JP2024127720A - Method and system for analyzing vibration data of rotating equipment - Google Patents

Abstract

[Problem] Even in the case of rolling bearings with large noise from rotating equipment, vibrations caused by bearing defects can be extracted, making it easier to perform abnormality diagnosis after the occurrence of a defect. [Solution] An ARMA model (autoregressive moving average model) is estimated for waveform data acquired using an acceleration sensor 1, the frequency response of the estimated ARMA model is calculated, the pass frequency of a bandpass filter is determined based on the peak frequency (Fpeak) of the calculated frequency response, noise is removed from the vibration waveform data using the determined pass frequency, the noise-removed vibration waveform data is envelope-analyzed to calculate the amplitude value of the peak frequency component, and if a tendency for the amplitude value of the peak frequency component to increase in a predetermined period is confirmed, it is determined that there is a possibility of an abnormality in a bearing 2. [Selected Figure] Figure 3

Description

The present invention relates to a method and system for analyzing vibration data from rotating equipment, such as generators, pumps, and fans, to analyze the presence or absence of abnormalities in rolling bearings and the types of bearing failures.

Traditionally, rolling bearings have been used in large numbers in the rotating parts of a wide range of equipment. If an abnormality occurs in these rolling bearings, various inconveniences occur, such as the shutdown of the machinery and equipment. Generally, rolling bearings are highly accurate, and under proper operating conditions, they can be used for long periods of time before fatigue failure occurs due to repeated strain. Therefore, it is important to diagnose the presence or absence of bearing abnormalities in rotating equipment and the type of bearing failure.

Unexpected failures in rolling bearings are caused by mechanical conditions that place stress on the bearing, such as improper lubrication, misalignment of the rotating shaft, or improper installation of the rolling bearing. The majority of failures in rolling bearings are caused by improper lubrication, such as the inclusion of dirt or deterioration of the lubricant. Various methods have been proposed for diagnosing failures in rolling bearings. For example, there is a method that uses an acceleration sensor to measure bearing vibration and issues an alarm if the bearing vibration value exceeds an allowable value. There is also a method that estimates the cause of the failure by analyzing the frequency of bearing vibration. There is also a method that predicts the bearing's lifespan as well as diagnosing abnormalities by predicting the trend of increasing bearing vibration values. Other methods that have been proposed include the shock pulse method and AE (Acoustic Emission) method.

Various technologies have been proposed for diagnosing the presence or absence of bearing abnormalities and the type of bearing failure using vibration acceleration signals for rotating equipment without disassembling the machinery and equipment. For example, Patent Document 1, JP 2018-155494 A, "Bearing Abnormality Diagnosis System and Bearing Abnormality Diagnosis Method," proposes a bearing abnormality diagnosis method that detects the vibration acceleration of a bearing or a bearing housing from a vibration acceleration sensor attached to a bearing or a bearing housing that supports the bearing, performs frequency analysis of a frequency spectrum based on the vibration acceleration signal from the vibration acceleration sensor, diagnoses the presence or absence of a bearing abnormality based on the frequency spectrum, and identifies the type of the bearing abnormality based on the occurrence order pattern of peak frequencies detected from the frequency spectrum.

Patent Document 1 also proposes a bearing abnormality diagnosis method in which a vibration acceleration sensor attached to a bearing or a bearing housing supporting a bearing detects the vibration acceleration of the bearing or bearing housing, performs frequency analysis of a frequency spectrum based on a vibration acceleration signal from the vibration acceleration sensor, diagnoses the presence or absence of a bearing abnormality based on the frequency spectrum, and, if a bearing abnormality is diagnosed, identifies the type of the bearing abnormality based on a crest factor, which is the ratio of the maximum vibration value to the vibration RMS value (root mean square) of the vibration acceleration signal from the vibration acceleration sensor.

