* Russian

* Ukrainian

* English

* Main

* Abstract

* Library

* Links

* Report about the search

* About turism

* Masters DonNTU

Surova Anna Development of specialized computer system for sheet renting defects research

The relevance of the work

Current production uses to detect surface defects during harvesting sheets visual methods which have several disadvantages:

  • labour-intensiveness;

  • poor accuracy due to the subjectivity of opinions of conducting diagnostics specialist;

  • long duration of the implementation of evaluation for the rolled products quality ;

  • high cost.

In comparison with the methods of visual assessment, automation of surface defect recognition process has a better performance and much less labour-intensiveness in determining of the number of physical characteristics surface layer of metal. Creating of operating defects diagnostics system offers great scope for improvement production processes, including detection of violations of production during the process and not at the stage of quality control, post factum, that provides an opportunity to correct the problem and prevent further production marriage.

Aims and objectives

The purpose of the study is searching for methods improving speed and accuracy recognition of defects, as well as the creation of a specialized operational computer decision support systems in the evaluation of defect products.

Research object - the surface of flat rolled products.

Diagnostic object - a set of surface defects in rolled products, arising because of irregularities in manufacturing technology.

Task of research - the creation of operational computer systems, automating process of identifying defects in investigating and classifying the defects found.

A review of existing development and scientific innovation

At this time, the development of automation systems for quality control of metallurgical products is conducted throughout the world.

Leading specialists of Magnitogorsk Metallurgical Combine, students, postgraduates and staff of Magnitogorsk State Technical University involved assessment of quality of templet macrostructure[1].

Experts of the company «Uralmash - Metallurgical Equipment» explore relationship quality blanks with the parameters characterizing the process of casting [2]. Sorokina L.I., employee OOO «VideoTest»(St. Petersburg) is working on standardization of automated techniques for monitoring the microstructure of the metal [3].

DonNTU graduate Gatilova Y.S. devoted her master's work to the problem of diagnosing such type of defect metal as captivity [4].

Science novelty of my work is studing of of «fast» image processing techniques to create the operating system od defects diagnosis, which is extremely important in the context of continuous production.

Diagnosis of superficial defects sheets

Rejection of defects at the stage of origin of steel sheets due to detection of surface defects is 1.5% [4].

Defects of the sheets are divided into 3 types depending on the reasons for their occurrence:

  • Defects of the surface due to the quality of cast billets (roll, crust, bubble, separation, unroll crack, accordion)

  • Surface defects formed during deformation (deformation flaw, flaw at the edges, extend the edge, rolled blister, crack, starling, scores, cuts, wrinkles, risk, through tears, nadryvy, prodir, nakoly-punctures, prints, roll fingerprints, footprints mesh, flake, overheated surfaces scale, sinks of scale, metal particles, sink-push retarded scum, traces of abrasive stripping, gray spotted carbonization, spots of pollution, clumping spots , burr, hack, crack face, strip-line sliding, strip autofrettage, excesses).

  • Surface defects generated during finishing operations (pickling cracks, flying slime, balances scale, rust, shades of etching, dents, scratches, matte surface) [5]

Depending on the defects impact all products are divided into four groups:

  • suitable, fully meet all the requirements of technical documentation and standards;

  • conditionally suitable, with minor deviations from the requirements (minor defects) does not have a significant impact on performance indicators castings or products in general, products are allowed for further processing and used for its intended purpose with the permission of the major specialists of industrial enterprises after careful evaluation of defects;

  • amendable waste - products that have one or more removable defects after the corrections which they may be admitted for further processing and intended using;

  • incorrigible or final waste - the product with such defects, that correction is technically impossible or economically impractical or quality corrections are impossible to verify.

Stages in diagnostics of surface defects:

  • Identification of surface defects on digitized images of the rolled products after processing the image:

A. Filtering,

B. Segmentation

  • Determination of physical characteristics of the surface defect (size, shape, color, location).

  • Classification of defects.

