Machining system, method of machining, method of training machine learning agent, aircraft and computer program

Abstract

The invention relates to a machining system, a method of machining, a method of training a machine learning agent, an aircraft and a computer program. According to one aspect of the invention, there is provided a machining system comprising: an automated manipulator configurable between a measurement configuration in which a position detector is operable and a machining configuration in which a tool is operable; a holder for holding the template or the workpiece; and a controller. The controller is configured to cause the automated manipulator to move the position detector to at least one reference feature of the template held in the fixture in the measurement configuration to determine coordinate data associated with the at least one reference feature. The controller then provides the determined coordinate data to a machine learning agent trained to provide an estimate of the machining coordinate deviation based on the determined coordinate data. If the estimated machining coordinate deviation is below the threshold coordinate deviation, the automated manipulator is allowed to continue machining the workpiece.

Description

Technical Field

[0001] The present invention relates to machining using a machining system. More particularly, but not exclusively, the present invention relates to a machining system comprising an automated manipulator configurable between a measurement configuration, in which the manipulator is capable of operating a position detector, and a machining configuration, in which the manipulator is capable of operating a tool. The present invention also relates to a method of machining using a machining system comprising such an automated manipulator. The present invention also relates to a method of training a machine learning agent for estimating machining accuracy of a machining process.

[0002] In the present invention, a machine learning agent is used to predict whether the accuracy of a downstream machining process is sufficient to allow the machining process to continue. Background Art

[0003] Machine learning is sometimes used to analyze and provide predictions, particularly in simulations or laboratory-based settings. However, machine learning techniques have not yet been fully applied to high-precision manufacturing processes, such as high-precision drilling. In particular, the use of machine learning techniques to assist in high-precision drilling on an industrial scale is unknown.

[0004] The present invention seeks to provide predictive capabilities to machining processes and machining systems through the use of machine learning agents. Summary of the Invention

[0005] According to a first aspect, the present invention provides a machining system. The machining system includes an automated manipulator that is configurable between a measurement configuration and a machining configuration, in which the position detector can be operated and in which a tool can be operated. The machining system also includes a fixture for holding a template or a workpiece and a controller. The controller is configured to cause the automated manipulator to move the position detector to at least one reference feature of the template held in the fixture in the measurement configuration to determine coordinate data associated with the at least one reference feature. The controller then provides the determined coordinate data to a machine learning agent. The machine learning agent is trained to provide an estimated machining coordinate deviation based on the determined coordinate data. In response to the estimated machining coordinate deviation being lower than a threshold coordinate deviation, the controller causes the automated manipulator to continue operating in the machining configuration.

[0006] One advantage of embodiments of the present invention is that, before the automated manipulator continues to operate in the machining configuration in which the workpiece is to be machined, a decision is made as to whether a subsequent machining coordinate deviation, indicative of machining accuracy, is likely to be acceptable, i.e., small enough, to continue machining the workpiece. This decision is based on the machining coordinate deviation predicted by the machine learning agent and an assessment of whether it is below a threshold coordinate deviation.

[0007] A machining coordinate deviation should be understood to mean the difference between the nominal machining coordinates (e.g. in X and Y Cartesian coordinates) and the actual machining coordinates. Differences may arise, for example, due to calibration drift, which is affected by factors such as temperature, wear and tear or changes in components of the automated manipulator and / or fixture. Ideally, such coordinate deviations would be zero. In practice, they are unlikely to be zero, but for a given machining process there will usually be an upper limit above which the deviation is considered too large to be acceptable. Therefore, a threshold coordinate deviation is defined, above which the machining process should not be continued and below which the machining process is allowed to continue.

