Test system and method for checking a strip of material positioned on a tire building drum

Description

[0001] The invention relates to a testing system for checking a material strip positioned on a tire building drum, comprising an image recording device configured to generate at least one image of a material strip positioned on a tire building drum, and an electronic data processing device configured to determine at least one property of the material strip positioned on the tire building drum on the basis of the image recorded by the image recording device.

[0002] Furthermore, the invention relates to a tire building machine with a tire building drum on which material strips for building a tire part, in particular for building a tire carcass or a belt package, can be positioned, and a testing system for checking the material strips positioned on the tire building drum.

[0003] Furthermore, the invention relates to a method for checking a material strip positioned on a tire building drum of a tire building machine by means of a testing system, comprising the steps of: generating at least one image of a material strip positioned on a tire building drum by means of an image recording device of the image system, and determining at least one property of the material strip positioned on the tire building drum on the basis of the image generated by the image recording device by means of an electronic data processing device of the testing system.

[0004] During tire production, various material layers are positioned as strips on a tire building drum. In the tire manufacturing process, it is common practice to produce the tire carcass and the tire's belt assembly on different tire building drums. During tire carcass production, for example, an inner ply, a textile cord ply, a turn-up, and side strips are positioned as strips on the tire building drum. During belt assembly production, one or more steel cord belt layers, a spool bandage, and the tread are positioned as strips on the tire building drum.

[0005] In automated or semi-automated tire production, it is necessary that all material layers, as well as the edge contours of the individual material layers, are detected by a testing system after the respective material layer has been positioned on the tire building drum before the next production sequence is initiated. In conventional testing systems, the material layer detection takes place in a detection mode of the testing system, while the edge detection takes place in a measurement mode of the testing system.

[0006] DE 102019211023 A1 relates to a method for measuring an open or overlapping joint of a material layer during the production of a green tire in a tire building machine. DE 102019110721 A1 relates to a workflow for annotating training data sets, and in particular to a computer-implemented method for expanding a training data set for machine learning, as well as a corresponding workflow system.

[0007] It has been shown that existing testing systems are not yet sufficiently capable of reliably handling the sometimes widely varying boundary conditions in different production facilities, so that manual intervention in the manufacturing process is often necessary. For example, the lighting conditions, the drum properties, the material properties of the individual material layers, and / or the design of the tire building machine can differ between production facilities. Regarding the drum properties, there may be differences, for example, in the color and / or reflective properties of the contact surface of the tire building drum. Furthermore, the material strips used may differ from one another in terms of their color and reflective properties.Due to the varying lighting conditions in production facilities, differing lighting conditions and / or mirroring and / or reflection effects can occur. These varying boundary conditions lead to errors in the detection of material layers and edge contours in camera-based inspection systems.

[0008] The object underlying the invention is therefore to improve the automated material strip inspection during the construction of a tire part in such a way that the inspection is carried out independently of environmental influences and boundary parameters in the production facility with the lowest possible error rate in the detection of the material layers and / or the detection of an edge profile.

[0009] The object is achieved by a testing system of the type mentioned at the outset, wherein the electronic data processing device of the testing system according to the invention is configured to evaluate the image recording generated by the image recording device to determine the property of the material strip by means of an algorithm based on machine learning.

[0010] The electronic data processing system thus has access to a machine learning-based algorithm for image analysis. Compared to an inspection system that evaluates the images captured by an image acquisition device using conventional image processing, the use of a machine learning-based algorithm can significantly reduce the error rate in automated material strip inspection, as the machine learning-based algorithm can cope with different boundary conditions in the production facilities and varying environmental influences.The influence of different lighting conditions, drum properties and / or material properties on the image properties is known to the machine learning-based algorithm through training, so that corresponding edge parameters and environmental influences do not lead to an increased error rate in the detection of material layers and / or the detection of an edge profile.

[0011] The inspection system can also have multiple image recording devices. The one or more image recording devices can be cameras, for example. In this case, the inspection system operates on a camera-based basis. To determine the property of the material strip, the electronic data processing device can also evaluate multiple images captured by the image recording device. The one or more image recording devices are preferably aligned with the tire building drum, and the recording area of ​​the image recording devices can be illuminated by an illumination device of the inspection system.

