Monitoring device and monitoring method

Abstract

A monitoring device (80) is provided with: an acquisition unit (81) that acquires a measurement value of a sensor (70); a detection unit (82) that detects the friction noise of the bearing (51) on the basis of the measured value; and a diagnosis unit (83) that diagnoses the state of the bearing (51) on the basis of the measured value, the diagnosis unit (83) diagnosing the bearing (51) using a measured value that does not include the friction sound detected by the detection unit (82).

Description

Technical Field

[0001] The present disclosure relates to a monitoring device and a monitoring method for monitoring the state of equipment including a bearing. Background Art

[0002] Conventionally, there is known a monitoring device for monitoring the state of equipment including a bearing. Japanese Patent Gazette No. 5917956 (Patent Document 1) describes a monitoring device that uses statistical methods to calculate effective values, peak values, average values, peak factors, effective values ​​after envelope processing, and peak values ​​after envelope processing based on measurement data from sensors installed on various devices, and compares the calculated values ​​with corresponding thresholds to determine bearing damage.

[0003] Japanese Patent No. 5146008 (Patent Document 2) describes a monitoring device that compares a frequency spectrum obtained by envelope analysis and frequency analysis with a threshold value for each part of the bearing, thereby identifying an abnormal part of the bearing.

[0004] Japanese Patent Gazette No. 6791770 (Patent Document 3) discloses a method for reducing erroneous judgments by dividing the entire measurement data into multiple segments and averaging the machine learning diagnosis results calculated in each segment when processing measurement data of a rotating machine affected by the operating conditions and noise of the equipment. Prior art literature Patent Literature

[0005] Patent Document 1: Japanese Patent No. 5917956 Patent Document 2: Japanese Patent No. 5146008 Patent Document 3: Japanese Patent No. 6791770 Summary of the invention Technical problem to be solved by the invention

[0006] Bearings used in various devices may produce an abnormal sound called friction noise (squeaking noise).

[0007] Friction noise is mainly measured in the frequency band above 1kHz. Friction noise is a rather harsh sound to humans, but it is not an abnormal sound caused by bearing damage. However, even in the case of bearing damage, there are cases where the amplitude value of vibration increases mainly in the frequency band above 1kHz.

[0008] Therefore, as shown in Patent Document 1, in the method of diagnosing the bearing based on the effective value and peak value without considering the friction noise, it is impossible to distinguish between the amplitude value increased by the friction noise and the amplitude value increased by the bearing damage. As a result, the method described in Patent Document 1 may cause misdiagnosis.

[0009] Friction noise often occurs during the period when the rolling element passes relative to the outer ring of the bearing. The period of friction noise is consistent with the period of damage vibration when the outer ring of the bearing is damaged by indentation or peeling. Therefore, in the method of diagnosing bearings by envelope analysis such as Patent Document 2, it is impossible to distinguish between the spectrum peaks caused by friction noise and the spectrum peaks caused by damage to the outer ring of the bearing. As a result, in the method described in Patent Document 2, misdiagnosis may occur. Regarding the method described in Patent Document 3, misdiagnosis may also occur due to the mixing of friction noise in the measurement data.

[0010] The present invention has been made to solve the above-mentioned technical problems, and an object of the present invention is to prevent the bearing diagnosis accuracy from being reduced due to the influence of friction noise. Technical solutions to technical problems

[0011] A monitoring device involved in one aspect of the present disclosure is a monitoring device for monitoring equipment including bearings, and a sensor for measuring a physical quantity that changes with the vibration of the equipment is installed on the equipment. The monitoring device includes: an acquisition unit for acquiring a measurement value of the sensor, a detection unit for detecting a friction sound of the bearing based on the measurement value, and a diagnosis unit for diagnosing a state of the bearing based on the measurement value, and the diagnosis unit uses the measurement value that does not include the friction sound to diagnose the bearing.

[0012] The monitoring device involved in other aspects of the present disclosure is a monitoring device for monitoring equipment including bearings, and a sensor for measuring a physical quantity that changes with the vibration of the equipment is installed on the equipment. The monitoring device includes: an acquisition unit for acquiring a measurement value of the sensor, a detection unit for detecting the friction sound of the bearing based on the measurement value, and a diagnosis unit for diagnosing the state of the bearing based on the measurement value. The detection unit inputs data generated based on the measurement value into a first estimation model for detecting friction sound to detect the friction sound, and the diagnosis unit inputs the detection result of the detection unit and the data generated based on the measurement value into a second estimation model for diagnosing the bearing to diagnose the bearing.

[0013] A monitoring method involved in one aspect of the present disclosure is a monitoring method for monitoring a device including a bearing, wherein a sensor for measuring a physical quantity that changes with the vibration of the device is installed on the device, and the monitoring method includes: a step of obtaining a measurement value of the sensor, a step of detecting a friction sound of the bearing based on the measurement value, and a step of diagnosing a state of the bearing based on the measurement value, wherein the diagnosis step includes a step of diagnosing the bearing using a measurement value that does not include friction sound. Effects of the Invention

[0014] According to the present disclosure, it is possible to prevent the bearing diagnosis accuracy from being reduced due to the influence of friction noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a diagram schematically showing the structure of a wind turbine generator to which the monitoring device is applied. Figure 2 This is a diagram showing an example of the hardware configuration of the monitoring device. Figure 3 This is a block diagram showing the functional structure of the monitoring device. Figure 4 Detailed description of the process flow of the friction noise detection process. Figure 5 is a diagram showing an example of measurement data. Figure 6 This is a diagram showing an example of a frequency spectrum obtained by performing Fourier transform on the measurement data. Figure 7 This is a diagram showing an example of smoothed data obtained by performing smoothing processing on a frequency spectrum. Figure 8 This is a diagram showing an example of segmented data obtained by performing segmentation processing on smoothed data. Fig. 9 This is a diagram showing an example of normalized data obtained by performing normalization processing on segmented data. Fig.10 Detailed description of the invention is a flowchart showing the processing steps of the bearing diagnosis process. Fig.11 It is a diagram for explaining a method of generating a first estimation model and a second estimation model. Fig.12 This is a flowchart showing the processing procedure of the first estimation model generation process. Fig.13 This is a conceptual diagram for explaining the method of friction noise detection processing involved in Modification Example 1. Fig.14 This is a flowchart showing the processing procedure of the friction noise detection process involved in Modification Example 1. Fig.15 : is a flowchart showing the processing procedure of the bearing diagnosis process according to the second modification. Fig.16 : is a flowchart showing the processing procedure of the bearing diagnosis process according to the third modification. Fig.17 This is a diagram for explaining a method in which the learning device according to Modification 4 generates the second estimation model. Fig.18: is a flowchart showing the processing procedure of the bearing diagnosis process according to the fourth modification. Fig.19 It is a diagram for explaining a specific example of a change in the structure of the monitoring device. DETAILED DESCRIPTION

[0016] Hereinafter, the embodiments of the present disclosure will be described with reference to the accompanying drawings. In the following drawings, the same reference numerals are used for the same or corresponding parts, and their description will not be repeated. In addition, the modified examples described below may also be appropriately and selectively combined.

[0017] <Structure of wind power generation equipment> Figure 1 1 is a diagram schematically showing the structure of a wind turbine generator 10 to which a monitoring device 80 according to the present embodiment is applied. Figure 1 The wind power generator 10 includes a main shaft 20, a hub 25, blades 30, a speed increaser 40, a generator 50, a main shaft bearing 60, a sensor 70, and a monitoring device 80. The speed increaser 40, the generator 50, the main shaft bearing 60, the sensor 70, and the monitoring device 80 are stored in a nacelle 90. The nacelle 90 is supported by a tower 100.

