A device health optimal monitoring point selection method and system

Abstract

The present application relates to a kind of equipment health optimal monitoring point selection method and system, it is related to the field of equipment monitoring, it includes the sound signal of the surrounding monitoring point of the monitored equipment is collected;Signal spectrum and signal attenuation are determined based on sound signal;Determine sound source position according to signal spectrum and signal attenuation;Based on signal spectrum, calculate signal-to-noise ratio, match to determine noise reduction level, and the comprehensive score of weighted according to noise reduction level and sound source position is obtained;Combining sound source position and noise reduction level, calculate comprehensive score;According to monitoring point and sound signal, calculate volatility, and the position of volatility less than stable threshold and comprehensive score higher than score threshold is screened out as secondary monitoring point;According to sound source position and equipment structure, determine the position distribution of secondary monitoring point;From position distribution, the point position of maximum coverage is selected, and the point position of highest coincidence degree is selected as optimal point position based on signal spectrum and fault spectrum match determination.This application has the effect of improving monitoring effect, realizes the effect of high-precision monitoring.

Description

Technical Field

[0001] This invention relates to the field of equipment monitoring, and in particular to a method and system for selecting the optimal monitoring point for equipment health. Background Technology

[0002] In industrial production, equipment is prone to malfunctions during operation, so the testing of industrial equipment is crucial to ensuring its stable operation.

[0003] When a malfunction occurs, the sound waves of the equipment will change, producing abnormal noise, unusual sounds, and frequency changes. Currently, the equipment status can be determined by listening with bare ears or by using a vibration and noise analyzer based on sound differences. This provides a convenient way to monitor and repair equipment malfunctions, helps to detect potential equipment problems in a timely manner, and improves production efficiency and safety.

[0004] In practical use, human listening is highly subjective, which increases the product failure rate and makes it impossible to achieve high-precision monitoring. On the other hand, for early faults such as slight bearing wear and loose parts, the sound waves generated by the analyzer may have non-linear characteristics, making it impossible to effectively extract these weak features, resulting in poor monitoring effect and failure to achieve high-precision monitoring. Summary of the Invention

[0005] To improve monitoring effectiveness and achieve high-precision monitoring, this invention provides a method and system for selecting the optimal monitoring point for equipment health.

[0006] In a first aspect, the present invention provides a method for selecting the optimal monitoring point for equipment health, which adopts the following technical solution:

[0007] A method for selecting the optimal monitoring point for equipment health includes:

[0008] Collect sound signals from preset monitoring points around the monitored equipment;

[0009] The signal spectrum and signal attenuation are determined based on the sound signal;

[0010] The location of the sound source is determined based on the signal spectrum and the signal attenuation.

[0011] Based on the signal spectrum, the signal-to-noise ratio of the preset monitoring point signal is calculated, the noise reduction level of each location is determined, and a comprehensive score is obtained by weighting the noise reduction level and the sound source location.

[0012] Based on the location of the sound source and the noise reduction level, calculate the comprehensive score of the monitoring point;

[0013] Based on the monitoring points and the corresponding sound signals, the volatility under various working conditions is calculated, and the locations with volatility less than a preset stability threshold and a comprehensive score higher than a preset score threshold are selected as secondary monitoring points.

[0014] The location distribution of the secondary monitoring points is determined based on the location of the sound source and the preset equipment structure.

[0015] Select the point with the largest coverage from the location distribution, and determine the point with the highest overlap as the optimal point based on the signal spectrum and the preset fault spectrum matching.

[0016] By adopting the above technical solution, sound signals are collected and the signal spectrum and attenuation are extracted. The signal-to-noise ratio and noise reduction level are calculated based on the spectrum. Then, by comprehensively considering the sound source location, operating condition fluctuation rate, structural coverage and fault spectrum overlap, the final location is selected as the optimal location. This improves the monitoring effect and achieves high-precision monitoring.

[0017] Optionally, determining the location of the sound source includes:

[0018] The signal amplitude and signal energy distribution are determined based on the signal spectrum.

[0019] The spectral similarity is determined based on the signal amplitude and a preset sound source spectrum library;

[0020] The energy focusing point is determined based on the signal energy distribution and the preset pickup array position;

[0021] The interval distance is determined based on the energy focusing point and the signal attenuation.

[0022] The location of the sound source is determined by combining the spectral similarity and the interval distance.

[0023] By adopting the above technical solution, the spectrum similarity is obtained by comparing the signal amplitude and energy distribution with the spectrum library, and the sound source location is determined by combining the coordinates of the pickup array and the attenuation. This shortens the positioning path and reduces environmental interference, ensuring the rapid convergence of the sound source coordinates, so as to maintain high robustness and real-time performance under complex working conditions and reduce interference.

[0024] Optionally, determining the energy focus point includes:

[0025] Based on the signal energy distribution, the energy value of the signal collected by each pickup unit is calculated;

[0026] A mapping coordinate system between energy and position is established based on the preset position coordinates of the pickup array;

[0027] Based on the mapped coordinate system, spatial focusing calculations are performed on the energy values ​​to determine the spatial spectrum of the energy distribution.

[0028] Based on the spatial spectrum, the energy peak point is identified, and the pickup distance is calculated based on the energy peak point and the preset pickup unit array.

[0029] The energy focusing point is determined by combining the preset signal attenuation parameters and the pickup distance.

[0030] By adopting the above technical solution, an energy and position mapping coordinate system is constructed, and a spatial spectrum is obtained through spatial focusing calculation. After picking the peak point, the energy focusing point is locked by combining the attenuation parameter. By replacing the traditional traversal with the spatial spectrum, the amount of computation is significantly reduced, taking into account both positioning accuracy and computational efficiency.