[0012] The inspection system according to the invention, thanks to the use of a machine learning-based algorithm, allows for material strip inspection with a low error rate, so that manual intervention is only required in exceptional cases. Furthermore, the use of the machine learning-based algorithm does not significantly increase cycle times.

[0013] In a preferred embodiment of the testing system according to the invention, the algorithm used by the electronic data processing device to determine the property of the material strip is an algorithm based on an artificial neural network. The machine learning-based algorithm used by the electronic data processing device is therefore based on an evaluation model artificially generated through a learning process. The machine learning-based algorithm is therefore a deep learning algorithm.

[0014] According to the invention, the algorithm used by the electronic data processing device to determine the property of the material strip comprises an edge segmentation algorithm based on machine learning. According to the invention, the electronic data processing device is configured to determine the edge profile of the material strip positioned on the tire building drum using the edge segmentation algorithm. The edge segmentation algorithm is preferably used to detect lateral longitudinal edges of the material strip. The data processing device can further be configured to detect the width and / or geometry of the material strip and / or to detect the position and / or orientation of the material strip on the tire building drum based on the detected edge profile.

[0015] In another preferred embodiment of the testing system according to the invention, the edge segmentation algorithm is an algorithm trained on the basis of training images depicting material strip edges and provided with an edge marker. The training images with which the edge segmentation algorithm is trained preferably show edge contours under different boundary conditions and / or environmental influences. The training images therefore show, for example, edge contours under different lighting conditions and / or different material properties of the material strips. The edge markers of the training images can have been set manually, so that the edge segmentation algorithm could be trained using the manually set edge markers. The training images can be real images or have been artificially generated.The artificially generated images may have been created by manipulating the real images. The artificially generated images artificially increase the dataset for machine learning, i.e., for training the edge segmentation algorithm.

[0016] In another preferred embodiment of the testing system according to the invention, the edge segmentation algorithm is an algorithm validated based on validation images depicting material strip edges and provided with an edge marker. The algorithm used by the electronic data processing device has thus undergone a validation process, through which the trained segmentation routines have been validated. Alternatively or additionally, the edge segmentation algorithm can be an algorithm tested for edge detection accuracy based on test images depicting material strip edges and provided with an edge marker.By testing the edge segmentation algorithm for its edge detection accuracy, an accuracy value can be assigned to the edge segmentation algorithm, for example, which can be used to describe the edge detection accuracy. The validation images and / or the test images show edge contours under different boundary conditions. The validation images and / or the test images can again be real images or artificially generated images. The artificially generated validation images and / or test images can be generated by manipulating the real images. The artificially generated validation images and / or test images increase the data volume for validation and accuracy detection.

[0017] In a further preferred embodiment of the testing system according to the invention, the algorithm used by the electronic data processing device to detect the properties of the material strip comprises an edge classification algorithm based on machine learning. Preferably, the electronic data processing device is configured to determine potential edge profiles of the material strip positioned on the tire building drum by means of image analysis and to classify the potential edge profiles to identify a specific edge profile using the edge classification algorithm. Therefore, if an edge classification algorithm is used, a potential edge profile is first detected by the electronic data processing device.However, a potential edge gradient may only give the visual impression of an edge gradient, for example, due to light reflections on the surface of the material strip, but this is not a real material edge. The edge classification algorithm detects real edge gradients.

[0018] According to the invention, the algorithm used by the electronic data processing device to determine the properties of the material strip comprises a stripe type detection algorithm based on machine learning. According to the invention, the electronic data processing device is configured to determine the stripe type of the material strip positioned on the tire building drum using the stripe type detection algorithm. The stripe type detection algorithm is preferably a classification algorithm. The electronic data processing device is preferably configured to classify the material strip to identify a specific stripe type using the stripe type detection algorithm.For example, the stripe type detection algorithm is configured to identify one, several, or all of the following stripe types: inner layer of a tire carcass, textile cord ply of a tire carcass, turnup of a tire carcass, side strip of a tire carcass, steel cord belt ply of a belt package, spool bandage of a belt package, and / or tread of a belt package.

