A compressor abnormal sound detection method and system based on acoustic spectrum pattern recognition

Abstract

The present application belongs to the technical field of compressor abnormal sound detection, and particularly relates to a compressor abnormal sound detection method and system based on acoustic spectrum pattern recognition, which comprises the following steps: extracting a characteristic signal from a sound signal through wavelet packet transform and kurtosis criterion, locating the occurrence time of an impact event contained in the characteristic signal and calculating the time interval of adjacent impact events, calculating a rhythm stability index based on all time intervals in an analysis time window, intercepting a time domain signal window containing each impact event and calculating its energy envelope, calculating the intrinsic damping factor of each energy envelope according to the logarithmic difference between the energy envelope peak value and the energy value after a set time length, calculating the fault authenticity score of each impact event according to the energy value of the impact event in the characteristic signal, the rhythm stability index corresponding to the analysis time window and the intrinsic damping factor, and performing early warning determination, so as to realize high-reliability and accurate early warning of compressor early faults.

Description

Technical Field

[0001] This invention relates to the field of compressor abnormal noise detection technology. More specifically, this invention relates to a compressor abnormal noise detection method and system based on acoustic spectrum pattern recognition. Background Technology

[0002] Reciprocating compressors are core power equipment in industrial processes such as petrochemicals, natural gas, and air separation. Their operating status directly affects the safety and efficiency of the entire production system. Due to the complex structure of the compressor and its long-term operation under high temperature and high pressure conditions, key components such as valves, bearings, and piston rings are prone to wear, cracks, or loosening, which can lead to abnormal impacts or noises. If these early signs of failure are not detected in time, the failure may evolve into a shutdown accident, causing significant economic losses and safety hazards.

[0003] Therefore, online condition monitoring and early fault warning of compressors have extremely high application value. Among the many monitoring methods, analysis based on vibration or acoustic signals is widely used, especially wavelet packet transform (WPT), which is often used to extract weak transient impact signals from strong background noise due to its excellent time-frequency analysis capabilities.

[0004] However, wavelet packet transform is highly sensitive to any transient or abrupt changes in a signal. While this helps to capture early faults, it also makes it impossible to distinguish the source of the signal. In a real factory environment, in addition to the impact that may be caused by compressor failures, there are also a large number of random transient interferences unrelated to the fault, such as the start-up and shutdown of other nearby equipment, accidental drops of production tools, sudden opening and closing of pipeline valves, and even electromagnetic pulse interference. These random interferences also manifest as transient impacts in the signal, and wavelet packet transform will faithfully capture and amplify them. If the monitoring system relies solely on a simple energy threshold for judgment, it will be completely unable to distinguish between such benign external random impacts and malignant, periodic impacts caused by internal faults.

[0005] This indiscriminate high sensitivity directly leads to an extremely high false alarm rate in the actual deployment of existing wavelet packet transform-based detection technologies. Frequent false alarms not only seriously disrupt the normal work schedule of equipment maintenance personnel, but more importantly, they gradually erode users' trust in the entire monitoring system, ultimately leading to the system being abandoned and losing its due protective function. Summary of the Invention

[0006] To address the technical problem of high false alarm rates caused by the inability of existing wavelet packet transform-based detection methods to effectively distinguish between fault impacts and random interference, this invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition, comprising: acquiring the sound signal and rotational speed of the compressor; performing wavelet packet transform on the sound signal to obtain a reconstructed signal with multiple frequency bands; calculating the kurtosis value of the reconstructed signal, and taking the reconstructed signal with the largest kurtosis value as a feature signal; within a set analysis time window, locating the occurrence time of the impact event contained in the feature signal and calculating the time interval between adjacent impact events; calculating a rhythm stability index based on all time intervals within the analysis time window, wherein the difference between the rhythm stability index and each time interval and the average time interval within the analysis time window is... The energy envelope of each impact event is negatively correlated with the energy value and positively correlated with the reciprocal of the compressor speed. The energy envelope is calculated by extracting a time-domain signal window containing each impact event. The intrinsic damping factor of each energy envelope is calculated; the intrinsic damping factor is directly proportional to the logarithmic difference between the peak value of the corresponding energy envelope and the energy value after a set time, and inversely proportional to the set time. The fault authenticity score of each impact event is calculated; the fault authenticity score is positively correlated with the product of the energy value of the corresponding impact event in the characteristic signal, the rhythmic stability index of the corresponding analysis time window, and the intrinsic damping factor. When the fault authenticity scores of a set number of impact events exceed a preset threshold, the compressor is determined to have abnormal noise.