As shown in Fig. 8, the bearing abnormality diagnosis system disclosed in Patent Document 1 is a system that includes a vibration acceleration sensor 52 that is attached to a bearing or a bearing housing 51 that supports the bearing and detects the vibration acceleration of the bearing or the bearing housing 51, a vibration analysis unit 53 that performs frequency analysis of the frequency spectrum based on the vibration acceleration signal from the vibration acceleration sensor 52, a first abnormality diagnosis unit 54 that diagnoses the presence or absence of a bearing abnormality based on the frequency spectrum, a diagnosis device 56 having a second abnormality diagnosis unit 55 that identifies the type of bearing abnormality based on the occurrence order pattern of the peak frequency detected from the frequency spectrum, and an output device 57 such as a monitor and an alarm. In this system, the first abnormality diagnosis unit shown in Fig. 9(a) calculates the occurrence order pattern of the peak frequency detected from the frequency spectrum, and the second abnormality diagnosis unit shown in Fig. 9(b) calculates the crest factor, which is the ratio of the vibration maximum value of the vibration acceleration signal to the vibration RMS (Root Mean Square), to determine the type of bearing abnormality, and diagnoses the presence or absence of a bearing abnormality and the type of bearing failure using a single vibration acceleration signal without disassembling the machinery.

An analysis method for diagnosing abnormalities called WCCS (Whitened Cross-Correlation Spectrum) has been proposed. This analysis method involves inputting a signal into a Whitening Filter calculated using AR model estimation as shown in the flow diagram of Figure 10, and diagnosing bearing abnormalities by performing frequency analysis on a waveform (Vω) in which only characteristic pulses are emphasized, and a waveform (Vcc) obtained by multiplying the waveform (Vω) by a waveform (Venv) that has been subjected to a Hilbert transform.

JP 2018-155494 A

In the "Bearing Abnormality Diagnosis System and Bearing Abnormality Diagnosis Method" of Patent Document 1, the acceleration vibration observed in the bearing of a rotating machine generally includes the shaft rotation speed component and its integer multiple, the vibration caused by any abnormality in the bearing, and also miscellaneous noise components. In order to extract the vibration caused by an abnormality in the bearing from among these, it is thought that various analyses should be performed after removing miscellaneous signals such as noise, and then the cause of the bearing abnormality should be identified. However, the bearing abnormality diagnosis method of Patent Document 1 has the problem that it is difficult to identify the cause of the abnormality because no prior noise processing is performed.

Furthermore, Patent Document 1 had the problem that in a noisy acceleration signal, the RMS value becomes large because the signal contains noise in addition to the signal originating from the abnormality, and the crest factor decreases, making it impossible to determine the type of bearing abnormality. For example, there was a problem in that it was impossible to determine what abnormality the crest factor represented.

The WCCS analysis method is an analysis method for diagnosing bearing abnormalities by performing frequency analysis on a waveform (Vω) that emphasizes only characteristic pulses and a waveform (Vcc) that is a combination of the waveform (Vω) and a waveform (Venv) that has been subjected to a Hilbert transform. However, when the amplitude of this characteristic pulse is low, about twice the noise amplitude, the characteristics cannot be emphasized, making it difficult to determine whether there is an abnormality.

The present invention was invented to solve these problems. That is, the object of the present invention is to provide a method and system for analyzing vibration data of rotating equipment that can extract vibrations caused by bearing defects even in the case of bearings that are noisy in rotating equipment, making it easier to diagnose abnormalities after defects occur.

