WO2020134210A1 - 一种冷轧带钢表面质量管理系统及方法 - Google Patents

一种冷轧带钢表面质量管理系统及方法 Download PDF

Info

Publication number
WO2020134210A1
WO2020134210A1 PCT/CN2019/106449 CN2019106449W WO2020134210A1 WO 2020134210 A1 WO2020134210 A1 WO 2020134210A1 CN 2019106449 W CN2019106449 W CN 2019106449W WO 2020134210 A1 WO2020134210 A1 WO 2020134210A1
Authority
WO
WIPO (PCT)
Prior art keywords
cold
defect
rolled strip
surface quality
strip surface
Prior art date
Application number
PCT/CN2019/106449
Other languages
English (en)
French (fr)
Inventor
夏志
何涛
周云根
Original Assignee
中冶南方工程技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中冶南方工程技术有限公司 filed Critical 中冶南方工程技术有限公司
Publication of WO2020134210A1 publication Critical patent/WO2020134210A1/zh

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
    • B21B—ROLLING OF METAL
    • B21B37/00—Control devices or methods specially adapted for metal-rolling mills or the work produced thereby
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
    • B21B—ROLLING OF METAL
    • B21B38/00—Methods or devices for measuring, detecting or monitoring specially adapted for metal-rolling mills, e.g. position detection, inspection of the product

