Wind turbine blade manufacturing using intelligent imaging

Abstract

A manufacturing system for a process of manufacturing a wind turbine blade portion, such as a blade half or a segment of a blade half, such as a composite fiber wind turbine blade. The imaging system has a camera (CM) arranged at a raised position to capture an image of the manufacturing area (AM). The processor system (PS) is arranged to process the captured image from the imaging system to analyze the captured image (IA) at least with respect to texture and / or contrast. Based on the image analysis, an anomaly (AD) on a portion of the wind turbine blade is detected during the manufacturing process, the anomaly is classified as one (ACL) of a plurality of predetermined types of anomalies according to the analysis of the texture, and a location (AL), such as both spanwise and chordwise positions, of the classified anomaly on the wind turbine blade is further determined. A user interface is connected to the processor system to communicate the anomaly type and the location on the wind turbine blade portion in real time to an operator of the manufacturing process.

Description

Technical Field

[0001] This invention relates to the field of manufacturing wind turbine blades or parts thereof. In particular, the invention provides a method for using intelligent imaging to assist operators in detecting errors or anomalies in the manufacturing of wind turbine blades during the manufacturing process. Background Technology

[0002] The manufacture of large wind turbine blades (such as composite fiber blades) is a complex process involving many activities or states. Due to their large size (e.g., lengths greater than 50m), it is a complex task for operators to monitor the entire object at any time during the manufacturing process and detect any possible errors or anomalies. Summary of the Invention

[0003] Therefore, based on the above description, the object of the present invention is to provide a system and method for manufacturing wind turbines that allows operators to easily monitor the manufacturing of large wind turbine blades and allows for rapid initialization of actions in the event of errors detected during the manufacturing process.

[0004] In a first aspect, the present invention provides a manufacturing system for a process of manufacturing wind turbine blade portions (such as blade halves or segments of blade halves), such as composite fiber wind turbine blade portions, the system comprising:

[0005] - An imaging system comprising at least one camera and arranged to capture images of the area where the wind turbine blade section is manufactured, such as being arranged to capture images covering the entire area of ​​the wind turbine blade section.

[0006] - The processor system, which is arranged as follows:

[0007] - Process images captured from the imaging system to analyze the captured images at least with respect to texture and / or contrast.

[0008] - Anomalies on a portion of a wind turbine blade are detected during the manufacturing process based on texture analysis of captured images using anomaly detection algorithms.

[0009] - Based on the analysis of texture and / or contrast of the captured image using an anomaly classification algorithm, the anomaly is classified into one of several predetermined anomaly types, and

[0010] - At least regarding the first axis, such as regarding both axes, determine the location of the classified anomalies on the wind turbine blades, and

[0011] - A user interface operably connected to the processor system, wherein the user interface is configured to transmit the anomaly type of detected anomalies and / or their location on a wind turbine blade section to the operator of the manufacturing process in real time.

[0012] Image segmentation methods can be used to perform texture and contrast analysis.

[0013] Real-time operation should be understood as transmission (communication) being performed instantaneously or within the inherent time delay of the processor system. It should also be understood that information such as anomaly type or location is transmitted, for example, intermittently displayed according to a certain update rate, rather than continuously transmitted. Real-time operation may also require performing some or all of the steps of processing captured images, detecting anomalies, classifying, and determining locations within certain time constraints (such as within the inherent time delay of the processor system). In this way, potential anomalies can be communicated to the operator during manufacturing, allowing corrections to be made as soon as an anomaly is detected, and of course, during manufacturing, rather than waiting until after manufacturing.

[0014] This system has been found to be a highly efficient tool for ensuring error-free (defect-free) manufacturing of wind turbine blades and large blades (e.g., composite fiber blades involving multiple manufacturing steps including layup, resin infusion, etc.). Anomalies can be alerted to the operator, for example, via a wireless tablet or smartwatch, or through other visual and / or auditory alarms, to draw the operator's attention to the anomaly and its location. It has been found that cameras can be positioned in an elevated position relative to the manufacturing area so that they do not interfere with manufacturing operations while still capturing images with sufficient detail to allow, for example, the detection of wrinkles during the layup process.

