State detection device and state detection system

Abstract

A state detection device (10) is provided with an analysis unit (12) that detects the state of an object to be detected. The analysis unit (12) acquires sound information corresponding to a sound generated by the detection target, detects the state of the detection target on the basis of a pattern included in an image indicating a change in the frequency component of the sound information over time, and outputs a detection result of the state of the detection target.

Description

[0001] Cross - reference to related applications

[0002] This application claims the priority of Japanese Patent Application No. 2022 - 185127 (filed on November 18, 2022), and the entire content disclosed in that application is incorporated herein by reference for reference purposes. Technical field

[0003] The present invention relates to a state detection device and a state detection system. Background art

[0004] Currently, the following method is known. That is, when determining an abnormality in a blade of a windmill, the presence of a Doppler shift component in which a frequency component with a higher sharpness moves over time is detected in the analysis result of sound information (breaking wind sound) emitted from the blade (for example, refer to Patent Document 1).

[0005] Patent Document 1: Japanese Unexamined Patent Application Publication No. 2010 - 281279 Summary of the invention

[0006] When diagnosing equipment such as a wind power generation device or its structural elements such as blades, it is required to detect various states such as abnormalities of the diagnosis object.

[0007] The present invention has been made in view of the above problems, and an object thereof is to provide a state detection device and a state detection system that can improve the detection accuracy of various states such as abnormalities of the diagnosis object when diagnosing equipment such as a wind power generation device or its structural elements such as blades.

[0008] The state detection device according to several embodiments has an analysis unit that detects the state (abnormality, etc.) of a detection object (such as a blade of a wind power generation device). The analysis unit acquires the measurement result of the sound generated by the detection object as sound information, and detects the state of the detection object based on a pattern included in an image representing the change over time of the frequency components of the sound information, and outputs a detection result of the state of the detection object. The state detection device detects the state of the detection object in the above - mentioned manner, so that the detection accuracy of various states such as abnormalities of the diagnosis object can be improved when diagnosing equipment such as a wind power generation device or its structural elements such as blades.

[0009] Based on the state detection device according to one embodiment, it can be configured such that the analysis unit analyzes the part of the detection object that generates the sound represented by the pattern. As a result, it is easy to perform inspections or repairs on the analyzed part. As a result, the safety of the detection object is improved.

[0010] Based on the state detection device according to an embodiment, it can be configured such that when the value representing the difference between the pattern and the normal pattern is greater than or equal to the difference threshold, the parsing unit outputs that the state of the part where the sound represented by the pattern is generated is abnormal. Thereby, the determination criterion for the state of the detection object becomes clear. As a result, the detection accuracy of the state is improved. In addition, this threshold can be set not only to a fixed value but also to a value using the operating state or the wind condition, etc. Thereby, the detection accuracy of the state is also improved.

[0011] Based on the state detection device according to an embodiment, it can be configured such that the parsing unit detects the state of the detection object based on a database that associates the pattern with the state of the detection object. The parsing unit can generate the normal pattern based on the sound information when the state of the detection object is normal and register it in the database. Thereby, the classification accuracy of the pattern is improved. As a result, the detection accuracy of the state of the detection object is improved.

[0012] The state detection system according to several embodiments may include: the above-mentioned state detection device; and a sound pickup device that measures the sound generated by the detection object and outputs the measurement result as sound information to the state detection device. The state detection system has the above-mentioned state detection device, so that the detection accuracy of various states such as abnormalities of the diagnostic object can be improved when diagnosing equipment such as a wind power generation device or its structural elements such as blades.

[0013] Based on the state detection system according to an embodiment, it can be configured such that the sound pickup device has a plurality of microphones. The sound pickup device can generate sound information representing the component of the sound arriving at the sound pickup device from a specified direction based on the sounds detected by the plurality of microphones respectively. Thereby, the influence of sounds generated other than the detection object, such as background noise, can be reduced. As a result, the detection accuracy of the state of the detection object is improved.

[0014] Based on the state detection system according to an embodiment, it can be configured such that the parsing unit of the state detection device obtains a plurality of pieces of sound information representing the components of the sounds arriving at the sound pickup device from a plurality of directions respectively. The parsing unit can detect the state of the detection object based on the plurality of pieces of sound information. By detecting the state of the detection object based on the plurality of pieces of sound information, the influence of sounds generated other than the detection object, such as background noise, can be reduced. As a result, the detection accuracy of the state of the detection object is improved.

[0015] Effects of the Invention

[0016] According to the present invention, there is provided a state detection device and a state detection system that can improve the detection accuracy of various states such as abnormalities of a diagnostic object when diagnosing devices such as wind power generation devices or their structural elements such as blades. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a diagram showing an analysis example of sound information related to a comparative example.

[0018] Figure 2 It is a schematic diagram showing a structural example of a state detection system according to an embodiment.

[0019] Figure 3 It is a block diagram showing a structural example of a state detection system according to an embodiment.

[0020] Figure 4 It is a diagram showing an example of the change over time of the frequency components of sound as an image in grayscale.

