DE102020107779A1 - Method and test system for testing a bipolar plate of a fuel cell - Google Patents

Abstract

A method for testing a bipolar plate of a fuel cell is provided in which an image (12) of a surface of a bipolar plate (14) is created, the image (12) is examined for possible defects (28) by an automated image processing-supported evaluation system and in which In the event that the evaluation system identifies an examined bipolar plate (14) as a potentially defective suspect plate, a detailed check (22) of an area of the suspect plate identified as potentially defective is carried out. The evaluation methods used in the image processing of the image (12) in the automated evaluation system can increase the test quality when testing a bipolar plate of a fuel cell without significant additional effort, so that a cost-effective and reliable test is made possible.

Description

The invention relates to a method for testing a bipolar plate of a fuel cell and a test system provided for this purpose, with the aid of which a bipolar plate of a fuel cell can be examined for defects during manufacture.

A fuel cell stack or stack typically includes a plurality of fuel cells in a stack configuration. The respective fuel cell comprises an electrolyte and electrodes, which are contacted via conductive plates. In the case of a polymer electrolyte fuel cell that is operated in the low temperature range, a polymer membrane-electrode unit is present. Electrically conductive bipolar plates, mostly made of metal, are used to separate the individual polymer electrolyte fuel cells in a stack. They not only serve to make electrical contact with the electrodes and conduct the electrical current to the neighboring cell, but also help to supply and distribute fuels and cooling media and to dissipate heat and reaction products. For this purpose, bipolar plates usually have gas distribution fields which, due to their structure, lead to an optimal distribution of the gaseous fuels, mostly hydrogen and oxygen, with regard to the membrane surfaces. The bipolar plate can be produced by two or more thin-walled metal sheets welded to one another, which form the desired flow paths, cooling channels and / or openings.

the end US 2019/0296379 A1 it is known to test a fuel cell by applying different voltages and evaluating them.

There is a constant need to test components of a fuel cell inexpensively and reliably before they are installed in a fuel cell or a fuel cell stack with further components.

It is the object of the invention to identify measures which enable the components of a fuel cell to be tested inexpensively and reliably.

The object is achieved by a method with the features of claim 1 and a test system with the features of claim 7. Preferred embodiments of the invention are specified in the subclaims and the following description, each of which can represent an aspect of the invention individually or in combination .

One embodiment relates to a method for testing a bipolar plate of a fuel cell, in which an image of a surface of a bipolar plate is created, the image is examined for possible defects by an automated image processing-assisted evaluation system and in the event that the evaluation system suspects an examined bipolar plate as potentially defective Plate is identified, a detailed inspection of an area of the suspect plate identified as potentially defective is carried out.

An essential manufacturing step in the manufacture of a fuel cell, in particular a polymer electrolyte fuel cell, is the manufacture of the bipolar plate, which takes place in particular by welding two or more metallic plates or sheets together. The weld seam produced during welding can be easily checked by means of a visual inspection. However, this visual inspection is carried out in an automated manner in that the surface of the bipolar plate, in particular both flat sides of the bipolar plate, is captured in the image with the aid of a camera and is fed to an image processing system of the automated evaluation system. The evaluation system can process the image and, for example, use optical parameters to identify certain shapes and / or colors and / or brightnesses and / or contrasts and check whether these are within certain target values. Such checking activities can be easily automated and implemented with the help of evaluation software in the evaluation system. It can also be checked here whether a weld seam runs in a correct position relative to features or structures that form, for example, media-carrying channels, for example in the area of a gas distribution field. The characteristics and structures of the welded bipolar plates can be identified in the image of the surface of the bipolar plate and used, for example, as a reference for the correct positioning of the weld seam.

In addition to defects relating to the welded connection on a bipolar plate, dimensional deviations of the plate and its structures, in particular the three-dimensional structures in the area of a gas distribution field, the formation of the openings in the area of the plate ends and the gas distribution field and the like, can also be checked in order to identify irregularities, sort out and in the Intervene correctively in the area of the respective process step that potentially generates the deviation.

However, if it should not be possible during the image processing evaluation of the image to identify, for example, the correct execution of the characteristics or structures in the area of a gas distribution field with the required reliability, this can indicate a manufacturing error be due, so that such defects of the bipolar plate can also be checked with the help of the evaluation system.