[0015] In a preferred embodiment, the image processing and anomaly detection and / or classification steps are trained using feedback from the operator. For example, the operator can reject anomalies as "non-anomaly," "unimportant," or "yes / important," or correct the anomaly type if it is not detected correctly. Therefore, the anomaly classification and localization capabilities can be improved.

[0016] In a preferred embodiment, images can be analyzed to provide automatic detection of which manufacturing step or stage is currently in progress. Therefore, anomaly detection can be improved because various algorithms or parameters of the same algorithm can be used to detect and classify anomalies in response to actually detected manufacturing steps or stages. For example, different anomalies are expected for laying and grouting processes.

[0017] For example, the system can be used as a planning tool. Activity tracking or detection of manufacturing steps can be combined with calculations of the average duration of different activities based on past blade counts and total blade build time to provide more accurate forecasts of upcoming blade production plans compared to traditional production planning.

[0018] Preferred features and embodiments will be described below.

[0019] The processor system is preferably arranged to automatically detect manufacturing activities or manufacturing states in response to analysis of images captured from the imaging system. Specifically, the system may be arranged to automatically detect a current manufacturing activity or state based on selecting one of a plurality of predefined manufacturing activities or states as current, such as selecting one of at least six predefined manufacturing activities or states, such as selecting one of the following: 1) mold preparation, 2) gel coating application, 3) outer layup, 4) core board, 5) inner layup, 6) vacuum bagging, 7) resin infusion, 8) shell curing heating blanket, 9) bag removal, 10) inspection and bonding surface preparation, 11) adhesive curing, and 12) demolding. In particular, the processor system may be arranged to select an anomaly detection algorithm and / or anomaly classification algorithm in response to the detected manufacturing activity or manufacturing state. Alternatively or additionally, the processor system may be arranged such that the imaging system adjusts its image capture with respect to at least one parameter (such as image resolution and / or image capture frequency) in response to the detected manufacturing activity or manufacturing state. Therefore, image capture and / or image analysis algorithms behind automated manufacturing activity or state detection, anomaly detection and classification can be applied to achieve optimal performance for a given activity or state in the manufacturing process.

[0020] The processor system is preferably arranged to classify anomalies into one of the following: 1) wrinkles in sheet or laminate layup, 2) ventilation or dry spots in post-injection laminates, 3) deviations from the centerline of the spar cap and adhesive, 4) wrinkles in vacuum bags and / or foils, 5) air bubbles in vacuum bags during injection, and 6) positioning deviations from the design specifications of components of the wind turbine blade section (such as web positioners, webs, lightning protection system wiring and fixtures).

[0021] In a preferred embodiment, the user interface is arranged to receive feedback from the operator on detected anomalies, and the user interface is arranged to transmit the feedback to the processor system. Such feedback can indicate the relevance of the anomaly detected after inspection by the operator; for example, the operator can provide feedback as “irrelevant,” “relevant,” “highly relevant,” or “non-anomaly” in response to an anomaly detected by the system and transmitted to the operator. Specifically, the processor system is arranged to receive feedback in response to detected anomalies and process the feedback according to a learning algorithm used to improve the accuracy of anomaly detection and / or anomaly classification algorithms in response to the feedback. In particular, the learning algorithm involves neural networks, such as convolutional neural networks and / or recurrent neural networks. In this way, the performance of anomaly detection and classification can be improved over time, and further, the system can adapt to different changes in visual input to the system that may alter the manufacturing setup.

[0022] The processor system can be arranged to determine the positioning of classified anomalies on the wind turbine blade section with respect to the longitudinal axis (span) of the wind turbine blade section and with respect to the axis perpendicular to the longitudinal axis (chord).

[0023] In some embodiments, the user interface includes a display, such as a display on a portable device (such as a tablet, smartphone, or smartwatch), wherein the type of anomaly detected and its location on a wind turbine blade section are conveyed via symbols and / or text overlays on a captured photograph of the wind turbine blade section. The display may include a large display visible to the operator from multiple locations. In particular, the system may also be arranged to calculate a value indicating the confidence level of the detected anomaly, and this value may also be indicated to the operator on the display.

[0024] The system can be configured to provide visual and / or auditory alarms to the operator when an anomaly is detected. This allows the operator to quickly assess the anomaly and determine if any action is necessary.