Definitions

  • the invention belongs to the technical field of cold-rolled strip steel, and in particular relates to a cold-rolled strip surface quality management system and method.
  • the strip steel surface quality online inspection system has two limitations: first, the number of defective samples, uniformity and marking accuracy are not enough, resulting in a low recognition rate of the strip steel surface quality online inspection system; second, the identification of the strip After the steel surface quality defects, the lack of strip surface quality evaluation and quality improvement suggestions, making the strip steel surface quality online detection system actually converted into benefits is not obvious.
  • the purpose of the present invention is to overcome the problem that the on-line detection system for strip steel surface quality in the prior art actually translates into unobvious benefits.
  • the present invention provides a cold-rolled strip surface quality management system, including a cold-rolled strip surface quality defect identification module, a cold-rolled strip surface quality defect storage module, a cold-rolled strip surface quality defect sample marking module, Cold rolled strip surface quality evaluation module and cold rolled strip surface quality improvement module;
  • the cold-rolled strip surface quality defect identification module is used to detect the surface defects of the cold-rolled strip
  • the cold-rolled strip surface quality defect storage module is used to save the detected defects of the cold-rolled strip surface
  • the cold-rolled strip surface quality defect sample marking module is used to mark and store defect samples
  • the cold-rolled strip surface quality evaluation module is used to match the cold-rolled strip surface defects with the defect samples to determine the types of cold-rolled strip surface defects and the strip surface quality grade;
  • the cold-rolled strip surface quality improvement module is used to provide corresponding improvement strategies according to the types of cold-rolled strip surface defects and strip surface quality grades, and to classify, improve strategies and treat cold-rolled strip surface defects The result is stored.
  • the surface defects of the cold-rolled strip and the defect samples include defect type, steel coil number, defect characteristic value and defect picture path, and the defect characteristic value includes defect area, defect length, defect perimeter, and defect Area slenderness ratio and defect shape factor.
  • the cold-rolled strip surface quality evaluation module builds a strip surface quality evaluation model based on the analytic hierarchy process based on the defect type and defect characteristic value data of each coil in the defect sample, and quantitatively evaluates the sample strip surface quality grade .
  • an abnormal sample rejection module is further included, and the abnormal sample rejection module is configured to remove the abnormal body in the defect sample according to the defect feature value of the defect sample.
  • the abnormal sample rejection module performs the rejection based on the abnormal sample rejection method of the angular distance of the defect, which specifically includes:
  • the system further includes a classifier training module, the classifier training module includes a classification trainer, the classifier training module configures a classification trainer according to the characteristics of the defect sample, and the classification trainer Defect samples are read.
  • the classifier training module includes a classification trainer
  • the classifier training module configures a classification trainer according to the characteristics of the defect sample, and the classification trainer Defect samples are read.
  • the classification trainer includes one or more of a decision tree algorithm, an Adaboost algorithm or an artificial neural network algorithm.
  • the invention also provides a surface quality management method for cold-rolled strip steel, which includes the following steps:
  • S400 Provide a corresponding improvement strategy according to the category of the surface defect of the cold rolled strip, and store the category, the improvement strategy and the processing result of the surface defect of the cold rolled strip.
  • the cold rolled strip surface quality management system and method provided by the present invention In view of the deficiencies of the existing cold-rolled strip surface quality online inspection system, the establishment of defect samples, the establishment of cold-rolled strip surface quality defects treatment countermeasures table, the cold-rolled strip surface quality inspection and quality improvement are related, the cold-rolled strip The surface defects are matched with the defect samples to determine the surface quality grade of the cold-rolled strip surface defects, so as to guide the improvement of the production process and improve the surface quality of the cold-rolled strip; on the other hand, the angular distance method is used to remove abnormal samples to improve the defects The labeling accuracy of the sample.
  • the sample marks of the cold-rolled strip surface quality management system cover the typical defects of the cold-rolled strip surface quality (holes, edge cracks, folds, scars, indentations, roll marks, indentations, scratches, warping, oxidation) Iron sheet, etc.), the category can be expanded;
  • the cold-rolled strip surface quality management system processes the samples in the sample table of the cold-rolled strip surface quality knowledge base through the angular distance method, and removes abnormal samples based on the sample characteristic values to ensure that the samples are clean;
  • the cold-rolled strip surface quality management system builds a strip surface quality evaluation model based on the AHP method based on all defect types and defect level data of each coil in the defect table to quantitatively evaluate the strip surface quality;
  • the cold-rolled strip surface quality management system provides surface quality treatment improvement methods and suggestions for the defects that affect the cold-rolled strip surface quality, and guides the on-site quality improvement.
  • FIG. 1 is a schematic diagram of a module of a surface quality management system for cold-rolled steel strip of the present invention
  • FIG. 2 is a schematic diagram of the logic principle of the surface quality management system of the cold-rolled strip of the present invention
  • 3 is a flow chart of the elimination algorithm of the cold rolled strip surface quality management system and method of the present invention.
  • FIG. 5 is a schematic diagram of a user interface of the cold rolled strip surface quality management system and method of the present invention.
  • first and second are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated.
  • the features defined as “first” and “second” may expressly or implicitly include one or more of the features; in the description of the present invention, unless otherwise stated, the meaning of "plurality” is Two or more.
  • An embodiment of the present invention provides a cold rolled strip surface quality management system, including a cold rolled strip surface quality defect identification module F0, a cold rolled strip surface quality defect storage module F1, a cold rolled strip surface quality defect sample marking module F2.
  • the cold-rolled strip surface quality defect identification module is used to detect the surface defects of the cold-rolled strip
  • the cold-rolled strip surface quality defect storage module F1 is used to save the detected defect of the cold-rolled strip surface
  • the cold-rolled strip surface quality defect sample marking module F2 is used to mark and store defect samples
  • the cold-rolled strip surface quality evaluation module F5 is used to match the cold-rolled strip surface defects with the defect samples to determine the types of cold-rolled strip surface defects and the strip surface quality grade;
  • the cold-rolled strip surface quality improvement module F6 is used to provide corresponding improvement strategies according to the types of cold-rolled strip surface defects and strip surface quality grades, and to classify the cold-rolled strip surface defects types, improvement strategies and Store the processing results.
  • the cold-rolled strip surface quality management system includes a cold-rolled strip surface quality defect identification module F0, a cold-rolled strip surface quality defect storage module F1, and a cold-rolled strip surface quality defect sample mark Module F2, Cold Rolled Strip Surface Quality Evaluation Module F5 and Cold Rolled Strip Surface Quality Improvement Module F6.
  • a cold-rolled steel strip surface quality defect sample marking module F2 retrieve and browse the cold-rolled steel strip surface quality defect pictures, classify the defects, and then store the classified defect samples, which may appear in the actual process of the project.
