EP4479595A1

EP4479595A1 - Teeth wear monitoring system for bucket wheel excavators

Info

Publication number: EP4479595A1
Application number: EP22720454.2A
Authority: EP
Inventors: Pradeep PRAJAPATI; Rohit Gupta; Stefan Ebert; Awadesh Kumar Tiwari; Wat KSHITIJ; Tushar Vaidya; Guido TRIEM
Original assignee: Koch Solutions GmbH
Current assignee: Koch Solutions GmbH
Priority date: 2022-02-17
Filing date: 2022-04-05
Publication date: 2024-12-25
Also published as: WO2023156027A1; AU2022441613A1

Abstract

The invention relates to a monitoring system for monitoring the teeth of a bucket wheel excavator.

Description

Teeth wear monitoring system for bucket wheel excavators

Bucket wheel excavators usually have teeth on the buckets to loosen the material, which is then transported away by the buckets. As a result, however, these teeth are subject to a very high degree of wear, which is all the greater the harder the material removed. Therefore, these teeth are very important for the removal of the material. On the other hand, complete wear of the teeth would lead to parts of the bucket wheel coming into direct contact with the solid material to be removed, which would damage them. Therefore, it is very important to replace the teeth in time. On the other hand, replacing the teeth naturally involves a downtime and thus a loss of production, so that the replacement is associated with high costs. This means that replacement should always be carried out as late as possible.

To make it easier to replace the teeth, they are often separably connected to the bucket wheel via sleeves. Furthermore, the teeth then have an area which is designed for fastening in the sleeve and another part which digs directly into the material to be conveyed as a wear area. In order to be able to remanufacture the teeth, it is necessary to replace them before the reusable part suffers irreversible damage.

Therefore, systems are increasingly being discussed which should enable automatic monitoring of the degree of wear of the teeth. For example, from WO WO 2019 / 034 691 A1 discloses a system for determining the wear of material-removing elements on a bucket wheel.

One problem here is to automatically assign the recorded data to a specific state of wear. It is therefore the task of the invention to ensure this allocation.

This task is solved by the method with the features given in claim 1 Advantageous further embodiments are shown from the dependent claims. The invention relates to a method for monitoring wear of teeth on a bucket wheel excavator. The method comprising the steps of: a) Providing a bucket wheel with fully intact teeth, b) acquiring at least a first image and at least a first angular position of the bucket wheel, c) Determining the first bucket detected in the first image, d) cropping the first bucket in the first image, e) contour detection of the first bucket in the captured first image, f) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket, g) determining a vector for each tooth from the first base point to the tip of the tooth in the first image, h) providing a model for a virtual bucket wheel with virtual teeth, i) Fitting the model to teeth with complete remaining life, j) further fitting the model so that the model matches the vectors determined in step g), k) operating the bucket wheel, l) acquiring at least a second image and at least a second angular position of the bucket wheel, m) determining the first bucket detected in the second image, n) cropping the first bucket in the second image, o) contour detection of the first bucket in the detected second image, p) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket in the second image, q) determining a vector for each tooth from the first base point to the tip of the tooth in the second image, r) varying the model fitted in step j) with one wear as a degree of freedom per tooth to fit the vectors obtained in step q), s) outputting the wear values per tooth.

Step a) involves providing a bucket wheel with fully intact teeth, both a new bucket wheel and retrofitting on an existing bucket wheel after fitting with new teeth. The status of the fully intact teeth is required as a starting point for the procedure to perform a subsequent analysis of the wear.

Although it is possible in step b) to detect only one blade and consider it representative of all blades, it is preferable to detect all blades and thus all teeth.

Steps c) and d) restrict the image data to be analysed to the region of interest. In particular, the adjacent buckets are also masked out.

The contour detection in step e) is done, for example, by a conversion to a 2-bit image, for example 1 component of the bucket including the teeth, 0 not component of the bucket.

For example, the colour of the pixels can be used for the contour detection in step e). Furthermore, edge detection can be performed directly. This is particularly preferred if only the bucket to be examined is in focus and the depth of field of the capturing camera is selected in such a way that no sharp edge can be captured for other objects captured with the camera.

The detection of the tips of the teeth in step f) is preferably done as an identification of the point of the contour which is furthest from the bottom of the bucket for each tooth.