[0025] In some embodiments, the user interface includes a projection system, such as a laser projector, arranged to provide a visual indication of the detected anomaly directly at a defined location on the surface of a wind turbine blade section. This type of anomaly indication can be a quick way to indicate anomalies on large objects.

[0026] In some embodiments, the user interface is configured to show or indicate the location of the detected anomaly. In particular, the user interface may be arranged to indicate the location along an axis (such as the spanwise axis) where the detected anomaly is located on a blade portion.

[0027] In particular, the user interface can be configured to indicate only the location of detected anomalies.

[0028] In particular, the user interface can be arranged to indicate both the location and type of the detected anomaly.

[0029] In some embodiments, the user interface includes a controllable line of multiple visual indicators (such as multiple LED light indicators arranged in a line (arranged in a row)) arranged along a portion of the wind turbine blade to provide a visual indication of the longitudinal location of a detected anomaly. Thus, the operator can be quickly notified of the detected anomaly and its longitudinal location, and the operator can move to inspect the anomaly. As described above, such a line of visual indicators (e.g., lights of different colors) can be used in conjunction with a projection of the precise location of the anomaly and / or by a display, for example, on a portable device. For example, the controllable line of visual indicators can be arranged adjacent to the mold, thus along the portion of the wind turbine blade, such as in a strip of lights extending along the spanwise direction of the mold. For example, the controllable line of visual indicators can be arranged above the mold for high visibility.

[0030] In some embodiments, the manufacturing system is arranged to perform automatic resin infusion tracking in response to captured images during a resin infusion process activity on a wind turbine blade section. This may include performing edge-tracking image analysis to track the resin leading edge in the captured images during the infusion process. Resin infusion can be understood as a process in which a vacuum-sealed foil is arranged on the blade section and the fiber layer, then infusion resin is supplied to the fiber layer under the vacuum foil, and a further vacuum is applied such that the resin is distributed into the fiber material. In particular, automatic resin infusion tracking may involve performing an analysis of the infusion fill percentage over time during the resin infusion process activity. In particular, automatic resin infusion tracking may involve transmitting information to an operator via a user interface when it is estimated that the shell infusion has reached 100% or nearly 100%. In particular, a processor system is operatively connected to control a resin infusion valve, such as controlling the resin infusion valve to close in response to automatic resin infusion tracking, such as controlling the resin infusion valve to close in response to an estimated shell infusion having reached 100% or nearly 100%.

[0031] Resin infusion tracking can be based on image analysis, where the coverage of infused and non-infused areas is determined to calculate the fill percentage. Additionally or alternatively, the processor system can be arranged to control the resin infusion valve (such as a valve with variable flow control) to control the resin flow rate based on the fill percentage, for example, causing the flow rate to automatically decrease to zero based on the fill percentage.

[0032] In one embodiment, the manufacturing system is specifically arranged to perform an automated resin infusion tracking process. According to this embodiment, a processor system is arranged to process captured images to determine edge tracking based on texture or contrast analysis. One or more of the steps of detecting anomalies on a portion of a wind turbine blade, classifying the anomalies into one of a plurality of predetermined types of anomalies, and determining the location of the classified anomaly may optionally be unnecessary. Similarly, the user interface may not be part of the manufacturing system, or the user interface may be configured to alternatively or additionally convey the infusion process, for example, by displaying the infusion leading edge to the operator.

[0033] In the above description, "resin infusion" is used, but it should be understood that other infusion processes may also be used, and the features of the processor system mentioned above still apply to the infusion of other fluids.

[0034] The imaging system may include at least one camera positioned and configured to provide an image of the entire width of a wind turbine blade section, such as by positioning the at least one camera at a fixed height above the surface of the wind turbine blade section, which is 0.8 to 2.0 times the width of the wind turbine blade section. This has been found to be suitable for providing a trade-off between undisturbed manufacturing areas and the level of detail in images captured with a low-cost camera.

[0035] Preferably, the imaging system includes a plurality of cameras positioned and configured to provide images of corresponding spanwise portions of a wind turbine blade section, such as providing images covering spanwise overlapping areas of the wind turbine blade section.

[0036] In a second aspect, the present invention provides the use of the manufacturing system according to the first aspect for manufacturing wind turbine blades.