[0008] In an embodiment, in a machining configuration, the controller is configured to return the automated manipulator to the position of at least one reference feature and machine a workpiece held in a fixture using a tool. The machined position corresponds to coordinate data measured by the position detector when the template is mounted on the fixture of the machining system. In other words, the automated manipulator repeats the measured feature position, but the tool is attached instead of the position detector. This takes advantage of the repeatability of the automated manipulator, i.e., its ability to return to the measured template position. However, it should be understood that once the predicted machining accuracy has been determined to be sufficient to continue machining, in an embodiment, the automated manipulator may machine a position different from or in addition to the position corresponding to the coordinate data measured by the position detector when the template is mounted on the fixture of the machining system, such as, for example, a position based on CAD data.

[0009] In an embodiment, in response to an estimated machining coordinate deviation exceeding a threshold coordinate deviation, the controller will not cause the automated manipulator to continue operating in the machining configuration. In this manner, if it is determined that downstream machining accuracy may be insufficient, the machining process is stopped or aborted. This has the advantage of reducing or eliminating scrap. Otherwise, continuing the machining process when the predicted machining accuracy is lower than the required accuracy could result in the final machined workpiece being outside the specified tolerances. This could result in an unusable workpiece, which would waste resources in terms of machine time and money.

[0010] In an embodiment, in response to the estimated machining coordinate deviation being above a threshold coordinate deviation, the controller is further configured to provide an alert to an operator of the machining system indicating that the estimated machining accuracy is insufficient to continue machining the workpiece. This alerts the operator, who may need to take steps to bring the machining system into a suitable and appropriate state to continue machining the workpiece. For example, this may involve recalibrating certain parts of the machining system. Optionally, the alert includes a hypothesis as to why the estimated machining coordinate deviation is above the threshold. This can help the operator troubleshoot the machining system or determine which component or aspect of the machining system requires intervention to restore machining accuracy to within a desired tolerance.

[0011] In an embodiment, the machining system further comprises one or more sensors configured to sense parameters of the machining system. For example, these parameters may be the temperature of the machining system or the forces present within the machining system. The controller reads the parameter values ​​from the sensors and provides the readings to the machine learning agent. Thus, in such an embodiment, the estimated machining coordinate deviation provided by the machine learning agent is based on the parameter values ​​in addition to the coordinate data described above. In this way, the accuracy or credibility of the predictions provided by the machine learning agent may be improved. This is because additional information relating to factors that may affect the machining coordinate deviation of the machining system is provided to the machine learning agent. It should be understood that the “accuracy of the predictions provided by the machine learning agent” refers to the extent to which the predictions of the machining coordinate deviations provided by the machine learning agent match the actual machining coordinate deviations of the real world that would be obtained if the machining process were allowed to continue.

[0012] In an embodiment, the machining system further comprises a memory for storing historical operating data associated with the machining system. “History” refers to operating data from previous machining operations on previous workpieces. For example, these may be hours, days, weeks, months or years prior to the current machining operation. The operating parameters may include previous coordinate data and / or previous sensor values ​​as described above. The controller reads the historical operating data from the memory and provides the historical operating data to the machine learning agent, which then further provides an estimate of the machining coordinate deviation based on the historical operating data. In this way, the accuracy of the predictions provided by the machine learning agent can be improved. This is because additional information about factors that may affect the machining coordinate deviation of the machining system over time is provided to the machine learning agent.

[0013] In an embodiment, the machining system further comprises a coordinate measuring machine (CMM). A CMM is a device that can measure the position of features on a workpiece or template using some form of contact or non-contact probing. The controller causes the CMM to measure the position of a feature or each feature of a workpiece being machined by the automated manipulator. The controller then compares the CMM measured position of the machined feature or each machined feature with the corresponding nominal position defined by the template. This comparison can be used as a verification step to verify that the machined workpiece does fall within a desired tolerance range. Additionally or alternatively, the training of the machine learning agent can be updated (improved) based on the comparison. This is because each machining run on the workpiece provides additional data, which can improve the quality of the training of the machine learning agent.

[0014] According to a second aspect of the present invention, a method of machining is also provided. The method includes first causing an automated manipulator of a machining system to move a position probe to at least one reference feature of a template held in a fixture of the machining system in a measurement configuration. Then, using the position probe, coordinate data associated with the at least one reference feature is determined. The determined coordinate data is then provided to a machine learning agent trained to provide an estimate of a machining coordinate deviation based on the determined coordinate data. Then, in response to the estimated machining coordinate deviation being below a threshold coordinate deviation, the automated manipulator is caused to continue operation in a machining configuration in which a tool can be operated to machine a workpiece.

