Title of Invention	A METHOD FOR THE DETECTION AND CLASSIFICATION OF SURFACE DEFECTS ON CONTINOUSLY CAST PRODUCTS
Abstract	The invention is directed to a method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces, wherein defects and/or flaws are determined with respect to their exact position, evaluated with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product, or are prevented by optimizing the process. The aim is a more reliable and only strictly necessary evaluation of defects. This aim is achieved in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and are stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the information from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product.

Title of Invention

A METHOD FOR THE DETECTION AND CLASSIFICATION OF SURFACE DEFECTS ON CONTINOUSLY CAST PRODUCTS

Abstract

The invention is directed to a method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces, wherein defects and/or flaws are determined with respect to their exact position, evaluated with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product, or are prevented by optimizing the process. The aim is a more reliable and only strictly necessary evaluation of defects. This aim is achieved in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and are stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the information from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product.

Full Text	METHOD FOR DETECTING AND CLASSIFYING SURFACE DEFECTS ON CONTINUOUSLY CAST SLABS The invention is directed to a method for detecting and classifying surface defects on continuously cast products using topogi-aphical information about the appearance of continuously cast surfaces in which defects and/or flaws are determined and evaluated with respect to their exact position. Numerous methods are known for detecting and eliminating surface defects on materials such as continuously cast products. For example, according to EP 0 880 023 Al, surface defects can be detected automatically and are removed subsequently by a grinding macliine before further processing, that is, before finish rolling in a roll mill. The grinding machine used for this purpose can operate reversibly so that successive defects or defects which are scattered over large surface areas can be determined by an inspection device arranged in front of and behind the grinding machine and can then be eliminated. In this method, defects are evaluated on the basis of comparisons with stored models so that the quality of defect detection and, therefore, elimination depends upon the material that has been stored. Accordingly, superfluous work steps cannot always be avoided. hi principle, only those surface defects which also lead to defects in the rolled product, e.g., hot strip or sheet, should be detected and evaluated. It is not neces.say to eliminate all of the other flaws on the slab surface. Therefore, it is the object of the invention to provide a method by which a more reliable and only strictly necessary evaluation and subsequent elimination of defects can be achieved. In so doing, the infomiation obtained in this way is used in conformity to the assessment to eliminate the defects before further machining of the product or for timely determination and classification of the possible quality of the finished product. In a method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces in which defects and/or flaws are determined with respect to their exact position, evaluated with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product, the above-stated object is met in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the information from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product. According to the invention, the surface topography of the continuously cast slabs is determined by suitable methods. Such raetlions are optical methods operating in the visible or invisible spectrum of light or microwave-based methods. In the visible range, fiinge projection methods and stereoscopic methods are used. Laser-based methods are also possible, The information about surface topography acquired by means of one or more of these methods is stored in a location-dependent manner, i.e., corresponding to position. The evaluation of detected changes in topography can be carried out by suitable classifying methods, e.g., neural networks, or the like. According to the invention, the method is characterized by a learning phase during which the classification methods are optimized so as to distinguish between irrelevant surface flaws and relevant surface defects which lead to, or have led to, surface defects on the finish-rolled products, in this case, hot strip or sheet. For this purpose, the results of a surface inspection on die finished produced are linked to the slab inspection system. In particular, the absolute position of the defects on the finished product is converted to the absolute position on the slab surface. For this purpose, the pass sequence data such as total degree of deformation and ratio of cross rolling to longitudinal rolling are incorporated in the model. The information on the position of a likely surface defect on tlie slab is stored and compared with the information determined by the inspection system following finish rolling so that the self-learning effect takes place. The relationships found by means of the classifying methods, e.g., neural networks, between the topography and the probability of the occurrence of surface defects are then used for a prediction. The method is able to measures both hot 'and cold slab surfaces, and either the slab or the measuring device is moved during measurement. The movement can