Detecting wear on a tool

Abstract

In a method for detecting wear on a tool (2), in particular on a mill- ing tool, the tool (2) is mounted in a rotatable tool holder (8) of a detection apparatus. A plurality of partial images of the surface of the tool (2) are recorded by means of a microscope (18) while rotating the tool (2) about a rotation axis (10) of the tool holder (8). The microscope (18) comprises a camera device (22) and camera optics (20), with the camera optics (20) projecting an object plane (24) onto the camera de- vice (22). The object plane (24) extends parallel to the rotation axis (10) and inter- sects the tool (2). Each partial image contains at least a strip-shaped section (38) of the object plane (24). These the strip-shaped sections (38) are assembled into a full image (50). The full image is shear-mapped and then analyzed for damaged areas (58).

Description

[0001] Detecting wear on a tool

[0002] Technical Field

[0003] The invention relates to a method and an apparatus for detecting wear on a tool, in particular wear on a milling cutter.

[0004] Background Art

[0005] Tool wear, e.g., on milling cutters, needs to be monitored closely to ensure a good quality of the machined parts and to optimize the tool lifecycle. However, manual inspection is both expensive and time consuming. Therefore, there is a need for automated monitoring techniques.

[0006] EP3881968 describes a method for detecting tool wear by recording image data of the tool.

[0007] Disclosure of the Invention

[0008] The problem to be solved by the present invention is to provide a method and apparatus of this type that allow an accurate, quick image -based detection of tool wear.

[0009] This problem is solved by the method and device of claim 1.

[0010] Accordingly, the method for detecting wear on a tool, in particular on a milling tool, comprises at least the following steps:

[0011] - Mounting the tool in a rotatable tool holder.

[0012] - Recording, by means of a microscope of the detection apparatus, a plurality of partial images of a surface of the tool while rotating the tool about a rotation axis of the tool holder. The microscope comprises a camera device and camera optics, with the camera optics projecting an object plane onto the camera device. The object plane intersects the tool or is tangential (within the depth of field of the camera optics) to the tool, i.e., the camera is able to record a well-focused image at least at locations where the object plane intersects / touches the tool. Each partial image contains at least a strip-shaped section of the object plane.

[0013] - Assembling the strip-shaped sections from the partial images into a full image. Each such section extends along a length L over the full image and over a width W perpendicularly thereto. The sections are strip-shaped in the sense that L > 50-W, in particular L > 100-W.

[0014] This technique is based on the understanding that assembling a full image of such strip-shaped sections allows obtain a well-focused representation of the curved outer surface region of the tool, which then can be processed for recognizing damage.

[0015] The tool may be illuminated by a light source while recording the partial images.

[0016] In this case, the light source may be adapted to generate a stripshaped illumination field covering said strip-shaped section of the object plane in order to concentrate the light on this section. Hence, advantageously, in the object plane the illumination field may have, along the length (i.e., the longitudinal extension) of the strip-shaped section, an extension A and, along a direction perpendicular to the direction X, an extension B, with A > 5-B, in particular A > 50 B.

[0017] In one embodiment, the strip-shaped section is laterally offset in respect to the rotation axis. This lateral offset is geometrically defined as follows: We assume that the camera optics projects the object plane along a projection direction onto the camera device. In this case, the projection of the rotation axis along the projection direction onto the object plane is parallel to but offset from the elongate center axis of the strip-shaped section.

[0018] Such a lateral offset allows to observe both lateral surfaces of an edge of the tool in the strip-shaped section of the object plane, which allows to gain a better image of the edge and any damages thereon.

[0019] Various types of image processing can be used to detect tool damage in the full image. However, a multi-step approach may be used for particularly efficient processing.

[0020] This approach starts from the full image, where we define directions u and v as orthogonal directions, with direction v extending parallel to the longitudinal axes of the strip-shaped sections.

[0021] In such an image, most conventional tools will have edges that extend transversally but not orthogonally nor parallel to direction v.

