A fault diagnosis method and system based on audio data

Abstract

This invention discloses a fault diagnosis method and system based on audio data. The method includes: acquiring cabin acoustic data and extracting multiple segments of audio data under the same stable operating condition; filtering and extracting the audio data to obtain vibration data; extracting time-frequency domain feature values ​​of the monitored object from the vibration data; and diagnosing faults in the monitored object based on the time-frequency domain feature values ​​through feature value threshold determination, octave amplitude ratio, and frequency shift. This invention utilizes cabin acoustic data (audio data) from existing data acquisition systems to extract vibration data of the monitored object, conducts vibration data analysis, and monitors damage to the monitored object through methods such as spectrum analysis, octave amplitude ratio, and frequency shift. This avoids the need to modify or add data acquisition units to equipment, effectively solving problems such as installation location, increased equipment size, weight, and power consumption caused by modifications, as well as the high costs associated with modifying existing batches of equipment.

Description

Technical Field

[0001] This invention relates to the field of equipment health management technology, and specifically to a fault diagnosis method and system based on audio data. Background Technology

[0002] Existing vibration data analysis methods are generally implemented through the data acquisition unit and data analysis system of a health management system.

[0003] The data acquisition unit and its associated vibration sensors acquire vibration data and other bus data of the monitored objects, perform data preprocessing, identify abnormal states, and record data. Based on the identification of abnormal states, the vibration data is classified and recorded. When an abnormality occurs, the raw data for 10 seconds before and after the abnormality (configurable) is recorded. When there is no abnormality, the preprocessed data is recorded. The unit also interacts with the data analysis system via Ethernet and other means to achieve data exchange and maintenance functions.

[0004] The data analysis system acquires vibration data and other bus data from the data acquisition unit to achieve data management, data analysis, fault identification and diagnosis. It enables a comprehensive assessment of the health status of the monitored system, as well as fusion analysis, trend analysis, component life analysis and service life analysis based on historical data. Through the online maintenance module, it enables functions such as data download, real-time parameter monitoring and online configuration, and pushes fault diagnosis results and maintenance suggestions to support maintenance decision-making.

[0005] The problems and shortcomings of the above vibration data analysis method are as follows:

[0006] Existing vibration data analysis methods require the acquisition of high-precision vibration signals from the monitored object. However, many existing equipment and devices did not achieve vibration data acquisition and recording in their early development plans, or the acquisition records were incomplete. Therefore, if vibration data analysis methods are to be used for health assessment of the monitored system, it is necessary to retrofit and install data acquisition units.

[0007] However, there are various constraints on adding or modifying data acquisition units for equipment and devices. These include issues such as installation location, increased equipment size, weight, and power consumption, as well as the high costs associated with modifying existing equipment. Summary of the Invention

[0008] In view of the above-mentioned technical defects in the prior art, the purpose of this invention is to provide a fault diagnosis method and system based on audio data.

[0009] To achieve the above objectives, in a first aspect, embodiments of the present invention provide a fault diagnosis method based on audio data, comprising:

[0010] Same operating condition data extraction steps: Obtain cabin sound data from the data acquisition system, and extract multiple segments of audio data under the same stable operating condition from the cabin sound data;

[0011] Digital filtering and vibration signal extraction steps: The audio data is filtered and extracted to obtain the vibration data of the monitored object;

[0012] Time-frequency domain feature extraction steps: Extract and process the vibration data of the monitored object to obtain the time-frequency domain feature values ​​of the monitored object;

[0013] Fault diagnosis steps: Based on the time-frequency domain feature values, fault diagnosis of the monitored object is achieved through feature value threshold determination, octave amplitude ratio, and frequency shift.

[0014] As a specific implementation of this application, the steps for extracting data under the same working conditions are as follows:

[0015] Acquire cabin sound data from the data acquisition system, and obtain the sampling frequency of the cabin sound data through the audio data file header information;

[0016] The working state is determined based on the stable operating frequency, resulting in multiple stable operating conditions;

[0017] Multiple audio segments under the same stable operating condition are extracted from the cabin audio data.