[0031] Optionally, after matching to determine the noise reduction level at each location, the following may also be included:

[0032] Based on the signal spectrum, determine the waveform characteristics and spectral envelope;

[0033] The waveform matching degree is determined based on the waveform characteristics and the preset waveform template library;

[0034] The set of voiceprint parameters is determined based on the spectral envelope;

[0035] Based on the waveform characteristics, the voiceprint characteristics are determined by optimizing the voiceprint parameter set;

[0036] The similarity between the voiceprint features and the preset device standard voiceprint library is calculated to determine the voiceprint matching degree;

[0037] The voiceprint features are optimized based on the voiceprint feature matching degree to update the voiceprint features;

[0038] The acquired sound signal is denoised based on the updated voiceprint features and the noise reduction level.

[0039] By adopting the above technical solution, waveform features and spectral envelopes are extracted to generate a set of voiceprint parameters, which are then iteratively optimized using a standard voiceprint library, followed by noise reduction. This transforms acoustic features into quantifiable voiceprint indicators, enabling adaptive noise suppression and providing continuous assurance for high-fidelity acquisition of monitoring data.

[0040] Optionally, determining the voiceprint parameter set includes:

[0041] The sound signal is preprocessed to determine the noise reduction signal;

[0042] Time-frequency analysis is performed on the denoised signal to extract time-domain features, frequency-domain features, and cepstral features;

[0043] Based on the extracted time-domain features, frequency-domain features, and cepstral features, the voiceprint feature parameters are determined.

[0044] The spectral envelope is compared with preset voiceprint feature parameters to calculate the envelope similarity.

[0045] By introducing running weight parameters, the waveform matching degree and the envelope similarity are weighted and fused to determine the voiceprint parameter set.

[0046] By adopting the above technical solution, integrating time domain, frequency domain, and cepstral features, and introducing weighted waveform matching degree and envelope similarity, a voiceprint parameter set is constructed. By improving the completeness of feature expression, the voiceprint's sensitivity to minor faults is enhanced, thereby maintaining high recognizability in multi-condition noise environments.

[0047] Optionally, determining waveform characteristics and spectral envelope includes:

[0048] The signal amplitude and signal period are determined based on the signal spectrum.

[0049] The phase difference is calculated based on the signal amplitude.

[0050] The frequency period segment is determined based on the signal period and the preset period error range;

[0051] The waveform characteristics are determined by combining the phase difference and the frequency period segment;

[0052] The time-domain matrix is ​​determined based on the waveform characteristics, and the amplitude sequence is determined based on the time-domain matrix;

[0053] After smoothing the amplitude sequence, the spectral envelope is determined by combining it with the signal period.

[0054] By adopting the above technical solution, the phase difference and frequency period segment are extracted based on the signal amplitude and period, and a time-domain matrix and amplitude sequence are generated. After smoothing, a spectral envelope is formed. Thus, the signal morphology is characterized in a matrix manner, ensuring high consistency between the waveform and the envelope, and laying a reliable foundation for subsequent feature comparison.

[0055] Optional noise reduction processing includes:

[0056] Based on the voiceprint features, a sequence of feature parameters is determined, and a preliminary curve is obtained through curve fitting.

[0057] The trend rate of change of the signal is determined based on the preliminary curve.

[0058] When the rate of change of the trend exceeds a preset threshold, the fitting curve is determined by combining the preset fitting correction coefficient;

[0059] The energy percentage of noise is calculated based on the fitted curve.

[0060] The noise interference level is determined based on the energy percentage and the sound signal.

[0061] When the noise interference level is greater than the noise reduction level, an over-reduction parameter is determined based on the noise interference level, and a smoothing parameter is determined based on the noise reduction level.

[0062] Noise reduction is performed based on the over-subtraction parameter and the smoothing parameter.

[0063] By adopting the above technical solution, the voiceprint feature curve is fitted, the trend change rate and noise energy ratio are calculated, and the over-subtraction parameter and smoothing parameter are adjusted accordingly to complete the noise reduction; thereby realizing dynamic threshold control, avoiding excessive noise reduction that leads to signal distortion, so as to retain key fault information even in a noisy environment.

[0064] Optional noise reduction includes:

[0065] The effective signal bandwidth is determined based on the signal spectrum and the preset bandwidth.

[0066] The sound signal is converted to the frequency domain to obtain the frequency band distribution;

[0067] Based on the effective signal bandwidth, frequencies exceeding the preset frequency range are identified, marked, and defined as noise distribution;

[0068] The noise ratio is determined based on the frequency band distribution and the noise distribution;

[0069] When the noise ratio is greater than the preset reference ratio, the voiceprint parameter set is adjusted;

[0070] Based on the voiceprint parameter set and the waveform matching degree, a fitting model is generated, and the effective signal after noise separation is reconstructed to complete the noise reduction.

[0071] By adopting the above technical solution, the effective signal bandwidth is defined, and the frequency range is marked as noise distribution, and the separated effective signal is reconstructed. Noise is precisely removed in the frequency domain range to maintain signal integrity, thereby significantly improving the signal-to-noise ratio and monitoring accuracy in high-frequency interference scenarios.

[0072] Optionally, generating a fitted model includes:

[0073] Based on the voiceprint parameter set and the waveform matching degree, an initial feature set is constructed;

[0074] The initial feature set is segmented into time series data to calculate statistical features;

[0075] The statistical features are fitted using the least squares method to determine the preliminary fitting curve;

[0076] During the fitting process, when the noise interference level is less than the preset interference level, a preliminary fitting curve is obtained by directly fitting.

[0077] When the noise interference level is not less than the preset interference level, the statistical features are first filtered, and then a preliminary fitting curve is obtained based on the filtered statistical features.

[0078] When the trend change rate is greater than a preset ratio, the trend change rate is adjusted to a corrected trend change rate.

[0079] Based on the corrected trend change rate and the preliminary fitted curve, a fitted model is generated.

[0080] By adopting the above technical solution, an initial feature set is constructed and the time series is segmented. Least squares polynomial fitting is used, supplemented by filtering or trend correction to generate the final fitted model. By coupling statistical features and trends in modeling, the generalization ability of the model is improved, providing stable algorithm support for long-term tracking of equipment health status.

