JP2022158570A - Abnormality detection device, abnormality detection method, and program - Google Patents

Abstract

To provide an abnormality detection device which can quickly diagnose the abnormality of a machinery plant.SOLUTION: An abnormality detection device for detecting the presence of abnormality in a machinery plant includes a processor that executes: a process of acquiring, on the basis of a detection signal outputted from a first vibration sensor that measures the vibration of machinery plant, a measured value composed of the amplitude and phase of the vibration at a predetermined frequency point; a process of calculating, with respect to a unit space composed of a plurality of measured values acquired at a plurality of past points of time, the Mahalanobis distance of the measured value acquired at a point of time when evaluating the machinery plant; and a process of assessing, when the calculated Mahalanobis distance exceeds a prescribed threshold, that abnormality has occurred in the machinery plant.SELECTED DRAWING: Figure 1

Description

The present disclosure relates to an abnormality detection device, an abnormality detection method, and a program.

As a technique for diagnosing the soundness of an inspection target such as mechanical equipment, a method of detecting an abnormality using the MT (Mahalanobis-Taguchi) method is known. In the MT method, the Mahalanobis distance is calculated using the inverse matrix of the covariance matrix of reference data (for example, a group of measured values of various characteristic items in a normal state). Then, if the calculated Mahalanobis distance exceeds a predetermined threshold, it can be determined that the state of the object to be inspected is abnormal.

For example, Patent Literature 1 describes an anomaly detection device that detects an anomalous vibration of a rotating machine based on the results of frequency analysis of a measurement signal y from a vibration sensor.

Japanese Patent Application Laid-Open No. 2020-144111

The results of frequency analysis are specifically Fourier coefficients obtained as a result of discrete Fourier transform. The anomaly detection device described in Patent Document 1 uses the time interval ΔT of the time history of the measured signal y {y(ΔT), y(2ΔT), . The complex Fourier coefficients {a(0Δf)+jb(0Δf),a(1Δf)+jb(1Δf),a( 2Δf)+jb(2Δf), . . . , a((N−1)Δf)+jb((N−1)Δf)}. j is the imaginary unit. A fast Fourier transform (FFT) is used for the discrete Fourier transform in terms of computation speed. Executing this in the control device of the machinery has problems in terms of the amount of computational processing and the immediacy of detection.

In FFT, the number of frequency samples is set to be a power of 2, such as 256 points or 1024 points per signal, and one Fourier coefficient is calculated for each set number of frequency samples. , the number of calculations of Fourier coefficients also increases to 256 points and 1024 points. That is, in FFT, the same number of Fourier coefficients as the number of frequency samples must be calculated.

If the signal is 1ch, frequency analysis by FFT is possible. However, it may become difficult when there are, for example, 16ch measurement signals. If the number of frequency samples is 256, 16ch×256 points, that is, 4096 points of Fourier coefficients will be calculated. In this case, even though the calculation speed has improved today, it cannot be processed by a general control device, and a dedicated arithmetic device (frequency analysis device) must be provided. This is not desirable from a product cost point of view.

Furthermore, even if the Fourier coefficients of 256 frequency points per channel are calculated, only some of them are often used for vibration diagnosis. Since the frequency points where anomalies are likely to appear are known in advance, it is only necessary to know the Fourier coefficients of the frequency points belonging to the frequency band. Most of the computed Fourier coefficients are usually not used. However, in calculating the Fourier coefficients, it is not possible to limit the frequency points and calculate the Fourier coefficients. For this reason, in the conventional technique, even the Fourier coefficients of frequency points that are not used are calculated, which is wasteful.

It's not just a waste of computation. FFT also has a problem in terms of immediacy. In FFT, if the time interval of the measurement signal is ΔT [s], and the number of measurement points to be subjected to FFT is N, the frequency resolution, that is, the increment of frequency points is calculated as Δf = (ΔTN) - 1 [Hz]. , the Fourier coefficients for the frequency points {0Δf, 1Δf, 2Δf, . . . , (N−1)Δf}. For example, ΔT=0.01 sec, and in order to evaluate the resolution at 0.01 Hz, a data length of N=1024 points is sufficient. At this time, the time required to accumulate data becomes a problem. It takes 10.24 seconds to accumulate 1024 points with ΔT=0.01 seconds. If the object of interest is, for example, a 0.1 Hz component (10 second period), it does not matter if the accumulation takes 10 seconds. However, if, for example, you want to check a frequency component of 1 Hz (1-second period), you want to check it in 1 second instead of 10 seconds. In order to quickly detect anomalies and symptoms, it is important to make judgments with short data. Therefore, the problem of the conventional FFT is that it takes time to accumulate data.

Furthermore, in FFT, only the average frequency characteristics of accumulated N points can be obtained. For example, assume that an abnormality appears in the 1 Hz component, and that the characteristic of the abnormality appears in one second in the middle of the accumulated N-point data interval. In the FFT, since the average frequency characteristic for 10 seconds is calculated, the feature of an abnormality appearing in a certain 1 second is diluted by the data of the remaining 9 seconds, making abnormality detection difficult. Thus, feature dilution is also a problem of FFT.

The present disclosure has been made in view of such problems, and provides an abnormality detection device, an abnormality detection method, and a program capable of performing abnormality diagnosis of mechanical equipment at high speed.

According to one aspect of the present disclosure, an abnormality detection device that detects the presence or absence of an abnormality in mechanical equipment includes a predetermined frequency point based on a detection signal output from a first vibration sensor that measures vibration of the mechanical equipment. Acquired at the time of evaluating the mechanical equipment based on a process of acquiring a measured value consisting of the amplitude and phase of the vibration in and a unit space composed of the plurality of measured values acquired at a plurality of points in the past. a processor that executes a process of calculating the Mahalanobis distance of the measured value that has been calculated, and a process of determining that an abnormality has occurred in the mechanical equipment if the calculated Mahalanobis distance exceeds a predetermined threshold .

Further, according to one aspect of the present disclosure, the abnormality detection device that detects the presence or absence of an abnormality in the mechanical equipment is based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment. A process of acquiring a measured value consisting of the amplitude and phase of the vibration at a frequency point, and a process of calculating a plurality of components by resolving the measured value vector consisting of the measured value in the direction of an index vector defined as an index of abnormality. and determining that an abnormality has occurred in the mechanical equipment when the calculated absolute value of the component exceeds a predetermined threshold.

Further, according to one aspect of the present disclosure, an abnormality detection device that detects the presence or absence of an abnormality in mechanical equipment acquires a measured value vector composed of measured values measured by a first vibration sensor that measures vibration of the mechanical equipment. and dividing an index vector determined as an abnormality index into a real part and an imaginary part using a phase angle of a predetermined frequency point, and calculating the inner product of each of the real part and the imaginary part and the measurement value vector a process of determining that an abnormality has occurred in the mechanical equipment when the absolute value of the time average for one cycle of the calculated inner product exceeds a predetermined threshold;
a processor that executes

Further, according to one aspect of the present disclosure, the abnormality detection method for detecting the presence or absence of an abnormality in the mechanical equipment is based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment. A step of obtaining measured values consisting of the amplitude and phase of the vibration at frequency points; calculating the Mahalanobis distance of the measured value obtained in , and determining that an abnormality has occurred in the mechanical equipment when the calculated Mahalanobis distance exceeds a predetermined threshold.