[0018] As a specific implementation of this application, the digital filtering and vibration signal extraction steps are as follows:

[0019] A low-pass filter is used to filter multiple audio data segments under the same stable operating condition to extract vibration data of the monitored object with a high signal-to-noise ratio.

[0020] As a specific implementation of this application, the time-frequency domain feature value extraction steps are as follows:

[0021] Perform a Fourier transform on the vibration data of the monitored object to obtain the spectral distribution map;

[0022] Extract the time-frequency domain feature values ​​of the monitored object from the spectrum distribution map.

[0023] As one specific implementation of this application, the fault diagnosis steps include:

[0024] Monitoring steps for the frequency amplitude ratio of the monitored object: Calculate the ratio of the frequency amplitudes of the N-fold frequency signal and the 2N-fold frequency signal in the spectrum distribution diagram. If the ratio is within the preset threshold range, the monitored object is determined to be in a normal state; otherwise, the monitored object is determined to have a fault symptom or mechanical fault.

[0025] Monitoring steps for frequency shift of the monitored object: Obtain the frequency value of the Nth harmonic component of the vibration data of the monitored object from the spectrum distribution diagram. If the frequency value remains basically unchanged over time, it is determined that the monitored object is in a normal state. If the frequency value shifts, it is determined that the monitored object has a fault symptom or mechanical fault.

[0026] The time-domain signal-assisted fault determination steps are as follows: Fourier transform is performed on the vibration data of the monitored object to obtain the frequency domain signal, and inverse Fourier transform is performed on the frequency domain signal to restore the time domain signal. Fault determination is then performed based on the periodic continuity and fluctuation of the time domain signal.

[0027] Accordingly, in a second aspect, embodiments of the present invention provide a fault diagnosis system based on audio data, comprising:

[0028] The same working condition data interception module is used to obtain cabin sound data from the data acquisition system and extract multiple segments of audio data under the same stable working condition from the cabin sound data;

[0029] A digital filtering and vibration signal extraction module is used to filter and extract the audio data to obtain the vibration data of the monitored object;

[0030] The time-frequency domain feature extraction module is used to extract and process the vibration data of the monitored object to obtain the time-frequency domain feature values ​​of the monitored object;

[0031] The fault diagnosis module is used to diagnose faults in the monitored object based on the time-frequency domain feature values, through feature value threshold determination, harmonic amplitude ratio, and frequency shift.

[0032] As a specific implementation of this application, the same working condition data interception module is specifically used for:

[0033] Acquire cabin sound data from the data acquisition system, and obtain the sampling frequency of the cabin sound data through the audio data file header information;

[0034] The working state is determined based on the stable operating frequency, resulting in multiple stable operating conditions;

[0035] Multiple audio segments under the same stable operating condition are extracted from the cabin audio data.

[0036] As one specific implementation of this application, the digital filtering and vibration signal extraction module is specifically used for:

[0037] A low-pass filter is used to filter multiple audio data segments under the same stable operating condition to extract vibration data of the monitored object with a high signal-to-noise ratio.

[0038] As a specific implementation of this application, the time-frequency domain feature extraction module is specifically used for:

[0039] Perform a Fourier transform on the vibration data of the monitored object to obtain the spectral distribution map;

[0040] Extract the time-frequency domain feature values ​​of the monitored object from the spectrum distribution map.

[0041] As one specific implementation of this application, the fault diagnosis module includes:

[0042] The monitoring module for monitoring the frequency ratio of the N-fold frequency signal and the 2N-fold frequency signal in the spectrum distribution diagram is used to calculate the ratio of the frequency amplitudes. If the ratio is within the preset threshold range, the monitoring object is determined to be in a normal state; otherwise, the monitoring object is determined to have a fault symptom or mechanical fault.

[0043] The monitoring module for frequency shift of the monitored object is used to obtain the frequency value of the Nth harmonic component of the vibration data of the monitored object from the spectrum distribution diagram. If the frequency value remains basically unchanged over time, the monitored object is determined to be in a normal state. If the frequency value shifts, the monitored object is determined to have a fault symptom or mechanical failure.

