METHOD FOR MACHINING AND INSPECTING OF WORKPIECES

Abstract

A method for the machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1), preferably a die sinking electrical discharge machine. The method uses at least one machining process interruption during which the processed surface of the workpiece (11) is inspected. Within said machining process interruption, at least one image of the processed workpiece surface is captured on the machine tool (1) by means of a digital camera (12). The images are processed by one or two pattern recognition algorithm (PRA D, PRA S), which were previously trained to determine the surface characteristics such as roughness parameters, functional surface features and/or characteristic defects of the processed workpiece surface captured on that at least one image. The determined surface characteristics are used to resume the processing of the workpiece surface with or without adjusting the processing parameters.

Claims

1. A method for the machining of workpieces (11) and inspection of a processed workpiece surface in a machine tool (1), the method comprising at least one machining process interruption during which the processed surface of the workpiece (11) is inspected, wherein within said machining process interruption at least one image of the processed workpiece surface is captured in situ by means of a digital camera (12), and that said at least one image is processed by a pattern recognition algorithm (PRA D, PRA S) which was previously trained by means of a predefined training dataset, whereas said predefined training dataset comprises one or more reference images of a plurality of reference surfaces, each reference image being associated to one or more known surface characteristics and/or to a known characteristic defect of a reference surface, and that at least one surface characteristic of the processed workpiece surface captured on the at least one image or the presence of one or more characteristic defects of the processed workpiece surface captured on the at least one image is estimated by a pattern recognition algorithm (PRA D, PRA S), assigning the corresponding surface characteristic and/or the characteristic defects determined by the pattern recognition algorithm.

2. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, the method comprising one or more pattern recognition algorithms, wherein: one pattern recognition algorithm (PRA-S) is trained to the estimation of at least one surface characteristic of the processed workpiece surface; and/or one pattern recognition algorithm (PRA-D) is trained to the estimation of the presence of characteristic defects on the processed workpiece surface.

3. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, comprising one or more predefined training datasets, wherein said at least one predefined training dataset comprises a list with a plurality of records, whereas each record comprises a reference image and at least one known surface characteristics associated to said reference image; or a reference image and a known characteristic defect.

4. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 3, wherein said one or more predefined training datasets and/or said one or more pattern recognition algorithms are stored: in a machine tool control unit of the machine tool (1); or remotely, and is accessible to a machine tool control unit of the machine tool (1) by means of a network.

5. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, wherein the pattern recognition algorithm PRA-S is trained to estimate at least one of the following surface characteristics: Profile roughness parameter (Ra, Rz, Rmax, RSm); or Areal roughness parameter (Sa, Sq, Sz); or Specific feature of a functional workpiece surface.

6. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 5, wherein said specific feature of functional surfaces is associated to one of the following workpiece surface properties: superhydrophobic-, oleophobic-, antibacterial.

7. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, wherein the pattern recognition algorithm PRA-D is trained to identify defects on the processed workpiece surface, the defects being one or more of: a pitting mark, a micro-crack, a burn mark, a chatter mark, a burr, a breakout, a micro-geometrical error, a material defect, a machining residue.

8. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, wherein the processed workpiece surface is inspected within the machining process interruption of a current machining process, whereas said machining process interruption is: a machining process pause; or the end of the machining process.

9. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 8, wherein the processed workpiece surface is inspected within a machining process pause, and that at least one characteristic of the machined workpiece surface which is estimated by the pattern recognition algorithm PRA-S and/or PRA-D and compared with a target characteristic, and that based on said comparison the machining process is either resumed, ended or aborted.

10. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 8, wherein the machining process is resumed and that machining parameters are either left unaltered or changed in consideration of the estimate of pattern recognition algorithm (PRA-D) and/or with the estimate of the pattern recognition algorithm (PRA-S).

11. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 9, wherein the processed workpiece surface is inspected within a process pause or upon process completion, and that at least one estimated characteristic of the machined workpiece surface is compared with a target characteristic, and that one or more of the following actions are taken: the at least one estimated characteristic and comparison with the target characteristic is documented; the workpiece is handled according to the result of the comparison as good or defective.