Further, according to one aspect of the present disclosure, the program causes a computer of an abnormality detection device that detects the presence or absence of an abnormality in the mechanical equipment based on a detection signal output from a first vibration sensor that measures vibration of the mechanical equipment. a step of acquiring a measured value consisting of the amplitude and phase of the vibration at a predetermined frequency point; a step of calculating the Mahalanobis distance of the measured value obtained at the time of evaluating the equipment; and a step of determining that an abnormality has occurred in the mechanical equipment when the calculated Mahalanobis distance exceeds a predetermined threshold. , is executed.

According to the abnormality detection device, the abnormality detection method, and the program according to the present disclosure, abnormality diagnosis of vibration of mechanical equipment can be performed at high speed.

1 is a diagram showing the overall configuration of an anomaly detection system according to a first embodiment of the present disclosure; FIG. It is a figure which shows an example of a process of the abnormality detection apparatus which concerns on 1st Embodiment of this indication. It is a figure which shows the whole structure of the abnormality detection system which concerns on 2nd Embodiment of this indication. FIG. 10 is a diagram illustrating an example of processing of an abnormality detection device according to a second embodiment of the present disclosure; FIG. FIG. 11 is a diagram showing the functional configuration of an abnormality detection device according to a third embodiment of the present disclosure; FIG. FIG. 11 is a diagram illustrating an example of processing of an abnormality detection device according to a third embodiment of the present disclosure; FIG. FIG. 11 is a diagram showing a functional configuration of an abnormality detection device according to a fourth embodiment of the present disclosure; FIG. FIG. 11 is a diagram illustrating an example of processing of an abnormality detection device according to a fourth embodiment of the present disclosure; FIG. FIG. 13 is a diagram illustrating an example of processing of an abnormality detection device according to a fifth embodiment of the present disclosure; FIG.

<First Embodiment>
An abnormality detection device 2 according to a first embodiment of the present disclosure will be described below with reference to FIGS. 1 and 2. FIG.

(overall structure)
FIG. 1 is a diagram showing the overall configuration of mechanical equipment according to the first embodiment of the present disclosure.
As shown in FIG. 1 , the abnormality detection system 100 includes mechanical equipment 10 , an abnormality detection device 2 and a control device 3 .

The mechanical equipment 10 is a machine or group of devices to be subjected to abnormality diagnosis by the abnormality detection device 2 . A first vibration sensor 11 (11A to 11D) is attached to each measurement point of the mechanical equipment 10, respectively. The first vibration sensor 11 is, for example, an accelerometer, a microphone that measures sound caused by vibration, a pressure gauge that measures pressure caused by vibration, a stress gauge that measures stress caused by vibration, or the like. The first vibration sensor 11 outputs a detection signal obtained by measuring the vibration of the mechanical equipment 10 to the abnormality detection device 2 .

The abnormality detection device 2 detects whether there is an abnormality in the mechanical equipment 10 based on the detection signal acquired from the first vibration sensor 11 of the mechanical equipment 10 . A functional configuration of the abnormality detection device 2 will be described later.

The control device 3 generates control signals for controlling the operation of the machinery 10 . For example, when the control device 3 receives a determination result indicating that an abnormal vibration is occurring in the mechanical equipment 10 from the abnormality detection device 2, a control signal for stopping the operation of the mechanical equipment 10, an output of the mechanical equipment 10 to generate a control signal for lowering

(Functional configuration of abnormality detection device)
Next, the functional configuration of the abnormality detection device 2 will be described with reference to FIG.
As shown in FIG. 1 , the abnormality detection device 2 includes a processor 21 , an input/output unit 22 , a main storage device 23 and an auxiliary storage device 24 .

The processor 21 controls the overall operation of the abnormality detection device 2 . The processor 21 functions as a measured value acquisition unit 210, a unit space generation unit 211, a Mahalanobis distance calculation unit 212, a determination unit 213, and a frequency point designation unit 214 by operating according to a predetermined program.

Based on the detection signal output from the first vibration sensor 11 of the mechanical equipment 10 , the measured value acquisition unit 210 acquires measured values including the amplitude and phase of vibration at predetermined frequency points.

The unit space generation unit 211 generates a unit space including a plurality of measured values acquired at a plurality of past points in time. Also, the unit space generation unit 211 stores the generated unit space in the auxiliary storage device 24 .

The Mahalanobis distance calculation unit 212 calculates the Mahalanobis distance of the measurement values acquired at the time of evaluating the mechanical equipment 10 with reference to the unit space.

The determination unit 213 determines whether or not an abnormality has occurred in the mechanical equipment 10 based on the calculated Mahalanobis distance. Specifically, the determination unit 213 determines that the mechanical equipment 10 is abnormal when the calculated Mahalanobis distance exceeds a predetermined threshold.

The frequency point designation unit 214 designates a predetermined frequency point for the measurement value acquisition unit 210 .

The input/output unit 22 receives an input of a detection signal from the first vibration sensor 11 of the mechanical equipment 10 . In addition, the input/output unit 22 outputs the determination result of the presence/absence of abnormality by the abnormality detection device 2 to the control device 3 .

The main storage device 23 has storage areas required for the operation of the processor 21 .

The auxiliary storage device 24 is a large-capacity storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The auxiliary storage device 24 stores various data acquired or generated by each part of the abnormality detection device 2 .

(Processing of anomaly detection device)
FIG. 2 is a diagram illustrating an example of processing of the abnormality detection device according to the first embodiment of the present disclosure;
The abnormality diagnosis processing of the abnormality detection device 2 will be described in detail below with reference to FIG.

The feature of this embodiment resides in means of frequency analysis executed in the abnormality detection device 2 . As mentioned above, FFT has the following problems: (a) the amount of calculation is heavy and requires a processing device, (b) it takes time to accumulate data points, and (c) the characteristics of abnormalities are diluted. . Therefore, in the abnormality detection device 2 according to the present embodiment, the measurement value acquisition unit 210 uses the principle equations (1A) and (1B) of the Fourier transform instead of the FFT to measure the frequency characteristics of each detection signal. get the value.

Here, n is a predetermined serial number of the detection signal. For example, as shown in FIG. 1, if detection signals are obtained from each of the four vibration sensors 11, n is any of {1, 2, 3, 4}, and the value of n determines the period { Calculate the Fourier coefficients of the frequency components of T ₁ , T ₂ , T ₃ , T ₄ }. The values of periods {T ₁ , T ₂ , T ₃ , T ₄ } may be the same.