[0081] Secondly, this application provides a system for selecting the optimal monitoring point for equipment health, which adopts the following technical solution:

[0082] A system for selecting optimal monitoring points for device health includes:

[0083] The acquisition module is used to acquire sound signals;

[0084] The memory is used to store the program for selecting the optimal health monitoring point for any device.

[0085] The processor loads and executes programs from memory.

[0086] In summary, this application includes at least one of the following beneficial technical effects:

[0087] 1. By adopting the above technical solution, sound signals are collected and the signal spectrum and attenuation are extracted. The signal-to-noise ratio and noise reduction level are calculated based on the spectrum. Then, by comprehensively considering the sound source location, operating condition fluctuation rate, structural coverage, and fault spectrum overlap, the final location is selected as the optimal location. This allows the monitoring point to be accurately converged from a large number of candidate options, significantly improving the pertinence and reliability of subsequent fault diagnosis, thereby improving the monitoring effect and achieving high-precision monitoring.

[0088] 2. By adopting the above technical solution, waveform features and spectral envelopes are extracted to generate a set of voiceprint parameters, which are then iteratively optimized using a standard voiceprint library, followed by noise reduction. By converting acoustic features into quantifiable voiceprint indicators, adaptive noise suppression is achieved, providing continuous assurance for high-fidelity acquisition of monitoring data.

[0089] 3. By adopting the above technical solution, an initial feature set is constructed and the time series is segmented. Least squares multinomial fitting is used, supplemented by filtering or trend correction to generate the final fitted model. By coupling statistical features with trends in modeling, the generalization ability of the model is improved, providing stable algorithm support for long-term tracking of equipment health status. Attached Figure Description

[0090] Figure 1 This is a flowchart of a method for selecting the optimal monitoring point for device health according to an embodiment of the present invention;

[0091] Figure 2 This is a flowchart of a noise reduction processing method according to an embodiment of the present invention;

[0092] Figure 3 This is a flowchart of the method for generating a fitting model according to an embodiment of the present invention. Detailed Implementation

[0093] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0094] This application discloses a method for selecting the optimal monitoring point for equipment health.

[0095] Reference Figure 1 A method for selecting the optimal monitoring point for equipment health includes the following steps:

[0096] Step S100: Collect sound signals from preset monitoring points around the monitored device.

[0097] The monitored equipment refers to the machine that needs to be monitored by sound pickup to identify the health of the equipment, such as motors and water pumps, which are pre-set monitoring objects.

[0098] The monitoring point refers to the location of the pre-placed sound pickup sensors around the monitored equipment.

[0099] A sound pickup sensor is a sensor that can collect the sound of the monitored device. It has an array of pickup units, each of which is a single acoustic sensor. The monitoring point location and the pickup sensor are preset by technicians according to the actual situation, and will not be described in detail here.

[0100] Sound signals refer to acoustic data acquired from multiple monitoring points using sound pickup sensors.

[0101] Step S101: Determine the signal spectrum and signal attenuation based on the sound signal.

[0102] The signal spectrum refers to the spectral data of a sound signal in the frequency domain.

[0103] Signal attenuation refers to the energy loss of sound during propagation.

[0104] After the sound signal is converted into a digital signal by an analog-to-digital converter, the digital signal is converted into a sound spectrum by a fast Fourier transform, which is the signal spectrum.

[0105] The locations of each monitoring point are different, resulting in variations in the received sound signals. There are also distance differences between the monitoring points, and the sound signals attenuate as the distance decreases. The sound signals and monitoring point locations of each point are input into a preset attenuation database to match and obtain the signal attenuation amount. The attenuation database is a database that is preset by technicians according to the actual situation. The attenuation database contains formulas for determining the signal attenuation amount based on the monitoring point location and the corresponding sound signal. The actual formulas are preset by technicians according to the actual situation and will not be elaborated here.

[0106] Step S102: Determine the location of the sound source based on the signal spectrum and the signal attenuation.

[0107] The location of a sound source refers to the coordinates of the sound source generated by equipment noise or abnormal vibration.

[0108] Sound attenuates during propagation due to factors such as diffusion and absorption, and the amount of attenuation is directly related to the propagation distance. The distance can be deduced from the attenuation of the sound source; the farther the distance, the more dispersed the sound energy distribution. By substituting the signal spectrum and the signal attenuation into a preset sound source formula, the corresponding sound source location can be obtained. The preset sound source formula will not be elaborated here.

[0109] Step S103: Based on the signal spectrum, calculate the signal-to-noise ratio of the preset monitoring point signal, match and determine the noise reduction level of each location, and obtain a comprehensive score by weighting the noise reduction level and the sound source location.

[0110] Signal-to-noise ratio (SNR) refers to the ratio of the energy of a sound signal to that of noise.

[0111] Noise reduction level refers to the required noise reduction intensity level for a corresponding location. The noise reduction intensity corresponding to each level is preset by technicians according to the actual situation, and will not be elaborated here.

[0112] The comprehensive score refers to the score used to judge the monitoring effect of the monitoring point location. It is obtained by multiplying the sound source location and noise reduction level by a preset scoring weight. The scoring weight is preset by technicians according to the actual situation, and will not be elaborated here.

[0113] The signal-to-noise ratio (SNR) is obtained by identifying the preset monitoring point signal in the signal spectrum and calculating the ratio between the monitoring point signal and the noise other than the monitoring point signal in the signal spectrum.

[0114] Different signal-to-noise ratios correspond to different noise reduction levels. The correspondence between signal-to-noise ratios and noise reduction levels is pre-stored. The corresponding noise reduction level can be determined based on the obtained signal-to-noise ratio. The correspondence between signal-to-noise ratios and noise reduction levels is pre-set by technicians according to the actual situation, and will not be elaborated here.

[0115] Step S104: Based on the monitoring points and the corresponding sound signals, calculate the volatility under various operating conditions, and select the locations where the volatility is less than a preset stability threshold and the comprehensive score is higher than a preset score threshold as secondary monitoring points.