The bipolar plates found to be potentially not in order by the evaluation system are subjected to the detailed examination as suspect plates, in which an examination is carried out, possibly with significantly greater effort. However, since only the suspect plates are subjected to the detailed examination, the examination effort can be reduced. For example, this means that an eddy current test is not carried out on every bipolar plate, but only on those suspect plates that were found to be suspicious during the automated visual inspection by the evaluation system. This makes use of the knowledge that experience shows that a weld seam that looks almost perfect is not a problem, while a weld seam that does not look perfect may still have a sufficient, in particular sufficiently tight, connection, but this cannot be guaranteed in all cases by a pure visual inspection . The bipolar plates assessed as harmless in automated image processing can thus be processed further without a further detailed check, while after the detailed check, only those suspicious plates are rejected as rejects that actually do not meet the intended criteria and those suspicious plates that do not meet the criteria They still meet the criteria and can be fed back for further processing. Separation of a suspect plate as scrap, which is unnecessary for actual reasons, is avoided, so that unnecessary costs are avoided. The evaluation method used in the image processing of the image in the automated evaluation system allows the inspection quality to be increased when inspecting the fuel cell without significant additional effort, so that a cost-effective and reliable inspection of a fuel cell is made possible.

In particular, in the detailed test, a non-destructive test, in particular penetrant test, magnetic particle test, ultrasonic test, radiographic test and / or eddy current test, is carried out on the suspect plate in the area identified as potentially defective. Destruction of the suspicion plate during the detailed inspection can thus be avoided. In the event that the detailed inspection should show that the suspicion plate does meet the requirement profile, the suspicion plate can be returned to the subsequent manufacturing steps as a regular bipolar plate. Unnecessary costs in the production of the fuel cell can thereby be avoided. At the same time, in comparison to a purely visual inspection, a measurement of the conditions present in the interior of the suspicion plate can take place, whereby a particularly reliable inspection is achieved. The detailed test can be carried out automatically and / or by a test person trained for this purpose.

The requirement profile can relate in particular to the tightness and position of a weld seam, the permeability and position of a structure in the gas distribution field, the arrangement of openings for fuel gas supply, the design and position of seals and the like.

The evaluation system preferably has a neural network trained on the basis of a plurality of non-defective and defective bipolar plates depicting images of bipolar plates for identifying a suspect plate. The neural network can be trained by machine learning so that the image evaluation of the images in the evaluation system can be improved. This makes it possible, in particular, to take into account a large number of different, possibly differently weighted parameters when evaluating the images. The weighting of the parameters during the training of the neural network is adapted, in particular automatically, in such a way that during the image evaluation of the image used for the training, for which it is already known in advance whether a bipolar plate with such an appearance is OK or not , the correct images are rated as OK or not OK. This makes it possible to automatically recognize complex damage patterns even with complex geometries. This can improve the quality of the exam.

The neural network is particularly preferably further trained with the results of the detailed check. The detailed examination of the suspect plates can lead to an additional gain in knowledge when evaluating the bipolar plates, which can be fed back to the evaluation system as additional training data. This leads to reinforcement learning of the neural network, in particular as part of a deep learning structure with several intermediate layers of the neural network. As a result, the quality of the test can be continually improved over time. Reinforcement learning can be continuous or incremental.

In particular, two or more metal sheets are welded to produce the bipolar plate, with several images of a weld seam occurring during welding being created and evaluated and / or an image of the entire weld seam being created and evaluated after welding. The image evaluation in the evaluation system can already take place during the welding process, see above that melt zones that arise when the areas to be welded are melted can also be taken into account in the evaluation. However, it is also possible to carry out the evaluation only after completion of the welding process, as a result of which the amounts of data to be taken into account in the evaluation, in particular the number of images, can be reduced. In addition, it is possible to illuminate the bipolar plate and create the image with the aid of a camera independently of the welding process and / or independently of light effects and / or reversible thermal expansion effects occurring during welding.

The evaluation system preferably uses the results of the evaluation to adapt a process control of the welding. In particular in the event that a bipolar plate was assessed as not in order, the evaluation system can classify the possible defect on the basis of the criterion used for this purpose. This makes it possible to assign certain identified defects to a certain classification and, if necessary, to certain causes.

If the defect of the suspect plate assigned to a certain class of defects can be assigned to an incorrect process control during welding, it is possible, by feeding back the results of the evaluation by the evaluation system with the process control of the welding, to adapt the process control so that such defects to be avoided in the future. The evaluation system can be part of a control loop for the welding process control, for example in order to make adjustments to control variables due to wear. This allows manufacturing errors during welding to be recognized very early on and corrected automatically, so that unnecessary rejects and unnecessary costs are avoided.