Note that the detection signals {y ₁ , y ₂ , y ₃ , y ₄ } may include the same signal obtained by branching the signal of the same vibration sensor 11 . However, when multiple identical signals are included in the detected signals {y ₁ , y ₂ , y ₃ , y ₄ }, the values of the periods {T ₁ , T ₂ , T ₃ , T ₄ } for the multiple signals are changed. is common. Specifically, for example, if _y1 and y2 _are branched signals from the same vibration sensor 11A, _T1 and T2 _are generally set to different values. Unless there is a special situation such as disconnection of the branch wiring, the Fourier coefficients of the same periodic components of y1 and y2 branched from the same vibration sensor _are the same, so it is sufficient to calculate only _one of them. be.

Equations (1A) and (1B) can be calculated from the time history data of one period, that is, the time history data from t-T _n to t, where t is the current time, so the above (a) and ( C) problem is solved. Moreover, since the equations (1A) and (1B) are simple integral equations, the above problem (a) is also solved. Furthermore, according to the formulas (1A) and (1B), the time history data need not be evenly spaced. In FFT, the period T _n had to be an integer multiple of the time interval ΔT of the measurement signal. However, in equations (1A) and (1B), this is not necessary, and calculation is possible for frequency points that are not integral multiples of ΔT as in equation (2) below. Suppose that the period _Tn is represented by the following equation (2).

where m _n is the quotient of T _n with respect to the divisor ΔT and r _n is its remainder.

For example, a developer, an operator, or the like of the mechanical equipment 10 has, as prior knowledge, frequency points at which characteristics of each detection signal are likely to appear. The periods {T ₁ , T ₂ , T ₃ , T ₄ } are set based on such prior knowledge. As shown in FIG. _{2 ,} the frequency point designation unit 214 _sets predetermined _periods {T ₁ , T ₂ , T ₃ , T ₄ }.

Then, the equations (1A) and (1B) are calculated by the following equations (3A) and (3B).

That is, the measured value acquiring unit 210 uses the above equations (3A) and (3B) for each of the detection signals y _n acquired from the vibration sensor 11 at the current time t, and the period T _n Obtain the Fourier coefficients {a _n ,b _n } in . The Fourier coefficients {a _n , b _n } correspond to measurements of amplitude and phase of vibration at predetermined frequency points.

In addition, the Mahalanobis distance calculation unit 212 calculates the Mahalanobis distance of the measurement values {a _n , b _n } at the time t acquired by the measurement value acquisition unit 210 based on the unit space stored in the auxiliary storage device 24. . Note that the unit space is generated in advance by the unit space generation unit 211 based on a plurality of measurement values (Fourier coefficients) acquired during the normal operation of the mechanical equipment 10 . As for the method of generating the unit space and the method of calculating the Mahalanobis distance, the existing technology (for example, the technology described in Patent Literature 1) can be used, so the description is omitted.

Next, when the Mahalanobis distance calculated by the Mahalanobis distance calculation unit 212 exceeds a predetermined threshold, the determination unit 213 determines that the mechanical equipment 10 is abnormal. On the other hand, when the Mahalanobis distance is equal to or less than the threshold, the determination unit 213 determines that the mechanical equipment 10 is not abnormal. Also, the determination unit 213 outputs the determination result to the control device 3 . When the control device 3 receives a determination result indicating the occurrence of an abnormality from the abnormality detection device 2, the control device 3 performs control such as stopping the operation of the mechanical equipment 10 or reducing the output.

The abnormality detection device 2 detects whether or not there is an abnormality in the mechanical equipment 10 by repeatedly executing the above-described processing.

(Effect)
As described above, the abnormality detection device 2 according to the present embodiment acquires the measured values (Fourier coefficients an, _{bn )} consisting of the amplitude and phase of the vibration in the predetermined cycle _Tn for the detection signal _yn . Then, when the Mahalanobis distance of the acquired measurement value exceeds a predetermined threshold, it is determined that the mechanical equipment 10 has an abnormality.

By doing so, the abnormality detection device 2 needs to calculate only Fourier coefficients at specific frequency points, so that the computational load can be greatly reduced compared to the conventional technology. As a result, the abnormality detection device 2 can perform abnormality diagnosis of the mechanical equipment 10 at high speed. Moreover, since the abnormality detection device 2 can calculate Fourier coefficients by itself without separately preparing a frequency analysis device, the cost of the abnormality detection system 100 as a whole can be reduced.

Further, the abnormality detection device 2 uses the principle formula of Fourier transform to specify one cycle _{Tn for one detection signal yn, and calculates Fourier coefficients {a n ,} b _{n }} .

By doing so, the abnormality detection device 2 can perform abnormality diagnosis from the time history data for one cycle, so that the time for accumulating data can be shortened. For example, if it is desired to examine the frequency component of 1 Hz (1-second cycle) for a certain detection signal, it is possible to make a diagnosis from time history data for the past 1 second without accumulating data for several seconds. In addition, since the abnormality detection device 2 can make a diagnosis with a small number of data points, it is possible to suppress the dilution of the characteristics of the abnormality. Thereby, the abnormality detection device 2 can improve the sensitivity to abnormality.

<Second embodiment>
Next, an abnormality detection device 2 according to a second embodiment of the present disclosure will be described with reference to FIGS. 3 and 4. FIG.
Components common to those of the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

(overall structure)
FIG. 3 is a diagram showing the overall configuration of an anomaly detection system according to the second embodiment of the present disclosure.
As shown in FIG. 3 , in the abnormality detection system 100 according to this embodiment, the abnormality detection device 2 further acquires a detection signal from the second vibration sensor 14 provided in the mechanical equipment 10 . The second vibration sensor 14 is, like the first vibration sensor 11, an accelerometer, a microphone, or the like.

For example, the machine 10 has a power unit 12 and gears 13 . The power plant 12 drives the machinery 10 . The gear 13 changes the rotation speed (frequency) of the power unit 12 and transmits it to the mechanical equipment 10 . As shown in FIG. 3, the second vibration sensor 14 measures vibration of the power plant 12, for example.

(Functional configuration of abnormality detection device)
Further, as shown in FIG. 3, the processor 21 of the abnormality detection device 2 according to this embodiment has a frequency detection section 215 as a functional section instead of the frequency point designation section 214 . The frequency detection unit 215 detects a characteristic frequency representing the characteristics of vibration of the mechanical equipment 10 .

The measured value acquisition unit 210 acquires the Fourier coefficient at the characteristic frequency detected by the frequency detection unit 215 for each detection signal of the first vibration sensor 11 .

(Processing of anomaly detection device)
FIG. 4 is a diagram illustrating an example of processing of the abnormality detection device according to the second embodiment of the present disclosure.
The abnormality diagnosis processing of the abnormality detection device 2 will be described in detail below with reference to FIG.

The feature of this embodiment is that the abnormality detection device 2 detects a characteristic frequency (characteristic frequency) of the mechanical equipment 10 to be diagnosed. The characteristic frequency is, for example, the number of rotations of the power unit 12 of the mechanical equipment 10 .