[0018] The main shaft 20 is connected to the input shaft of the speed increaser 40 and is rotatably supported by the main shaft bearing 60. The main shaft 20 transmits the rotation torque generated by the blades 30 subjected to the wind force to the input shaft of the speed increaser 40. The blades 30 are provided on the hub 25, convert the wind force into the rotation torque and transmit it to the main shaft 20. The main shaft bearing 60 is provided in the nacelle 90 and rotatably supports the main shaft 20.

[0019] The speed increaser 40 is provided between the main shaft 20 and the generator 50, and increases the rotation speed of the main shaft 20 and outputs it to the generator 50. As an example, the speed increaser 40 is composed of a gear speed increase mechanism including a planetary gear, an intermediate shaft, a high-speed shaft, and the like.

[0020] The generator 50 is connected to the output shaft of the speed-increasing gearbox 40 and generates electricity by the rotation torque received from the speed-increasing gearbox 40. The generator 50 is constituted by, for example, an induction generator.

[0021] The generator 50 is provided with a plurality of bearings 51 (see Figure 3 ). The generator 50 is an example of a device including bearings. The plurality of bearings 51 are composed of, for example, rolling bearings, and have an outer ring (fixed wheel), rolling elements, and an inner ring (rotating wheel). The rolling bearings are composed of, for example, self-aligning roller bearings, tapered roller bearings, cylindrical roller bearings, and ball bearings. The rolling bearings may be composed of a single row or a plurality of rows.

[0022] A sensor 70 is mounted on the generator 50. The sensor 70 measures the vibration of the generator 50 and outputs the measured value to the monitoring device 80. Since the bearing 51 is provided in the generator 50, the measured value of the sensor 70 includes the vibration element of the bearing 51. As physical quantities that change with vibration, acceleration, velocity, displacement, sound, AE (Acoustic Emission) and power can be cited. In the present embodiment, the sensor 70 is, for example, a vibration sensor (acceleration pickup) using a piezoelectric element. As the sensor 70, an acoustic sensor, an AE sensor, etc. can also be used.

[0023] The monitoring device 80 acquires the measured value from the sensor 70. The monitoring device 80 monitors the state of the generator 50 based on the acquired measured value. In particular, the monitoring device 80 has a function of diagnosing whether the bearing 51 has an abnormality based on the acquired measured value. Generally, when the bearing is damaged, the amplitude value of the vibration mainly increases in the frequency band higher than 1 kHz depending on the type and size of the bearing. The monitoring device 80 can diagnose whether the bearing 51 has an abnormality using the measured value of the sensor 70.

[0024] <Relationship between bearing damage and friction noise> Here, the relationship between bearing damage and friction noise is explained. When the bearing has multiple conditions such as clearance, sliding, vibration, oil film change, etc., an abnormal sound called friction noise (squeaking) may be generated. The friction noise is believed to be caused by the friction or collision between the inner and outer rings and the rolling elements.

[0025] In particular, in generators and motors that use bearings, friction noise is clearly recognized. Friction noise is mainly measured in the frequency band above 1kHz, and is accompanied by strong vibration (large, sharp sound). Friction noise is quite harsh to people. However, friction noise is not an abnormal sound caused by damage to the bearing (such as wear, peeling, cracks, notches, etc.). In fact, friction noise can sometimes be alleviated by performing maintenance such as greasing the bearing.

[0026] However, even in the case of bearing damage, as described above, there is a case where the amplitude value of vibration increases in the frequency band higher than 1 kHz. Therefore, as described in Japanese Patent No. 5917956, etc., in the method of diagnosing bearings based on effective values ​​and peak values ​​without considering friction noise, it is impossible to distinguish between the amplitude value increased due to friction noise and the amplitude value increased due to bearing damage.

[0027] Friction noise is often generated during the period of the rolling element passing relative to the outer ring of the bearing, and its generation period is consistent with the generation period of damage vibration when the outer ring of the bearing is damaged by indentation or peeling. Therefore, in the method of diagnosing bearings by envelope analysis described in Japanese Patent Gazette No. 5146008, it is impossible to distinguish between the spectrum peaks generated by friction noise and the spectrum peaks generated by damage to the outer ring of the bearing. In addition, if a small amount of friction noise is mixed in the measurement data, there is also a problem that the data cannot be used for bearing diagnosis.

[0028] Therefore, in the monitoring device 80 according to the present embodiment, as described below, bearing diagnosis taking friction noise into consideration is realized, thereby preventing the bearing diagnosis accuracy from being reduced due to the influence of friction noise.

[0029] Here, as the bearing to be monitored by the monitoring device 80, the bearing 51 in the generator 50 is exemplified. The bearing in the speed increaser 40 and the main shaft bearing 60 may be added to the monitoring object of the monitoring device 80. In addition, the monitoring object of the monitoring device 80 is not limited to the bearing in the wind turbine generator 10. For example, as the monitoring object of the monitoring device 80, bearings included in various equipment installed in factories and power plants, and bearings included in railway vehicles, etc. may also be used. In short, the monitoring device 80 can be applied to any equipment as long as it includes a bearing that supports a rotating shaft.

[0030] <Hardware structure of monitoring device> Figure 2 1 is a diagram showing an example of the hardware configuration of the monitoring device 80. The monitoring device 80 is constituted by, for example, a general-purpose computer (processing device) that acquires the measurement value of the sensor 70 and performs calculation processing.

[0031] like Figure 2 As shown, the monitoring device 80 includes a CPU (Central Processing Unit) 801 , a RAM (Random Access Memory) 802 , a storage 803 , and a communication interface 804 . The CPU 801 , the RAM 802 , the storage 803 , and the communication interface 804 are connected via a bus 805 .

[0032] The CPU 801 executes the monitoring program 806 stored in the storage 803. The RAM 802 provides a work area for storing data required to execute the monitoring program 806. The storage 803 is composed of, for example, a HDD (Hard Disk Drive) or an SSD (Flash Solid State Drive). The monitoring program 806 includes programs required for the CPU 801 to execute various flowcharts described below.

[0033] The communication interface 804 has an input / output port for inputting and outputting various signals. For example, the communication interface 804 receives a measurement value from the sensor 70. The communication interface 804 can output various signals generated by executing the monitoring program 806 to an external device. The external device is provided outside the wind turbine generator 10. The communication interface 804 communicates with the external device via a wireless or wired line.

[0034] <Functional structure of monitoring device> Figure 3 80 is a block diagram showing the functional structure of the monitoring device 80. Figure 3 As shown, the monitoring device 80 includes an acquisition unit 81, a friction noise detection unit 82, a bearing diagnosis unit 83, an alarm output unit 84, and a storage unit 85. The storage unit 85 is composed of, for example, Figure 2 RAM 802 and storage 803 implementations are shown.

[0035] The acquisition unit 81, the friction noise detection unit 82, the bearing diagnosis unit 83, and the alarm output unit 84 are configured to Figure 2 This is achieved by the CPU 801 shown executing the monitoring program 806 stored in the storage unit 85 .

[0036] Each structural element such as the acquisition unit 81, the friction sound detection unit 82, the bearing diagnosis unit 83, and the alarm output unit 84 can also be implemented by dedicated hardware such as a processing circuit (Processing Circuitry). The above-mentioned structural elements can also be configured as multiple processors and multiple memories that function in collaboration. Each structural element such as the acquisition unit 81, the friction sound detection unit 82, the bearing diagnosis unit 83, and the alarm output unit 84 can also be implemented by multiple independent processing devices. That is, the monitoring device 80 can also be composed of multiple processing devices that are communicably connected. In this case, the "monitoring device" should be understood as a concept that includes a monitoring system composed of multiple processing devices that are communicably connected.