The vibration data analysis method of the present invention is a method for analyzing vibration data of a rotating device for analyzing the presence or absence of an abnormality in a bearing provided in the rotating device, comprising the steps of:
Obtaining waveform data of vibrations generated in the bearing using an acceleration sensor;
An ARMA model is estimated using an autoregressive model and a moving average model for the waveform data of the vibration acquired by the acceleration sensor, and a frequency response of the estimated ARMA model is calculated;
determining a pass frequency of a band pass filter that extracts only a specific frequency band based on the calculated peak frequency of the frequency response of the ARMA model, and removing noise from the waveform data of the vibration acquired by the acceleration sensor using the determined pass frequency;
performing an envelope analysis on the vibration waveform data from which noise has been removed, and calculating an amplitude value of a peak frequency component obtained by the envelope analysis;
The method is characterized in that, using the amplitude value of the peak frequency component calculated from the vibration waveform data measured over a predetermined period, if a tendency for the amplitude value of the peak frequency component to increase over time is confirmed, it is determined that there is a possibility of an abnormality in the bearing.
The method for analyzing vibration data of a rotating device may include determining the passing frequency based on a peak frequency when a peak frequency of a frequency response of the ARMA model is within a predetermined range for a natural frequency calculated based on a specification of the bearing.
Furthermore, the method for analyzing vibration data of a rotating machine may include calculating a slope of the amplitude value of the peak frequency component in a first period of the specified time period and in a second period immediately before the first period, and determining that there is a possibility of an abnormality in the bearing if the slope in the second period is larger.

The vibration data analysis system of the present invention is a vibration data analysis system for a rotating device that analyzes the presence or absence of an abnormality in a bearing provided in the rotating device,
an acceleration sensor for acquiring waveform data of vibrations generated in the bearing;
an ARMA model estimation unit that estimates an ARMA model using an autoregressive model and a moving average model for the vibration waveform data acquired by the acceleration sensor, and calculates a frequency response of the estimated ARMA model;
a band-pass filter unit that determines a pass frequency of a band-pass filter that extracts only a specific frequency band based on the calculated peak frequency of the frequency response of the ARMA model, and removes noise from the waveform data of the vibration acquired by the acceleration sensor using the pass frequency;
an envelope processing unit that performs an envelope analysis on the vibration waveform data from which noise has been removed by the band pass filter unit, and calculates an amplitude value of a peak frequency component obtained by the envelope analysis;
and an abnormality determination unit which, using the amplitude value of the peak frequency component calculated in the envelope processing unit for the vibration waveform data measured over a predetermined period, determines that there is a possibility of an abnormality in the bearing if a tendency for the amount of increase in the amplitude value of the peak frequency component over time to be increasing is confirmed.

The vibration data analysis method of the present invention can extract vibrations caused by defects even in the case of bearings (2) with high noise levels, making it easier to perform abnormality diagnosis after a defect occurs in a rotating device. Abnormality diagnosis is possible by setting only the bearing information and rotation speed information of the bearing (2).

The replacement period or timing of the bearings (2) installed in the rotating equipment is determined, enabling more efficient maintenance of the rotating equipment. For example, for rotating equipment in a power plant, a plan can be easily made to replace the bearings (2) in autumn, avoiding the summer when the operating rate is high. In addition, for equipment that previously required regular inspections, early detection of abnormalities is now possible, making it easier to lengthen the inspection period and collect deterioration data, and streamlining the transition from the conventional regular inspection system to a condition-based maintenance system in which maintenance is performed according to the condition of the rotating equipment.
In addition, the compact device can easily and accurately estimate the lifespan of a bearing (2) installed in a rotating device with a high degree of precision, and by connecting the device to an Internet line, abnormalities in a specific bearing (2) can be easily determined from a remote location.

1 is a schematic block diagram showing a configuration of a vibration data analysis system according to the present invention. FIG. 2 is a flow chart illustrating an operation procedure of the vibration data analysis method of the present invention. FIG. 2 is a first half of a flow chart illustrating an operation procedure for determining that “anomaly exists” in the vibration data analysis method of the present invention. FIG. 4 is a schematic diagram showing an example of a peak frequency in the vibration data analysis method of the present invention. FIG. 2 is a schematic diagram showing an example of an envelope analysis result in the vibration data analysis method of the present invention. FIG. 2 is a schematic diagram showing a frequency distribution of defect frequencies in the vibration data analysis method of the present invention. FIG. 2 is a schematic diagram showing an example of an increase (slope) in the amplitude value of a defect frequency in the vibration data analysis method of the present invention. FIG. 1 is a schematic block diagram showing the configuration of a conventional bearing abnormality diagnosis system. 1A and 1B are flow diagrams illustrating the operation procedures of a conventional bearing abnormality diagnosis system, in which (a) shows the operation procedures of a first abnormality diagnosis unit, and (b) shows the operation procedures of a second abnormality diagnosis unit. FIG. 1 is a flow chart showing a WCCS (Whitened Cross-Correlation Spectrum) abnormality diagnosis analysis method.