[0022] The multi-step approach comprises at least the following steps:

[0023] - Automatically detecting, in the full image, a set of parallel tool edges extending transversally and non-perpendicularly to the direction v.

[0024] - Applying a shear mapping to the full image, thereby generating a transformed image where the tool edges extend parallel to the direction v. As a result of this step, the shear edges become parallel independently of the pitch of the edges of the tool, which simplifies the subsequent steps.

[0025] - Automatically generating edge images from the transformed image, with each edge image comprising part of the transformed image including only one tool edge.

[0026] - Feeding individual ones of the edge images to a classifier for identifying damage areas of the tool.

[0027] The last two steps allow to break down the task of finding damages in all of the edges of the whole transformed image to a series of smaller tasks identifying damages in a single edge, which allows to use a simpler, faster, and / or more reliable classifier.

[0028] The tool may have a curved, non-cylindrical surface, and not all of its parts may be recorded, in the partial images, with perfect focus. To alleviate this problem, the method may further comprise the following steps:

[0029] - Recording a first set of the partial images for a first distance between the microscope and the rotation axis.

[0030] - Recording a second set of the partial images for a second distance between the microscope and the rotation axis, wherein the second distance is different from the first distance.

[0031] - Combining partial images from the first set and the second set to obtain a full image of better focus than the full image from the first set and the second set.

[0032] This sequence of steps allows to use, in each combination, the better focused parts of the two sets of partial images to be combined, thereby obtaining a combined full image with improved focus.

[0033] The scheme can be further improved by recording additional sets of partial images for additional distances between the microscope and the rotation axis, with these additional distances being different form the first and second distances, and then combining three partial images from the different sets to obtain a combined partial image with even better focus, etc.

[0034] For superior measurements, the camera optics is at least object-side telecentric. This is particularly useful if the two or more sets of partial images are recorded for different distances between the microscope and the tool holder as described above because it makes it easier to combine the partial images of the different sets.

[0035] The object plane is advantageously arranged to extend parallel to the rotation axis, in particular when the tool to be analyzed has edges running along a cylinder surface when then tool is rotated about the rotation axis. The invention also relates to a detection apparatus for detecting wear on a tool, in particular on a milling tool. This apparatus comprises at least the following elements:

[0036] - A rotatable tool holder. This holder is adapted to hold the tool to be examined and to rotate it.

[0037] - A camera device and camera optics, with the camera optics projecting an object plane onto the camera device. The object plane extends parallel to the rotation axis and is located to intersect the tool when the tool mounted to the rotatable tool holder.

[0038] - A control unit adapted and structured to execute the steps of the method as described herein.

[0039] The invention also relates to a computer program product comprising instructions to cause this type of apparatus to execute the steps of the method as described herein.

[0040] Brief Description of the Drawings

[0041] The invention will be better understood and objects other than those set forth above will become apparent when consideration is given to the following detailed description thereof. Such description makes reference to the annexed drawings, wherein:

[0042] Fig. 1 shows components of an embodiment of a detection apparatus,

[0043] Fig. 2 shows the tool, the microscope, and the light source when viewed along the axis of rotation,

[0044] Fig. 3 illustrates the geometric relationship of some of the elements of the device,

[0045] Fig. 4 is an example of a part of an assembled full image,

[0046] Fig. 5 is an example of a part of a transformed image,

[0047] Fig. 6 is the transformed image with two detected edge regions,

[0048] Fig. 7 shows an edge image as extracted from the transformed image,

[0049] Fig. 8 shows the edge image with the classified damage areas,

[0050] Fig. 9 shows a reassembled full image with the classified damage areas, Fig. 10 is an example of an edge image with a manually marked damage area as used for training the classifier, and

[0051] Fig. 11 is an example of a transformed image with manually marked edge regions for training the edge detector.

[0052] Note: Grayscale images are dithered for black-and-white reproduction.

[0053] Modes for Carrying Out the Invention

[0054] Definitions

[0055] The term "transversal" as used herein is to be understood as "nonparallel". For example, two directions extend transversally in respect to each other if they are non-parallel.