[0015] In an embodiment of the second aspect, the method further comprises - in the machining configuration and based on the determined coordinate data - returning the automated manipulator to the position of the at least one reference feature and machining the workpiece held in the fixture using the tool.

[0016] In an embodiment of the second aspect, the method further comprises causing the CMM to measure the position of a feature or each feature of a workpiece machined by the automated manipulator. The CMM measured position of the machined feature or each machined feature is then compared to a corresponding nominal position defined by the template. The training of the machine learning agent is then updated based on the comparison.

[0017] According to a third aspect of the present invention, there is also provided a method for training a machine learning agent for estimating machining coordinate deviations of a machining process. The method comprises causing an automated manipulator of a machining system to move a position detector to at least one reference feature of a template held in a fixture of the machining system in a measurement configuration. Coordinate data associated with the at least one reference feature is then determined using the position detector. The automated manipulator is then caused to return to the position of the at least one reference feature in the machining configuration and based on the determined coordinate data, and a workpiece held in the fixture is machined using a tool. The position of a feature or each feature of the workpiece machined by the automated manipulator is then measured. The measured position of the machined feature or each machined feature is then compared with a corresponding nominal position defined by the template. The training of the machine learning agent is then updated based on the comparison. In an embodiment, the training of the machine learning agent may be updated based on data from a temperature sensor and / or a force sensor.

[0018] According to a fourth aspect of the present invention, a machining system is provided, comprising: an automated manipulator including a tool; a position-sensing probe; a sensor configured to sense a parameter of the machining system; and a controller. The controller is configured to receive at least one of: position data from the position-sensing probe; and parameter data from the sensor. The controller provides the received data to a machine learning algorithm, which is trained to provide an estimate of machining coordinate deviation based on the received data. In response to the estimated machining coordinate deviation being below a threshold coordinate deviation, the controller causes the automated manipulator to process a workpiece using the tool.

[0019] The sensor may comprise a temperature sensor and / or a force sensor.There may be a plurality of sensors.

[0020] According to a fifth aspect of the present invention, there is provided an aircraft comprising an aircraft structure machined by the machining system of the first aspect and / or the method of the second aspect. The aircraft structure may comprise all or part of a wing structure.

[0021] According to a sixth aspect of the present invention, there is provided a computer program comprising a set of instructions, which, when executed by a computer, causes the computer to perform the method of the second aspect and / or the method of the third aspect.

[0022] Of course, it will be understood that features described with respect to one aspect of the invention may be incorporated into other aspects of the invention. For example, the method of the invention may incorporate any of the features described with reference to the apparatus of the invention, and vice versa. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Embodiments of the invention will now be described, by way of example only, with reference to the accompanying schematic drawings, in which:

[0024] Figure 1 shows a schematic diagram of a processing system including an automated manipulator according to an embodiment of the present invention;

[0025] Figure 2 shows a comparison of the estimated X-axis coordinate deviation and the measured X-axis coordinate deviation;

[0026] Figure 3 is a flow chart illustrating a method of processing according to an embodiment of the present invention;

[0027] Figure 4 is a flowchart illustrating a method of training a machine learning agent for estimating machining accuracy of a machining process according to an embodiment of the present invention; and

[0028] Figure 5is an illustration of an aircraft according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] Figure 1 A schematic diagram of a machining system 100 according to an embodiment of the present disclosure is shown. The machining system 100 includes an automated manipulator 102, a template 104, a fixture 106, and a workpiece 108. The template 104 and the workpiece 108 may include planar surfaces to be machined (e.g., drilled) by the automated manipulator 102. For example, the workpiece 108 may form part of a wing structure for an aircraft.