For the identification of the vectors in step g), the lowest point of the bucket on the underside, the side of the bucket opposite the teeth, is preferably selected as the reference point. This point is particularly suitable due to the position and usual symmetry of the bucket.

The model provided in step h) is preferably a 3D model. The 3D model preferably consists of the bucket body, the teeth and the mounting elements that attach the teeth to the bucket body. This makes it possible to take variations into account when inserting the teeth into the mounting elements. The model can also take into account different states of wear of the teeth. Furthermore, the model can also take into account deformations of the bucket body, if these are considered possible during operation. In step i), the wear of all teeth is first set to non-existent for the model, since completely intact teeth were provided in step a). Thus, a simple synchronicity between reality and model can be achieved.

In step j) the model is now fitted to the vectors determined in step g). This creates a direct relationship between the real bucket and the virtual bucket in the model. This relationship can also be used as a constant for the further steps, so that the only remaining variable is the wear of the teeth.

The operation of the bucket wheel in step k) can be regarded as a continuous process, which is also preferably not interrupted for the subsequent steps. In particular, the operation includes the mining of material, which causes the teeth to dig into the material to be mined and wear out in the process.

In step I), the first bucket with the teeth worn out by the operation in step k) is now detected in comparison with step b).

Steps m), n), o), p) and q) are analogous to steps c), d), e), f) and g) respectively. Preferably, the same algorithms are used.

In step r), only the wear is adjusted in the model, so there is only one degree of freedom per tooth. Preferably, this degree of freedom is between 100 % remaining service life and 0 % remaining service life.

In step s), the output is preferably in the form of a percentage value of the service life, particularly preferably with the information that a replacement should take place soon if the value falls below a first threshold value or the information that a replacement should take place immediately if the value falls below a second threshold value. The last two information items can also be displayed in colour, for example in a graphical output (green - no replacement necessary; yellow - replacement soon; red - replacement immediately).

In a further embodiment of the invention, the steps b) to g) are performed for each bucket. The adjustment in step j) is performed for all buckets. The steps I) to s) are performed for each bucket. In particular, a large number of images are taken and the appropriate images are assigned to the respective buckets using the determined or interpolated angular positions. By capturing all buckets and thus all teeth, the accuracy is naturally at its highest. This applies in particular because differences in density and/or hardness in the material to be extracted can lead to very different wear of the teeth.

In a further embodiment of the invention the acquisition of the first image in step b) as well as the acquisition of the second image and in step I) is taken place continuously. For example, the capture can take the form of a video from which the individual images are extracted. In this case, exactly those images are extracted that optimally depict a specific bucket due to the associated angular position.

In a further embodiment of the invention the acquisition of the first image and the first angular position in step b) and the acquisition of the second image and the second angular position in step I) are each performed separately. For example, as in the above embodiment, a video stream is received. In parallel, the control system either receives an angular position from a clinometer, whereby there is no temporal synchronicity between the two. Alternatively, the system can request the angular position from a clinometer, for example, at regular intervals and then receive the angular position as a response, preferably with time information. For the times between two pieces of angular information, the angular position is linearly extrapolated from the available information.

In an alternative embodiment, the angular positions of the individual buckets are specified and these are given as a request to a clinometer. The clinometer then triggers the acquisition of an image by a camera when the predefined angles are reached. This ensures that the images to be evaluated are always taken in the optimum position. A disadvantage is the higher complexity of the acquisition system.

In a further embodiment of the invention, the images are captured on the opposite side of the removal side of the bucket wheel. In the case of a rotational movement usual for mining, in which the buckets dig through the material from the bottom upwards, the buckets are thus captured with the teeth pointing downwards. This can make it advantageous to rotate the images before further processing so that the teeth of the buckets point upwards.

In a further embodiment of the invention the contour detection in step e) and in step o) is pixel-based. Here, the assignment of the pixels to the bucket or to the background can be done, for example, via the colour relationship. If necessary, the shape of the bucket can be used as a rough grid to determine an average colour value of the bucket. An abrupt colour change starting from the centre of the bucket thus most likely indicates the edge of the bucket and thus the outer contour.