[0008] This invention utilizes wavelet packet transform and kurtosis criterion to extract the most sensitive feature signals, and introduces a rhythm stability index to quantify the periodicity of impact events, thereby eliminating random interference unrelated to equipment rotation speed. Simultaneously, it analyzes the physical attenuation pattern of impact signals by calculating the intrinsic damping factor, providing physical-level authentication that the impact originates from mechanical impacts within the equipment. Finally, it integrates energy, rhythm stability, and intrinsic damping into a fault authenticity score. Through multi-dimensional feature fusion, it constructs a decision-making basis far more reliable than a single energy threshold, thereby significantly reducing the false alarm rate and significantly improving the accuracy and reliability of early fault diagnosis, achieving precise identification of compressor abnormal noises.

[0009] Preferably, the method for calculating the rhythm stability index is as follows: ;in: This is the rhythm stability index for the current analysis window; This represents the total number of impact events detected within the current analysis window; For the current analysis window The first shock event and the first The impact interval of a single impact event; The average period of the current analysis window is equal to the average of all impact intervals within the current analysis window; The reference rotation period, and , This refers to the real-time speed of the compressor. It is a natural exponential function; This indicates taking the absolute value.

[0010] This invention provides an accurate index for evaluating the regularity of impact sequences by accumulating the deviations of each impact interval from the average cycle and normalizing them with the compressor's reference rotation cycle. This makes the assessment of rhythm more objective and robust, and can adapt to changes in compressor speed, thereby more accurately identifying periodic faults related to rotating parts of the equipment.

[0011] Preferably, the intrinsic damping factor is calculated as follows: ;in: The intrinsic damping factor for shock events; The energy envelope of the shock event The peak value; The energy envelope of the shock event From peak attenuation Energy value after time; This is the preset analysis duration.

[0012] This invention models the attenuation process of the energy envelope of an impact event as a linear attenuation in logarithmic coordinates and calculates its slope, providing a clear quantitative parameter for judging the physical properties of the impact source. It can accurately capture the characteristics of structural damped vibrations caused by internal mechanical impacts, thereby effectively eliminating external interferences that are rhythmic but do not match the physical form, and enhancing the accuracy of fault source identification.

[0013] Preferably, the fault authenticity score is also positively correlated with the intrinsic damping characteristic matching factor, and the intrinsic damping characteristic matching factor is calculated as follows: ;in: The intrinsic damping factor for shock events; and These are the mean and standard deviation derived from health data; Indicates taking the absolute value; This indicates taking the absolute value.

[0014] This invention transforms a physical parameter into a confidence score within the [0,1] interval by comparing the intrinsic damping factor calculated in real time with the benchmark value and standard deviation learned from health equipment data. This enables the system to not only determine the impact pattern but also assess its deviation from the normal pattern, achieving an upgrade from feature extraction to intelligent assessment and making fault judgment more statistically significant and adaptive.

[0015] Preferably, the method for calculating the fault authenticity score is as follows: ;in: Scoring the authenticity of the failure in the impact event; For impact events in characteristic signals The corresponding energy value; The rhythmic stability index of the analysis window where the impact event occurs; The intrinsic damping characteristic matching factor of the impact event.

[0016] This invention ensures that a high score can only be obtained when an impact event simultaneously meets the three conditions of significant energy, stable rhythm, and matching physical form by multiplying the impact energy, rhythm stability, and intrinsic damping characteristic matching factor. This sets a higher fault judgment threshold and can most effectively suppress various interference signals with obvious single characteristics. It is the core calculation step for achieving low false alarm rate and high reliability detection.

[0017] Preferably, locating the occurrence time of the impact events contained in the feature signal includes: applying the Teager-Kaiser energy operator to the feature signal to obtain an energy sequence; performing peak detection on the energy sequence to locate the occurrence time of all impact events.