[0021] Figure 5 It is a diagram showing an example of the change over time of the frequency components of sound information corresponding to the components of sound coming from the first direction as an image in grayscale.

[0022] Figure 6 It is a diagram showing an example of the change over time of the frequency components of sound information corresponding to the components of sound coming from the second direction as an image in grayscale.

[0023] Figure 7 It is a diagram showing an example of a periodic pattern included in an image in grayscale.

[0024] Figure 8 It is a diagram showing an example of a granular pattern included in an image in grayscale.

[0025] Figure 9 It is a diagram showing an example of a linear pattern included in an image in grayscale.

[0026] Figure 10 It is a flowchart showing a flow example of a state detection method according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0027] (Comparative Example)

[0028] The system related to the comparative example uses a microphone provided at a fixed position to detect the cracking sound generated by the rotating blades of a windmill. As Figure 1 shown in the curve graph, the sound is represented as the change over time of the frequency components. Figure 1The horizontal axis of the curve graph represents time. The vertical axis represents frequency. The frequency components of the sound of the blades of the windmill are represented as region 90. In addition, in the case where there are characteristic parts such as scars on the blades, the sound generated at the characteristic parts is distinguished as region 91 that is also prominent in region 90.

[0029] Here, the blades of the rotating windmill have a speed relative to the fixed microphone. When the sound source moves together with the blades, the sound generated by the abnormal unevenness of the blades changing the air flow is detected by the microphone together with the sound of the blades. The system according to the comparative example determines the abnormality of the blades based on the presence of the component that the frequency component with higher sharpness in the sound signal shifts over time (Doppler shift). In addition, the system according to the comparative example determines that the blades are abnormal when the shift of the frequency component with higher sharpness has a specified slope.

[0030] However, the system according to the comparative example can only detect the state in which the frequency component with higher sharpness in the sound signal appears over time or as a slope.

[0031] Therefore, the present invention will describe a state detection system 1 (refer to Figure 2 and Figure 3 ), that is, even if it is impossible to detect the shift over time or the slope of the frequency component with higher sharpness in the sound signal of the sound generated by the detection object such as the blade 42 (refer to Figure 2 ) of the windmill 40, various states of the detection object can be detected.

[0032] (Embodiment of the present invention)

[0033] As Figure 2 and Figure 3 shown, the state detection system 1 according to one embodiment includes a state detection device 10, a sound pickup device 20, and a display device 30. The state detection system 1 detects the state of the detection object based on the sound generated by the detection object. In the present embodiment, the detection object is set as the blade 42 of the windmill 40 shown in Figure 2 . In addition to the blade 42, the windmill 40 also includes a fairing 44 and a support column 46. In the present embodiment, it is assumed that the blade 42 has an abnormality. The abnormality generated by the blade 42 can include various forms such as scars, cracks, fractures, protrusions or depressions on the surface of the blade 42, or cavities, cracks or fractures inside the blade 42. The state detection system 1 can set the blade 42 as the detection object and detect the abnormality generated by the blade 42. The part where the abnormality occurs in the blade 42 is also referred to as the abnormal part 48.

[0034] (Structural example of the state detection system 1)

[0035] Next, a structural example of the state detection system 1 will be described.

[0036] <State Detection Device 10>

[0037] The state detection device 10 detects the state of the blade 42 based on the acoustic information corresponding to the waveform of the sound generated by the blade 42 as the detection object. The detection object is not limited to the blade 42 and may also include other parts such as the fairing 44. The state detection device 10 may detect the state of the blade 42 without relying on the frequency components of the acoustic information or the change of the frequency components of the acoustic information over time. The state detection device 10 may obtain the acoustic information, the frequency components of the acoustic information, or the change of the frequency components of the acoustic information over time from the sound pickup device 20 described later. The state detection device 10 may obtain the frequency components of the acoustic information or the change of the frequency components of the acoustic information over time by obtaining the acoustic information from the sound pickup device 20 and performing frequency analysis of the acoustic information. The state detection device 10 includes an analysis unit 12, a storage unit 14, and an interface 16.

[0038] The analysis unit 12 controls each structural part of the state detection device 10. The analysis unit 12 may be configured to include a processor such as a CPU (Central Processing Unit). The analysis unit 12 may implement a specified function by causing the processor to execute a specified program.

[0039] The storage unit 14 may store various information for the operation of the analysis unit 12 or programs for implementing the functions of the analysis unit 12. The storage unit 14 may act as a working memory for the analysis unit 12. The storage unit 14 may be constituted by, for example, a semiconductor memory. The storage unit 14 may be configured to include a volatile memory or a non-volatile memory. The storage unit 14 may be configured as a non-temporary computer-readable storage medium. The storage unit 14 may be included in the analysis unit 12.

[0040] The interface 16 is configured to include a communication device that connects the state detection device 10 to the sound pickup device 20 or the display device 30 in a communicable manner. The communication device may be configured to communicate based on mobile communication standards such as 4G (4th Generation), LTE (Long Term Evolution), or 5G (5th Generation), for example. The communication device may be configured to communicate based on the communication standard of a LAN (Local Area Network), for example. The communication device may be configured to communicate in a wired or wireless manner.