12. The method for machining of workpieces (11) and inspection of the processed workpiece surface in a machine tool (1) according to claim 1, wherein said machining process interruption comprises one or more cleaning and drying cycles in view of the image capturing with a digital camera (12), whereas the cleaning and drying cycle is repeated if the pattern recognition algorithm (PRA-D, PRA-D) estimates one or more defects as to be machining residues.

13. A machine tool (1) for the machining of workpieces (11), including a computer numerical control system (CNC) for controlling the processing of a workpiece (11), wherein the computed numerical control system (CNC) is equipped with an integrated inspection system for the determination of the surface characteristics and/or characteristic defects of the processed workpiece surface, comprising wherein the computer numerical control system (CNC) with its integrated inspection system machines the workpiece (11) in accordance with one of the preceding method claims, the inspection system being equipped with a digital camera (12), preferably removable mounted inside the machine tool (1), most preferably the machine tool (1) is an electric discharge machine, a laser processing machine, a grinding machine or a milling machine.

14. The machine tool (1) for the machining of workpieces (11) according to claim 13, including an associated tool magazine (2), wherein the digital camera (12) is a wireless digital camera, and that said wireless digital camera (12) is stored in the tool magazine (2) associated to the machine tool (1).

15. The manufacturing cell including more than two machine tools (1) according to claim 13, the manufacturing cell further including a tool magazine (2) shared by the machine tools of the manufacturing cell, wherein the digital camera is deployable by a plurality of machine tools of said manufacturing cell.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Embodiments of the invention will now be described, by way of example, and with reference to the accompanying drawings, in which:

[0069] FIG. 1a is a simplified schematic illustration of an EDM machine with an optical measuring system stored in a rack

[0070] FIG. 1b is a simplified schematic illustration of the EDM machine with the optical measuring system placed in the machine head to take images of the surface

[0071] FIG. 2 is an illustration of a manufacturing cell with the optical measuring system placed in the rack

[0072] FIGS. 3-4 are images of workpiece surfaces processed by EDM with different machining parameters

[0073] FIGS. 5-6 are images of workpiece surface processed by EDM with different parameters, captured by a digital camera in view of estimation of surface characteristics

[0074] FIG. 7 is an image of a functional surface, having superhydrophobic properties

[0075] FIG. 8 is an image of a micro-crack in a surface

[0076] FIG. 9 is an image of a burn in a surface

[0077] FIG. 10 is a flow chart of the training dataset building phase

[0078] FIG. 11 is a flow chart of the algorithm for the estimation of defects and surface characteristics

DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0079] According to FIG. 1a, there is illustrated a die sinking electrical discharge machine 1, equipped with a digital camera 12. The optical axis of the digital camera is parallel with the Z-axis of the machine. The digital camera acquires a 2D-image of the machining surface by means of a 2D-digital camera or by a linear camera which acquires the surface while camera and surface move linearly in the orthogonal direction with respect to each other. The tool electrode 10 is clamped to the chuck on the machining head 16 whereas the digital camera 12 is safely stored in the rack 15 of an electrode changer, at the border of the machining area or outside thereof, in a large capacity magazine (cp. FIG. 2).

[0080] After an initial machining phase, the machining process is paused, the processing liquid is drained and the workpiece 11 is prepared for a measurement. The area of interest is cleaned and dried, either manually or automatically.

[0081] The tool electrode 10 is removed from its working position by means of the electrode changer, stored in rack 15 and replaced by the digital camera 12, which is now precisely clamped to the chuck on the machining head 16, as shown in FIG. 1b. A reclamping error of for instance 2 m may normally occur with standard EDM electrode tool holders. Normally said error is not relevant for the present invention, however a correction cycle may be used to correct the offset of the optical axis of the digital camera with respect to the machine axis.

[0082] FIG. 2 illustrates a manufacturing cell comprising a die sinking electrical discharge machine 1, a large capacity magazine 2, a robotic manipulator 3 mounted on rails, and further machine tools or stations (not shown). As illustrated the digital camera 12 is stored in magazine 2, where it can be grasped by the robotic manipulator 3 and clamped on machine 1 or another machine. A particular position within magazine 2 is dedicated to the digital camera 12, said position comprising a charging unit 4 by which the battery of the digital camera is charged. The charging occurs either inductively or by physically contacting the digital camera.

