DE102009059092A1 - Method for distinguishing and identifying workpieces made of ferromagnetic material by means of nondestructive testing - Google Patents

Abstract

The invention relates to a method for differentiating and identifying workpieces made of ferromagnetic material, in which the workpiece is brought into an alternating magnetic field by means of a coil, as a result of which a voltage is induced in a second coil and its amplitude and phase are evaluated and the test result with in sorting limits for test parts of the same material and the same heat treatment are compared to a previously carried out calibration process, the calibration being carried out with defined test parameter sets consisting of the excitation and measurement frequency and the response signal being evaluated for each test parameter set according to amplitude and phase and assigned to test classes. During the calibration, the measurement results of all test parameter sets used are evaluated simultaneously using a multivariate optimization algorithm in a 2 × n-dimensional space (n: number of test parameter sets and n ≥ 2) and the affiliation to a class is calculated using an n - 1 dimensional hypersurface shown as a dividing surface between the classes.

Description

The invention relates to a method for distinguishing and identifying workpieces made of ferromagnetic material, in particular steel, according to the preamble of claim 1.

The invention will be described below on steel blocks for the production of seamless tubes, but it is applicable to all distinguishable or identifiable components or Werkstücke-, such as sheets, rods, etc., of ferromagnetic material.

For the quality assurance and cost optimization in the production of hot-finished steel tubes, the traceability and clear distinction or identification of the manufactured products, ie when what is where and how has been manufactured, is of decisive importance. For example, the traceability of pipes by means of data matrix codes after the DE 10 2004 051 124 B3 applied.

Directly linked to a complete quality assurance in the production but also the unique identification or distinction of the input material used for manufacturing.

For the production of seamless tubes round blocks of different alloy composition are used as starting material. These blocks are stored on a so-called block space. There may occasionally come to confusion with the production to be used round blocks, if z. B. round blocks of different alloy composition are used.

To detect confusion, therefore, a confounding test of the finished tubes is usually made. For this purpose, z. B. an emission spectrometer can be used with which, however, not all materials can be safely distinguished.

In particular, it is difficult to separate materials that differ substantially in carbon content. Therefore, in some cases additionally the so-called spark test is applied. For this test, sparks are generated with a grinding machine and their flight and characteristics are evaluated visually.

With these methods, however, no distinction between pipes of the same steel types can be made with, z. B. different heat post-treatments have been subjected. In addition, this method is not automatable and depends on the experience of the examiner. For safe identification under quality assurance and cost aspects, this method is therefore not useful.

Also an automated on-line measurement method of identifying workpieces made of ferromagnetic material under the name Magnatest ^® is known ( www.foerstergroup.de ), which can be used commercially for identification or sorting in test classes of ferromagnetic materials. According to this known method, the workpieces are placed in an alternating magnetic field for testing, which is produced in a continuous coil. Due to the relative permeability and the electrical conductivity of the material of the specimen, the magnetic field is influenced, so that in a second coil, a voltage is induced. This is evaluated with regard to its amplitude and phase and the result is compared with limits defined in a calibration process as a threshold value for the test group. With this method, in principle, different metallic materials, as well as the same metals with different alloy composition or different structural states distinguishable.

Before the materials can be tested, the device must first be set up and calibrated. The basis for a test order is the test definition, which contains the settings for the test, in particular the test parameters. These include the test frequency including the frequency (s) at which the signal is evaluated. The calibration is used to define the sorting areas. For this purpose, parts of the first test class, i. H. Class of test pieces with the same material and heat treatment, transported through the coil and made measurements. After testing a certain number of parts (test part group) with at least 5 and a maximum of 30 test specimens, a tolerance field is formed as a sorting limit around the results of the respective test part group and separately for each set of test parameters. It must be ensured by suitable means that only calibration parts made of the same material and with the same heat treatment are used.

The tolerance fields are created automatically in the known calibration method according to fixed specifications and can then be manually changed in size and shape. As a selection for the shapes circles, ellipses, rectangles and polygons are available.

After determining the first check class, the process can be repeated for a further (6 total) classes. This completes the calibration and the actual test of the DUTs can be started.

