Intelligent fault diagnosis method and system for dynamic voiceprint of valve cooling equipment

Abstract

The invention belongs to the technical field of industrial equipment state monitoring and fault diagnosis, and particularly discloses an intelligent fault diagnosis method and system for dynamic voiceprint of valve cooling equipment, and the method comprises the steps: fusing sound and vibration dual-mode signals, taking a strong correlation vibration signal as a weight to enhance voiceprint features, and effectively amplifying a homologous fault signal; the method comprises the following steps: binding physical space coordinates of a sensor in a data acquisition stage, reserving position information in a whole diagnosis process, and synchronously outputting a fault type and sensor coordinates of an abnormal signal source when a system identifies fault features; a dynamic signal processing strategy is adopted, low-frequency voiceprints are focused in a high-vibration period to capture structural faults, and high-frequency noise is suppressed in a low-vibration period to reduce environmental interference; through real-time dynamic segmentation analysis of equipment operation data, a complete dynamic process of a fault from germination to development is captured, decision support is provided for predictive maintenance, and non-planned shutdown is effectively avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial equipment state monitoring and fault diagnosis, and relates to an intelligent fault diagnosis method and system for dynamic voiceprints of valve cooling equipment. BACKGROUND

[0002] As a key component in industrial production, the running state of valve cooling equipment directly affects the stability and safety of the entire system. In the long-term high-load operation process, valve cooling equipment will inevitably have various faults such as leakage, jamming and abnormal vibration due to wear, fatigue or medium impact. If these early faults are not discovered and handled in time, it will lead to performance degradation, energy consumption increase, and even equipment damage, production interruption and safety accidents, causing huge economic losses and safety hazards.

[0003] The traditional fault diagnosis method of valve cooling equipment mainly relies on manual inspection and regular maintenance. The inspection personnel make subjective judgments on the equipment through listening and touching, which is not only inefficient and labor-intensive, but also the diagnosis result is seriously dependent on the personal experience of the inspection personnel, lacking objectivity and consistency. Regular maintenance usually adopts a "one-size-fits-all" strategy, replacing parts according to the plan regardless of the equipment state, which is easy to cause unnecessary waste and cannot effectively prevent the occurrence of sudden faults. Some automatic monitoring systems only use single temperature or pressure sensors, which are difficult to fully reflect the complex running state of the equipment.

[0004] Some existing monitoring technologies based on a single signal, such as pure vibration analysis or sound analysis, have obvious drawbacks in complex industrial site environments. Industrial sites are usually full of strong background noise and vibration interference from other equipment, which will drown out the effective fault characteristic signals, resulting in low signal-to-noise ratio of data collected by single sensors, and easy to produce false positives or false negatives. In addition, some specific faults, such as small leaks, have very weak signal characteristics, which are difficult to effectively capture and identify by a single signal source, thereby limiting the accuracy and reliability of fault diagnosis. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the technical scheme is as follows: an intelligent fault diagnosis method for dynamic voiceprints of valve cooling equipment, comprising: S1, obtaining equipment running data: activating a sound sensor and a vibration sensor, synchronously reading real-time sound signals and vibration signals, and outputting original sound stream and original vibration stream.

[0006] S2, extracting sound state features: processing the original sound stream, segmenting into continuous period sound data segments, and outputting a device voiceprint state sequence.

[0007] S3, convert vibration energy parameters: process the original vibration flow, calculate the time domain intensity value of acceleration data, and output the device vibration intensity sequence.

[0008] S4, perform double signal space-time alignment: input the device voiceprint state and the device vibration intensity, match the time scale and the physical position coordinates, and output the space-time aligned voiceprint-vibration combination parameters.

[0009] S5, generate a joint vibration signal: apply the device vibration intensity to the device voiceprint state to perform weighted fusion, and output a joint vibration signal containing vibration-sound coupling features.

[0010] S6, diagnose fault state: compare the joint vibration signal with the preset fault feature library, output the fault type and positioning coordinates, and generate a diagnostic report containing the fault type and positioning coordinates.

[0011] The second aspect of the present application provides an intelligent fault diagnosis system for dynamic voiceprints of valve cooling equipment, comprising the following contents: a device operation data acquisition module, which activates a sound sensor and a vibration sensor, synchronously reads real-time sound signals and vibration signals, and outputs original sound flow and original vibration flow.

[0012] A sound state feature extraction module processes the original sound flow, divides it into continuous time period sound data segments, and outputs a device voiceprint state sequence.

[0013] A vibration energy parameter conversion module processes the original vibration flow, calculates the time domain intensity value of acceleration data, and outputs a device vibration intensity sequence.

[0014] A double signal space-time alignment module inputs the device voiceprint state and the device vibration intensity, matches the time scale and the physical position coordinates, and outputs the space-time aligned voiceprint-vibration combination parameters.

[0015] A joint vibration signal generation module applies the device vibration intensity to the device voiceprint state to perform weighted fusion, and outputs a joint vibration signal containing vibration-sound coupling features.