[0116] Volatility refers to the rate of change of the amplitude of a sound signal under different operating conditions, and is used to judge the stability of a sound signal.

[0117] The preset stability threshold refers to the maximum allowable fluctuation range, which is preset by technicians according to the actual situation, and will not be elaborated here.

[0118] The scoring threshold refers to the cut-off score used to determine whether the comprehensive score of the monitoring point is qualified. It is set in advance by technical personnel according to the actual situation and will not be elaborated here.

[0119] The system pre-sets different sound data under various operating conditions, compares the sound signals and corresponding sound data at different monitoring points with parameters, calculates the difference between the corresponding sound signals and sound data under the same operating condition, and then determines the corresponding fluctuation rate based on the sound data.

[0120] For example, if the parameter of the sound data is 10 and the parameter of the sound signal is 9, the difference between the two is 10-9=1, then the corresponding volatility is 1 / 10*100%=10%.

[0121] Step S105: Determine the location distribution of the secondary monitoring points based on the sound source location and the preset equipment structure.

[0122] The equipment structure refers to the positional arrangement of the various parts of the monitored equipment, which is pre-set by technicians according to the actual situation and will not be elaborated here.

[0123] Since the location of the monitoring points is known when they are set up, the location of the secondary monitoring points on the equipment structure is determined by combining the equipment structure, and the distribution of the secondary monitoring points around the sound source location is determined by the sound source location.

[0124] For example, let the coordinates of the sound source location be (1, 2) in the initial state (first coordinate system), and the coordinates of one of the secondary monitoring points be (3, 5). Then, take the coordinates of the sound source location as the origin of the coordinate system, and take the sound source location as the reference point (second coordinate system) to determine the coordinates of the secondary monitoring point (2, 3). The equipment structure is known, the distance between the locations is known, and the coverage of the covering equipment at each location can also be determined, so as to obtain the distribution of the secondary monitoring points at the sound source location.

[0125] Step S106: Select the point with the largest coverage from the location distribution, and determine the point with the highest overlap as the optimal point based on the signal spectrum and the preset fault spectrum matching.

[0126] The fault spectrum refers to the spectral characteristics of the equipment when it fails. It is preset by technicians according to the actual situation and will not be elaborated here.

[0127] Coverage refers to the richness of the types of sound signals that a monitoring point at a given location can collect. The coverage of each location can be determined by analyzing the sound signals on the signal spectrum.

[0128] Overlap ratio refers to the similarity between the signal spectrum and a preset fault spectrum, used to screen for the optimal position. By comparing the signal spectrum and the fault spectrum and analyzing their matching degree in dimensions such as frequency components, amplitude distribution, and harmonic characteristics, the point in the sound signal with the most complete fault spectrum is selected, which is the optimal position.

[0129] Determining the location of a sound source involves the following steps:

[0130] Step S200: Determine the signal amplitude and signal energy distribution based on the signal spectrum.

[0131] Signal amplitude refers to the maximum range of fluctuations in the spectrum over a period of time.

[0132] Signal energy distribution refers to the distribution of energy in the frequency domain.

[0133] From the spectrum data of the signal spectrum, identify the frequency points with the largest amplitude at each frequency in the spectrum, record their amplitude, and thus obtain the signal amplitude.

[0134] The signal energy distribution E(f) is proportional to the square of the signal amplitude A(f). If the spectrum is the amplitude spectrum A(f), the energy spectrum can be obtained by squaring it. (k is a preset constant, which is related to the signal sampling parameters).

[0135] Step S201: Determine the spectral similarity based on the signal amplitude and the preset sound source spectrum library.

[0136] The sound source spectrum library refers to a database of pre-stored characteristic amplitudes of various sound sources. It is pre-set by technicians according to the actual situation and will not be elaborated on here.

[0137] Spectral similarity refers to the degree of matching between the signal amplitude and the characteristic amplitude of the sound source in the spectral library, and is used to determine the changes in the propagation of the sound signal.

[0138] The amplitude of each frequency point on the signal amplitude is compared with the amplitude of various sound source features pre-stored in the sound source spectrum library to determine the degree of matching between the amplitudes, and the spectrum similarity is obtained by weighting.

[0139] Step S202: Determine the energy focusing point based on the signal energy distribution and the preset pickup array position.

[0140] The position of the pickup array refers to the spatial coordinates of the pickup sensor array arrangement, which is preset by technicians according to the actual situation and will not be elaborated here.

[0141] An energy focal point refers to the location in space where energy converges and overlaps most intensely.

[0142] When sound propagates through space, its energy attenuates with distance. By analyzing the differences in signal energy received by each pickup unit in the pickup array, it is determined that the energy received by a pickup unit is related to its distance from the energy focal point. Combining this with the known positions of the pickup units, the spatial coordinates of the most concentrated sound energy can be determined, which is the energy focal point. Thus, the spatial coordinates of the energy concentration are obtained, which will not be elaborated here. The specific method is described in steps S300 to S304.

[0143] Step S203: Determine the interval distance based on the energy focusing point and the signal attenuation.

[0144] The gap distance refers to the physical distance from the energy focusing point to the pickup sensor.

[0145] Since the energy of a sound signal attenuates with distance during propagation, the distance calculated by inputting the signal attenuation amount into the sound source formula based on the energy focal point is the interval distance.

[0146] Step S204: Determine the location of the sound source by combining the spectral similarity and the interval distance.

[0147] By eliminating sources with low spectral similarity and retaining only the focal point that matches the high-similarity sound source, and with the positions of each unit in the pickup array known, combined with the calculated interval distance, the coordinate range of the focal point can be narrowed down through spatial geometric positioning. Thus, the sound source position can be determined based on the interval distance, thereby reducing interference.

[0148] Determining the energy focal point involves the following steps:

[0149] Step S300: Based on the signal energy distribution, calculate the energy value of the signal collected by each pickup unit.