[0056] Device

[0057] Fig. 1 shows some of the components of a detection apparatus for detecting wear on a tool 2.

[0058] The tool 2 as shown is a milling cutter having helical teeth 4 with edges 5 that, when the tool 2 is rotated about its longitudinal axis, run along a cylinder 6 shown in dashed lines in Fig. 1.

[0059] For examination, tool 2 is mounted in a rotatable tool holder 8 with its longitudinal axis coinciding with a rotation axis 10 of tool holder 8.

[0060] Tool holder 8 comprises a chuck 12 for holding the tool and a rotational drive 14 for rotating chuck 12 about rotation axis 10.

[0061] In the shown embodiment, tool holder 8 further comprises a linear displacement drive 16 for displacing tool 2 along a projection 26 direction as described below.

[0062] The shown apparatus further comprises a microscope 18 having camera optics 20 and a camera device 22, with camera optics 20 projecting an object plane 24 along a projection direction 26 onto camera device 22, i.e., an image of object plane 24 is generated at the plane of camera device 22.

[0063] Microscope 18 is positioned such that object plane 24 intersects cylinder 6.

[0064] The shown apparatus further comprises a light source 28 that generates a strip-shaped illumination field. The device further comprises a control unit 29 controlling the operation of the drives 14, 16, of camera device 22, and of light source 28. Control unit 29 is programmable and runs a computer program comprising the instructions to operate the detection apparatus as described herein.

[0065] The geometrical arrangement of the components of the apparatus is now discussed with reference to Figs. 1 to 3.

[0066] As can be seen, microscope 18 is arranged such that its central viewing axis 30, which extends along projection direction 26 and intersects a center of the pixel array 35 of a camera device 22, intersects object plane 24 at a viewing line 34, with viewing line 34 being the line where object plane 24 intersects cylinder 6.

[0067] Camera optics 20 projects viewing line 34 onto the pixels of the pixel array 35 of camera device 22.

[0068] As described below, only strip-shaped sections of the partial images recorded by camera device 22 are used when assembling a full image of tool 2. Hence, advantageously, pixel array 35 of camera device 22 is a one-dimensional pixel array only because this allows faster read-out and processing of the image data.

[0069] Light source 28 is arranged to generate a strip-shaped illumination field 36 of object plane 24 at the location of viewing line 34 (see Fig. 3). Along the direction of rotation axis 10, illumination field 36 has a full-width-half-maximum (FWHM) extension A, and perpendicularly thereto it has a FWHM extension B, with A > 5-B, in particular with A > 50 B.

[0070] As mentioned, camera device 22 and camera optics 20 are arranged to take an image of at least a strip-shaped section 38 of object plane 24. In Fig. 3, this strip-shaped section 38 is, for simplicity, shown to coincide with illumination field 36, but it is typically narrower than the FWHM-width of illumination field 36 in order to have it homogeneously illuminated.

[0071] Strip-shaped section 38 has, along the direction of rotation axis 10, a length L and, perpendicularly thereto, a width W, again with L > 50-W, in particular with L > 100-W.

[0072] The center axis 40 of strip-shaped section 38 along its longitudinal axis advantageously coincides with viewing line 34, i.e., with the intersecting line of cylinder 10 and object plane 24.

[0073] When seen along projection direction 26, center axis 40 is laterally offset from rotation axis 10. In other words, the projection 42 of rotation axis 10 along projection direction 26 onto object plane 24 is parallel to but offset, by an offset width M, from center axis 40. In this way, camera device 22 is able to take images of both surfaces adjacent to an edge 5 of tool 2.

[0074] Advantageously, the offset M is between 0.50-R and 0.98-R, with R being the radius of cylinder 6, in particular between 0.64-R and 0.94-R.

[0075] In other words, the angle a (see Fig. 2) between projection direction 26 and a connecting line 44 that runs perpendicularly between center axis 40 and rotation axis 10 is between 30° and 80°, in particular between 40° and 70°. Again, this geometry allows a good view of both lateral surfaces of a tool edge, assuming that one of them extends in substantially tangential direction and the other extends roughly perpendicularly thereto.