[0037] The friction noise detection unit 82 detects the friction noise generated in the bearing 51 using the measurement value obtained from the sensor 70. In order to detect the friction noise, the friction noise detection unit 82 reads the learned first estimation model 821 from the storage unit 85. The first estimation model 821 generates data including the measurement data of the friction noise as a feature amount through machine learning.

[0038] The bearing diagnosis unit 83 uses the measurement value from the sensor 70 to diagnose abnormalities including damage to the bearing 51. For diagnosis, the bearing diagnosis unit 83 reads the learned second estimation model 831 from the storage unit 85. The second estimation model 831 generates data including the measurement data including the vibration when the bearing abnormality occurs as feature quantity through machine learning.

[0039] The alarm output unit 84 outputs an alarm related to the occurrence of friction noise and abnormality of the bearing 51. The alarm output unit 84 is connected to an external alarm device 15 via a network 5 such as the Internet. The alarm device 15 is, for example, a personal computer, a smart phone, or a tablet computer. The alarm device 15 can be a display that displays an alarm or a speaker system that outputs an alarm sound. The alarm output unit 84 itself can also have the function of displaying an alarm or outputting an alarm sound. The alarm output unit 84 can also output an alarm to an external system such as a cloud.

[0040] The acquisition unit 81 acquires the measured value from the sensor 70. Each time the measured value is acquired from the sensor 70, the acquisition unit 81 writes data corresponding to the measured value and the time when the measured value was acquired into the storage unit 85. Thus, the storage unit 85 generates and stores measurement data indicating the temporal change of the measured value of the sensor 70.

[0041] The friction noise detection unit 82 reads the measurement data from the storage unit 85. After the friction noise detection unit 82 performs various processes including Fourier transform processing on the measurement data read, the obtained data is input to the first estimation model 821. The first estimation model 821 outputs an estimation result on whether there is friction noise. The friction noise detection unit 82 outputs the output of the first estimation model 821 as a detection result to the bearing diagnosis unit 83.

[0042] The bearing diagnosis unit 83 obtains the detection result from the friction noise detection unit 82. When the detection result obtained from the friction noise detection unit 82 indicates that there is friction noise, the bearing diagnosis unit 83 does not perform the bearing diagnosis using the measurement data. As a result, by including the measurement data of friction noise, it is possible to prevent the output of an erroneous bearing diagnosis result from the bearing diagnosis unit 83.

[0043] The bearing diagnosis unit 83 sets an alarm for the friction noise (first alarm) when the detection result obtained from the friction noise detection unit 82 indicates that there is friction noise. The alarm output unit 84 outputs the set alarm for the friction noise to the alarm device 15. The operator who maintains the wind turbine generator 10 can determine the occurrence of the friction noise based on the alarm issued by the alarm device 15.

[0044] The bearing diagnosis unit 83 does not perform bearing diagnosis using the measurement data when the detection result obtained from the friction noise detection unit 82 is that there is no friction noise. In this case, the bearing diagnosis unit 83 reads the measurement data from the storage unit 85. The bearing diagnosis unit 83 diagnoses the state of the bearing using the read measurement data and the second estimation model 831. The bearing diagnosis unit 83 may also input the obtained data into the second estimation model 831 after performing various processing including Fourier transform processing on the measurement data.

[0045] The bearing diagnosis unit 83 sets an alarm (second alarm) for bearing abnormality when the bearing is abnormal. The alarm output unit 84 outputs the set alarm for bearing abnormality to the alarm device 15. The operator who maintains the wind turbine generator 10 can determine that the bearing 51 is abnormal based on the alarm issued by the alarm device 15.

[0046] When the alarm device 15 is constituted by a personal computer, the alarm output unit 84 sends a message (email) or a report notifying the alarm to the alarm device 15. The alarm device 15 displays the message or report on the monitor. The alarm device 15 can generate an alarm sound corresponding to the friction sound and an alarm sound corresponding to the bearing abnormality from the speaker system, and can also light up a lamp corresponding to the friction sound and a lamp corresponding to the bearing abnormality.

[0047] In this way, the monitoring device 80 detects the friction sound and bearing abnormality (bearing damage) with high accuracy and outputs an alarm. As a result, the operator can effectively maintain the equipment (wind turbine 10) including the bearing 51. As a result, the operating rate of the equipment can be improved. In particular, when the friction sound is output as an alarm, the operator can also perform additional maintenance for reducing the friction sound. By performing detailed maintenance by the operator, the frequency of collecting data that can be used for bearing diagnosis can also be increased.

[0048] <Processing steps of friction noise detection processing> Figure 4 1 is a flowchart showing the processing steps of the friction noise detection process. The friction noise detection process is performed by the monitoring device 80 (mainly the friction noise detection unit 82). After the monitoring device 80 reads the measurement data from the storage unit 85 (step S101), it sequentially performs Fourier transform processing (step S102), smoothing processing (step S103), segmentation processing (step S104) and normalization processing (step S105).

[0049] Figures 5 to 9 An example of waveforms or data associated with each process (step S101 to step S105 ) is shown.

[0050] Figure 5 2 is a diagram showing an example of measurement data. The measurement data is waveform data showing a change in a measurement value of the sensor 70 over time. Figure 6 This is a diagram showing an example of a frequency spectrum obtained by performing Fourier transform on the measurement data. Figure 7 This is a diagram showing an example of smoothed data obtained by performing smoothing processing on a frequency spectrum. Figure 8 This is a diagram showing an example of segmented data obtained by performing segmentation processing on smoothed data. Fig. 9This is a diagram showing an example of normalized data obtained by performing normalization processing on segmented data.

[0051] When friction noise occurs, the amplitude values ​​of multiple frequency bands of the spectrum obtained by Fourier transforming the measured data show characteristic changes. This change has nothing to do with whether the bearing is damaged, and is a change that is unique to friction noise. Such a change that is unique to friction noise is observed not only in the amplitude value of the spectrum, but also when observing the power spectrum density (PSD). In the friction noise detection process, the monitoring device 80 performs various processes including Fourier transform processing on the measured data, appropriately captures the changes that are unique to friction noise, and detects the presence or absence of friction noise.

[0052] Then, as needed, refer to Figures 5 to 9 Detailed description Figure 4 The friction sound detection process is shown.

[0053] <Fourier transform processing> like Figure 5 As shown in FIG. 1 , the measurement data is time series data indicating the change of the measurement value of the sensor 70 over time. The monitoring device 80 performs Fourier transform processing on the measurement data (step S102). In the Fourier transform processing, the monitoring device 80 generates a spectrum by performing frequency analysis on the measurement data. Thus, Figure 6 The spectrum (Fourier waveform) shown. Figure 6 In the graph of FIG. 1 , the horizontal axis represents frequency (Hz) and the vertical axis represents amplitude value. As the vertical axis, power spectrum density (PSD) may also be used. By performing Fourier transform processing, the waveform data of the sensor 70 is converted from time domain data to frequency domain data.

[0054] <Smoothing> Next, the monitoring device 80 performs a smoothing process on the spectrum (step S103). The monitoring device 80 smoothes the spectrum during the smoothing process. Here, it is important to reduce the sensitivity to the operating conditions of the equipment, the bearing model, and the differences in equipment without eliminating the changes in the spectrum that are unique to the friction sound. Thereby, it is possible to improve the robustness to differences in external environments such as the operating conditions of the equipment during data collection (rotation speed and load changes, etc.), the bearing model and the differences in equipment. For example, when the shaft speed of the equipment changes, a phenomenon in which part of the frequency band of the spectrum shifts can be seen. In order to blunt the detection sensitivity to such subtle changes, smoothing is an effective method.