[0037] In a third aspect, the present invention provides a method for manufacturing a wind turbine blade portion (such as a blade half or a segment of a blade half), such as a composite fiber wind turbine blade portion, the method comprising:

[0038] - Capture images of the area where the wind turbine blade section is manufactured, such as capturing images covering the entire area of ​​the wind turbine blade section.

[0039] -Analyze the captured image at least regarding texture and / or contrast.

[0040] - Anomalies on a portion of a wind turbine blade were detected during the manufacturing process based on texture analysis of captured images.

[0041] - Based on the analysis of the texture and / or contrast of the captured image, the anomaly is classified into one of several predefined anomaly types, and

[0042] - At least regarding the first axis, such as regarding both axes, determine the location of the classified anomalies on the wind turbine blade section, and

[0043] - The detected anomalies and their locations on the wind turbine blade sections are transmitted in real time to the operators in the manufacturing process.

[0044] The method may include initiating actions during the manufacturing process to remedy detected anomalies, such as actions initiated automatically by the system, such as stopping an ongoing process, or actions initiated by an operator.

[0045] The method may also include receiving feedback from the operator to the processor system in response to a transmitted anomaly. Specifically, the method may further include training a neural network involved in at least detecting or classifying the anomaly in response to feedback from the operator.

[0046] The method may also include detecting manufacturing activity or manufacturing status in response to analysis of the captured images.

[0047] In a fourth aspect, the present invention provides a method for manufacturing wind turbine blades, the method comprising the steps of the method according to the third aspect.

[0048] In a fifth aspect, the present invention provides a wind turbine rotor blade manufactured according to the method of the fourth aspect.

[0049] In a sixth aspect, the invention provides a wind turbine comprising at least one rotor blade according to the fifth aspect, such as comprising three rotor blades. Specifically, the wind turbine blades can be arranged to drive a generator located within a nacelle disposed on top of a tower. Specifically, the wind turbine can generate at least 1 MW, such as 2-10 MW or greater than 10 MW of electrical power.

[0050] In a seventh aspect, the present invention provides a manufacturing system for a process of manufacturing wind turbine blade portions (such as blade halves or segments of blade halves), such as composite fiber wind turbine blade portions, the system comprising:

[0051] - An imaging system comprising at least one camera and arranged to capture images of the area where the wind turbine blade section is manufactured, such as being arranged to capture images covering the entire area of ​​the wind turbine blade section.

[0052] - The processor system, which is arranged as follows:

[0053] - Process the images captured from the imaging system to analyze the captured images, such as analyzing the captured images with at least some regard to edge detection, and

[0054] - Automatic infusion tracking is performed in response to the analysis of captured images during the resin infusion process on the wind turbine blade section, such as resin leading edge tracking by edge tracking in response to captured images during the infusion process.

[0055] Specifically, automated resin filling tracking can involve performing an analysis of the filling percentage over time during resin filling process activities. Specifically, automated resin filling tracking can involve transmitting information to the operator via a user interface when it is estimated that the housing filling has reached 100% or nearly 100%. Specifically, a processor system can be operatively connected to control the resin filling valve, such as controlling the resin filling valve to close in response to automated resin filling tracking, such as controlling the resin filling valve to close when it is estimated that the housing filling has reached 100% or nearly 100%.

[0056] In the above description, "resin infusion" is used, but it should be understood that other infusion processes may also be used, and the features of the processor system mentioned above still apply to the infusion of other fluids.

[0057] It should be understood that the same advantages, preferred embodiments, and features described with respect to the first aspect also apply to the second aspect, and these aspects can be combined in any way. Attached Figure Description

[0058] The invention will now be described in more detail with reference to the accompanying drawings, in which:

[0059] Figure 1a and Figure 1b The diagram illustrates a simplified setup for wind turbine blade manufacturing, with cameras installed to capture images of the entire manufacturing area.

[0060] Figure 2a and Figure 2b The diagram illustrates block diagrams of two different system configuration embodiments.

[0061] Figure 3 The diagram illustrates a simplified representation of the components in a system embodiment.

[0062] Figure 4 The illustration shows an example of a photograph of a wind turbine blade during the installation process, where wrinkles have been detected and their location transmitted to the operator.

[0063] Figure 5 The illustration shows an example of a user interface that includes a row of light sources arranged along the spanwise direction of the wind turbine blades to indicate the location of anomalies to the operator.

