JP2017173321A - State monitoring system and wind power generator - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a state monitoring system and a wind power generator that realize accurate diagnosis of abnormality.SOLUTION: The state monitoring system comprises a vibration sensor 70 for measuring the vibration waveform of a machine element constituting a device, and a processing device 80 for diagnosing abnormality of the machine element. In the processing device 80, an evaluation value computation unit 160 is constituted so as to compute continuously in time an evaluation value that characterizes vibration waveform data outputted from the vibration sensor 70 in a fixed time. A diagnosis unit 150 is constituted so as to start measurement of vibration waveform data as triggered by a change in the tendency of a temporal change of the evaluation value computed by the evaluation value computation unit 160 and thereby diagnose abnormality of the machine element using the vibration waveform data.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a state monitoring system that monitors the state of machine elements that constitute a device, and more particularly, to a state monitoring system that monitors the state of machine elements that constitute a wind turbine generator.

  In a wind turbine generator, power is generated by rotating a main shaft connected to a blade that receives wind power, rotating the main shaft with a speed increaser, and then rotating a rotor of the power generator. Each of the main shaft, the speed increaser, and the rotating shaft of the generator is rotatably supported by a rolling bearing, and a condition monitoring system (CMS: Condition Monitoring System) for diagnosing such a bearing abnormality is known. . In such a state monitoring system, whether or not the bearing is damaged is diagnosed using vibration waveform data measured by a vibration sensor fixed to the bearing.

  As a state monitoring system for monitoring the state of an internal device of a general apparatus incorporating a computing unit or the like, for example, Japanese Patent Laid-Open No. 7-159469 (Patent Document 1) discloses a circuit for measuring the internal state of a device under measurement. When an abnormality of the device under test is detected based on a signal output from the device, a configuration is shown in which a trigger for starting measurement and recording is generated for each measurement element circuit and recording unit in response to the detection signal. ing.

JP-A-7-159469

  However, in the state monitoring system disclosed in Patent Document 1 described above, when noise is superimposed on a signal output from a circuit that measures the internal state of the device under measurement, an abnormality is detected in the device under measurement and a trigger is generated. The case where it is done arises. In such a case, since a trigger is frequently generated under the influence of noise, measurement and recording of the internal state of the device under test is started each time the trigger is generated. As a result, a huge amount of data is accumulated in the recording unit in a state where data when the device under measurement is normal and data when the device under measurement is abnormal are mixed. As a result, there is a possibility that accurate abnormality diagnosis based on data accumulated in the recording unit cannot be performed.

  Accordingly, the present invention has been made to solve such a problem, and an object of the present invention is to provide a state monitoring system and a wind turbine generator that realize an accurate abnormality diagnosis.

  A state monitoring system according to the present invention is a state monitoring system for monitoring a state of a machine element constituting an apparatus, and includes a vibration sensor for measuring a vibration waveform of the machine element, and a diagnosis of an abnormality of the machine element. And a processing device. The processing device includes an evaluation value calculation unit and a diagnosis unit. The evaluation value calculation unit is configured to continuously calculate an evaluation value characterizing the vibration waveform data output from the vibration sensor within a predetermined time. The diagnosis unit uses the vibration waveform data to diagnose abnormalities in machine elements by starting measurement of vibration waveform data triggered by a change in the tendency of the evaluation value calculated by the evaluation value calculation unit over time. Configured to do.

  According to the present invention, vibration waveform data when an abnormality has occurred in the mechanical elements constituting the apparatus can be acquired reliably and appropriately, so that an accurate abnormality diagnosis can be realized.

It is the figure which showed schematically the structure of the wind power generator to which the state monitoring system which concerns on Embodiment 1 of this invention was applied. It is a functional block diagram which shows the structure of the data processor shown in FIG. 1 functionally. It is a figure explaining operation | movement of the vibration waveform data storage part and evaluation value calculating part which were shown in FIG. It is a figure explaining operation | movement of the evaluation value trend memory | storage part shown in FIG. It is a figure explaining operation | movement of the measurement trigger generation | occurrence | production part and vibration waveform data output part shown in FIG. 6 is a flowchart illustrating a control process for storing bearing vibration waveform data in the state monitoring system according to the first embodiment. 6 is a flowchart illustrating a control process for generating a measurement trigger for vibration waveform data of a bearing in the state monitoring system according to the first embodiment. 10 is a flowchart illustrating a control process for generating a measurement trigger for vibration waveform data of a bearing in the state monitoring system according to the second embodiment. 10 is a flowchart illustrating a control process for generating a measurement trigger for vibration waveform data of a bearing in the state monitoring system according to the third embodiment. It is a functional block diagram which shows functionally the structure of the data processor in the state monitoring system which concerns on Embodiment 4. FIG. It is a conceptual diagram which shows the relationship between the vibration waveform data of a bearing in a fixed time, and a segment. It is a figure for demonstrating a feature-value vector. It is a figure for demonstrating the basic concept of OC-SVM.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals, and the description thereof will not be repeated.

[Embodiment 1]
FIG. 1 is a diagram schematically showing a configuration of a wind turbine generator to which a state monitoring system according to Embodiment 1 of the present invention is applied. Referring to FIG. 1, a wind turbine generator 10 includes a main shaft 20, a blade 30, a speed increaser 40, a generator 50, a main shaft bearing (hereinafter simply referred to as “bearing”) 60, and vibration. A sensor 70 and a data processing device 80 are provided. The step-up gear 40, the generator 50, the bearing 60, the vibration sensor 70, and the data processing device 80 are stored in the nacelle 90, and the nacelle 90 is supported by the tower 100.

  The main shaft 20 enters the nacelle 90 and is connected to the input shaft of the speed increaser 40 and is rotatably supported by the bearing 60. The main shaft 20 transmits the rotational torque generated by the blade 30 receiving the wind force to the input shaft of the speed increaser 40. The blade 30 is provided at the tip of the main shaft 20 and converts wind force into rotational torque and transmits it to the main shaft 20.

  The bearing 60 is fixed in the nacelle 90 and rotatably supports the main shaft 20. The bearing 60 is composed of a rolling bearing, and is composed of, for example, a self-aligning roller bearing, a tapered roller bearing, a cylindrical roller bearing, or a ball bearing. These bearings may be single row or double row.

  The vibration sensor 70 is fixed to the bearing 60. The vibration sensor 70 measures the vibration waveform of the bearing 60 and outputs the measured vibration waveform data to the data processing device 80. The vibration sensor 70 is constituted by, for example, an acceleration sensor using a piezoelectric element.

  The speed increaser 40 is provided between the main shaft 20 and the generator 50, and increases the rotational speed of the main shaft 20 to output to the generator 50. As an example, the speed increaser 40 is configured by a gear speed increasing mechanism including a planetary gear, an intermediate shaft, a high speed shaft, and the like. In addition, although not specifically illustrated, a plurality of bearings that rotatably support a plurality of shafts are also provided in the speed increaser 40.