In another embodiment of the invention, the acquisition of the first image in step b) and the acquisition of the second image in step I) is performed in such a way that the acquisition of the bucket is approximately perpendicular. In the sense of the invention, approximately perpendicular means that the imaginary line from the camera and the bucket passes through the axis of rotation of the bucket wheel or has a smallest distance to the axis of rotation of the bucket wheel, which is less than 20 %, preferably less than 10 %, of the diameter of the bucket wheel.

In a further embodiment of the invention the detection of the tips of the teeth in step f) and step p) is performed by means of a neural network. Particularly preferably, the neural network is a convolutional neural network. Particularly preferably, the neural network is trained in a supervisor-supervised manner. Particularly preferably, the neural network is trained in a supervisor-supervised manner. This is done in such a way that first an automatic recognition takes place and this is then checked by the supervisor and corrected if necessary. When sufficient recognition accuracy is achieved, i.e. no or only very little need for correction, the training is terminated. Preferred the convolutional neural network is a region based convolutional neural network, especially fast convolutional neural network.

In a further embodiment of the invention the further adjustment in step j) comprises in particular arranging the model in space for conversion into a two-dimensional image captured by a camera. This step ensures an exact alignment and distance between the camera and the blade to create a virtual two-dimensional image corresponding to a camera shot. This geometric relationship will also be the same for all subsequent captured images, so this adjustment only needs to be made the first time.

In a further embodiment of the invention the further adjustment in step j) comprises adjusting the position of the teeth relative to the bucket. This step primarily takes into account a variation of the teeth during assembly, but above all also corresponds in the later analogue step r) due to wear.

In a further embodiment of the invention steps I) to s) are performed and repeated during step k). This embodiment is particularly preferred. This allows the teeth to be monitored continuously or at predefined intervals during regular operation. This also has the great advantage that a relatively large number of data points relating to the closing of the teeth are recorded, which in turn enables the remaining time to be predicted. In addition, this also allows the wear to be recorded very precisely in terms of time, whereby, for example, inclusions in the material to be removed with a higher hardness, which leads to more abrasion, can be easily detected. As a result, such wear events can easily be calculated out of an averaged wear progression. It is also possible to detect a high current wear and thus shorten the prognosis compared to the time average and thus minimise the risk of damage to the bucket wheel due to completely worn teeth.

In a further embodiment of the invention the model in step h) is a 3D model.

In a further embodiment of the invention a first wear value and a second wear value are predetermined. The first wear value recommends a replacement shortly for example, during the next inspection or relocation of the bucket wheel to another position or another planned interruption. This means that no additional interruption of the mining operation is necessary. The second wear value recommends an immediate replacement. Of course, this involves an interruption of the operation and thus a loss of operating time, but this is necessary to protect the paddle wheel itself and must therefore be accepted. For each tooth the wear determined in step r) is compared with the first wear value and the second wear value before step s). Since the wear is different for each tooth, each must be considered separately. An averaged consideration only makes sense insofar as there may be an increased urgency for replacement when a certain number of teeth exceed the first wear value. Likewise, a conspicuousness in the local frequency can also be taken into account. For example, if all left teeth are particularly heavily worn, a corresponding warning can be given, as this can indicate a systematic problem. In step s) an exceeding of the first wear value or of the second wear value is also output.

In a further embodiment of the invention in step I) a plurality of second images is acquired. The acquired second images are compared with a last stored and last evaluated image of the same bucket. The second image with the smallest deviation from the last stored image is selected and used for steps m) to q). During normal operation of the bucket wheel, the buckets will constantly show contamination, whereby the contamination usually changes with each rotation. If, for example, 20 images of the same bucket are captured during 20 consecutive rotations, the dirt build-up will most likely be different in each image. Assuming that the first image in step f) with new teeth captures a clean bucket, the next image will again have the highest similarity with at least the dirt. In the following check, the bucket with the least dirt will again show the best match, and so on.

In the following, the method according to the invention is illustrated by example with reference to the figures.

Fig. 1 Bucket wheel excavator

Fig. 2 Flow chart

Fig. 1 shows a bucket wheel 10 of a bucket wheel excavator for carrying out the method according to the invention. The bucket wheel 10 usually has a fixed inner area 20 on which the bucket area 30 rotates. The bucket area has a plurality of buckets with teeth not shown here. The removed material 60 is transferred via a chute 80 onto a conveyor belt 70 and transported away. In addition, the bucket wheel 10 has a camera 40 that captures images of the buckets 32.