[0016] A fault state diagnosis module compares the joint vibration signal with the preset fault feature library, outputs the fault type and positioning coordinates, and generates a diagnostic report containing the fault type and positioning coordinates.

[0017] Compared with the prior art, the present application has the following advantages: (1) the present application fuses sound and vibration dual-mode signals, utilizes the direct reflection characteristics of vibration signals on structural abnormalities and the high sensitivity of sound signals to fluid leakage and other specific faults, enhances the voiceprint features by using strongly correlated vibration signals as weights, effectively amplifies the homologous fault signals, helps to improve the fault feature clarity of the joint signal, greatly reduces the false positive and false negative rates, and thus enhances the accuracy and reliability of the diagnosis.

[0018] (2) The application binds the sensor physical space coordinates in the data acquisition stage, and retains the position information in the whole diagnosis process. When the system identifies the fault characteristics, the fault type and the sensor coordinates of the abnormal signal source are output synchronously, so that the maintenance personnel can directly lock the fault physical area, significantly shorten the troubleshooting time and reduce the maintenance cost.

[0019] (3) The application adopts a dynamic signal processing strategy, which adaptively adjusts the signal processing mode according to the vibration intensity: focusing on low-frequency voiceprints in high-vibration periods to capture structural faults, and suppressing high-frequency noise in low-vibration periods to reduce environmental interference. This adaptive noise reduction and feature enhancement mechanism enables the system to effectively cope with changes in working conditions and noise fluctuations, ensuring stable and reliable diagnostic performance under different operating conditions, and significantly improving the adaptability and robustness of the system to complex industrial environments.

[0020] (4) The application captures the complete dynamic process from the inception to the development of the fault through real-time dynamic segmentation and analysis of equipment operation data, such as the weak rise of acoustic characteristics in the early stage of leakage or the periodic pulses in the early stage of bearing wear. This sensitive perception of early weak features enables the system to issue an early warning before the fault causes serious consequences, providing decision support for predictive maintenance and effectively avoiding unplanned downtime. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 The method embodiment of the application is shown in the flowchart.

[0023] Figure 2 The system module connection diagram of the application is shown. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only some embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0025] Embodiment one

[0026] Please refer to Figure 1As shown, the application proposes an intelligent fault diagnosis method for dynamic voiceprint of valve cooling equipment, which comprises the following steps: S1, obtaining equipment operation data: activating sound sensors and vibration sensors, synchronously reading real-time sound signals and vibration signals, and outputting original sound stream and original vibration stream.

[0027] In a preferred embodiment, the step of obtaining equipment operation data specifically comprises the following steps: S1-1, collecting continuous sound waveform data during equipment operation by sound sensors.

[0028] Specifically, the sound sensor is defined as a device for converting sound signals into electrical signals, which is installed on the side of the valve cooling equipment and used to continuously record sound waveform data during equipment operation.

[0029] The valve cooling equipment includes but is not limited to cooling valves, coolers / heat exchangers, pumps and pipeline systems.

[0030] The sound waveform data is represented as a time series, wherein each data point contains a sound amplitude value, and 44100 points are sampled per second. The sampling rate is set according to the industrial audio acquisition standard 44.1 kHz, which is based on the test verification of the common sound frequency range 20Hz-20kHz of valve cooling equipment, to ensure that key sound characteristics can be completely captured.

[0031] S1-2, synchronously collecting three-axis acceleration data of the equipment shell by vibration sensors and integrating them into vibration data.

[0032] Specifically, the vibration sensor is defined as a piezoelectric device for measuring acceleration, which is installed on the shell of the valve cooling equipment and used to collect acceleration data of the equipment shell in three orthogonal directions x, y and z in the standard coordinate system. The vibration sensor synchronously collects data at a sampling rate of 1000 points per second. The sampling rate is set according to the requirement of equipment vibration frequency analysis, based on the Nyquist sampling theorem and industrial test data, to ensure coverage of the 0-500Hz vibration range.

[0033] S1-3, aligning the time stamps of the continuous sound waveform data and the vibration data to generate time-synchronized original sound stream and original vibration stream.

[0034] Specifically, after obtaining the sound waveform data and the vibration data, the time stamps of each data point are extracted, the time stamp sequences are compared, and a deviation threshold of 0.1 milliseconds is set. The threshold is set according to the industrial sensor response time specification and determined by 50 sets of valve cooling equipment test data to ensure that the alignment accuracy meets the subsequent processing requirements. If the time deviation between the sound waveform data and the vibration data exceeds 0.1 milliseconds, a linear interpolation method is applied to adjust the data sequence. The linear interpolation is based on the estimation of intermediate values between adjacent data points to eliminate time deviation. After alignment, the original sound stream and the original vibration stream are generated.

[0035] The original sound stream is a time-aligned continuous sound waveform sequence, and the original vibration stream is a time-aligned three-axis acceleration sequence, both of which share the same time scale.

