JP7408519B2 - Condition monitoring device and condition monitoring method - Google Patents

Description

The present disclosure relates to a condition monitoring device and a condition monitoring method that monitor the condition of an object.

BACKGROUND ART Conventionally, the state of an object such as a rotating machine is monitored based on a signal from a sensor installed on the object. Specifically, an abnormality in the object is diagnosed by analyzing a temporal waveform indicating a change in signal intensity over time.

For example, Japanese Patent Laid-Open No. 2011-154020 (Patent Document 1) discloses a technique for diagnosing an abnormality in an object based on an envelope waveform generated by performing envelope processing on a vibration waveform measured by a vibration sensor. are doing. For example, low-pass filter processing is used as the envelope processing.

Japanese Unexamined Patent Publication No. 2016-75563 (Patent Document 2) diagnoses an abnormality in an object by analyzing a plurality of composite waveforms obtained by combining vibration waveform data of the object and a plurality of noise data. The technology is disclosed.

Japanese Patent Application Publication No. 2011-154020 JP 2016-75563 Publication Japanese Patent Application Publication No. 2018-179977

It is known that when the rotational speed of a rotating machine is low, the magnitude of vibration of the rotating machine becomes smaller compared to when the rotational speed is high. Therefore, as the rotation speed decreases, the noise component included in the signal output from the sensor becomes relatively large. Furthermore, as the rotational speed decreases, the magnitude of vibrations caused by abnormalities in the rotating machine decreases, and the period of occurrence of vibrations caused by abnormalities increases. As a result, in the time waveform of the measured signal strength, the component caused by the abnormality becomes smaller than the noise component. In other words, the S/N ratio decreases.

In the technique described in Patent Document 1, a component in a specific frequency band caused by an abnormality in a target object is extracted from a time waveform of signal strength by envelope processing such as low-pass filter processing. However, as the rotational speed decreases, the influence of noise existing in the same band as the component caused by the abnormality increases, and the S/N ratio in the band decreases, making it impossible to accurately diagnose the abnormality of the object.

In the technique described in Patent Document 2, in order to improve the S/N ratio, it is necessary to increase the number of times of synthesis processing of vibration waveform data and noise data. Therefore, the calculation cost increases. Computation cost is computation time and memory capacity required for computation.

The present disclosure has been made in order to solve the above-mentioned problems, and the purpose is to reduce measurement accuracy even when the S/N ratio of the signal acquired from the sensor installed on the object is low. It is an object of the present invention to provide a condition monitoring device and a condition monitoring method capable of diagnosing an abnormality in an object at low cost.

The condition monitoring device of the present disclosure monitors the condition of an object. The condition monitoring device includes a correction section and a diagnosis section. The correction unit generates the second signal by multiplying the intensity of the first signal acquired from the sensor installed on the object in the first period by a correction coefficient corresponding to the intensity. The diagnosis unit diagnoses an abnormality in the object based on a time waveform indicating a change over time in the intensity of the second signal. If the expected value of the intensity distribution of the third signal acquired from the sensor in the second period in the past than the first signal or the first period is μ, and the intensity of the first signal is x _raw , then the correction coefficient is The setting is made to satisfy at least one of the condition and the second condition. The first condition is that when μ<x _raw , the correction coefficient increases as (x _raw −μ) increases. The second condition is that when μ>x _raw , the correction coefficient increases as (μ−x _raw ) increases.

A state monitoring method of the present disclosure monitors the state of an object. The condition monitoring method includes the steps of: generating a second signal by multiplying the intensity of the first signal acquired from a sensor installed on the object in a first period by a correction coefficient corresponding to the intensity; diagnosing an abnormality in the object based on a time waveform showing a change in intensity over time. If the expected value of the intensity distribution of the third signal acquired from the sensor in the second period in the past than the first signal or the first period is μ, and the intensity of the first signal is x _raw , then the correction coefficient is The setting is made to satisfy at least one of the condition and the second condition. The first condition is that when μ<x _raw , the correction coefficient increases as (x _raw −μ) increases. The second condition is that when μ>x _raw , the correction coefficient increases as (μ−x _raw ) increases.

According to the present disclosure, even if the S/N ratio of a signal acquired from a sensor installed on the target object is low, an abnormality in the target object can be diagnosed at low measurement cost.

1 is a diagram schematically showing the configuration of a wind power generator to which a state monitoring device according to an embodiment is applied. FIG. 1 is a diagram illustrating an example of a hardware configuration of a state monitoring device according to an embodiment. 1 is a block diagram showing a functional configuration of a state monitoring device according to an embodiment. FIG. 7 is a flowchart showing the flow of processing in the correction function generation unit according to the embodiment. It is a figure showing an example of probability density function PDF (x). It is a figure which shows an example of cumulative distribution function CDF(x). It is a figure which shows an example of the time waveform of a sensor signal when the bearing in a speed increaser is damaged. It is a figure which shows an example of the correction function p (xraw(t)). 9 is a diagram showing a first-order derivative of the correction function p(xraw(t)) shown in FIG. 8. FIG. FIG. 3 is a diagram showing an example of a corrected signal output from a correction section. FIG. 7 is a diagram schematically showing update timing of a correction function of a state monitoring device according to a second modification. 12 is a flowchart showing a process flow of a correction function generation unit according to modification example 3. FIG. 7 is a diagram showing changes in a correction function during a transition period.