Fig. 2 shows the flow chart of the method according to the invention. The method has the following steps: a) Providing a bucket wheel with fully intact teeth, b) acquiring at least a first image and at least a first angular position of the bucket wheel, c) Determining the first bucket detected in the first image, d) cropping the first bucket in the first image, e) contour detection of the first bucket in the captured first image, f) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket, g) determining a vector for each tooth from the first base point to the tip of the tooth in the first image, h) providing a model for a virtual bucket wheel with virtual teeth, i) Fitting the model to teeth with complete remaining life, j) further fitting the model so that the model matches the vectors determined in step g), k) operating the bucket wheel, l) acquiring at least a second image and at least a second angular position of the bucket wheel, m) determining the first bucket detected in the second image, n) cropping the first bucket in the second image, o) contour detection of the first bucket in the detected second image, p) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket in the second image, q) determining a vector for each tooth from the first base point to the tip of the tooth in the second image, r) varying the model fitted in step j) with one wear as a degree of freedom per tooth to fit the vectors obtained in step q), s) outputting the wear values per tooth.

As shown, the steps I) to s) are repeated continuously during to operation in step k).

Reference list

10 bucket wheel

20 inner area

30 bucket area 32 bucket

40 camera

60 excavation material

70 conveyor belt 80 chute

Claims

1 . A method of detecting wear on teeth of a bucket wheel excavator, the method comprising the steps of: a) Providing a bucket wheel with fully intact teeth, b) acquiring at least a first image and at least a first angular position of the bucket wheel, c) Determining the first bucket detected in the first image, d) cropping the first bucket in the first image, e) contour detection of the first bucket in the captured first image, f) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket, g) determining a vector for each tooth from the first base point to the tip of the tooth in the first image, h) providing a model for a virtual bucket wheel with virtual teeth, i) Fitting the model to teeth with complete remaining life, j) further fitting the model so that the model matches the vectors determined in step g), k) operating the bucket wheel, l) acquiring at least a second image and at least a second angular position of the bucket wheel, m) determining the first bucket detected in the second image, n) cropping the first bucket in the second image, o) contour detection of the first bucket in the detected second image, p) detecting the tips of the teeth of the first bucket and detecting a first base point of the first bucket in the second image, q) determining a vector for each tooth from the first base point to the tip of the tooth in the second image, r) varying the model fitted in step j) with one wear as a degree of freedom per tooth to fit the vectors obtained in step q), s) outputting the wear values per tooth.

2. Method according to claim 1 , wherein steps b) to g) are performed for each bucket, wherein the adjustment in step j) is performed for all buckets, wherein steps I) to s) are performed for each bucket.

3. Method according to any one of the preceding claims, wherein the acquisition of the first image in step b) as well as the acquisition of the second image and in step I) each take place continuously.

4. Method according to any one of the preceding claims, wherein the acquisition of the first image and the first angular position in step b) and the acquisition of the second image and the second angular position in step I) are each performed separately.

5. Method according to any one of the preceding claims, wherein the contour detection in step e) and in step o) is pixel-based.

6. Method according to any one of the preceding claims, wherein the detection of the tips of the teeth in step f) and step p) is performed by means of a neural network.

7. Method according to any one of the preceding claims, wherein the further adjustment in step j) comprises in particular arranging the model in space for conversion into a two-dimensional image captured by a camera.

8. Method according to any one of the preceding claims, wherein the further adjustment in step j) comprises adjusting the position of the teeth relative to the bucket.

9. Method according to any one of the preceding claims, wherein steps I) to s) are performed and repeated during step k).

10. Method according to any one of the preceding claims, wherein the model in step h) is a 3D model.

11 . Method according to any one of the preceding claims, wherein a first wear value and a second wear value are predetermined, wherein the first wear value recommends a replacement shortly, wherein the second wear value recommends an immediate replacement, wherein for each tooth the wear determined in step r) is compared with the first wear value and the second wear value before step s), wherein in step s) an exceeding of the first wear value or of the second wear value is also output. Method according to any one of the preceding claims, wherein in step I) a plurality of second images is acquired, wherein the acquired second images are compared with a last stored and last evaluated image of the same bucket, wherein the second image with the smallest deviation from the last stored image is selected and used for steps m) to q).