[0150] Based on the signal energy distribution, the approximate energy level of each pickup unit is initially determined. Then, the final energy value is obtained by weighted calculation with the signal energy values ​​collected by each unit. The weighted calculation method is common knowledge to those skilled in the art and will not be elaborated here.

[0151] Step S301: Establish a mapping coordinate system of energy and position based on the preset pickup array position coordinates.

[0152] The position coordinates of the pickup array refer to the coordinate data of each pickup unit array, which are preset by technicians according to the actual situation and will not be elaborated here.

[0153] A mapped coordinate system is a mathematical model that maps energy values ​​to spatial coordinates.

[0154] Based on the energy value of each pickup unit and the position of the pickup array, the energy and the corresponding position are matched one by one to establish a mapping coordinate system between energy and position. The specific method is common knowledge to those skilled in the art and will not be elaborated here.

[0155] Step S302: Perform spatial focusing calculation on the energy value based on the mapped coordinate system to determine the spatial spectrum of the energy distribution.

[0156] A spatial spectrum is a diagram of the distribution of energy in space.

[0157] Based on the energy value of the pickup array, mathematical calculations are used to deduce the corresponding spatial location of the energy emission point received by different pickup units. Each point is marked in space, and the location of the point is different for different energy values. The spatial location of the point is transformed into an intuitive image, thus obtaining a spatial spectrum.

[0158] Step S303: Based on the spatial spectrum, identify the energy peak point, and calculate the pickup distance according to the energy peak point and the preset pickup unit array.

[0159] Each pickup unit array refers to the arrangement and distance between each pickup unit array, which is preset by technicians according to the actual situation and will not be elaborated here.

[0160] The energy peak point refers to the position in the spectrum where the focus energy points have the highest degree of overlap.

[0161] The pickup distance refers to the distance from the peak point to the sensor.

[0162] Based on the spatial spectrum, the peak points are extracted using MATLAB software to obtain the energy peak points. The specific method is common knowledge to those skilled in the art and will not be elaborated here.

[0163] The pickup distance is calculated by taking the distance between the points in the pickup array and the peak energy point. For example, if the pickup array has points (X2, Y2, Z2) and the peak energy point is (X1, Y1, Z1), then the calculated pickup distance L = √((X1-X2)). 2 +(Y1-Y2) 2 +(Z1-Z2) 2 ).

[0164] Step S304: Determine the energy focusing point by combining the preset signal attenuation parameters and the pickup distance.

[0165] Signal attenuation parameters refer to the mathematical relationship between signal attenuation and distance. These parameters are preset by technicians based on actual conditions and will not be elaborated upon here.

[0166] The energy peak point may contain false points caused by environmental interference, which need to be verified by the signal attenuation parameter. Therefore, based on the signal attenuation parameter and the pickup distance, it is determined whether the energy peak point conforms to the signal attenuation parameter. If it does, the energy peak point is the energy focus point. If it does not, the energy peak point is corrected according to the signal attenuation parameter and the pickup distance to obtain the energy focus point.

[0167] After matching and determining the noise reduction level at each location, the following steps are included:

[0168] Step S400: Determine waveform characteristics and spectral envelope based on the signal spectrum.

[0169] Waveform characteristics refer to the shape and periodicity of the fluctuations in a sound signal in the time domain.

[0170] The spectral envelope refers to the envelope of a signal's spectrum, used to describe the distribution trend of energy with frequency.

[0171] The specific method for determining waveform characteristics and spectral envelope based on the signal spectrum is described in steps S600 to S605, and will not be repeated here.

[0172] Step S401: Determine the waveform matching degree based on the waveform characteristics and the preset waveform template library.

[0173] The waveform template library refers to a database that stores standard waveform features. It is pre-set by technicians according to actual conditions and will not be elaborated on here.

[0174] Waveform matching degree refers to the similarity between the current waveform features and the standard waveforms in the template library.

[0175] The waveform matching degree is obtained by comparing the waveform features with the same standard waveform features in the waveform template library.

[0176] Step S402: Determine the voiceprint parameter set based on the spectral envelope.

[0177] The voiceprint parameter set refers to the set of parameter vectors extracted from the spectral envelope for voiceprint recognition.

[0178] The specific method for determining the voiceprint parameter set is described in steps S500 to S504, and will not be repeated here.

[0179] Step S403: Based on the waveform features, determine the voiceprint features by optimizing the voiceprint parameter set.

[0180] Voiceprint features refer to the voiceprint data of the target signal, which is used to match fault sounds and identify the core fault conditions of the equipment.

[0181] By using parameters from the voiceprint parameter set, the temporal morphological characteristics of the waveform (such as period, peak value, and waveform details corresponding to the spectral envelope) are modified to form voiceprint features that are both stable and distinctive.

[0182] Step S404: Calculate the similarity between the voiceprint features and the preset device standard voiceprint library to determine the voiceprint matching degree.

[0183] The device standard voiceprint library refers to a database that stores the voiceprint characteristics of health devices. It is pre-set by technicians according to the actual situation and will not be elaborated here.

[0184] Voiceprint matching degree refers to the similarity between the current voiceprint features and the standard voiceprint database.

[0185] The similarity between the voiceprint features and the standard voiceprints matched in the device's standard voiceprint library is calculated to determine the voiceprint matching degree.

[0186] Step S405: Optimize the voiceprint features based on the voiceprint feature matching degree to update the voiceprint features.

[0187] By analyzing the matching degree between the current voiceprint features and the standard voiceprint, the bias components in the features that lead to low matching degree are located. Based on the selected bias parameters, the corresponding features of the current voiceprint are directly corrected, thereby obtaining more accurate voiceprint features.

[0188] Step S406: Perform noise reduction processing on the acquired sound signal based on the updated voiceprint features and the noise reduction level.

[0189] The specific methods for noise reduction processing are described in steps S700 to S805, and will not be repeated here.

[0190] Determining the voiceprint parameter set includes the following steps:

[0191] Step S500: Preprocess the sound signal to determine the noise reduction signal.

[0192] Preprocessing refers to operations such as filtering and normalizing audio signals. Preprocessing methods are common knowledge to those skilled in the art and will not be elaborated here.

