Method and control system for controlling a physical vapor deposition machine

Abstract

A method of controlling a physical vapor deposition (PVD) machine is described. The method comprises the steps of capturing, by means of at least one camera (32), an image of at least one evaporator boat (14) of a plurality of evaporator boats (14); extracting, by means of an image processing module, at least one image portion of the image corresponding to the at least one evaporator boat (14); detecting, by means of a machine-learning module, a pool of molten material, a supply wire, and / or a contact point between a supply wire and the at least one evaporator boat based on the at least one image portion; and generating, by means of a control module, a control signal for an energy source and / or for supply wire drives based on the detected pool, supply wire, and / or contact point. Further, a control system (26) and a PVD machine are described.

Description

[0001] Method and control system for controlling a physical vapor deposition machine

[0002] The present invention generally relates to a method of controlling a physical vapor deposition (PVD) machine. The present invention further relates to a control system for a PVD machine, and to a PVD machine.

[0003] A PVD machine is a machine with which a material is deposited on a substrate. The material forms a layer consisting of a metal and / or a metal oxide, such as aluminum oxide, on the substrate. The substrate can be a thin film (such as a plastic, foil, paper or card) and be used in many different application, e.g. as a packaging material, electronic capacitors, batteries, insulations and many more. Such a packaging material may be used for packaging food. The layer may be used to provide a barrier against ingress of gas and / or water and / or light. The layer provided on the foil, paper or card may be transparent and / or mechanically dense and / or mechanically stable.

[0004] In some types of PVD machines, evaporator boats are used for melting and evaporating the deposition material. The evaporator boats are arranged in a process chamber so that a vacuum can be established for conducting the deposition process. The deposition material is applied to the evaporator boats by supply wires that are advanced towards the evaporator boats by supply wire drives. The evaporator boats are heated in order to melt and evaporate the deposition material.

[0005] Typically, the evaporator boats are controlled manually by the operator. The manual process involves the operator looking through a transparent window to visually inspect a pool of molten material on the evaporator boat and adjusting electrically via power, voltage or current to ensure an optimized pool shape of evaporant material based on variables such as wire feed rate, evaporant material, metal deposition as well as different processes (e.g. AIOx, Dark Night, AluBond, etc.) and evaporator boat age. Failure to do so will result in poor product quality and a reduction in life of the evaporator boats.

[0006] In a typical metallizing process the operator requires to make adjustments every 5 minutes to many evaporators (up to 60 evaporators or even more) for a 1hr cycle time to ensure good product quality. This operator dependent task requires a very experienced machine operator to ensure product quality, maximizing yield and increasing the life of the evaporator boat, as the contrast between the pool and boat surface is typically low and thus difficult to assess correctly.

[0007] The object of the present invention is to facilitate the control of PVD machines.

[0008] According to the present invention, the problem is solved by a method of controlling a physical vapor deposition (PVD) machine. The PVD machine comprises a plurality of evaporator boats arranged in a fixture, wherein the fixture is configured to supply energy from an energy source to the evaporator boats. The PVD machine further comprises a plurality of supply wire drives being configured to advance supply wires to the evaporator boats. The method comprises the steps of capturing, by means of at least one camera, an image of at least one evaporator boat of the plurality of evaporator boats; extracting, by means of an image processing module, at least one image portion of the image corresponding to the at least one evaporator boat; detecting, by means of a machine-learning module, a pool of molten material, a supply wire, and / or a contact point between the supply wire and the at least one evaporator boat based on the at least one image portion; and generating, by means of a control module, a control signal for the energy source and / or for the supply wire drives based on the detected pool, supply wire, and / or contact point.

[0009] Therein and in the following, the term “module” is understood to describe suitable hardware, suitable software, or a combination of hardware and software that is configured to have a certain functionality.

[0010] The hardware may, inter alia, comprise a CPU, a GPU, an FPGA, an ASIC, or other types of electronic circuitry. Moreover, the term “detect a pool of molten material” may be understood to denote detecting a size and / or shape of the pool of molten material, namely by detecting which portions of the at least one image portion are covered by the pool of molten material.