[0055] Figure 7 It shows that through Figure 6 An example of smoothed data obtained by smoothing the spectrum shown. In particular, Figure 7The smoothed data is obtained by performing moving median processing on the amplitude value of the spectrum with a certain frequency width. The smoothing method is not limited to the moving median processing. For example, any smoothing method can be used as long as it can smooth the original shape of the data to a degree that is not seriously damaged, such as moving average processing and smoothing processing based on local regression.

[0056] <Segment processing> Next, the monitoring device 80 performs segmentation processing on the waveform obtained by the smoothing processing (step S104). In the segmentation processing, the monitoring device 80 divides the frequency band of the spectrum into a plurality of intervals, and then adjusts the amplitude value in each interval to a single representative value.

[0057] Figure 8 It shows that through Figure 7 The smoothed data shown in FIG. 1 is segmented into segmented data. Figure 7 The smoothed data shown is divided into 50 intervals, and the average value of the amplitude value in each interval is obtained as the segmented amplitude value of each interval. As the segmented amplitude value, any value may be used as long as it is not the average value of the amplitude value in the interval of the segment, but a value that gives a row characteristic to the amplitude value in the interval, such as the sum of the amplitude values ​​in the interval of the segment or the median value.

[0058] The purpose of segmentation is to reduce the resolution of the spectrum axis. By performing segmentation, similar to smoothing, the sensitivity to the operating conditions of the equipment, bearing model, and equipment differences can be reduced without canceling the changes in the spectrum unique to the friction noise.

[0059] Therefore, by performing segmentation processing, it is possible to improve robustness against differences in the external environment. In particular, by performing smoothing processing and segmentation processing, the amount of data can be reduced. In addition, when the measurement data is used for machine learning, since the number of input variables to the machine learning processing is reduced, it also has the effect of improving the processing time.

[0060] <Normalization Process> Next, the monitoring device 80 performs normalization processing on the waveform obtained by the division processing (step S105 ), thereby obtaining normalized data.

[0061] The purpose of normalization is to eliminate the influence of the magnitude (absolute value) of the amplitude value of the segmented data. For example, in the generator 50, as the load on the bearing 51 increases, the amplitude value of the measured data increases. At this time, since the amplitude value of the spectrum obtained from the measured data also increases, the feature quantity extracted from the segmented data also increases.

[0062] When the spectrum obtained when the load of the bearing 51 is small is compared with the spectrum obtained when the load of the bearing 51 is large, although the overall shape characteristics of the two are the same, the scale of the amplitude value changes. In this case, the feature value may not be properly evaluated. This has an adverse effect on the detection accuracy of the friction sound. By performing normalization processing, the difference in the scale of the amplitude value can be eliminated while retaining the shape characteristics of the spectrum. Therefore, the monitoring device 80 performs normalization processing to make the amplitude value of each segment dimensionless.

[0063] Fig. 9 It shows that through Figure 8 An example of normalized data obtained by normalizing the segmented data shown in FIG. Figure 8 The segmented data shown is normalized (scaled) to obtain Fig. 9 The normalized data shown is such that the amplitude value is within the range of a minimum value of 0 to a maximum value of 1. As the normalization processing, a processing in which the average value or the median value of the amplitude value is 1 may be adopted.

[0064] As described above, the monitoring device 80 performs various processes such as smoothing and segmentation on the frequency spectrum. If the unique changes of the friction sound are evaluated only by the increase or decrease of a single frequency band, it may be difficult to accurately distinguish the increase or decrease of the frequency band and the friction sound caused by the operating conditions of the equipment, the difference in bearing models, interference noise, and the occurrence of bearing damage.

[0065] Therefore, the monitoring device 80 removes noise by smoothing and segmenting the frequency spectrum, thereby generating data for detecting the friction noise with higher accuracy.

[0066] Next, the monitoring device 80 reads the first estimation model 821 from the storage unit 85 (step S106 ).

[0067] Next, the monitoring device 80 inputs the normalized data to the first estimation model 821 (step S107). The first estimation model 821 detects the friction noise from the normalized data. The first estimation model 821 outputs a result indicating the presence or absence of the friction noise. The monitoring device 80 saves the output of the first estimation model 821 as the detection result (step S108). The saved detection result is used in the bearing diagnosis process described below.

[0068] <Processing steps of friction noise detection processing> Fig.10 2 is a flowchart showing the processing procedure of the bearing diagnosis process. First, the monitoring device 80 refers to the detection result of the first estimation model 821 (step S201). Next, the monitoring device 80 determines whether friction noise is detected (step S202).

[0069] When the monitoring device 80 detects the friction noise, it sets the friction noise alarm (first alarm) (step S208). The alarm output unit 84 outputs the set friction noise alarm to the alarm device 15. When the monitoring device 80 detects the friction noise, it does not perform the bearing abnormality diagnosis process, but ends the shaft diagnosis process.

[0070] When the friction sound is not detected, the monitoring device 80 reads the second estimation model 831 from the storage unit 85 (step S203 ).

[0071] Next, the monitoring device 80 inputs the measurement data to the second estimation model 831 (step S204). The second estimation model 831 diagnoses the state of the bearing based on the measurement data. The diagnosis result of the bearing is output from the second estimation model 831. The monitoring device 80 refers to the output (bearing diagnosis result) of the second estimation model 831 (step S205) and determines whether the bearing has an abnormality (step S206).

[0072] If the bearing is abnormal, the monitoring device 80 sets an alarm (second alarm) for bearing abnormality (step S207). The alarm output unit 84 outputs the set friction noise alarm to the alarm device 15. If the bearing is not abnormal, the monitoring device 80 ends the bearing diagnosis process.

[0073] As described above, in the bearing diagnosis process, the monitoring device 80 does not perform bearing diagnosis when friction noise is detected, and performs bearing diagnosis only when friction noise is not detected. In this way, by excluding data containing friction noise from the measurement data for bearing diagnosis, misdiagnosis due to friction noise can be prevented. As a result, the diagnostic accuracy of the bearing can be improved.

[0074] <Generation of Estimation Model> Fig.11 is a diagram for explaining a method of generating the first estimation model 821 and the second estimation model 831. The learning device 8 is provided with Figure 2 The monitoring device 80 may be a processing device having the same hardware structure as the monitoring device 80 shown in FIG.

[0075] The learning device 8 generates a first estimation model 821 and a second estimation model 831, respectively. The first estimation model 821 and the second estimation model 831 are generated by, for example, trained machine learning based on any algorithm such as a decision tree, a random forest, a support vector machine, and a neural network. The learning device that generates the first estimation model 821 and the learning device that generates the second estimation model 831 may be different.

[0076] The first estimation model 821 is generated for the purpose of estimating whether there is friction sound in the measurement data. The learning data of the first estimation model 821 is composed of measurement data and answer data. The learning device 8 generates the first estimation model 821 after learning by repeating machine learning using a plurality of learning data. The learning data preferably includes a plurality of measurement data in which the characteristic features of the friction sound often appear. The learning device 8 outputs the generated first estimation model 821 to the monitoring device 80. The monitoring device 80 stores the first estimation model 821 in the storage unit 85.