[0041] The interface 16 can be configured to include a display device. The display device can include various displays such as a liquid crystal display. The interface 16 can be configured to include a voice output device such as a speaker. The interface 16 is not limited thereto and can also be configured to include various other output devices.

[0042] The interface 16 can be configured to include an input device for receiving input from a user. The input device can include, for example, a keyboard or physical keys, and can also include a touch panel or touch sensor or a pointing device such as a mouse. The input device is not limited to the above examples and can also be configured to include various other devices.

[0043] The state detection device 10 can be configured as a PC (Personal Computer) or can be configured as at least one server device. The state detection device 10 can be implemented by a cloud computing system.

[0044] The state detection device 10 can be configured, for example, to be able to input parameters for detecting the state of a detection object including the blade 42, etc.

[0045] <Pickup device 20>

[0046] The pickup device 20 has one or more microphones. The pickup device 20 uses one or more microphones to detect the sound generated from the detection object or its surrounding environment. That is, the pickup device 20 uses one or more microphones to detect the sound generated in the environment where the detection object exists. The pickup device 20 outputs the detection result of the sound detected by one or more microphones as sound information to the state detection device 10. The sound information can be represented as a waveform showing the change of sound pressure over time. The sound information can be represented by the amplitude and phase of the sound pressure.

[0047] When the sound pickup device 20 has a plurality of microphones, the plurality of microphones may be configured as an array microphone arranged in an array. The sound pickup device 20 may output the measurement result of the waveform of the sound detected by each microphone constituting the array microphone to the state detection device 10 as sound information. The state detection device 10 may generate sound information corresponding to the waveform of the component of the sound arriving from a specified direction with respect to the sound pickup device 20 by analyzing the amplitude and phase of the waveforms of the measurement results of the respective microphones obtained as the sound information. For example, the state detection device 10 may generate sound information corresponding to the waveform of the component of the sound arriving from a specified part of the detection target. The direction from the specified part of the detection target toward the sound pickup device 20 is also referred to as the first direction. The sound information representing the component of the sound arriving at the sound pickup device 20 from the first direction is also referred to as the first sound information. The state detection device 10 may generate sound information corresponding to the waveform of the component of the sound arriving from a specified part other than the detection target. The direction from the specified part other than the detection target toward the sound pickup device 20 is also referred to as the second direction. The sound information representing the component of the sound arriving at the sound pickup device 20 from the second direction is also referred to as the second sound information. The state detection device 10 may generate sound information corresponding to the waveforms of the components of the sounds arriving from the first direction and the second direction, respectively. The state detection device 10 may generate a plurality of pieces of sound information corresponding to the waveforms of the components of the sounds arriving from a plurality of directions, respectively.

[0048] The sound pickup device 20 may be arranged Figure 2 as an example, around the support column 46 of the windmill 40 serving as the detection target. The sound pickup device 20 may be arranged near the intersection of the extension line of the line (vertically downward line) from the rotation axis of the blade 42 to the tip of the blade 42 with the ground when the blade 42 rotates to the lowest position. It is not limited to Figure 2 the position shown as an example, and the sound pickup device 20 may also be arranged at various other positions such as the fairing 44. The sound pickup device 20 may be arranged near the intersection of the line from the tip of the blade 42 toward the vertical direction with the ground when the tip of the blade 42 rotates from the rotation axis to the horizontal position.

[0049] <Display device 30>

[0050] The display device 30 displays the detection result of the state of the detection target based on the state detection device 10. The display device 30 may display various data such as sound information, frequency components of the sound information, or changes in the frequency components of the sound information over time. The display device 30 may include various displays such as a liquid crystal display, for example. The state detection system 1 may have a speaker that outputs the sound itself measured by the sound pickup device 20 or a warning sound or the like generated according to the detection result of the state of the detection target as voice.

[0051] (Operation Example of State Detection System 1)

[0052] Regarding the state detection system 1, the state detection device 10 can detect the state of the detection object based on the sound information corresponding to the waveform of the sound generated in the environment where the detection object exists. Hereinafter, a structural example of the state detection system 1 will be described.

[0053] <Analysis of Sound Information>

[0054] Regarding the sound information, the analysis unit 12 of the state detection device 10 can, for example, use methods such as FFT (Fast Fourier Transform) or 1 / 1 or 1 / 3 octave analysis to numericalize the sound pressure of each frequency. The analysis unit 12 can numericalize the change amount of the sound pressure of each frequency. The analysis unit 12 can calculate the frequency at which the sound pressure is a maximum value with a specified constant time amplitude as the peak frequency. The analysis unit 12 can calculate the duration of the peak, the movement amount of the peak frequency, the magnitude of the sound pressure of the peak frequency, and the dispersion of the sound pressure in the frequency band including the peak frequency by tracking the peak frequency. The analysis unit 12 can store the analysis result of the sound information together with the sound information in the storage unit 14.

