SYSTEM AND METHOD FOR EXAMINING OBJECTS FOR ERRORS

Abstract

A system (1) for examining an object (2) for errors comprises a monitoring device (3), a processing module (5), a capturing device (4) and a recognition module (6). The monitoring device (3) is designed to monitor at least one parameter. A specified range of the parameter defines a context within which a result of a recognition of at least parts of the object (2) is expected. The processing module (5) is designed to prove whether the monitored parameter is within the specified range and in this case to trigger the capturing device (4) which is designed to capture input data associated with the object (2). The recognition module (6) is pre-trained for recognizing the object (2) and to perform the recognition based on the input data. The recognition module (6) is designed to detect an error if a result of the recognition is not corresponding to the expected result.

Claims

1. A system (1) for examining at least one object (2) for errors comprising: a monitoring device (3), a processing module (5), a capturing device (4) and a recognition module (6), wherein the monitoring device (3) is configured to monitor at least one parameter, wherein a specified range of the at least one parameter defines a context within which a result of a recognition of at least parts of the at least one object (2) is expected, wherein the processing module (5) is configured to prove whether the at least one parameter that has been monitored is within the specified range and in this case to trigger the capturing device (4), wherein the capturing device (4) is configured to capture input data associated with at least parts of the at least one object (2), wherein the recognition module (6) is pre-trained for recognizing at least parts of the at least one object (2) and to perform the recognition of at least parts of the at least one object (2) based on the input data, wherein the recognition module (6) is configured to detect an error if a result of the recognition is not corresponding to the result that is expected.

2. The system (1) according to claim 1, wherein the monitoring device (3) is configured to monitor a location of the system (1), wherein the recognition module (6) is configured to detect an error if the system (1) is within a specified region of the at least one object (2) and a result of the recognition is not corresponding to the result that is expected.

3. The system (1) according to claim 1, wherein the monitoring device (3) is configured to monitor an orientation of the capturing device (4), wherein the recognition module (6) is configured to detect an error if the capturing device (4) is oriented according to a specified orientation range with respect to the at least one object (2) and a result of the recognition is not corresponding to the result that is expected.

4. The system (1) according to claim 1, wherein the monitoring device (3) is configured to monitor a time, wherein the recognition module (6) is configured to detect an error if the recognition module (6) is performing the recognition within a specified time slot and a result of the recognition is not corresponding to the result that is expected.

5. The system (1) according to claim 1, further comprising a communication module (7) configured to transmit the input data to a controlling station (21) if an error is detected.

6. The system (1) according to claim 1, wherein the system (1) further comprises a plurality of error modules (8), wherein each error module (8) of said plurality of error modules (8) is pre-trained for a recognition of a specific error, wherein the system (1) is configured to perform an error recognition by triggering the plurality of error modules (8) sequentially.

7. The system (1) according to claim 1 wherein the system (1) is configured as a mobile apparatus.

8. The system (1) according to claim 7 wherein the mobile apparatus is a drone (10), wherein the capturing device (4) is a camera (18), wherein the recognition module (6) is a neuronal network pre-trained for recognizing at least one intact wind turbine (19) based on at least one image captured by the camera (18), wherein the monitoring device (3) is configured to monitor a location of the drone (10) and an orientation of the camera (18), wherein the recognition module (6) is configured to detect an error if the drone (10) is within a specified region of the at least one intact wind turbine (19), the camera (18) is oriented according to a specified orientation range with respect to the at least one intact wind turbine (19) and a result of the recognition is not corresponding to the result that is expected of the at least one intact wind turbine (19).

9. A method (9) for examining at least one object for errors comprising: monitoring at least one parameter, wherein a specified range of the at least one parameter defines a context within which a result of a recognition of at least parts of the at least one object (2) is expected, proving whether the at least one parameter that has been monitored is within the specified range, capturing input data associated with at least parts of the at least one object (2) if the at least one parameter is within the specified range, performing a recognition of at least parts of the at least one object (2) based on the input data, detecting an error if a result of the recognition is not corresponding to the result that is expected.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In the following, the invention is described in connection with schematic figures, wherein

[0022] FIG. 1 shows a system for examining at least one object for errors;

[0023] FIG. 2 shows a method for examining at least one object for errors; and

[0024] FIG. 3 shows an exemplary embodiment of the system.