[0077] The second estimation model 831 is generated for the purpose of estimating whether the bearing has abnormalities based on the measurement data. The learning data of the second estimation model 831 is composed of measurement data and answer data. The learning device 8 generates the second estimation model 831 after learning by repeating machine learning using a plurality of learning data. In the learning data, it is preferred to include a plurality of measurement data in which characteristics specific to bearing abnormalities frequently appear. The learning device 8 outputs the generated second estimation model 831 to the monitoring device 80. The monitoring device 80 stores the second estimation model 831 in the storage unit 85.

[0078] <Processing Procedure for Generating First Estimation Model> Fig.12 is a flowchart showing the processing steps of the first estimation model generation process. Fig.12 The processing steps shown generate a first estimation model 821 for detecting friction noise.

[0079] The process of generating the first estimation model 821 includes reading Fig.11 The learning data processing shown is (step S331), Fourier transform processing (step S332), smoothing processing (step S333), segmentation processing (step S334) and normalization processing (step S335).

[0080] The details of the Fourier transform process, smoothing process, segmentation process, and normalization process have been described using Figure 4 The learning device 8 reads the learning data, performs various processes on the measurement data (steps S332 to S335), and then repeatedly learns the first estimation model 821 using the learning data (step S336). The learning device 8 outputs the learned first estimation model 821 to the monitoring device 80 (step S337). The monitoring device 80 stores the first estimation model 821 in the storage unit 85.

[0081] Here, the processing steps for generating the first estimation model 821 are described, but the learning device 8 may generate the second estimation model 831 in the same steps. When generating the second estimation model 831, the learning device 8 may not execute each process (steps S332 to S335) but use Figure 5 The second estimation model 831 is generated based on the measurement data itself.

[0082] As described above, according to this embodiment, since the friction noise can be detected with high accuracy before the bearing is diagnosed, the frequency of misdiagnosis of bearing damage and friction noise can be reduced. As a result, the diagnostic performance of the bearing can be improved.

[0083] Here, a method of detecting friction noise using the first estimation model 821 is exemplified. However, the friction noise detection unit 82 may detect friction noise by a method different from the method using the first estimation model 821. As already described, when friction noise occurs, the amplitude values ​​of multiple frequency bands of the spectrum obtained by Fourier transforming the measurement data show characteristic changes. This change is not related to whether the bearing is damaged, but is a change unique to friction noise.

[0084] Therefore, the friction noise detection unit 82 may detect the friction noise by comparing the amplitude values ​​of several specific frequency bands of the spectrum with threshold values ​​predetermined corresponding to each frequency band. However, in order to further improve the detection accuracy, it is preferable to use a method of detecting the friction noise using the first estimation model 821.

[0085] Here, a method of performing bearing diagnosis using the second estimation model 831 is illustrated. However, the bearing diagnosis unit 83 may perform bearing diagnosis using a method different from the method using the second estimation model 831. For example, the bearing diagnosis unit 83 may also calculate the effective value, peak value, average value, peak factor, effective value after envelope processing, and peak value after envelope processing, etc. based on the measurement data of the sensor using a statistical method as described in Patent Gazette No. 5917956, and compare the calculated value with the threshold value to detect bearing abnormality.

[0086] Alternatively, the bearing diagnosis unit 83 may calculate a frequency spectrum by performing envelope analysis and frequency analysis on the signal obtained from the sensor as described in Patent Gazette No. 5146008, and compare the frequency spectrum with a threshold value for each bearing portion to identify the portion of the bearing where an abnormality has occurred.

[0087] <Variation 1> exist Figure 4 In the friction noise detection process shown above, an example is described in which the entire generated measurement data (waveform data) is processed collectively. Here, as a modification 1, an example is described in which the measurement data is divided into a plurality of segments and processed in units of the segments.

[0088] Friction noise does not always occur. Friction noise occurs when multiple conditions such as clearance, sliding, vibration, and oil film changes in the bearing are consistent. Sometimes friction noise occurs continuously in the measurement data representing the change of the measured value over time, and sometimes friction noise only occurs for 1 or 2 seconds. Therefore, the waveform data of the part of the measurement data that is not mixed with friction noise can be used for bearing diagnosis.

[0089] Here, a method is proposed to divide a measurement data into multiple segments, and after determining the segment that does not contain friction noise, use the measurement data to perform bearing diagnosis. Specifically, the segment containing friction noise can be removed, and only the segment containing no friction noise can be used to perform bearing diagnosis. Alternatively, a method can be adopted in which the value of the segment containing friction noise is replaced with a specified value (for example, zero).

[0090] Fig.13 2 is a conceptual diagram for explaining a method of detecting friction noise according to Modification 1. Fig.13 The measurement data is data of a time length T1 generated based on the measurement value of the sensor 70 . The measurement data is stored in the storage unit 85 .

[0091] The friction sound detection unit 82 divides the measurement data into segments of a time length T2 shorter than the time length T1 before performing Fourier transform processing on the measurement data. Thus, one measurement data is divided into a plurality of segments (variable i=1 to N). For example, the time length T2 can be set to 1 second or 2 seconds. The monitoring device 80 can also be designed to accept the user's setting input of the time length T2. The friction sound detection unit 82 detects the friction sound for each segment.

[0092] Fig.14 4 is a flowchart showing the processing steps of the friction noise detection process according to Modification 1. After reading the measurement data from the storage unit 85 (step S411), the friction noise detection unit 82 divides the measurement data into segments (step S412). Thus, the measurement data is divided into a plurality of segments (variable i=1 to N).

[0093] Next, the friction noise detection unit 82 sets an initial value (=1) for the variable i (step S413). Next, the friction noise detection unit 82 performs the following operations in sequence: Figure 4 The Fourier transform processing, smoothing processing, segmentation processing and normalization processing shown in steps S102 to S105 are performed (step S414).

[0094] Thus, normalized data is generated from the waveform data in segments. Next, the friction sound detection unit 82 reads the first estimation model 821 from the storage unit 85 (step S415). Next, the friction sound detection unit 82 inputs the normalized data in segments to the first estimation model 821 (step S416). The first estimation model 821 detects friction sound from the normalized data in segments. A result indicating the presence or absence of friction sound is output from the first estimation model 821.

[0095] Next, the friction noise detection unit 82 determines whether the friction noise is detected in the normalized data in segments based on the detection result of the first estimation model 821 (step S417). If the friction noise is detected in the normalized data in segments, the segment i is stored in the storage unit 85 as a correction object (step S418). If the friction noise is not detected in the normalized data in segments, step S418 is not performed. Therefore, the segment that does not contain the friction noise is not used as a correction object.

[0096] Next, the friction noise detection unit 82 updates the variable i (step S419), and determines whether the updated variable i exceeds the maximum value N (step S420). If the updated variable i does not exceed N, the friction noise detection unit 82 repeats the processing of steps S414 to S419. Thus, the designated values ​​(i) of all the segments containing the friction noise are stored in the storage unit 85.

[0097] Next, the friction sound detection unit 82 performs measurement data correction processing (step S421). Thus, the value of the portion of the measurement data corresponding to the segment containing the friction sound is corrected. As a correction method, two methods are proposed here. One method is to remove the segment containing the friction sound and only use the segment not containing the friction sound as data. Another method is to replace the value of the segment containing the friction sound with a specified value (for example, zero).

[0098] Then, the friction noise detection unit 82 outputs the corrected measurement data to the bearing diagnosis unit 83, causing the bearing diagnosis unit 83 to perform bearing diagnosis ( Fig.10 Thus, the friction noise detection unit 82 ends the friction noise detection process according to the modification example 1.