  • the surface quality defects of the cold-rolled strip are classified and stored, mainly in the form of pictures.
  • the cold-rolled strip surface quality defect identification module F0 performs defect detection on the cold-rolled strip surface, and saves the detected defects of the cold-rolled strip surface to the cold-rolled strip surface quality defect storage module F1, and then cold
  • the surface quality evaluation module F5 of the rolled steel strip compares the detected surface defects of the cold rolled steel strip with the defect samples to determine the surface quality grade of the steel strip of the surface defects of the cold rolled steel strip, and finally passes the surface of the cold rolled steel strip
  • the quality improvement module F6 provides corresponding improvement strategies according to the types and grades of the surface defects of the cold-rolled strip, and stores the types, improvement strategies, and processing results of the surface defects of the cold-rolled strip.
  • the cold-rolled strip surface quality management system includes a cold-rolled strip surface quality knowledge base and an on-line surface quality inspection system.
  • the former is a data stream
  • the latter is a software and hardware carrier, mainly including User interface layer 1, logic layer 2, data access layer 3, and data and folder 4, user interface layer 1 receives user input and performs human-computer interaction; logic layer 2 performs logical and mathematical operations to process data; data access layer 3 pairs
  • the data is read/operated; the database and the picture folder 4 organize and store data, in which the database organizes and stores data, and the picture folder stores defective pictures.
  • the defects detected and identified by the surface quality online inspection system are stored in the defect table and the defect picture folder, and the samples are read from the sample table for classifier training and verification.
  • the user interface layer 1 can be a browser, which realizes the sharing of network data, obtains samples, removes sample marks and abnormal samples through the logic layer 2, and then enters the data access layer 3 to read and write tables to read defective samples Write operation, and then form the defect table and sample table in the database and the picture folder, and form the quality evaluation table and defect picture correspondingly.
  • the surface quality online detection system corresponds to the cold-rolled steel strip surface quality defect identification module F0, including online surface quality detection Components, hardware, identification software and online detection classifier software.
  • the surface defects of the cold-rolled steel strip and the defect sample include defect type, steel coil number, defect characteristic value and defect picture path, and the defect characteristic value includes defect area, defect length, defect perimeter, Defect area slenderness ratio and defect shape factor.
  • the defect samples and the defects of the surface of the cold-rolled strip are all defect types including defect type, steel coil number, defect characteristic value and defect picture path.
  • the defect characteristic value includes defect area, defect length, defect cycle Long, defective area slenderness ratio and defect shape factor.
  • the types of defect samples include holes, edge cracks, folds, scars, indentation, roller printing, indentation, scratches, peeling, iron oxide, etc.
  • the number of each type of defect samples is 300 to 500, which is the online inspection of the surface quality of the strip steel
  • the system classifier that is, the cold rolled strip surface quality evaluation module F5 learning and verification provides data support.
  • the cold-rolled strip surface quality evaluation module F5 builds a strip surface quality evaluation model based on the analytic hierarchy process (AHP) based on the defect type and defect characteristic value data of each coil in the defect sample, and quantitatively evaluates the samples Strip steel surface quality grade.
  • AHP analytic hierarchy process
  • the preferred solution is to build a strip surface quality evaluation model based on the AHP method based on all the defect types and defect level data for each coil of strip in the defect table to quantitatively evaluate the strip surface quality, as shown in Figure 4.
  • Layer D is a different level within each defect, and each level has a weight.
  • the weight of the defect level mainly considers the defect area, defect length, defect perimeter, defect area slenderness ratio and defect shape factor
  • the criterion layer C is the strip Various defect types of steel, each of which has a defect weight, which mainly considers the degree of influence of the defect on the surface quality.
  • the defect weight is solved by constructing a defect judgment matrix.
  • the scheme layer P is each steel plate with defects, including the volume Raw data of all surface quality defects of strip steel.
  • the surface quality grade of the strip steel defect characteristic value ⁇ level weight
  • the preferred solution further includes an abnormal sample rejection module F3, the abnormal sample rejection module F3 is configured to remove the abnormal body in the defect sample according to the defect feature value of the defect sample. It can be seen that by analyzing the samples, abnormal samples can be found and then eliminated to avoid affecting the accuracy of the samples.
  • the abnormal sample culling module F3 performs culling according to the abnormal sample culling method of the angular distance of the defect, which specifically includes:
  • the system further includes a classifier training module F4.
  • the classifier training module F4 includes a classification trainer.
  • the classifier training module configures a classification trainer according to the characteristics of the defect sample.
  • the classification trainer Read the defective sample.
  • the classifier training module F4 includes one or more of a decision tree algorithm, an Adaboost algorithm, or an artificial neural network algorithm.
  • the classifier training module F4 selects the characteristics of the surface defects of the preserved cold-rolled strip according to the process experience, configures the classifier parameters, then reads the defect sample table, and then trains and verifies the constructed classifier for later
  • the cold rolled strip surface quality evaluation module F5 performs call evaluation.
  • Classifiers include decision trees, Adaboost and artificial neural networks.
  • An embodiment of the present invention also provides a surface quality management method for cold-rolled strip steel, including the following steps:
  • S400 Provide a corresponding improvement strategy according to the category of the surface defect of the cold rolled strip, and store the category, the improvement strategy and the processing result of the surface defect of the cold rolled strip.
  • the detected value is matched with the sample and analyzed, and the case description of the detected strip is given, and the processing method and the processing result are given.
  • the cold rolled strip surface quality management system and method provided by the present invention Aiming at the deficiencies of the existing cold-rolled strip surface quality online inspection system, the establishment of defect samples, the establishment of cold-rolled strip surface quality defect treatment countermeasures table, the cold-rolled strip surface quality inspection and quality improvement are related, the cold-rolled strip
  • the surface defects are matched with the defect samples to determine the surface quality grade of the cold-rolled strip surface defects, so as to guide the improvement of the production process and improve the surface quality of the cold-rolled strip; on the other hand, the angular distance method is used to remove abnormal samples to improve the defects The labeling accuracy of the sample.
  • the sample marks of the cold-rolled strip surface quality management system cover the typical defects of the cold-rolled strip surface quality (holes, edge cracks, folds, scars, indentations, roll marks, indentations, scratches, warping, oxidation) Iron sheet, etc.), the category can be expanded;
  • the cold-rolled strip surface quality management system processes the samples in the sample table of the cold-rolled strip surface quality knowledge base through the angular distance method, and removes abnormal samples based on the sample characteristic values to ensure that the samples are clean;
  • the cold-rolled strip surface quality management system builds a strip surface quality evaluation model based on the AHP method based on all defect types and defect level data of each coil in the defect table to quantitatively evaluate the strip surface quality;
  • the cold-rolled strip surface quality management system provides surface quality treatment improvement methods and suggestions for the defects that affect the cold-rolled strip surface quality, and guides the on-site quality improvement.