In the conventional FFT, the frequency points to be subjected to frequency analysis are discrete values depending on the number of data points N of the detection signal and the time interval ΔT of measurement. That is, in FFT, if the time interval of the detection signal is ΔT [s] and the number of measurement points to be subjected to FFT is N, the frequency resolution is Δf = (ΔTN) ^-1 [Hz], and the Fourier coefficient is a multiple of the frequency resolution are calculated for the frequency points {0Δf, 1Δf, 2Δf, . . . , (N−1)Δf}. As in the above description, ΔT=0.01 [s] and N=1024 will be described as an example. At this time, the frequency resolution Δf is approximately 0.098 Hz. For example, suppose that the characteristic component of the vibration of the measured value appears at 1.025 Hz. Since the frequency points are constrained to multiples of the resolution in the FFT, the closest to 1.025 Hz are the 11th frequency point, 0.98 Hz, and the 12th frequency point, 1.07 Hz. In such a case, it is easy to substitute the frequency characteristic of 0.98 Hz or 1.025 Hz for the frequency characteristic of 1.025 Hz. However, with such treatment, the 1.025 Hz vibration is bisected into a 0.98 Hz component and a 1.07 Hz component. As a result, the FFT cannot correctly measure the characteristic frequency component of 1.025 Hz.

In contrast, the abnormality detection device 2 according to the first embodiment has the advantage that the frequency points can be continuously set independently of the frequency resolution Δf. But what if the dominant frequency component is not constant at 1.025 Hz, but fluctuates over time, say between 0.9 Hz and 1.1 Hz? Since the characteristics of the actual mechanical equipment 10 often change due to operating conditions and deterioration over time, it is valuable to deal with changes in dominant frequency components.

To cope with this, the abnormality detection device 2 according to the present embodiment has a frequency detection section 215 instead of the frequency point designation section 214 according to the first embodiment. The frequency detection unit 215 detects a characteristic frequency (characteristic frequency) of the mechanical equipment 10 by performing processing described below.

First, the frequency detection unit 215 measures a signal z for detecting characteristic vibrations, and Fourier-transforms it using equations (4A) and (4B).

Specifically, z includes vibration of the power unit 12 such as an electric motor. For example, considering four detection signals as in the first embodiment, the signal z _n is any of {z ₁ , z ₂ , z ₃ , z ₄ }. The elements of z may be outputs of separate vibration sensors 14 or may be branched output signals of the same vibration sensor 14 .

Here, θ _n in equations (4A) and (4B) is adjusted by a PI controller as in equation (5), for example, so as to match the phase of the characteristic frequency component of _zn . In equation (5), atan is the inverse of the tangent function, atan(dn/ _cn ) is an approximation of the phase [rad] of the characteristic frequency component of _zn , and atan( _d The approximation error becomes zero when _n /c _n )=0.

The frequency detection unit 215 outputs _ωn obtained by Equation (5) as a signal representing the characteristic frequency of _zn .

The measured value acquisition unit 210 uses ω _n obtained from the frequency detection unit 215 to obtain the phase angle θ _n used for Fourier transform from Equation (6).

Also, the measured value acquiring unit 210 determines m _n and r _n so as to satisfy equation (7) instead of equation (2) in the first embodiment.

Then, the measured value acquisition unit 210 calculates the Fourier coefficients a _n and b _n of the detection signal y _n using equations (8A) and (8B).

The processing after obtaining the Fourier coefficients a _n and b _n is the same as in the first embodiment.

(Effect)
As described above, the abnormality detection device 2 according to the present embodiment detects the characteristic frequency ωn of the mechanical equipment 10 based on the detection signal _zn acquired from the second vibration sensor 14, and detects the detected characteristic frequency _ω Obtain a measurement (Fourier coefficients a _n , b _n ) consisting of the amplitude and phase of the oscillation at _n .

By doing so, the abnormality detection device 2 can automatically detect the frequency component that changes according to the operating conditions of the mechanical equipment 10, deterioration over time, and the like. As a result, the abnormality detection device 2 can acquire more accurate Fourier coefficients, thereby further improving the accuracy of abnormality detection.

<Third Embodiment>
Next, an abnormality detection device 2 according to a third embodiment of the present disclosure will be described with reference to FIGS. 5 and 6. FIG.
Components common to those of the first and second embodiments are denoted by the same reference numerals, and detailed description thereof is omitted.

(Functional configuration of abnormality detection device)
FIG. 5 is a diagram showing a functional configuration of an abnormality detection device according to a third embodiment of the present disclosure;
As shown in FIG. 5, the processor 21 of the abnormality detection device 2 according to this embodiment further has an estimation section 216 . The estimation unit 216 estimates the characteristic frequency of each measurement point of the mechanical equipment 10 based on the detection signal acquired from the second vibration sensor 14 and the vibration frequency model of the mechanical equipment 10 .

In the mechanical equipment 10, for example, the rotational speed of the power unit 12 and its integral multiple harmonics appear as characteristic frequencies. In addition to the number of rotations of the power unit 12, features such as a natural frequency determined by the rigidity and mass of the mechanical equipment 10, and a resonance frequency determined by the density and compressibility of the fluid in the piping appear. These vibrations are not independent for each location of the mechanical equipment 10 but work together. The linkages most likely to occur are those that vibrate at the same frequency. Alternatively, a value obtained by multiplying the characteristic frequency of a certain site by an integer may appear at another location.

The anomaly detection device 2 according to the second embodiment inputs a plurality of Fourier coefficients, and compares how the plurality of Fourier coefficients are interlocked at each time with a predetermined normal state interlocking method using the Mahalanobis distance scale. to detect whether there is an abnormality. At this time, if there is an error in the frequency at which the Fourier coefficient is evaluated, the calculation result of the Fourier coefficient includes an error, which may make it difficult to correctly detect an abnormality.

For example, it is known that four measurement points have the property of vibrating at the same frequency around 1 Hz, and {1.07 Hz, 1.07 Hz, 0.98 Hz, 1.07 Hz} is output from the frequency detection unit 215 Let's say it was Using these frequencies to evaluate the Mahalanobis distance, the third Fourier coefficient has a different frequency and is not matched. In such a case, priority should be given to vibrating at the same frequency at each measurement point, and all should be Fourier transformed at a common frequency {1.07 Hz, 1.07 Hz, 1.07 Hz, 1.07 Hz}. . This embodiment is a technique for doing that.

(Processing of anomaly detection device)
FIG. 6 is a diagram illustrating an example of processing of an abnormality detection device according to the third embodiment of the present disclosure;
The abnormality diagnosis processing of the abnormality detection device 2 will be described in detail below with reference to FIG.

For example, if there are four detection signals (four measurement points), the interlocking method is expressed by Equation (10). where A ₁ , A ₂ , A ₃ , A ₄ are column vectors of size 3×1, respectively.