be carried out in discrete increments or continuously. The topographical information obtained in this way can be stored with its absolute position in accordance with the results of the classification. As was mentioned above, the decision about whether or not a surface defect exists which can be removed prior to further processing of the slab surface or which leads to a loss of quality on the finished product makes a learning phase necessary. This learning can be carried out manually through inspector guidelines, but is then subject to subjective errors. Therefore, according to the invention, this problem is solved by linking to surface inspection systems on the finished product. When the method is used in an unlinked installation (slab continuous casting plant and rolling mill) without direct use of the slab or with externally purchased slabs, this learning algorithm must be implemented using a long-term database. Another possibility consists in die use of this system within the framework of an automated slab inspection. In this case, the defects are marked by a marking robot to facilitate location for repair purposes. The position information can also be forwarded to a machining center for automatic repairs. The method according to the invention will be described in the following with reference to the drawings. The drawings show: Fig. 1 a sample in which artificial flaws, i.e., defects, have been incorporated; Fig. 2 the results of a measurement of the sample by means of a fringe projection method, specifically line 5 in Figure 1; Fig. 3 the topographical evaluation of this line; Fig. 4 an embodiment example for a measuring device; and Fig. 5 the principle of the inspecting and evaluating system according to the invention. Flaws were artificially created in a sample for illustrating the method. This is shown in Figure 1, The lines are shown on the right-hand side of the illustration. Line 5 is referred to by way of example for the following description. The sample was measured by means of a fringe projection method and the results for line 5 are shown in Figure 2. The topogrphical information can be correlated with a punctifonn defect as is shown in Figure 3. Figure 4 shows an example for the arrangement of the measuring device with projector and camera above a table roller serving to transport the slab. Finally, Figure 5 shows the basic concept, wherein the defects and/or flaws on the preliminary product, i.e., the slab, are detected by the first surface inspection, and a second inspection device then detects defects and/or flaws on the roiled finished product. Based on the comparison that is then earned out, conclusions can be reached concerning which of the defects initially detected have also led to a defect on the finished product so that a learning process can be initiated which leads to an improved evaluation of defects on the preliminary product with the result that only those defects which are disadvantageous for the finished product need be eliminated. 1. Method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces, wherein defects and/or flaws are determined with respect to their exact position, evaluated with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product or are prevented by optimizing tire process, characterized in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the infonnation from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product. 2. Method according to claim 1, characterized in that the topographical information is determined by means of optical methods operating in the visible or invisible region, 3. Method according to claim 1, characterized in that the topographical infoiTnation is detennined by means of laser-based or microwave-based metliods. 4. Method according to claim 2, characterized in that the topographical infonnation is obtained by fringe projection methods or stereoscopic methods. 5. Method according to claim 1, characterized in that the topographical infonnation is determined by means of visible or invisible sources of electromagnetic radiation. 6. Method according to one of the preceding claims, characterized in that the acquired information which is stored in a location-dependent manner is evaluated by classifying methods such as neural networks. 7. Method according to one of the preceding claims, characterized in that during the learning phase only those areas on the slab surface which lead to surface defects on the finish-rolled product, such as hot strip or sheet, are assessed or accounted for as defective. 8. Method according to one of the preceding claims, characterized in that the absolute position of the defects on the finished product is converted to the absolute position on the slab surface, wherein the pass sequence data such as total degree of deformation and ratio of transverse rolling to longitudinal rolling are incorporated in the calculation for this purpose. 9. Method according to one of the preceding claims, characterized in that the relationships found by means of neural networks or other methods between the topography and the probability of the occurrence of surface defects are used for a prediction. 10. Method according to claim 8, characterized in that the acquired topographical infomiation is stored with its absolute position depending upon the results. 11. Method according to one of the preceding claims, characterized by the use thereof within the framework of an automated slab inspection, wherein the defects are marked by a marking robot. 12. Method according to one of the preceding claims, characterized in that the machining information obtained from the comparison of information is forwarded to a machining center for automatic repair.