DETAILED DESCRIPTION OF THE INVENTION

[0025] FIG. 1 schematically shows a system 1 for examining at least one object 2 for errors.

[0026] The object 2 may be any apparatus or any system, e.g. the object 2 can be a wind turbine. However, the object 2 can also e.g. be any industrial production facility and/or industrial product or e.g. a building. There may be different errors associated with the object 2, e.g. mechanical damages and/or other technical failures may occur. An error can also consist in another undesirable condition of the object 2, e.g. an error can consist in a delayed arrival of a product.

[0027] The system 1 can e.g. be part of a mobile apparatus. The system can e.g. be part of a vehicle or a drone which can be used for the purpose of the detection of errors associated with the object 2. However, the system 1 may be also part of an immobile apparatus, e.g. a stationary surveillance system.

[0028] The system 1 comprises a monitoring device 3. The monitoring device 3 is designed to monitor at least one parameter. A specified range of the parameter defines a context within which a result of a recognition of at least parts of the object 2 is expected. The context can be used for the recognition as there is an expected recognition result which is linked to the context given by the specified range of the parameter.

[0029] In one embodiment the monitoring device 3 is designed to monitor a location of the system 1. For this, the monitoring device 3 can be a positioning and navigation system, e.g. a GPS based system. In this case, the context is given by a specified region of the object 2, i.e. the object 2 is expected to be present within the specified region. However, depending on the object 2 to be examined the capturing device 4 may also be another sensor.

[0030] Alternatively or additionally, the monitoring device 3 is designed to monitor a time. In this case, the context is given by a specified time slot, e.g. within which a certain event is expected, e.g. the arrival of the object 2. If the object 2 is a wind turbine the context may be given by a time slot within which win conditions allow for proper and efficient operation.

[0031] Alternatively or additionally, the monitoring device 3 is designed to monitor an orientation of a capturing device 4 of the system 1. The capturing device 4 is designed to capture input data associated with at least parts of the object 2. The capturing device 4 can e.g. be a camera designed to provide images and videos of the object 2.

[0032] The system 1 comprises a processing module 5 designed to prove whether the monitored parameter is within the specified range. If this is the case, the processing module 5 is designed to trigger the capturing device 4 to capture the input data associated with the object 2. E.g. the capturing device 4 captures images and/or videos of the object 2.

[0033] For the examination of the object 2 for errors, the system 1 comprises a recognition module 6. The recognition module 6 comprises a machine learning based algorithm, e.g. a neuronal network. For the purpose of recognition based on images, e.g. deep convolutional neuronal networks may be used. The recognition module 6 is pre-trained for recognizing at least parts of the object 2 and to perform the recognition of at least parts of the object 2 based on the input data captured and provided by the capturing device. The input data may be adapted prior to the recognition process. E.g. an image may be adapted to a neuronal network by scaling the image to a size which is compatible with the neuronal network.

[0034] The recognition module 6 is designed to detect an error based on the input data if a result of the recognition is not corresponding to the expected result. E.g. the recognition module 6 is designed to detect an error if the system 1 is within a specified region of the object 2 and a result of the recognition is not corresponding to the expected result. Also, the recognition module 6 can be designed to detect an error if the capturing device 4 is oriented according to a specified orientation range with respect to the object 2 and a result of the recognition is not corresponding to the expected result. Furthermore, the recognition module 6 can be designed to detect an error if the recognition module 6 is performing the recognition within a specified time slot and a result of the recognition is not corresponding to the expected result.

[0035] Performing the recognition within the context enables an efficient detection of errors associated with the object 2. Also, there is no need to train the recognition module for all potential situations and errors which may occur, whereby a training time may be reduced. Furthermore, the necessity to provide training data for a plurality of errors is avoided. As the training of the recognition module 6 is kept simple and smaller than a complex training, the system 1 can run on edge devices or devices with limited computing capacity. Thus, the system 1 can be part of a mobile apparatus.

[0036] Optionally, the system 1 may also comprise a communication module 7. The communication module 1 can be designed to transmit information about detected errors and/or the captured input data to a controlling station for further examination.