[0076] Light source 28 may be placed to provide a good illumination of the edges at center axis 40. If we assume, as shown in Fig. 2, that light source casts 28 light along a central illumination direction 46 onto object plane 24, the angle 0 between projection direction 26 and illumination direction 46 is advantageously smaller than or equal to the angle a, which allows to illuminate both lateral surfaces of a tool edge 5.

[0077] Further, when seen along projection direction 26, light source 28 is best placed opposite to rotation axis 10 and closer to microscope 18 than the rotation axis 10. In other words, and as shown in Fig. 3, if we define a projection plane 48 as being the plane extending parallel to projection direction 26 with center axis 40 lying in said projection plane 48, then:

[0078] - light source 28 and rotation axis 10 lie on opposite sides of projection plane 48 and

[0079] - light source 28 and rotation axis 10 lie on opposite sides of object plane 24.

[0080] In this case, the edges of tool 2 are turned towards light source 28 as they are being recorded in strip-shaped section 38 of object plane 24, which makes it easier to recognize defects.

[0081] In must be noted, though, that a may also be 0°, i.e., viewing axis 30 intersects rotation axis 10. In this case, object plane 24 may, for good focus, be arranged tangentially to cylinder 6, i.e., tool 2, within an accuracy of the depth of field of camera optics 20.

[0082] Operation

[0083] This section describes how the present apparatus is used to detect defects on tool 2, in particular along the edges 5 of tool 2. In a first step, tool 2 is mounted in tool holder 8 with its longitudinal axis coinciding with rotation axis 10.

[0084] Next, while illuminating tool 2 with light source 28 and rotating tool 2 about rotation axis 10, a series of partial images of the tool surface are recorded. Each such partial image contains at least strip-shaped section 38 of object plane 24.

[0085] If camera device 22 contains a one-dimensional array 35 of pixels only, each such partial image contains a single row of pixels. If camera device 22 contains a two-dimensional array 38 of pixels, each partial image may contain several rows of pixels.

[0086] The strip-shaped sections 38 of the partial images advantageously represent all of the outer surface (i.e., the surface facing cylinder 6) of tool 2, i.e., the tool is rotated over 360° while recording the partial images.

[0087] The number of partial images taken over one 360° rotation of tool 2 may be fairly large for a good resolution of the full image mentioned above. In particular, it may be at least 100, in particular at least 1000.

[0088] Next, the strip-shaped sections 38 of the partial images are assembled into a full image by placing them side by side. Each strip -shaped section 38 extends along its length L over the whole full image. An example of such a full image 50 is shown in Fig. 4.

[0089] In the following, we define u and v as orthogonal directions along the edges of full image 50, with direction v extending parallel to the longitudinal axes of the strip-shaped sections 38.

[0090] As can be seen, the edges 5 of the tool extend transversally to the directions u and v, with the angle being a function of the pitch of the tool.

[0091] In a next step, a shear mapping is applied to full image 50 in order to re-align the tool edges 5 to make them run parallel and perpendicular to the directions v and u, respectively.

[0092] To do so, the set of parallel tool edges 5, which extend transversally and non-perpendicularly to direction v, are identified, and then full image 50 is shearmapped to generate a transformed image 52 where the tool edges extend parallel to direction v. This step is performed by means of an image realigner 29a of control unit 29 as described in more detail below.

[0093] For the transformed image 52, the orthogonal directions u, v are again defined as directions extending parallel to the edges of the transformed image.

[0094] In a next step, edge images 54 as shown in Fig. 7 are generated from the transformed image 52, with each edge image 54 including only one tool edge 5 extending along a defined location in the edge image. As mentioned above, this allows to simplify the subsequent steps.

[0095] To generate the edge images 54, the areas 56 of the edges 5 in transformed image 52 are detected, see Fig. 6. Advantageously, these areas are assumed to be rectangular. This task can be performed by means of an edge detector 29b of control unit 29 as described in more detail below.