[0030] The term "formwork" refers to an object that supports the manufacturing process but does not form part of the manufactured product. Examples of manufactured products include vehicle structures, such as aircraft structures (e.g., wing structures), and vehicles (e.g., aircraft). Conversely, the term "workpiece" refers to an object that forms part of a manufactured product. Workpieces are machined with the support of a formwork and can be assembled to other workpieces to form assemblies of workpieces. Formworks are tools that assist in the manufacture of workpieces. Fixtures 106 maintain the orientation of at least one object, such as formwork 104 or workpiece 108. Fixtures 106 are bolted to a workshop floor 110.

[0031] The automated manipulator 102 includes a carrier / spindle 112, a position detector 114, a controller 116, and a memory 118. The controller 116 is configured to operate the carrier 112. Although not shown, the carrier 112 includes an articulated arm and several joints for operating the articulated arm. The carrier 112 is configured to carry each of the position detector 114 and a tool 120, such as a drill bit. The detector 114 is unloaded when not needed, and the drill bit 120 is loaded to replace the detector 114. The process of switching between the detector 114 and the drill bit 120 is performed automatically, but the process can be performed manually. The detector 114 is a touch-sensitive detector that indicates where the detector 114 contacts a surface. The carrier 112 is a multi-axis positioning system.

[0032] Such automated manipulators 102 as described above are sometimes referred to as automated devices or automated systems and may include, but are not limited to, robots, machining tools, and parallel kinematic machines (PKMs), for example. Controller 116 may precisely control the position of carrier 112 and, therefore, the position of probe 114 or drill head 120, depending on the configuration of automated manipulator 102.

[0033] The carrier 112, the detector 114, the controller 116, and the memory 118 are electrically connected to each other to enable electrical signals to be transmitted between each other and to allow internal communication of the automated manipulator 102. The controller 116 determines coordinate data based on information received from the detector 114. The position of the detector 114 includes the position of the articulated arm about the joint and the position of the joint.

[0034] The template 104 includes a reference hole pattern 122, which in the illustrated example is formed of eleven through-holes 124, 126. Once machined, the workpiece 108 will include a matching hole pattern 128. The arrangement and number of through-holes 124, 126 in the reference hole pattern 122 may vary, and eleven through-holes 124, 126 are shown by way of example. In practice, the through-holes 124, 126 will be dispersed in a non-linear manner and will not be as uniform as the through-holes 124, 126 shown in FIG. Figure 1 The through holes 124 , 126 are arranged as a grid as shown, and the number of through holes 124 , 126 will likely exceed one hundred, but this depends on the application of the corresponding workpiece 108 .

[0035] The automated manipulator 102 is configurable between a first, measurement configuration, in which it can operate the probe 114, and a second, machining configuration, in which it can operate the drill 120. In the measurement configuration, the automated manipulator 102 collects information about the template 104, such as coordinate data for a plurality of points around the edge of each hole 124, 126. This coordinate data can be used to determine other coordinate data associated with the center of each hole. In the machining configuration, the automated manipulator 102 uses the determined coordinate data collected by measuring the template 104 to machine the workpiece 108.

[0036] In the measurement configuration, the automated manipulator 102 learns the position of each hole 124, 126 in the reference hole pattern 122 relative to a reference point or origin. First, the controller 116 causes the carrier 112 to move the probe 114 to the nominal position of the first hole 124, for example, as indicated by the CAD data. In this embodiment, the first hole 124 is located in a corner of the template 104.

[0037] For example, due to tolerances inherent in the automated manipulator 102 and / or the template 104, the actual location of the first hole 124 may differ from the nominal location of the first hole 124. Nevertheless, the automated manipulator 102 learns the actual location of the first hole 124 when starting from the nominal location. The automated manipulator 102 can accomplish this by measuring the surface surrounding the first hole 124 before automatically finding the first hole 124, entering the first hole 124, and measuring the perimeter of the first hole 124 at a fixed depth of the first hole 124, for example, at any depth up to the thickness of the template 104 (e.g., 65 mm). The controller 116 is programmed to cause the carrier 112 to automatically move the probe 114 as needed to find the first hole 124 and take the necessary measurements.