[0044] The time-domain signal-assisted fault determination module is used to perform Fourier transform on the vibration data of the monitored object to obtain a frequency domain signal, perform inverse Fourier transform on the frequency domain signal to restore the time domain signal, and perform fault-assisted determination based on the periodic continuity and fluctuation of the time domain signal.

[0045] By implementing embodiments of the present invention, vibration data of the monitored object is extracted using cabin acoustic data (audio data) from existing data acquisition systems. Vibration data analysis is then conducted, and damage to the monitored object is monitored through methods such as spectrum analysis, octave amplitude ratio, and frequency shift. This avoids the need to modify or upgrade data acquisition units on equipment and facilities, effectively solving problems such as increased installation location, increased equipment size, weight, and power consumption associated with modifications, as well as the high costs associated with modifying existing batches of equipment. Furthermore, embodiments of the present invention improve the maintenance and support capabilities of equipment and facilities at a relatively low cost, demonstrating technical feasibility, low risk, and high military-economic benefits. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below.

[0047] Figure 1 This is a structural diagram of the fault diagnosis system based on audio data provided in an embodiment of the present invention;

[0048] Figure 2 Based on Figure 1 The flowchart of the fault diagnosis system is shown below.

[0049] Figure 3 This is a flowchart of a fault diagnosis method based on audio data provided in an embodiment of the present invention;

[0050] Figure 4 This is a flowchart for fault diagnosis of the rotor system;

[0051] Figure 5 This is a spectrogram of the raw cabin acoustic data;

[0052] Figure 6 This is a spectrum of the extracted vibration data;

[0053] Figure 7 This is a spectrum diagram showing the amplitude of the rotor's 6th and 12th harmonics;

[0054] Figure 8 This is a diagram illustrating the frequency shift of the rotor at 6 times its normal frequency.

[0055] Figure 9 This is a basic mechanism diagram of rotor-induced vibration. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0058] The inventive concept of this invention is to bypass the need for modified equipment and data acquisition units, and instead collect and record cabin acoustic data using existing data acquisition systems. This data is then analyzed using a data analysis system. Since the operating frequency and fault frequency of the monitored object differ significantly from the human voice frequency, the vibration signals at these frequencies can be separated and extracted. Furthermore, methods such as spectrum analysis, octave ratio, and frequency shift are used to monitor damage to the monitored object, thereby achieving system health assessment.

[0059] Based on the above inventive concept, the technical means of this invention can be summarized as follows: acquiring cabin sound data collected and recorded by an existing data acquisition system; extracting vibration signals and harmonic signals of the monitored object from the cabin sound data; filtering the data using a digital filter to obtain vibration data of the monitored object with a high signal-to-noise ratio; performing spectral analysis on the extracted vibration data; extracting time-frequency domain characteristic values ​​and harmonic information of the vibration signal of the monitored object for fault diagnosis of different types of equipment; extracting different harmonic information of the vibration data of the monitored object according to requirements; and realizing fault diagnosis of the monitored object through methods such as characteristic threshold determination, harmonic frequency-to-weight ratio, and frequency shift, thereby achieving system health assessment.

[0060] Please refer to Figure 1 The fault diagnosis system based on audio data provided in this embodiment of the invention includes:

[0061] The same working condition data interception module is used to obtain cabin sound data from the data acquisition system and extract multiple segments of audio data under the same stable working condition from the cabin sound data;

[0062] A digital filtering and vibration signal extraction module is used to filter and extract the audio data to obtain the vibration data of the monitored object;

[0063] The time-frequency domain feature extraction module is used to extract and process the vibration data of the monitored object to obtain the time-frequency domain feature values ​​of the monitored object;

[0064] The monitoring module for monitoring the frequency ratio of the N-fold frequency signal and the 2N-fold frequency signal in the spectrum distribution diagram is used to calculate the ratio of the frequency amplitudes. If the ratio is within the preset threshold range, the monitoring object is determined to be in a normal state; otherwise, the monitoring object is determined to have a fault symptom or mechanical fault.