[0096] In a next step, each edge image 54 is fed to a classifier 29c of control unit 29 for identifying damage areas 58 of tool 5 in the edge image. This classifier 29c is described in more detail below. Since each edge image 54 contains but a single edge 5 in a defined location and orientation, classification is made easier.

[0097] Fig. 8 shows an edge image 54 with the classified damage areas 58.

[0098] The damage areas 58 may then be used as an indicator of the state of tool 2. For example, a quality score of the tool may be calculated based on the total area of the damaged areas 58 and / or the count of damage areas 58. Such a damage score may also be derived with a neural network that has been trained with images containing manually scored damage.

[0099] In addition or alternatively, the quality score may also be linked with a threshold value that determines if the tool can be used further. The quality score may be differently weighted for normal wear (signs of damage along the edges of the tool) and other damage, such as chipping.

[0100] In a further step, the edge images 54 with the marked damage areas 58 as shown in Fig. 8 may be re-inserted into the transformed image 52 (or mapped back into full image 50) in order to generate a transformed image 52' (or full image) as shown in Fig. 9, which shows the whole lateral surface of tool 2 with the marked damage areas 58.

[0101] Image Realigner

[0102] As mentioned above, control unit 29 may comprise an image realigner 29a to transform full image 50 (Fig. 4) into transformed image 52 (Fig. 5).

[0103] There are various ways to implement such an image realigner.

[0104] For example, Canny edge detection may be used to detect and extract edges in full image 50. Next, a Hough transform may be used to determine the slope of the tool's edges 5 in respect to the coordinates u, v. In a next step, shear mapping can be applied to full image 50, with the shear angle being calculated from the angle of the slope. In this shear mapping step, the lines of the image are offset (along a horizontal direction in Figs. 4, 5), and the parts of the offset lines that extend over one side of the image are moved to the other side of the image.

[0105] In an alternative embodiment, cross-correlations may be calculated between different lines of the image, thereby detecting an average inter-line offset between the edges in neighboring lines, which again allows to calculate the slope of the edges in the coordinates u, v.

[0106] Edge detector

[0107] As mentioned above, control unit 29 may comprise an edge detector 29b to detect the areas 56 of the edges in the transformed image (cf. Fig. 6).

[0108] Such an edge detector may, e.g., be implemented as a trained neural network, such as a " faster-RCNN-ResnetlOl" (see, e.g., He et al., "Deep Residual Learning for Image Recognition", arXiv:1512.03385vl).

[0109] The network may be trained by means of transformed images where the areas 56 of the edges 5 are marked manually, such as shown in Fig. 11.

[0110] In an alternative embodiment, the edges 5 may, e.g., be detected by image convolution techniques, Hough transforms, and / or Fourier transform.

[0111] Classifier.

[0112] As mentioned above, control unit 29 may comprise a classifier 29c to detect damage areas 58 in the edge images 54 as shown in Fig. 8.

[0113] Such a classifier may, e.g., be implemented as a trained neural network, such as a " Resnetl52" (see, e.g., He et al., "Deep Residual Learning for Image Recognition", arXiv: 1512.03385vl).

[0114] The network may be trained by means of partial images where damaged areas 58 are marked manually, such as shown in Fig. 10.

[0115] In an alternative embodiment, the damaged areas 58 may, e.g., be detected by means of edge detection algorithms and / or any of the techniques mentioned above. Further methods are, e.g., described by R. Schmitt et al., Machine Vision Systems for Inspecting Flank Wear on Cutting Tools, ACEEE Int. J. on Control System and Instrumentation, Vol. 03, No. 01, pp. 27 - 31 (DOI: 01.IJCSI.03.01.13). Improving depth of focus

[0116] For good resolution, camera optics 20 may have a small depth of focus, i.e., if parts of the surface of tool 2 lie outside object plane 24, their image on camera device 22 may not be in good focus.

[0117] Hence, several sets of the partial images may be recorded for different distances between microscope 18 and tool holder 8.