[0038] Using the information from the detector 114, the controller determines coordinate data associated with the first hole 124, such as the orientation of the first hole 124 (e.g., the orientation of the central axis of the first hole 124, although this may be derived from the overall orientation of the template 104), the location of the center of the first hole 124, and / or the diameter of the first hole 124. Controller 116 can thus bring the carrier 112 into close proximity with the location associated with the first hole 124 (and do the same for each subsequent hole 126). The automated manipulator 102 repeats the process of learning the actual location of each subsequent hole 126 in the reference hole pattern 122 in a similar manner.

[0039] In an embodiment, before any machining of the workpiece 108 begins, the controller 116 first uses a machine learning algorithm / agent to analyze coordinate data associated with one or more template holes 124, 126, so-called reference features. The machine learning agent comprises any suitable model, such as, for example, a tree-based ensemble method, a neural network, or a deep learning method. In an embodiment, the controller comprises the machine learning agent, while in other embodiments, the machine learning agent may be executed on a remote server under instructions from the controller. The machine learning agent is trained to provide, as output, an estimate of machining coordinate deviation based on the coordinate data collected by the detector 114. Machining coordinate deviation is understood to mean the difference between the nominal machining coordinates (e.g., nominal X and Y Cartesian coordinates) and the actual machining coordinates. For example, the difference may arise due to calibration drift, which is affected by factors such as temperature or variations within components of the automated manipulator 102 and / or fixture 106. If the coordinate deviation predicted by the machine learning agent is below a threshold (i.e., sufficiently small), the process of machining the workpiece 108 is allowed to continue. On the other hand, if the coordinate deviation predicted by the machine learning agent is above the threshold (ie, too large), the process of machining the workpiece 108 is not allowed to continue.

[0040] As an example, assuming that the workpiece 108 has a substantially flat machined surface in the X, Y plane, a tolerance of + / - 0.010 mm on the X and Y axes may be specified for a given machining process. If the machine learning agent estimates the coordinate deviation on the X axis to be +0.008 mm and the coordinate deviation on the Y axis to be -0.005 mm, then the machining process is allowed to continue - because the machining coordinate deviation (the magnitude) is below the threshold coordinate deviation. On the other hand, if the machine learning agent estimates the coordinate deviation on the X axis to be +0.012 mm and the coordinate deviation on the Y axis to be -0.017 mm, then the machining process is not allowed to continue. It should be understood that different tolerances / thresholds may be allowed for the X, Y, and Z axes, and that the term "machining coordinate deviation" as used herein refers to deviations on one or more of the X, Y, and Z axes.

[0041] If the process of machining workpiece 108 is permitted to continue, automated manipulator 102 is placed in a machining configuration, enabling drill head 120 to operate to machine workpiece 108. Furthermore, template 104 is removed from fixture 106, and workpiece 108 is mounted on fixture 106 in place of template 104. Once workpiece 108 is in position on the fixture and automated manipulator 102 is in the machining configuration, controller 116 causes carrier 112 to move to the location of workpiece 108 for drilling first hole 124'. Thus, carrier 112 is placed in the same position associated with first hole 124 in template 104 based on the determined coordinates of the hole center of first hole 124 in template 104. This allows automated manipulator 102 to replay the position of carrier 112, thereby leveraging the repeatability of the automated manipulator's 102 mechanical work. Once carrier 112 is in position, controller 116 causes drill head 120 to operate to drill hole 124' in workpiece 108. Once drilled, the first hole 124 of the template 104 is effectively transferred to the workpiece 108. The automated manipulator 102 repeats these steps for each of the plurality of holes 124, 126 of the reference hole pattern 122 to produce a matching hole pattern 128 in the workpiece 108.