[0065] The monitoring module for frequency shift of the monitored object is used to obtain the frequency value of the Nth harmonic component of the vibration data of the monitored object from the spectrum distribution diagram. If the frequency value remains basically unchanged over time, the monitored object is determined to be in a normal state. If the frequency value shifts, the monitored object is determined to have a fault symptom or mechanical failure.

[0066] The time-domain signal-assisted fault determination module is used to perform Fourier transform on the vibration data of the monitored object to obtain a frequency domain signal, perform inverse Fourier transform on the frequency domain signal to restore the time domain signal, and perform fault-assisted determination based on the periodic continuity and fluctuation of the time domain signal.

[0067] Specifically, the same working condition data interception module is used for:

[0068] Acquire cabin sound data from the data acquisition system, and obtain the sampling frequency of the cabin sound data through the audio data file header information;

[0069] The working state is determined based on the stable operating frequency, resulting in multiple stable operating conditions;

[0070] Multiple audio segments under the same stable operating condition are extracted from the cabin audio data.

[0071] In this embodiment, the digital filtering and vibration signal extraction module is specifically used for:

[0072] A low-pass filter is used to filter multiple audio data segments under the same stable operating condition to extract vibration data of the monitored object with a high signal-to-noise ratio.

[0073] In this embodiment, the time-frequency domain feature extraction module is specifically used for:

[0074] Perform a Fourier transform on the vibration data of the monitored object to obtain the spectral distribution map;

[0075] Extract the time-frequency domain feature values ​​of the monitored object from the spectrum distribution map.

[0076] Please refer to this again. Figure 2 and Figure 3 The fault diagnosis method based on audio data provided in this embodiment of the invention may include:

[0077] S1, Same operating condition data extraction step: Obtain cabin sound data from the data acquisition system, and extract multiple segments of audio data under the same stable operating condition from the cabin sound data.

[0078] Specifically, the cabin audio data is imported, and the sampling frequency of the cabin audio data is obtained through the audio data file header information. The operating state is determined based on the stable operating frequency, resulting in multiple stable operating conditions. Multiple segments of audio data under the same stable operating condition are extracted from the cabin audio data. In this embodiment, there are 5 stable operating conditions. The sampling frequency mentioned above is set by the hardware in the acquisition system, and the stable operating frequency is the operating frequency of the monitored object.

[0079] It should be noted that in this step, the fundamental frequency and the frequency values ​​of each harmonic signal are calculated based on the operating frequency of the monitored object, thus obtaining the least common multiple of the periods of each harmonic signal. If this least common multiple is less than the length of the cabin sound data, the cabin sound data is truncated to an integer period and windowed; if the least common multiple is longer than the length of the observed signal, or it is difficult to obtain the period of the vibration signal of the monitored object in the cabin sound data, a Hanning window is added when performing a Fourier transform on the signal. The purpose of this processing method is to improve the accuracy of converting the time-domain signal to the frequency-domain signal and to minimize frequency leakage.

[0080] S2, Digital Filtering and Vibration Signal Extraction Steps: The audio data is filtered and extracted to obtain the vibration data of the monitored object.

[0081] Specifically, using a low-pass filter to filter audio data and setting a reasonable cutoff frequency can effectively filter noise, human voices, and high-frequency signals generated by other devices, thereby extracting vibration data of the monitored object with a high signal-to-noise ratio.

[0082] It should be noted that the effective vibration signals of the monitored object in the cabin acoustic data are all low-frequency signals. Therefore, setting a low-pass filter can filter out frequency components higher than the cutoff frequency. After filtering, this embodiment mainly analyzes the frequency response of the Nth harmonic component and the 2Nth harmonic component in the vibration signal spectrum of the monitored object.

[0083] S3, Time-frequency domain feature extraction step: Extract and process the vibration data of the monitored object to obtain the time-frequency domain feature values ​​of the monitored object.

[0084] Time-domain characteristics are relatively intuitive features of mechanical vibration signals. The main time-domain analysis features include minimum value, maximum value, root mean square value, waveform index, peak value, etc.