  The generator 50 is connected to the output shaft of the speed increaser 40, and generates power by the rotational torque received from the speed increaser 40. The generator 50 is constituted by, for example, an induction generator. A bearing for rotatably supporting the rotor is also provided in the generator 50.

  The data processing device 80 is provided in the nacelle 90 and receives vibration waveform data of the bearing 60 from the vibration sensor 70. The data processing device 80 diagnoses the abnormality of the bearing 60 using the vibration waveform data of the bearing 60 according to a preset program.

  FIG. 2 is a functional block diagram functionally showing the configuration of the data processing device 80 shown in FIG. Referring to FIG. 2, data processing device 80 includes a band pass filter (hereinafter referred to as “BPF (Band Pass Filter)”) 110, an effective value calculation unit 120, a storage unit 130, and vibration waveform data output. Unit 140, diagnosis unit 150, evaluation value calculation unit 160, and measurement trigger generation unit 170.

  The BPF 110 receives vibration waveform data of the bearing 60 from the vibration sensor 70. The BPF 110 includes, for example, a high pass filter (HPF (High Pass Filter)). The HPF passes a signal component higher than a predetermined frequency with respect to the received vibration waveform data, and blocks a low frequency component. The HPF is provided to remove a direct current component included in the vibration waveform data of the bearing 60. Note that the HPF may be omitted if the output of the vibration sensor 70 does not include a DC component.

  The BPF 110 may include a low pass filter (LPF (Low Pass Filter)) in addition to the HPF or instead of the HPF. The LPF passes a signal component lower than a predetermined frequency for the received vibration waveform data.

  Further, an envelope processing unit may be provided between the vibration sensor 70 and the BPF 110. When the envelope processing unit receives vibration waveform data of the bearing 60 from the vibration sensor 70, the envelope processing unit performs envelope processing on the received vibration waveform data to generate an envelope waveform of the vibration waveform data of the bearing 60. Various known methods can be applied to the envelope processing calculated in the envelope processing unit. For example, the vibration waveform data of the bearing 60 received from the vibration sensor 70 is rectified to an absolute value and passed through the LPF. Thus, an envelope waveform of the vibration waveform data of the bearing 60 is generated.

  In this case, in the BPF 110, when the HPF receives the envelope waveform of the vibration waveform data of the bearing 60 from the envelope processing unit, the HPF passes a signal component higher than a predetermined frequency with respect to the received envelope waveform, and the low frequency component Shut off. That is, the HPF is configured to remove a direct current component included in the envelope waveform and extract an alternating current component of the envelope waveform.

  The effective value calculation unit 120 receives vibration waveform data of the bearing 60 subjected to the filter processing from the BPF 110. The effective value calculation unit 120 calculates an effective value (also referred to as “RMS (Root Mean Square) value)) of the vibration waveform data of the bearing 60, and stores the calculated effective value of the vibration waveform data in the storage unit 130. Output to.

  Since the effective value of the vibration waveform of the bearing 60 calculated by the effective value calculation unit 120 is the effective value of the raw vibration waveform that has not been subjected to the envelope processing, for example, separation occurs in a part of the bearing ring of the bearing 60. However, although the increase in the value is small for the impulse-like vibration in which the vibration increases only when the rolling element passes through the peeled portion, it occurs when the surface of the contact portion between the race ring and the rolling element is rough or poorly lubricated. For continuous vibrations, the value increases.

  On the other hand, as described above, when the envelope processing unit is provided, the effective value of the alternating current component of the envelope waveform calculated by the effective value calculating unit 120 is the continuous value generated when the raceway surface is rough or poorly lubricated. The increase in value is small for vibration, and the increase is large for impulse vibration.

  The storage unit 130 includes a vibration waveform data storage unit 132 and an evaluation value trend storage unit 134. The vibration waveform data storage unit 132 and the evaluation value trend storage unit 134 are configured by, for example, a readable / writable nonvolatile memory.

  The vibration waveform data storage unit 132 stores the effective value of the vibration waveform data of the bearing 60 calculated by the effective value calculation unit 120 every moment. The vibration waveform data storage unit 132 is configured to store an effective value of vibration waveform data of the bearing 60 within a predetermined time. As will be described later, the effective value of the vibration waveform data of the bearing 60 stored in the vibration waveform data storage unit 132 is read, and the abnormality of the bearing 60 is diagnosed using the read effective value. In the following description, the effective value of the vibration waveform data is simply referred to as “vibration waveform data”.

  When the evaluation waveform calculation unit 160 reads out the vibration waveform data of the bearing 60 within a certain time from the vibration waveform data storage unit 132, the evaluation value calculation unit 160 calculates an evaluation value that characterizes the read vibration waveform data of the bearing 60 within the certain time. The evaluation value calculation unit 160 is configured to calculate evaluation values continuously in time.

  The evaluation value trend storage unit 134 receives the evaluation value calculated by the evaluation value calculation unit 160. The evaluation value trend storage unit 134 stores the evaluation value given from the evaluation value calculation unit 160 every moment. In other words, the evaluation value trend storage unit 134 is configured to store an evaluation value trend indicating a tendency of the evaluation value to change over time.

  FIG. 3 is a diagram for explaining operations of the vibration waveform data storage unit 132 and the evaluation value calculation unit 160 shown in FIG.

  Referring to FIG. 3, vibration waveform data storage unit 132 receives vibration waveform data of bearing 60 from effective value calculation unit 120 at predetermined time intervals. In the example of FIG. 3, the predetermined time interval is 1 second. D0 to D11 in FIG. 3 represent vibration waveform data given to the vibration waveform data storage unit 132 at intervals of one second.

  The vibration waveform data storage unit 132 stores vibration waveform data of the bearing 60 within a predetermined time. The fixed time can be set according to the rotational speed of the main shaft 20. In the example of FIG. 3, the fixed time is 10 seconds. For example, at time t0, the vibration waveform data storage unit 132 stores a total of ten vibration waveform data D0 to D9 (that is, for 10 seconds) continuous in time.

  When the vibration waveform data storage unit 132 receives the vibration waveform data D10 from the effective value calculation unit 120 at time t1 when 1 second has elapsed from time t0, the oldest vibration waveform among the vibration waveform data D0 to D9 for 10 seconds. While deleting the data D0 and adding the newly input vibration waveform data D10, the vibration waveform data within a predetermined time is updated.

  Furthermore, at time t2 when 1 second has elapsed from time t1, the vibration waveform data storage unit 132 erases the oldest vibration waveform data D1 from among the vibration waveform data D1 to D10 for 10 seconds and is newly input. By adding the vibration waveform data D11, the vibration waveform data of the bearing 60 within a predetermined time is updated.