For example, if one method of interlocking these measurement points is determined, if any _one of _ω1 to ω4 is known, the others can be estimated. A case of estimating ω ₂ , ω ₃ and ω ₄ from ω ₁ will be described. Denoting the frequency estimates by the symbols {̂ω ₁ ,̂ω ₂ ,̂ω ₃ ,̂ω ₄ }, the values are calculated by equation (11). Equation (11) is a vibration frequency model representing the relationship between the characteristic frequencies representing the vibration characteristics of each measurement point of the mechanical equipment 10 .

As shown in FIG. 5, the frequency detection unit 215 according to the present embodiment acquires the detection signal z1 from the second vibration sensor 14 provided at _one measurement point (for example, the power unit 12) of the mechanical equipment 10. to detect the characteristic frequency _ω1 . Then, the estimation unit 216 estimates the characteristic frequencies ω ₂ , ω ₃ , and ω ₄ of the other measurement points from ω ₁ detected by the frequency detection unit 215 using Equation (11).

In this way, the estimated values of the characteristic frequencies ω ₂ , ω ₃ , and ω ₄ at each measurement point are values linked to the characteristic frequency ω ₁ detected by the frequency detection section 215 . This makes it possible to eliminate errors in the frequency detector according to the second embodiment. The phase angle θ _n used for the Fourier transform is determined by Equation (12), and the quantities (m _n , r _n ) for integral calculation are determined so as to satisfy Equation (13).

Then, the measured value acquisition unit 210 calculates the Fourier coefficients a _n and b _n of the detection signal y _n using the above-described equations (8A) and (8B), as in the second embodiment. Also, the processing after obtaining the Fourier coefficients a _n and b _n is the same as in the first and second embodiments.

(Effect)
As described above, the anomaly detection device 2 according to the present embodiment uses the vibration frequency model and the characteristic frequency ω1 detected at _one measurement point to determine the characteristic frequency Estimate ω ₂ , ω ₃ , ω ₄ and obtain measurements (Fourier coefficients a _n , b _n ) consisting of the amplitude and phase of the oscillation at the estimated characteristic frequencies.

By doing so, the abnormality detection device 2 can suppress calculation errors of the Fourier coefficients caused by measurement errors of the vibration sensor or the like. Thereby, the abnormality detection device 2 can further improve the accuracy of abnormality detection.

<Fourth Embodiment>
Next, an abnormality detection device 2 according to a fourth embodiment of the present disclosure will be described with reference to FIGS. 7 and 8. FIG.
Components common to those of the first to third embodiments are denoted by the same reference numerals, and detailed description thereof is omitted.

The abnormality detection device 2 according to the first to third embodiments described above calculates the Mahalanobis distance of the measurement value composed of the Fourier coefficients acquired by the measurement value acquisition unit 210, and when the Mahalanobis distance exceeds the threshold, the mechanical equipment 10 were determined to be abnormal. However, the method for detecting an abnormality in the mechanical equipment 10 is not limited to the Mahalanobis distance. In this embodiment, a mode will be described in which the anomaly detection device 2 detects the presence or absence of an anomaly based on the component of the measurement value vector instead of the Mahalanobis distance.

(Functional configuration of abnormality detection device)
FIG. 7 is a diagram showing a functional configuration of an abnormality detection device according to a fourth embodiment of the present disclosure;
As shown in FIG. 7 , in this embodiment, the processor 21 of the abnormality detection device 2 has a component calculator 217 instead of the Mahalanobis distance calculator 212 .

The component calculation unit 217 calculates a plurality of components by resolving the measurement value vector composed of the measurement values acquired by the measurement value acquisition unit 210 in the direction of the index vector determined as the abnormality index. In this embodiment, the index vector is a singular vector obtained by performing singular value decomposition of the unit space.

(Processing of anomaly detection device)
FIG. 8 is a diagram illustrating an example of processing of an abnormality detection device according to the fourth embodiment of the present disclosure;
The abnormality diagnosis processing of the abnormality detection device 2 will be described in detail below with reference to FIG.

The measured value acquisition unit 210 according to this embodiment performs the same processing as in the first to third embodiments, and acquires Fourier coefficients (measured values) at predetermined frequency points of each measurement point. In the example of FIG. 8, the measured value acquisition unit 210 performs the same processing as in the third embodiment.

The Fourier coefficients acquired by the measured value acquisition unit 210 are represented as a vector by a Fourier coefficient vector x (measured value vector). Equation (14) is for the case where the number of Fourier coefficients is M and the size of the Fourier coefficient vector x is M×1.

The unit space generator 211 generates a unit space using the normal state values of the Fourier coefficient vector x. The unit space is the covariance matrix Q of the Fourier coefficient vector x obtained when the mechanical equipment is in a normal state, and is represented by the following Equations (15) and (16). Note that X ₀ in equation (15) is a matrix in which a plurality of Fourier coefficient vectors x are arranged in the column direction. L is the number of Fourier _coefficient vectors used to create the unit space, and X0 is a complex matrix of size M×L. X ₀ ^* in formula (16) represents the conjugate transpose of X ₀ .

Here, it is assumed that the unit space generator 211 has already generated the unit space based on the Fourier coefficient vector x acquired at the past time, and the unit space is stored in the auxiliary storage device 24. .

In the first to third embodiments, the abnormality detection device 2 uses the Mahalanobis distance calculator 212 to calculate the Mahalanobis distance MD at time t using the following equation (17).

In contrast, the component calculator 217 of the anomaly detection device 2 according to the present embodiment calculates the components ρ _k (k=1, 2, . . . , M) of the Fourier coefficient vector x. Specifically, the component calculator 217 calculates the component ρ _k in the following procedure.

The unit space can be singular value decomposed as in Equation (18) below.

Here, M is the number of first vibration sensors 11 . u _k (k=1, 2, . . . , M) are singular vectors, of size M×1, and generally column vectors whose elements are complex numbers. σ _k (k=1, 2, . . . , M) is a singular value and a non-negative real number.

The component calculation unit 217 calculates components ρ _k (k=1, 2, . . . , M) obtained by resolving the Fourier coefficient vector x in the direction of singular vectors uk ( _k =1, 2, . Obtained using (19).

Note that the component calculation unit 217 according to this modification outputs the component ρ _{k obtained} by decomposing the Fourier coefficient vector x in the direction of the singular vector u _k to the determination unit 213 . Note that component calculation section 217 may output the absolute value of component ρ _k to determination section 213 . The absolute value of component ρ _k is represented by the following equation (20).

The determination unit 213 compares the absolute value of the component _{ρk obtained} by decomposing the Fourier coefficient vector x in the direction of the singular vector _uk with the square root of the singular value _σk corresponding to the singular vector _uk for each singular vector. perform contrast processing. Since the unit space Q is the covariance matrix of a plurality of Fourier coefficient vectors x obtained when the mechanical equipment 10 is in the normal state, its singular value σ _k is the absolute value of the component ρ _k in the normal state. represents the variation of the squared value of . Therefore, if the mechanical equipment 10 is normal at a certain time, the expected value of ρ _k ρ _k ^* is close to the singular value σ _k . That is, if the expected value is represented by E, it becomes Formula (21).