[0083] The digital camera 12 takes one or more images of the surface of the workpiece 11. FIG. 3 shows a typical crater-landscape as generated by a die sinking process. The example has an Ra value of 1.26 m, and RSm=78 m, which may be a achieved in a finishing step. The machine memory, or a remote memory being accessible to the machine, contains a pre-trained pattern recognition algorithm. Such algorithm was previously trained using a training dataset that has been created in laboratory by machining a big number of reference surfaces and by measuring them with sophisticated reference instrumentation and obtaining several reference values, e.g., Ra, Rz, RSm, Rt, Sa, Sz for each specific machining condition, i.e. a regime, electrode pairing. The training dataset includes one or more images of each reference surface with one or more reference values associated to said one or more image of the reference surface.

[0084] The superior value of the invention can be understood by observing FIGS. 3 and 4, having the same magnification. Although these two images look different, they have the same Ra value. Therefore describing the surface characteristics by only one surface characteristic value, e.g., Ra, might not be sufficient. A holistic approach that considers the totality of the surface, estimates not only the frequently used surface characterization parameters Ra, Rz, RSm, Rt, Sa, Sz, but is also able to provide a correct information about specific functional characteristics of said surface (e.g., hydrophobicity, oleophobic-, antibacterial).

[0085] FIGS. 5 and 6 are further images of a surface which has been processed by EDM. The images are captured by a digital camera in view of the estimation of the surface characteristics. It is to be noted that the surfaces in FIGS. 5 and 6 have same arithmetic roughness Ra, but different width of profile elements RSm. By applying a pattern recognition algorithm according to the invention, the surface characteristics are easily and quickly estimated.

[0086] The FIG. 7 is an image of a functional surface, having superhydrophobic properties. Such surfaces have very complex structures, which are impossible to be described with the existing methods. The invention provides however a cost effective solution for the surface inspection, which is affordable by all machine tool users, whereas the sophisticated, very expensive measuring instruments necessary in the state of the art are out of reach for most of them.

[0087] FIG. 8 is a polished micrograph section of the machined workpiece, illustrating an important crack occurred in a recast layer. FIG. 9 is a picture of the machined workpiece showing burning marks. Such damages can occur for several reasons, e.g., with unsuitable processing parameters, but also due to difficult flushing conditions and the like. It is important to avoid such defects; should they occur anyway it is imperative to locate them safely.

[0088] The sequence to build up a training dataset for a pre-trained pattern recognition algorithm is illustrated in FIG. 10: Firstly, reference surfaces are produced with defined regimes. One or preferably more images of each of the reference surfaces are captured with a suitable digital camera and illumination setup, which are similar to those implemented in the machine. The surface characteristics of these surfaces are determined by using reference measuring equipment, for instance the mentioned and known roughness measuring instruments of Taylor Hobson Ltd. The images of the produced reference surfaces and the associated measured surface characteristics are listed in a dataset and constitute the training dataset that is then used in a learning phase to train the pattern recognition algorithms used by the invention.

[0089] A possible operational sequence of the inventive method is illustrated in FIG. 11 for a simplified machining task, i.e. erosion in Z axis (vertical). At a predetermined intermediate point the machining process is paused, and the workpiece 11 is prepared and an image is taken by means of the digital camera 12, as explained further above. A pre-trained pattern recognition algorithm estimates the surface characteristics or the presence of surface defects based only on the image of the machining surface. This can be done in one oras FIG. 11 illustratesin two steps: a first pattern recognition algorithm (PRA-D) estimates the presence of surface defects (see e.g., FIG. 8). A second pattern recognition algorithm (PRA-S) estimates the surface characteristics.

[0090] If the estimates determined by one or both pattern recognition algorithms PRA D and PRA S do not correspond to the desired value, the machine can adjust the processing parameters and resume therewith the machining process to correct this inaccuracy. Therefore it is advisable to check the surface when there is still room to allow an eventual correction. The above mentioned predetermined point of the machining is preferably a predetermined machining depth. For example, if the machining depth is Z=10 mm, the check may be done at 9.95 mm, to allow 50 m of processing by using a correction regime. Alternatively the above mentioned predetermined point of the machining may be a predetermined processing time, or a predetermined regime, or other criteria. If both pattern recognition algorithms PRA D and PRA S do not determine any surface defect or surface quality problem, the processing of the workpiece can continue orin case the inventive measuring method is applied at the end of the machining processthe machining process is ended with that determination of the surface characteristic.