The handling of the test specimens during the test is identical to that during the calibration process. After the measurement, the measured values for each test parameter set are compared individually with the respective tolerance fields. If a test specimen is not in any of the tolerance fields for at least one parameter set, it is rejected as being confused.

In addition to the unambiguous assignment of a part to a class, however, a test value in different parameter sets can also fall into different classes and thus can not be assigned unambiguously. Especially with materials similar alloy composition with different heat treatment, z. As remuneration, this evaluation method can not ensure an accurate identification or distinction and therefore lead to false positives.

The object of the invention is to improve the known method for identification or differentiation of ferromagnetic materials so that a higher selectivity between the determined test classes is achieved and thus an identification or distinction is possible even with similar alloy composition and different heat treatment.

According to the invention, the object is achieved by simultaneously evaluating the measurement results of all test parameter sets used via a multivariate optimization algorithm in a 2 × n-dimensional space (n: number of test parameter sets and n ≥ 2) and assigning them to a class via an The measured n - 1d dimensional hypersurface is displayed as the interface between the classes.

In an advantageous embodiment of the invention, the profile of the separating surface between the different classes is optimized so that the distance between the result values of the different test classes which are closest to the separating surface is greatest. The n-1-dimensional separating surface can have any shape, such as a hyperplane.

• • Beim Kalibrieren (Lernen) werden die Klassen unter gleichzeitiger Berücksichtigung der Messergebnisse aller Prüfparametersätze unterschieden. Damit ist prinzipiell eine bessere Trennung der Messergebnisse möglich. Es kann außerdem eine Bewertungsskala der für die Klassifikation wichtigsten Prüfparametersätze erstellt werden, um damit gezielt die Prüfung zu verbessern. Die einzelnen Prüfparametersätze einer Prüfdefinition werden dabei im Zuge der Auswertung der Messergebnisse einer Bewertung im Hinblick auf ihre Eignung zur Unterscheidung der Prüfklassen unterzogen, wobei entsprechend der Bewertung eine Auswahl an geeigneten Prüfparametersätzen vorgenommen wird.
• • Der erfindungsgemäß verwendete multivariate Algorithmus optimiert automatisch die Trennung der Prüfklassen. Ein manueller Eingriff ist nicht nötig. Die Form der Trennfläche ist flexibel und wird ebenfalls automatisch optimiert.
• • Da erfindungsgemäß für jeden Messwert ein Vertrauenswert für die Zugehörigkeit zu einer Prüfklasse angegeben wird, ergibt sich als Ergebnis der Teileprüfung auf Basis der simultanen Bewertung der Messergebnisse aller Prüfparameter eine genauere Beurteilung der Prüfklassenzugehörigkeit, als bei einem festen Schwellenwert.

The present invention overcomes the disadvantages of the known Magnatest -Kalibrierverfahrens ^®, and in particular benefits with regard to a more accurate classification (discrimination) of ferromagnetic materials. Instead of the known method for calibration and evaluation in a two-dimensional space, the present invention uses a multivariate algorithm (data mining) in a 2 × n-dimensional space with n ≥ 2, which optimizes the separation of the test classes. In particular, the following advantages of the invention result:

• • During calibration (learning), the classes are distinguished by taking into account the measurement results of all test parameter sets. Thus, a better separation of the measurement results is possible in principle. In addition, a rating scale of the most important test parameter sets for the classification can be created in order to improve the test in a targeted manner. In the course of evaluating the measurement results, the individual test parameter sets of a test definition are subjected to an evaluation with regard to their suitability for distinguishing the test classes, with a selection of suitable test parameter sets being made according to the evaluation.
• The multivariate algorithm used according to the invention automatically optimizes the separation of the test classes. A manual intervention is not necessary. The shape of the interface is flexible and is also automatically optimized.
• • Since, according to the invention, a confidence value for belonging to a test class is given for each measured value, the result of the parts test based on the simultaneous evaluation of the measurement results of all test parameters is a more precise assessment of the test class membership than for a fixed threshold value.