[0011] Likewise, “detect a supply wire” may be understood to denote detecting which portions of the at least one image portion correspond to the supply wire.

[0012] The method according to the present invention is based on the idea to apply for example artificial intelligence, machine-learning and / or deep learning techniques or any other similar technique to pre-processed images of the at least one evaporator boat, namely the at least one extracted image portion, in order to detect the pool of molten material, the supply wire, and / or the contact point between the supply wire and the at least one evaporator boat.

[0013] In fact, the machine-learning module is pre-trained to detect the pool, supply wire, and / or contact point, as will be described in more detail below.

[0014] The control module automatically generates a control signal for the energy source and / or for the supply wire drives based on the detected pool, supply wire, and / or contact point. Accordingly, the energy supplied to the at least one evaporator boat and thus the temperature of the at least one evaporator boat can be controlled automatically. Alternatively or additionally, the speed with which the corresponding supply wire is advanced towards the at least one evaporator boat can be controlled automatically.

[0015] Particularly, the control module may automatically control the energy source and / or the supply wire drives such that an optimal pool shape, an optimal pool size, and / or an optimal contact point is maintained.

[0016] While it may be still possible that a human operator may manually adjust individual settings of the PVD machine, such as the energy supplied to the evaporator boats and the supply wire speed, controlling the machine is facilitated significantly by the method according to the present invention.

[0017] Thus, a uniform metallizing process with a maximum of quality, product yield and consumable utilization is ensured. According to an aspect of the present invention, obscured areas are detected by means of the machine-learning module based on the at least one image portion. In general, the obscured areas are portions of the image that are obscured by components of the PVD machine or other objects. For example, the obscured areas may be caused by deposits or rather debris on a portion of the fixture holding the evaporator boats in place.

[0018] Correctly identifying the obscured areas can lead to an enhanced accuracy of detecting the pool, the supply wire, and / or the contract point.

[0019] Further, it is also conceivable that the machine-learning module may generate a warning signal if more than predefined threshold of the at least one image portion corresponds to an obscured area, and / or if the obscured areas cover the pool, wire, and / or (presumed) contact point. Thus, an operator may be warned that the PVD machine has to be cleaned of the debris.

[0020] In an embodiment of the present invention, the machine-learning module comprises a first machine-learning sub-module and a second machine-learning sub-module, wherein the first machine-learning sub-module is trained to detect the pool of molten material, and wherein the second machine-learning submodule is trained to detect the supply wire and / or the contact point. In other words, different sub-modules of the machine-learning module are specifically pretrained to detect the pool of molten material on hand, and the supply wire and / or the contact point on the other hand. This way, the detection accuracy of the individual machine-learning sub-modules is enhanced, as specialized training can be applied to the individual sub-modules.

[0021] According to another aspect of the present invention, the at least one extracted image portion is corrected for perspective by means of the image processing module, thereby obtaining at least one corrected image portion. Thus, before the extracted image portions are provided to the machine-learning module for detection of the pool, wire, and / or contact point (and / or the obscured areas), the at least one image portion is pre-processed by the image processing module in order to account for perspective distortion (e.g. homography) of the at least one evaporator boat depicted in the at least one image portion. Particularly, the pool, the supply wire, and / or the contact point is detected in the at least one corrected image portion. Of course, the obscured areas described above may also be detected in the at least one corrected image portion. This way, the detection accuracy for detecting the pool, wire, contact point, and / or obscured areas is enhanced.

[0022] As will be described in more detail below, the at least one extracted image portion may also be displayed on a display. Correcting the at least one extracted image portion for perspective distortions also makes it easier for an operator to assess whether the PVD machine is operating within nominal parameters.

[0023] Particularly, the at least one extracted image portion may be transformed to a standard format. Therein, the term “format” relates to the geometry of the extracted image portion, particularly to the dimensions of the extracted image portion. As the at least one corrected image portion has a standard format, the machine-learning module can be specifically trained for that format, which may further enhance the achievable detection accuracy.

[0024] For example, the standard format may be a rectangle having a predefined length and a predefined width.