[0018] Preferably, obtaining the energy envelope includes: applying Hilbert transform to calculate the energy envelope of the time-domain signal window centered on the moment of the impact event in the characteristic signal.

[0019] Preferably, the method further includes: calculating the intrinsic damping factor and fault authenticity score only for impact events within the analysis window where the rhythm stability index is greater than 0.5.

[0020] Preferably, determining that the compressor has abnormal noise includes: in an evaluation sequence containing 20 consecutive impact events, if the failure authenticity score of 5 or more impact events exceeds the preset threshold, then the compressor is determined to have abnormal noise.

[0021] Secondly, the present invention provides a compressor abnormal noise detection system based on acoustic spectrum pattern recognition, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned compressor abnormal noise detection method based on acoustic spectrum pattern recognition is implemented.

[0022] By adopting the above technical solution, a computer program for detecting abnormal noise of a compressor based on acoustic spectrum pattern recognition is generated and stored in a memory so that it can be loaded and executed by a processor. A terminal device can then be made based on the memory and the processor for convenient use.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention utilizes wavelet packet transform and kurtosis criterion to extract the most sensitive feature signals, and introduces a rhythm stability index to quantify the periodicity of impact events, thereby eliminating random interference unrelated to equipment rotation speed. Simultaneously, it analyzes the physical attenuation pattern of impact signals by calculating the intrinsic damping factor, providing physical-level authentication that the impact originates from mechanical impacts within the equipment. Finally, it integrates energy, rhythm stability, and intrinsic damping into a fault authenticity score. Through multi-dimensional feature fusion, it constructs a decision-making basis far more reliable than a single energy threshold, thereby significantly reducing the false alarm rate and significantly improving the accuracy and reliability of early fault diagnosis, achieving precise identification of compressor abnormal noises. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a compressor abnormal noise detection method based on acoustic spectrum pattern recognition according to the present invention;

[0026] Figure 2 It is a schematic diagram illustrating a sound signal;

[0027] Figure 3 This is a schematic diagram illustrating the characteristic signals;

[0028] Figure 4 This is a schematic diagram illustrating a detected impact event;

[0029] Figure 5 This is a schematic diagram illustrating the intrinsic damping factors of various impact events;

[0030] Figure 6 This is a schematic diagram illustrating the failure authenticity score for each impact event. Detailed Implementation

[0031] 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.

[0032] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] This invention discloses a method for detecting abnormal noise in compressors based on acoustic spectrum pattern recognition, referring to... Figure 1 This includes steps S1-S4:

[0034] S1: Acquire the compressor's sound signal and speed; perform wavelet packet transform on the sound signal to obtain reconstructed signals in multiple frequency bands; calculate the kurtosis value of the reconstructed signal, and take the reconstructed signal with the largest kurtosis value as the feature signal.

[0035] First, high-fidelity acoustic sensors are deployed near key components of the compressor to synchronously collect sound signals from the compressor during operation, which are then recorded as sound signals. These sensors are then connected to the compressor's speed sensor to obtain the compressor's real-time operating speed. The key components include, but are not limited to, bearing housings and cylinder heads.

[0036] Collected sound signals This typically includes steady-state noise from normal equipment operation, such as airflow noise and motor electromagnetic noise, as well as potential early fault impact signals and random interference in the environment; in order to effectively capture weak early fault impacts, the sound signal... conduct Layer wavelet packet transform, then the sound signal Decomposed into These are non-overlapping frequency bands, covering the entire frequency range from 0 to half the sampling rate.

[0037] Because the impact signals generated by early mechanical failures such as pitting and microcracks are typically low in energy but highly transient, and their energy is mainly concentrated in the high-frequency region, this step, based on the kurtosis maximization criterion, iterates through the obtained frequency bands to adaptively identify which frequency band best highlights this impact characteristic. For each frequency band, calculate the kurtosis value of its reconstructed signal.

[0038] Kurtosis is a statistical measure characterizing the impulsiveness of a signal; the larger the kurtosis value, the more impulsive components the signal contains. Therefore, the frequency band with the largest kurtosis value is selected as the optimal fault characteristic frequency band, and its reconstructed signal is denoted as the characteristic signal. Compared to the original signal , Background noise was significantly suppressed, while transient impact components were effectively highlighted.