Tests with data obtained in the laboratory from pipe sections made of different starting material gave excellent results. This meant that almost all materials made available could be sorted correctly. Even a confusion test between pipes made of the same material but with different remuneration was possible.

For a calibration it must be ensured that the test specimens used for the calibration are not of different material or different temper.

In a first step, for the calibration z. B. 5 tubes of the same material used. Thus, the pipes to be tested below can be distinguished from other materials for whose testing the spark test has hitherto been used, which is therefore superfluous.

In a second step, the calibration is carried out with a larger number of tubes, so as to be able to distinguish pipes with only different heat treatments, such as, for example, tempering. This is done outlier tests according to the known calibration procedure with the Magnatest ^® software by means of a "classic", again each set of parameters is examined separately.

In contrast, the Java software developed in the context of the invention becomes a multivariate one Outlier test performed with the above-described advantages of the evaluation in a 2 × n-dimensional space with n: number of parameter sets and n ≥ 2.

Further features, advantages and details of the invention will become apparent from the following description of the figure shown.

Shown in a two-dimensional projection is the simultaneous evaluation according to the invention using the example of two test parameter sets used for testing via a multivariate optimization algorithm and the evaluation in a four-dimensional space.

Shown are the measurement results on several test specimens of two different ferromagnetic materials, which were measured by the known Magnatest ^® method.

The measurements were carried out with two sets of test parameters, in particular at two different frequencies f ₁ and f ₂ . For each frequency, the known evaluation method yields a two-component vector (x, y) in a two-dimensional space as a result. The measured values for the two metal types are marked with +/- in the diagram. The resulting classes are thus M ⁺ and M ^- . The axes respectively represent the two-dimensional space of the response signals for the frequencies f ₁ and f ₂ , ie the spaces U ₁ = {x ₁ , y ₁ } and U ₂ = {x ₁ , y ₂ }.

The calibration is carried out with the known method as follows: The tolerance fields are determined individually for each frequency, ie these fields result from the projection of M ⁺ or M ^- on the axes and are marked with P in the figure. It is easily seen that the projections of the class M ^- (P ^- ₁ and P ^- ₂₎ in the corresponding projections of the class M + (P ⁺ ₁ and P ⁺ ₂₎ are included. Thus, a separation of these two classes is not possible.

However, considering both frequencies simultaneously according to the invention, d. H. for this example in a four-dimensional space, the classes z. B. clearly separated by the dividing line T. The dividing line is optimized so that the distance between the next points and the dividing line is greatest.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• DE 102004051124 B3 [0003]

Cited non-patent literature

• www.foerstergroup.de [0009]

Claims (7)

Method for distinguishing and identification of workpieces made of ferromagnetic material, in particular steel, in which the workpiece is brought into a magnetic alternating field by means of a coil, whereby a voltage is induced in a second coil and evaluated in terms of their amplitude and phase and the test result in a before the calibration process is carried out for test parts of the same material and the same heat treatment, the calibration is performed with set test parameter sets of exciter and measuring frequency and evaluated for each test parameter set the response signal amplitude and phase and assigned test classes, characterized in that during calibration simultaneously evaluate the measurement results of all test parameter sets used via a multivariate optimization algorithm in a 2 × n-dimensional space (n: number of test parameter sets and n ≥ 2) t and the membership of a class is displayed as an interface between the classes via an n - 1dimensional hypersurface calculated from the measured values. A method according to claim 1, characterized in that the course of the interface between the different classes is optimized so that the distance of the separation surface closest to the result values of the different test classes is greatest. A method according to claim 2, characterized in that the n - 1dimensionale separating surface has an arbitrary shape. A method according to claim 3, characterized in that the separating surface is a hyperplane. Method according to one of claims 1-4, characterized in that the individual Prüfparametersätze a test definition in the course of evaluation of the measurement results of a rating in terms of their suitability for distinguishing the test classes are subjected. A method according to claim 5, characterized in that according to the evaluation, a selection of suitable test parameter sets for the test is made. Method according to one of claims 1-6, characterized in that for each measured value, a confidence value for belonging to a test class is specified.

Images