[0064] Figure 6 The steps of an embodiment of the method are illustrated.

[0065] The accompanying drawings illustrate specific ways of implementing the invention and should not be construed as limiting oneself to other possible embodiments falling within the scope of the appended claims. Detailed Implementation

[0066] Figure 1a The diagram illustrates a simplified cross-sectional view of the manufacturing area M_A where two wind turbine blade halves, BL1 and BL2, are being manufactured. The blade halves are manufactured, for example, in a blade mold (not shown) via a vacuum-assisted resin transfer molding (VARTM) process, in which a layer of fibrous material is initially laid in the mold. A vision system, including cameras CM, is arranged at an elevated position above the blade halves BL1 and BL2. Two lines from each camera CM indicate the image coverage area, and as shown, the left camera CM can capture an image that completely covers the chordal direction (i.e., along the Y-axis) of the left blade halves BL1, while the right camera CM can capture an image that completely covers the chordal direction of the right blade halves BL2.

[0067] Figure 1b The diagram illustrates a simplified cross-sectional view of the manufacturing region M_A, where the blade portion BL is seen along the spanwise direction (i.e., along axis X). Here, the lines from the cameras CM are used to cover the entire blade portion BL along the spanwise direction, and as shown by the two lines from each camera CM, the entire spanwise blade portion BL is covered by an overlapping image from the camera CM.

[0068] use Figure 1a and Figure 1b The camera setup shown has a vision system that completely covers the manufacturing area M_A of the blade sections BL, BL1, and BL2.

[0069] Figure 2a A block diagram of a system embodiment is shown. A vision system, including a camera (CM), provides captured images to a processor system (PS). The processor system is programmed to execute image analysis algorithms (IA), which preferably include analysis of the captured images regarding texture and contrast, for example, where each image is analyzed individually in multiple segments. The image analysis results are applied to an anomaly detection algorithm (AD), an anomaly classification algorithm (ACL), and an anomaly localization algorithm (AL). The functions of these algorithms (AD, ACL, AL) are to determine the presence of one of several predetermined types of anomalies based on the input from the image analysis IA.

[0070] If an anomaly is detected, its type and location on the blade section are determined, and then the anomaly type (AT) and location (LC) are transmitted to the user interface (UI) to notify the operator of the anomaly. This allows the operator to investigate the anomaly and initiate appropriate actions to remedy it if necessary.

[0071] The operator can provide feedback (UFB) on transmitted anomalies (AT, LC) via the user interface (UI). The operator can indicate whether the anomaly was detected correctly and can also rank anomalies based on relevance, providing feedback to the processor system (PS) for application to the neural network-based learning algorithm (NN). This learning algorithm (NN) can improve the performance of at least the algorithms (AD, ACL) over time and allows anomaly detection and classification to adapt to various changes in the physical manufacturing setup.

[0072] The user interface (UI) can be implemented using one or more of the following: a smartwatch, tablet, large display, warning light, acoustic alarm, a row of lights indicating the location of anomalies, or a projection system that visually indicates anomalies on the surface of a wind turbine blade section. In the case of a smartwatch, tablet, or large display, the display can be used to indicate a photograph, which uses graphics and optionally text indicating the type of anomaly to show the entire blade section or only a portion of the blade section indicating anomalies.

[0073] Based on the captured images, the processor system PS is able to determine the actual manufacturing activity or state. This allows the system to tune various parameters, such as those for monitoring specific anomalies associated with a particular manufacturing activity or state, and thus select a specific algorithm in response to the determined manufacturing activity or state. Furthermore, the resolution, optical lens, or frame rate of the images captured by the camera CM can be specifically selected for each manufacturing activity or step to provide the best possible performance for anomaly detection and classification.

[0074] Figure 2b Another system embodiment is shown that provides monitoring functions related to blade manufacturing steps or resin infusion activities. This embodiment can be used with... Figure 2a The embodiments can be a combination of examples, or they can be independent.