Abstract

一种冷轧带钢表面质量管理系统及方法，涉及冷轧带钢技术领域，针对现有冷轧带钢表面质量在线检测系统的不足，建立缺陷样本，建立冷轧带钢表面质量缺陷处理对策表，将冷轧带钢表面质量检测与质量改进关联起来，将冷轧带钢表面缺陷与缺陷样本进行匹对判断冷轧带钢表面缺陷的带钢表面质量等级，从而指导生产工艺改进，提升冷轧带钢表面质量；另一方面，利用角度距离法剔除异常样本，提高缺陷样本的标记准确率。

Description

一种冷轧带钢表面质量管理系统及方法 技术领域
本发明属于冷轧带钢技术领域，具体涉及一种冷轧带钢表面质量管理系统及方法。
背景技术
随着汽车、家电等产品对带钢表面质量要求越来越高，企业对冷轧带钢表面质量更加重视，国内各大型钢企在冷轧生产线上均配置有带钢表面质量在线检测系统，在线检测识别冷轧带钢表面质量。
在运行过程中，带钢表面质量在线检测系统存在两个方面局限：一、缺陷样本数量、均匀性和标记准确率不够，导致带钢表面质量在线检测系统识别率偏低；二、识别出带钢表面质量缺陷后，缺少带钢表面质量评价及质量改进建议，使得带钢表面质量在线检测系统实际转化成效益不明显。
发明内容
本发明的目的是克服现有技术中带钢表面质量在线检测系统实际转化成效益不明显问题。
为此，本发明提供了一种冷轧带钢表面质量管理系统，包括冷轧带钢表面质量缺陷识别模块、冷轧带钢表面质量缺陷存储模块、冷轧带钢表面质量缺陷样本标记模块、冷轧带钢表面质量评价模块及冷轧带钢表面质量改进模块；
所述冷轧带钢表面质量缺陷识别模块用于检测冷轧带钢的表面缺陷情况；
所述冷轧带钢表面质量缺陷存储模块用于将检测到有缺陷的冷轧带钢 表面缺陷进行保存；
所述冷轧带钢表面质量缺陷样本标记模块用于标记并存储缺陷样本；
所述冷轧带钢表面质量评价模块用于将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的类别及带钢表面质量等级；
所述冷轧带钢表面质量改进模块用于根据所述冷轧带钢表面缺陷的类别及带钢表面质量等级提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
优选地，所述冷轧带钢表面缺陷和所述缺陷样本均包括缺陷类别、钢卷编号、缺陷特征值和缺陷图片路径，所述缺陷特征值包含缺陷面积、缺陷长度、缺陷周长、缺陷区域细长比和缺陷形状因子。
优选地，所述冷轧带钢表面质量评价模块依据缺陷样本中每卷带钢缺陷类别和缺陷特征值数据，并基于层次分析法构建带钢表面质量评价模型，量化评定样品带钢表面质量等级。
优选地，所述带钢表面质量评价模型为：样品带钢表面质量等级＝缺陷特征值×级别权，所述级别权为所述缺陷特征值中对应的权重的大小。
优选地，还包括异常样本剔除模块，所述异常样本剔除模块用于根据所述缺陷样本的缺陷特征值对所述缺陷样本中的异常体进行剔除。
优选地，所述异常样本剔除模块根据缺陷的角度距离的异常样本剔除法进行剔除，具体包括：
S1、选取缺陷样本中第i点A，i＝1；
S2、任选缺陷样本中非第i点的B和C两点，得到以A为顶点的向量积<AB，AC>、向量模|AB|及向量模|AC|，并得到向量夹角余弦cos＝<AB，AC>/，且在所述缺陷样本中遍历所有B、C点组合；
S3、计算所述缺陷样本中第i点所有组合向量夹角余弦值方差
Figure PCTCN2019106449-appb-000001
Figure PCTCN2019106449-appb-000002
其中σ 2为方差，X为向量夹角余弦，μ为向量夹角余弦均值，N为缺陷样本总数；
S4、比较i与N大小，如果i≤N，则i＝i+1，返回S1，如果i＝N，则进入下一步；
S5、对所有点的向量夹角余弦值方差进行排序；
S6、取其中向量夹角余弦值方差最小的k个点为异常点，k为正整数；
S7、结束。
优选地，所述系统还包括分类器训练模块，所述分类器训练模块包含分类训练器，所述分类器训练模块根据所述缺陷样本的特征配置分类训练器，所述分类训练器对所述缺陷样本进行读取。
优选地，所述分类训练器包含决策树算法、Adaboost算法或人工神经网络算法中的一种或多种。
本发明还提供了一种冷轧带钢表面质量管理方法，包括以下步骤：
S100：建立缺陷样本；
S200：检测冷轧带钢的表面缺陷，将检测到有缺陷的冷轧带钢表面缺陷进行保存；
S300：将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的带钢表面质量等级；
S400：根据所述冷轧带钢表面缺陷的类别提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
本发明的有益效果：本发明提供的这种冷轧带钢表面质量管理系统及方法。针对现有冷轧带钢表面质量在线检测系统的不足，建立缺陷样本， 建立冷轧带钢表面质量缺陷处理对策表，将冷轧带钢表面质量检测与质量改进关联起来，将冷轧带钢表面缺陷与缺陷样本进行匹对判断冷轧带钢表面缺陷的带钢表面质量等级，从而指导生产工艺改进，提升冷轧带钢表面质量；另一方面，利用角度距离法剔除异常样本，提高缺陷样本的标记准确率。
本发明提供的冷轧带钢表面质量管理系统及方法具有以下优点：
(1)冷轧带钢表面质量管理系统的样本标记覆盖冷轧带钢表面质量典型缺陷(孔洞，边裂，折叠，结疤，压痕，辊印，压入，划伤，翘皮，氧化铁皮等)，类别可拓展；