[0085] Frequency domain characteristics first involve calculating the operating frequency point value of the monitored object. Different components have different vibration frequencies, making the frequency point a crucial monitoring feature. Key frequency domain analysis parameters include average frequency, frequency center, root mean square frequency, and frequency standard deviation.

[0086] Specifically, the vibration data of the monitored object is subjected to Fourier transform to obtain a spectrum distribution map, and then the time-frequency domain feature values ​​of the vibration signal of the monitored object are extracted from the spectrum map.

[0087] S4, Fault diagnosis steps: Based on the time-frequency domain feature values, fault diagnosis of the monitored object is achieved through feature value threshold determination, harmonic frequency amplitude ratio and frequency shift.

[0088] Specifically, fault diagnosis mainly includes three sub-steps:

[0089] 1. Monitoring of frequency ratio of monitored objects

[0090] The principle behind this step is to diagnose faults by comparing the proportional relationships between the amplitudes of different harmonic components of the vibration signal from the monitored object. The amplitude ratios of the harmonic components from the same source signal must be within a certain threshold range. In other words, this step uses an amplitude comparison method to diagnose faults in the monitored object using the same harmonic signal from different data segments.

[0091] It should be noted that, considering the accuracy of Fourier transform, the frequency of the same frequency component may differ slightly in the spectral distribution of different data segments of the cabin sound data. Therefore, before comparing the amplitudes, it is necessary to appropriately shift the frequency components in the frequency domain signal of each data segment so that the frequencies corresponding to the Nth harmonic and 2Nth harmonic components of the monitored object are the same in the spectral distribution of different data segments.

[0092] In the aforementioned steps, cabin sound data is extracted based on the operation of the same equipment to obtain multiple data segments under the same working condition. Vibration data is obtained based on these data segments. The extracted vibration data of the monitored object is then converted into a frequency domain signal using Fourier transform, and the spectral distribution map of different cabin sound data segments is obtained.

[0093] The ratio of the frequency amplitudes of the Nth harmonic signal and the 2Nth harmonic signal in the spectral distribution diagram is calculated. If the ratio of the amplitude of the Nth harmonic component to the amplitude of the 2Nth harmonic component of the vibration signal of the monitored object is within a certain threshold range, the monitored object system is in a normal state; otherwise, the monitored object system is considered to have fault symptoms or mechanical failures.

[0094] 2. Frequency shift monitoring of monitored objects

[0095] The principle behind this step is as follows: fault diagnosis is achieved by monitoring the operating frequency values ​​of the vibration signal of the monitored object. When the equipment and facilities are operating stably, the operating spectrum of the vibration signal of the monitored object remains unchanged, and the frequency values ​​of its Nth harmonic components also remain essentially unchanged. First, the vibration signal of the monitored object extracted from the cabin acoustic data is converted into a frequency domain signal using a Fourier transform; then, the frequency values ​​of the Nth harmonics of the data at different time periods under stable operating conditions are obtained, thereby deriving the trend of the change in the frequency values ​​of the Nth harmonics of the data at different time periods, and realizing fault diagnosis of the monitored object system.

[0096] Specifically, the frequency values ​​of the Nth harmonic components of the vibration signal of the monitored object can be obtained from the spectral distribution diagram of the cabin acoustic data at different time periods. If the frequency value of the Nth harmonic component of the vibration signal of the monitored object remains basically unchanged over time, the monitored object system is considered to be in a normal state. If the Nth harmonic component of the vibration signal of the monitored object shows a frequency shift, the monitored object system is considered to have a fault symptom or mechanical failure.

[0097] 3. Time-domain signal-assisted fault diagnosis

[0098] Fourier transform is performed on the vibration data of the monitored object to obtain the frequency domain signal. Inverse Fourier transform is performed on each frequency domain signal to restore the time domain form of the filtered cabin sound data. Faults can be assisted in by analyzing the periodicity and fluctuation of the time domain signal.