[0036] For example, in one embodiment, the sound sensor outputs continuous sound waveform data at a sampling rate of 44.1 kHz, with an amplitude range of -1.0V to 1.0V; the vibration sensor outputs three-axis acceleration data at a sampling rate of 1 kHz, with a range of ±2g. After aligning the time stamps, a maximum deviation of 0.05 milliseconds (below the 0.1 millisecond threshold) is detected, and no interpolation is required. The original sound stream contains 10,000 sampling points, and the original vibration stream contains 1,000 sampling points (each containing x, y, z acceleration values), and the time series are completely synchronized. A person skilled in the art can directly reproduce the subsequent steps based on this output, verifying the effectiveness of the generation of the original sound stream and the original vibration stream.

[0037] S2, extracting sound state features: processing the original sound stream, segmenting into continuous time period sound data segments, and outputting device voiceprint state sequences.

[0038] In a preferred embodiment, the processing of the original sound stream, the segmentation into continuous time period sound data segments, and the output of the device voiceprint state sequence comprise: S2-1, dividing the original sound stream into multiple continuous equal-length time pieces of sound data segments.

[0039] Specifically, after receiving the original sound stream, a fixed time window length of 0.1 seconds is set as the segmentation reference, and the setting is based on valve cooling equipment fault voiceprint feature change rate analysis, verified by 100 groups of industrial measurement data, 0.1 second window can capture more than 90% of fault feature mutation; the original sound stream is divided into continuous segments in time sequence, each segment contains 4410 sampling points, and the calculation is based on the original sound stream sampling rate 44100 Hz x 0.1 seconds.

[0040] S2-2, performing frequency analysis on each sound data segment to extract a list of main frequency band amplitude values. The specific steps are as follows: the time domain sound waveform is converted into frequency domain energy distribution through Fourier transform mathematical method, Fourier transform can decompose complex time domain signal into a series of sine and cosine waves of different frequencies, so as to obtain the frequency domain representation of the signal.

[0041] In the frequency domain, the three continuous frequency bands with the highest energy are identified as the main frequency bands, which contain the key sound feature information generated during equipment failure. The main frequency band setting is based on the statistics of 200 groups of valve cooling equipment fault voiceprint database, and the fault features are concentrated in the three main frequency bands.

[0042] For each identified main frequency band, the arithmetic mean of the amplitudes of all frequency points in the band is calculated, and the mean represents the average intensity of the sound signal in the corresponding frequency band.

[0043] For each sound data segment, a main frequency band amplitude value list is output, which contains the average amplitude values of the three main frequency bands, and the elements in the list are arranged in descending order of frequency band energy.

[0044] By outputting the main frequency band amplitude value list of each time segment, the energy changes of the key frequency components in the device sound signal at different time points can be intuitively seen, thereby providing an important basis for subsequent fault diagnosis.

[0045] S2-3, the main frequency band amplitude values are summarized in time sequence to generate a device voiceprint state sequence.

[0046] Specifically, 1 second is set as the current processing period, which contains 10 time segments, and the amplitude value list of each segment is concatenated in time sequence. The continuous 10 lists are stored through a data buffer to form a three-dimensional matrix structure in the time dimension: the first dimension represents the time segment number, the second dimension represents the main frequency band number, and the third dimension stores the amplitude value. The matrix is the device voiceprint state, which directly reflects the dynamic characteristics of the evolution of voiceprint features over time.

[0047] For example, the input original sound stream contains 10000 sampling points (about 0.2268 seconds), which are divided into two complete time segments (each 0.1 second), and the remaining 0.0268 seconds of data is discarded. The first segment frequency analysis extracts the main frequency band amplitude value list as 0.5 volts, 0.3 volts, and 0.2 volts; the second segment extracts 0.6 volts, 0.25 volts, and 0.15 volts. The device voiceprint state is generated by summarizing: a dynamic sequence containing two time points, and the sequence elements are three-element amplitude lists. Those skilled in the art can reproduce the subsequent steps based on this output to verify the integrity and effectiveness of the voiceprint state generation.

[0048] S3, convert vibration energy parameters: process the original vibration stream, calculate the time domain intensity value of the acceleration data, and output the device vibration intensity sequence.

[0049] In a preferred embodiment, the processing of the original vibration stream, the calculation of the time domain intensity value of the acceleration data, and the output of the device vibration intensity sequence include: S3-1, based on the original vibration stream, a synthesized vector amplitude sequence is calculated.

[0050] Specifically, the original vibration flow is defined as a time-aligned three-axis acceleration sequence. For each sampling point's timestamp, the acceleration values of the x-axis, y-axis, and z-axis are extracted, and a mathematical formula is applied to calculate the synthetic vector amplitude, which is equal to the sum of the square of the x-axis acceleration, the square of the y-axis acceleration, and the square of the z-axis acceleration, and then the square root of the sum is taken.