[0025] Image processing can be done with classical methods using raw codes. A further aspect of the present disclosure provides that the image processing module is a sub-module of the machine-learning module. In other words, the image processing module may be a machine-learning sub-module of the machine-learning module, wherein the image processing module is pre-trained to extract the at least one image portion from the image captured by the at least one camera.

[0026] More precisely, the image processing module may be pre-trained to detect evaporator boats in the image, and to cut out the detected evaporator boats from the image, thereby obtaining the at least one image portion.

[0027] The machine-learning module may comprise at least one artificial neural network, wherein the at least one artificial neural network is trained to detect the pool, the supply wire, and / or the contact point. Of course, the at least one artificial neural network may also be trained to detect the obscured areas described above. In fact, the machine-learning module may comprise a plurality of artificial neural networks, wherein the individual artificial neural networks may be pretrained to at least one of detect the pool of molten material in the at least one image portion, detect the supply wire in the at least one image portion, detect the contact point in the at least one image portion, detect obscured areas in the at least one image portion, extract the at least one image portion, and correct the at least one image portion for perspective.

[0028] Dependent upon the web width of the machine (or the web width that is being processed) there may only be one boat in a camera’s image (i.e. The camera’s at the edge of the film). According to an aspect of the present invention, the at least one image comprises at least two evaporator boats. Thus, less cameras are necessary for capturing images of all evaporator boats, such that the costs for controlling the PVD machine are reduced. Further the quality of the film may be enhanced, in particular if lower skilled operators are employed.

[0029] According to a variant of the present invention, the at least one image portion is augmented with additional information regarding the detected pool, wire, and / or contact point by means of the machine-learning module, thereby obtaining at least one augmented image portion. Of course, the at least one image portion may further be augmented with additional information regarding the detected obscured areas described above. In general, the additional information facilitates recognizing the pool, wire, and / or contact point (and / or obscured areas) in the at least one image portion. Thus, supervision and control of the PVD machine is facilitated.

[0030] Particularly, the at least one augmented image portion is displayed on a display. Thus, instead of the raw images captured by the at least one camera, the augmented images comprising the additional information may be displayed to an operator of the PVD machine, thereby facilitating supervision and control of the PVD machine.

[0031] For example, the augmented image portion comprises colors, markers, lines, and / or text labels marking the detected pool, wire, and / or contact point. Thus, the pool, wire, and / or contact point are easier to discern for an operator or another technician of the PVD machine, as the pool, wire, and / or contact point may be clearly marked. Of course, the augmented image portion may further comprise colors, markers, lines, and / or text labels marking the detected obscured areas described above.

[0032] However, it is to be understood that the pool, wire, contact point, and / or obscured areas may be marked in any other suitable way.

[0033] According to a further variant of the present invention, at least one artificial image portion is generated by means of the machine-learning module based on the at least one image portion and based on the detected pool, wire, and / or the contact point, wherein the detected pool, wire, and / or contact point are represented in the at least one artificial image portion. In general, the at least one artificial image portion may be generated such that the detected pool, wire, contact point, and / or obscured areas are easily discernable in the at least one artificial image portion, thereby facilitating supervision and control of the PVD machine.

[0034] For example, the at least one artificial image portion may comprise colors, markers, lines, and / or text labels marking the detected pool, wire, contact point, and / or obscured areas.

[0035] Therein and in the following, the term “artificial image portion” is understood to denote an image portion that is generated by the machine-learning module from scratch, while the term “augmented image portion” is understood to denote an image portion captured by the camera that is being adapted by the machinelearning module.

[0036] Particularly, the at least one artificial image portion is displayed on a display. Thus, instead of the raw images captured by the at least one camera, the artificial images comprising the additional information may be displayed to an operator of the PVD machine, thereby facilitating supervision and control of the PVD machine.

[0037] According to the invention, the problem further is solved by a control system for a PVD machine. The control system comprises at least one camera, an image processing module, a machine-learning module, and a control module. The control system is configured to perform the method described above. Regarding the advantages and further properties of the control system, reference is made to the explanations given above with respect to the method, which also hold for the control system and vice versa.

[0038] According to the invention, the problem further is solved by a PVD machine. The PVD machine comprises a control system described above.