[0042] If the process of machining workpiece 108 is not permitted to continue, i.e., because the coordinate deviation predicted by the machine learning agent is above a threshold, the workpiece 108 is not machined. Instead, the controller 116 provides some form of alert to the operator of the machining system 100. The alert may include a hypothesis as to why the coordinate deviation is outside of acceptable limits. This, in turn, prompts the operator to investigate the potential cause and implement corrective procedures to bring the machining system back into tolerance.

[0043] In addition to the probe 114, in an embodiment, one or more additional sensors 130, 132 are provided, for example, placed in and around the automated manipulator 102 and / or fixture 106. These additional sensors may include temperature sensors and / or force sensors. The sensors 130, 132 communicate with the controller 116 so that readings from the sensors 130, 132 can be fed to the machine learning agent along with the coordinate data obtained by probing the template 104. In such an embodiment, the machine learning agent is trained to further provide an estimate of the process coordinate deviation based on the sensor readings.

[0044] In some embodiments, memory 118 stores historical operational data associated with the machining system. This may include data such as historical temperature data from temperature sensors 130, 132; historical force data from force sensors 130, 132; and historical coordinate data obtained from previous probes of template 104 by position probe 114. Controller 116 reads the historical operational data from memory 116 and feeds it to the machine learning agent along with current / real-time data, such as current probe data received from probe 114 probing template 104 and current temperature data received from one of sensors 130, 132. In such embodiments, the machine learning agent is trained to further provide an estimate of machining coordinate deviation based on the historical data.

[0045] In an embodiment, the machining system includes a coordinate measuring machine (CMM) 140. The CMM can be contact-based (e.g., based on capacitive inspection) or non-contact (e.g., based on optical measurement technology). The CMM 140 is typically used to verify the workpiece 108 once it has been machined by the automated manipulator 102. It should be understood that not every workpiece 108 must be verified. For example, every second or tenth workpiece 108 may be transferred for verification by the CMM 140. Using the CMM 140, features 124', 128 machined in the workpiece 108 by the automated manipulator 102 can be measured, and the positions of these measurements can be compared with nominal positions, such as, for example, the nominal positions defined by the template 104. Ideally, the measured positions will correspond exactly to the nominal positions, but in practice there may be small offsets. These measured offsets can be fed back to the machine learning agent to supplement the training of the machine learning agent, thereby improving its ability to make future estimates of machining coordinate deviations.

[0046] Reference Figure 2To evaluate the effectiveness of the estimated / predicted coordinate deviations provided by the machine learning agent, an experiment was conducted in which a machined workpiece 108 was analyzed using a CMM 140 to determine the true coordinate deviation of each machined feature (indexed 0 to 25) relative to the nominal value. Figure 2 The dark grey line in (labeled “A”) corresponds to the X-axis deviation (in mm) of each feature predicted by the machine learning agent. Figure 2 The light grey lines in FIG (labeled “B”) correspond to the true X-axis deviation of each feature obtained by measuring the workpiece 108 by the CMM 140. As can be seen, the predicted deviations and the measured deviations are in close agreement, with the deviations being within acceptable tolerances.

[0047] Figure 3 A method 300 of machining according to an embodiment of the present disclosure is shown. The method 300 includes causing 301 an automated manipulator 102 of a machining system 100 to move a position detector 114 to at least one reference feature 124, 126 of a template 104 held in a fixture 106 of the machining system 100 in a measurement configuration. The method 300 includes determining 302, using the position detector 114, coordinate data associated with the at least one reference feature 124, 126. The method 300 includes providing 303 the determined coordinate data to a machine learning agent trained to provide an estimate of a machining coordinate deviation based on the determined coordinate data. The method 300 includes, in response 304 to the estimated coordinate deviation being below a threshold coordinate deviation, causing the automated manipulator 102 to continue operation in a machining configuration in which the tool 120 is operable to machine the workpiece 108.