[0019] The object underlying the invention is further achieved by a tire building machine of the type mentioned above, wherein the testing system of the tire building machine according to the invention is designed according to one of the embodiments described above. Regarding the advantages and modifications of the tire building machine according to the invention, reference is made to the advantages and modifications of the testing system according to the invention.

[0020] The object underlying the invention is further achieved by a method of the type mentioned above, wherein the electronic data processing device evaluates the image generated by the image recording device to determine the property of the material strip using an algorithm based on machine learning. The method according to the invention is preferably carried out for inspecting a material strip positioned on a tire building drum using a testing system according to one of the embodiments described above. With regard to the advantages and modifications of the method according to the invention, reference is therefore first made to the advantages and modifications of the testing system according to the invention.

[0021] In a preferred embodiment of the method according to the invention, the algorithm is generated based on an artificial neural network. The artificial neural network is preferably trained during training using a plurality of training images.

[0022] Preferred embodiments of the invention are explained and described in more detail below with reference to the accompanying drawings. Fig. 1 shows an embodiment of the tire building machine according to the invention in a schematic representation; Fig. 2 shows an edge profile detectable with the testing system according to the invention; Fig. 3 shows the training of the algorithm used by the testing system according to the invention; and Fig. 4 shows the generation of evaluable images in a schematic representation.

[0023] The Fig. 1shows a tire building machine 100 with a tire building drum 102, on which a material strip M is positioned for building a tire part, namely a tire carcass or a belt package. The tire building machine 100 further comprises a testing system 10, by means of which the material strip M positioned on the drum surface 104 of the tire building drum 102 can be inspected. Furthermore, the testing system 10 can also be used to inspect additional material strips that are placed on the illustrated material strip M during the subsequent tire building process.

[0024] If the tire building machine 100 is a machine used to build the tire carcass, the inner layer, the textile cord ply, the turn-up, and / or the side strips can be inspected using the inspection system 10. If the tire building machine 100 is a machine for building a belt package, the steel cord belt plies, the spool bandage, and / or the tread can be inspected using the inspection system 10.

[0025] The inspection system 10 has at least one, in this case two, image recording devices 12a, 12b designed as cameras, by means of which images of the material strip M positioned on the tire building drum 102 can be generated. The recording area of ​​the image recording devices 12a, 12b is illuminated by lighting devices 16a, 16b of the inspection system 10.

[0026] Furthermore, the inspection system 10 comprises an electronic data processing device 14, by means of which, based on the images generated by the image recording devices 12a, 12b, properties of the material strip M positioned on the tire building drum 102 can be determined. For example, the electronic data processing device 14 can record the strip type of the material strip M and the edge contours K of the material strip M. Via the edge contour detection, the electronic data processing device 14 can further determine the material strip width and the orientation of the material strip M on the tire building drum 102.

[0027] The electronic data processing device 14 is configured to evaluate the images generated by the image recording devices 12a, 12b to determine the stripe type and the edge profiles K of the material strip M using a machine learning-based algorithm A. The algorithm A used by the electronic data processing device 14 is an algorithm A based on an artificial neural network. The algorithm A used is therefore based on training data from which the algorithm A has derived patterns and regularities for stripe type recognition and edge profile recognition.

[0028] The algorithm A used comprises a machine learning-based edge segmentation algorithm, wherein the data processing device 14 is configured to determine the edge profile of the material strip M positioned on the tire building drum 102 using the edge segmentation algorithm. The algorithm used by the electronic data processing device 14 can further comprise a machine learning-based edge classification algorithm. In this case, potential edge profiles of the material strip M positioned on the tire building drum 102 are first determined using image analysis, before the electronic data processing device 14 uses the edge classification algorithm to classify the potential edge profiles to identify a desired edge profile using the edge classification algorithm.