ρ _k ρ _k ^* is a component of the Mahalanobis distance as shown in Equation (22).

Therefore, abnormality determination based on the Mahalanobis distance is based on the sum of the components on the right side of Equation (22). In the following, a method of determining abnormality by checking values at the stage of components prior to summing will be described.

Based on such a premise, the determination unit 213 compares the absolute value of the component ρ _k and the square root of the singular value σ _k for each singular vector u _k (k=1, 2, . . . , M), and Output the judgment result.

For example, the determination unit 213 outputs a determination result indicating normal when the absolute value of the component ρ _k is less than the threshold value based on the singular value σ _k . On the other hand, the determination unit 213 outputs a determination result indicating abnormality when the absolute value of the component _ρk is equal to or greater than the threshold. Specifically, the determination unit 213 determines that there is an abnormality if the following equation (23) holds.

Here, γ _1k and γ _2k in Equation (23) are positive constants determined for each singular vector u _k (k=1, 2, . . . , M).

7 and 8 show an example in which a component calculation unit 217 is added in place of the Mahalanobis distance calculation unit 212 of the third embodiment, but the present invention is not limited to this. In another embodiment, a component calculator 217 may be added instead of the Mahalanobis distance calculator 212 of the first or second embodiment.

(Effect)
As described above, the abnormality detection apparatus 2 according to the present embodiment obtains the measurement value vector (Fourier coefficient vector x) obtained by the measurement obtained by the measurement value obtaining unit 210 by singular value decomposition of the unit space. is calculated, and when the absolute value of the calculated ρ exceeds a predetermined threshold value, it is determined that the mechanical equipment 10 has an abnormality.

By doing so, the anomaly detection device 2 can monitor the presence or absence of an anomaly for each component instead of the total Mahalanobis distance of the components, so that the sensitivity to anomalies can be improved. In addition, as in the first to third embodiments, the abnormality detection device 2 only needs to calculate the Fourier coefficient at the characteristic frequency of the mechanical equipment 10, so that the abnormality can be diagnosed without separately preparing a frequency analysis device. It can be done at high speed.

<Fifth Embodiment>
Next, an abnormality detection device 2 according to a fifth embodiment of the present disclosure will be described with reference to FIG.
Components common to those of the first to fourth embodiments are denoted by the same reference numerals, and detailed description thereof is omitted.

The abnormality detection device 2 according to the first to third embodiments described above calculates the Fourier coefficient for each first vibration sensor 11 . That is, equations (3A) and (3B) are used in the abnormality detection device 2 according to the first embodiment, and equations (8A) and (8B) are used in the abnormality detection devices 2 according to the second and third embodiments. , the Fourier coefficient is calculated for each first vibration sensor 11 . According to such a method, the amount of computation for calculating Fourier coefficients increases as the number of sensors (measurement points) increases.

On the other hand, the anomaly detection device 2 according to the present embodiment is characterized by using prior knowledge of anomalies to reduce the calculation of Fourier coefficients. This has practical advantages, such as reducing the cost of the computer.

(Processing of anomaly detection device)
FIG. 9 is a diagram illustrating an example of processing of an abnormality detection device according to the fifth embodiment of the present disclosure;
The abnormality diagnosis processing of the abnormality detection device 2 will be described in detail below with reference to FIG.

As shown in FIG. 9, the measured value acquisition unit 210 according to the present embodiment obtains a measured value vector {y ₁ , y ₂ , . . . , y _M }.

The component calculation unit 217 divides the singular vector _uk determined as an index of abnormality into a real part and an imaginary part using a phase angle θ of a predetermined frequency point (characteristic frequency ω), and calculates the real part and the imaginary part, respectively. Calculate the inner product with the measured value vector. Note that FIG. 9 shows an example in which the characteristic frequency ω estimated by the frequency detection unit 215 and the estimation unit 216 according to the third embodiment is input to the component calculation unit 217, but is limited to this. never. The component calculation unit 217 may receive the period T _n specified by the frequency point specifying unit according to the first embodiment, or the characteristic frequency ω _n detected by the frequency detection unit 215 according to the second embodiment. may be entered.

The abnormality diagnosis process executed by the abnormality detection device 2 according to this embodiment is a further improvement of the process described in the fourth embodiment. In the fourth embodiment, the abnormality is determined by Equation (23) based on the component ρ _{k obtained} by decomposing the Fourier coefficient vector x in the direction of the singular vector u _k (k=1, 2, . . . , M). Abnormalities of the mechanical equipment 10 include those with high frequency and those with low frequency. Since the type of anomaly can be associated with a singular vector u _k (k=1, 2, . There are things that are not. Then, it is rational to determine whether only the components that are likely to conflict are abnormal at each time step, and to determine whether the other components are abnormal by thinning out once every 100 time steps, for example, when there is a constraint on the computing power.

The present embodiment is a technology that makes it possible to perform such an abnormality determination regarding _one specific _singular vector at _high speed. } On the other hand, M times of Fourier transformation is reduced to one time. However, as described in the third embodiment, the characteristic frequencies at which the M vibration characteristics appear are related to each other, and the second to M-th characteristic frequencies are known from the first characteristic frequency of vibration. In the following, for simplicity of explanation, it is assumed that all characteristic frequencies are the same as in Equation (24).

Therefore, the estimator 216 estimates the characteristic frequency of each measurement point from the characteristic frequency ω1 detected by the frequency detector 215 using Equation (24).

Then, the case where the abnormality detection device 2 determines abnormality with respect to the _k -th singular vector uk will be described. k is any one value of {1, 2, . . . , M}.

The component calculator 217 calculates the phase angle θ of the characteristic frequency of each measurement point (first vibration sensor 11) using Equation (12) as in the third embodiment. The component calculation unit 217 divides the singular vector u _k into a real part and an imaginary part, and, as shown in equations (25A) and (25B), the measurement values {y1, _y2 , . Compute the inner product. Here, the symbol [u _k ] ₁ represents the first element of the singular vector u _k .

Next, the component calculation unit 217 time-averages the inner product for one cycle as shown in Equation (26). Since the characteristic frequencies of M vibrations are equal to each other as shown in Equation (24), the length of _one cycle is T1. If there is a characteristic frequency with a different value among the M vibrations, the length of one cycle is calculated as the least common multiple of the cycles.

Apply ρ _k obtained in equation (26) to equation (20) above. Also, the determination unit 213 determines an abnormality using the above equation (23).

In the calculation of formula (26), {α(t−ΔT), α(t−2ΔT), . . . , α(t−m ₁ ΔT)} and {β(t−ΔT), β( t−2ΔT), . . . , β(t−m ₁ ΔT)}. The abnormality detection device 2 according to the fourth embodiment had to store a total of M×2m ₁ values. On the other hand, the abnormality detection device 2 according to the present embodiment only needs to store _a total of 2m1 past values. Therefore, according to this embodiment, it is possible to further save the amount of storage of past values and the sum-of-products operation. Thereby, for example, the function of the abnormality detection device 2 can be incorporated into a normal control device and executed.