[0048] Figure 4A method 400 of training a machine learning agent for estimating machining accuracy of a machining process according to an embodiment of the present disclosure is shown. The method 400 includes causing 401 an automated manipulator 102 of a machining system 100 to move a position probe 114 to at least one reference feature 124, 126 of a template held 104 in a fixture 106 of the machining system 100 in a measurement configuration. The method 400 includes determining 402, using the position probe 114, coordinate data associated with the at least one reference feature 124, 126. The method 400 includes causing 403 the automated manipulator 102 to return to the position of the at least one reference feature 124, 126 in the machining configuration and based on the determined coordinate data, and machining a workpiece 108 held in the fixture 106 using a tool 120. The method 400 includes measuring 404 the position of the or each feature 124′ of the workpiece 108 being machined by the automated manipulator 102. The method 400 comprises comparing 405 the measured position of the or each machined feature 124' with a corresponding nominal position defined by the template 104. The method 400 comprises updating 406 the training of the machine learning agent based on the comparison.

[0049] Figure 5 An aircraft 500 according to an embodiment of the present disclosure is shown. Aircraft 500 includes a wing 501, a center wing box (CWB) location 503, an outer wing box (OWB) location 509, a root interface location 505 between the CWB and the OWB, and a fuselage 507. The CWB is part of the fuselage 507, and the OWB 505 is part of the wing 501. The wing 501 is secured to the fuselage 507 by coupling the CWB and the OWB at the root interface. Thus, each of the OWB and the CWB is an example of a wing structure. All or part of a wing structure can be manufactured using the machining systems disclosed herein and / or according to embodiments of the machining methods disclosed herein.

[0050] Although the present invention has been described and illustrated with reference to specific embodiments, it will be understood by those skilled in the art that the invention lends itself to many different variations not specifically described herein. Some possible variations will now be described by way of example only.

[0051] Each of the above methods may be included in a device including a processor or a processing system or implemented in these devices. The processing system may include one or more processors and / or memories. One or more aspects of the embodiments described herein include processes performed by a device. In some examples, the device includes one or more processing systems or processors configured to perform these processes. In this regard, the embodiments may be implemented at least in part by computer software stored in a (non-transitory) memory and executable by a processor, or by hardware, or by a combination of tangibly stored software and hardware (and tangibly stored firmware). The embodiments also extend to computer programs, particularly computer programs on or in a carrier, which are suitable for putting the above embodiments into practice. The program may be in the form of non-transitory source code, object code, or any other non-transitory form suitable for use in the implementation of the process according to the embodiment. The carrier may be any entity or device capable of carrying a program, such as a RAM, ROM, or optical storage device.

Claims

1. A processing system comprising: an automated manipulator configurable between a measurement configuration in which it can operate a position detector and a machining configuration in which it can operate a tool; a holder for holding a template or a workpiece; and A controller configured to: causing the automated manipulator to move the position probe to at least one reference feature of the template held in the fixture in the measurement configuration to determine coordinate data associated with the at least one reference feature; providing the determined coordinate data to a machine learning agent trained to provide an estimate of a process coordinate deviation based on the determined coordinate data; and In response to the estimated coordinate deviation being below a threshold coordinate deviation, continuing operation of the automated manipulator in the machining configuration.

2. The processing system according to claim 1, wherein: In the machining configuration, the controller is configured to return the automated manipulator to the location of the at least one reference feature based on the determined coordinate data and machine the workpiece held in the fixture using the tool.

3. The processing system according to claim 1 or 2, wherein: In response to the estimated coordinate deviation being above the threshold coordinate deviation, the controller does not cause the automated manipulator to continue operation in the machining configuration.

4. The processing system according to claim 3, wherein: The controller is further configured to provide an alert to an operator of the machining system that the estimated coordinate deviation is above the threshold coordinate deviation.

5. The processing system according to claim 4, wherein: The alert includes an indication of a hypothesis as to why the estimated coordinate deviation is above the threshold coordinate deviation.

6. The processing system according to any preceding claim, further comprising at least one sensor configured to sense a parameter of the processing system, wherein The controller is further configured to: reading a parameter value from the at least one sensor; and The parameter values ​​are provided to the machine learning agent, which is trained to provide an estimate of the process coordinate deviation further based on the parameter values.