Full Text

METHOD FOR DETECTING AND CLASSIFYING SURFACE DEFECTS ON CONTINUOUSLY CAST SLABS
The invention is directed to a method for detecting and classifying surface defects on continuously cast products using topogi-aphical information about the appearance of continuously cast surfaces in which defects and/or flaws are determined and evaluated with respect to their exact position.
Numerous methods are known for detecting and eliminating surface defects on materials such as continuously cast products. For example, according to EP 0 880 023 Al, surface defects can be detected automatically and are removed subsequently by a grinding macliine before further processing, that is, before finish rolling in a roll mill. The grinding machine used for this purpose can operate reversibly so that successive defects or defects which are scattered over large surface areas can be determined by an inspection device arranged in front of and behind the grinding machine and can then be eliminated.
In this method, defects are evaluated on the basis of comparisons with stored models so that the quality of defect detection and, therefore, elimination depends upon the material that has been stored.
Accordingly, superfluous work steps cannot always be avoided.
hi principle, only those surface defects which also lead to defects in the rolled product, e.g., hot strip or sheet, should be detected and evaluated. It is not neces.say to eliminate all of the other flaws on the slab surface.
Therefore, it is the object of the invention to provide a method by which a more reliable and only strictly necessary evaluation and subsequent elimination of defects can be achieved.
In so doing, the infomiation obtained in this way is used in conformity to the assessment to eliminate the defects before further machining of the product or for timely determination and classification of the possible quality of the finished product.
In a method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces in which defects and/or flaws are determined with respect to their exact position, evaluated

with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product, the above-stated object is met in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the information from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product.
According to the invention, the surface topography of the continuously cast slabs is determined by suitable methods. Such raetlions are optical methods operating in the visible or invisible spectrum of light or microwave-based methods. In the visible range, fiinge projection methods and stereoscopic methods are used. Laser-based methods are also possible, The information about surface topography acquired by means of one or more of these methods is stored in a location-dependent manner, i.e., corresponding to position. The evaluation of detected changes in topography can be carried out by suitable classifying methods, e.g., neural networks, or the like.
According to the invention, the method is characterized by a learning phase during which the classification methods are optimized so as to distinguish between irrelevant surface flaws and relevant surface defects which lead to, or have led to, surface defects on the finish-rolled products, in this case, hot strip or sheet.
For this purpose, the results of a surface inspection on die finished produced are linked to the slab inspection system. In particular, the absolute position of the defects on the finished product is converted to the absolute position on the slab surface. For this purpose, the pass sequence data such as total degree of deformation and ratio of cross rolling to longitudinal rolling are incorporated in the model. The information on the position of a likely surface defect on tlie slab is stored and compared with the information determined by the inspection system following finish rolling so that the self-learning effect takes place.
The relationships found by means of the classifying methods, e.g., neural networks, between the topography and the probability of the occurrence of surface defects are then used for a prediction.

The method is able to measures both hot 'and cold slab surfaces, and either the slab or the measuring device is moved during measurement. The movement can be carried out in discrete increments or continuously.
The topographical information obtained in this way can be stored with its absolute position in accordance with the results of the classification.
As was mentioned above, the decision about whether or not a surface defect exists which can be removed prior to further processing of the slab surface or which leads to a loss of quality on the finished product makes a learning phase necessary. This learning can be carried out manually through inspector guidelines, but is then subject to subjective errors. Therefore, according to the invention, this problem is solved by linking to surface inspection systems on the finished product.
When the method is used in an unlinked installation (slab continuous casting plant and rolling mill) without direct use of the slab or with externally purchased slabs, this learning algorithm must be implemented using a long-term database.
Another possibility consists in die use of this system within the framework of an automated slab inspection. In this case, the defects are marked by a marking robot to facilitate location for repair purposes. The position information can also be forwarded to a machining center for automatic repairs.
The method according to the invention will be described in the following with reference to the drawings.
The drawings show:
Fig. 1 a sample in which artificial flaws, i.e., defects, have been incorporated;
Fig. 2 the results of a measurement of the sample by means of a fringe projection method, specifically line 5 in Figure 1;
Fig. 3 the topographical evaluation of this line;
Fig. 4 an embodiment example for a measuring device; and
Fig. 5 the principle of the inspecting and evaluating system according to the invention.

Flaws were artificially created in a sample for illustrating the method. This is shown in Figure 1, The lines are shown on the right-hand side of the illustration. Line 5 is referred to by way of example for the following description.
The sample was measured by means of a fringe projection method and the results for line 5 are shown in Figure 2. The topogrphical information can be correlated with a punctifonn defect as is shown in Figure 3.
Figure 4 shows an example for the arrangement of the measuring device with projector and camera above a table roller serving to transport the slab.
Finally, Figure 5 shows the basic concept, wherein the defects and/or flaws on the preliminary product, i.e., the slab, are detected by the first surface inspection, and a second inspection device then detects defects and/or flaws on the roiled finished product. Based on the comparison that is then earned out, conclusions can be reached concerning which of the defects initially detected have also led to a defect on the finished product so that a learning process can be initiated which leads to an improved evaluation of defects on the preliminary product with the result that only those defects which are disadvantageous for the finished product need be eliminated.