[0039] Regarding the advantages and further properties of the PVD machine, reference is made to the explanations given above with respect to the method and the control system, which also hold for the PVD machine and vice versa.

[0040] The foregoing aspects and many of the attendant advantages of the claimed subject matter will become more readily appreciated as the same become better understood by reference to the following detailed description, when taken in conjunction with the accompanying drawings, wherein:

[0041] Figure 1 schematically shows a first variant of a PVD machine according to the present invention;

[0042] Figure 2 schematically shows a second variant of a PVD machine according to the present invention;

[0043] Figure 3 is a schematic view of a fixture for evaporator boats as used in the machine of Figures 1a and 1b;

[0044] Figure 4 schematically shows a control system according to the present invention;

[0045] Figure 5 shows a flow chart of a method of controlling a PVD machine according to the present disclosure;

[0046] Figures 6 to 8 show images illustrating individual steps of the method of Figure 5; and

[0047] Fig. 9 shows an image illustrating a training method for a machinelearning module.

[0048] The detailed description set forth below in connection with the appended drawings, where like numerals reference like elements, is intended as a description of various embodiments of the disclosed subject matter and is not intended to represent the only embodiments. Each embodiment described in this disclosure is provided merely as an example or illustration and should not be construed as preferred or advantageous over other embodiments. The illustrative examples provided herein are not intended to be exhaustive or to limit the claimed subject matter to the precise forms disclosed.

[0049] For the purposes of the present disclosure, the phrase “at least one of A, B, and C”, for example, means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C), including all further possible permutations when more than three elements are listed. In other words, the term “at least one of A and B” generally means “A and / or B”, namely “A” alone, “B” alone or “A and B”.

[0050] Figures 1 and 2 show essential components of a PVD machine. The PVD machine comprises a process drum 10 or free span rollers 11 around which a substrate 12 in the form of a web is guided. The substrate 12 can be a thin plastic foil which is used for packaging food but could also well be paper, card, fabric etc. for any suitable application.

[0051] Without restriction of generality, the exemplary embodiment of the PVD machine shown in Figure 1 is described hereinafter. However, it is to be understood that the explanations given hereinafter likewise apply to the embodiment shown in Figure 2.

[0052] For providing a deposition material to be deposited on the substrate 12, a plurality of evaporator boats 14 is provided in the vicinity of the process drum 10. Evaporator boats 14 are arranged in a fixture 16 (please see Figure 3 so as to form a row of adjacent evaporator boats, the row being arranged in parallel with the axis of rotation of the process drum 10 so that the entirety of the evaporator boats 14 extends over the entire width of substrate 12. The arrangement of the evaporator boats does not necessarily need to be in a row. This could be in any other form.

[0053] As is shown in Figure 3, the evaporator boats 14 may be arranged in a staggered arrangement relative to one another. Other arrangements are possible, e.g. arranging the evaporator boats in line.

[0054] The fixture 16 is adapted for supplying electric energy from an energy source

[0055] 18 (schematically depicted in Figures 1 and 2) to the evaporator boats 14. The amount of energy supplied to the evaporator boats 14 can be controlled separately for each evaporator boat 14.

[0056] For example, depending on the width of the substrate 12, up to 60 or more evaporator boats 14 can be arranged adjacent to each other in the fixture 16.

[0057] The deposition material is supplied to each of the evaporator boats 14 in the form of a supply wire 20 which is stored on a supply reel 22. For each of the evaporator boats 14, a supply wire drive 24 is provided which controls the speed with which the supply wire 20 is advanced towards the respective evaporator boat 14.

[0058] The supply wire drive 24 may be implemented in the form of a stepper motor.

[0059] A control system 26 is provided for controlling various functions of the PVD machine.

[0060] In general, the control system 26 is configured to control the amount of energy provided to each of the evaporator boats 14, and the respective speed of the supply wire drives 24.

[0061] A surface inspection system 28 may be provided which inspects the surface of substrate 12 downstream of process drum 10, wherein the surface inspection system 28 may be connected to the control system 26.

[0062] Surface defects of the substrate provided with the deposition material as well as other quality issues can be detected by the surface inspection system 28.