[0099] According to variant 1, even if one piece of measurement data contains some friction noise, the measurement data can be used for bearing diagnosis. Therefore, more measurement data can be effectively used for diagnosis. As a result, the frequency of bearing diagnosis can be increased. By increasing the frequency of bearing diagnosis, bearing damage can be discovered at an early stage. In addition, when performing so-called trend analysis, the reliability of diagnosis can also be improved due to the increase in the number of data points.

[0100] <Variation 2> Fig.15 80 is a flowchart showing the processing steps of the bearing diagnosis process involved in Modification Example 2. When the monitoring device 80 determines the presence or absence of friction noise based on the output of the first estimation model 821, it may use the output of the probability (%) of occurrence of friction noise instead of the output of the presence or absence of friction noise (binary classification). Fig.15 In the bearing diagnostic process shown, Fig.10 Compared with the bearing diagnosis process shown in FIG. 1 , the determination step of step S202a is used instead of step S202. First, the monitoring device 80 refers to the output of the friction sound output from the first estimation model 821 (step S201). In particular, in the second modification, the monitoring device 80 refers to the occurrence probability of the friction sound output from the first estimation model 821.

[0101] Next, the monitoring device 80 compares the probability of occurrence of the friction sound with a predetermined threshold value (step S202a). When the probability of occurrence of the friction sound exceeds the threshold value, the monitoring device 80 determines that there is friction sound and executes the processing of step S208. When the probability of occurrence of the friction sound does not exceed the threshold value, the monitoring device 80 determines that there is no friction sound and executes the processing of steps S203 to S207. The threshold value can be appropriately set to 70%, etc. The monitoring device 80 can also be designed to accept user input for setting the threshold value. The details of the processing of steps S203 to S207 have been used. Fig.10 have been explained, so their description will not be repeated here.

[0102] According to Modification 2, the presence or absence of friction noise can be determined based on the evaluation of the occurrence probability output from the first estimation model 821 using a threshold value.

[0103] <Variation 3> Fig.16 : is a flowchart showing the processing procedure of the bearing diagnosis process according to Modification 3. As Modification 3, an example is proposed in which the monitoring device 80 outputs the occurrence frequency of friction noise as an alarm to the outside.

[0104] exist Fig.16 In the bearing diagnostic process shown, Fig.10 Compared with the bearing diagnosis process shown in FIG. 1 , the process of steps S217 to S219 is used instead of step S208. Step S202 of "whether friction noise is detected" is a branch for evaluating the detection result of friction noise. Therefore, the monitoring device 80 may also refer to the history of past detection results when friction noise is detected, and output an alarm indicating that the occurrence frequency of friction noise is high when the detection frequency of friction noise exceeds a predetermined frequency.

[0105] Specifically, the monitoring device 80 executes Fig.16The monitoring device 80 records the occurrence of the friction noise and the detection date and time in the history data when the friction noise is detected (step S217). The monitoring device 80 stores the history data in the storage unit 85 and updates the history data every time the friction noise is detected.

[0106] Next, the monitoring device 80 refers to the historical data to determine the detection frequency (high / medium / low) of the friction sound (step S218). The monitoring device 80 uses the first threshold and the second threshold (the first threshold > the second threshold) to determine the "high", "medium" and "low" of the detection frequency. The monitoring device 80 determines that the detection frequency is "high" when the detection frequency exceeds the first threshold; when the detection frequency is below the second threshold, it determines that the detection frequency is "low"; and when the detection frequency is other than these, it determines that the detection frequency is "medium". The first threshold and the second threshold can be set appropriately. The monitoring device 80 can also be designed to accept user input for setting the first threshold and the second threshold.

[0107] Next, the monitoring device 80 sets an alarm (third alarm) indicating the frequency corresponding to the determination result (step S219). In addition to the alarm indicating the frequency corresponding to the determination result, the monitoring device 80 may also set an alarm for the friction sound (see Fig.10 The alarm output unit 84 outputs the set alarm to the alarm device 15.

[0108] According to the third modification, a worker who maintains the wind turbine generator 10 can identify the frequency of occurrence of the friction noise based on the alarm issued by the alarm device 15 .

[0109] <Variation 4> Fig.17 This is a diagram for explaining a method in which the learning device 8 generates the second estimation model 832 according to Modification 4. Here, a method for generating the second estimation model 832 is proposed using not only measurement data but also the output of the learned first estimation model 821 as a feature amount.

[0110] like Fig.17 As shown, the friction noise detection unit 82 reads the first estimation model 821 that has been learned, and the learning device 8 reads the second estimation model 832 that is the object of machine learning. A plurality of measurement data are synchronously input to the friction noise detection unit 82 and the learning device 8. The plurality of data includes measurement data obtained when friction noise is generated, measurement data obtained when bearing abnormality occurs, and measurement data obtained when friction noise and bearing abnormality are generated at the same time.

[0111] like Fig.17As shown, the learning data of the second estimation model 832 is composed of the output of the learned first estimation model 821, the measurement data, and the answer data (the presence or absence of bearing abnormality) to the measurement data.

[0112] When receiving the measurement data, the friction sound detection unit 82 detects the friction sound using the first estimation model 821. The first estimation model 821 outputs the detection result to the learning device 8 as one of the learning data of the second estimation model 832. The detection result of the first estimation model 821 can be information indicating only the presence or absence of the friction sound (binary classification) or the probability (%) of the occurrence of the friction sound. In addition to the detection result of the first estimation model 821, the measurement data and the answer data are input to the learning device 8. The measurement data input here is the same as the measurement data for which the first estimation model 821 calculates the detection result.

[0113] The learning device 8 generates a learned second estimation model 832 by repeating machine learning using a plurality of learning data. In this way, the second estimation model 832 is generated by machine learning based on the measurement data and the output value of the learned first estimation model 821. The learned second estimation model 832 is stored in the storage unit 85 of the monitoring device 80.

[0114] Fig.18 FIG. 4 is a flowchart showing the processing steps of the bearing diagnosis process involved in the modification example 4. In the bearing diagnosis process involved in the modification example 4, the use of Fig.17 The second inferred model 832 that has been learned is described. Fig.18 In the bearing diagnosis process shown, the monitoring device 80 reads the second estimation model 832 from the storage unit 85 (step S231 ).

[0115] Next, the monitoring device 80 inputs the output (detection result) and measurement data of the first estimation model 821 into the second estimation model 832 (step S232). The measurement data input into the second estimation model 832 is the same as the measurement data input into the first estimation model 821 that has completed learning via the friction sound detection unit 82.

[0116] The second estimation model 832 diagnoses the state of the bearing based on the measurement data and the output of the first estimation model 821. The second estimation model 832 outputs the diagnosis result of the bearing. Processing and use of steps S233 to S235 Fig.10 The processes of step S205 to step S207 described above are the same, and therefore, the description of these processes will not be repeated here.

[0117] The second estimation model 832 involved in Modification 4 is generated using not only the measurement data but also the learning data including the information on the presence or absence of friction noise. Therefore, for example, it is possible to improve the diagnostic accuracy of the monitoring device 80 when performing bearing diagnosis based on the measurement data obtained when friction noise and bearing damage occur simultaneously.

[0118] <Changes in the structure of the monitoring device 80> Here, a change in the structure of the monitoring device 80 is described. Figure 1 As shown, the monitoring device 80 may be arranged on the wind turbine generator 10 including the bearing 51 to be monitored. Alternatively, the monitoring device 80 may be connected to the wind turbine generator 10 via a network such as the Internet. In this case, the measurement value of the sensor 70 in the wind turbine generator 10 is transmitted to the monitoring device 80 via the network.