[0091] In a preferred embodiment the machine further comprises a means or method for the dimensional measurements, such as measurement of the effective (intermediate or final) depth of the cavity. For instance, the top surface and bottom of cavity may be touched by the electrode or by a touch probe, or, the digital camera used to take the images comprises a feature to determine the exact Z position of the surface and can thus be used to determine the depth of the cavity. In case of deviation with respect to expected intermediate depth values, the machining depth can be corrected by resuming the machining. The remaining amount of material is hereby identified more precisely in view of an eventual correction to achieve the desired surface characteristics respectively the desired roughness at the final workpiece surface.

[0092] As just explained, the inventive method includes in a preferred embodiment that a second pattern recognition algorithm (PRA-D) is used to detect, i.e., estimate the presence of a number of possible surface defects. These defects can be pitting, cracks, burns, etc. (see FIGS. 8 and 9).

[0093] Preferably, the second pattern recognition algorithm (PRA-D) to detect surface defects is applied before the pattern recognition algorithm (PRA-S) to estimate the surface characteristics. If a surface defect is determined by the second pattern recognition algorithm (PRA-D), the process parameters are adjusted accordingly.

[0094] This quality control is very important e.g., for the aerospace applications. Micro-cracks as illustrated in FIG. 8 can have catastrophic consequences. At least they reduce the lifetime of the parts, so that they have be avoided. By using the inventive method, defective workpieces can automatically be discarded by separating them from the good parts, and non-conformities are automatically documented.

[0095] The defect check can also be process driven, e.g., in case the acquired EDM machining process parameters signal a deteriorated or dangerous situation (e.g., short circuit or arcing rate). In such a case the machining process can be stopped and a defect check according to the inventive method can be carried out. Since the position of the electrical discharge can be detected in modern EDM machine tools, a map of the zones with deteriorated discharges can be generated. In this way the inspection according to the invention can possibly be reduced to the zones where such a deteriorated or dangerous situation has occurred.

[0096] Some minor defects, like burns (see FIG. 9), can be corrected by applying a reversed polarity (electrode negative, workpiece positive) and by generating very short pulses, or positive polarity and low frequency pulses.

[0097] In this case, according to the invention, first the minor defects are removed, then the correct roughness is achieved as explained further above.

[0098] The invention addresses the specific deviations of the current machining with respect to target values and tolerances, thus providing an adaptive machining process.

[0099] The invention can be applied to most types of machine tools and machining operations of workpieces, such as electric discharge machines, laser processing machines, a grinding machine or milling machines.

[0100] The invention relates also to machine tools, which are accordingly equipped to apply the inventive method. Therefore prepared machine tools for the machining of workpieces include a computer numerical control system (CNC) for controlling the processing of a workpiece. The computed numerical control system (CNC) is equipped with an integrated inspection system for the determination of the surface characteristics and/or characteristic defects of the processed workpiece surface. The computer numerical control system (CNC) with its integrated inspection system machines and inspects the workpiece in accordance with the inventive method. The inspection system is equipped with a digital camera, preferably removable mounted inside the machine tool.

[0101] Preferably the machine tool is an electric discharge machine, a laser processing machine, a grinding machine or a milling machine.

[0102] In a preferred embodiment, the machine tool includes an associated tool magazine in which the digital camera is stored. Preferably, the digital camera is a wireless digital camera.

[0103] Further, the invention relates also to manufacturing cells including two or more of the inventive machine tools. The manufacturing cell includes a tool magazine, which is shared by the machine tools of the manufacturing cell. Preferably, the digital camera is deployable by a plurality of machine tools of said manufacturing cell. Most preferably the digital camera is store in the tool magazine and deployable by the machine tools of the manufacturing cell.

[0104] The present invention is not limited to the explicitly explained examples and embodiments. The illustrated alternatives are rather to be considered suggestions intended to motivate the person skilled in the art to implement the invention in a most favorable way.

REFERENCES

[0105] 1 machine tool [0106] 2 tool magazine, electrode magazine [0107] 3 robotic manipulator, robot [0108] 4 charging unit [0109] 10 tool electrode [0110] 11 workpiece [0111] 12 digital camera [0112] 15 rack, tool holder [0113] 16 machining head