[0039] Thus, by using wavelet packet decomposition and kurtosis maximization optimization, the fault feature signal with the highest signal-to-noise ratio can be adaptively extracted from the complex original signal, laying the foundation for subsequent accurate analysis.

[0040] S2: Within the set analysis time window, locate the occurrence time of the impact event contained in the feature signal and calculate the time interval between adjacent impact events; calculate the rhythm stability index based on all time intervals within the analysis time window.

[0041] In real industrial environments, relying solely on characteristic signals The energy amplitude of the impact cannot distinguish between fault impact and random interference; this step aims to start from the regularity of the time series and analyze the characteristic signals. The impact events included are screened; failure impacts are usually related to the movement of the rotating parts of the compressor and are periodic or quasi-periodic, while environmental disturbances are mostly random and isolated events.

[0042] To assess the stability of this rhythm, it is first necessary to accurately identify the timing of each impact event; firstly, by analyzing characteristic signals... Instantaneous energy calculations were performed using the Teager-Kaiser energy operator to obtain the energy sequence. Among them, the Teager-Kaiser energy operator is highly sensitive to instantaneous amplitude and frequency changes in the signal, and can instantaneously amplify the impact point; then, the peak detection algorithm is used in the energy sequence The occurrence times of all impact events are located, forming an impact time series. , This represents the total number of impact events detected within the current analysis window.

[0043] Based on the impact time series, the impact interval sequence of adjacent impact events is calculated. ,in, .

[0044] A real failure caused by a rotating component, with its impact interval. It should stably revolve around the fundamental frequency period. Fluctuations; to assess whether the interval sequence is related to the compressor's own operating rhythm, a rhythm stability index is introduced. The index is calculated based on a sliding analysis window, the length of which is set. seconds, sliding step Second.

[0045] The system calculates and updates the rhythmic stability index of all impact events falling within the analysis window in real time. The rhythm stability index of the current analysis window The calculation formula is as follows:

[0046]

[0047] in: This is the rhythm stability index for the current analysis window; This represents the total number of impact events detected within the current analysis window; For the current analysis window The first shock event is related to the previous one, namely the... The impact interval of a single impact event; The average period of the current analysis window is equal to the average of all impact intervals within the current analysis window, representing the main period of the sequence corresponding to the current analysis window; The reference rotational cycle is based on the real-time speed of the compressor. Calculated, and ; It is a natural exponential function; This indicates taking the absolute value.

[0048] For example, when the compressor speed =10Hz, or 600 revolutions per minute, then the reference rotation period =1 / 10=0.1 seconds:

[0049] (1) Case 1: Within a 3-second analysis window, 5 impact events were detected, and the impact interval sequence is as follows: The units are all seconds; therefore, the average period of the calculated analysis window is... =0.100 seconds; further calculation yields the rhythm stability index of the analysis window. This value is very close to 1, indicating that the impact sequence corresponding to the analysis window has a very strong inherent regularity and is highly correlated with the machine speed, which is likely due to equipment failure.

[0050] (2) Case 2: In another 3-second analysis window, 5 impact events were also detected, with the impact interval sequence as follows: The units are all seconds; therefore, the average period of the calculated analysis window is... =0.364 seconds; further calculation yielded the rhythm stability index of the analysis window. This value is very close to 0, indicating that the shock sequence corresponding to the analysis window is an irregular random disturbance.

[0051] Thus, by calculating and analyzing the rhythmic stability index of the impact sequence corresponding to the time window, it is possible to distinguish between potential fault signals with periodic patterns and randomly occurring environmental disturbances.

[0052] S3: Extract the time-domain signal window containing each impact event and calculate its energy envelope; calculate the intrinsic damping factor of each energy envelope based on the logarithmic difference between the peak value of the energy envelope and the energy value after a set duration.

[0053] Step S2 can initially filter out a large number of random interferences. However, there may be other periodic interference sources in the environment, such as another nearby device. In order to further confirm that the impact source is the physical process inside the target compressor, this step starts with the morphology of a single impact event for authentication.

[0054] A mechanical impact originating from within the equipment, whether it is a normal valve opening and closing or a faulty crack impact, will excite the compressor structure to produce damped vibrations. Its energy envelope will show a pattern of rapid rise followed by decay according to a specific law. The energy envelope pattern of external acoustic interference is usually different, such as a periodic knocking sound.