The present invention is a method for analyzing vibration data of rotating equipment that determines and analyzes the presence or absence of abnormalities in bearings installed in rotating equipment such as agitators, pumps, and fans.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
<Configuration of Rotating Equipment Vibration Data Analysis System>
FIG. 1 is a block diagram showing an example of a vibration data analysis device constituting a vibration data analysis system for a rotating machine according to the present invention.
The vibration data analysis system for rotating equipment of the present invention is a system mainly comprising an acceleration sensor 1, an ARMA model estimation unit 3, a band pass filter (BPF) unit 4, an envelope processing unit 5, a database 6, an abnormality determination unit 7, and a BPF change determination unit 8. Using this vibration data analysis system, the presence or absence of an abnormality in a bearing 2 provided in the rotating equipment and the type of bearing failure are analyzed.

An acceleration sensor 1 acquires waveform data of vibrations generated in a bearing 2 provided in a rotating device. An ARMA model estimation unit 3 performs ARMA model (autoregressive moving average model) estimation on the waveform data acquired by this acceleration sensor 1, and calculates the frequency response of the estimated ARMA model. Data related to the natural frequency calculated by this ARMA model estimation unit 3 is stored in a database 6.
The estimation of the ARMA model in the ARMA model estimating unit 3 is a method of estimating a numerical model using an autoregressive model (AR model) and a moving average model (MA model), and existing technology can be used. For example, by using a command (e.g., armax) set in a program of MATLAB (registered trademark) by MathWorks (registered trademark), vibration waveform data can be input to calculate each parameter of the ARMA model.

Existing technology can also be used to calculate the frequency response of the ARMA model. For example, a command (e.g., freqz) for the frequency response of a digital filter set in MATLAB (registered trademark) can be used to calculate the frequency response of an ARMA model in which each parameter has been estimated. This makes it possible to calculate a frequency response in which the horizontal axis represents frequency and the vertical axis represents amplitude, and to obtain data related to the natural frequency, such as by calculating the peak frequency of the frequency response as the natural frequency.
The BPF change determination unit 8 determines whether or not to change the pass frequency estimated by the ARMA model estimation unit 3 to a band pass filter (BPF) unit 4 in order to remove noise.

A band pass filter (BPF) unit 4 removes noise from the waveform data collected by the acceleration sensor 1 .
The envelope processing unit 5 analyzes the periodicity of the repeated vibration of the bearing 2 for the waveform data from which noise has been removed by the band pass filter (BPF) unit 4. The envelope processing unit 5 also stores data on defect frequencies calculated from the analysis results in a database 6.

The abnormality determination unit 7 determines whether or not there is an abnormality in the bearing 2 using the natural frequency calculated by the ARMA model estimation unit 3 and the defect frequency analyzed by the envelope processing unit 5, using the data (vibration data) of the natural frequency, defect frequency, and the amplitude value of the defect frequency stored in the database 6.

Although not shown in the figure, by connecting this system to an Internet line, it is also possible to easily determine abnormalities in a specific bearing from a remote location.

<Configuration of vibration data analysis method for rotating equipment>
Fig. 2 is a flow chart for explaining the operation procedure of the vibration data analysis method of the present invention, and Fig. 3 is a schematic explanatory diagram for explaining specific criteria for judging a bearing abnormality in the vibration data analysis method.
In the method for analyzing vibration data of a rotating device of the present invention, an acceleration sensor 1 is attached to a rotating device such as a generator, a pump, or a fan, and waveform data of vibrations occurring in a bearing 2 of the rotating device is obtained. An ARMA model (autoregressive moving average model) is estimated for the waveform data obtained by the acceleration sensor 1. The frequency response of this estimated ARMA model is calculated.