Note that Equation (26) may be replaced with a first-order low-pass filter. This allows the number of past values to be two. In this case, it takes several _times longer than T1 for the low-pass filter to settle, and it takes a long time to determine whether there is an abnormality. For this reason, the method using equation (26) as described above is more effective for applications that require quick response.

(Effect)
As described above, the abnormality detection device 2 according to the present embodiment divides the singular vector u _k into the real part and the imaginary part using the phase angle θ of the characteristic frequency ω, and the real part and the imaginary part and the measured value An inner product with the vector y is calculated, and if the absolute value of the time average ρ _k for one cycle of the calculated inner product exceeds a predetermined threshold, it is determined that the mechanical equipment 10 has an abnormality.

By doing so, the abnormality detection device 2 can reduce the number of times of Fourier transform for each diagnosis to one, so that the computation load can be significantly reduced. In addition, since the abnormality detection device 2 can reduce the past values used for abnormality diagnosis, it is possible to reduce the amount of storage and the load of sum-of-products calculation.

<Modification>
In the above-described fourth and fifth embodiments, abnormality is determined based on the singular vector uk _obtained by singular value decomposition of the unit space Q regarding the Mahalanobis distance. However, if the abnormal mode shape of the vibration of the mechanical equipment 10 is known in advance, the abnormality can be determined as soon as the mode shape appears. This modification realizes it. Here, an example in which this modified example is applied to the abnormality detection device 2 according to the fifth embodiment will be described.

In the abnormality detection device 2 according to this modification, the component calculator 217 uses a predetermined mode shape vector φ as an index vector instead of the singular vector _uk . That is, equations (27A) and (27B) are used instead of equations (25A) and (25B) of the fifth embodiment. Note that the mode shape vector φ can be obtained using a known technique such as modal analysis.

Then, the component calculation unit 217 calculates the component ρ _k of the mode shape vector by Equation (28).

In the equation (23) in which the determination unit 213 determines abnormality, σ _k is determined by the equation (27) from α _k and β _k during normal operation.

(Effect)
As described above, the abnormality detection device 2 according to this modification uses the mode shape vector φ as the index vector.

By doing so, the abnormality detection device 2 can omit the calculation for obtaining the singular vector _uk , so that the calculation load can be further reduced.

In the above-described first to fifth embodiments and modifications thereof, the various processes of the abnormality detection device 2 described above are stored in a computer-readable recording medium in the form of a program. are read out and executed by the computer to perform the various processes described above. Computer-readable recording media include magnetic disks, magneto-optical disks, CD-ROMs, DVD-ROMs, semiconductor memories, and the like. Alternatively, the computer program may be distributed to a computer via a communication line, and the computer receiving the distribution may execute the program.

The program may be for realizing part of the functions described above. Furthermore, it may be a so-called difference file (difference program) that can realize the above functions by combining with a program already recorded in the computer system.

Further, in another embodiment, the control device 3 may have each functional part of the abnormality detection device 2 described in the first to fifth embodiments and their modifications.

Although several embodiments of the present invention have been described above, all these embodiments are provided by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. These embodiments and their modifications are included in the gist and technical scope of the invention.

<Appendix>
The abnormality detection device, abnormality detection method, and program described in the above-described embodiments are understood as follows, for example.

(1) According to the first aspect, the abnormality detection device (2) that detects the presence or absence of an abnormality in the mechanical equipment (10) outputs from the first vibration sensor (11) that measures the vibration of the mechanical equipment (10) A process of acquiring measured values (a, b) consisting of the amplitude and phase of vibration at a predetermined frequency point based on the detected signal (y), and a plurality of measured values acquired at a plurality of past points A process of calculating the Mahalanobis distance of the measurement value acquired at the time of evaluating the mechanical equipment (10) based on the unit space configured by 10) and a processor (21) for determining that an abnormality has occurred.

By doing so, the anomaly detection apparatus needs to calculate only Fourier coefficients (amplitude and phase) at specific frequency points, so that the computational load can be significantly reduced compared to conventional techniques. As a result, the abnormality detection device can perform abnormality diagnosis of mechanical equipment at high speed. Moreover, since the abnormality detection device can calculate Fourier coefficients by itself without separately preparing a frequency analysis device, the cost of the entire abnormality detection system can be reduced.

(2) According to the second aspect, in the abnormality detection device (2) according to the first aspect, the processor (21) outputs from the second vibration sensor (14) that measures the vibration of the mechanical equipment (10) Based on the detected signal, a process of detecting the characteristic frequency (ω) of the mechanical equipment (10) is further executed, and in the process of acquiring the measured value, the detected characteristic frequency (ω) is set as a frequency point and measured. Get the value (a,b).

By doing so, the abnormality detection device can automatically detect the frequency component that changes according to the operating conditions of the mechanical equipment, aged deterioration, and the like. As a result, the anomaly detection device can acquire more accurate Fourier coefficients, thereby further improving the accuracy of anomaly detection.

(3) According to the third aspect, in the anomaly detection device (2) according to the second aspect, the processor (21) generates a vibration signal representing the relationship of characteristic frequencies at a plurality of measurement points of the mechanical equipment (10). Based on the frequency model and the characteristic frequency (ω ₁ ) of one measurement point of the mechanical equipment (10) detected in the process of detecting the characteristic frequency, the characteristic frequency (^ω _n ) of each of the plurality of measurement points is calculated as follows: In the process of further executing the estimation process and acquiring the measured values, the estimated characteristic frequencies (^ω _n ) of each of the plurality of measurement points are set as frequency points, and the measured values corresponding to each of the plurality of measurement points are obtained. Get (a,b).

By doing so, the abnormality detection device can suppress calculation errors of the Fourier coefficients caused by measurement errors of the vibration sensor or the like. Thereby, the abnormality detection device can further improve the accuracy of abnormality detection.

(4) According to the fourth aspect, the abnormality detection device (2) for detecting the presence or absence of abnormality in the mechanical equipment (10) outputs from the first vibration sensor (11) for measuring the vibration of the mechanical equipment (10) Based on the detected signal, a process of acquiring measured values (a, b) consisting of the amplitude and phase of vibration at a predetermined frequency point, and a measured value vector consisting of the measured values is defined as an index of abnormality A process of calculating a plurality of components (ρ _k ) resolved in the direction of the vector (uk or φ), and when the absolute value of the calculated component ( _{ρ k )} exceeds a predetermined threshold, the mechanical equipment (10) and a processor for determining that an abnormality has occurred.

By doing so, the abnormality detection device can monitor the presence or absence of abnormality for each component, so that the sensitivity to abnormality can be improved. In addition, as in the first to third aspects, the abnormality detection device needs to calculate only the Fourier coefficient at the characteristic frequency of the mechanical equipment, so the abnormality diagnosis can be performed at high speed without separately preparing a frequency analysis device. It is possible.