[0118] To do so, a first set of the partial images is recorded for a first distance DI (Fig. 2) between microscope 18 and rotation axis 10. This first set of partial images may include a full 360° scan of the outer surface of tool 2.

[0119] Then, the distance between microscope 18 and rotation axis 10 is changed by means of displacement drive 16, to a second distance D2, e.g., with D2 f DI, and a second set of the partial images is recorded. This second set of partial images may again include a full 360° scan of the outer surface of tool 2.

[0120] The two sets are assembled into a first and second full image, and the two full images are then combined to obtain an image having better focus than full images that were obtained from the first and / or the second set of partial images alone.

[0121] This may, e.g., be implemented by combining the first set of partial images into a first full image and combining the second set of partial images into a second full image. Subsequently, the first and second full image are combined by means of a focus stacking algorithm, see, e.g., https: / / en.wikipedia.org / wiki / Fo- cus_stacking and the references cited therein.

[0122] As mentioned, further sets of partial images may be recorded for further distances and then be combined for an even better depth of focus.

[0123] Camera optics

[0124] As mentioned, camera optics 20 projects at least strip-shaped section 38 of object plane 24 onto camera device 22.

[0125] Advantageously, for a quantitative assessment of the size of the damage areas 58, camera optics 20 is at least object-side telecentric. This is particularly important when recording several sets of partial images for different distances DI, D2, etc. and stacking them as described in the previous section. Notes

[0126] An important application of the present technique is to detect defects in tools 2 that are milling cutters (milling tools). Such tools have lateral edges 5 running along a cylinder 6 when rotated.

[0127] The present technique may, however, also be used for assessing damage in other types of tools 2, e.g., in in particular in cutting tools having one or more edges, such as drill bits, reamers, insert milling cutters, or grinding tools.

[0128] If the edges do not run along a cylinder surface when the tool is rotated about the rotation axis, but, e.g., along a conical surface, object plane 24 may be tilted in respect to rotation axis 10 for better aligning strip-shaped section 38 with the edges.

[0129] In the embodiments above, displacement drive 16 is adapted to displace tool 2 along projection 26 direction. In addition or alternatively thereto, it may be adapted to displace tool 2 in at least one direction perpendicular to projection direction 26.

[0130] In one embodiment, for example, displacement drive 16 may be adapted to displace tool 2 along a direction perpendicular to projection 26 direction and perpendicular to rotation axis 10, which allows to vary angle a described above.

[0131] In another embodiment, for example, displacement drive 16 may be adapted to displace tool 2 rotation axis 10, which allows to view different sections of the tool.

[0132] While there are shown and described presently preferred embodiments of the invention, it is to be distinctly understood that the invention is not limited thereto but may be otherwise variously embodied and practiced within the scope of the following claims.

Claims

Claims1. A method for detecting wear on a tool (2), in particular on a milling cutter, comprising the steps of mounting the tool (2) in a rotatable tool holder (8) of a detection apparatus, recording, by means of a microscope (18) of the detection apparatus, a plurality of partial images of a surface of the tool (2) while rotating the tool (2) about a rotation axis (10) of the tool holder (8), wherein the microscope (18) comprises a camera device (22) and camera optics (20), with the camera optics (20) projecting an object plane (24) onto the camera device (22), wherein the object plane (24) intersects the tool (2) or is tangential to the tool (2), and wherein each partial image contains at least a strip-shaped section (38) of the object plane (24), assembling the strip-shaped sections (38) from the partial images into a full image (50), wherein each strip-shaped section (38) extends along a length L over the whole full image (50) and over a width W perpendicularly thereto, with L > 50-W, in particular with L > 100-W.

2. The method of claim 1 further comprising the step of illuminating the tool (2) by means of a light source (28) while recording the partial images.

3. The method of claim 2 wherein the light source (28) generates a strip-shaped illumination field (36) covering the strip-shaped section (38) of the object plane (24), with the illumination field (36) having, along a length of the stripshaped section (38), an extension A and, along a direction perpendicular to the length of the strip-shaped section (38), an extension B, with A > 5-B, in particular with A > 50 B.