[0055] Therefore, a stability threshold of 0.5 is set, and only the rhythm stability index within the analysis window is considered. Impact events that pass the initial screening are considered; for each impact event that passes the initial screening for rhythm stability, a time-domain signal window is extracted before and after its occurrence time, the size of which is equal to... , The preset analysis duration is used; then, the energy envelope of the time-domain signal window is calculated using Hilbert transform. .

[0056] in, Based on the preset analysis duration according to the compressor's structural characteristics, the method is as follows: Under the compressor's healthy state, collect transient impact signals in its background noise, such as the normal opening and closing sound of valves, and analyze the average time required for the energy envelope to decay to the background noise level. Take 80% of this average time as the analysis duration. This ensures that the analysis window covers the main decay process of the impact energy.

[0057] To analyze the energy envelope The decay process starting from the peak introduces an intrinsic damping factor. To assess this physical characteristic, the damped vibration decay process of a mechanical structure can be physically approximated as an exponential decay model. These are the characteristic parameters used to estimate this decay process.

[0058] The intrinsic damping factor of the shock event The calculation formula is as follows:

[0059]

[0060] in: The intrinsic damping factor for shock events; The energy envelope of the shock event The peak value; The energy envelope of the shock event From peak attenuation Energy value after time; This is the preset analysis duration.

[0061] Among them, the intrinsic damping factor Essentially, it is an estimate of the decay slope of the energy envelope on logarithmic coordinates, directly reflecting the rate of energy decay. For a specific compressor, its structure and materials are fixed; therefore, damped vibrations caused by internal mechanical shocks... The values ​​will be distributed within a relatively stable and narrow range.

[0062] For example,

[0063] (1) Case 1: The energy envelope of the impact event is detected. Peak value within the signal window During the preset analysis time Seconds later, energy value The calculated intrinsic damping factor of the impact event =0.916 / 0.005=183.2.

[0064] (2) Case 2: The energy envelope of another impact event is detected. Peak value within the signal window However, its decay is very slow, within the preset analysis time. Seconds later, energy value The calculated intrinsic damping factor of the impact event =0.357 / 0.005=71.4.

[0065] Thus, by calculating and comparing the intrinsic damping factor, this step provides physical source identification for the impact event, effectively eliminating external disturbances that are highly rhythmic but do not conform to the physical form.

[0066] S4: Calculate the fault authenticity score of each impact event based on the energy value of the impact event in the characteristic signal, the rhythm stability index of the corresponding analysis time window, and the intrinsic damping factor; when the fault authenticity score of a set number of impact events exceeds the preset threshold, it is determined that the compressor has abnormal noise.

[0067] Finally, to comprehensively assess whether each impact event is a genuine early fault signal, this step integrates the impact energy, rhythmic stability, and intrinsic physical characteristics to construct a fault authenticity score. A genuine early fault signal should simultaneously meet the following requirements: it must have a certain amount of energy, its appearance must be highly rhythmic, and its physical form must conform to the characteristics of internal mechanical impact. All three are indispensable.

[0068] First, based on the intrinsic damping factor of the shock event Calculate its intrinsic damping characteristic matching factor It is essentially a confidence factor between [0,1], and its specific calculation formula is as follows:

[0069]

[0070] in: The intrinsic damping factor for shock events; and These are the mean and standard deviation derived from health data; Indicates taking the absolute value; This indicates taking the absolute value.

[0071] Among them, the average value and standard deviation The acquisition method is as follows: Acoustic signals of the compressor under various typical operating conditions during long-term healthy operation are collected. These typical operating conditions include, but are not limited to, different speeds and different loads. All impact events are extracted through S1 and S2. The impact events correspond to benign interference or background noise. The intrinsic damping factor of each impact event is calculated. ; Statistical analysis of all intrinsic damping factors The average value, and calibrated as Simultaneously, calculate all intrinsic damping factors. The standard deviation of is , and is calibrated as This is used for determining the confidence interval in subsequent steps.