For the calculated peak frequency (Fpeak) of the frequency response, it is assumed that the natural frequency (Fn) of bearing 2 is the natural frequency when the natural frequency (Fn) calculated using Equation 1 is within ±5% of this natural frequency (Fn).

Note that Equation 1 is a simplified formula for calculating the natural frequency of the bearing 2 in the radial direction, and is used to check whether the peak frequency (Fpeak) is significantly different from the natural frequency (Fn) calculated from the specifications of the bearing 2, such as the inner diameter and outer diameter. If the peak frequency (Fpeak) is significantly different beyond a predetermined range, such as a numerical range of ±5% with the natural frequency (Fn) as the median, other processing can be performed, such as using other waveform data as the waveform data. The formula for calculating this natural frequency (Fn) is not limited to Equation 1, and other calculation formulas for calculating the natural frequency (Fn) can be used. Also, it is not necessary to determine whether the natural frequency (Fn) is significantly different from the peak frequency (Fpeak) by calculating it from the specifications of the bearing 2, and the peak frequency (Fpeak) of the calculated frequency response can be assumed to be the natural frequency as it is.

The pass frequency (Fpass) is determined by Equation 2 in Equation 2, and the determined peak frequency (Fpeak) is stored in database 6.

The pass frequency (Fpass) does not necessarily have to be determined by Equation 2, and may be determined by another method capable of extracting only a specific frequency band based on the pass frequency (Fpass). For example, the pass frequency (Fpass) may be determined by a calculation formula such as multiplication or division using a coefficient other than Equation 2, such as one-third or more of the peak frequency (Fpeak), or by adding or subtracting an absolute value such as ±20 Hz to the peak frequency (Fpeak), or by adding or subtracting a percentage (e.g., 20%) of the peak frequency (Fpeak), or by using table data that predefines a specific frequency band corresponding to the value of the peak frequency (Fpeak).

The envelope processing unit 5 performs envelope analysis processing to identify abnormal parts of the bearing 2 by analyzing the periodicity of the repeated vibration of the bearing 2. It analyzes whether the vibration occurs repeatedly (periodicity) with a certain cycle, and if it is dependent on the rotation of the rotating equipment, it determines whether there is an abnormality in the bearing 2 based on the frequency. The signal processing of the envelope analysis performs frequency analysis on the envelope of the vibration waveform.
In order to identify the abnormal portion of the bearing 2, table data is prepared in advance in which the frequency of abnormal vibration (natural frequency) when each portion of the bearing 2 shows an abnormality is associated with the portion, and the frequency obtained by the envelope analysis process and the rotation speed information of the rotating equipment (operating rotation speed) are compared against the table data for diagnosis, whereby the abnormal portion of the bearing 2 can be identified.

FTT analysis (Fast Fourier Transform) is a method of quickly calculating the discrete Fourier transform, which represents a sequence of samples of a digital signal arranged in time series as a set of frequency components, using a computer. The Fourier transform is a process that extracts how much of each frequency component is contained in a signal, but in FTT analysis, the input waveform is divided into several groups, and by devising the calculation order, it is possible to perform the transformation with significantly less calculation effort than with normal calculation methods.

The acquired waveform data is subjected to noise removal using the band pass filter (BPF) unit 4 for the pass frequency (Fpass) calculated using Equation 1. The vibration waveform data is subjected to envelope analysis, the peak frequency is regarded as the defect frequency, and the defect frequency and the amplitude value of the defect frequency (amplitude value of the peak frequency component) are calculated and then stored in the database 6.

<How to determine bearing abnormalities>
Fig. 4 is a schematic diagram showing an example of a peak frequency in the vibration data analysis method of the present invention, and Fig. 5 is a schematic diagram showing an example of an envelope analysis result in the vibration data analysis method of the present invention.
The vibration data analysis method of the present invention judges whether or not there is an abnormality in a bearing by the procedure shown in Figures 3 to 7. The presence or absence of an abnormality in the bearing 2 is judged by using data on the natural frequency, defect frequency and amplitude value (amplitude value of the peak frequency component) stored in the database 6.