The processing of digital images

In the experiments conducted in the Matlab, the most appropriate method was the median filtering:

The original image

Pic.1 The original image

The result of median filtering

Pic.2 The result of median filtering

The use of other filters

Animation 1. The use of other filters:

The size of the animation: 113 Kb

Number of frames: 5

The frame display : 2 sec (1st frame), 5 sec (2nd-5th frame)

Number of recurrence cycles : 8

The principle of the median filter

For each pixel in a certain environment (the window), the median value is searched and assigned to this pixel. Determination of the median values: if the array of pixels is sorted by their values, the median is middle element of the array. The size of the window respectively must be odd to existing the middle element.

You can also determine median by the formula:

Median filtering formula

where W is the set of pixels, in which we looks for the median, and fi is a brightness of these pixels.

For color images we use the vector median filter (VMF):

Vector median filter

where Fi are the values of pixels in the three-dimensional color space, and d is an arbitrary metric (eg Euclidean). [6]

Opportunities of accelerate for filtering algorithms:

• to make a few quick steps to get the mid-sort

• specific implementation for each radius of the window [7]

Segmentation

Segmentation of the image is called a splitting image on unlike for some signs of the area. It is anticipated that the area corresponds to the real objects or their parts, and the border areas appropriate boundaries of objects. Segmentation plays an important role in problems of image processing and computer vision. Segmentation methods can be divided into two classes: automatic - do not require interaction and interactive with the user - using the user input directly into the process.

For rough assessment of the quality of a specific task we usually fix several properties, which should have a good segmentation. The quality of the method is evaluated depending on whether the received segmentation has these properties. The most frequently used the following properties:

  • regions homogeneity (uniformity of color or texture)

  • neighboring regions otherness

  • region smoothness

  • just a few of small «holes» in regionalization, etc. [8]

After filtering, we have the approximate boundary defects rolled (80-90% of true size of the defect). Then we select the pixel in the detected area and use the of rising of areas around it. Taking this result as a basis for further processing,we use an algorithm for specifying, for example, Kanni method of edges detecting.

Time advantage is also due to the fact that if after the rapid filtration we did not found a defect of hire on the picture , then further long procedure of segmentation is not performed. Besides, implementation of the streaming mode of detection program is possible. The main stream monitors each frame, performs filtering of the image and transmits the image of a founded defect to a new stream, which serves detailed studing of the defect.

Expected results

Writing of this Master's abstract work is not complete yet. Final results are expected in December 2009. Full text and materials can be obtained by e-mail after that date.

Conclusion

Using of defects detection systems in production and operation of products will give a big economic impact by reducing the time required for processing workpieces with defects, saving the metal.

List of sources

  1. Капцан Ф.В., Суспицын В.Г., Логунова О.С., Павлов В.В., Нуров X.X. Организация автоматизированного рабочего места в системе оценки качества макроструктуры заготовок в ЭСПЦ ММК// Сталь. 2006 № 11. С. 80 - 82
  2. Паршин В. М., Чертов Л. Д. Интеллектуальные системы управления качеством непрерывнолитой заготовки// Сталь. 2005 № 2. С. 37 - 43.
  3. http://lityo.com.ua/li/s_174.html Сорокина Л.И. (ООО «ВидеоТесТ», г.Санкт-Петербург) Стандартизация автоматизированных методик контроля микроструктуры
  4. Гатилова Ю.С. Специализированная компьютерная система диагностики поверхностных дефектов листового проката// Автоматизация технологических объектов и процессов. Поиск молодых. Сборник научных работ VII Международной научно-технической конференции аспирантов и студентов в г. Донецке 26-28 апреля 2007 г. - Донецк, ДонНТУ, 2007.- С.251-253. 3 с.
  5. Дефекты проката http://eugene-motors.kiev.ua/wiev.php?partname=%EF%EE%EB%E5%E7%ED%EE%E5&id=34
  6. Проблема подавления шума на изображениях и видео и различные подходы к ее решению http://cgm.computergraphics.ru/content/view/74
  7. Лекции по обработке изображений http://graphics.cs.msu.ru/courses/cg02b/lectures/lection5/sld019.htm
  8. Методы сегментации изображений: автоматическая сегментация http://cgm.computergraphics.ru/content/view/147
  9. Возможности цифровой обработки изображений в Matlab http://matlab.exponenta.ru/imageprocess/book2/80.php
  10. Компьютерное зрение http://ru.wikipedia.org/wiki/Компьютерное_зрение

Up