Another embodiment relates to a test system for testing a bipolar plate of a fuel cell, with a camera for creating at least one image of a surface of the bipolar plate, an automated image processing-supported evaluation system for examining the image for possible defects and a switch that can be actuated by the evaluation system for deriving one from the evaluation system Bipolar plate identified as a potentially defective suspect plate to a detailed inspection station. The test system is especially designed to carry out the method described above. The test system can preferably be designed and developed, as explained above with reference to the method.

The bipolar plates found to be potentially not in order by the evaluation system are subjected to the detailed examination as suspect plates, in which an examination is carried out, possibly with significantly greater effort. However, since only the suspect plates are subjected to the detailed inspection in the detailed inspection station, the inspection effort can be reduced. The bipolar plates assessed as harmless in the automated image processing in the evaluation system can thus be further processed without a further detailed check, while of the suspect plates after the detailed check, only those are rejected as rejects that actually do not meet the intended criteria. Those suspicious plates that still meet the criteria are sent back for further processing. An unnecessary separation of a suspicion plate as scrap is avoided, so that unnecessary costs are avoided. The evaluation methods used in the image processing of the image in the automated evaluation system can increase the quality of the test when testing the bipolar plate without significant additional effort, so that a cost-effective and reliable test is made possible.

Particularly preferably, the evaluation system has a neural network trained on the basis of a large number of non-defective and defective bipolar plates depicting images of bipolar plates for identifying a suspect plate, the evaluation system having an interface coupled to the neural network, in particular communicating with the detailed inspection station, for feeding in the Has results of the detailed test for the purpose of further training the neural network with the results of the detailed test. The detailed examination of the suspect plates in the detailed examination station can lead to an additional gain in knowledge when evaluating the bipolar plates, which can be fed back to the evaluation system as training data via the interface of the evaluation system. This leads to reinforcement learning of the neural network, in particular as part of a deep learning structure with several intermediate layers of the neural network. As a result, the quality of the test can be continually improved over time.

In particular, the evaluation unit has an output port that can be coupled to a production unit, in particular a laser welding system, for producing the bipolar plate, for adapting a process control of the production unit as a function of the results of the evaluation by the evaluation system. The evaluation system can, in particular in the event that a bipolar plate has been assessed as not in order, based on the criterion used for this purpose, a classification the possible defect. This makes it possible to assign certain identified defects to a certain classification and, if necessary, to certain causes. If the defect of the suspect plate assigned to a certain class of defects can be assigned to a faulty process management in the production unit, it is possible, by feeding back the results of the evaluation via the output port of the evaluation system to the production unit, to adapt the process management so that such defects to be avoided in the future. The evaluation system can be part of a control loop for the process control of the production unit, for example in order to make adjustments to control variables due to wear. Manufacturing errors can be recognized very early and automatically corrected, so that unnecessary rejects and unnecessary costs are avoided.

The detailed testing station preferably has a testing device for non-destructive testing, in particular penetrant testing, magnetic particle testing, ultrasonic testing, radiographic testing and / or eddy current testing, of the suspect plate. Destruction of the suspicion plate during the detailed inspection can thus be avoided. In the event that the detailed inspection should show that the suspicion plate does meet the requirement profile, the suspicion plate can be returned to the subsequent manufacturing steps as a regular bipolar plate. Unnecessary costs in the production of the components of a fuel cell can thereby be avoided. At the same time, in comparison to a purely visual inspection, a measurement of the conditions present in the interior of the suspicion plate can take place, whereby a particularly reliable inspection is achieved. The detailed test can be carried out automatically and / or by a test person trained for this purpose.

The method according to the invention and the testing system according to the invention are particularly suitable for testing bipolar plates for use in a fuel cell, in particular a polymer electrolyte fuel cell, the bipolar plate being formed from two or more metal sheets.

• 1: einen schematischen Ablauf für die Durchführung einer Prüfung einer Bipolarplatte, und
• 2: eine schematische perspektivische Ansicht eines Teils einer Prüfanlage für die die Durchführung einer Prüfung bei einer Bipolarplatte.

In the following, the invention is explained by way of example with reference to the attached drawings using preferred exemplary embodiments, the features shown below being able to represent an aspect of the invention both individually and in combination. Show it:

• 1 : a schematic procedure for carrying out a test of a bipolar plate, and
• 2 : a schematic perspective view of part of a test system for performing a test on a bipolar plate.