[0051] The synthetic vector amplitude represents the overall vibration intensity in space, with the same unit as the input acceleration (g). The calculation principle is based on the vector synthesis rule in Newtonian mechanics, which is used to simplify three-axis data into a single intensity indicator.

[0052] The calculation is performed in real time by an embedded microprocessor, and the output is a sequence of synthetic vector amplitudes consistent with the original vibration flow timestamps.

[0053] S3-2, in the time period matching the sound data segment, the values of the synthetic vector amplitude sequence are accumulated to generate a vibration intensity sequence matching the time period of the sound data segment.

[0054] Specifically, the accumulation operation steps are as follows: all values belonging to the same time segment are extracted from the sequence, and then all values within the segment are added to obtain the accumulated value, which represents the cumulative vibration energy intensity in that period. The calculation process is realized through an accumulator register, and a single accumulated value for each time segment is output.

[0055] The accumulated values of each time segment are arranged in time sequence to form a sequence, and the time points of the sequence are completely aligned with the time segments of the device voiceprint state, i.e. each 0.1 second period corresponds to a value. This sequence is the device vibration intensity.

[0056] For example, the input original vibration flow contains 1000 sampling points (sampling rate 1 kHz, duration 1 second). The synthetic vector amplitude sequence (1000 points, unit g) is calculated. The time segments are divided: 100 sampling points correspond to a 0.1 second segment, and a total of 10 complete segments are calculated (1 second / 0.1 second = 10). The vector amplitude values are accumulated in each segment: for example, the first segment has an accumulated value of 120g, and the second segment has an accumulated value of 110g (data depends on actual measurement). The device vibration intensity sequence is generated: a sequence containing 10 elements, each element corresponding to the accumulated value of a time segment, matching the time period of the voiceprint state sequence. Those skilled in the art can directly reproduce the subsequent steps based on this output, verifying the completeness and effectiveness of the generation of the device vibration intensity sequence.

[0057] S4, perform double signal space-time alignment: input device voiceprint state and device vibration intensity, match time scale and physical location coordinates, output space-time aligned voiceprint-vibration combination parameters.

[0058] In a preferred embodiment, the matching time scale and physical position coordinates, outputting the spatio-temporally aligned voiceprint-vibration combination parameter, comprises: S4-1, checking the maximum time deviation value of the device voiceprint state sequence and the device vibration intensity sequence.

[0059] Specifically, the device voiceprint state sequence is a time sequence, each element corresponds to a 0.1-second time segment, and contains a list of main frequency band amplitude values; the device vibration intensity is a time sequence, each element corresponds to the same 0.1-second time segment, and contains an accumulated value.

[0060] The start time difference and the end time difference of the two sequences are calculated to determine the maximum time deviation value, and the time deviation threshold is set to 0.05 milliseconds, and the setting basis is: based on statistical analysis of 50 sets of valve cooling device operation data, to ensure the subsequent fusion processing accuracy, the measured data shows that this threshold can cover more than 95% of the working conditions.

[0061] If the maximum time deviation value is less than or equal to the set time deviation threshold, directly proceed to step S4-3; otherwise, proceed to step S4-2.

[0062] S4-2, when the maximum time deviation value exceeds the set time deviation threshold, interpolate the data.

[0063] Specifically, when the maximum time deviation value exceeds the set time deviation threshold, a linear interpolation method is applied to adjust the sequence, the linear interpolation estimates the intermediate value based on adjacent data points, and the mathematical principle is to calculate the new value according to the time interval ratio. For example, data points are added to the shorter sequence to make the time points of the two sequences completely consistent. The interpolation operation is performed in a digital signal processor, and the corrected device voiceprint state sequence and device vibration intensity sequence are output to ensure the alignment of the time scale.

[0064] S4-3, define the fixed physical position coordinates of the sound sensor and the vibration sensor, bind the device voiceprint state data and the device vibration intensity data of each time segment to the fixed physical position coordinates of the corresponding sensor, and form the spatio-temporally aligned voiceprint-vibration combination parameter.

[0065] Specifically, the position coordinates are determined based on the equipment installation drawing and stored as three-dimensional coordinate values (x, y, z), for example, the sound sensor coordinates (0, 0, 0) and the vibration sensor coordinates (0.5, 0, 0). The coordinate setting basis is the structure layout of the valve cooling device, and is calibrated by a three-dimensional measuring instrument.

[0066] The combination parameter includes a timestamp, a device voiceprint state sequence, a device vibration intensity sequence, a sound sensor coordinate, and a vibration sensor coordinate.