[0119] like Figure 2 As shown, the monitoring device 80 may be constituted by a general-purpose computer (processing device). In this case, each component such as the acquisition unit 81, the friction noise detection unit 82, the bearing diagnosis unit 83, and the alarm output unit 84 for realizing the monitoring device 80 is realized by a single processing device.

[0120] Alternatively, the monitoring device 80 may be configured by distributing the components by a plurality of processing devices. In this case, the monitoring device 80 is realized by a collection of a plurality of processing devices. When the monitoring device 80 is realized by a collection of a plurality of processing devices, such a monitoring device 80 may also be referred to as a monitoring system. The monitoring system includes the first processing device 80A and the second processing device 80B, and includes the sensor 70, and may be referred to as a monitoring system.

[0121] Fig.19 It is a diagram for explaining a specific example of a change in the structure of the monitoring device 80. Fig.19 The first mode in which the monitoring device 80 is composed of the first processing device 80A and the second to fourth modes in which the monitoring device 80 is composed of the first processing device 80A and the second processing device 80B are shown. The second processing device 80B is an external device that is communicably connected to the first processing device 80A via a network 5 such as the Internet. The second processing device 80B may also be composed of a cloud server. The configuration of the monitoring device 80 is not limited to Fig.19 The situation shown.

[0122] The first mode is an example in which the acquisition unit 81, the friction noise detection unit 82, the bearing diagnosis unit 83, and the alarm output unit 84 are provided in the first processing device 80A. Figure 2 and Figure 3 The structure shown.

[0123] The second mode is an example in which the first processing device 80A is provided with the acquisition unit 81 and the friction noise detection unit 82 , and the second processing device 80B is provided with the bearing diagnosis unit 83 and the alarm output unit 84 . In the second mode, the friction noise detection unit 82 sends the detection result to the bearing diagnosis unit 83 via the network 5 .

[0124] The third mode is an example in which the first processing device 80A is provided with the acquisition unit 81 and the second processing device 80B is provided with the friction noise detection unit 82 , the bearing diagnosis unit 83 and the alarm output unit 84 . In the third mode, the acquisition unit 81 sends the measurement value to the friction noise detection unit 82 via the network 5 .

[0125] The fourth mode is an example in which the acquisition unit 81 and the alarm output unit 84 are provided in the first processing device 80A, and the friction noise detection unit 82 and the bearing diagnosis unit 83 are provided in the second processing device 80B. In the third mode, the acquisition unit 81 sends the measured value to the friction noise detection unit 82 via the network 5, and the bearing diagnosis unit 83 sends the diagnosis result to the alarm output unit 84 via the network 5.

[0126] Thus, in this embodiment, in addition to the structure (edge ​​computing) in which a single first processing device 80A performs a plurality of computational processing required for monitoring the bearing, a structure in which a second processing device 80B or a plurality of processing devices performs a portion of these computational processing in a distributed manner is also envisioned. The second processing device 80B may be composed of an external terminal such as a personal computer, a smart phone, and a tablet computer. The second processing device 80B may also be composed of a cloud server.

[0127] <Other Modifications> The monitoring device 80 involved in this embodiment sequentially performs Fourier transform processing, smoothing processing, segmentation processing, and normalization processing. However, this embodiment does not exclude the application of other processing as needed between each processing. For example, before or after any of these processing, a certain processing for more detailed noise removal or a certain supplementary processing may be applied.

[0128] The monitoring device 80 according to the present embodiment detects the friction noise using the normalized data obtained by the normalization process. Thus, the friction noise can be detected with higher accuracy. However, the present embodiment does not exclude the method of detecting the friction noise using segmented data or smoothed data instead of the normalized data.

[0129] In addition, this embodiment does not exclude the method of using the spectrum before smoothing to replace the normalized data to detect the friction noise. One of the characteristics of this embodiment is that before the bearing diagnosis is performed, the process of detecting the friction noise is performed separately from the process of bearing diagnosis, and the bearing diagnosis is performed based on the data without the influence of the friction noise.

[0130] The machine learning of the second estimation model 831 may be performed using the measurement data itself that has not been subjected to Fourier transform processing, or using the spectrum that has been subjected to Fourier transform processing. The machine learning of the second estimation model 831 may be performed using normalized data, as in the case of performing machine learning on the first estimation model 821. Data in a format used for machine learning is input to the learned second estimation model 831.

[0131] The monitoring device 80 may perform friction noise detection and bearing diagnosis on a plurality of types of bearings provided in the wind turbine generator 10 .

[0132] [Way] The following lists the aspects of the present disclosure.

[0133] (Item 1) The monitoring device described in Item 1 is a monitoring device for monitoring equipment including bearings, and a sensor for measuring a physical quantity that changes with the vibration of the equipment is installed on the equipment. The monitoring device includes an acquisition unit for acquiring a measurement value of the sensor, a detection unit for detecting a friction sound of the bearing based on the measurement value, and a diagnosis unit for diagnosing a state of the bearing based on the measurement value, and the diagnosis unit uses the measurement value that does not include the friction sound to diagnose the bearing.

[0134] (Item 2) In the monitoring device described in Item 1, the diagnosis unit determines the measurement value that does not include the friction sound detected by the detection unit by determining the measurement value of the friction sound detected by the detection unit.

[0135] (Item 3) In the monitoring device as described in Item 1, the detection unit divides the measurement data representing the time-dependent variation of the measurement value of the sensor into a plurality of segmented data in the direction of the time axis, detects the friction sound for each segmented data, the detection unit corrects the segmented data portion of the measurement data in which the friction sound is detected, and the diagnosis unit diagnoses the bearing using the corrected measurement data as the measurement value that does not include the friction sound detected by the detection unit.

[0136] (Item 4) In the monitoring device as described in Item 3, the correction is to delete the segment data part containing the friction sound from the measurement data.

[0137] (Item 5) In the monitoring device described in Item 3, the correction is to change the measured value of the segmented data part including the friction sound to a specified value.

[0138] (Item 6) In the monitoring device described in any one of Items 1 to 3, the detection unit sequentially performs Fourier transform processing, smoothing processing, segmentation processing, and normalization processing on measurement data representing the time-dependent changes in the measurement values ​​of the sensor, and the detection unit inputs the normalized data generated by sequentially performing each processing into the first estimation model for detecting friction sound, thereby detecting the friction sound.

[0139] (Item 7) In the monitoring device as described in Item 6, the detection unit divides the measurement data representing the temporal variation of the measurement value of the sensor into a plurality of segmented data in the direction of the time axis, and inputs the normalized data in segments generated by sequentially performing each process on the segmented data into the first inference model to detect friction sound.

[0140] (Item 8) In the monitoring device described in Item 6 or Item 7, the first estimation model is generated by using measurement data including fricative sounds as feature data and by machine learning based on any algorithm of decision tree, random forest, support vector machine, and neural network.

[0141] (Item 9) In the monitoring device according to any one of Items 6 to 8, the detection unit outputs the presence or absence of friction noise or the probability of occurrence of friction noise as the detection result of friction noise.

[0142] (Item 10) In the monitoring device according to any one of Items 1 to 9, the sensor is any one of a vibration sensor, an acoustic sensor, and an AE sensor.

[0143] (Item 11) In the monitoring device described in any one of Items 1 to 10, the monitoring device includes a first processing device and a second processing device, the first processing device includes an acquisition unit, and the second processing device includes at least one of a detection unit and a diagnosis unit.

[0144] (Item 12) In the monitoring device described in Item 11, the first processing device and the second processing device are connected to each other so as to be communicable via a network.