7. The processing system according to claim 6, wherein: The at least one sensor comprises a temperature sensor, and the parameter comprises a temperature of the processing system.

8. The processing system according to claim 6 or 7, wherein: The at least one sensor comprises a force sensor, and the parameter comprises a force present within the machining system.

9. The processing system of any preceding claim, further comprising a memory for storing historical operating data associated with the processing system, wherein The controller is further configured to: reading the historical operation data from the memory; and The historical operating data is provided to the machine learning agent, which is trained to provide an estimate of the machine coordinate deviation further based on the historical operating data.

10. The processing system according to claim 9, wherein: The historical operational data includes one or more of the following: Historical temperature data; Historical force data; and Historical coordinate data associated with the at least one reference feature.

11. A machining system according to claim 2 or any claim directly or indirectly dependent thereon, further comprising a coordinate measuring machine (CMM), wherein The controller is further configured to: causing the coordinate measuring machine to measure the position of the or each feature of the workpiece machined by the automated manipulator; comparing a CMM measured position of the or each machined feature with a corresponding nominal position defined by the template; and Based on the comparison, the training of the machine learning agent is updated.

12. A processing system according to any preceding claim, wherein: The automated manipulator comprises a parallel kinematic machine, or PKM.

13. A processing method, comprising: causing an automated manipulator of a machining system to move a position probe to at least one reference feature of a template held in a fixture of the machining system in a measurement configuration; determining coordinate data associated with the at least one reference feature using the position detector; providing the determined coordinate data to a machine learning agent trained to provide an estimate of a process coordinate deviation based on the determined coordinate data; and In response to the estimated coordinate deviation being below a threshold coordinate deviation, the automated manipulator is caused to continue operation in a machining configuration in which a tool is operable to machine a workpiece.

14. The method of claim 13, further comprising returning the automated manipulator to the location of the at least one reference feature in the machining configuration and based on the determined coordinate data, and machining the workpiece held in the fixture using the tool.

15. The method according to claim 14, further comprising: causing a coordinate measuring machine (CMM) to measure the position of the or each feature of the workpiece machined by the automated manipulator; comparing a CMM measured position of the or each machined feature with a corresponding nominal position defined by the template; and The training of the machine learning agent is updated based on the comparison.

16. A method for training a machine learning agent for estimating machining accuracy of a machining process, the method comprising: causing an automated manipulator of a machining system to move a position probe to at least one reference feature of a template held in a fixture of the machining system in a measurement configuration; determining coordinate data associated with the at least one reference feature using the position detector; in a machining configuration and based on the determined coordinate data, returning the automated manipulator to the location of the at least one reference feature and machining a workpiece held in the fixture using a tool; measuring the position of the or each feature of the workpiece machined by the automated manipulator; comparing a measured position of the or each machined feature with a corresponding nominal position defined by the template; and The training of the machine learning agent is updated based on the comparison.

17. A processing system comprising: an automated manipulator comprising a tool; Position sensing detectors; a sensor configured to sense a parameter of the machining system; as well as A controller configured to: receiving at least one of position data from a position detector and parameter data from the sensor; providing the received data to a machine learning algorithm, the machine learning algorithm being trained to provide an estimate of the process coordinate deviation based on the received data; and In response to the estimated coordinate deviation being below a threshold coordinate deviation, the automated manipulator is caused to process a workpiece using the tool.

18. The processing system according to claim 17, wherein: In response to the estimated coordinate deviation being above the threshold coordinate deviation, the controller does not cause the automated manipulator to process the workpiece using the tool.

19. The processing system according to claim 17 or 18, wherein: The sensor includes a temperature sensor and / or a force sensor.

20. An aircraft comprising an aircraft structure machined by the machining system according to any one of claims 1 to 12 and / or the method according to any one of claims 13 to 15.

21. The aircraft according to claim 20, wherein: The aircraft structure includes all or part of a wing structure.

22. A computer program comprising a set of instructions which, when executed by a computer, cause the computer to perform the method of any one of claims 13 to 16.

Images