[0063] A process chamber 30 is formed which allows establishing a vacuum in the area in which the deposition material is deposited on the substrate 12.

[0064] At least one camera 32 is provided for capturing an image of at least one of the evaporator boats 14. The term „camera“ here designates each and every device which is able convert optical information within the viewing area of the device into electronic information.

[0065] It is possible to use one camera 32 for each of the evaporator boats 14. In order to reduce the number of necessary cameras 32, it however is preferred to use cameras 32 which each cover a plurality of evaporator boats 14. As an example, each of the cameras 32 can capture the images of three evaporator boats 14.

[0066] The cameras 32 are arranged outside or protected inside of the process chamber 30. A viewing window 34 is provided in a wall of the process chamber 30 so as to allow the cameras to capture the images of the evaporator boats 14.

[0067] The images captured by the cameras 32 (infrared and visible light emission) are supplied to the control system 26 which analyzes the captured images as described in more detail hereinafter.

[0068] Figure 4 schematically shows the control system 26 in more detail.

[0069] Besides the at least one camera 32, the control system 26 comprises a machine-learning module 36 that is connected to the at least one camera 32.

[0070] The machine-learning module 36 comprises an image processing module 38 that is connected to the at least one camera 32 so as to receive the images captured by the at least one camera 32.

[0071] It is noted that the image processing module 38 may be established separately from the machine-learning module 36.

[0072] Without restriction of generality, the exemplary case of the image processing module 38 being established as a sub-module of the machine-learning module 36 is described in the following.

[0073] The machine-learning module 36 further comprises a plurality of machinelearning sub-modules. In the exemplary embodiment shown in Figure 4, the machine-learning module 36 comprises a first machine-learning sub-module 40, a second machine-learning sub-module 42, and a third machine-learning submodule 44.

[0074] Downstream of the machine-learning sub-modules 40, 42, 44, a visualization module 46 is provided.

[0075] The image processing module 38, the machine-learning sub-modules 40, 42, 44, and / or the visualization module 46 may comprise or be established as an artificial neural network, respectively. The control system 26 further comprises a control module 48 that is connected to outputs of the machine-learning sub-modules 40, 42, 44.

[0076] It is noted that while the control module 48 is illustrated to be established separately from the machine-learning module 36 in Figure 4, it is also conceivable that the control module 48 may be integrated into the machinelearning module 36.

[0077] For example, the control module 48 may comprise or be established as an artificial neural network.

[0078] The control system 26 is configured to perform a method of controlling the PVD machine that is described hereinafter with reference to Figure 5.

[0079] An image of at least one of the evaporator boats 14, particularly of all evaporator boats 14, is captured by the at least one camera 32 (step S1).

[0080] In the exemplary embodiment illustrated in Figure 6, the captured image comprises three evaporator boats 14. However, it is to be understood that the captured image may comprise any other number of evaporator boats 14.

[0081] The captured image is forwarded to the image processing module 38.

[0082] At least one image portion corresponding to the at least one evaporator boat 14 is extracted by means of the image processing module 38, and the at least one image portion is corrected for perspective by means of the image processing module 38, thereby obtaining at least one corrected image portion (step S2).

[0083] As is illustrated in Figure 7, the individual image portions are transformed to a standard format, i.e. to a standard length and width, wherein perspective distortions caused by different angles between the at least one camera 32 and the evaporator boats 14 are accounted for.

[0084] The at least one corrected image portion is forwarded to the machine-learning sub-modules 40, 42, 44, respectively.

[0085] A pool of molten material, a supply wire, a contact point between the supply wire and the at least one evaporator boat 14 and / or obscured areas are detected by the machine-learning sub-modules 40, 42, 44 based on the at least one corrected image portion (step S3). This is illustrated in Figure 8, which shows the at least one corrected image portion (“Source”), the detected pool of molten material (“Pool”), the detected contact point (“Wire”), and the detected obscured areas (“Obscured Area Mask”).

[0086] For example, the pool of molten material may be detected by the first machine-learning sub-module 40. The supply wire and / or the contact point may be detected by the second machine-learning sub-module 42. The obscured areas may be detected by the third machine-learning sub-module 44.