(2)冷轧带钢表面质量管理系统的标记每类缺陷样本数量为300～500张，样本均匀、数量充足且方便更新；
(3)冷轧带钢表面质量管理系统通过角度距离法对冷轧带钢表面质量知识库样本表中样本进行处理，基于样本特征值剔除异常样本，确保样本干净；
(4)冷轧带钢表面质量管理系统依据缺陷表中每卷带钢所有缺陷种类和缺陷级别数据，基于AHP方法构建带钢表面质量评价模型，量化评定带钢表面质量；
(5)冷轧带钢表面质量管理系统对影响冷轧带钢表面质量的缺陷，提供表面质量处理改进方法与建议，指导现场进行质量改进。
以下将结合附图对本发明做进一步详细说明。
附图说明
图1是本发明冷轧带钢表面质量管理系统模块示意图；
图2是本发明冷轧带钢表面质量管理系统逻辑原理示意图；
图3是本发明冷轧带钢表面质量管理系统及方法的剔除算法流程图；
图4是本发明冷轧带钢表面质量管理系统及方法的基于AHP的带钢质量评价模型图；
图5是本发明冷轧带钢表面质量管理系统及方法的用户界面示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
在本发明的描述中，需要理解的是，术语“中心”、“上”、“下”、“前”、“后”、“左”、“右”、“竖直”、“水平”、“顶”、“底”、“内”、“外”等指示的方位或位置关系为基于附图所示的方位或位置关系，仅是为了便于描述本发明和简化描述，而不是指示或暗示所指的装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。
术语“第一”、“第二”仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征；在本发明的描述中，除非另有说明，“多个”的含义是两个或两个以上。
本发明实施例提供了一种冷轧带钢表面质量管理系统，包括冷轧带钢表面质量缺陷识别模块F0、冷轧带钢表面质量缺陷存储模块F1、冷轧带钢表面质量缺陷样本标记模块F2、冷轧带钢表面质量评价模块F5及冷轧带钢表面质量改进模块F6；
所述冷轧带钢表面质量缺陷识别模块用于检测冷轧带钢的表面缺陷情况；
所述冷轧带钢表面质量缺陷存储模块F1用于将检测到有缺陷的冷轧带钢表面缺陷进行保存；
所述冷轧带钢表面质量缺陷样本标记模块F2用于标记并存储缺陷样本；
所述冷轧带钢表面质量评价模块F5用于将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的类别及带钢表面质量等级；
所述冷轧带钢表面质量改进模块F6用于根据所述冷轧带钢表面缺陷的类别及带钢表面质量等级提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
由此可知，如图1所示，冷轧带钢表面质量管理系统包括冷轧带钢表面质量缺陷识别模块F0、冷轧带钢表面质量缺陷存储模块F1、冷轧带钢表面质量缺陷样本标记模块F2、冷轧带钢表面质量评价模块F5及冷轧带钢表面质量改进模块F6。先通过冷轧带钢表面质量缺陷样本标记模块F2检索和浏览冷轧带钢表面质量缺陷图片，对缺陷进行标记分类，然后将标记分类好的缺陷样本进行存储，将工程实际过程中可能会出现的冷轧带钢表面质量缺陷的情况进行分类并保存，主要是以图片的形式保存。然后冷轧带钢表面质量缺陷识别模块F0对冷轧带钢表面进行缺陷检测，并将检测到有缺陷的冷轧带钢表面缺陷进行保存至冷轧带钢表面质量缺陷存储模块F1，然后冷轧带钢表面质量评价模块F5将上述检测到的冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的带钢表面质量等级，最后通过冷轧带钢表面质量改进模块F6根据所述冷轧带钢表面缺陷的类别及等 级提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
该系统的总体框架如图2所示，冷轧带钢表面质量管理系统包括冷轧带钢表面质量知识库和表面质量在线检测系统，前者是数据流，后者是软硬件承载体，主要包括用户界面层1、逻辑层2、数据访问层3以及数据与文件夹4，用户界面层1接收用户输入，进行人机交互；逻辑层2进行逻辑和数学运算，处理数据；数据访问层3对数据进行读/操作；数据库与图片文件夹4组织和存储数据，其中数据库组织和存储数据，图片文件夹存储缺陷图片。表面质量在线检测系统检测与识别到的缺陷存储到缺陷表和缺陷图片文件夹，并从样本表读取样本进行分类器训练与验证。其中，用户界面层1可以是浏览器，实现网络数据的共享，获取样本，通过逻辑层2进行样本标记和异常样本剔除，然后进入数据访问层3进行读表和写表来对缺陷样本进行读写操作，然后在数据库与图片文件夹中形成缺陷表和样本表，并对应形成质量评价表和缺陷图片，表面质量在线检测系统对应冷轧带钢表面质量缺陷识别模块F0，包括表面质量在线检测元件、硬件、识别软件及在线检测分类器软件。
优选的方案，所述冷轧带钢表面缺陷和所述缺陷样本均包括缺陷类别、钢卷编号、缺陷特征值和缺陷图片路径，所述缺陷特征值包含缺陷面积、缺陷长度、缺陷周长、缺陷区域细长比和缺陷形状因子。由此可知，缺陷样本及有缺陷的冷轧带钢表面缺陷，缺陷属性都包括缺陷类别、钢卷编号、缺陷特征值和缺陷图片路径，所述缺陷特征值包含缺陷面积、缺陷长度、缺陷周长、缺陷区域细长比和缺陷形状因子。缺陷样本种类包括孔洞，边裂，折叠，结疤，压痕，辊印，压入，划伤，翘皮，氧化铁皮等，每类缺 陷样本数量为300～500，为带钢表面质量在线检测系统分类器(即冷轧带钢表面质量评价模块F5学习与验证提供数据支持。
