution of mold

fluxes in the gap between mold wall and strand shell.

Continuous casting mold fluxes are added to the top of liquid steel in the mold. These solid

slags, accumulating on the surface of liquid steel as a powder layer, can prevent liquid steel

level crusting due to excessive temperature drop of liquid steel. The temperature of mold fluxes

then gradually rise to the melting point, and mold fluxes are melted to form a sintered layer.

Raw materials of mold fluxes form low-melting-point substances and then liquid slag through

chemical reactions, and the composition of mold fluxes will change to a certain extent. It is

melting process.

As mold fluxes are completely fused, a liquid slag layer will be formed and covering on

the surface of liquid steel. The slag film will be formed when the liquid slag infiltrates from the

slag pool at the steel liquid surface into the gap between the shell and the copper mold wall.

The slag against the strand still maintains liquid phase, because of the high temperature of the

strand surface. However, as the temperature of liquid slag decreases with that of the strand

surface in the longitudinal direction of the mold, the slag film, against the copper mold wall, is

quenched and solidified to form a glassy solid slag film (solidification behavior of slag film),

with mold’s forced cooling, while slag film will crystallize at certain areas and form a

crystalline layer (crystallization behavior of slag film) under suitable conditions, finally

creating a typical three-layer slag film structure: glassy layer, crystalline layer and liquid slag

layer. This process is crystallization.

Because of the high temperature, transient fluid flow, complex phase transitions and

chemical reactions as well as the opacity of mold wall, it is difficult to observe the phase

changes of mold fluxes directly. The SHTT II tester of melting and crystallization temperature

2022 Asia and Pacific Mathematical Contest in Modeling2

is now widely applied to observe the crystallization behaviors of mold fluxes. After the

experiment is finished, experimenters demonstrate the images one by one, record the

information in the upper left corner of the images, and identify the key node images with naked

eyes and experience (see Figure 1), so as to guide the design of mold fluxes to meet the

solidification requirements of steel grades. This process wastes manpower and hinders the

development of experimental process information. It is urgent to develop automatic feature

extraction and mathematical modeling technology of sequence images.

Attachment 1 has 562 sequence images of mold fluxes’ melting and crystallization. These

sequence images are collected from the 110th to 671st seconds when the experiment starts. The

file serial numbers follow the collection time sequence, and images are collected every 1s. The

information is presented by digital images in Attachment 1 (see Figure 1). The upper left corner

of each image is marked with the corresponding time of the image and the temperature values

of No.1 thermocouple and No.2 thermocouple.

(a) The sample

(b) The sample begins to melt

at 110s

(c) The sample is completely

melted at 141s

(d) The sample sees steels

melting at 149s

(e) Crystallization begins at

150s

(f) Crystallization ends at

671s

Fig. 1 Melting and crystallization of mold fluxes

To realize automatic feature extraction and mathematical modeling of sequence images of

mold fluxes melting and crystallization, please answer the following three questions.

Question 1: With image segmentation and recognition or other technologies, please

automatically extract temperatures of No.1 thermocouple and No.2 thermocouple in the upper

left corner of each image and import them automatically into the corresponding table in 3

Attachment 2 (please write a step-by-step technical operation document), and please make a

temperature-time curve diagram (1#wire temperature-2#wire temperature-time diagram;

1#wire average temperature-2#wire average temperature-time diagram). In addition, the test

result of 1#wire or 2 #wire is inaccurate. Please point it out and explain.

Question 2: According to the six node images in Fig.1, please study and quantify the

dynamic differences between adjacent sequence images in the process of mold fluxes melting

and crystallization by applying digital image processing technology. On this basis, please make

a time series modeling of the quantified different characteristics, and discuss the melting and

crystallization process curve of mold fluxes based on the simulation results of mathematical

model.

Question 3: Given the changes of temperature and time, as well as the research results of

Question 2, please make a mathematical model to discuss the functional relationship between

the changes of temperature and time as well as the melting and crystallization process of mold

fluxes, and discuss the kinetics of melting and crystallization of mold fluxes (the relationship

between temperature, melting rate and crystallizing rate) based on numerical simulation results.

Your PDF solution of no more than 25 total pages should include:

l One-page Summary Sheet.

l Table of Contents.

l Your complete solution.

Note: The APMCM Contest has a 25-page limit. All aspects of your submission count toward

the 25-page limit (Summary Sheet, Table of Contents, your complete solution). However, the

pages of Reference List and Appendices are not limited.