(5) According to the fifth aspect, the abnormality detection device (2) for detecting the presence or absence of an abnormality in the mechanical equipment (10) is measured by the first vibration sensor (11) for measuring the vibration of the mechanical equipment (10). A process of acquiring a measured value vector (y) composed of the measured values obtained, and an index vector ( _uk or φ) determined as an index of abnormality is obtained by using the phase angle (θ) of the predetermined frequency point to obtain the real part and Divided into imaginary parts, a process of calculating the inner product of each of the real part and the imaginary part and the measured value vector (y), and the absolute value of the time average (ρ _k ) for one cycle of the calculated inner product is a predetermined threshold and a processor for determining that an abnormality has occurred in the mechanical equipment (10) when it exceeds.

By doing so, the abnormality detection device can reduce the number of times of Fourier transform for each diagnosis to one, so that the computational load can be significantly reduced. In addition, since the abnormality detection device can reduce the past values used for abnormality diagnosis, it is possible to reduce the amount of storage and the load of sum-of-products calculation.

(6) According to the sixth aspect, in the abnormality detection device (2) according to the fourth or fifth aspect, the processor (21) is composed of a plurality of measured values obtained at a plurality of points in the past. A singular vector (u _k ) obtained from a unit space is used as an index vector.

By doing so, the anomaly detection device can monitor the presence or absence of an anomaly for each component instead of the Mahalanobis distance that is the sum of the components, so that it is possible to improve the sensitivity to anomalies.

(7) According to the seventh aspect, in the abnormality detection device (2) according to the fourth or fifth aspect, the processor (21) indexes the mode shape vector (φ) specific to the mechanical equipment (10) Used as a vector.

By doing so, the abnormality detection device 2 can omit the calculation for obtaining the singular vector, so that the calculation load can be further reduced.

(8) According to the eighth aspect, the anomaly detection method for detecting the presence or absence of an anomaly in the mechanical equipment (10) includes the detection output from the first vibration sensor (11) for measuring the vibration of the mechanical equipment (10). A step of obtaining measured values (a, b) consisting of the amplitude and phase of vibration at a predetermined frequency point based on the signal, and a unit space composed of a plurality of measured values obtained at a plurality of past points in time. a step of calculating the Mahalanobis distance of the measured value acquired at the time of evaluating the mechanical equipment (10) with reference to, and if the calculated Mahalanobis distance exceeds a predetermined threshold, an abnormality occurs in the mechanical equipment (10) and determining that

(9) According to the ninth aspect, the program causes the computer of the abnormality detection device (2) for detecting the presence or absence of an abnormality in the machinery (10) to cause the first vibration sensor for measuring the vibration of the machinery (10). (11) based on the detection signal output from; A step of calculating the Mahalanobis distance of the measured value obtained at the time of evaluating the mechanical equipment (10) based on the unit space configured by the measured value of; and a step of determining that an abnormality has occurred in the mechanical equipment (10).

100 Abnormality detection system 10 Mechanical equipment 11, 11A, 11B, 11C, 11D First vibration sensor 12 Power unit 13 Gear 14 Second vibration sensor 2 Abnormality detection device 21 Processor 210 Measurement value acquisition unit 211 Unit space generation unit 212 Mahalanobis distance calculation Unit 213 Determination unit 214 Frequency point designation unit 215 Frequency detection unit 216 Estimation unit 217 Component calculation unit 22 Input/output unit 23 Main storage device 24 Auxiliary storage device 3 Control device

Claims (9)

An abnormality detection device for detecting the presence or absence of an abnormality in mechanical equipment,
a process of acquiring a measurement value consisting of the amplitude and phase of the vibration at a predetermined frequency point based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment;
A process of calculating the Mahalanobis distance of the measured values obtained at the time of evaluating the mechanical equipment with reference to a unit space composed of the plurality of measured values obtained at a plurality of points in the past;
a process of determining that an abnormality has occurred in the mechanical equipment when the calculated Mahalanobis distance exceeds a predetermined threshold;
comprising a processor that executes
Anomaly detector. The processor
further executing a process of detecting a characteristic frequency of the mechanical equipment based on a detection signal output from a second vibration sensor that measures vibration of the mechanical equipment;
obtaining the measured value by setting the detected characteristic frequency to the frequency point in the process of obtaining the measured value;
The abnormality detection device according to claim 1. The processor
A plurality of Further executing a process of estimating the characteristic frequency of each of the measurement points of
In the process of acquiring the measured value, the estimated characteristic frequency of each of the plurality of measurement points is set as the frequency point, and the measured value corresponding to each of the plurality of measurement points is acquired.
The abnormality detection device according to claim 2. An abnormality detection device for detecting the presence or absence of an abnormality in mechanical equipment,
a process of acquiring a measurement value consisting of the amplitude and phase of the vibration at a predetermined frequency point based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment;
A process of calculating a plurality of components by resolving the measured value vector composed of the measured values in the direction of an index vector defined as an index of abnormality;
a process of determining that an abnormality has occurred in the mechanical equipment when the calculated absolute value of the component exceeds a predetermined threshold;
comprising a processor that executes
Anomaly detector. An abnormality detection device for detecting the presence or absence of an abnormality in mechanical equipment,
a process of acquiring a measurement value vector composed of measurement values measured by a first vibration sensor that measures vibration of the mechanical equipment;
a process of dividing an index vector determined as an abnormality index into a real part and an imaginary part using a phase angle of a predetermined frequency point, and calculating an inner product of each of the real part and the imaginary part and the measurement value vector; ,
a process of determining that an abnormality has occurred in the mechanical equipment when the absolute value of the time average for one cycle of the calculated inner product exceeds a predetermined threshold;
comprising a processor that executes
Anomaly detector. The processor uses, as the index vector, a singular vector obtained from a unit space composed of a plurality of the measured values obtained at a plurality of past points in time.
The abnormality detection device according to claim 4 or 5. the processor uses a mode shape vector specific to the machinery as the index vector;
The abnormality detection device according to claim 5. An abnormality detection method for detecting the presence or absence of an abnormality in mechanical equipment,
a step of acquiring a measured value consisting of the amplitude and phase of the vibration at a predetermined frequency point based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment;
A step of calculating the Mahalanobis distance of the measured values obtained at the time of evaluating the mechanical equipment, with reference to a unit space composed of the plurality of measured values obtained at a plurality of points in the past;
determining that an abnormality has occurred in the mechanical equipment when the calculated Mahalanobis distance exceeds a predetermined threshold;
An anomaly detection method comprising: In the computer of the abnormality detection device that detects the presence or absence of abnormality in the mechanical equipment,
a step of acquiring a measured value consisting of the amplitude and phase of the vibration at a predetermined frequency point based on the detection signal output from the first vibration sensor that measures the vibration of the mechanical equipment;
A step of calculating the Mahalanobis distance of the measured values obtained at the time of evaluating the mechanical equipment, with reference to a unit space composed of the plurality of measured values obtained at a plurality of points in the past;
determining that an abnormality has occurred in the mechanical equipment when the calculated Mahalanobis distance exceeds a predetermined threshold;
program to run.

Images