优选的方案，所述冷轧带钢表面质量评价模块F5依据缺陷样本中每卷带钢缺陷类别和缺陷特征值数据，并基于层次分析法(AHP)构建带钢表面质量评价模型，量化评定样品带钢表面质量等级。
优选的方案，依据缺陷表中每卷带钢所有缺陷种类和缺陷级别数据，基于AHP方法构建带钢表面质量评价模型，量化评定带钢表面质量，如图4所示。目标层O用一个最终质量评分作为衡量冷轧带钢表面质量的指标，∑得分＝∑缺陷权×级别权，得分即为带钢表面质量等级，可以量化评定冷轧带钢表面质量水平，准则层D为每种缺陷内的不同级别，每个级别都有权值，缺陷级别权重主要考虑缺陷面积、缺陷长度、缺陷周长、缺陷区域细长比和缺陷形状因子，准则层C为板带钢的各种缺陷类型，每种缺陷都有缺陷权值，主要考虑缺陷对表面质量影响程度，通过构造缺陷判断矩阵求解缺陷权值，方案层P是每一卷带缺陷的钢板，包含该卷带钢所有表面质量缺陷原始数据。
优选的方案，品带钢表面质量等级＝缺陷特征值×级别权，所述级别权为所述缺陷特征值中对应的权重的大小。即对应上述∑得分＝∑缺陷权×级别权。
优选的方案，还包括异常样本剔除模块F3，所述异常样本剔除模块F3用于根据所述缺陷样本的缺陷特征值对所述缺陷样本中的异常体进行剔除。由此可知，通过分析样本，可以查找出异常样本，然后剔除，避免影响样本的精度。
优选的方案，所述异常样本剔除模块F3根据缺陷的角度距离的异常样 本剔除法进行剔除，具体包括：
S1、选取缺陷样本中第i点A，i＝1；
S2、任选缺陷样本中非第i点的B和C两点，得到以A为顶点的向量积<AB，AC>、向量模|AB|及向量模|AC|，并得到向量夹角余弦cos(AB，AC)＝<AB，AC>/(|AB|·|AC|)，且在所述缺陷样本中遍历所有B、C点组合；
S3、计算所述缺陷样本中第i点所有组合向量夹角余弦值方差
Figure PCTCN2019106449-appb-000003
Figure PCTCN2019106449-appb-000004
其中σ 2为方差，X为向量夹角余弦，μ为向量夹角余弦均值，N为缺陷样本总数；
S4、比较i与N大小，如果i≤N，则i＝i+1，返回S1，如果i＝N，则进入下一步；
S5、对所有点的向量夹角余弦值方差进行排序；
S6、取其中向量夹角余弦值方差最小的k个点为异常点，k为正整数；
S7、结束。
由此可知，如图3所示，绘制逻辑层算法实现流程图，实现逻辑层算法，并基于角度距离来对异常本进行剔除。
优选的方案，所述系统还包括分类器训练模块F4，所述分类器训练模块F4包含分类训练器，所述分类器训练模块根据所述缺陷样本的特征配置分类训练器，所述分类训练器对所述缺陷样本进行读取。所述类器训练模块F4包含决策树算法、Adaboost算法或人工神经网络算法中的一种或多种。分类器训练模块F4依据工艺经验选择保存好的有缺陷的冷轧带钢表面缺陷的特征，配置分类器参数，再读取缺陷样本表，然后对构建的分类器进行训练与验证，供后面的冷轧带钢表面质量评价模块F5进行调用评估。分类器包括决策树、Adaboost和人工神经网络。
本发明实施例还提供了一种冷轧带钢表面质量管理方法，包括以下步骤：
S100：建立缺陷样本；
S200：检测冷轧带钢的表面缺陷，将检测到有缺陷的冷轧带钢表面缺陷进行保存；
S300：将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的带钢表面质量等级；
S400：根据所述冷轧带钢表面缺陷的类别提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
如图5所示，将检测值与样本进行匹对分析，并对检测的带钢进行案例描述，给出处理方法，以及处理结果。
本发明的有益效果：本发明提供的这种冷轧带钢表面质量管理系统及方法。针对现有冷轧带钢表面质量在线检测系统的不足，建立缺陷样本，建立冷轧带钢表面质量缺陷处理对策表，将冷轧带钢表面质量检测与质量改进关联起来，将冷轧带钢表面缺陷与缺陷样本进行匹对判断冷轧带钢表面缺陷的带钢表面质量等级，从而指导生产工艺改进，提升冷轧带钢表面质量；另一方面，利用角度距离法剔除异常样本，提高缺陷样本的标记准确率。
本发明提供的冷轧带钢表面质量管理系统及方法具有以下优点：
(1)冷轧带钢表面质量管理系统的样本标记覆盖冷轧带钢表面质量典型缺陷(孔洞，边裂，折叠，结疤，压痕，辊印，压入，划伤，翘皮，氧化铁皮等)，类别可拓展；
(2)冷轧带钢表面质量管理系统的标记每类缺陷样本数量为300～500 张，样本均匀、数量充足且方便更新；
(3)冷轧带钢表面质量管理系统通过角度距离法对冷轧带钢表面质量知识库样本表中样本进行处理，基于样本特征值剔除异常样本，确保样本干净；
(4)冷轧带钢表面质量管理系统依据缺陷表中每卷带钢所有缺陷种类和缺陷级别数据，基于AHP方法构建带钢表面质量评价模型，量化评定带钢表面质量；
(5)冷轧带钢表面质量管理系统对影响冷轧带钢表面质量的缺陷，提供表面质量处理改进方法与建议，指导现场进行质量改进。
以上例举仅仅是对本发明的举例说明，并不构成对本发明的保护范围的限制，凡是与本发明相同或相似的设计均属于本发明的保护范围之内。