[0072] Furthermore, the failure authenticity score of the impact event is calculated. The specific calculation formula is as follows:

[0073]

[0074] in: Scoring the authenticity of the failure in the impact event; For impact events in characteristic signals The corresponding energy value; The rhythmic stability index of the analysis window where the impact event occurs; The intrinsic damping characteristic matching factor of the impact event.

[0075] The logic for constructing the fault authenticity score is as follows: the base score is the energy value of the impact event. Then, the baseline score is modulated by two confidence factors; the first factor is the rhythm stability index. If the shock sequence is random, i.e. If the value approaches 0, then even with higher energy levels, the final fault accuracy score will be low. It will also approach 0; the second factor is the intrinsic damping characteristic matching factor. The intrinsic damping factor of the impact event was calculated. Compared with normal benchmark The degree of deviation, when Falling to 3 standard deviations from the health benchmark At that time, It decreases linearly as the deviation increases; when the deviation exceeds this range... It is always equal to 0.

[0076] For example, the baseline of step S3 is followed. , The confidence interval is And the energy value of the impact event :

[0077] (1) Case 1: The rhythmic stability index of the analysis window in which the impact event occurs The intrinsic damping factor of the shock event It fell within the confidence interval. If the value is within the range of 180 and very close to the baseline value, then the intrinsic damping characteristic matching factor of the impact event is calculated. The failure authenticity score of the impact event was further calculated. Ultimately, the system determined that the impact likely originated from inside the device.

[0078] (2) Case 2: The rhythmic stability index of the analysis window in which the impact event occurs The intrinsic damping factor of the shock event It deviated significantly from the confidence interval. Then the intrinsic damping characteristic matching factor of the impact event is calculated. The failure authenticity score of the impact event was further calculated. The score was and Simultaneously, the suppression is 0, and the system ultimately determines it as external interference.

[0079] As the examples above demonstrate, ultimately, a fault authenticity score is only achieved when an impact energy is significant, the rhythm is stable, and the physical characteristics match. Only then will you get a high score.

[0080] Finally, set a reasonable early warning threshold. Warning threshold At lower levels, it has high sensitivity and can detect very early-stage faults, but the risk of false alarms is slightly higher; warning threshold At higher levels, reliability is high and the risk of false alarms is extremely low, but it may miss the golden window for detecting early, minor faults; therefore, the warning threshold... The value range is [0.001, 0.02]. In this embodiment, the warning threshold is... Set to 0.01.

[0081] To prevent false alarms caused by a single, sporadic high-scoring event, this embodiment employs an M-out-of-N early warning strategy: in an evaluation sequence containing M=20 consecutive impact events, if N=5 or more impact events are found to have fault authenticity scores... Exceeded The system then confirms the fault status and triggers an alert.

[0082] Thus, by constructing a comprehensive score that integrates energy, rhythm, and physical characteristics, and adopting an M-out-of-N decision-making mechanism, this method can achieve early fault warning with extremely high reliability and sensitivity, fundamentally solving the problem of high false alarm rate.

[0083] This invention constructs a comprehensive diagnostic framework for compressor abnormal noise that integrates three dimensions: periodic rhythm, physical form, and energy. It abandons the traditional technology that relies solely on the single criterion of energy threshold and proposes rhythmic stability index and intrinsic damping factor as key features. The former is used to identify the periodic pattern of the fault, and the latter is used to verify the internal mechanical properties of the impact source. By integrating these multi-dimensional features into a comprehensive fault authenticity score and combining it with the M-out-of-N decision mechanism, this invention can effectively distinguish between real early equipment faults and complex random field interference, fundamentally solving the problems of high false alarm rate and insufficient reliability of existing technologies.

[0084] For example, for such Figure 2 The diagram shows the feature signals extracted from the sound signal through wavelet packet transform and kurtosis criterion. Figure 3 As shown; Positioning Figure 3 The characteristic signal shown contains impact events, and a schematic diagram of the detected impact events is shown below. Figure 4 As shown; the time-domain signal window containing each impact event is extracted and its energy envelope is calculated. Based on the logarithmic difference between the peak value of the energy envelope and the energy value after a set duration, the intrinsic damping factor of each impact event is calculated. (See schematic diagram below.) Figure 5 As shown; finally, based on the energy value of each impact event in the characteristic signal, the rhythmic stability index of the corresponding analysis time window, and the intrinsic damping factor, the fault authenticity score of each impact event is calculated, as illustrated in the diagram. Figure 6 As shown, around 5.25 seconds, out of 20 consecutive impact events, the fault authenticity scores of 5 impact events exceeded the preset threshold. At this point, it was determined that the compressor had abnormal noise.