Embodiments of the present disclosure will be described below with reference to the drawings. In the following drawings, the same or corresponding parts are given the same reference numerals, and the description thereof will not be repeated. Further, the modified examples described below may be selectively combined as appropriate.

In the following, a speed increaser of a wind power generator will be described as an example of an object whose state is monitored by the state monitoring device. However, the target object is not limited to the speed increaser of a wind power generator, and abnormalities can cause changes in the temporal waveform that indicates changes in signal strength over time such as vibration, sound, acoustic emission (AE), etc. It is fine as long as it is something. For example, the objects include various types of equipment installed in factories, power plants, etc., and railway vehicles.

<Configuration of wind power generator>
FIG. 1 is a diagram schematically showing the configuration of a wind power generator to which a condition monitoring device according to an embodiment is applied. Referring to FIG. 1, the wind power generator 10 includes a main shaft 20, a hub 25, a blade 30, a speed increaser 40, a generator 50, a main shaft bearing 60, a sensor 70, and a condition monitoring device 80. Equipped with. The speed increaser 40, the generator 50, the main shaft bearing 60, the sensor 70, and the condition monitoring device 80 are housed in the nacelle 90. Nacelle 90 is supported by tower 100.

The main shaft 20 enters the nacelle 90, 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 rotational torque generated by the blades 30 receiving the wind force to the input shaft of the speed increaser 40 . The blade 30 is provided on the hub 25 and converts wind power into rotational torque and transmits it to the main shaft 20. The main shaft bearing 60 is fixed within the nacelle 90 and rotatably supports the main shaft 20.

The speed increaser 40 is provided between the main shaft 20 and the generator 50, increases the rotational speed of the main shaft 20, and outputs it to the generator 50. As an example, the speed increaser 40 is configured by a gear speed increase mechanism including a planetary gear, an intermediate shaft, a high speed shaft, and the like. A plurality of bearings are provided in the speed increaser 40 to rotatably support a plurality of shafts. The plurality of bearings are configured, for example, by rolling bearings, and include an outer ring (fixed ring), rolling elements, and an inner ring (rotating ring).

The sensor 70 is fixed to the speed increaser 40 and measures a physical quantity indicating the state of the speed increaser 40 . In the present embodiment, the sensor 70 measures the vibration of the speed increaser 40 and sends a signal (hereinafter referred to as a "sensor signal") having an intensity corresponding to the magnitude of the measured vibration to the condition monitoring device 80. Output. That is, the intensity of the sensor signal output from the sensor 70 indicates the magnitude of vibration of the speed increaser 40. The sensor 70 is, for example, an acceleration pickup using a piezoelectric element.

The generator 50 is connected to the output shaft of the speed increaser 40 and generates electricity using rotational torque received from the speed increaser 40 . The generator 50 is configured by, for example, an induction generator. Note that a bearing that rotatably supports the rotor is also provided within the generator 50.

Condition monitoring device 80 is provided within nacelle 90 and receives sensor signals from sensor 70 . The condition monitoring device 80 monitors the condition of the speed increaser 40 based on sensor signals. The condition monitoring device 80 and the sensor 70 constitute a condition monitoring system that monitors the condition of the speed increaser 40.

In the wind power generator 10, the rotational speed of the main shaft 20 varies depending on the wind force. Therefore, the rotational speed of the main shaft 20 can be 100 min ^-1 or less.

<Hardware configuration of status monitoring device>
FIG. 2 is a diagram illustrating an example of the hardware configuration of the status monitoring device according to the embodiment. The status monitoring device 80 is configured by, for example, a general-purpose computer.

As shown in FIG. 2, the state monitoring device 80 includes a CPU (Central Processing Unit) 801, a RAM (Random Access Memory) 802, a storage 803, and a communication interface 804. These components are connected for data communication via bus 805.

The CPU 801 executes various programs stored in the storage 803. The RAM 802 provides a work area for storing data necessary for the CPU 801 to execute programs. The storage 803 is configured with, for example, an HDD (Hard Disk Drive) or an SSD (Flash Solid State Drive), and stores a monitoring program 806 for monitoring the state of the speed increaser 40.

Communication interface 804 has an input/output port for inputting/outputting various signals. For example, communication interface 804 receives sensor signals from sensor 70 . Furthermore, the communication interface 804 may output various signals generated by executing the monitoring program 806 to an external device. The external device is installed outside the wind power generation device 10. A communication interface 804 communicates with an external device via a wireless or wired line.