[0067] For example, the input device voiceprint state sequence contains 10 time segments (0.1 seconds each), such as the first segment amplitude list 0.5 volts, 0.3 volts, 0.2 volts; the device vibration intensity sequence contains 10 elements, such as the first segment cumulative value 120 g. Check the maximum time deviation value: the start time is the same, the end time is the same, the maximum time deviation is 0.01 milliseconds (lower than the threshold 0.05 milliseconds), no interpolation is required. Correlation position coordinates: sound sensor position (0, 0, 0), vibration sensor position (0.5, 0, 0). Output the spatiotemporally aligned voiceprint-vibration combination parameter sequence: each element contains a timestamp, a voiceprint amplitude list (such as the first element [0.5, 0.3, 0.2] volts), a vibration intensity value (such as the first element 120 g), and a position coordinate pair ((0, 0, 0) and (0.5, 0, 0)). Those skilled in the art can directly reproduce the subsequent steps according to this output, verifying the integrity and effectiveness of the generation of the spatiotemporally aligned voiceprint-vibration combination parameters.

[0068] S5, generate a joint vibration signal: apply device vibration intensity to device voiceprint state to perform weighted fusion, output a joint vibration signal containing vibration-sound coupling features.

[0069] In a preferred embodiment, the application device vibration intensity to device voiceprint state to perform weighted fusion, output a joint vibration signal containing vibration-sound coupling features, specifically: S5-1, take the vibration intensity in the spatiotemporally aligned voiceprint-vibration combination parameter as a weight factor, amplify the homologous features in the voiceprint state by time segment, and generate a weighted amplified voiceprint state sequence.

[0070] Specifically, receive the spatiotemporally aligned voiceprint-vibration combination parameter, multiply the device vibration intensity value in each time segment of the parameter by the amplitude value list of the main frequency band in the device voiceprint state as a weight factor, indicating that when the vibration intensity of a certain time segment is large, the voiceprint features (amplitude value) in that time segment are amplified accordingly, and then output the weighted amplified voiceprint state sequence, wherein the signal intensity of the homologous features is enhanced.

[0071] The amplification operation is realized by a digital multiplier circuit, and the mathematical principle is to multiply each amplitude value by the vibration intensity value of the corresponding time segment. For example, the amplitude value of a certain time segment is 0.5 volts and the vibration intensity is 120 g, so the amplified amplitude value is 60 volts·g.

[0072] The homologous features refer to features generated by the same device event within the same time segment of the sound signal and the vibration signal, which can enhance the associated signal intensity after amplification, and help improve the accuracy of subsequent fault diagnosis.

[0073] S5-2, according to the weighted amplified voiceprint state sequence, when the vibration intensity exceeds the preset vibration intensity threshold, the low frequency voiceprint amplitude is weighted and focused, and the voiceprint state sequence after low frequency focusing is generated.

[0074] Specifically, the vibration intensity threshold is set to 200g, and the basis is set as follows: based on statistical analysis of 300 sets of valve cooling equipment operation data, 95% of normal vibration is lower than the value.

[0075] When the equipment vibration intensity value exceeds 200g, the low frequency range is defined as 0-150Hz, and the basis is set as follows: analysis of valve cooling equipment leakage fault characteristic frequency, 100 sets of measured data verify that the leakage fault mainly generates low frequency signals.

[0076] For the weighted amplified voiceprint state sequence, the low frequency component in the main frequency band range completely contained in 0-150Hz is identified, and its amplitude value is multiplied by a focusing coefficient 1.5, and the focusing coefficient is determined by industrial test, which can optimize the fault detection rate.

[0077] The voiceprint sequence after low frequency focusing is output, wherein the amplitude value of the low frequency component is enhanced.

[0078] When the equipment vibration intensity exceeds the preset vibration intensity threshold, the amplitude value belonging to the preset low frequency range in the main frequency band amplitude value is gain amplified, the gain amplification is combined with the weight factor, and the low frequency fault acoustic characteristics coupled with high intensity mechanical vibration are synergistically enhanced, so that the detection signal-to-noise ratio of the specific fault mode is improved.

[0079] S5-3, according to the weighted amplified voiceprint state sequence, the high frequency noise fluctuation in the period when the vibration intensity is lower than the preset vibration intensity threshold is suppressed, and the voiceprint state sequence after high frequency suppression is generated.

[0080] Specifically, when the equipment vibration intensity value is lower than 200g, the high frequency range is defined as greater than 1500Hz, and the basis is set as follows: analysis of valve cooling equipment environmental noise spectrum, 80 sets of background noise are mainly distributed in the high frequency band.

[0081] For the weighted amplified voiceprint state sequence, the high frequency component in the main frequency band with a frequency greater than 1500Hz is identified, and its amplitude value is multiplied by a suppression coefficient 0.3, and the suppression coefficient is determined by noise attenuation experiment, which can reduce noise interference while retaining effective signals.

[0082] The suppression operation is realized by a digital attenuator circuit, and then the voiceprint sequence after high frequency suppression is output, wherein the amplitude value of the high frequency noise is attenuated.