[0085] This invention also discloses a compressor abnormal noise detection system based on acoustic spectrum pattern recognition, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a compressor abnormal noise detection method based on acoustic spectrum pattern recognition according to the present invention.

[0086] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition, characterized in that, include: Acquire the compressor's sound signal and speed; The sound signal is subjected to wavelet packet transform to obtain a reconstructed signal with multiple frequency bands; Calculate the kurtosis value of the reconstructed signal, and take the reconstructed signal with the largest kurtosis value as the feature signal; Within the set analysis time window, the occurrence time of the impact event contained in the feature signal is located and the time interval between adjacent impact events is calculated; based on all time intervals within the analysis time window, the rhythm stability index is calculated. The rhythm stability index is negatively correlated with the sum of the differences between each time interval and the average time interval within the analysis time window, and positively correlated with the reciprocal of the compressor speed. Extract the time-domain signal window containing each impact event and calculate its energy envelope; Calculate the intrinsic damping factor for each energy envelope. The intrinsic damping factor is directly proportional to the logarithmic difference between the peak value of the corresponding energy envelope and the energy value after a set duration, and inversely proportional to the set duration. The fault authenticity score of each impact event is calculated. The fault authenticity score is positively correlated with the product of the energy value of the corresponding impact event in the characteristic signal, the rhythm stability index of the corresponding analysis time window, and the intrinsic damping factor. When the fault authenticity scores of a set number of impact events exceed a preset threshold, it is determined that the compressor has abnormal noise.

2. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, The method for calculating the rhythm stability index is as follows: ； in: This is the rhythm stability index for the current analysis window; This represents the total number of impact events detected within the current analysis window; For the current analysis window The first shock event and the first The impact interval of a single impact event; The average period of the current analysis window is equal to the average of all impact intervals within the current analysis window; The reference rotation period, and , This refers to the real-time speed of the compressor. It is a natural exponential function; This indicates taking the absolute value.

3. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, The method for calculating the intrinsic damping factor is as follows: ； in: The intrinsic damping factor for shock events; The energy envelope of the shock event The peak value; The energy envelope of the shock event From peak attenuation Energy value after time; This is the preset analysis duration.

4. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 3, characterized in that, The fault authenticity score is also positively correlated with the intrinsic damping feature matching factor, which is calculated as follows: ； in: The intrinsic damping factor for shock events; and These are the mean and standard deviation derived from health data; Indicates taking the absolute value; This indicates taking the absolute value.

5. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 4, characterized in that, The method for calculating the fault authenticity score is as follows: ； in: Scoring the authenticity of the failure in the impact event; For impact events in characteristic signals The corresponding energy value; The rhythmic stability index of the analysis window where the impact event occurs; The intrinsic damping characteristic matching factor of the impact event.

6. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, Locating the time of occurrence of the impact event contained in the characteristic signal includes: The Teager-Kaiser energy operator is applied to the characteristic signal to obtain an energy sequence; Peak detection is performed on the energy sequence to pinpoint the occurrence time of all impact events.

7. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, Obtaining the energy envelope includes: For the time-domain signal window centered on the moment of the impact event in the characteristic signal, the energy envelope of the time-domain signal window is calculated by applying the Hilbert transform.

8. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, The method further includes: calculating the intrinsic damping factor and fault authenticity score only for impact events within the analysis window where the rhythm stability index is greater than 0.

5.

9. The method for detecting abnormal noise in a compressor based on acoustic spectrum pattern recognition according to claim 1, characterized in that, The compressor is identified as making abnormal noises, including: In an evaluation sequence containing 20 consecutive impact events, if the failure authenticity score of 5 or more impact events exceeds the preset threshold, the compressor is determined to have abnormal noise.

10. A compressor abnormal noise detection system based on acoustic spectrum pattern recognition, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a compressor abnormal noise detection method based on acoustic spectrum pattern recognition according to any one of claims 1-9.

Images