Claims (9)

  1. 一种冷轧带钢表面质量管理系统，其特征在于：包括冷轧带钢表面质量缺陷识别模块(F0)、冷轧带钢表面质量缺陷存储模块(F1)、冷轧带钢表面质量缺陷样本标记模块(F2)、冷轧带钢表面质量评价模块(F5)及冷轧带钢表面质量改进模块(F6)；
    所述冷轧带钢表面质量缺陷识别模块用于检测冷轧带钢的表面缺陷情况；
    所述冷轧带钢表面质量缺陷存储模块(F1)用于将检测到有缺陷的冷轧带钢表面缺陷进行保存；
    所述冷轧带钢表面质量缺陷样本标记模块(F2)用于标记并存储缺陷样本；
    所述冷轧带钢表面质量评价模块(F5)用于将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的类别及带钢表面质量等级；
    所述冷轧带钢表面质量改进模块(F6)用于根据所述冷轧带钢表面缺陷的类别及带钢表面质量等级提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
  2. 根据权利要求1所述的冷轧带钢表面质量管理系统，其特征在于：所述冷轧带钢表面缺陷和所述缺陷样本均包括缺陷类别、钢卷编号、缺陷特征值和缺陷图片路径，所述缺陷特征值包含缺陷面积、缺陷长度、缺陷周长、缺陷区域细长比和缺陷形状因子。
  3. 根据权利要求2所述的冷轧带钢表面质量管理系统，其特征在于：所述冷轧带钢表面质量评价模块(F5)依据缺陷样本中每卷带钢缺陷类别和缺陷特征值数据，并基于层次分析法(AHP)构建带钢表面质量评价模型， 量化评定样品带钢表面质量等级。
  4. 根据权利要求3所述的冷轧带钢表面质量管理系统，其特征在于，所述带钢表面质量评价模型为：样品带钢表面质量等级＝缺陷特征值×级别权，所述级别权为所述缺陷特征值中对应的权重的大小。
  5. 根据权利要求2所述的冷轧带钢表面质量管理系统，其特征在于：还包括异常样本剔除模块(F3)，所述异常样本剔除模块(F3)用于根据所述缺陷样本的缺陷特征值对所述缺陷样本中的异常体进行剔除。
  6. 根据权利要求5所述的冷轧带钢表面质量管理系统，其特征在于，所述异常样本剔除模块(F3)根据缺陷的角度距离的异常样本剔除法进行剔除，具体包括：
    S1、选取缺陷样本中第i点A，i＝1；
    S2、任选缺陷样本中非第i点的B和C两点，得到以A为顶点的向量积<AB，AC>、向量模|AB|及向量模|AC|，并得到向量夹角余弦cos(AB，AC)＝<AB，AC>/(|AB|·|AC|)，且在所述缺陷样本中遍历所有B、C点组合；
    S3、计算所述缺陷样本中第i点所有组合向量夹角余弦值方差
    Figure PCTCN2019106449-appb-100001
    Figure PCTCN2019106449-appb-100002
    其中σ 2为方差，X为向量夹角余弦，μ为向量夹角余弦均值，N为缺陷样本总数；
    S4、比较i与N大小，如果i≤N，则i＝i+1，返回S1，如果i＝N，则进入下一步；
    S5、对所有点的向量夹角余弦值方差进行排序；
    S6、取其中向量夹角余弦值方差最小的k个点为异常点，k为正整数；
    S7、结束。
  7. 根据权利要求1所述的冷轧带钢表面质量管理系统，其特征在于： 所述系统还包括分类器训练模块(F4)，所述分类器训练模块(F4)包含分类训练器，所述分类器训练模块(F4)根据所述缺陷样本的特征配置分类训练器，所述分类训练器对所述缺陷样本进行读取。
  8. 根据权利要求7所述的冷轧带钢表面质量管理系统，其特征在于：所述分类训练器包含决策树算法、Adaboost算法或人工神经网络算法中的一种或多种。
  9. 一种冷轧带钢表面质量管理方法，其特征在于，包括以下步骤：
    S100：建立缺陷样本；
    S200：检测冷轧带钢的表面缺陷，将检测到有缺陷的冷轧带钢表面缺陷进行保存；
    S300：将所述冷轧带钢表面缺陷与所述缺陷样本进行匹对判断所述冷轧带钢表面缺陷的带钢表面质量等级；
    S400：根据所述冷轧带钢表面缺陷的类别提供相应的改进策略，并将冷轧带钢表面缺陷的类别、改进策略及处理结果进行存储。
PCT/CN2019/106449 2018-12-29 2019-09-18 一种冷轧带钢表面质量管理系统及方法 WO2020134210A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811632987.3A CN109692877B (zh) 2018-12-29 2018-12-29 一种冷轧带钢表面质量管理系统及方法
CN201811632987.3 2018-12-29

Publications (1)

Publication Number Publication Date
WO2020134210A1 true WO2020134210A1 (zh) 2020-07-02

Family

ID=66232276

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/106449 WO2020134210A1 (zh) 2018-12-29 2019-09-18 一种冷轧带钢表面质量管理系统及方法

Country Status (2)

Country Link
CN (1) CN109692877B (zh)
WO (1) WO2020134210A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI766737B (zh) * 2021-06-24 2022-06-01 中國鋼鐵股份有限公司 絞乾輥品質檢測系統與其方法

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109692877B (zh) * 2018-12-29 2020-11-10 中冶南方工程技术有限公司 一种冷轧带钢表面质量管理系统及方法
CN110175274A (zh) * 2019-05-22 2019-08-27 山西太钢不锈钢股份有限公司 基于表检仪的缺陷分级方法及系统
CN110320337A (zh) * 2019-06-05 2019-10-11 广州Jfe钢板有限公司 一种带钢表面质量自动判级系统和方法
CN110400099A (zh) * 2019-08-09 2019-11-01 马鞍山钢铁股份有限公司 一种带钢产品表面质量分级方法
CN110717455B (zh) * 2019-10-10 2021-05-18 北京同创信通科技有限公司 一种收储中的废钢等级分类检测方法
CN111223078B (zh) * 2019-12-31 2023-09-26 富联裕展科技(河南)有限公司 瑕疵等级判定的方法及存储介质
CN111299318B (zh) * 2020-03-02 2022-04-12 马鞍山钢铁股份有限公司 一种热轧板带产品表面质量的自动判定方法
CN112547807B (zh) * 2020-10-30 2022-01-04 北京科技大学 一种基于决策树算法的热轧带钢质量精准判定方法
CN113578972B (zh) * 2021-04-08 2022-09-27 华院计算技术(上海)股份有限公司 一种热轧产品质量追溯方法及装置
CN113578976B (zh) * 2021-07-28 2023-09-01 北京首钢股份有限公司 一种控制边部翘皮的热轧电工钢的生产方法和生产系统
CN114722973B (zh) * 2022-06-07 2022-08-26 江苏华程工业制管股份有限公司 一种钢管热处理的缺陷检测方法及系统
CN115983687B (zh) * 2022-12-22 2023-09-29 北京弥天科技有限公司 一种冷轧带钢质量智能检测管理系统及方法