<Functional configuration of status monitoring device>
FIG. 3 is a block diagram showing the functional configuration of the condition monitoring device according to the embodiment. As shown in FIG. 3, the condition monitoring device 80 includes a storage processing section 81, a correction function generation section 82, a correction section 83, a diagnosis section 84, and a storage section 87. The storage processing unit 81, the correction function generation unit 82, the correction unit 83, and the diagnosis unit 84 are realized, for example, by the CPU 801 shown in FIG. 2 executing the monitoring program 806. Alternatively, the storage processing section 81, the correction function generation section 82, the correction section 83, and the diagnosis section 84 may be realized by dedicated hardware (electronic circuit). The storage unit 87 is realized, for example, by the RAM 802 and storage 803 shown in FIG.

Every time the storage processing unit 81 acquires a sensor signal from the sensor 70, it writes into the storage unit 87 unit data in which the intensity of the sensor signal is associated with the time at which the sensor signal is acquired. As a result, unit data for each time is accumulated in the storage unit 87. Hereinafter, the intensity of the sensor signal acquired at time t will be x _raw (t).

The correction function generation unit 82 generates a correction function p(x _raw (t)) using the sensor signal intensity x _raw (t) as an explanatory variable and the correction coefficient k(t) as an objective variable. k(t) represents a correction coefficient for the intensity x _raw (t) of the sensor signal acquired at time t. The correction coefficient k(t) satisfies the following first and second conditions, assuming that the expected value of the intensity distribution of the sensor signal is μ. The first condition is that when μ<x _raw (t), the correction coefficient k(t) increases as (x _raw (t)−μ) increases. The second condition is that when μ>x _raw (t), the correction coefficient k(t) increases as (μ−x _raw (t)) increases. Detailed processing of the correction function generation unit 82 will be described later.

The correction unit 83 multiplies the intensity x _raw (t) of the sensor signal by a correction coefficient k(t) corresponding to the intensity x _raw (t) to generate a new signal (hereinafter referred to as a “corrected signal”). ) is generated.

Specifically, the correction unit 83 generates a corrected signal corresponding to a sensor signal acquired during a period to be diagnosed (hereinafter referred to as a "diagnosis period"). The correction unit 83 substitutes the intensity x _raw (t) of the sensor signal acquired during the diagnosis period into the correction function p (x _raw (t)), thereby obtaining a correction coefficient k (t) (=p (x _raw ( t))). The correction unit 83 generates a corrected signal whose intensity is k(t)×x _raw (t).

The correction unit 83 may generate the corrected signal in real time during the diagnosis period. Alternatively, after the diagnostic period ends, the correction unit 83 reads the intensity x _raw (t) corresponding to time t within the diagnostic period from the storage unit 87, and applies a correction coefficient k(t) to the intensity x _raw (t). The corrected signal may be generated by multiplication.

The diagnosis unit 84 diagnoses an abnormality based on a time waveform indicating a change over time in the intensity of the corrected signal during the diagnosis period. The diagnosis unit 84 includes an evaluation value calculation unit 85 that calculates an evaluation value from the time waveform, and a determination unit 86 that uses the evaluation value to determine whether or not there is an abnormality in the speed increaser 40.

<Correction function generation unit>
FIG. 4 is a flowchart showing the flow of processing in the correction function generation unit according to the embodiment.

(Step S1)
First, the correction function generation unit 82 selects the type of probability distribution that represents the probability of the value that the intensity of the sensor signal takes (step S1). Types of probability distributions include normal distribution, uniform distribution, and Laplace distribution. The correction function generation unit 82 selects one type of probability distribution from among the plurality of types of probability distributions.

The correction function generation unit 82 may select one predetermined type of probability distribution from among a plurality of types of probability distributions. For example, if the type of probability distribution of the value taken by the intensity of the sensor signal has already been analyzed, that type of probability distribution is predetermined as a selection target.

The correction function generation unit 82 may select one type of probability distribution from a plurality of types of probability distributions according to user input. In this case, the user may input the type of probability distribution according to the sensor 70.

When a corrected signal corresponding to the sensor signal acquired within the diagnosis period is generated after the end of the diagnosis period, the correction function generation unit 82 generates a shape of the distribution of the intensity of the sensor signal acquired from the sensor 70 during the diagnosis period. One type of probability distribution may be selected that fits . Specifically, the correction function generation unit 82 extracts the intensity corresponding to the time included in the diagnosis period from the storage unit 87, and generates one type of probability distribution that has the highest degree of similarity to the shape of the extracted intensity distribution. Select.

When a corrected signal is generated in real time during the diagnosis period, the correction function generation unit 82 generates one type of signal that fits the shape of the distribution of the intensity of the sensor signal acquired from the sensor 70 in a period past the diagnosis period. A probability distribution may also be selected. Specifically, the correction function generation unit 82 extracts intensities corresponding to times included in the past period from the storage unit 87, and selects one type of probability that has the highest degree of similarity to the shape of the distribution of the extracted intensities. Select a distribution.

Hereinafter, a case where the selected probability distribution is a normal distribution will be explained as an example. The probability density function PDF(x) corresponding to the normal distribution is expressed by the following equation (1). The probability density function PDF(x) is a function representing the probability density that the random variable X becomes x. In equation (1), μ represents the expected value and σ represents the standard deviation.