[0037] Additionally, the system 1 can comprise a plurality of error modules 8. Each error module comprises a machine learning based algorithm, e.g. a neuronal network, which is pre-trained for a recognition of a specific error. The system 1 is designed to perform an error recognition by triggering the error modules 8 sequentially. The error modules 8 can be triggered by the processing module 5 after the recognition module 6 has detected an error. If a first error module detects its error type, information about the error can be transmitted to the controlling station, too. If the first error module fails to recognize the error it has been trained for, a second error module is triggered to perform its own error recognition. Thus, an efficient error recognition is enabled.

[0038] FIG. 2 schematically shows method steps 11, 12, 13, 14, 15, 16, 17 of a method 9 for examining at least one object 2 for errors.

[0039] In a first method step 11 of the method 9 the at least one parameter is monitored. The specified range of the parameter defines the context within which a result of a recognition of at least parts of the object is expected. In a second method step 12 it is proved whether the monitored parameter is within the specified range. In a third method step 13 the input data associated with at least parts of the object 2 is captured if the parameter is within the specified range. In a fourth method step 14 the recognition of at least parts of the object 2 is performed based on the input data. In a fifth method step 15 an error is detected if a result of the recognition is not corresponding to the expected result.

[0040] Optionally, the method may also comprise a sixth step 16 within which information about detected errors and/or the captured input data are transmitted to a controlling station for further examination. In another optional seventh method step 17 an error recognition is performed by triggering the error modules 8 sequentially.

[0041] FIG. 3 schematically shows an exemplary embodiment of the system 1 and an exemplary object 2 which shall be examined for errors.

[0042] The system 1 is part of a drone 10. The capturing device 4 is a camera 18 designed to capture images. The recognition module 6 which is not shown in FIG. 3 is a neuronal network pre-trained for recognizing at least one intact wind turbine 19 based on at least one image captured by the camera 18. The monitoring device 3 is designed to monitor a location of the drone 10 and an orientation of the camera 18. The recognition module 6 is designed to detect an error if the drone 10 is within a specified region of the wind turbine 19, the camera 18 is oriented according to a specified orientation range with respect to the wind turbine 19 and a result of the recognition is not corresponding to the expected result of at least one intact wind turbine 19.

[0043] E.g. the recognition module 6 can be pre-trained for detecting a wind turbine 19 only if no parts of the wind turbine 19 are missing. If e.g. a blade 20 of the wind turbine 19 is missing, the recognition module 6 is not able to detect the wind turbine 19 as such. In this case, an error can be detected if the recognition is performed within the context given by the specified region of the wind turbine 19 and the specified orientation range with respect to the wind turbine 19 and a result of the recognition is not corresponding to the expected result of at least one intact wind turbine 19. The recognition module 6 e.g. may also be pre-trained for recognizing a proper operation of the wind turbine 19 within a specified time slot, e.g. when wind conditions are expected to enable an efficient operation of the wind turbine.

[0044] The system 1 may also comprise error modules 8 not shown in FIG. 3 to perform an error recognition. E.g. one error module may be pre-trained for recognizing a wind turbine 19 with at least one missing blade 20. Another error module 8 may be pre-trained to differentiate wind turbines 19 which are in operation from those which are not, e.g. due to bad wind conditions.

[0045] Information about a detected error can be sent to the controlling station 21 for further examination of the error. Also, information about recognized types of errors may be sent to the controlling station 21 in order to initiate a maintenance.

LIST OF REFERENCE NUMERALS

[0046] 1 system for examining at least one object for errors [0047] 2 object [0048] 3 monitoring device [0049] 4 capturing device [0050] 5 processing module [0051] 6 recognition module [0052] 7 communication module [0053] 8 error module [0054] 9 method for examining at least one object for errors [0055] 10 drone [0056] 11 first method step [0057] 12 second method step [0058] 13 third method step [0059] 14 fourth method step [0060] 15 fifth method step [0061] 16 sixth method step [0062] 17 seventh method step [0063] 18 camera [0064] 19 wind turbine [0065] 20 blade of the wind turbine [0066] 21 controlling station