The in 1 illustrated sequence 10 will be several in a first step provided, in particular the surface of a bipolar plate 14th after a welding process. In a second step, the for example divided into two categories, one of which is 16 for is provided which has a bipolar plate 14th which is OK, and the other category 18th for is provided which has a bipolar plate 14th that is not OK (NOK). In particular, it can be used for defective bipolar plates 14th intended category 18th have several sub-categories, which are assigned different causes for different defect patterns. In a third step, a neural network 20th with the help of in the categories 16 , 18th divided be trained. The neural network 20th can then in a fourth step during a manufacturing process of the fuel cell 14th evaluate independently as OK or NOK. The bipolar plates found to be OK 14th can be fed to a further production step. The suspicion plates that have been assessed as potentially not in order can be checked in detail in a fifth step 22nd be subjected to the by the neural network 20th Confirm or correct an assessment classified as NOK. In particular, the final assessment is during the detailed review 22nd A trained person is included, who can also record features that cannot be automatically recorded and incorporate them into the assessment. During the detailed inspection 22nd A result is automatically produced in which with a very high level of reliability from the neural network 20th at least classified in the NOK category as a case of doubt classified as actually NOK or OK. This corresponds to the state after the second step, in the training data for the neural network 20th generated. This as a result of the detailed examination 22nd automatically generated data can be sent as additional training data to the neural network 20th are fed back to reinforcement learning of the neural network 20th to reach.

As in 2 is shown, the test system 24 a camera 26th have, which from a surface of a, in particular welded, bipolar plate 14th one generated. The camera can also have a plurality of or even create a movie sequence. The bipolar plate 14th can be suitably illuminated for this purpose, so that in the all relevant features can be seen and a possible defect 28 of one the neural network 20th having evaluation system can be recognized by an optical evaluation.

In addition to defects relating to the welded connection on a bipolar plate, dimensional deviations of the plate and its structures, in particular the three-dimensional structures in the area of a gas distribution field, the formation of the openings in the area of the plate ends and the gas distribution field and the like, can also be checked in order to identify irregularities, sort out and in the Intervene correctively in the area of the respective process step that potentially generates the deviation.

List of reference symbols

10

sequence

12th

Illustration

14th

Bipolar plate

16

a category

18th

other category

20th

neural network

22nd

Detailed examination

24

Testing system

26th

camera

28

possible defect

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was 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.

Patent literature cited

• US 2019/0296379 A1 [0003]

Claims (10)

Method for testing a bipolar plate of a fuel cell, in which an image (12) of a surface of a bipolar plate (14) is created, the image (12) is examined for possible defects (28) by an automated image processing-supported evaluation system and in the event that the Evaluation system, an examined bipolar plate (14) is identified as a potentially defective suspect plate, a detailed check (22) of an area of the suspect plate identified as potentially defective is carried out. Procedure according to Claim 1 In the detailed test (22) a non-destructive test, in particular penetrant test, magnetic particle test, ultrasonic test, radiographic test and / or eddy current test, is carried out on the suspect plate in the area identified as potentially defective. Procedure according to Claim 1 or 2 , in which the evaluation system has a neural network (20) trained on the basis of a plurality of non-defective and defective bipolar plates (14) representing images (12) of bipolar plates (14) for the identification of a suspect plate. Procedure according to Claim 3 , in which the neural network (20) continues to be trained with the results of the detailed test (22). Method according to one of the Claims 1 until 4th , in which two or more metal sheets are welded to produce the bipolar plate (14), with several images (12) of a weld seam occurring during welding being created and evaluated and / or an image (12) of the whole after welding Weld seam is created and evaluated. Procedure according to Claim 5 , in which the results of the evaluation by the evaluation system are used to adapt a process control of the welding. Test system for testing a bipolar plate of a fuel cell, with a camera (26) for creating at least one image (12) of a surface of the bipolar plate (14), an automated image processing-assisted evaluation system for examining the image (12) for possible defects (28) and a switch that can be actuated by the evaluation system to divert a bipolar plate identified by the evaluation system as a potentially defective suspect plate to a detailed inspection station. Test system according to Claim 7 , characterized in that the evaluation system has a neural network (20) trained on the basis of a plurality of non-defective and defective bipolar plates (14) depicting images (12) of bipolar plates (14) for identifying a suspect plate, the evaluation system having one with the neural network (20) coupled, in particular communicating with the detailed test station, interface for feeding in the results of the detailed test (22) for the purpose of further training the neural network (20) with the results of the detailed test (22). Test system according to Claim 8 , characterized in that the evaluation unit has at least one output port that can be coupled to at least one production unit, in particular a welding system and / or a stamping system and / or a forming system, for the production of the bipolar plate (14) for adapting a process control of the production unit as a function of the results of the evaluation has the evaluation system. Test system according to one of the Claims 7 until 9 , characterized in that the detailed testing station has a testing device for non-destructive testing, in particular for penetrant testing, magnetic particle testing, ultrasonic testing, radiographic testing and / or eddy current testing, of the suspect plate.

Images