[0055] The analysis unit 12 can detect the state of the detection object based on the change of the frequency component of the sound information over time. The change of the frequency component of the sound information over time is as Figure 4 shown in the example, and can be represented as an image including pixels representing the magnitude of the component of a certain frequency of the sound information at a certain time in gray scale. The vertical direction in which the pixels are arranged in the image is also called a column. That is, the image includes multiple columns of pixels. In addition, the horizontal direction in which the pixels are arranged in the image is also called a row. That is, the image includes multiple rows of pixels. Figure 4 Each column of pixels included in the image shown in the example represents the magnitude of the component of each frequency of the sound information at the same time. Figure 4 Each row of pixels included in the image shown in the example represents the magnitude of the component of the same frequency of the sound information at each time. In Figure 4 the image, the closer the color of the point representing a certain frequency component at a certain time is to black, it means that the sound pressure of the frequency component at that time is greater. The closer the color of the point is to black can also be rephrased as the color of the point is darker. On the contrary, the closer the color of the point representing a certain frequency component at a certain time is to white, it means that the sound pressure of the frequency component at that time is smaller. The closer the color of the point is to white is also rephrased as the color of the point is lighter. That is, Figure 4 the image is substantially represented by a three-dimensional curve graph showing the relationship between time and the magnitude of the component of each frequency of the sound information at each time.

[0056] Figure 4The gray-scale image shown as an example represents the change over time of the frequency components of the sound information after the sounds arriving from various directions converge toward the sound pickup device 20 in an environment where the blade 42 is an object to be detected. In other words, Figure 4 The gray-scale image shown as an example represents the change over time of the frequency components of the sound information detected by the non-directional sound pickup device 20. The change over time of the frequency components of the sound information is not limited to a gray-scale image and can be represented by an RGB image or in various other ways.

[0057] The sound pickup device 20 can detect sound by using the microphone array in the above-described manner, thereby generating sound information corresponding to the waveform of the component of the sound arriving from a specific direction. That is, the sound pickup device 20 can detect sound in a directional manner to generate sound information.

[0058] Figure 5 The gray-scale image shown as an example represents the change over time of the frequency components of the sound information corresponding to the waveform of the component of the sound arriving at the sound pickup device 20 from the first direction. Figure 6 The gray-scale image shown as an example represents the change over time of the frequency components of the sound information corresponding to the waveform of the component of the sound arriving at the sound pickup device 20 from the second direction. In Figure 5 and Figure 6 In the gray-scale image shown as an example, each column of pixels, that is, the pixels arranged in the vertical direction, represents the magnitude of the components of the respective frequencies of the sound information at the same time. Figure 4 Each row of pixels, that is, the pixels arranged in the horizontal direction, included in the image shown as an example represents the magnitude of the components of the same frequency of the sound information at each time.

[0059] <Filtering>

[0060] The analysis unit 12 can perform filtering on the sound information. The analysis unit 12 can perform filtering at a higher frequency or a lower frequency, or a frequency obtained by combining a higher frequency and a lower frequency. The analysis unit 12 can perform filtering to remove the sudden increase in sound pressure that occurs within a short period of time. The sudden increase in sound pressure within a short period of time is also referred to as a pulse. The analysis unit 12 can perform digital filtering using, for example, a one-lag or moving average. The analysis unit 12 can perform background sound filtering using the sound pressure data of the recorded or defined background sound. The analysis unit 12 can perform filtering before the analysis of the above-described sound information.

[0061] <Detection of Feature Pattern>

[0062] An image representing the change in the frequency components of sound information over time contains characteristic patterns. The analysis unit 12 can identify the sound information synchronized with the passage of the blade 42 as the sound information of the blade 42. Here, in the image representing the change in the frequency components of sound information over time, the axis extending in the left - right direction is also called the time axis. Additionally, the axis extending in the up - down direction is also called the frequency axis. In the sound information of the blade 42, the frequency that is significantly larger compared to the surrounding part including the pixels adjacent to or near the pixel on the time axis or frequency axis of the sound pressure is also called the peak frequency. The analysis unit 12 can detect the pattern including the peak frequency as a characteristic pattern. The analysis unit 12 can detect, as a characteristic pattern, a range in the image of the sound information that includes the point reaching the peak frequency and points whose sound pressure is greater than or equal to the sound pressure obtained by multiplying the sound pressure of the point reaching the peak frequency by a prescribed ratio.

[0063] As Figure 7 For example, the analysis unit 12 can detect patterns 51, 52, and 53 as patterns corresponding to the whooshing sounds of the three blades 42 to be detected respectively. The analysis unit 12 can calculate the rotational speed of the blade 42 based on the time intervals between the detections of patterns 51, 52, and 53.

[0064] The analysis unit 12 can calculate the frequency ranges respectively showing patterns 51, 52, and 53. The analysis unit 12 can determine whether the state of each blade 42 changes based on the change over time of the frequency ranges showing each pattern.