[0075] A vision system, including a camera (CM), provides captured images to a processor system (PS). The processor system is programmed to execute an image analysis algorithm (IA), which preferably includes analysis of the captured images regarding at least edge tracking or edge detection. The image analysis results are applied to a resin infusion tracking algorithm (INF_TR), which determines which part of the resin infusion process has been completed, for example, by indicating the percentage of resin infusion completion via a user interface, thereby informing the operator of the resin infusion progress. In this embodiment, the output of the infusion tracking algorithm (INF_TR) is used to control the function of the infusion valve (INF_V), i.e., opening or closing the infusion valve (INF_V), which controls the application of resin during the infusion process. In this way, an automated infusion process can be provided because the system can identify when the infusion process is close to 100% via edge tracking, and therefore control the infusion valve (INF_V) to close accordingly.

[0076] In addition, the INF_TR algorithm for perfusion tracking can also identify dry spots, bubbles or other anomalies during the perfusion process and notify the operator accordingly.

[0077] Figure 3 A simplified diagram of the possible components of the system is shown. Cameras CM or camera arrays are used to cover the manufacturing area M_A where the wind turbine blades are manufactured. Each camera CM may have an optical lens, such as a controllable optical lens. The camera CMs are connected to a server SRV, which performs image processing to detect, classify, and locate anomalies. The user interface is here designated as the visual monitor V_M, and the feedback path is here designated as a PLC connected to the server SRV and wirelessly connected to a smartwatch FB_D worn by the operator.

[0078] The smartwatch FB_D can present text labels TL to the operator to indicate anomalies, such as "CW fold @R15,5" indicating a chordal fold detected at a spanwise position of 15,5m from the reference point. The user can then provide feedback via PLC to the server SRV, which can then feed the feedback into a learning algorithm to improve anomaly detection and classification accordingly.

[0079] A visual monitor (V_M) can be used to present a photograph of a setup with indications of detected anomalies and their types. For example, a visual monitor (V_M) can be a display on a tablet, smartphone, large monitor, or computer monitor.

[0080] Figure 4 An example of output to be presented to the operator on a display is shown, such as a photograph (indicated by dashed lines) of a surface of a wind turbine blade section BL with a portion of a text message overlaid on it. Figure 4 For simplicity, only a simplified diagram of the blade section BL is shown, not a photograph.

[0081] In this example, wrinkles are indicated by a black dashed box on the photo, and the related text “Wri154%” indicates that the detected anomaly is a wrinkle, and the system has calculated the confidence level of detecting the wrinkle to be 54%.

[0082] Figure 5 Another example of the operator's output is shown, namely a row of light indicators LLS set along the spanwise direction of the wind turbine blade section BL, illustrated here in a simplified diagram. The light sources (e.g., LEDs) are controlled to indicate the location of an anomaly in the spanwise direction. For example, all light sources could be green or turned off in all normal spanwise positions, with anomalies indicated by light or red light. In the example shown, light indicators L1, L2, and L3 indicate the detected spanwise location of an anomaly, indicated here by a black dashed box.

[0083] This type of light array LLS will allow operators to easily locate areas with abnormalities L1, L2, L3 and move to that location for manual inspection.

[0084] For example, the user interface could simply indicate a detected anomaly via a light indicator (LLS), allowing the operator to move around to search for the anomaly. In other variations, additional information is provided to the operator via the user interface, such as through a mobile terminal or a wall-mounted display. This additional information could indicate the type of anomaly detected.

[0085] Figure 6 The illustration depicts steps of an embodiment of a method for manufacturing a wind turbine blade portion (such as a blade half or a segment of a blade half), such as a composite fiber wind turbine blade portion. The method includes capturing an image C_I of an area where the wind turbine blade portion is manufactured, such as capturing an image covering the entire area of ​​the wind turbine blade portion. Furthermore, the method includes analyzing the captured image A_I at least with respect to texture and / or contrast, such as analyzing the texture and / or contrast in multiple segments of the captured image.

[0086] Furthermore, based on the analysis of the texture of the captured image, anomalies D_A are detected on a portion of the wind turbine blade section during the manufacturing process. Additionally, based on the analysis of the texture and / or contrast of the captured image, the anomalies are classified into one of several predetermined types of anomalies C_A. Furthermore, at least with respect to a first axis, such as the spanwise and chordwise positions, the location D_L of the classified anomalies on the wind turbine blade section is determined. Furthermore, this includes detecting manufacturing activity or manufacturing state D_M_A in response to the analysis of the captured image. This can be performed to adapt to at least one or all of the algorithms following steps A_I, D_A, and CL_A. Furthermore, step D_M_A can be further or alternatively performed to adjust the parameters of the image capture C_I step, such as camera resolution, frame rate, etc.