(Step S2)
Next, the correction function generation unit 82 sets parameter values according to the type of probability distribution selected in step S1, and generates a probability density function PDF(x) (step S2). In the case of the normal distribution shown in equation (1), the expected value μ and the standard deviation σ are set.

The correction function generation unit 82 may set a predetermined value as the value of the parameter. For example, if the shape of the probability distribution of the values taken by the intensity of the sensor signal has already been analyzed, the value of the parameter according to the shape is determined in advance.

The correction function generation unit 82 may set parameter values according to user input. In this case, the user may input the value of the parameter according to the shape of the probability distribution of the values taken by the intensity of the sensor signal.

When a corrected signal corresponding to the sensor signal acquired within the diagnosis period is generated after the end of the diagnosis period, the correction function generation unit 82 generates a shape of the distribution of the intensity of the sensor signal acquired from the sensor 70 during the diagnosis period. You may set the parameter values to fit. Alternatively, when a corrected signal is generated in real time during the diagnosis period, the correction function generation unit 82 generates a signal so as to fit the shape of the distribution of the intensity of the sensor signal acquired from the sensor 70 in a period past the diagnosis period. You may also set the value of the parameter. Specifically, the correction function generation unit 82 extracts the intensity corresponding to the time included in the diagnosis period or the past period from the storage unit 87, and calculates the shape of the distribution of the extracted intensity and the probability density function PDF(x). Set the value of the parameter so that the degree of similarity is the highest. The value of the parameter may be calculated using maximum likelihood estimation, kernel density estimation, etc., for example. However, the method of calculating the parameter value is not limited to this.

In equation (1), when the expected value μ=0 and the variance σ ² =1 are set, the probability density function PDF(x) is expressed by the following equation (2).

In step S1, when a continuous probability distribution as shown in equation (1) is selected, the correction function generation unit 82 may set a range that the probability variable X can take. For example, in equation (2), if the range from the lower limit x_min to the upper limit x_max is set as the possible range of the random variable X, the probability density function PDF(x) is expressed by the following equation (3). be done.

FIG. 5 is a diagram showing an example of the probability density function PDF(x). FIG. 5 shows the probability density function PDF(x) according to the above equation (3). As shown in equation (3) and FIG. 5, when the random variable X takes a value within the range from the lower limit value x_min to the upper limit value x_max, the probability density function PDF(x) is expressed by equation (2), When the random variable X takes a value outside this range, the probability density function PDF(x) is zero.

The lower limit value x_min is set to a value that is smaller than the expected value μ and that makes the probability density sufficiently small in the probability density function PDF(x). The upper limit value x_max is set to a value that is larger than the expected value μ and that makes the probability density sufficiently small in the probability density function PDF(x).

(Step S3)
Next, as shown in FIG. 4, the correction function generation unit 82 calculates a cumulative distribution function CDF(x) from the probability density function PDF(x) generated in step S2 (step S3). The cumulative distribution function CDF(x) is a function of the probability that the random variable X is less than or equal to x.

For example, the correction function generation unit 82 calculates the cumulative distribution function CDF(x) expressed by the following equation (4) from the probability density function PDF(x) according to the above equation (3).

FIG. 6 is a diagram showing an example of the cumulative distribution function CDF(x). FIG. 6 shows the cumulative distribution function CDF(x) according to the above equation (4).

Note that depending on the type of probability distribution, the cumulative distribution function CDF(x) may become nonlinear when the random variable X is near the lower limit value x_min or near the upper limit value x_max. However, by setting the upper limit value x_max and the lower limit value x_min to values at which the probability density is sufficiently small, the nonlinearity of the cumulative distribution function CDF(x) does not become a problem.

(Step S4)
Next, the correction function generation unit 82 calculates the correction function p(x _raw (t)) using the cumulative distribution function CDF(x) and the threshold values th1, th2 (th1>th2) (step S4).

FIG. 7 is a diagram showing an example of a time waveform of a sensor signal when a bearing in the speed increaser is damaged. In FIG. 7, vibrations caused by damage to the bearing appear at a period of about 0.3 seconds. As shown in FIG. 7, the threshold th1 is preferably set to be slightly larger than the positive peak intensity of the noise component. It is preferable that the threshold th2 is set to be slightly smaller than the negative peak intensity of the noise component.

The threshold values th1 and th2 may be determined in advance depending on the type of sensor 70. When a corrected signal corresponding to a sensor signal acquired within the diagnosis period is generated after the end of the diagnosis period, the correction function generation unit 82 generates a signal according to the intensity distribution of the sensor signal acquired from the sensor 70 during the diagnosis period. You may set threshold values th1 and th2. Alternatively, when the corrected signal is generated in real time during the diagnosis period, the correction function generation unit 82 sets the thresholds th1 and th2 according to the intensity distribution of the sensor signal acquired from the sensor 70 in a period past the diagnosis period. May be set.

For example, the correction function generation unit 82 uses the standard deviation σ and the average value AVE calculated from the intensity distribution of the sensor signal acquired from the sensor 70 during the diagnosis period or the past period, and sets AVE+cσ as the threshold value th1, AVE-cσ may be set as the threshold th2. c is a predetermined constant.