  In this way, the vibration waveform data storage unit 132 updates the vibration waveform data of the bearing 60 within a predetermined time at predetermined time intervals. The evaluation value calculation unit 160 reads vibration waveform data of the bearing 60 from the vibration waveform data storage unit 132 within a fixed time that is updated at predetermined time intervals. The evaluation value calculation unit 160 calculates an evaluation value by statistically processing the vibration waveform data of the bearing 60 within the read fixed time.

  In the example of FIG. 3, when the evaluation value calculation unit 160 reads vibration waveform data D0 to D9 for 10 seconds from the vibration waveform data storage unit 132 at time t0, statistical processing is performed on the read vibration waveform data D0 to D9. As a result, the evaluation value E0 is calculated. The evaluation value E0 is a value (representative value) that characterizes the vibration waveform data D0 to D9 for 10 seconds immediately before the time t0. Therefore, in the statistical process, for example, an average value of the vibration waveform data D0 to D9 can be calculated as the evaluation value E0. Alternatively, the median value, mode value, minimum value, etc. of the vibration waveform data D0 to D9 can be calculated as the evaluation value E0.

  At time t1, the evaluation value calculation unit 160 calculates the evaluation value E1 by statistically processing the vibration waveform data D1 to D10 for 10 seconds. At time t2, the evaluation value calculation unit 160 calculates the evaluation value E2 by statistically processing the vibration waveform data D2 to D11 for 10 seconds.

  In this way, the evaluation value calculation unit 160 calculates the evaluation value of the vibration waveform data of the bearing 60 within a predetermined time at predetermined time intervals. The evaluation value calculation unit 160 outputs the calculated evaluation value to the evaluation value trend storage unit 134.

FIG. 4 is a diagram for explaining the operation of the evaluation value trend storage unit 134 shown in FIG.
Referring to FIG. 4, evaluation value trend storage unit 134 receives evaluation values from evaluation value calculation unit 160 at predetermined time intervals. In the example of FIG. 4, the predetermined time interval is 1 second. E0 to E6 in FIG. 4 represent evaluation values given from the evaluation value calculation unit 160 to the evaluation value trend storage unit 134 at intervals of 1 second.

  The evaluation value trend storage unit 134 stores a predetermined number of evaluation values that are temporally continuous. The predetermined number of evaluation values that are continuous in time corresponds to an “evaluation value trend” that represents a tendency of the evaluation value to change over time.

  In the example of FIG. 4, the predetermined number is 5. For example, the evaluation value trend storage unit 134 stores the evaluation value E0 at time t0, stores the evaluation value E1 at time t1 when 1 second has elapsed from time t0, and at time t2 when 1 second has elapsed from time t1. The evaluation value E2 is stored, the evaluation value E3 is stored at time t3 when 1 second has elapsed from time t2, and the evaluation value E4 is stored at time t4 when 1 second has elapsed from time t3. The evaluation value E3 is calculated by statistically processing vibration waveform data D3 to D12 for 10 seconds immediately before time t3. The evaluation value E4 is calculated by statistically processing vibration waveform data D4 to D13 for 10 seconds immediately before time t4.

  In the example of FIG. 4, the evaluation value trend storage unit 134 stores a total of five evaluation values E0 to E4, for example, at time t5. When the evaluation value trend storage unit 134 receives the evaluation values E5 of the vibration waveform data D5 to D14 for 10 seconds immediately before the time t5 from the evaluation value calculation unit 160 at the time t6 when 1 second has elapsed from the time t5, a total of 5 is obtained. The oldest evaluation value E0 among the evaluation values E0 to E4 is deleted, and a newly input evaluation value E5 is added to update a predetermined number of evaluation values.

  Furthermore, at the time t6 when 1 second has elapsed from the time t5, the evaluation value trend storage unit 134 receives the evaluation values E6 of the vibration waveform data D6 to D15 for 10 seconds immediately before the time t6 from the evaluation value calculation unit 160. The oldest evaluation value E1 out of the total of five evaluation values E1 to E5 is deleted, and a newly input evaluation value E6 is added to update a predetermined number of evaluation values.

  In this way, the evaluation value trend storage unit 134 updates a predetermined number of evaluation values (evaluation value trends) that are continuous in time at predetermined time intervals.

  Returning to FIG. 2, when the measurement trigger generation unit 170 reads a predetermined number of evaluation values (evaluation value trend) that are temporally continuous from the evaluation value trend storage unit 134, the vibration waveform is generated based on the read evaluation value trend. A trigger for starting data measurement (hereinafter also referred to as “measurement trigger”) is generated. The measurement trigger generation unit 170 outputs the generated measurement trigger to the vibration waveform data output unit 140.

  FIG. 5 is a diagram for explaining operations of the measurement trigger generator 170 and the vibration waveform data output unit 140 shown in FIG. FIG. 5 shows the evaluation value trend stored in the evaluation value trend storage unit 134, the measurement trigger generated from the measurement trigger generation unit 170 based on the evaluation value trend, and the bearing 60 output from the vibration waveform data output unit 140. An example of vibration waveform data is shown.

  Ei-4, Ei-3,... Ei, Ei + 1 in FIG. 5 represent evaluation values given to the evaluation value trend storage unit 134 at predetermined time intervals. In the example of FIG. 5, the predetermined time interval is 1 second. Ei represents an evaluation value given to the evaluation value trend storage unit 134 at time ti, Ei-1 represents an evaluation value given to the evaluation value trend storage unit 134 at time ti-1, and Ei + 1 represents an evaluation value trend at time ti + 1. An evaluation value given to the storage unit 134 is shown.

  The measurement trigger generation unit 170 determines whether or not the evaluation value trend representing the tendency of the evaluation value over time has changed. If it is determined that the evaluation value trend has changed, the measurement trigger generator 170 generates a measurement trigger. Specifically, the measurement trigger generation unit 170 determines whether or not the evaluation value trend has changed based on the temporal change rate of the evaluation value, that is, the amount of change within the unit time.

  In the example of FIG. 5, when the evaluation value Ei is stored in the evaluation value trend storage unit 134 at time ti, the measurement trigger generation unit 170 calculates the evaluation value Ei at time ti and the evaluation value Ei−1 at time ti + 1. Based on the difference, the temporal change rate of the evaluation value is calculated. Assuming that the time change rate of the evaluation value at time ti is dEi, dEi is expressed by Expression (1).

dEi = (Ei−Ei + 1) / (ti−ti + 1) (1)
The measurement trigger generator 170 compares the calculated temporal change rate dEi with a predetermined threshold value α. When the temporal change rate dEi is equal to or greater than the threshold value α, the measurement trigger generation unit 170 sets the measurement trigger to ON. On the other hand, when the temporal change rate dEi is smaller than the threshold value α, the measurement trigger generation unit 170 sets the measurement trigger to OFF. FIG. 5 shows a case where the temporal change rate dEi + 1 at time ti + 1 is greater than or equal to the threshold value α. In this case, at time ti + 1, measurement trigger generation unit 170 switches the measurement trigger from off to on. The measurement trigger generator 170 outputs the measurement trigger to the vibration waveform data output unit 140.