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006242906A (ja) * 2005-03-07 2006-09-14 Jfe Steel Kk 表面欠陥検査方法及びその装置
CN102253049A (zh) * 2011-06-30 2011-11-23 东北大学 带钢生产过程表面质量在线精准检测方法
CN103778445A (zh) * 2014-01-21 2014-05-07 武汉科技大学 一种冷轧带钢表面缺陷原因分析方法及系统
CN105531581A (zh) * 2013-09-10 2016-04-27 蒂森克虏伯钢铁欧洲股份公司 对用于识别表面缺陷的检查系统进行检验的方法和设备
CN106845825A (zh) * 2017-01-18 2017-06-13 西安交通大学 一种基于改进pca的带钢冷轧质量问题溯源及控制方法
CN106875104A (zh) * 2017-01-21 2017-06-20 西安交通大学 一种冷轧带钢质量综合评价方法
CN108021938A (zh) * 2017-11-29 2018-05-11 中冶南方工程技术有限公司 一种冷轧带钢表面缺陷在线检测方法以及检测系统
CN109692877A (zh) * 2018-12-29 2019-04-30 中冶南方工程技术有限公司 一种冷轧带钢表面质量管理系统及方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008055443A (ja) * 2006-08-29 2008-03-13 Kobe Steel Ltd 金属材料の材質分析方法および材質安定化方法
JP5942395B2 (ja) * 2011-11-29 2016-06-29 東芝三菱電機産業システム株式会社 製品欠陥情報追跡装置
CN102554167A (zh) * 2012-02-14 2012-07-11 首钢总公司 H型钢缺陷的控制方法
CN202606510U (zh) * 2012-04-19 2012-12-19 中冶南方工程技术有限公司 一种冷轧带钢板形板厚综合控制系统
CN103399016A (zh) * 2013-07-26 2013-11-20 齐鲁工业大学 冷轧铝板表面缺陷在线检测系统及其检测方法
CN104985002B (zh) * 2015-05-27 2017-02-22 北京首钢股份有限公司 一种热轧带钢边部缺陷报警方法及装置
CN108445008A (zh) * 2018-02-27 2018-08-24 首钢京唐钢铁联合有限责任公司 一种带钢表面缺陷的检测方法
CN109030503A (zh) * 2018-07-11 2018-12-18 无锡赛默斐视科技有限公司 带钢在线表面缺陷检测系统及其检测方法

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006242906A (ja) * 2005-03-07 2006-09-14 Jfe Steel Kk 表面欠陥検査方法及びその装置
CN102253049A (zh) * 2011-06-30 2011-11-23 东北大学 带钢生产过程表面质量在线精准检测方法
CN105531581A (zh) * 2013-09-10 2016-04-27 蒂森克虏伯钢铁欧洲股份公司 对用于识别表面缺陷的检查系统进行检验的方法和设备
CN103778445A (zh) * 2014-01-21 2014-05-07 武汉科技大学 一种冷轧带钢表面缺陷原因分析方法及系统
CN106845825A (zh) * 2017-01-18 2017-06-13 西安交通大学 一种基于改进pca的带钢冷轧质量问题溯源及控制方法
CN106875104A (zh) * 2017-01-21 2017-06-20 西安交通大学 一种冷轧带钢质量综合评价方法
CN108021938A (zh) * 2017-11-29 2018-05-11 中冶南方工程技术有限公司 一种冷轧带钢表面缺陷在线检测方法以及检测系统
CN109692877A (zh) * 2018-12-29 2019-04-30 中冶南方工程技术有限公司 一种冷轧带钢表面质量管理系统及方法

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI766737B (zh) * 2021-06-24 2022-06-01 中國鋼鐵股份有限公司 絞乾輥品質檢測系統與其方法

Also Published As

Publication number Publication date
CN109692877B (zh) 2020-11-10
CN109692877A (zh) 2019-04-30

Similar Documents

Publication Publication Date Title
WO2020134210A1 (zh) 一种冷轧带钢表面质量管理系统及方法
CN105389593B (zh) 基于surf特征的图像物体识别方法
CN104794491B (zh) 基于预分类的模糊聚类钢板表面缺陷检测方法
CN106952258A (zh) 一种基于梯度方向直方图的瓶口缺陷检测方法
CN100464332C (zh) 一种图片查询方法及系统
CN107341688A (zh) 一种客户体验的采集方法及系统
CN106228137A (zh) 一种基于关键点定位的atm机异常人脸检测方法
US20180322416A1 (en) Feature extraction and classification method based on support vector data description and system thereof
CN108268838A (zh) 人脸表情识别方法及人脸表情识别系统
CN108846831B (zh) 基于统计特征和图像特征相结合的带钢表面缺陷分类方法
CN111476307B (zh) 一种基于深度领域适应的锂电池表面缺陷检测方法
CN103034838A (zh) 一种基于图像特征的特种车辆仪表类型识别与标定方法
CN115063423B (zh) 基于计算机视觉的机械铸件冷热裂纹自适应辨别方法
Wu et al. Detection of crack eggs by image processing and soft-margin support vector machine
CN111932510A (zh) 一种图像清晰度的确定方法及装置
CN107292879A (zh) 一种基于图像分析的金属板材表面异常检测方法
CN115294109A (zh) 基于人工智能的实木板材生产缺陷识别系统、电子设备
WO2021109011A1 (zh) 一种基于超声图像的电容内部缺陷智能检测方法
CN107730486A (zh) 一种酸连轧生产线轧机保护方法及装置
CN103854011A (zh) 一种虹膜图像的质量评价方法
Barman et al. Development of a real-time home textile fabric defect inspection machine system for the textile industry
CN112597865A (zh) 热轧带钢边部缺陷智能识别方法
CN115359412B (zh) 一种盐酸中和实验评分方法、装置、设备及可读存储介质
CN113926865B (zh) 铸坯夹渣预报方法、机清控制方法、计算设备及存储介质
Xin et al. Detection of apple surface defect based on YOLOv3

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19902656

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19902656

Country of ref document: EP

Kind code of ref document: A1