[0083] The application attenuates and suppresses the amplitude values in the preset high frequency range among the main frequency band amplitude values when the device vibration intensity is lower than the preset vibration intensity threshold. The attenuation and suppression and gain amplification together constitute an adaptive filtering mechanism driven by the device vibration intensity, which cooperatively suppresses the high frequency environmental noise generated by non-device failure sources, thereby reducing the false positive rate of the diagnostic system.

[0084] S5-4, reorganize the voiceprint state sequence processed by steps S5-1 to S5-3 in time sequence, and fuse to generate a joint vibration signal containing vibration-sound coupling characteristics.

[0085] Specifically, the voiceprint state sequence processed by sub-steps S5-1 to S5-3 is reorganized in time sequence, and the fusion process retains the vibration intensity weighting feature, the low frequency focusing feature and the high frequency suppression feature to form a vibration-sound coupling characteristic set, and then outputs a joint vibration signal through a data integrator. The signal is in time sequence form, and each element contains the amplitude values of the three main frequency bands after processing.

[0086] The weighted focused low frequency voiceprint amplitude is combined with the suppressed high frequency noise fluctuation to generate a joint signal that integrates sound and vibration information, providing more comprehensive and accurate data support for subsequent fault diagnosis.

[0087] For example, the input voiceprint-vibration combined parameter sequence has 10 time segments. The fifth segment has a device vibration intensity of 250g (above the 200g threshold), an original voiceprint state low frequency amplitude of 0.5 volts (frequency band 50-100Hz), a weighted value of 75 volts·g, and a focused value of 112.5 volts·g. The first segment has a vibration intensity of 120g (below the threshold), a high frequency amplitude of 0.2 volts (frequency band 2000-2500Hz), a weighted value of 24 volts·g, and a suppressed value of 7.2 volts·g. The fusion generates a joint vibration signal sequence: 10 time segments, each containing three processed amplitude values (such as the first segment [60 volts·g, 36 volts·g, 7.2 volts·g]). A person skilled in the art can directly reproduce the subsequent steps based on this output to verify the effectiveness and noise reduction enhancement effect of the joint vibration signal generation.

[0088] S6, diagnose the fault state: compare the joint vibration signal with the preset fault feature library, output the fault type and positioning coordinates, and generate a diagnostic report containing the fault type and positioning coordinates.

[0089] In a preferred embodiment, the comparison of the joint vibration signal with the preset fault feature library and the output of the fault type and positioning coordinates comprise: S6-1.1, detecting the amplitude change rate of adjacent time segments in the joint vibration signal, combining with the preset mutation judgment threshold, determining the amplitude mutation period and frequency band, and integrating into a mutation period list.

[0090] Specifically, the amplitude mutation determination threshold is set to 40%, and the basis is that, based on 200 sets of valve cooling equipment fault data statistics, the amplitude fluctuation under normal working conditions is less than 30%.

[0091] For each main frequency band amplitude value, the change rate of the adjacent time segment is calculated, and the formula is the absolute value of the current amplitude value minus the amplitude value of the previous time segment divided by the amplitude value of the previous time segment.

[0092] When the amplitude change rate exceeds 40%, the corresponding time segment and frequency band sequence number are recorded, that is, the amplitude mutation time period and frequency band.

[0093] S6-1.2, match the mutation time period list with the fault mode stored in the preset fault feature library, when the similarity meets the preset matching condition, confirm that the matching is successful, and determine the fault type according to the leakage fault feature and the bearing wear fault feature.

[0094] Specifically, the preset fault feature library is stored in a read-only memory, and contains two typical fault features, i.e. leakage fault feature and bearing wear fault feature.

[0095] Among them, the leakage fault feature is that the rising amplitude of low-frequency band amplitude on three consecutive time segments exceeds 50%, and the setting basis is that 50 times of leakage experiment shows that liquid leakage will cause continuous enhancement of low-frequency vibration energy; the bearing wear fault feature is that the amplitude of medium-frequency band appears periodic pulse, the pulse interval is 0.3 seconds and the peak value exceeds the average value by 100%, and the setting basis is that 30 sets of bearing wear case statistics show that the impact of wear particles will produce high-energy pulse of fixed frequency.

[0096] The preset matching condition is that when the rising amplitude of low-frequency amplitude of three consecutive time segments reaches 55%, it is matched as a leakage fault feature; when the fixed interval pulse is detected and the peak value exceeds the average value by 110%, it is matched as a bearing wear feature.

[0097] S6-1.3, mark the sensor coordinate position associated with the fault type.

[0098] Specifically, for the matched fault type, the corresponding position coordinate pair is extracted, including the sound sensor coordinate and the vibration sensor coordinate.

[0099] The marking operation is to associate the fault type with the position coordinates of the corresponding time segment, generate a fault event record through a data marker, and each record contains four items: fault type name, time segment start time, sound source three-dimensional coordinates, and vibration source three-dimensional coordinates.

[0100] In a further preferred embodiment, the diagnostic report is generated to include fault type and location coordinates, wherein: all fault event records are acquired, arranged in chronological order to generate the diagnostic report, and ensure that the fault events are displayed in the order of occurrence.