The correction function generation unit ₈₂ generates a correction function p( x _raw (t)). Specifically, the correction function generation unit 82 calculates the correction function p(x _raw (t)) according to the following equation (5).
p(x _raw (t))=1-CDF(th1-x _raw (t))+CDF(th2-x _raw (t))
...Equation (5).

FIG. 8 is a diagram showing an example of the correction function p(x _raw (t)). As shown in FIG. 8, the correction function p(x _raw (t)) becomes approximately 0 when the intensity x _raw (t) of the sensor signal is around the expected value μ (0 in FIG. 8). However, when the intensity x _raw (t) is larger than the expected value μ, the correction function p(x _raw (t)) increases as (x _raw (t)−μ) increases. The slope of the correction function p(x _raw (t)) is maximum at the threshold value th1. Further, when the intensity x _raw (t) is smaller than the expected value μ, the correction function p(x _raw (t)) increases as (μ−x _raw (t)) increases. The slope of the correction function p(x _raw (t)) is minimum at the threshold value th2.

FIG. 9 is a diagram showing the first-order derivative of the correction function p(x _raw (t)) shown in FIG. 8. As shown in FIG. 9, the first-order derivative of the correction function p(x _raw (t)) has a maximum value at the threshold th1 and a minimum value at the threshold th2.

<Example of corrected signal>
FIG. 10 is a diagram showing an example of a corrected signal output from the correction section. FIG. 10 shows a time waveform of a corrected signal obtained by correcting the sensor signal shown in FIG. 7 using the correction function p(x _raw (t)) shown in FIG.

As shown in FIG. 10, signal components with a high probability of being noise are effectively reduced, and the S/N ratio is improved. As a result, the peak of vibration (impact vibration with a period of about 0.3 seconds) caused by damage to the bearing becomes clear.

<Evaluation value calculation section>
The evaluation value calculation unit 85 calculates an evaluation value indicating the characteristics of the time waveform of the corrected signal during the diagnosis period. When an abnormality occurs in the speed increaser 40, the time waveform of the corrected signal changes. Therefore, the evaluation value numerically represents the characteristics of the abnormality of the speed increaser 40. By using the time waveform of the corrected signal with an improved signal-to-noise ratio, the sensitivity of the evaluation value to abnormalities is improved. As a result, abnormality detection can be detected earlier and with higher accuracy. In particular, even under low rotational speed conditions (for example, 100 min ^-1 or less) where the SN ratio tends to decrease, the sensitivity of the evaluation value to abnormalities is good.

As evaluation values, for example, the root mean square value (RMS) calculated by the following formula (6), the crest factor (CF) calculated by the formula (7), and the formula (8) Kurtosis calculated by the above method may be used.

By using a corrected signal with an improved signal-to-noise ratio, the effective value representing the vibration level has good sensitivity to abnormal vibrations occurring at the damaged area of the bearing. For the same reason, the crest factor, which is the ratio between the peak value and the effective value of the signal, and the kurtosis, which indicates the sharpness of the frequency distribution of the signal (also called ``sharpness''), also have good sensitivity to abnormal vibrations. have

Alternatively, the evaluation value calculation unit 85 may calculate the evaluation value from the frequency spectrum obtained by frequency analysis (Fourier transform) of the corrected signal. As a method of calculating the evaluation value from the frequency spectrum, for example, a known technique described in Japanese Patent Application Laid-Open No. 2018-179977 (Patent Document 3) may be adopted. Since noise components are also reduced on the frequency spectrum, frequency peaks related to bearing damage clearly appear. Therefore, the evaluation value has good sensitivity to abnormal vibrations.

The evaluation value calculation unit 85 may calculate multiple types of evaluation values. For example, the evaluation value calculation unit 85 may calculate an effective value and kurtosis.

<Judgment section>
The determination unit 86 uses the evaluation value to determine the presence or absence of an abnormality. Typically, the determination unit 86 compares the evaluation value with a predetermined threshold, and determines the presence or absence of an abnormality according to the comparison result. If a plurality of types of evaluation values are calculated, the determination unit 86 may compare each of the plurality of types of evaluation values with a threshold value and determine the presence or absence of an abnormality according to a combination of the plurality of comparison results.

Furthermore, when determining that there is an abnormality, the determination unit 86 may estimate the location where the abnormality has occurred. As a method for estimating a site where an abnormality has occurred, a known technique described in Patent Document 3, for example, may be adopted.

The determination unit 86 may output the determination result to an external device. Thereby, the operator operating the external device can check whether there is any abnormality in the speed increaser 40 of the wind power generator 10.