[0193] Preprocessing is used to obtain a noise-reduced signal, thereby reducing noise interference.

[0194] Step S501: Perform time-frequency analysis on the noise-reduced signal to extract time-domain features, frequency-domain features, and cepstral features.

[0195] Time-domain characteristics refer to the features of a signal in the time dimension, such as amplitude and period.

[0196] Frequency domain characteristics refer to the features of a signal in the frequency dimension, such as the distribution of spectral energy.

[0197] Cepstral characteristics refer to the inverse transform characteristics of the logarithmic spectrum of a signal.

[0198] By analyzing the dynamic changes of the signal in the time and frequency domains, the non-stationary characteristics of the signal are fully captured, and time-domain, frequency-domain, and cepstral features are extracted from them. The extraction methods are common knowledge to those skilled in the art and will not be elaborated here.

[0199] Step S502: Determine the voiceprint feature parameters based on the extracted time-domain features, frequency-domain features, and cepstral features.

[0200] The voiceprint feature parameters refer to the set of parameters obtained from the time-domain features, the frequency-domain features, and the cepstral features.

[0201] Time-domain features reflect the dynamic amplitude changes of sound, yielding statistical characteristics of short-time energy and peak intervals, as well as the mean zero-crossing rate. Frequency-domain features reflect the frequency composition of sound, providing statistical characteristics of the fundamental frequency, formant characteristics, and the mean spectral entropy. Cepstral features efficiently capture the macroscopic characteristics of duct resonance, providing static coefficients, dynamic coefficients, and the mean cepstral value of the MFCC.

[0202] The various features are obtained from the time-domain features, the frequency-domain features, and the cepstral features. The voiceprint feature parameters are constructed based on the obtained features. The specific methods are common knowledge to those skilled in the art and will not be elaborated here.

[0203] Step S503: Compare the spectral envelope with the preset voiceprint feature parameters and calculate the envelope similarity.

[0204] Envelope similarity refers to the similarity between the spectral envelope and the standard voiceprint feature parameters.

[0205] After extracting the corresponding envelope from the voiceprint feature parameters, the similarity between the spectral envelope and the extracted envelope is calculated to obtain the envelope similarity.

[0206] Step S504: Introduce running weight parameters to perform weighted fusion of the waveform matching degree and the envelope similarity to determine the voiceprint parameter set.

[0207] The running weight parameter refers to the weight ratio used to adjust the waveform matching degree and envelope similarity. It is preset by the technicians and will not be elaborated here.

[0208] The waveform matching degree and envelope similarity are calculated by weighted average based on the running weight parameters to obtain the voiceprint parameter set.

[0209] Determining waveform characteristics and spectral envelope involves the following steps:

[0210] Step S600: Determine the signal amplitude and signal period based on the signal spectrum.

[0211] Signal amplitude refers to the peak amplitude of a signal waveform.

[0212] The signal period refers to the time interval between repetitions of a signal waveform.

[0213] In the spectrum data of the signal spectrum, the amplitude of the peak value on the extracted signal waveform is the signal amplitude; the signal period is obtained by passing through the time interval of the signal waveform repetition. The method of determination is common knowledge to those skilled in the art and will not be elaborated here.

[0214] Step S601: Calculate the phase difference based on the signal amplitude.

[0215] Phase difference refers to the phase offset between signal waveforms.

[0216] Based on the peak amplitude of the signal and the mathematical relationship between the superposition of the signal amplitude and the phase difference, the phase difference is derived. The calculation method is common knowledge to those skilled in the art and will not be elaborated here.

[0217] Step S602: Determine the frequency period segment based on the signal period and the preset period error range.

[0218] The period error range refers to the allowable period deviation range, which is used to define the period range and avoid excessively large periods. It is set in advance by technicians and will not be elaborated here.

[0219] The frequency period segment refers to the effective range of the signal frequency.

[0220] Within a signal period, signals exceeding the period error range are removed, and corresponding intervals of continuous frequency signals that meet the period stability requirements are identified in the frequency domain. The period error range is used to filter stable periodic signal components and eliminate unstable frequency components caused by excessive period fluctuations.

[0221] Step S603: Determine waveform characteristics by combining the phase difference and the frequency period segment.

[0222] The frequency period segment determines the stable frequency components of the signal, and then the phase difference between these components is used to further obtain the waveform's morphological properties, which are the waveform characteristics.

[0223] Step S604: Determine the time domain matrix based on the waveform characteristics, and determine the amplitude sequence according to the time domain matrix.

[0224] A time-domain matrix is ​​a matrix form in which signals are stored in a structured manner along the time dimension.

[0225] An amplitude sequence refers to a sequence in which the amplitude of a signal changes over time.

[0226] The time-domain features of the waveform are segmented and feature extracted through time windows, and then transformed into matrix data. The amplitude information that changes with time is then extracted from the matrix to form an ordered amplitude sequence. The specific method is common knowledge to those skilled in the art and will not be elaborated here.

[0227] Step S605: After smoothing the amplitude sequence, determine the spectral envelope by combining it with the signal period.

[0228] Smoothing refers to filtering the amplitude sequence. Smoothing methods are common knowledge to those skilled in the art and will not be elaborated here.

[0229] The spectral envelope is the overall profile of amplitude variation with frequency in the frequency domain. Based on the smoothed amplitude sequence and signal period, the time domain is first transformed to the frequency domain through Fourier transform, and then the frequency domain of the transformed signal is extracted by periodic feature matching. The method for determining this is common knowledge to those skilled in the art and will not be elaborated here.

[0230] Noise reduction processing includes the following steps:

[0231] Step S700: Determine the feature parameter sequence based on the voiceprint features, and obtain a preliminary curve through curve fitting.

[0232] The feature parameter sequence refers to the parameter sequence extracted from the voiceprint features, which is used to fit the noise reduction curve.

[0233] The initial curve refers to the initial noise reduction curve obtained through fitting, which facilitates subsequent correction.