[0029] Furthermore, algorithm A used by the electronic data processing device 14 comprises a stripe type detection algorithm based on machine learning, wherein the electronic data processing device 14 is configured to determine the stripe type of the material strip M positioned on the tire building drum 102 using the stripe type detection algorithm. The electronic data processing device 14 uses a machine learning-based algorithm A because conventional image processing is not sufficiently tolerant to environmental influences and varying boundary parameters, such as changing lighting conditions.Light reflections, the material properties and / or the drum properties lead to a comparatively high error rate in stripe type and edge pattern detection in inspection systems based on conventional image processing, so that manual intervention is required comparatively frequently.

[0030] The Fig. 2shows two material strips M 1 , M 2 , which are positioned on a tire building drum 102 of a tire building machine 100. The material strip M 1 is located directly on the drum surface 104. The material strip M 2 is located above the material strip M 1 . Due to unfavorable lighting conditions, a light reflection R occurs on the surface of the material strip M 2. Due to the light reflection R, reliable edge detection is not possible using conventional image processing. When evaluating images of the drum area shown, there is a high risk when using conventional image analysis that the edge K' of the light reflection R is detected as a material edge.This faulty edge detection would either lead to an interruption of the automated tire building process or, if the tire building process continues based on the faulty edge tracking data, to the production of a defective tire part, for example a defective carcass or a defective belt package.

[0031] By using a machine learning-based algorithm A, the risk of incorrect edge detection is significantly reduced or even eliminated. Using algorithm A, the edge profile K of the material strip M 2 can be determined either directly or indirectly.

[0032] For the direct determination of the edge profile K, an edge segmentation algorithm and a material classification algorithm are used, which were previously trained on training images BT. The training images BT show material strip edges and are provided with edge markings, showing edge profiles under different boundary conditions. The machine learning-based algorithm A has derived patterns and regularities for direct edge detection from the training images BT.

[0033] In indirect edge detection, a potential edge profile K is first determined using conventional image processing. Using a classification algorithm, the electronic data processing device 14 then identifies the desired material edge profile K, thus avoiding incorrect edge detection.

[0034] The Fig. 3shows the conceptual idea of ​​training algorithm A. Algorithm A is based on a dataset D, where dataset D comprises images of edge profiles under different boundary conditions and environmental influences. The images of dataset D can be real images or artificially generated images. The artificially generated images can be generated by manipulating the real images, thus providing a large amount of data for machine learning. The images show material strip edges and are provided with an edge marker. The edge marker can comprise a plurality of edge presence points that were manually set for the purpose of algorithm training.

[0035] The data set D is first divided into training images BT , validation images Bv and test images B Test .

[0036] Using the training images BT, which depict the edges of the material strips and are provided with edge markings, algorithm A is trained for more precise and reliable edge detection. Using the validation images Bv, algorithm A, trained using the training images BT, is validated. The test images Btest are used to derive a characteristic value KW, where the characteristic value KW describes the edge detection accuracy of algorithm A.

[0037] The trained, validated, and tested algorithm A can then be used in an electronic data processing device 14 of a testing system 10 for detecting properties E of a material strip M positioned on a tire building drum 102. For this purpose, the electronic data processing device 14 evaluates one or more image recordings B using the machine learning-based algorithm A for determining properties E of a material strip M. The properties E can, for example, relate to the edge profile of the side edges of the material strip M.

[0038] The Fig. 4shows that the image recordings B evaluated by algorithm A can still be modified before the actual evaluation. First, an image recording B' is generated of a partial area of ​​a material strip M, wherein the material strip M is positioned on the drum surface 104 of a tire building drum 102. Before the actual evaluation by algorithm A, the image recording B' is scaled or cropped in order to provide algorithm A with an image recording in a specified image format and / or with a specified image resolution. List of reference symbols

[0039] 10Test system 12a, 12bImage recording devices 14Data processing device 16a, 16bLighting devices 100Tire building machine 102Tire building drum 104Drum surface AAlgorithm B, B'Image acquisition BT Training images BvValidation images B Test Test images DData set EProperty M, M 1 , M 2 Material strips KKedge profile K'potential edge profile KWCharacteristic value RLight reflection