Attachment:

Attachment 1.zip, download on the website: https://share.weiyun.com/ubtXPGz0

Attachment 2.xlsx

相关文章
|
3月前
|
存储 算法 计算机视觉
【博士每天一篇文献-模型】Meta-Learning Based Tasks Similarity Representation for Cross Domain Lifelong Learning
本文提出了一种基于元学习的跨域终身学习框架,通过跨域三元组网络(CDTN)学习任务间的相似性表示,并结合自注意模块与软注意网络(SAN)来增强特征提取和任务权重分配,以提高学习效率并减少对先前知识的遗忘。
49 1
【博士每天一篇文献-模型】Meta-Learning Based Tasks Similarity Representation for Cross Domain Lifelong Learning
|
1月前
|
机器学习/深度学习 Web App开发 人工智能
轻量级网络论文精度笔(一):《Micro-YOLO: Exploring Efficient Methods to Compress CNN based Object Detection Model》
《Micro-YOLO: Exploring Efficient Methods to Compress CNN based Object Detection Model》这篇论文提出了一种基于YOLOv3-Tiny的轻量级目标检测模型Micro-YOLO,通过渐进式通道剪枝和轻量级卷积层,显著减少了参数数量和计算成本,同时保持了较高的检测性能。
33 2
轻量级网络论文精度笔(一):《Micro-YOLO: Exploring Efficient Methods to Compress CNN based Object Detection Model》
|
机器学习/深度学习 人工智能 自然语言处理
OneIE:A Joint Neural Model for Information Extraction with Global Features论文解读
大多数现有的用于信息抽取(IE)的联合神经网络模型使用局部任务特定的分类器来预测单个实例(例如,触发词,关系)的标签,而不管它们之间的交互。
183 0
|
人工智能 自然语言处理 算法
【论文精读】AAAI 2022 - OneRel Joint Entity and Relation Extraction with One Module in One Step
联合实体和关系提取是自然语言处理和知识图构建中的一项重要任务。现有的方法通常将联合提取任务分解为几个基本模块或处理步骤,以使其易于执行
207 0
|
数据挖掘
PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation-ppt版学习笔记
PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation-ppt版学习笔记
129 0
|
机器学习/深度学习 存储 自然语言处理
Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation 论文笔记
Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation 论文笔记
|
机器学习/深度学习 人工智能 计算机视觉
Re2:读论文 CS-GNN Measuring and Improving the Use of Graph Information in Graph Neural Networks
Re2:读论文 CS-GNN Measuring and Improving the Use of Graph Information in Graph Neural Networks
Re2:读论文 CS-GNN Measuring and Improving the Use of Graph Information in Graph Neural Networks
|
算法 数据挖掘 知识图谱
Self-supervised graph convolutional network for multi-view clustering(论文阅读记录)
IEEE TRANSACTIONS ON MULTIMEDIA 2区top 2022 影响因子:6.005
434 0
|
机器学习/深度学习 存储 自然语言处理
【推荐系统论文精读系列】(八)--Deep Crossing:Web-Scale Modeling without Manually Crafted Combinatorial Features
人工制作的组合特征是许多成功模型背后的 "秘诀"。然而,对于网络规模的应用来说,特征的种类和数量使得这些手工制作的特征在创建、维护和部署时成本高昂。本文提出了Deep Crossing模型,它是一个深度神经网络,可以自动结合特征来产生卓越的模型。Deep Crossing的输入是一组单独的特征,可以是密集的也可以是稀疏的。重要的交叉特征是由网络隐含地发现的,网络由嵌入和堆叠层以及残余单元的级联组成。
263 0
|
机器学习/深度学习 自然语言处理 搜索推荐
【推荐系统论文精读系列】(九)--Product-based Neural Networks for User Response Prediction
预测用户的反应,如点击和转换,是非常重要的,并已发现它在许多网络应用程序中使用,包括推荐系统、网络搜索和在线广告。这些应用程序中的数据大多是分类的,包含多个字段;典型的表示是通过一个热编码将其转换为高维稀疏二进制特征表示。面对极端稀疏性,传统模型可能会限制其从数据中挖掘浅层模式的能力,即低阶特征组合。另一方面,像深度神经网络这样的深度模型由于其巨大的特征空间而不能直接应用于高维输入。在本文中,我们提出了一种基于产品的神经网络(PNN),其嵌入层用于学习分类数据的分布式表示,产品层用于捕获场间类别之间的交互模式,以及进一步的全连接层用于探索高阶特征交互。我们在两个大规模真实世界广告点击数据集上的
266 0
【推荐系统论文精读系列】(九)--Product-based Neural Networks for User Response Prediction