<Effect>
As described above, the state monitoring device 80 of the present disclosure monitors the state of the speed increaser 40, which is an example of a target object. The condition monitoring device 80 includes a correction section 83 and a diagnosis section 84. The correction unit 83 adds a correction coefficient k( corresponding to the intensity x _raw (t) to the intensity x _raw (t) of the sensor signal acquired from the sensor 70 installed in the speed increaser 40 during the diagnosis period (first period). t) to generate a corrected signal (second signal). The diagnosis unit 84 diagnoses an abnormality in the speed increaser 40 based on a time waveform indicating a change over time in the intensity of the corrected signal. When μ is the expected value of the intensity distribution of the sensor signal acquired from the sensor 70 during the diagnosis period or a period past the diagnosis period (second period), the correction coefficient satisfies the first condition and the second condition. The first condition is that when μ<x _raw (t), the correction coefficient k(t) increases as (x _raw (t)−μ) increases. The second condition is that when μ>x _raw (t), the correction coefficient k(t) increases as (μ−x _raw (t)) increases.

The intensity of many noise components is close to the expected value μ. Therefore, when μ<x _raw (t), the intensity of the component caused by the abnormality of the speed increaser 40 is emphasized more than the noise component. Similarly, when μ>x _raw (t), the intensity of the component caused by the abnormality of the speed increaser 40 is emphasized more than the noise component. Therefore, even if the S/N ratio of the sensor signal is low, an abnormality in the speed increaser 40 can be diagnosed with high accuracy based on the time waveform showing the change over time in the intensity of the corrected signal.

Furthermore, according to the above configuration, unlike Patent Document 2, there is no need to perform the synthesis process of vibration waveform data and noise data multiple times. Therefore, measurement costs are suppressed.

In this way, even if the S/N ratio of the sensor signal obtained from the sensor 70 is low, an abnormality in the speed increaser 40 can be diagnosed at low measurement cost.

The condition monitoring device 80 includes a correction function generation unit 82 that generates a correction function p(x _{raw (t)) with the intensity x raw (} t) of the sensor signal as an explanatory variable and the correction coefficient k(t) as an objective variable. Be prepared for more. The correction unit 83 determines the correction coefficient k(t) by substituting the intensity x _raw (t) of the sensor signal into the correction function p(x _raw (t)). Thereby, the correction coefficient k(t) can be easily determined using the correction function p(x _raw (t)).

The first-order derivative of the correction function p(x _raw (t)) has a maximum value at a threshold value th1 that is larger than the expected value μ. As a result, as the intensity x _raw (t) approaches the threshold value th1 from the expected value μ, the correction coefficient k(t) sharply increases. Therefore, noise components having an intensity from the expected value μ to the threshold th1 are reduced, and components with an intensity exceeding the threshold th1, which are caused by an abnormality in the speed increaser 40, are more emphasized.

Similarly, the first-order derivative of the correction function p(x _raw (t)) has a minimum value at a threshold value th2 that is smaller than the expected value μ. As a result, as the intensity x _raw (t) approaches the threshold value th2 from the expected value μ, the correction coefficient k(t) sharply increases. Therefore, noise components having an intensity from the expected value μ to the threshold th2 are reduced, and components with an intensity exceeding the threshold th2, which are caused by an abnormality in the speed increaser 40, are more emphasized.

The correction function generation unit 82 sets a probability density function PDF(x) according to the intensity distribution of sensor signals acquired during the diagnosis period or a period past the diagnosis period. Further, the correction function generation unit 82 sets at least one of a threshold value th1 that is larger than the expected value μ and a threshold value th2 that is smaller than the expected value μ. When both the thresholds th1 and th2 are set, the correction function generation unit 82 converts the above equation (5) into a correction function using the cumulative distribution function CDF(x) calculated from the probability density function PDF(x). Set as p(x _raw (t)).

According to the above configuration, the correction function p(x _raw (t)) for calculating the correction coefficient k(t) that satisfies the first and second conditions described above can be easily set.

<Modification 1>
In the above description, the correction function generation unit 82 sets the threshold values th1 and th2, and calculates the correction function p(x _raw (t)) using the threshold values th1 and th2. However, the correction function generation unit 82 may set only one of the threshold values th1 and th2.

When setting only the threshold value th1 larger than the expected value μ, the correction function generation unit 82 calculates the correction function p(x _raw (t)) according to the following equation (9).
p(x _raw (t)) = 1-CDF(th1-x _raw (t))...Equation (9)
The correction function p(x _raw (t ₎ ) calculated according to equation (9) _is calculated by the correction coefficient The condition that k(t) becomes large) is satisfied. However, the correction function p(x _raw (t)) calculated according to equation ( ₉ ) _is (the condition that the correction coefficient k(t) becomes large) is not satisfied.

When setting only the threshold value th2 smaller than the expected value μ, the correction function generation unit 82 calculates the correction function p(x _raw (t)) according to the following equation (10).
p(x _raw (t))=1+CDF(th2-x _raw (t))...Equation (10)
The correction function p(x _raw (t)) calculated according to equation (10) satisfies the above second condition. However, the correction function p(x _raw (t)) calculated according to equation (10) does not satisfy the above first condition.