[0087] Based on the detected pool, supply wire, contact point, and / or obscured areas, a control signal for the energy source 18 and / or for the supply wire drives 24 is generated by the control module 48 (step S4).

[0088] In general, the energy supplied to the individual evaporator boats 14 and thus the temperature of the individual evaporator boats 14 is controlled by means of the control signal. Alternatively or additionally, the speed with which the corresponding supply wire is advanced towards the individual evaporator boats 14 is controlled by means of the control signal.

[0089] In fact, the control module 48 may automatically control the energy source 18 and / or the supply wire drives 24 such that an optimal pool shape, an optimal pool size, and / or an optimal contact point is maintained.

[0090] Further, at least one augmented image portion and / or at least one artificial image portion may be generated by the visualization module 46 based on the detected pool, wire, contact point, and / or obscured areas (step S5).

[0091] Therein, the at least one augmented image portion corresponds to the at least one corrected image portion, but augmented with additional information regarding the detected pool, wire, contact point, and / or obscured areas.

[0092] For example, the augmented image portion comprises colors, markers, lines, and / or text labels marking the detected pool, wire, contact point, and / or obscured areas.

[0093] It is also conceivable that the at least one image portion may comprise additional information on operational parameters of the individual evaporator boats 14, such as a temperature of the evaporator boats 14 and / or a temperature of the pool of molten material. In a particular example, the contact point may be marked with a cross, as is illustrated in the left portion of Figure 8.

[0094] The pool of molten material, particularly portions of the pool of molten material having different temperatures, may be marked with different colors, as is illustrated in the portion of Figure 8 labeled “Pool”.

[0095] Likewise, the detected wire and / or the obscured areas may be marked with different colors.

[0096] The explanations given above likewise apply to the at least one artificial image portion, wherein the at least one artificial image portion corresponds to an artificially generated image of the at least one corrected image portion, augmented with additional information regarding the detected pool, wire, contact point, and / or obscured areas.

[0097] The at least one augmented image portion and / or the at least one artificial image portion may be displayed on a display (step S6).

[0098] The display may be integrated into the PVD machine or may be established separately from the PVD machine.

[0099] As far as the steps described above are performed by the machine-learning module 36 or its sub-modules, it is understood that the machine-learning module 36 or rather its sub-modules are pre-trained to perform the described functionality.

[0100] The training may be performed by any suitable machine-learning technique. For example, supervised deep learning machine-learning techniques based on labeled training data may be performed, wherein the training data may be labeled by a human expert.

[0101] However, it is to be understood that any other suitable machine-learning technique may be applied.

[0102] Figure 9 shows a specific example of training data for training the first machine-learning sub-module 40 to detect the pool of molten material.

[0103] The training data comprises images of evaporator boats, wherein the pool of molten material has been manually marked by a human expert. In order to train the machine-learning module 36 or its respective submodules, the machine-learning module 36 may be fed with the raw images of the evaporator boats, and the machine-learning module 36 detects the pool of molten material, i.e. determines the size and / or shape of the pool of molten material.

[0104] Based on the labels provided by the human expert, errors between the pool detected by the machine-learning module 36 and the actual pool of molten material marked by the human expert can be determined.

[0105] Based on the errors determined, weighting factors of the machine-learning module 36 can be adapted, for example by backpropagation of the errors.

[0106] The same type of training data, i.e. supervised and / or unsupervised training, may be provided for training the detection of the wire, the contact point, and / or the obscured areas.

[0107] Certain embodiments disclosed herein, particularly the respective module(s) and / or unit(s), utilize circuitry (e.g., one or more circuits) in order to implement standards, protocols, methodologies or technologies disclosed herein, operably couple two or more components, generate information, process information, analyze information, generate signals, encode / decode signals, convert signals, transmit and / or receive signals, control other devices, etc. Circuitry of any type can be used.

[0108] In an embodiment, circuitry includes, among other things, one or more computing devices such as a processor (e.g., a microprocessor), a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a system on a chip (SoC), or the like, or any combinations thereof, and can include discrete digital or analog circuit elements or electronics, or combinations thereof. In an embodiment, circuitry includes hardware circuit implementations (e.g., implementations in analog circuitry, implementations in digital circuitry, and the like, and combinations thereof).