[0234] Relevant parameters such as fundamental frequency and cepstral features are extracted from the voiceprint features. Then, discrete relevant parameters are input into fitting software through mathematical fitting methods to transform them into continuous curves, thus obtaining a preliminary curve.

[0235] Step S701: Determine the trend change rate of the signal based on the preliminary curve.

[0236] The rate of change of trend refers to the rate at which a signal changes over time.

[0237] The rate of change of the trend is obtained by dividing the data of each point on the initial curve by the corresponding time.

[0238] Step S702: When the trend change rate exceeds a preset threshold, determine the fitting curve by combining the preset fitting correction coefficient.

[0239] The preset threshold refers to the critical value at which the rate of change of the trend needs to be corrected. It is set in advance by technical personnel and will not be elaborated here.

[0240] Fit correction coefficients refer to the coefficients used to correct the fitted curve.

[0241] Based on the rate of change of the trend, the intervals in the location curve where the rate of change exceeds a preset threshold are identified. The fitted point data of these intervals are then multiplied by a fitting correction parameter to ensure the overall continuity of the curve and its fit with the data.

[0242] When the trend change rate exceeds the preset threshold, it indicates that the trend change of the current fitted curve is too steep, or that the curve is distorted due to noise interference or abnormal data fluctuations. The values ​​that are out of range in the initial curve are corrected to ensure the reliability of the fitted curve.

[0243] Step S703: Calculate the energy percentage of noise based on the fitted curve.

[0244] Energy percentage refers to the proportion of noise energy to the total signal energy.

[0245] By further processing the fitting residuals through fitting curves and combining filtering or transformation algorithms (such as wavelet transform and FFT), the target signal and noise signal are separated, and the energy of each is calculated separately. Finally, the proportion of noise energy in the total signal energy is obtained, which is the energy proportion. The specific method is common knowledge to those skilled in the art and will not be elaborated here.

[0246] Step S704: Determine the noise interference level based on the energy ratio and the sound signal.

[0247] Noise interference level refers to the severity of noise interference with signals.

[0248] Based on the sound signal, the higher the energy ratio, the higher the noise interference level. The energy ratio and sound signal are input into a preset noise interference level database to obtain the noise interference level. The noise interference level database is a database that is preset by technicians according to the actual situation. The noise interference level database is equipped with a lookup table to determine the noise interference level based on the energy ratio and sound signal. The actual lookup table is preset by technicians according to the actual situation, which will not be elaborated here.

[0249] Step S705: When the noise interference level is greater than the noise reduction level, determine the over-reduction parameter based on the noise interference level, and determine the smoothing parameter based on the noise reduction level.

[0250] Over-reduction parameters refer to parameters used for excessive noise reduction.

[0251] Smoothing parameters refer to parameters used for smoothing noise reduction.

[0252] The noise interference level and over-attenuation parameter, as well as the noise reduction level and smoothing parameter, are all positively correlated. The noise interference level and noise reduction level are input into the corresponding mapping database to match the corresponding over-attenuation parameter and smoothing parameter. The mapping database stores the mapping relationship between the noise interference level and over-attenuation parameter, and the mapping relationship between the noise reduction level and smoothing parameter in advance. The mapping relationship is preset by the technicians according to the actual situation, and will not be elaborated here.

[0253] Step S706: Perform noise reduction based on the over-subtraction parameter and the smoothing parameter.

[0254] The obtained over-subtraction and smoothing parameters are used to denoise the audio signal using spectral subtraction. Spectral subtraction is common knowledge to those skilled in the art and will not be elaborated here.

[0255] Noise reduction includes the following steps:

[0256] Step S800: Determine the effective signal bandwidth based on the signal spectrum and the preset bandwidth width.

[0257] Bandwidth refers to the width range of the set signal frequency band, which is used to extract and determine the effective signal bandwidth. It is preset by technicians and will not be elaborated here.

[0258] Effective signal bandwidth refers to the frequency band range that contains the target signal, and is used to define the frequency range that contains the main useful signal energy.

[0259] Signals that exceed the range in the signal spectrum are removed by using the bandwidth width, thus obtaining the effective signal bandwidth.

[0260] Step S801: Convert the sound signal to the frequency domain to obtain the frequency band distribution.

[0261] Frequency band distribution refers to the distribution of signal energy in the frequency domain.

[0262] The sound signal is transformed into the frequency domain through Fourier transform to obtain the frequency band distribution.

[0263] Step S802: Based on the effective signal bandwidth, identify frequencies that exceed the preset frequency range, mark them, and define them as noise distribution.

[0264] The frequency range refers to the frequency boundary of the effective signal, which is preset by technicians and will not be elaborated here.

[0265] Noise distribution refers to the frequency range that is labeled as noise.

[0266] By identifying all frequency components whose frequency values ​​exceed the effective signal bandwidth within the frequency range, and marking the frequencies that exceed the range, the noise distribution can be determined.

[0267] Step S803: Determine the noise ratio based on the frequency band distribution and the noise distribution.

[0268] The noise ratio is the ratio of the noise frequency to the signal frequency.

[0269] The noise ratio is calculated by dividing the noise frequency of the noise distribution by the signal frequency of the frequency distribution.

[0270] Step S804: When the noise ratio is greater than the preset reference ratio, adjust the voiceprint parameter set.

[0271] The reference ratio refers to the upper limit of the permissible noise ratio. It is a benchmark value used to judge whether the noise is too high. It is preset by technicians and will not be elaborated here.

[0272] When the noise ratio is greater than the preset benchmark ratio, it indicates that the noise will interfere with the fault diagnosis and analysis.

[0273] Step S805: Generate a fitting model based on the voiceprint parameter set and the waveform matching degree, reconstruct the effective signal after noise separation, and complete the noise reduction.

[0274] A fitting model is generated based on the voiceprint parameter set and the waveform matching degree. The specific fitting method is described in steps S900 to S906.

[0275] The noise is separated based on the fitted model, thus obtaining the effective signal and completing the noise reduction.