  When receiving the measurement trigger from the measurement trigger generating unit 170, the vibration waveform data output unit 140 reads the vibration waveform data of the bearing 60 within a certain time stored in the vibration waveform data storage unit 132. The vibration waveform data of the bearing 60 within the certain time corresponds to the vibration waveform data of the bearing 60 within the certain time immediately before the time when the measurement trigger is generated. In the example of FIG. 5, vibration waveform data Di-9 to Di for 10 seconds immediately before time ti + 1 are read from the vibration waveform data storage unit 132.

  The vibration waveform data output unit 140 further receives vibration waveform data of the bearing 60 after the time ti + 1 from the effective value calculation unit 120. In the example of FIG. 5, the vibration waveform data output unit 140 receives vibration waveform data Di + 1 to Di + 10 for 10 seconds after time ti + 1 from the effective value calculation unit 120.

  The vibration waveform data output unit 140 includes vibration waveform data Di-9 to Di of the bearing 60 within a certain time immediately before time ti + 1, which is the time when the measurement trigger is generated, and time ti + 1 and later, which is the time when the measurement trigger is generated. The vibration waveform data Di + 1 to Di + 10 of the bearing 60 are collectively output to the diagnosis unit 150.

  Upon receiving the combined vibration waveform data Di-9 to Di + 10 of the bearing 60, the diagnosis unit 150 diagnoses the abnormality of the bearing 60 based on the received vibration waveform data Di-9 to Di + 10. That is, the diagnosis unit 150 is configured to start measurement of vibration waveform data when a measurement trigger is generated due to a change in the tendency of the evaluation value to change over time.

  As described above, according to the state monitoring system according to the first embodiment, the evaluation value characterizing the vibration waveform data within a predetermined time is calculated, and the evaluation value temporally indicating the tendency of the evaluation value to change over time is calculated. When the change rate changes, a trigger for starting measurement of vibration waveform data is generated. In this way, the trigger can be generated based on the evaluation value in which the influence of noise superimposed on the vibration waveform data is appropriately eliminated, thus preventing frequent triggers due to the influence of noise. can do. As a result, since vibration waveform data when a failure occurs in the bearing 60 can be reliably and efficiently measured, an accurate abnormality diagnosis can be realized.

  Here, the combined vibration waveform data Di-9 to Di + 10 of the bearing 60 given to the diagnosis unit 150 is before and after the time point when the measurement trigger occurs, that is, the time point when the tendency of the evaluation value changes. This corresponds to the vibration waveform data of the bearing 60 acquired over the entire time. Therefore, the diagnosis unit 150 can analyze the state of the bearing 60 before and after the time point when the tendency of the evaluation value changes with time by analyzing these vibration waveform data.

  The vibration waveform data storage unit 132 stores the vibration waveform data of the bearing 60 within a certain time immediately before the time when the measurement trigger is generated, while deleting the vibration waveform data of the bearing 60 when the measurement trigger is not generated. Configured to do. According to this, since the vibration waveform data storage unit 132 only needs to have a storage capacity that can store only data useful for the subsequent investigation, the storage capacity of the memory built in the data processing device 80 is large. It is possible to avoid becoming too much.

  In the data processing device 80, the vibration waveform data storage unit 132 and the evaluation value trend storage unit 134 are independent from the storage unit for storing the vibration waveform data output from the effective value calculation unit 120 momentarily. Can be configured. In this way, the vibration waveform data storage unit 132 and the evaluation value trend storage unit 134 can be easily added or removed according to the use and situation of the wind power generator 10.

  Next, with reference to FIG. 6 and FIG. 7, a control process for abnormality diagnosis of the bearing 60 in the state monitoring system according to the first embodiment will be described.

  FIG. 6 is a flowchart illustrating a control process for storing vibration waveform data of the bearing 60 in the state monitoring system according to the first embodiment. The control process shown in FIG. 6 is repeatedly executed by the vibration waveform data storage unit 132 at predetermined time intervals. For example, in the example of FIG. 3, the control process shown in FIG. 6 is repeatedly executed at intervals of 1 second.

  Referring to FIG. 6, vibration waveform data storage unit 132 receives vibration waveform data of bearing 60 from effective value calculation unit 120 in step S01. In step S02, the vibration waveform data storage unit 132 determines whether or not the number of stored vibration waveform data of the bearing 60 is equal to or greater than a predetermined number X. This predetermined number X is equivalent to the number of vibration waveform data of the bearing 60 acquired within a certain time. In the example of FIG. 3, since the predetermined time is 10 seconds, the predetermined number X is set to “10” which is a value obtained by dividing the predetermined time by a predetermined time interval.

  When the number of stored vibration waveform data of the bearing 60 is greater than or equal to the predetermined number X (when YES in S02), the vibration waveform data storage unit 132 proceeds to step S03 and includes the predetermined number X of vibration waveform data. Delete the oldest vibration waveform data. Then, the vibration waveform data storage unit 132 adds the vibration waveform data acquired in step S01 in step S04.

  On the other hand, when the number of stored vibration waveform data of the bearing 60 is smaller than the predetermined number X (NO determination in S02), the vibration waveform data storage unit 132 proceeds to step S04, and the vibration acquired in step S01. Add waveform data. In this way, the vibration waveform data storage unit 132 updates the vibration waveform data within a predetermined time at predetermined time intervals.

  FIG. 7 is a flowchart for explaining a control process for generating a measurement trigger for vibration waveform data of the bearing 60 in the state monitoring system according to the first embodiment. The control process shown in FIG. 7 is repeatedly executed at predetermined time intervals by the evaluation value calculation unit 160, the evaluation value trend storage unit 134, the measurement trigger generation unit 170, and the vibration waveform data output unit 140.

  Referring to FIG. 7, evaluation value calculation unit 160 reads vibration waveform data of bearing 60 within a predetermined time from vibration waveform data storage unit 132 in step S11.

  In step S12, the evaluation value calculation unit 160 performs statistical processing on the vibration waveform data of the bearing 60 within the fixed time read out in step S11, thereby calculating the evaluation value of the vibration waveform data of the bearing 60 within the fixed time. The evaluation value calculation unit 160 outputs the calculated evaluation value to the evaluation value trend storage unit 134.