[0109] In an embodiment, circuitry includes combinations of circuits and computer program products having software or firmware instructions stored on one or more computer readable memories that work together to cause a device to perform one or more protocols, methodologies or technologies described herein. In an embodiment, circuitry includes circuits, such as, for example, microprocessors or portions of microprocessor, that require software, firmware, and the like for operation. In an embodiment, circuitry includes one or more processors or portions thereof and accompanying software, firmware, hardware, and the like. The present application may reference quantities and numbers. Unless specifically stated, such quantities and numbers are not to be considered restrictive, but exemplary of the possible quantities or numbers associated with the present application. Also in this regard, the present application may use the term "plurality" to reference a quantity or number. In this regard, the term "plurality" is meant to be any number that is more than one, for example, two, three, four, five, etc. The terms "about", "approximately”, "near" etc., mean plus or minus 5% of the stated value.

Claims

Claims1. A method of controlling a physical vapor deposition (PVD) machine, wherein the PVD machine comprises a plurality of evaporator boats (14) arranged in a fixture (16), wherein the fixture (16) is configured to supply energy from an energy source (18) to the evaporator boats (14), wherein the PVD machine further comprises a plurality of supply wire drives (24) being configured to advance supply wires (20) to the evaporator boats (14), wherein the method comprises the steps of capturing, by means of at least one camera (32), an image of at least one evaporator boat (14) of the plurality of evaporator boats (14); extracting, by means of an image processing module (38), at least one image portion of the image corresponding to the at least one evaporator boat (14); detecting, by means of a machine-learning module (36), a pool of molten material, a supply wire (20), and / or a contact point between the supply wire and the at least one evaporator boat (14) based on the at least one image portion; and generating, by means of a control module (48), a control signal for the energy source (18) and / or for the supply wire drives (24) based on the detected pool, supply wire, and / or contact point.

2. The method of claim 1 , wherein obscured areas are detected by means of the machine-learning module (36) based on the at least one image portion.

3. The method according to any one of the preceding claims, wherein the machine-learning module (36) comprises a first machine-learning sub-module (40) and a second machine-learning sub-module (42), wherein the first machinelearning sub-module (40) is trained to detect the pool of molten material, and wherein the second machine-learning sub-module (42) is trained to detect the supply wire and / or the contact point.

4. The method according to any one of the preceding claims, wherein the at least one extracted image portion is corrected for perspective by means of theimage processing module (38), thereby obtaining at least one corrected image portion.

5. The method of claim 4, wherein the pool, the supply wire, and / or the contact point is detected in the at least one corrected image portion.

6. The method of claim 4 or 5, wherein the at least one extracted image portion is transformed to a standard format.

7. The method according to any one of the preceding claims, wherein the image processing module (38) is a sub-module of the machine-learning module (36).

8. The method according to any one of the preceding claims, wherein the at least one image comprises at least two evaporator boats (14).

9. The method according to any one of the preceding claims, wherein the at least one image portion is augmented with additional information regarding the detected pool, wire, and / or contact point by means of the machine-learning module (36), thereby obtaining at least one augmented image portion.

10. The method of claim 9, wherein the at least one augmented image portion is displayed on a display.

11. The method of claim 9 or 10, wherein the augmented image portion comprises colors, markers, lines, and / or text labels marking the detected pool, wire, and / or contact point.

12. The method according to any one of the preceding claims, wherein at least one artificial image portion is generated by means of the machine-learning module (36) based on the at least one image portion and based on the detected pool, wire, and / or the contact point, wherein the detected pool, wire, and / or contact point are represented in the at least one artificial image portion.

13. The method according to any one of the preceding claims, wherein alarm, status or guidance messages are displayed for an operator based upon the pool, wire, contact point or other information derived from the artificial image portion.

14. A control system for a PVD machine, wherein the control system (26) comprises at least one camera (32), an image processing module (38), amachine-learning module (36), and a control module (48), wherein the control system (26) is configured to perform the method according to any one of the preceding claims.

15. A PVD machine, comprising a control system (26) according to claim 14.

Images