The peak frequency (Fpeak) is a natural frequency stored in database 6 and is calculated from the frequency response of the ARMA model. The calculated peak frequency (Fpeak) is compared with the natural frequency (Fn) calculated from the specifications of bearing 2. If the difference is large, it is determined that there is no abnormality in bearing 2, and if the difference is small, it is determined that there is a possibility of an abnormality. The envelope analysis result is then compared with the numerical value of the defect frequency previously stored in database 6.

FIG. 6 is a schematic diagram showing an example of a frequency distribution of defect frequencies in the vibration data analysis method of the present invention.
When determining whether or not there is an abnormality in the bearing 2 using the data of the natural frequency, defect frequency and amplitude value stored in the database 6, if the standard deviation (σ _d ) of the defect frequencies measured over a specified period is smaller than 5% of the average value (Fave) of the defect frequencies, it is determined that there is a possibility of an abnormality in the bearing 2. If this value is 5% or more, it is determined that there is no abnormality in the bearing 2. It is also possible to use a standard value.

FIG. 7 is a schematic diagram showing an example of an increase (slope) in the amplitude value of a defect frequency in the vibration data analysis method of the present invention.
When determining whether or not there is an abnormality in the bearing 2 using the data on natural frequency, defect frequency, and amplitude value stored in the database 6, the amplitude value (amplitude data) is compared to see whether it is tending to increase from previous measured values. If there is a large increase in the amplitude value, the slope of the amplitude values of the most recently measured points is greater than the average slope, in which case it is determined that there is a possibility of an abnormality in the bearing 2. If it is smaller than the average slope, it is determined that there is no abnormality in the bearing 2. It is preferable to calculate the slope using a minimum of three points, and the slope of the amplitude value may be calculated using a normal scale or a scale using a logarithm.

7 shows an example of comparing the slope of the amplitude value of the defect frequency in a first period (a period excluding the most recent four days) and a second period (a period of the most recent four days) that is closer to the first period as a predetermined period in the judgment using the amplitude value of the peak frequency component (defect frequency). The slope can be calculated by linear regression of the increase in the amplitude value of the peak frequency component over time, and if the slope in the most recent second period is larger than the slope in the first period, it can be confirmed that the amplitude value in the most recent period is increasing, and it can be judged that there is a possibility of an abnormality in the bearing 2.
Other methods can be used to check the tendency of the amplitude value of the peak frequency component to increase over time. For example, instead of linear regression, the slopes of the first period and the second period can be compared using curved regression, or a value corresponding to the difference between the amplitude value measured previously and the current amplitude value can be recorded and the tendency of the value to fluctuate can be checked.

When determining whether or not there is an abnormality in the bearing 2 using the data on the natural frequency, defect frequency, and amplitude value stored in the database 6, the period for measuring the latest value from the time when the waveform data was acquired may be one week. However, this period of one week is merely an example and is not limited to this, and it is possible to set it longer than one week. However, if it is shorter than one week, there is a risk that accurate analysis will not be possible. In this case, two or more waveform data may be acquired in one day, and the amplitude values of two or more peak frequency components for one day may be analyzed. For example, if one week is the specified period, multiple (e.g., three) waveform data may be acquired at intervals of several hours in one day, and the amplitude values of the three peak frequency components may be acquired to determine whether or not there is an abnormality.