When an abnormality occurs in the speed increaser 40, the amplitude of the intensity of the sensor signal increases. Therefore, when the correction function p(x _raw (t)) is calculated according to equation (9), a positive peak resulting from the occurrence of an abnormality clearly appears in the time waveform of the intensity of the corrected signal. On the other hand, when the correction function p(x _raw (t)) is calculated according to equation (10), a negative peak resulting from the occurrence of an abnormality clearly appears in the time waveform of the intensity of the corrected signal. Therefore, even if the correction function p(x _raw (t)) is calculated using Equation (9) or Equation (10), an abnormality in the speed increaser 40 can be diagnosed with high accuracy.

<Modification 2>
In the wind power generation device 10, the magnitude of the load and the rotation speed may change from moment to moment due to changes in the environment such as wind direction and wind power. In such a case, the magnitude of the vibration of the speed increaser 40 and the magnitude of the noise of the sensor signal may also change from moment to moment. Therefore, it is preferable that the correction function generation unit 82 periodically updates the correction function p(x _raw (t)).

FIG. 11 is a diagram schematically showing the update timing of the correction function of the condition monitoring device according to the second modification. FIG. 11 shows an example in which the correction function p(x _raw (t)) is updated every cycle T.

The correction function generation unit 82 updates the correction function p(x _raw (t)) for determining the correction coefficient for each section having the same time length as the period T. As shown in FIG. 11, the correction function generation unit 82 stores data indicating the intensities x _raw (t1) to x _raw (t2) of the sensor signals acquired in the interval A from time t1 to time t2 in the storage unit 87. Then, a correction function p _a (x _raw (t)) is generated using the intensities x _raw (t1) to x _raw (t2).

Specifically, in step S1 shown in FIG. 4, the correction function generation unit 82 selects one type of probability distribution that fits the shape of the distribution of the intensities x _raw (t1) to x _raw (t2). In step S2, the correction function generation unit 82 sets the values of the parameters of the probability distribution so as to fit the shape of the distribution of the intensities x _raw (t1) to x _raw (t2), and generates the probability density function PDF(x). Calculate. Then, the correction function generation unit 82 calculates the cumulative distribution function CDF(x) from the probability density function PDF(x). In step S4, the correction function generation unit 82 sets AVE+cσ as the threshold th1 using, for example, the standard deviation σ and the average value AVE calculated from the distribution of the intensities x _raw (t1) to x _raw (t2), AVE-cσ is set as the threshold th2. The correction function generation unit 82 calculates the correction function p _a (x _raw (t)) according to the above equation (5) using the cumulative distribution function CDF(x) and the thresholds th1 and th2.

When the correction function p _a (x _raw (t)) is calculated, the correction unit 83 uses the correction function _{p a} (x _raw (t)) to calculate the correction coefficient k (t ) to determine. The correction unit 83 generates a corrected signal at times t1 to t2 by multiplying the intensity x _raw (t) by the correction coefficient k(t).

When the corrected signal from time t1 to t2 is generated, the diagnostic unit 84 uses the corrected signal to diagnose an abnormality in the object. Note that the time length of section A may be the same as the diagnosis period, or may be multiple times the diagnosis period. When the time length of section A is multiple times the diagnosis period, the diagnosis unit 84 may divide section A into a plurality of diagnosis periods and diagnose the abnormality of the object for each diagnosis period.

Similarly, for the next section B after section A, after the end of section B, _the correction function p _b ( _{x raw (} t)) is generated. Thereafter, the correction unit 83 uses the correction function p _b (x _raw (t)) to determine the correction coefficient k(t) at each time from time t2 to t4. The correction unit 83 generates a corrected signal at times t2 to t4 by multiplying the intensity x _raw (t) by the correction coefficient k(t). The diagnostic unit 84 diagnoses an abnormality in the object using the corrected signal.

According to the state monitoring device according to the second modification, the correction function p(x _raw (t)) is periodically updated depending on the state of the object (eg, rotational speed). As a result, abnormalities can be diagnosed with high accuracy.

<Modification 3>
In the state monitoring device according to the second modification, when changing from section A to section B, the correction function is updated from p _a (x _raw (t)) to p _b (x _raw (t)). Therefore, the correction coefficient also changes nonlinearly. In order to alleviate such nonlinear changes, the correction function generation unit 82 may perform smoothing processing on the correction function p(x _raw (t)).

FIG. 12 is a flowchart showing the process flow of the correction function generation unit according to the third modification. The flowchart shown in FIG. 12 is different from the flowchart shown in FIG. 4 in that it includes a smoothing process shown in step S5.

In step S5, the correction function generation unit 82 sets a transition period when updating the correction function. For example, when updating the correction function p _a (x _raw (t)) shown in FIG. 11 to the correction function p _b (x _raw (t)), the correction function generation unit 82 updates A period within section B up to time t3 is set as a transition period. Time t3 is, for example, the middle point in section B.

The correction function generation unit 82 gradually changes the correction function p _a (x _raw (t)) to the correction function p _b (x _raw (t)) during the transition period. For example, the correction function generation unit 82 calculates the correction function p(x _raw (t)) at time t within times t2 to t3 according to the following equation (11).