[0087] Finally, the anomaly type and location on the wind turbine blade section of the detected anomaly are transmitted in real time to the operator C_A_T_L in the manufacturing process, allowing the operator to initiate actions during manufacturing to remedy the detected anomaly. The method includes receiving feedback R_O_FB from the operator to the processor system in response to the transmitted anomaly, thereby allowing the neural network involved in at least detecting or classifying the anomaly to be trained in response to the feedback from the operator, in order to improve anomaly detection and classification.

[0088] In summary, the present invention provides a manufacturing system for manufacturing wind turbine blade portions (such as blade halves or segments of blade halves), for example, composite fiber wind turbine blades. The imaging system has a camera (CM) arranged in an elevated position to capture images of a manufacturing area (A_M). A processor system (PS) is arranged to process the captured images from the imaging system to analyze the captured images (IA) at least with respect to texture and / or contrast. Based on the image analysis, anomalies (AD) on a portion of the wind turbine blade are detected during the manufacturing process, classified into one of several predetermined types of anomalies (ACL) based on the analysis of texture, and further determined to be located (AL) of the classified anomaly on the wind turbine blade, such as both spanwise and chordwise positions. A user interface is connected to the processor system to transmit the anomaly type and its location on the wind turbine blade portion to the operator of the manufacturing process in real time.

[0089] Although the invention has been described in conjunction with specified embodiments, it should not be construed as being limited in any way to the presented examples. The scope of the invention will be interpreted in accordance with the appended set of claims. In the context of the claims, the terms "comprising" or "including" do not exclude other possible elements or steps. Furthermore, references such as "a" or "an" should not be construed as excluding multiples. The use of reference numerals in the claims with respect to elements indicated in the drawings should also not be construed as limiting the scope of the invention. Moreover, various features mentioned in different claims may be advantageously combined, and the mention of these features in different claims does not preclude the possibility and advantage of combining features.

Claims

1. A manufacturing system for manufacturing wind turbine blade portions such as composite fiber wind turbine blade sections, blade halves, or segments thereof, said system comprising: - An imaging system comprising at least one camera (CM) and arranged to capture images of the area (A_M) where the wind turbine blade section is manufactured, such as being arranged to capture images covering the entire area of ​​the wind turbine blade section. - Processor system (PS), which is arranged as follows: - Process the captured images from the imaging system to analyze the captured images (IA) at least with respect to texture and / or contrast. - Anomalies (ADs) on a portion of a wind turbine blade are detected during the manufacturing process based on texture analysis of captured images using an anomaly detection algorithm. - Based on the analysis of texture and / or contrast of the captured image using an anomaly classification algorithm, the anomalies are classified into one of several predefined anomaly types (ACL), and - At least with respect to the first axis (X), such as with respect to both axes (X, Y), determine the location (AL) of the classified anomalies on the wind turbine blades, and - A user interface operably connected to the processor system, wherein the user interface is configured to transmit the anomaly type of detected anomalies and / or their location on a wind turbine blade section to the operator of the manufacturing process in real time.

2. The manufacturing system according to claim 1, wherein, The processor system is configured to automatically detect manufacturing activities or manufacturing states in response to analysis of images captured from the imaging system, such as being configured to automatically detect the current manufacturing activity or manufacturing state based on selecting one of a plurality of predefined manufacturing activities or manufacturing states as the current one, such as selecting one of at least six predefined manufacturing activities or manufacturing states, such as selecting one of the following: 1) mold preparation, 2) gel coating application, 3) outer layer, 4) core board, 5) inner layer, 6) vacuum bagging, 7) resin infusion, 8) shell curing heating blanket, 9) bag removal, 10) inspection and bonding surface preparation, 11) adhesive curing, and 12) demolding.

3. The manufacturing system according to claim 1 or 2, wherein, The processor system is configured to select an anomaly detection algorithm and / or anomaly classification algorithm in response to detected manufacturing activities or manufacturing states.

4. The manufacturing system according to claim 2 or 3, wherein, The processor system is arranged such that the imaging system adjusts its image capture with respect to at least one parameter, such as image resolution and / or image capture frequency, in response to detected manufacturing activity or manufacturing state.