[0065] The analysis unit 12 can detect the characteristic pattern 54 as a characteristic pattern that does not exist in patterns 51 and 53 in pattern 52. The analysis unit 12 can detect the possibility of an abnormal state of the blade 42 that detected the characteristic pattern 54 by detecting the characteristic pattern 54. The analysis unit 12 can analyze which part of the blade 42 is likely to be in an abnormal state based on the change amount of the frequency per unit time of the characteristic pattern 54. By analyzing the part that is likely to be in an abnormal state, it is easy to perform inspections, repairs, etc. on that part. As a result, the safety of the object to be detected is improved.

[0066] The characteristic sound varies according to the size of foreign matter, scratches, etc. at the part of the blade 42 that emits the characteristic sound. The analysis unit 12 can detect a change in the state of the part of the blade 42 that emits the characteristic sound by monitoring the expansion of a characteristic pattern including the change in the frequency of the characteristic sound, the magnitude of the sound pressure of the peak frequency, or the peak frequency. The expansion of the characteristic pattern including the frequency of the characteristic sound or the magnitude of the sound pressure of the peak frequency or the peak frequency varies according to the change in the relative speed between the blade 42 and the wind or the occurrence of turbulence. When the same characteristic sound is detected multiple times during one rotation cycle of the blade 42 by the sound pickup device 20, the analysis unit 12 can compare the sound information of the characteristic sound detected each time, and can reduce the influence caused by the change in the wind or the generation of turbulence according to the characteristic sound, and improve the detection accuracy of the change in the state of the part that emits the characteristic sound.

[0067] As another pattern, the analysis unit 12 can detect a pattern 55 existing in a region with a lower frequency. When the pattern 55 is detected, the analysis unit 12 can determine that the sound of electrical equipment existing around the detection target, the sound of waves on the coast or at sea, etc. are detected. As another pattern, the analysis unit 12 can detect a pattern 56 existing in a region with a higher frequency. When the pattern 56 is detected, the analysis unit 12 can determine that the sound of the wind blowing around the detection target, the noise of an aircraft, etc. are detected.

[0068] As Figure 8 shown by way of example, the analysis unit 12 can detect a granular characteristic pattern 57. When the granular characteristic pattern 57 is detected in the pattern corresponding to the blade 42, the analysis unit 12 can detect a slight vibration generated in the blade 42. The slight vibration generated in the blade 42 is also called flutter. As Figure 9 shown by way of example, the analysis unit 12 can detect a linear characteristic pattern 58. When the linear characteristic pattern 58 is detected in the pattern corresponding to the blade 42, the analysis unit 12 can detect the generation of unevenness on the surface of the blade 42.

[0069] The analysis unit 12 can not only detect shapes such as granular or linear shapes, but also detect the state of the blade 42 based on other matters. For example, the analysis unit 12 can detect the state of the blade 42 based on in which frequency band of the sound information a characteristic pattern appears for how long. The analysis unit 12 can calculate in which frequency band of the sound information a characteristic pattern appears for how long as the area of the characteristic pattern of the image of the sound information, and detect the state of the blade 42 based on the calculated area of the characteristic pattern. The analysis unit 12 can detect the state of the blade 42 based on the magnitude of the frequency components included in the characteristic pattern. The magnitude of the frequency components included in the characteristic pattern corresponds to the color density of the pixels when the sound pressure magnitude or the magnitude of the components is represented by gray scale in the image of the sound information.

[0070] The analysis unit 12 can identify the sound information that is not synchronized with the passage of the blade 42 as the sound generated in the surrounding environment other than the blade 42, or can also identify it as the sound generated by the collision of a bird or a flying object with at least a part of the blade 42 as the detection object.

[0071] <Classification of Characteristic Patterns>

[0072] The patterns included in the image representing the sound information are classified into a category representing the state of the detection object and a category not related to the state of the detection object. The category representing the state of the detection object is further classified into a category representing the normal state of the detection object and a category representing the abnormal state of the detection object. The category representing the abnormal state of the detection object is further classified into a category representing the manner of the abnormality generated in the detection object. The category not related to the state of the detection object is further classified into a category representing natural sounds such as wind or waves in the environment where the detection object exists and a category representing artificial sounds generated around the detection object. The category representing artificial sounds can be further classified into categories representing various artificial sounds such as the sound of a car running, the sound of a ship sailing, or the sound of an aircraft flying. The categories are not limited to the above examples.

[0073] The analysis unit 12 can preset categories and classify the detected patterns into the preset categories. When the category into which the pattern is classified represents the state of the detection object, the analysis unit 12 can detect the state represented by the category as the state of the detection object when the characteristic pattern is detected. For example, when the characteristic pattern is classified into the category representing the holes existing on the surface of the blade 42, the analysis unit 12 can detect whether there are holes on the surface of the blade 42 as the state of the detection object. For example, when the characteristic pattern is classified into a category not related to the state of the detection object or a category representing the normal state of the detection object, the analysis unit 12 can detect the state of the detection object as a non-abnormal state.

[0074] The analysis unit 12 can pre-register data associating the category of the characteristic pattern with the state of the detection object in the database. The analysis unit 12 can classify the characteristic pattern into the categories registered in the database. By classifying the characteristic pattern using the database, the classification accuracy of the characteristic pattern is improved. The analysis unit 12 can detect the state associated with the classified category as the state of the detection object. By detecting the state of the detection object on the basis of the improved classification accuracy of the characteristic pattern, the detection accuracy of the state of the detection object is improved. The database can be stored in the storage unit 14 or in an external storage device connected to the state detection device 10.