[0099] In addition, this embodiment can also monitor the vibration characteristic threshold of the monitored object, as follows:

[0100] Vibration characteristic values ​​are assessed using an exceedance threshold, which can be obtained through statistical analysis. For example, the threshold can be calculated by collecting predefined data points and then performing statistical analysis. The number of data points can be preset to 40. The statistical threshold is set to m+10D, where m is the average of the 40 data points and D is the standard deviation of the 40 data points.

[0101] The vibration over-limit judgment method can combine one or more characteristic values ​​for comprehensive judgment. When they exceed the limit at the same time, the vibration over-limit is valid. For a certain important characteristic value, the vibration over-limit fault is judged by exceeding the limit 5 times consecutively.

[0102] As described above, implementing this embodiment of the invention utilizes cabin acoustic data (audio data) from existing data acquisition systems to extract vibration data of the monitored object, conducts vibration data analysis, and monitors damage to the monitored object through methods such as spectrum analysis, octave amplitude ratio, and frequency shift. This avoids the need to modify or upgrade equipment and devices by adding data acquisition units, effectively solving problems such as installation location, increased equipment size, weight, and power consumption associated with modifications, as well as the high costs associated with modifying existing batches of equipment. Furthermore, this embodiment of the invention improves the maintenance and support capabilities of equipment and devices at a relatively low cost, demonstrating technical feasibility, low risk, and high military-economic benefits.

[0103] Furthermore, embodiments of the present invention have been applied to fault diagnosis of a certain type of helicopter rotor system; the specific process can be found in [reference needed]. Figure 4 As shown in the figure, fault diagnosis of a helicopter rotor system mainly includes:

[0104] 1. Extract rotor vibration signals from cabin acoustic data

[0105] Import cabin audio data (such as...) Figure 5 As shown, the audio data sampling frequency is obtained through the audio data file header information. The working state is determined based on the stable operating frequency, resulting in multiple stable operating conditions. Multiple segments of audio data under the same stable operating condition are extracted from the cabin audio data. A low-pass filter is used to filter the audio data. Setting a reasonable cutoff frequency can effectively filter noise, human voices, and high-frequency signals generated by other equipment, thereby obtaining rotor vibration sixth harmonic data with a high signal-to-noise ratio.

[0106] 2. Vibration signal feature extraction

[0107] Time-frequency domain eigenvalues ​​are relatively intuitive characteristic information of rotor vibration signals. The main characteristic parameters of time-frequency domain analysis include minimum value, maximum value, root mean square value, waveform index, peak index, average frequency, frequency center, root mean square frequency, and frequency standard deviation.

[0108] Specifically, Fourier transform was performed on the 6th harmonic data of rotor vibration with a high signal-to-noise ratio to obtain, as shown below. Figure 6 The vibration signal spectrum distribution diagram shown can be used to extract the time-frequency domain characteristics of the rotor vibration signal.

[0109] Furthermore, in Figure 6 In the spectrum distribution diagram shown, the amplitudes of the rotor's 6th and 12th harmonics are as follows: Figure 7 As shown. Frequency shifting from its 6th harmonic yields... Figure 8 .

[0110] 3. Fault diagnosis based on vibration signal characteristics

[0111] The basic mechanism of rotor-induced vibration is as follows Figure 9 As shown, please refer to the detailed fault diagnosis steps. Figure 4 .

[0112] The above scheme, through the analysis of the vibration data spectrum and eigenvalues ​​extracted from the cabin sound data, concludes that the rotor operating frequency signal shows a frequency shift before the occurrence of an obvious fault, and that the ratio of the harmonic amplitude of the rotor main frequency shows an abnormality in the early stable operating state of the rotor system. Therefore, the rotor system fault can be diagnosed.