  The evaluation value trend storage unit 134 determines whether or not the number of stored evaluation value data is greater than or equal to a predetermined number Y in step S13. The predetermined number Y corresponds to the number of data necessary for obtaining an evaluation value trend representing a tendency of the evaluation value to change over time. The predetermined number Y is set to a number of 2 or more. In the example of FIG. 4, the predetermined number Y is set to 5.

  When the number of stored evaluation value data is greater than or equal to the predetermined number Y (when YES is determined in S13), the evaluation value trend storage unit 134 proceeds to step S14, and the oldest evaluation among the predetermined number Y of evaluation values. Erase the value. And the evaluation value trend memory | storage part 134 adds the evaluation value calculated by step S12 by step S15.

  On the other hand, when the number of stored evaluation value data is smaller than the predetermined number Y (NO determination in S13), the evaluation value trend storage unit 134 proceeds to step S15 and adds the evaluation value calculated in step S12. . In this way, the evaluation value trend storage unit 134 updates the predetermined number Y of evaluation values at predetermined time intervals.

  When the measurement trigger generation unit 170 reads out a predetermined number Y of evaluation values (evaluation value trend) that are continuous in time from the evaluation value trend storage unit 134 in step S16, the temporal change in the evaluation value in the read evaluation value trend. Calculate the rate. In step S <b> 16, the measurement trigger generation unit 170 reads the evaluation value added in step S <b> 15 and the evaluation value added immediately before this evaluation value from the evaluation value trend storage unit 134. Then, the measurement trigger generation unit 170 calculates the temporal change rate of the evaluation value based on the difference between the two evaluation values that have been read using the above equation (1).

  In step S17, the measurement trigger generation unit 170 compares the evaluation value temporal change rate calculated in step S16 with the threshold value α. When the temporal change rate of the evaluation value is smaller than the threshold value α (NO determination in S17), the subsequent processes S18 to S21 are skipped. On the other hand, when the temporal change rate of the evaluation value is greater than or equal to the threshold value α (when YES is determined in S17), the measurement trigger generation unit 170 generates a measurement trigger in step S18, and outputs the generated measurement trigger as vibration waveform data. Output to the unit 140.

  When the vibration waveform data output unit 140 receives the measurement trigger from the measurement trigger generation unit 170, the vibration waveform data output unit 140 is stored in the vibration waveform data storage unit 132 momentarily in step S19. Read waveform data.

  Further, in step S20, the vibration waveform data output unit 140 reads the vibration waveform data of the bearing 60 stored in the vibration waveform data storage unit 132 within a predetermined time. As described with reference to FIG. 5, the vibration waveform data of the bearing 60 within the certain time corresponds to the vibration waveform data of the bearing 60 within the certain time immediately before the time when the measurement trigger is generated.

  The vibration waveform data output unit 140 reads the vibration waveform data of the bearing 60 within a certain period of time immediately before the time when the measurement trigger is generated, which is read in Step S20 in Step S21, and the measurement trigger acquired in Step S19. The vibration waveform data of the bearing 60 after the point of occurrence of the failure is collectively output to the diagnosis unit 150. As a result, the diagnosis unit 150 diagnoses the abnormality of the bearing 60 using the vibration waveform data of the bearings 60 collected together.

(Modification of Embodiment 1)
In Embodiment 1 described above, a temporal change rate (for example, dEi in FIG. 5) between two temporally continuous evaluation values (for example, evaluation values Ei−1 and Ei in FIG. 5) is calculated, The configuration for generating the measurement trigger has been described based on the result of comparing the calculated temporal change rate and the threshold value α. However, in this configuration, when noise is superimposed on one of the two evaluation values, the temporal change rate of the evaluation value may temporarily exceed the threshold value α due to the influence of the noise. In such a case, there is a possibility that the measurement trigger generator 170 erroneously generates a measurement trigger.

  Therefore, in order to prevent the measurement trigger from being erroneously generated due to the influence of such noise, the measurement trigger generation unit 170 continuously receives a determination result that the temporal change rate of the evaluation value is greater than or equal to the threshold value α. If obtained, it can be configured to generate a measurement trigger. For example, in the example of FIG. 5, the measurement trigger generation unit 170 generates a measurement trigger when it is determined that the temporal change rate dEi + 1 at time ti + 1 and the temporal change rate dEi + 2 at time ti + 2 are both greater than or equal to the threshold value α. It can be set as the structure to do.

  In this way, when the temporal change rate dEi + 1 at time ti + 1 is equal to or greater than the threshold value α, and when the temporal change rate dEi + 2 at time ti + 2 is smaller than the threshold value α, the measurement trigger generation unit 170 sets the measurement trigger. Does not occur. Since the increase in the temporal change rate dEi + 1 is considered to be temporary due to the influence of noise, it is possible to prevent the measurement trigger from being erroneously generated.

[Embodiment 2]
In the first embodiment described above, a measurement trigger is generated by determining that the tendency of the temporal change in the evaluation value of the vibration waveform data of the bearing 60 within a certain time has changed based on the temporal change rate of the evaluation value. The configuration to be described has been described. In the second embodiment, a configuration will be described in which a measurement trigger is generated by determining that a tendency of a temporal change in an evaluation value has changed, based on the magnitude of the evaluation value.

  FIG. 8 is a flowchart for explaining a control process for generating a measurement trigger for vibration waveform data of the bearing 60 in the state monitoring system according to the second embodiment. The control process shown in FIG. 8 is repeatedly executed at predetermined time intervals by the evaluation value calculation unit 160, the evaluation value trend storage unit 134, the measurement trigger generation unit 170, and the vibration waveform data output unit 140.

  8 is compared with FIG. 7, in the state monitoring system according to the second embodiment, step S17A is executed instead of steps S16 and S17 after the processing of steps S11 to S15 similar to FIG.

  That is, when the evaluation value trend storage unit 134 adds the evaluation value calculated in step S12 in step S15, the measurement trigger generation unit 170 compares the evaluation value added in step S15 with the threshold value β in step S17A. . When the evaluation value is smaller than the threshold β (when NO is determined in S17A), the subsequent processes S18 to S21 are skipped. On the other hand, when the evaluation value is equal to or greater than the threshold value β (when YES is determined in S17A), the measurement trigger generation unit 170 generates a measurement trigger by the process of step S18 similar to FIG.

  The vibration waveform data output unit 140 performs the processing of steps S19 to S21 similar to FIG. 7, so that the vibration waveform data of the bearing 60 and the measurement trigger are generated within a certain time immediately before the time when the measurement trigger is generated. The vibration waveform data of the bearing 60 after the time is collectively output to the diagnosis unit 150.

  As described above, according to the second embodiment, the evaluation value characterizing the vibration waveform data within a predetermined time is calculated, and when the magnitude of the evaluation value becomes equal to or larger than the threshold value β, the temporal change of the evaluation value is calculated. It is determined that the tendency has changed, and a trigger for starting measurement of vibration waveform data is generated. In this way, the trigger can be generated based on the evaluation value from which the influence of noise superimposed on the vibration waveform data is appropriately eliminated. Therefore, also in the second embodiment, the same effect as in the first embodiment can be obtained.