[0075] A database can be generated for each individual blade 42 of the windmill 40. Due to factors such as the repair history of each individual blade 42, errors in components during manufacturing, or errors during assembly, the patterns in the case of abnormal states of each blade may be different for each blade 42. The analysis unit 12 can register the information of the repair marks as normal patterns in the database for each blade 42 by measuring the sound immediately after repair for each blade 42 using the sound pickup device 20 and obtaining the waveform. In addition, the analysis unit 12 can register the information of the repair marks as normal patterns in the database for each blade 42 by extracting the characteristic pattern from the image of the sound information.

[0076] A database can be generated commonly for multiple blades 42. In the database generated commonly for multiple blades 42, the individual differences of the above-mentioned blades 42 are not considered. Under this assumption, even when the characteristic pattern included in the image representing the sound information of each blade 42 corresponds to the normal state of the blade 42, the analysis unit 12 can determine the state of the blade 42 as an abnormal state. In addition, under this assumption, conversely, even when the characteristic pattern corresponds to the abnormal state of the blade 42, the analysis unit 12 can determine the state of the blade 42 as a normal state. That is, when determining, the individual differences of the blades 42 can be not considered, and the ratio of accurately identifying the state corresponding to the characteristic pattern can be reduced. The ratio of accurately identifying the state corresponding to the characteristic pattern is also called the recognition rate. The analysis unit 12 regards the blades 42 corresponding to the normal state or the abnormal state with the same or similar patterns as a group when the windmill 40 starts a new operation or when the blades 42 are restarted after repair in such a way that the recognition rate can be maintained. The analysis unit 12 can generate a common database by setting the characteristic pattern common to the blades 42 included in this group as the standard pattern and determine the state of the blades 42 considering the individual differences of the blades 42.

[0077] The analysis unit 12 can generate a trained model for classifying the feature pattern by performing machine learning using teacher data indicating which class the feature pattern belongs to. The analysis unit 12 can obtain a trained model for classifying the feature pattern from an external device. The analysis unit 12 can classify the feature pattern using the trained model.

[0078] The normal state of the blade 42 to be detected includes various states. For example, the normal state may include the state when the blade 42 is repaired or replaced. The analysis unit 12 can obtain the sound information that measures the sound generated from the blade 42 when the blade 42 to be detected is in a normal state. The analysis unit 12 can register various patterns included in the sound information when the blade 42 is in a normal state in association with the normal state or the abnormal state in the database. The patterns included in the sound information when the blade 42 to be detected is in a normal state are also referred to as normal patterns. The analysis unit 12 can register the normal patterns in association with the normal state of the blade 42 in the database. The analysis unit 12 can update the database by directly retaining the normal patterns already registered in the database and adding new normal patterns. The analysis unit 12 can update the database by replacing the normal patterns already registered in the database with new normal patterns. Even when the state of the blade 42 changes, the analysis unit 12 can re-register the patterns included in the sound information when the blade 42 is known to be in a normal state as normal patterns in the database. As a result, for example, a pattern corresponding to a sound caused by a repair mark is not erroneously detected as abnormal. As a result, the detection accuracy of the state of the detection object is improved.

[0079] The analysis unit 12 can extract a pattern having a difference from the normal pattern from the patterns included in the sound information of the blade 42. When a pattern having a difference from the normal pattern can be extracted, the analysis unit 12 can detect that the state of the blade 42 is not a normal state. The state where the blade 42 is not in a normal state may include, for example, a scar on the surface of the blade 42.

[0080] The analysis unit 12 can calculate the difference between the extracted pattern and the normal pattern as a numerical value in order to determine whether the pattern extracted from the sound information of the blade 42 corresponds to a normal pattern. The analysis unit 12 can detect that the state of the part where the sound represented by the feature pattern is generated is normal when the numerical value representing the difference is less than the difference threshold. The analysis unit 12 can detect that the state of the part where the sound represented by the feature pattern is generated is abnormal when the numerical value representing the difference is greater than or equal to the difference threshold. As a result, the determination criterion for the state of the detection object becomes clear. As a result, the detection accuracy of the state is improved.

[0081] The analysis unit 12 may determine whether the state of the blade 42 is a normal state without calculating the difference between the pattern included in the sound information of the blade 42 and the normal pattern. For example, the analysis unit 12 may register, as an abnormal pattern, the pattern included in the sound information when the blade 42 to be detected is not in a normal state or the pattern included in the sound information when the state of the blade 42 is an abnormal state in a database. When the pattern included in the sound information of the blade 42 matches the abnormal pattern, the analysis unit 12 may determine that the state of the blade 42 is not a normal state or the state of the blade 42 is an abnormal state.