[0101] The diagnostic report is in a table structure: the first column is the time stamp, the second column is the fault type, the third column is the sound sensor coordinates, the fourth column is the vibration sensor coordinates, and an example is as follows:

[0102] Table 1, diagnostic report example:

[0103] Time stamp (ms) Fault type Sound sensor coordinates (m) Vibration sensor coordinates (m) 5000 Leakage (1.2,3.5,0.8) (1.2,3.5,0.0) 12000 Bearing wear (2.1,4.0,1.0) (2.1,4.0,0.0)

[0104] The report is output through an industrial display and stored in a non-volatile memory, and the device alarm indicator is triggered after the output is completed.

[0105] For example, the input joint vibration signal sequence (10 time segments) is detected, the low-frequency amplitude value of the fifth segment is 112.5 volts·g, the fourth segment is 75 volts·g, the change rate is 50% above the threshold; the rising amplitude of the third to fifth segments reaches 55%, matching the leakage fault characteristics, and the fault type is marked as leakage, and the position coordinates are (0, 0, 0) and (0.5, 0, 0). The diagnostic report is generated: time stamp 0.4-0.5 seconds, fault type leakage, sound source (0, 0, 0), vibration source (0.5, 0, 0). Those skilled in the art can verify the fault location accuracy according to this report to meet the closed evidence specification required by the Patent Law.

[0106] Example two

[0107] Please refer to Figure 2 As shown, based on the basis of example one, the second aspect of the present application provides an intelligent fault diagnosis system for dynamic soundprint of valve cooling equipment, which comprises: an equipment operation data acquisition module, a sound state feature extraction module, a vibration energy parameter conversion module, a dual signal space-time alignment module, a joint vibration signal generation module and a fault state diagnosis module.

[0108] The equipment operation data acquisition module is connected with the sound state feature extraction module and the vibration energy parameter conversion module, the sound state feature extraction module is connected with the vibration energy parameter conversion module, the dual signal space-time alignment module is connected with the sound state feature extraction module, the vibration energy parameter conversion module and the joint vibration signal generation module, and the joint vibration signal generation module is connected with the fault state diagnosis module.

[0109] The equipment operation data acquisition module activates the sound sensor and the vibration sensor, synchronously reads the real-time sound signal and the vibration signal, and outputs the original sound stream and the original vibration stream.

[0110] The sound state feature extraction module processes the original sound stream, divides it into sound data segments of continuous time periods, and outputs a device voiceprint state sequence.

[0111] The vibration energy parameter conversion module processes the original vibration stream, calculates the time-domain intensity value of the acceleration data, and outputs a device vibration intensity sequence.

[0112] The dual-signal space-time alignment module inputs the device voiceprint state and the device vibration intensity, matches the time scale and the physical position coordinates, and outputs a voiceprint-vibration combined parameter aligned in space-time.

[0113] The joint vibration signal generation module applies the device vibration intensity to perform weighted fusion on the device voiceprint state, and outputs a joint vibration signal containing vibration-sound coupling features.

[0114] The fault state diagnosis module compares the joint vibration signal with a preset fault feature library, outputs a fault type and a positioning coordinate, and generates a diagnosis report containing the fault type and the positioning coordinate.

[0115] The above is merely an example and a description of the present application, and those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the concept of the present application or exceed the scope defined by the present application.

Claims

1. An intelligent fault diagnosis method for dynamic voice prints of valve cooling equipment, characterized in that, The method comprises the following steps: S1, obtaining device running data: activating sound sensors and vibration sensors, synchronously reading real-time sound signals and vibration signals, and outputting original sound streams and original vibration streams; S2, extracting sound state features: processing original sound streams, segmenting sound data segments in continuous time periods, and outputting device voiceprint state sequences; S3, converting vibration energy parameters: processing original vibration streams, calculating time-domain intensity values of acceleration data, and outputting device vibration intensity sequences; S4, performing double-signal space-time alignment: inputting device voiceprint states and device vibration intensities, matching time scales and physical position coordinates, and outputting space-time aligned voiceprint-vibration combination parameters; S5, generating a joint vibration signal: applying device vibration intensities to device voiceprint states to perform weighted fusion, and outputting a joint vibration signal containing vibration-sound coupling features; S6, diagnosing fault states: comparing the joint vibration signal with a preset fault feature library, outputting fault types and positioning coordinates, and generating a diagnostic report containing fault types and positioning coordinates.

2. The intelligent fault diagnosis method for dynamic voice prints of a valve cooling device according to claim 1, characterized in that, The method for obtaining device running data specifically comprises the following steps: collecting continuous sound waveform data during device operation through sound sensors; synchronously collecting three-axis acceleration data of the device shell through vibration sensors and integrating the data into vibration data; aligning the time stamps of the continuous sound waveform data and the vibration data to generate time-synchronized original sound streams and original vibration streams.