4. The method of any of the preceding claims wherein the camera optics (20) projects the object plane (24) along a projection direction (26) onto the camera device (22) wherein a projection (42) of the rotation axis (10) along the projection direction (26) onto the object plane (24) is parallel to but offset from an elongate center axis (40) of the strip-shaped section (38).

5. The method of claim 4 wherein an angle a between the projection direction (26) and a connecting line (44) connecting the center axis (40) perpendicularly to the rotation axis (10) is between 10° and 60°, in particular between 20° and 50°.

6. The method of any of claim 2 or 3 and of claim 5 wherein the light source (28) casts light along a central illumination direction (46) onto the object plane (24), and wherein an angle 0 between the projection direction (26) and the illumination direction (46) is smaller than or equal to the angle a.

7. The method of any of the claims 2 or 3 and any of the claims 4 or 5, or the method of claim 6 wherein, the center axis (40) of the strip-shaped section (38) lies in a projection plane (48) that extends parallel to the projection direction (26), the light source (28) and the rotation axis (10) lie on opposite sides of the projection plane (48) and the light source (28) and the rotation axis (10) lie on opposite sides of the object plane (24).

8. The method of any of the preceding claims wherein, in said full image (50), directions u and v are orthogonal directions, wherein direction v extends parallel to the longitudinal axes of the stripshaped sections (38), and wherein said method further comprises the steps of- automatically detecting, in the full image (50), a set of parallel tool edges (5) extending transversally and non-perpendicularly to the direction v,- applying a shear mapping to the full image (50), thereby generating a transformed image (52) where the tool edges (5) extend parallel to the direction v,- automatically generating edge images (54) from the transformed image (52), with each edge image (54) including only one tool edge (5),- feeding individual ones of the edge images (54) to a classifier (19c) for identifying damage areas (58) of the tool (2).

9. The method of any of the preceding claims wherein the method comprises the steps ofrecording a first set of the partial images for a first distance (DI) between the microscope (18) and the rotation axis (10), recording at least a second set of the partial images for a second distance (D2) between the microscope (18) and the rotation axis (10), wherein the second distance (D2) is different form the first distance (DI), and combining partial images from the first set and the second set to obtain a full image (50) with better focus than the full image (50) from the first set and the second sets alone.

10. The method of any of the preceding claims wherein the camera optics (20) is object-side telecentric.

11. The method of any of the preceding claims wherein, when rotating the tool (2) in the tool holder (8), edges (5) of the tool run along a cylinder (6), wherein the object plane (24) intersects the cylinder (6) along a viewing line (34).

12. The method of claim 11 wherein a central viewing axis (30) of the microscope (18) extending along a projection direction (26) of the microscope (18) and intersecting a center of a pixel array (35) of the camera device (22) intersects the object plane (24) at the viewing line (34), wherein the camera optics (20) projects the viewing line (24) onto pixels of said pixel array (35).

13. The method of claim 12 wherein the pixel array (35) is a onedimensional pixel array.

14. The method of any of the preceding claims wherein the object plane (24) extends parallel to the rotation axis (10).

15. The method of any of the preceding claims wherein the object plane (24) intersects the tool (2).

16. A detection apparatus for detecting wear on a tool (2), in particular on a milling tool, comprising a rotatable tool holder (8),a camera device (22) and camera optics (20), with the camera optics (20) projecting an object plane (24) onto the camera device (22), wherein the object plane (24) is located to intersect the tool (2) when mounted on the rotatable tool holder (8), and a control unit (19) adapted and structured to execute the steps of the method of any of the preceding claims.

17. The apparatus of claim 16 further comprising a displacement drive (16) is adapted to displace the tool holder (8) in respect to the camera device along at least one of a projection direction (26) of the camera device (22), a rotation axis (10) of the tool holder, and a direction perpendicular to the projection direction (26) and the rotation axis (10).

18. A computer program product comprising instructions to cause the detection apparatus of any of the claims 16 or 17 to execute the steps of any of the claims 1 to 15.

Images