(Modification of Embodiment 2)
In the second embodiment, the configuration for generating the measurement trigger based on the result of comparing the evaluation value and the threshold value β has been described. However, in this configuration, when the evaluation value temporarily exceeds the threshold value β due to noise superimposed on the evaluation value, the measurement trigger generation unit 170 may erroneously generate the measurement trigger.

  In order to prevent the measurement trigger from being erroneously generated due to the influence of such noise, the measurement trigger generation unit 170 performs measurement when the determination result that the evaluation value is equal to or greater than the threshold value β is continuously obtained a plurality of times. It can be configured to generate a trigger. For example, in the example of FIG. 5, the measurement trigger generator 170 generates a measurement trigger when it is determined that the evaluation value Ei + 1 at time ti + 1 and the evaluation value Ei + 2 at time ti + 2 are both equal to or greater than the threshold value β. be able to.

  In this way, when the evaluation value Ei + 1 at time ti + 1 is equal to or greater than the threshold value β, but the evaluation value Ei + 2 at time ti + 2 is smaller than the threshold value β, the measurement trigger generation unit 170 does not generate a measurement trigger. Since the increase in the evaluation value Ei + 1 is considered to be temporary due to the influence of noise, it is possible to prevent the measurement trigger from being erroneously generated.

[Embodiment 3]
In the third embodiment, a configuration will be described in which a change in the tendency of the evaluation value over time is determined based on the time change rate and magnitude of the evaluation value and a measurement trigger is generated.

  FIG. 9 is a flowchart for explaining a control process for generating a measurement trigger for vibration waveform data of the bearing 60 in the state monitoring system according to the third embodiment. The control process shown in FIG. 9 is repeatedly executed at predetermined time intervals by the evaluation value calculation unit 160, the evaluation value trend storage unit 134, the measurement trigger generation unit 170, and the vibration waveform data output unit 140.

  9 is compared with FIG. 7, in the state monitoring system according to the third embodiment, step S <b> 17 </ b> A is executed in addition to step S <b> 17 after the processing of steps S <b> 11 to S <b> 16 similar to FIG. 7.

  That is, when the evaluation value trend storage unit 134 adds the evaluation value calculated in step S12 in step S15, the measurement trigger generation unit 170 performs the temporal change rate and threshold value α of the evaluation value calculated in step S16 in step S17. (First threshold) is compared. When the temporal change rate of the evaluation value is smaller than the threshold value α (NO determination in S17), the subsequent processes S17A to S21 are skipped. On the other hand, when the temporal change rate of the evaluation value is equal to or greater than the threshold value α (when YES is determined in S17), the measurement trigger generation unit 170 proceeds to step S17A, and the evaluation value added in step S15 and the threshold value β (second The threshold). When the evaluation value is smaller than the threshold β (when NO is determined in S17A), the subsequent processes S18 to S21 are skipped.

  On the other hand, when the evaluation value is equal to or greater than the threshold value β (when YES is determined in S17A), the measurement trigger generation unit 170 generates a measurement trigger by the process of step S18 similar to FIG.

  The vibration waveform data output unit 140 performs the processing of steps S19 to S21 similar to FIG. 7, so that the vibration waveform data of the bearing 60 and the measurement trigger are generated within a certain time immediately before the time when the measurement trigger is generated. The vibration waveform data of the bearing 60 after the time is collectively output to the diagnosis unit 150.

  As described above, according to the third embodiment, the evaluation value characterizing the vibration waveform data within a predetermined time is calculated, the temporal change rate of the evaluation value is equal to or greater than the threshold value α, and the magnitude of the evaluation value is When it becomes more than threshold value (beta), it determines with the tendency of the time change of an evaluation value having changed, and the trigger which starts the measurement of vibration waveform data is generated. Even when the temporal change rate of the evaluation value is equal to or greater than the threshold value α, when the evaluation value is smaller than the threshold value β, the degree of vibration, which is the magnitude of vibration of the bearing 60, is small, so It can be determined that the degree of influence has increased. In such a case, by adopting a configuration in which the trigger is not generated, the trigger can be generated based on the evaluation value from which the influence of noise superimposed on the vibration waveform data is appropriately eliminated. Therefore, also in the third embodiment, the same effect as in the first embodiment can be obtained.

(Modification of Embodiment 3)
As described above, when noise is superimposed on the evaluation value, the temporal change rate of the evaluation value temporarily becomes greater than or equal to the threshold value α, or the evaluation value temporarily becomes greater than or equal to the threshold value β. 170 may generate a measurement trigger in error.

  Therefore, in the measurement trigger generation unit 170, the determination result that the temporal change rate of the evaluation value is equal to or greater than the threshold value α is continuously obtained a plurality of times, and the determination result that the evaluation value is equal to or greater than the threshold value β is obtained multiple times. It can be configured to generate a measurement trigger when obtained continuously. According to this, since the increase in the evaluation value is considered to be temporary due to the influence of noise, it is possible to prevent the measurement trigger from being erroneously generated.

[Embodiment 4]
In the fourth embodiment, a configuration for calculating the degree of abnormality of vibration waveform data within a certain time as an evaluation value characterizing vibration waveform data within a certain time will be described.

  FIG. 10 is a functional block diagram functionally showing the configuration of the data processing device 80 in the state monitoring system according to Embodiment 4 of the present invention. Referring to FIG. 10, the data processing device 80 includes a BPF 110, an effective value calculation unit 120, a storage unit 130, a vibration waveform data output unit 140, an evaluation value calculation unit 160, a measurement trigger generation unit 170, And a diagnosis unit 150.

  Storage unit 130 includes vibration waveform data storage unit 132, evaluation value trend storage unit 134, and normal data storage unit 136. The vibration waveform data storage unit 132, the evaluation value trend storage unit 134, and the normal data storage unit 136 are configured by, for example, a readable / writable nonvolatile memory.

  As described with reference to FIG. 3, the vibration waveform data storage unit 132 is configured to store an effective value (vibration waveform data) of the vibration waveform data of the bearing 60 within a predetermined time.

  The normal data storage unit 136 is an effective value (vibration waveform data) of the vibration waveform data of the bearing 60 measured when normal operation of the wind turbine generator 10 (see FIG. 1) is guaranteed (for example, in an initial state). Configured to store. The vibration waveform data stored in the normal data storage unit 136 is used for setting a classification boundary in the learning unit 162 described later. In the following description, the vibration waveform data stored in the normal data storage unit 136 is also referred to as “learning data”.