[0082] Artificial sounds are generated for various reasons. The analysis unit 12 may register, as a normal pattern, a characteristic pattern such as an operation sound that is known to be generated during normal times in a database. When a characteristic pattern that does not correspond to the normal pattern is detected, the analysis unit 12 may, for example, detect thunder near the object to be detected. When a characteristic pattern that does not correspond to the normal pattern is detected, the analysis unit 12 may determine that a bird, a flying object, etc. have collided with the blade 42. When a characteristic pattern that does not correspond to the normal pattern is detected, the analysis unit 12 may determine that an abnormal phenomenon such as the intrusion of a suspicious person has occurred near the windmill 40 having the blade 42 as the object to be detected.

[0083] A trained model can be generated in such a way that characteristic patterns corresponding to various states when the blade 42 to be detected is in a normal state are classified into a class indicating a normal state.

[0084] The analysis unit 12 may detect the state of the object to be detected based on first sound information indicating a component of the sound arriving at the sound pickup device 20 from a first direction and second sound information indicating a component of the sound arriving at the sound pickup device 20 from a second direction. For example, the analysis unit 12 may extract a noise component from the second sound information and detect the state of the object to be detected based on the information obtained by removing the extracted noise component from the first sound information. By detecting the state of the object to be detected based on the sound information from which the noise component has been removed, the detection accuracy of the state of the object to be detected is improved. The analysis unit 12 may detect a characteristic pattern based only on the first sound information indicating a component of the sound coming from the object to be detected. As a result, the influence of sounds generated other than the object to be detected such as background noise is reduced. As a result, the detection accuracy of the state of the object to be detected is improved.

[0085] The analysis unit 12 may remove or attenuate the sound information in a specified frequency band in the sound information using a band-pass filter. For example, the analysis unit 12 may remove or attenuate the sound information in a frequency band other than the frequency of the sound of the blade 42 using a band-pass filter. As a result, the detection accuracy of the state of the blade 42 based on the sound information of the blade 42 is improved.

[0086] The analysis unit 12 can perform filtering using a reference object on data regarding the change over time of the frequency components of the sound information. Specifically, the analysis unit 12 can calculate a difference spectrum obtained by subtracting the past spectrum at a specified time from the spectrum at a certain time. In the difference spectrum, the spectrum of the background sound is removed or reduced. As a result, the detection accuracy of sudden sounds included in the background sound is improved.

[0087] <Tracking>

[0088] The analysis unit 12 can monitor the peak frequency that changes over time and regard it as a series of phenomena. The analysis unit 12 can calculate, based on the series of phenomena, the threshold of the sound pressure detected as a characteristic pattern including the peak frequency, the peak frequency at the start or end of the series of phenomena, the duration of the series of phenomena, or the spread of the characteristic pattern. The spread means the width of the frequencies greater than or equal to the sound pressure threshold.

[0089] <Notification of State>

[0090] The analysis unit 12 can output, as the detection result of the state detection device 10, the state indicated by the category for classifying the characteristic pattern to the display device 30. The display device 30 can notify the main body of the windmill 40 that manages the blade 42 that is the detection object by displaying the detection result of the state detection device 10. Not limited to the display device 30, the analysis unit 12 can output the detection result of the state detection device 10 to various other devices such as a speaker or a lamp. The speaker can notify the management main body of the windmill 40 having the blade 42 of the state of the blade 42 that is the detection object by outputting voice. The lamp can notify the management main body of the windmill 40 having the blade 42 of the state of the blade 42 that is the detection object by emitting light or changing the light emission state.

[0091] <Example Flow of State Detection Method>

[0092] The analysis unit 12 of the state detection device 10 can execute a state detection method including the Figure 10 flow of the flowchart shown as an example. The state detection method can be implemented as a state detection program that causes the processor constituting the analysis unit 12 to execute. The state detection program can be stored in a non-temporary computer-readable medium.

[0093] The analysis unit 12 acquires the sound information of the blade 42 that is the detection object (step S1). The analysis unit 12 analyzes an image representing the frequency components of the sound information or the change over time of the frequency components (step S2). The analysis unit 12 determines whether a pattern is detected based on the image of the sound information (step S3). When no pattern is detected from the image of the sound information (step S3: NO), it endsFigure 10 Execution of the process of the flowchart

[0094] When the analysis unit 12 detects a pattern from the image of the sound information (step S3: YES), it determines whether the detected pattern corresponds to the pattern corresponding to the sound of the blade 42 (step S4). In other words, the analysis unit 12 determines whether to classify the detected pattern into the category representing the sound generated from the blade 42.