Claims

1. Checking system (10) for checking a material strip (M, M1, M2) positioned on a tyre-building drum (102), comprising - an image-recording device (12a, 12b), which is designed to generate at least one image recording (B, B') of a material strip (M, M1, M2) positioned on a tyre-building drum (102); and - an electronic data-processing device (14), which is designed to determine, on the basis of the image recording (B, B') generated by the image-recording device (12a, 12b), at least one property (E) of the material strip (M, M1, M2) positioned on the tyre-building drum (102); characterized in that the electronic data-processing device (14) is designed to evaluate the image recording (B, B') generated by the image-recording device (12a, 12b) for determining the property (E) of the material strip (M, M1, M2) by means of an algorithm (A) based on machine learning, wherein the algorithm (A) used by the electronic data-processing device (14) for determining the property (E) of the material strip (M, M1, M2) comprises an edge segmentation algorithm based on machine learning and the electronic data-processing device (14) is designed to determine the edge profile (K) of the material strip (M, M1, M2) positioned on the tyre-building drum (102) by means of the edge segmentation algorithm, and wherein the algorithm (A) used by the electronic data-processing device (14) for determining the property (E) of the material strip (M, M1, M2) comprises a strip-type detection algorithm based on machine learning and the electronic data-processing device (14) is designed to determine the strip type of the material strip (M, M1, M2) positioned on the tyre-building drum (102) by means of the strip-type detection algorithm.

2. Checking system (10) according to Claim 1, characterized in that the algorithm (A) used by the electronic data-processing device (14) for determining the property (E) of the material strip (M, M1, M2) is an algorithm (A) based on an artificial neural network.

3. Checking system (10) according to either of Claims 1 and 2, characterized in that the edge segmentation algorithm is an algorithm (A) trained on the basis of training image recordings (BT) reproducing material strip edges and provided with an edge marking.

4. Checking system (10) according to one of Claims 1 to 3, characterized in that the edge segmentation algorithm - is an algorithm (A) validated on the basis of validation image recordings (Bv) reproducing material strip edges and provided with an edge marking; and / or - is an algorithm (A) tested with regard to its edge-sensing accuracy on the basis of test image recordings (BTest) reproducing material strip edges and provided with an edge marking.

5. Checking system (10) according to one of the preceding claims, characterized in that the algorithm (A) used by the electronic data-processing device (14) for determining the property (E) of the material strip (M, M1, M2) comprises an edge classification algorithm based on machine learning and the electronic data-processing device (14) is designed to determine potential edge profiles of the material strip (M, M1, M2) positioned on the tyre-building drum (102) by means of image evaluation and to classify the potential edge profiles (K) for the identification of an edge profile (K) belonging to the material by means of the edge classification algorithm.

6. Tyre-building machine (100), with - a tyre-building drum (102), on which material strips (M, M1, M2) for building a tyre part, in particular for building a tyre carcass or a breaker belt assembly, can be positioned; and - a checking system (10) for checking the material strips (M, M1, M2) positioned on the tyre-building drum (102); characterized in that the checking system (10) is formed according to one of the preceding claims.

7. Method for checking a material strip (M, M1, M2) positioned on a tyre-building drum (102) of a tyre-building machine (100) by means of a checking system (10), according to one of Claims 1 to 5, with the steps of: - generating at least one image recording (B, B') of a material strip (M, M1, M2) positioned on a tyre-building drum (102) by means of an image-recording device (12a, 12b) of the checking system (10); and - determining at least one property (E) of the material strip (M, M1, M2) positioned on the tyre-building drum (102) on the basis of the image recording (B, B') generated by the image-recording device (12a, 12b) by means of an electronic data-processing device (14) of the checking system (10); characterized in that the electronic data-processing device (14) evaluates the image recording (B, B') generated by the image-recording device (12a, 12b) for determining the property (E) of the material strip (M, M1, M2) by means of an algorithm (A) based on machine learning.

8. Method according to Claim 7, characterized by the step of: - generating the algorithm (A) on the basis of an artificial neural network; wherein the algorithm (A) is trained during the generating, preferably by means of a multiplicity of training image recordings (BT).