[0113] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A fault diagnosis method based on audio data, applied to fault diagnosis of helicopter rotor systems, characterized in that, include: Same operating condition data extraction steps: Obtain cabin sound data from the existing data acquisition system, and extract multiple segments of audio data under the same stable operating condition from the cabin sound data; Digital filtering and vibration signal extraction steps: Use a low-pass filter to filter multiple audio data segments under the same stable operating condition to extract vibration data of the monitoring object with a high signal-to-noise ratio; Time-frequency domain feature extraction steps: Perform Fourier transform on the vibration data of the monitored object to obtain a spectrum distribution map; extract the time-frequency domain feature values ​​of the monitored object from the spectrum distribution map; Fault diagnosis steps: Based on the time-frequency domain feature values, fault diagnosis of the monitored object is achieved through feature value threshold determination, harmonic amplitude ratio, and frequency shift; Monitoring steps for the frequency amplitude ratio of the monitored object: Calculate the ratio of the frequency amplitudes of the N-fold frequency signal and the 2N-fold frequency signal in the spectrum distribution diagram. If the ratio is within the preset threshold range, the monitored object is determined to be in a normal state; otherwise, the monitored object is determined to have a fault symptom or mechanical fault. Monitoring steps for frequency shift of the monitored object: Obtain the frequency value of the Nth harmonic component of the vibration data of the monitored object from the spectrum distribution diagram. If the frequency value remains basically unchanged over time, it is determined that the monitored object is in a normal state. If the frequency value shifts, it is determined that the monitored object has a fault symptom or mechanical fault. The time-domain signal-assisted fault determination steps are as follows: Fourier transform is performed on the vibration data of the monitored object to obtain the frequency domain signal, and inverse Fourier transform is performed on the frequency domain signal to restore the time domain signal. Fault determination is then performed based on the periodic continuity and fluctuation of the time domain signal.

2. The fault diagnosis method as described in claim 1, characterized in that, The specific steps for extracting data under the same working conditions are as follows: Acquire cabin sound data from an existing data acquisition system, and obtain the sampling frequency of the cabin sound data through the audio data file header information; The working state is determined based on the frequency of stable operating conditions, resulting in multiple stable operating conditions. Multiple audio segments under the same stable operating condition are extracted from the cabin audio data.

3. A fault diagnosis system based on audio data, applied to fault diagnosis of helicopter rotor systems, characterized in that, include: The same working condition data interception module is used to obtain cabin sound data from the existing data acquisition system and extract multiple segments of audio data under the same stable working condition from the cabin sound data. The digital filtering and vibration signal extraction module is used to filter multiple audio data segments under the same stable operating condition using a low-pass filter, and extract the vibration data of the monitoring object with a high signal-to-noise ratio. The time-frequency domain feature extraction module is used to perform Fourier transform on the vibration data of the monitored object to obtain a spectrum distribution map, and extract the time-frequency domain feature values ​​of the monitored object from the spectrum distribution map; The fault diagnosis module is used to diagnose faults in the monitored object based on the time-frequency domain feature values, through feature value threshold determination, harmonic frequency amplitude ratio, and frequency shift. The fault diagnosis module includes: The monitoring module for monitoring the frequency amplitude ratio of the monitored object is used to calculate the ratio of the frequency amplitude of the N-fold frequency signal and the 2N-fold frequency signal in the spectrum distribution diagram. If the ratio is within the preset threshold range, the monitored object is determined to be in a normal state; otherwise, the monitored object is determined to have a fault symptom or mechanical fault. The monitoring module for frequency shift of the monitored object is used to obtain the frequency value of the Nth harmonic component of the vibration data of the monitored object from the spectrum distribution diagram. If the frequency value remains basically unchanged over time, it is determined that the monitored object is in a normal state. If the frequency value shifts, it is determined that the monitored object has a fault symptom or mechanical fault. The time-domain signal-assisted fault determination module is used to perform Fourier transform on the vibration data of the monitored object to obtain a frequency domain signal, perform inverse Fourier transform on the frequency domain signal to restore the time domain signal, and perform fault-assisted determination based on the periodic continuity and fluctuation of the time domain signal.

4. The fault diagnosis system as described in claim 3, characterized in that, The same working condition data interception module is specifically used for: Acquire cabin sound data from an existing data acquisition system, and obtain the sampling frequency of the cabin sound data through the audio data file header information; The working state is determined based on the frequency of stable operating conditions, resulting in multiple stable operating conditions. Multiple audio segments under the same stable operating condition are extracted from the cabin audio data.

Images