3. The method according to claim 1, wherein, The method for processing original sound streams, segmenting sound data segments in continuous time periods, and outputting device voiceprint state sequences comprises the following steps: cutting the original sound stream into multiple continuous and equal-length time pieces of sound data segments; performing frequency analysis on each sound data segment to extract a list of main frequency band amplitude values; sequentially collecting all main frequency band amplitude values to generate a device voiceprint state sequence.

4. The intelligent fault diagnosis method for dynamic voice prints of a valve cooling device according to claim 3, characterized in that, The method for processing original vibration streams, calculating time-domain intensity values of acceleration data, and outputting device vibration intensity sequences comprises the following steps: calculating a synthetic vector amplitude sequence based on the original vibration stream; accumulating the values of the synthetic vector amplitude sequence within a time period matching the sound data segment to generate a vibration intensity sequence matching the sound data segment time period.

5. The intelligent fault diagnosis method for dynamic voice prints of a valve cooling device according to claim 1, characterized in that, The method for matching time scales and physical position coordinates to output space-time aligned voiceprint-vibration combination parameters comprises the following steps: checking the maximum time deviation value of the device voiceprint state sequence and the device vibration intensity sequence; interpolating and completing the data when the maximum time deviation value exceeds a set time deviation threshold; defining fixed physical position coordinates of the sound sensors and the vibration sensors, binding the device voiceprint state data and the device vibration intensity data of each time segment to the fixed physical position coordinates of the corresponding sensors to form space-time aligned voiceprint-vibration combination parameters; The combination parameters include time stamps, device voiceprint state sequences, device vibration intensity sequences, sound sensor coordinates, and vibration sensor coordinates to form position coordinate pairs.

6. The intelligent fault diagnosis method for dynamic voice prints of a valve cooling device according to claim 1, characterized in that, The method for applying device vibration intensities to device voiceprint states to perform weighted fusion and outputting a joint vibration signal containing vibration-sound coupling features specifically comprises the following steps: S5-1, taking the vibration intensity in the spatio-temporally aligned voiceprint-vibration combined parameter as a weight factor, amplifying the homologous features in the voiceprint state by time segment, and generating a weighted and amplified voiceprint state sequence; S5-2, according to the weighted and amplified voiceprint state sequence, weighting and focusing on the low-frequency voiceprint amplitude in the time segment where the vibration intensity exceeds the preset vibration intensity threshold, and generating a low-frequency focused voiceprint state sequence; S5-3, according to the weighted and amplified voiceprint state sequence, suppressing the high-frequency noise fluctuation in the time segment where the vibration intensity is lower than the preset vibration intensity threshold, and generating a high-frequency suppressed voiceprint state sequence; S5-4, recombining the voiceprint state sequence processed by steps S5-1 to S5-3 in time sequence, and fusing to generate a joint vibration signal containing vibration-sound coupling features.

7. The method according to claim 1, wherein, The joint vibration signal is compared with the preset fault feature library, and the fault type and positioning coordinates are output, including: Detecting the amplitude change rate of adjacent time segments in the joint vibration signal, combining the preset mutation judgment threshold, determining the time segment and frequency band of amplitude mutation, and integrating into a mutation time segment list; Matching the mutation time segment list with the fault mode stored in the preset fault feature library, when the similarity meets the preset matching condition, confirming the matching success, and determining the fault type, including leakage fault features and bearing wear fault features; Marking the sensor coordinate position associated with the fault type.

8. The method according to claim 1, wherein, The diagnostic report containing the fault type and positioning coordinates is generated, wherein: All fault records are obtained and arranged in time sequence to generate a diagnostic report; The diagnostic report format is a table structure: the first column is the time stamp, the second column is the fault type, the third column is the sound sensor coordinate, and the fourth column is the vibration sensor coordinate.

9. An intelligent fault diagnosis system for dynamic voice prints of valve cooling equipment, characterized in that, Including: A device operation data acquisition module activates the sound sensor and the vibration sensor, synchronously reads real-time sound signals and vibration signals, and outputs original sound streams and original vibration streams; A sound state feature extraction module processes the original sound stream, divides it into continuous time period sound data segments, and outputs a device voiceprint state sequence; A vibration energy parameter conversion module processes the original vibration stream, calculates the time domain intensity value of the acceleration data, and outputs a device vibration intensity sequence; A dual-signal spatio-temporal alignment module inputs the device voiceprint state and the device vibration intensity, matches the time scale and the physical position coordinates, and outputs a spatio-temporally aligned voiceprint-vibration combined parameter; A joint vibration signal generation module applies the device vibration intensity to perform weighted fusion on the device voiceprint state, and outputs a joint vibration signal containing vibration-sound coupling features; A fault state diagnosis module compares the joint vibration signal with the preset fault feature library, outputs the fault type and positioning coordinates, and generates a diagnostic report containing the fault type and positioning coordinates.

Images