In the following situations, it may be determined that the possibility of an abnormality is low or that there is no abnormality. For example, a case may be exemplified where, for vibration waveform data measured over a predetermined period using data on natural frequencies, defect frequencies, and amplitude values stored in the database 6, the standard deviation (σ _d ) of the defect frequencies is greater than 5% of the average defect frequency (Fave). Another example may be where there is a small number of waveform data pieces with defect frequencies within a certain numerical range relative to the average defect frequency (Fave). Another example may be where there is no confirmed tendency for the increase in amplitude values in the most recent period (second period) to be greater than that in the previous period (first period). In these cases, it may be determined that there is a low possibility of an abnormality in the bearing 2 and that there is no abnormality.
The timing for storing the data on the natural frequency, defect frequency and amplitude value in the database 6 does not necessarily have to be the date or timing when the vibration waveform data was acquired. The database 6 only needs to store the acquired vibration waveform data, and data such as the natural frequency and amplitude value for a plurality of waveform data may be calculated collectively.

Note that the present invention is not limited to the above-described embodiment of the invention, and can of course be modified in various ways without departing from the spirit of the invention, as long as it is possible to extract vibrations caused by defects in the bearings, even in the case of bearings that generate a lot of noise from rotating equipment, and it is possible to perform an abnormality diagnosis immediately after the defect occurs.

The present invention can be used to diagnose abnormalities in bearings installed in rotating equipment such as generators, pumps, and fans.

REFERENCE SIGNS LIST 1 Acceleration sensor 2 Bearing 3 ARMA model estimation unit 4 Band pass filter (BPF) unit 5 Envelope processing unit 6 Database 7 Abnormality determination unit 8 BPF change determination unit

Claims (4)

A method for analyzing vibration data of a rotating device for analyzing the presence or absence of an abnormality in a bearing provided in the rotating device, comprising:
Obtaining waveform data of vibrations generated in the bearing using an acceleration sensor;
An ARMA model is estimated using an autoregressive model and a moving average model for the waveform data of the vibration acquired by the acceleration sensor, and a frequency response of the estimated ARMA model is calculated;
determining a pass frequency of a band pass filter that extracts only a specific frequency band based on the calculated peak frequency of the frequency response of the ARMA model, and removing noise from the waveform data of the vibration acquired by the acceleration sensor using the determined pass frequency;
performing an envelope analysis on the vibration waveform data from which noise has been removed, and calculating an amplitude value of a peak frequency component obtained by the envelope analysis;
a method for analyzing vibration data of a rotating machine, comprising: using the amplitude value of the peak frequency component calculated from the vibration waveform data measured over a predetermined period, determining that there is a possibility of an abnormality in the bearing if a tendency for an increase in the amplitude value of the peak frequency component over time to be increasing can be confirmed. The method for analyzing vibration data of rotating equipment according to claim 1, characterized in that, if the peak frequency of the frequency response of the ARMA model is within a predetermined range for the natural frequency calculated based on the specifications of the bearing, the passing frequency is determined based on the peak frequency. The method for analyzing vibration data of a rotating device according to claim 1 or 2, characterized in that the slope of the amplitude value of the peak frequency component is calculated for a first period of the predetermined period and a second period immediately before the first period, and if the slope for the second period is greater, it is determined that there is a possibility of an abnormality in the bearing. A vibration data analysis system for a rotating device that analyzes the presence or absence of an abnormality in a bearing provided in the rotating device, comprising:
an acceleration sensor for acquiring waveform data of vibrations generated in the bearing;
an ARMA model estimation unit that estimates an ARMA model using an autoregressive model and a moving average model for the vibration waveform data acquired by the acceleration sensor, and calculates a frequency response of the estimated ARMA model;
a band-pass filter unit that determines a pass frequency of a band-pass filter that extracts only a specific frequency band based on the calculated peak frequency of the frequency response of the ARMA model, and removes noise from the waveform data of the vibration acquired by the acceleration sensor using the pass frequency;
an envelope processing unit that performs an envelope analysis on the vibration waveform data from which noise has been removed by the band pass filter unit, and calculates an amplitude value of a peak frequency component obtained by the envelope analysis;
and an abnormality determination unit that determines that there is a possibility of an abnormality in the bearing when a tendency for an increase in the amplitude value of the peak frequency component over time to be increasing can be confirmed using the amplitude value of the peak frequency component calculated in the envelope processing unit for the vibration waveform data measured over a predetermined period.

Images