  When the evaluation waveform calculation unit 160 reads out the vibration waveform data of the bearing 60 within a certain time from the vibration waveform data storage unit 132, the evaluation value calculation unit 160 calculates an evaluation value that characterizes the read vibration waveform data of the bearing 60 within the certain time. In the present embodiment, evaluation value calculation unit 160 includes a learning unit 162 and an abnormality degree calculation unit 164.

  When the learning unit 162 reads the learning data from the normal data storage unit 136, the learning unit 162 sets a classification boundary for classifying normal and abnormal based on the read learning data.

  The abnormality degree calculation unit 164 calculates the abnormality degree of the vibration waveform data by applying the classification boundary to the vibration waveform data of the bearing 60 within a certain time read from the vibration waveform data storage unit 132. The degree of abnormality corresponds to the distance from the classification boundary. The calculated degree of abnormality is an evaluation value that characterizes the vibration waveform data of the bearing 60 within the predetermined time. The abnormality degree calculation unit 164 is configured to calculate the abnormality degree continuously in time.

  In the present embodiment, when processing the vibration waveform data of the bearing 60 within a predetermined time, the evaluation value calculation unit 160 divides the vibration waveform data into a plurality of segments and processes each segment. Hereinafter, the segment will be described.

  FIG. 11 is a conceptual diagram showing the relationship between the vibration waveform data of the bearing 60 and the segments within a certain time. In the example of FIG. 11, the vibration waveform data within a predetermined time is vibration waveform data D0 to D9 for 10 seconds, as in the example of FIG. The vibration waveform data D0 to D9 for 10 seconds are divided into 10 segments.

  The degree-of-abnormality calculation unit 164 generates a feature vector for each segment for the vibration waveform data D0 to D9 for 10 seconds. Similarly, the learning unit 162 generates a feature vector for each segment of the learning data read from the normal data storage unit 136.

  FIG. 12 is a diagram for explaining the feature amount vector. FIG. 12 shows an example in which vibration waveform data D0 to D9 for 10 seconds are divided into 10 segments and the feature quantity is n.

  The feature amount is, for example, an effective value (OA), maximum value (Max), peak value (Crest factor), kurtosis, skewness, and values after signal processing (FFT processing, quefrency processing) of vibration waveform data. It can be. The feature vector handles n feature values as a set of vectors. The feature vector is used for abnormality determination. In the example of FIG. 12, ten feature quantity vectors are generated for vibration waveform data D0 to D9 for 10 seconds.

  If extraction of feature amounts and generation of feature amount vectors are performed collectively for the entire vibration waveform data within a certain period of time, the entire data may not be usable for diagnosis when a sudden abnormality occurs. For this reason, in the present embodiment, vibration waveform data within a predetermined time is divided into a plurality of segments, and feature amounts are extracted and feature amount vectors are generated in units of segment data. For example, when the wind power generation apparatus 10 is monitored by the vibration sensor 70, sudden vibration may be temporarily detected by the vibration sensor 70. Even in such a case, a correct feature amount can be extracted at a time other than the sudden abnormality. Therefore, the segment data corresponding to the sudden abnormality can be excluded, and the feature amount can be compared and evaluated for each segment data.

  As illustrated in FIG. 12, the degree of abnormality 0 to 9 is calculated for the feature amount vectors 0 to 9 based on the classification boundary. The classification boundary is an index for performing abnormality determination used in a known abnormality detection method (One Class Support Vector Machine: OC-SVM).

  FIG. 13 is a diagram for explaining the basic concept of OC-SVM. The vertical axis and horizontal axis in FIG. 13 are different feature quantities. In FIG. 13, circles “◯” are learning data, and squares “□” and triangles “Δ” are vibration waveform data. In the vibration waveform data, “□” is data indicating abnormality, and “Δ” is data indicating normal.

  For example, as shown in FIG. 13A, on the two-dimensional scatter diagram in the case where there are two feature quantities, a case where a boundary line that can classify normal / abnormal cannot be drawn in the learning data and the vibration waveform data. Think.

  On the other hand, useful feature amounts differ depending on the diagnosis target and operating conditions. Therefore, an appropriate feature amount is selected based on the diagnosis target and the operating conditions. By mapping each learning data and vibration waveform data to a multidimensional feature space including appropriate feature amounts, a classification boundary surface that can classify normal / abnormal can be generated as shown in FIG. 13B. .

  For each learning data and vibration waveform data, the degree of abnormality, which is the distance from the classification boundary, can be calculated. The degree of abnormality is zero on the classification boundary, the degree of abnormality is negative (−) on the normal side of the classification boundary, and the degree of abnormality is positive (+) on the abnormal side of the classification boundary.

  Such a method is called machine learning by OC-SVM, and can be evaluated by converting many feature values into one index (abnormality).

  Returning to FIG. 10, the learning unit 162 sets the classification boundary described above using the learning data. The degree of abnormality calculation unit 164 calculates the degree of abnormality 0 to 9, which is the distance from the classification boundary in the feature section, for each of the feature vectors 0 to 9.

  The abnormality degree calculation unit 164 calculates the evaluation value E0 by performing statistical processing on the calculated abnormality degrees 0 to 9. The calculation of the evaluation value corresponds to the process of step S12 in the control process shown in FIGS. The evaluation value E0 is a value (representative value) that characterizes the abnormalities 0 to 9 of vibration waveform data for 10 seconds. Therefore, in the statistical process, an average value of the degree of abnormality 0 to 9 can be calculated as the evaluation value E0. Alternatively, as the evaluation value E0, a median value, a mode value, a minimum value, etc. of the degree of abnormality 0 to 9 can be calculated.

  Alternatively, in step S12 of FIGS. 7, 8, and 9, the abnormality rate may be calculated by comparing the abnormality determination threshold value set based on the learning data with each abnormality degree as the evaluation value E0. it can. The abnormality rate is obtained by dividing the number of abnormality degrees 0 to 9 in which the abnormality degree exceeds a predetermined abnormality determination threshold by the total number of segments (10).

  In this way, the evaluation value calculation unit 160 calculates the evaluation value (abnormality) of the vibration waveform data of the bearing 60 within a predetermined time at predetermined time intervals. The evaluation value calculation unit 160 outputs the calculated evaluation value to the evaluation value trend storage unit 134.

  As described with reference to FIG. 4, the evaluation value trend storage unit 134 stores the evaluation value (abnormality) given from the evaluation value calculation unit 160 every moment. In the present embodiment, a predetermined number of abnormalities that are temporally continuous correspond to an “evaluation value trend” that represents a tendency of temporal change in the abnormal degree. The evaluation value trend storage unit 134 updates the evaluation value trend at predetermined time intervals.