FIG. 13 is a diagram showing changes in the correction function during the transition period. As shown in FIG. 13, during the transition period, the correction function gradually changes from p _a (x _raw (t)) to p _b (x _raw (t)).

Note that the transition period is not limited to the period from time t2 to time t3. For example, the period from the middle time of section A to time t3 may be set as the transition period. In this case, correction and diagnosis for the sensor signal intensity x _raw (t) acquired during the diagnosis period in section A are performed after the correction function p _b (x _raw (t)) corresponding to section B is calculated. be done.

<Modification 4>
In the above description, it is assumed that the condition monitoring device 80 includes the correction function generation section 82. However, the correction function generation unit 82 may be provided in an external device. In this case, the condition monitoring device 80 obtains the correction function p(x _raw (t)) from the external device. The correction unit 83 of the condition monitoring device 80 may determine the correction coefficient using the obtained correction function p(x _raw (t)).

The embodiments disclosed this time should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is indicated by the claims rather than the description of the embodiments described above, and it is intended that all changes within the meaning and scope equivalent to the claims are included.

10 wind power generator, 20 main shaft, 25 hub, 30 blade, 40 speed increaser, 50 generator, 60 main shaft bearing, 70 sensor, 80 condition monitoring device, 81 storage processing section, 82 correction function generation section, 83 correction section , 84 diagnosis section, 85 evaluation value calculation section, 86 judgment section, 87 storage section, 90 nacelle, 100 tower, 801 CPU, 802 RAM, 803 storage, 804 communication interface, 805 bus, 806 monitoring program.

Claims (11)

A condition monitoring device that monitors the condition of an object,
a correction unit that generates a second signal by multiplying the intensity of the first signal acquired from the sensor installed on the object in the first period by a correction coefficient according to the intensity;
a diagnostic unit that diagnoses an abnormality in the object based on a time waveform indicating a change in intensity of the second signal over time;
If the expected value of the intensity distribution of the first signal or the third signal acquired from the sensor in a second period in the past than the first period is μ, and the intensity of the first signal is x _raw , then the correction The first condition is that the correction coefficient increases as (x _raw - μ) increases when μ<x _raw , and the correction coefficient increases as (μ-x _raw ) increases when μ>x _raw . A condition monitoring device configured to satisfy at least one of the second conditions that the coefficient becomes large. Further comprising a generation unit that generates a correction function using the intensity of the first signal as an explanatory variable and the correction coefficient as an objective variable,
The state monitoring device according to claim 1, wherein the correction unit determines the correction coefficient by substituting the intensity of the first signal into the correction function. the correction coefficient satisfies the first condition;
The condition monitoring device according to claim 2, wherein the first-order derivative of the correction function has a maximum value at a first threshold value that is larger than the expected value. the correction coefficient satisfies the second condition;
The condition monitoring device according to claim 2 or 3, wherein the first-order derivative of the correction function has a minimum value at a second threshold value that is smaller than the expected value. The generation unit is
Setting a probability density function according to the intensity distribution,
setting at least one of a first threshold larger than the expected value and a second threshold smaller than the expected value,
Assuming that the cumulative distribution function calculated from the probability density function is CDF(x), the first threshold is th1, and the second threshold is th2, when both the first threshold and the second threshold are set, 1-CDF(th1-x _raw )+CDF(th2-x _raw ) is set as the correction function,
When the first threshold is set and the second threshold is not set, 1-CDF (th1-x _raw ) is set as the correction function,
The condition monitoring device according to claim 2, wherein when the first threshold value is not set and the second threshold value is set, 1+CDF(th2-x _raw ) is set as the correction function. The condition monitoring device according to claim 2 , wherein the generation unit periodically updates the correction function. The generation unit is
Setting a transition period when updating the first correction function to the second correction function,
The condition monitoring device according to claim 6, wherein the first correction function is gradually changed to the second correction function during the transition period. The diagnostic department includes:
an evaluation value calculation unit that calculates an evaluation value indicating the characteristics of the time waveform;
The condition monitoring device according to any one of claims 1 to 7, further comprising a determination unit that uses the evaluation value to determine whether or not there is an abnormality in the object. The condition monitoring device according to claim 8, wherein the evaluation value includes at least one of an effective value, a crest factor, and a kurtosis of the time waveform. The condition monitoring device according to claim 8, wherein the evaluation value is calculated from a frequency spectrum obtained by frequency analyzing the time waveform. A condition monitoring method for monitoring the condition of an object, the method comprising:
generating a second signal by multiplying the intensity of the first signal acquired from the sensor installed on the object in the first period by a correction coefficient according to the intensity;
diagnosing an abnormality in the object based on a time waveform indicating a change in intensity of the second signal over time,
If the expected value of the intensity distribution of the first signal or the third signal acquired from the sensor in a second period in the past than the first period is μ, and the intensity of the first signal is x _raw , then the correction The first condition is that the correction coefficient increases as (x _raw - μ) increases when μ<x _raw , and the correction coefficient increases as (μ-x _raw ) increases when μ>x _raw . A state monitoring method configured to satisfy at least one of the second conditions that the coefficient becomes large.

Images