[0095] When the detected pattern corresponds to the pattern corresponding to the sound of the blade 42 (step S4: YES), the analysis unit 12 determines whether the detected pattern matches the normal pattern (step S5). In other words, the analysis unit 12 determines whether to classify the pattern into the category representing the normal state of the blade 42 as the detection object. When the detected pattern matches the normal pattern (step S5: YES), it is determined that the state of the blade 42 as the detection object is not abnormal and the process ends Figure 10 Execution of the process of the flowchart. When the detected pattern does not match the normal pattern (step S5: NO), the analysis unit 12 determines that the state of the blade 42 as the detection object is abnormal, and notifies the abnormal state of the blade 42 as the detection object using the display device 30 (step S6). After the execution of the process in step S6, the analysis unit 12 ends Figure 10 Execution of the process of the flowchart

[0096] When the detected pattern does not correspond to the pattern corresponding to the sound of the blade 42 (step S4: NO), the analysis unit 12 classifies the detected pattern into the category of ambient sound, and notifies the type of sound associated with the classified category using the display device 30 (step S7). After the execution of the process in step S7, the analysis unit 12 ends Figure 10 Execution of the process of the flowchart

[0097] <Subsection>

[0098] As described above, in the state detection system 1 according to the present embodiment, the state detection device 10 detects a characteristic pattern based on an image representing the change over time of the frequency components of the sound information generated based on the sound generated by the detection object. The state detection device 10 classifies the detected characteristic pattern into categories representing various states and detects the states represented by the categories. Thereby, different states are detected from each other. As a result, the detection accuracy of various states is improved.

[0099] Even if the main body (such as an operator, etc.) that manages the windmill 40 with the blade 42 to be detected is not at the site, the state of the detection object can be grasped by the state detection system 1. As a result, the frequency of the operator, etc. accessing the site of the detection object and the access cost are reduced. In addition, continuous monitoring is achieved. The safety of the detection object is improved through continuous monitoring.

[0100] The determination criteria for the state are made clear by the state detection system 1. In addition, the analysis result of the sound information becomes clear as a numerical value. As a result, the safety of the detection object is improved.

[0101] The state detection system 1 can store the sound information. By jointly confirming the sound information stored by the operator, etc. and the detection result of the state of the detection object based on the state detection system 1, the skills or knowledge of the operator, etc. are improved.

[0102] (Other embodiments)

[0103] The following describes other embodiments.

[0104] ><Detection of state based on instantaneous frequency components>

[0105] As described above, the analysis unit 12 can detect the state of the detection object based on an image representing the change of the frequency components of the sound information over time. The analysis unit 12 can detect the state of the detection object based on the spectrum of the instantaneous frequency components of the sound information at a certain time. For example, when the sound pressure of a specific frequency component in the spectrum exceeds a threshold, the analysis unit 12 can determine that the state of the detection object is a specific state.

[0106] ><Other examples of detection objects>

[0107] The detection object is not limited to the blade 42 of the above-mentioned windmill 40, and the state detection system 1 can also detect the states of various other devices or equipment. As the detection object, the state detection system 1 can, for example, detect the state of a device or equipment that generates sound. As the detection object, the state detection system 1 can, for example, detect the state of a fan or a blower that generates sound in a periodic pattern. As the detection object, the state detection system 1 can, for example, detect the state of a compressor that generates sound in a discrete or burst pattern.

[0108] The embodiments related to the present invention have been described based on the respective drawings and embodiments. It should be noted that those skilled in the art can make various deformations or changes based on the present invention. Therefore, it should be noted that the above-mentioned deformations or changes are included in the scope of the present invention. For example, the functions, etc. included in each structural part can be reconfigured in a logically non-contradictory manner, and multiple structural parts can be combined into one or divided.

[0109] Description of reference numerals

[0110] 1 Status detection system

[0111] 10 Abnormality detection device (12: Analysis unit, 14: Storage unit, 16: Interface)

[0112] 20 Sound pickup device

[0113] 30 Display device

[0114] 40 Object to be detected (42: Blade, 44: Fairing, 46: Strut, 48: Abnormal part)

[0115] 51 - 53, 55, 56 Patterns

[0116] 54, 57, 58 Characteristic patterns

Claims

1. A state detection device, wherein, the state detection device has an analysis unit for detecting the state of a detection object, the analysis unit obtains a measurement result of a sound generated by the detection object as sound information, detects the state of the detection object based on a pattern included in an image representing the change over time of the frequency components of the sound information, and outputs a detection result of the state of the detection object.

2. The state detection device according to claim 1, wherein, the analysis unit analyzes the part of the detection object that generates the sound represented by the pattern.

3. The state detection device according to claim 2, wherein, when a value representing the difference between the pattern and a normal pattern is greater than or equal to a difference threshold, the analysis unit outputs a case where the state of the part that generates the sound represented by the pattern is abnormal.

4. The state detection device according to claim 3, wherein, the analysis unit detects the state of the detection object based on a database that associates the pattern with the state of the detection object, generates the normal pattern based on the sound information when the state of the detection object is normal, and registers it in the database.

5. A state detection system, wherein, the state detection system includes: the state detection device according to any one of claims 1 to 4; and a sound pickup device that measures a sound generated by a detection object and outputs the measurement result as sound information to the state detection device.

6. The state detection system according to claim 5, wherein, the sound pickup device has a plurality of microphones, and the sound pickup device generates sound information representing a component of the sound arriving at the sound pickup device from a specified direction based on the sounds detected by the plurality of microphones respectively.

7. The state detection system according to claim 6, wherein, the analysis unit of the state detection device obtains a plurality of pieces of sound information representing components of the sounds arriving at the sound pickup device from a plurality of directions respectively, and detects the state of the detection object based on the plurality of pieces of sound information.

Images