1. Method for the detection and classification of surface defects on continuously cast products using topographical information about the appearance of continuously cast surfaces, wherein defects and/or flaws are determined with respect to their exact position, evaluated with respect to their location and dimensions, and eliminated in accordance with the evaluation prior to further machining of the product or are prevented by optimizing tire process, characterized in that, on the one hand, the defects and/or flaws on the slab surface of the continuously cast preliminary product are detected and stored with respect to their exact position and, on the other hand, a detection of defects and/or flaws on the finished product is carried out and stored with respect to their exact position, and in that the information from the preliminary product is then compared with the infonnation from the surface inspection on the finished product, and only the information which has led to, or can lead to, defects on the finished product is taken into account for the elimination of defects and/or flaws on the preliminary product.
2. Method according to claim 1, characterized in that the topographical information is determined by means of optical methods operating in the visible or invisible region,
3. Method according to claim 1, characterized in that the topographical infoiTnation is detennined by means of laser-based or microwave-based metliods.
4. Method according to claim 2, characterized in that the topographical infonnation is obtained by fringe projection methods or stereoscopic methods.
5. Method according to claim 1, characterized in that the topographical infonnation is determined by means of visible or invisible sources of electromagnetic radiation.
6. Method according to one of the preceding claims, characterized in that the acquired information which is stored in a location-dependent manner is evaluated by classifying methods such as neural networks.

7. Method according to one of the preceding claims, characterized in that during the learning phase only those areas on the slab surface which lead to surface defects on the finish-rolled product, such as hot strip or sheet, are assessed or accounted for as defective.
8. Method according to one of the preceding claims, characterized in that the absolute position of the defects on the finished product is converted to the absolute position on the slab surface, wherein the pass sequence data such as total degree of deformation and ratio of transverse rolling to longitudinal rolling are incorporated in the calculation for this purpose.
9. Method according to one of the preceding claims, characterized in that the relationships found by means of neural networks or other methods between the topography and the probability of the occurrence of surface defects are used for a prediction.
10. Method according to claim 8, characterized in that the acquired topographical infomiation is stored with its absolute position depending upon the results.
11. Method according to one of the preceding claims, characterized by the use thereof within the framework of an automated slab inspection, wherein the defects are marked by a marking robot.
12. Method according to one of the preceding claims, characterized in that the machining information obtained from the comparison of information is forwarded to a machining center for automatic repair.

Documents:

http://ipindiaonline.gov.in/patentsearch/GrantedSearch/viewdoc.aspx?id=qpgEOKmcZezNpIYSH4AHFA==&loc=egcICQiyoj82NGgGrC5ChA==

« Previous Patent

Next Patent »

Patent Number

277952

Indian Patent Application Number

6254/CHENP/2009

PG Journal Number

51/2016

Publication Date

09-Dec-2016

Grant Date

07-Dec-2016

Date of Filing

22-Oct-2009

Name of Patentee

SMS GROUP GMBH.

Applicant Address

EDUARD-SCHLOEMANN-STRASSE 4, 40237 DUSSELDORF

Inventors:

#	Inventor's Name	Inventor's Address
1	SUDAU, PETER	WILHELM-MUNKER-STRASSE 8, 57271 HILCHENBACH
2	WEINERT, ANDREAS	BYFANGER STRASSE 70, 45257 ESSEN
3	ROSENTHAL, DIETER	BERGSTRASSE 22A, 57572 NIEDERFISCHBACH,
4	SCHULZE, STEPHAN	ZUR HEIDE 10, 40880 RATINGEN,
5	SCHUSTER, INGO	BONNENRING 83, 47877 WILLICH,
6	FACKERT, RAINER	GRABENSTRASSE 5, 56584 MEINBORN
7	SCHUMACHER, WILFRIED	GUSTAV-HEINEMANN STRASSE 3, 42579 HEILIGENHAUS,

PCT International Classification Number

B21B37/00

PCT International Application Number

PCT/DE08/00582

PCT International Filing date

2008-04-02

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	10 2007 020 240.9	2007-04-24	Germany