  Returning to FIG. 10, when the measurement trigger generation unit 170 reads a predetermined number of abnormality degrees (evaluation values) that are temporally continuous from the evaluation value trend storage unit 134, the measurement trigger generation unit 170 generates a measurement trigger based on the read evaluation value trend. To do. As described with reference to FIG. 5, the measurement trigger generator 170 generates a measurement trigger when it is determined that the evaluation value trend has changed. The generated measurement trigger is given to the vibration waveform data output unit 140.

  As described in FIG. 5, when the vibration waveform data output unit 140 receives the measurement trigger from the measurement trigger generation unit 170, the vibration waveform data output unit 140 is stored in the vibration waveform data storage unit 132 and is constant immediately before the time when the measurement trigger is generated. The vibration waveform data of the bearing 60 within the time is read. The vibration waveform data output unit 140 further receives vibration waveform data of the bearing 60 from the effective value calculation unit 120 after the time when the measurement trigger is generated. The vibration waveform data output unit 140 collectively outputs these vibration waveform data to the diagnosis unit 150.

  The diagnosis unit 150 diagnoses an abnormality in the bearing 60 based on the vibration waveform data of the bearing 60 that is collected.

  As described above, according to the state monitoring system according to the fourth embodiment, one index generated from a plurality of feature values extracted from the vibration waveform data is used as an evaluation value characterizing the vibration waveform data within a predetermined time. By using (abnormality), even when a plurality of feature quantities are individually changing, when changes occur in a different combination, the changes can be captured. According to this, it is possible to record vibration waveform data even in a state where it is not possible to clearly define what kind of change is used as a measurement trigger.

  Moreover, even if the vibration waveform data at the time of normality has variations, all of the machine learning by OC-SVM is recognized as a normal state, so that a measurement trigger can be prevented from being erroneously generated. . And, since the measurement trigger is generated when the trend of temporal change in the degree of abnormality is different from the normal time, a huge amount of data is accumulated and the vibration waveform data when the abnormality occurs is reliably and efficiently Therefore, accurate abnormality diagnosis can be realized.

  In each of the above-described embodiments, the vibration sensor 70 is installed in the bearing 60 that is one of the mechanical elements constituting the wind power generator 10, and the abnormality of the bearing 60 is diagnosed. It will be described in a definite manner that the mechanical element is not limited to the bearing 60. For example, a vibration sensor is installed in a bearing provided in the speed increaser 40 or in the generator 50 together with or in place of the bearing 60, and the speed increaser 40 is obtained by the same method as in each of the above embodiments. An abnormality of a bearing provided in the generator 50 or in the generator 50 can be diagnosed.

  Further, in each of the above embodiments, the configuration for calculating the evaluation value characterizing the vibration waveform data within a certain time by calculating the effective value of the vibration waveform data within the certain time has been described. The evaluation value may be calculated by statistically processing the peak value of the vibration waveform data within a certain time. In this configuration, the peak value of the vibration waveform data corresponds to the absolute value of the maximum value or the minimum value of the vibration waveform. Alternatively, the evaluation value may be calculated by statistically processing the crest factor of the vibration waveform data within the predetermined time. In this configuration, the crest factor of the vibration waveform data corresponds to the ratio of the effective value to the maximum value of the vibration waveform.

  In each of the above-described embodiments, one evaluation value characterizing vibration waveform data within a predetermined time is set as one, and measurement of vibration waveform data is triggered by a change in the tendency of temporal change of the one evaluation value. However, a configuration in which a change in the tendency of a plurality of evaluation values over time is used as a trigger may be used.

  In each of the above-described embodiments, the data processing device 80 corresponds to an example of the “processing device” in the present invention, and the storage unit 130, the evaluation value calculation unit 160, and the diagnosis unit 150 correspond to “ This corresponds to an embodiment of “storage unit”, “evaluation value calculation unit”, and “diagnosis unit”.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is shown not by the above description of the embodiments but by the scope of claims for patent, and is intended to include meanings equivalent to the scope of claims for patent and all modifications within the scope.

  DESCRIPTION OF SYMBOLS 10 Wind power generator, 20 Main shaft, 30 Blade, 40 Booster, 50 Generator, 60 Bearing, 70 Vibration sensor, 80 Data processing device, 90 Nacelle, 100 Tower, 110 HPF, 120 RMS value calculation part, 130 Storage part 132 Vibration waveform data storage unit 134 Evaluation value trend storage unit 136 Normal data storage unit 140 Vibration waveform data output unit 150 Diagnosis unit 160 Evaluation value calculation unit 162 Learning unit 164 Abnormality calculation unit 170 Measurement Trigger generator.

Claims (8)

A state monitoring system for monitoring the state of machine elements constituting the apparatus,
A vibration sensor for measuring a vibration waveform of the machine element;
A processing device for diagnosing an abnormality of the machine element,
The processor is
An evaluation value characterizing the vibration waveform data output from the vibration sensor within a predetermined time, an evaluation value calculation unit configured to continuously calculate in time,
By starting measurement of the vibration waveform data triggered by a change in the tendency of temporal change in the evaluation value calculated by the evaluation value calculation unit, the abnormality of the mechanical element is detected using the vibration waveform data. A condition monitoring system, comprising: a diagnosis unit configured to diagnose. The processing device further includes a storage unit for storing the vibration waveform data within the predetermined time,
2. The state monitoring system according to claim 1, wherein the diagnosis unit acquires the vibration waveform data within the predetermined time immediately before the time point when the tendency of the evaluation value changes with time from the storage unit.   The state monitoring system according to claim 1, wherein the diagnosis unit detects that the temporal change rate of the evaluation value is equal to or greater than a threshold value and starts measurement of the vibration waveform data.   The diagnostic unit detects that the temporal change rate of the evaluation value is equal to or greater than a first threshold and the magnitude of the evaluation value is equal to or greater than a second threshold; The state monitoring system according to claim 1 or 2, wherein measurement of vibration waveform data is started.   5. The state monitoring system according to claim 1, wherein the evaluation value calculation unit calculates the evaluation value by statistically processing the vibration waveform data within the predetermined time. The evaluation value calculator is
Dividing the vibration waveform data within the predetermined time into a plurality of segment data, and generating a feature quantity vector including a plurality of feature quantities for each segment data;
Using the feature vector generated for the vibration waveform data when the device is normal, setting a classification boundary for classifying normal and abnormal,
For each of the plurality of feature vectors generated with respect to the vibration waveform data within the predetermined time, an abnormality degree that is a distance from the classification boundary is calculated,
The state monitoring system according to any one of claims 1 to 4, wherein the evaluation value is generated by statistically processing a plurality of the abnormalities.   The storage unit stores the vibration waveform data provided from the vibration sensor at a predetermined time interval, and erases the oldest vibration waveform data among the stored vibration waveform data within the predetermined time. The state monitoring system according to claim 2.   A wind turbine generator comprising the state monitoring system according to any one of claims 1 to 7.