5. The manufacturing system according to any one of the preceding claims, wherein, The processor system is configured to classify anomalies into one of the following: 1) wrinkles in sheet or laminate layup, 2) ventilation or dry spots in post-injection laminates, 3) deviations from the centerline of the spar cap and adhesive, 4) wrinkles in vacuum bags and / or foils, 5) air bubbles in vacuum bags during injection, and 6) positioning deviations from the design specifications of components of the wind turbine blade section, such as web positioners, webs, lightning protection system wiring, and fixtures.

6. The manufacturing system according to any one of the preceding claims, wherein, The user interface is configured to receive feedback from the operator on detected anomalies, such as feedback indicating the relevance of the detected anomalies, and wherein the user interface is configured to transmit the feedback to the processor system.

7. The manufacturing system according to claim 6, wherein, The processor system is configured to receive the feedback in response to a detected anomaly, and to process the feedback according to a learning algorithm in response to the feedback, the learning algorithm being used to improve the accuracy of the anomaly detection algorithm and / or the anomaly classification algorithm, such as the learning algorithm involving neural networks, such as convolutional neural networks and / or recurrent neural networks.

8. The manufacturing system according to any one of the preceding claims, wherein, The processor system is arranged to determine the location of the classified anomalies on the wind turbine blade section with respect to the longitudinal axis (X), such as the spanwise direction, and with respect to the axis (Y), such as the chordwise direction, perpendicular to the longitudinal axis (X).

9. The manufacturing system according to any one of the preceding claims, wherein, The user interface includes a display, such as a display on a portable device, such as a tablet, smartphone, or smartwatch, and the anomaly type and location of the detected anomaly on a wind turbine blade section are conveyed by symbols and / or text overlays on a captured photograph of the wind turbine blade section.

10. The manufacturing system according to any one of the preceding claims, wherein, The user interface includes a projection system, such as a laser projector, or multiple projectors arranged along the spanwise direction of the wind turbine blade section, the projection system being configured to provide visual indications of detected anomalies directly at defined locations on the surface of the wind turbine blade section.

11. The manufacturing system according to any one of the preceding claims, wherein, The user interface includes multiple visual indicators arranged along the wind turbine blade section, such as controllable lines of multiple LED light indicators arranged in a line, to provide visual indication of the longitudinal position of detected anomalies.

12. The manufacturing system according to any one of the preceding claims, wherein, The manufacturing system is configured to perform automatic resin infusion tracking in response to captured images during the resin infusion process on the wind turbine blade section, such as the automatic resin infusion tracking involving the analysis of the infusion fill percentage over time during the resin infusion process.

13. The manufacturing system according to any one of claims 12, wherein, The processor system is operatively connected to control the resin infusion valve, such as controlling the resin infusion valve to close in response to automatic resin infusion tracking, such as controlling the resin infusion valve to close in response to an estimate that the housing infusion has reached 100% or nearly 100%.

14. The manufacturing system according to any one of the preceding claims, wherein, The imaging system includes at least one camera positioned and configured to provide an image of the entire width of the wind turbine blade section, such as the at least one camera being positioned at a fixed height above the surface of the wind turbine blade section, the fixed height being 0.8 to 2.0 times the width of the wind turbine blade section.

15. A method for manufacturing a wind turbine blade portion, such as a composite fiber wind turbine blade section, a segment such as a blade half or a blade half, the method comprising: - Capture an image (C_I) of the area where the wind turbine blade section is manufactured, such as capturing an image covering the entire area of ​​the wind turbine blade section. - At least regarding texture and / or contrast analysis of the captured image (A_I), such as analyzing texture and / or contrast in multiple segments of the captured image, - Anomalies (D_A) were detected on a portion of a wind turbine blade during the manufacturing process based on texture analysis of the captured images. - Based on the analysis of the texture and / or contrast of the captured image, the anomaly is classified into one of several predefined anomaly types (C_A), and - At least regarding the first axis (X), such as regarding both axes (X, Y), determine the location (D_L) of the classified anomaly on the wind turbine blade section, and - The detected anomalies and their location on the wind turbine blade section are transmitted in real time to the operators in the manufacturing process (C_A_T_L).

Images