[0145] (Item 13) The monitoring device according to any one of Items 1 to 12 further includes an output unit that outputs an alarm, wherein the output unit outputs a first alarm when the detection unit detects the friction sound.

[0146] (Item 14) In the monitoring device described in Item 13, the output unit outputs a second alarm when the diagnosis unit diagnoses an abnormality in the bearing.

[0147] (Item 15) In the monitoring device described in Item 13 or 14, the output unit outputs a third alarm when the detection frequency of the friction sound exceeds a threshold value.

[0148] (Item 16) In the monitoring device according to any one of Items 13 to 15, the alarm includes one or more forms of an electric signal, a message, and a data file.

[0149] (Item 17) The monitoring device described in Item 17 is a monitoring device for monitoring equipment including bearings, and a sensor for measuring a physical quantity that changes with the vibration of the equipment is installed on the equipment. The monitoring device includes: an acquisition unit for acquiring a measurement value of the sensor, a detection unit for detecting the friction sound of the bearing based on the measurement value, and a diagnosis unit for diagnosing the state of the bearing based on the measurement value, the detection unit inputs data generated based on the measurement value into a first estimation model for detecting friction sound to detect the friction sound, and the diagnosis unit inputs the detection result of the detection unit and the data generated based on the measurement value into a second estimation model for diagnosing the bearing to diagnose the bearing.

[0150] (Item 18) In the monitoring device described in Item 17, a second estimation model is generated by machine learning to diagnose the bearing based on the detection results and data of the detection unit.

[0151] (Item 19) The monitoring method described in Item 19 is a monitoring method for monitoring the state of a device including a bearing, wherein a sensor for measuring a physical quantity that changes with the vibration of the device is installed on the device, and the monitoring method comprises: a step of acquiring a measurement value of the sensor, a step of detecting a friction sound of the bearing based on the measurement value, and a step of diagnosing the state of the bearing based on the measurement value, wherein the diagnosis step comprises a step of diagnosing the bearing using a measurement value that does not include a friction sound.

[0152] All contents of the embodiments disclosed this time should be considered as illustrative rather than limiting. The scope of the present disclosure should be considered to be indicated not by the description of the above embodiments but by the claims, and include all changes within the meaning and scope equivalent to the claims. Description of symbols

[0153] 1 monitoring device, 5 network, 8 learning device, 10 wind turbine generator, 15 alarm device, 20 main shaft, 25 hub, 30 blade, 40 speed increaser, 50 generator, 60 main shaft bearing, 70 sensor, 80 monitoring device (processing device), 80A first processing device, 80B second processing device, 81 acquisition unit, 82 friction sound detection unit, 83 bearing diagnosis unit, 84 alarm output unit, 85 storage unit, 90 nacelle, 100 tower, 401 bearing, 801 CPU, 802 RAM, 803 storage, 804 communication interface, 805 bus, 806 monitoring program, 821 first estimation model, 831, 832 second estimation model.

Claims

1. A monitoring device for monitoring the state of a device including a bearing, the monitoring device being characterized in that: A sensor for measuring a physical quantity that changes with vibration of the device is installed on the device. The monitoring device comprises: An acquisition unit for acquiring a measurement value of the sensor; a detection unit for detecting the friction noise of the bearing according to the measured value; as well as a diagnosis unit for diagnosing a state of the bearing based on the measured value; The diagnosis unit diagnoses the bearing using the measurement value that does not include the friction noise.

2. The monitoring device according to claim 1, characterized in that The diagnosis unit determines the measurement values ​​that do not include the friction sound detected by the detection unit by determining the measurement values ​​in which the friction sound is detected by the detection unit.

3. The monitoring device according to claim 1, wherein: The detection unit divides the measurement data representing the temporal change of the measurement value of the sensor into a plurality of segmented data in the time axis direction, detects the friction sound for each segmented data, and corrects the segmented data portion of the measurement data in which the friction sound is detected. The diagnosis unit diagnoses the bearing using the corrected measurement data as the measurement value excluding the friction noise detected by the detection unit.

4. The monitoring device according to claim 3, characterized in that The correction is to delete the segment data portion containing the friction sound from the measurement data.

5. The monitoring device according to claim 3, characterized in that The correction is to change the measured value of the segment data portion including the friction sound to a predetermined value.

6. The monitoring device according to any one of claims 1 to 3, characterized in that The detection unit sequentially performs Fourier transform processing, smoothing processing, segmentation processing, and normalization processing on measurement data indicating temporal changes in the measurement value of the sensor. The detection unit detects the friction sound by inputting normalized data generated by sequentially executing the respective processes into a first estimation model for detecting the friction sound.

7. The monitoring device according to claim 6, characterized in that The detection unit divides the measurement data representing the temporal variation of the measurement value of the sensor into a plurality of segmented data in the time axis direction, and inputs the normalized data in segments generated by sequentially performing the respective processes on the segmented data into the first estimation model to detect the friction sound.

8. The monitoring device according to claim 6 or 7, characterized in that The first estimation model is generated by machine learning based on any algorithm of decision tree, random forest, support vector machine, and neural network, using the measurement data including the friction sound as feature amount data.

9. The monitoring device according to any one of claims 6 to 8, characterized in that The detection unit outputs the presence or absence of the friction noise or the probability of occurrence of the friction noise as a detection result of the friction noise.

10. The monitoring device according to any one of claims 1 to 9, characterized in that The sensor is any one of a vibration sensor, an acoustic sensor and an AE sensor.

11. The monitoring device according to any one of claims 1 to 10, characterized in that: The monitoring device includes a first processing device and a second processing device, The first processing device includes the acquisition unit, The second processing device includes at least one of the detection unit and the diagnosis unit.

12. The monitoring device according to claim 11, characterized in that The first processing device and the second processing device are connected to each other so as to be communicable via a network.

13. The monitoring device according to any one of claims 1 to 12, characterized in that: Also includes an output unit for outputting an alarm, The output unit outputs a first alarm when the detection unit detects the friction sound.

14. The monitoring device according to claim 13, characterized in that The output unit outputs a second alarm when the diagnosis unit diagnoses that the bearing has an abnormality.

15. The monitoring device according to claim 13 or 14, characterized in that The output unit outputs a third alarm when the detection frequency of the friction sound exceeds a threshold value.

16. The monitoring device according to claim 13 or 14, characterized in that The alarm includes any one or more forms of an electrical signal, a message, or a data file.

17. A monitoring device for monitoring the status of a device including a bearing, the monitoring device being characterized in that: A sensor for measuring a physical quantity that changes with vibration of the device is installed on the device. The monitoring device comprises: An acquisition unit for acquiring a measurement value of the sensor; a detection unit for detecting the friction noise of the bearing according to the measured value; as well as a diagnosis unit for diagnosing a state of the bearing based on the measured value; The detection unit detects the friction sound by inputting data generated based on the measurement value into a first estimation model for detecting the friction sound, The diagnosis unit inputs the detection result of the detection unit and the data generated based on the measurement value into a second estimation model for diagnosing the bearing, and diagnoses the bearing.

18. The monitoring device according to claim 17, characterized in that The second estimation model is generated by machine learning to diagnose the bearing based on the detection result of the detection unit and data generated based on the measurement value.

19. A monitoring method for monitoring the status of a device including a bearing, the monitoring method being characterized in that: A sensor for measuring a physical quantity that changes with vibration of the device is installed on the device. The monitoring method comprises: A step of obtaining a measurement value of the sensor; A step of detecting the friction noise of the bearing according to the measured value; and The step of diagnosing the state of the bearing based on the measured values, The diagnosing step includes the step of diagnosing the bearing using the measured value that does not include the friction noise.

Images