[0276] Generating a fitted model involves the following steps:

[0277] Step S900: Construct an initial feature set based on the voiceprint parameter set and the waveform matching degree.

[0278] The initial feature set values ​​refer to the set of preliminary parameter features used to fit the generated model.

[0279] The initial feature set is constructed by using multiple parameter features and waveform matching degree in the voiceprint parameter set as preliminary parameters.

[0280] Step S901: Perform time series segmentation on the initial feature set and calculate the statistical features.

[0281] Statistical characteristics refer to the set of parameters obtained by statistically analyzing signals over time.

[0282] The initial feature set is divided according to time, thereby obtaining the statistical features of each time period according to the time sequence. The original continuous time series data is transformed into a point set-like feature vector, which is the statistical feature.

[0283] Step S902: Use the least squares method to perform polynomial fitting on the statistical features to determine the preliminary fitting curve.

[0284] A set of coefficients is found using the least squares method to minimize the sum of squared errors between the predicted values ​​of the fitted curve and the actual statistical feature values. The statistical features are then input into MATLAB software, and feature lines are extracted based on the point cloud characteristics. The corresponding fitted curve is generated by selecting the parametric curve modeling (fitting regular geometric shapes) method.

[0285] Step S903: During the fitting process, when the noise interference level is less than the preset interference level, a preliminary fitting curve is obtained by direct fitting.

[0286] The preset interference level refers to the critical level of noise interference, which is used to determine whether noise interference will affect the fitting. It is set in advance by technicians and will not be elaborated here.

[0287] When the noise interference level is less than the preset interference level, it means that the noise will not interfere with the fitting results, so fitting can be performed directly.

[0288] Step S904: When the noise interference level is not less than the preset interference level, the statistical features are first filtered, and then a preliminary fitting curve is obtained based on the filtered statistical features.

[0289] Filtering refers to filtering a signal to remove noise and reduce noise interference.

[0290] When the noise interference level is not less than the preset interference level, it indicates that the noise is too large, which will cause the fitting curve to be distorted. From the signal containing noise interference, retain the target frequency signal and remove other unwanted components (such as noise, clutter, interference frequencies, etc.), perform filtering, and then perform fitting.

[0291] Step S905: When the trend change rate is greater than the preset ratio, the trend change rate is adjusted to the corrected trend change rate.

[0292] The preset ratio refers to the critical ratio of the trend change rate, which is used as a benchmark value to judge whether the trend change rate is too large. It is preset by technical personnel and will not be elaborated here.

[0293] The corrected rate of change of trend refers to the adjusted rate of change of trend, which is obtained by multiplying the rate of change of trend by 1.2.

[0294] When the trend change rate is greater than the preset ratio, it indicates that the signal trend has a significant fluctuation or reversal. At this time, by increasing the trend change rate for targeted correction, the fitting model can follow the changes in the signal trend more quickly, reduce the fitting deviation caused by the sudden change in trend, and enhance the model's ability to capture dynamic signals.

[0295] Step S906: Generate a fitting model based on the corrected trend change rate and the preliminary fitting curve.

[0296] The corrected trend change rate is multiplied as a weighting factor and multiplied with the coefficients in the initial fitted curve to change the curvature of the initial fitted curve, thereby obtaining the fitted model. This allows for synchronous adjustment based on the actual situation, reducing deviations.

[0297] Based on the same inventive concept, embodiments of the present invention provide a system for selecting optimal monitoring points for device health, comprising:

[0298] The acquisition module is used to acquire sound signals and their energy values.

[0299] The memory is used to store the program for selecting the optimal monitoring point for the health of any device.

[0300] The processor loads and executes programs from memory.

[0301] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0302] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for selecting the optimal monitoring point for equipment health, characterized in that, include: Collect sound signals from preset monitoring points around the monitored equipment; The signal spectrum and signal attenuation are determined based on the sound signal; The location of the sound source is determined based on the signal spectrum and the signal attenuation. Based on the signal spectrum, the signal-to-noise ratio of the preset monitoring point signal is calculated, the noise reduction level of each location is determined, and a comprehensive score is obtained by weighting the noise reduction level and the sound source location. Based on the monitoring points and the corresponding sound signals, the volatility under various working conditions is calculated, and the locations with volatility less than a preset stability threshold and a comprehensive score higher than a preset score threshold are selected as secondary monitoring points. The location distribution of the secondary monitoring points is determined based on the location of the sound source and the preset equipment structure. Select the point with the largest coverage from the location distribution, and determine the point with the highest overlap as the optimal point based on the signal spectrum and the preset fault spectrum matching.

2. The method for selecting the optimal monitoring point for equipment health according to claim 1, characterized in that, Determining the location of the sound source includes: The signal amplitude and signal energy distribution are determined based on the signal spectrum. The spectral similarity is determined based on the signal amplitude and a preset sound source spectrum library; The energy focusing point is determined based on the signal energy distribution and the preset pickup array position; The interval distance is determined based on the energy focusing point and the signal attenuation. The location of the sound source is determined by combining the spectral similarity and the interval distance.

3. The method for selecting the optimal monitoring point for equipment health according to claim 2, characterized in that, Determining the energy focal point includes: Based on the signal energy distribution, the energy value of the signal collected by each pickup unit is calculated; A mapping coordinate system between energy and position is established based on the preset position coordinates of the pickup array; Based on the mapped coordinate system, spatial focusing calculations are performed on the energy values ​​to determine the spatial spectrum of the energy distribution. Based on the spatial spectrum, the energy peak point is identified, and the pickup distance is calculated based on the energy peak point and the preset pickup unit array. The energy focusing point is determined by combining the preset signal attenuation parameters and the pickup distance.

4. A system for selecting optimal monitoring points for equipment health, characterized in that, include: The acquisition module is used to acquire sound signals; A memory for storing a program for a method of selecting the optimal monitoring point for device health as described in any one of claims 1 to 3; The processor loads and executes programs from memory.

Images