Industrial equipment health assessment method and system based on acoustic analysis

Abstract

The application discloses an industrial equipment health evaluation method and system based on acoustic analysis, and relates to the technical field of equipment evaluation, comprising: setting a plurality of excitation reference points and a plurality of groups of microphone arrays, applying a reference response, calculating a transfer function matrix for each frequency to obtain a reference impulse response; calculating a deviation index according to the transfer function matrix, the reference impulse response and the measured impulse response waveform respectively; spatially fusing the deviation indexes of multiple microphone channels to calculate a single-action component combined abnormality and a component comprehensive abnormality, setting a screening threshold range, screening suspicious components and calculating a confidence degree, and performing abnormal identification; by establishing a transfer function matrix and a reference impulse response template under the health state of the equipment, the acoustic propagation characteristics of each individual equipment are taken as personalized reference standards, and the individual difference false alarm problem caused by the use of a general threshold in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of equipment evaluation technology, and in particular to a method and system for health assessment of industrial equipment based on acoustic analysis. Background Technology

[0002] With the continuous improvement of industrial automation, various equipment in production workshops are in a state of high load and continuous operation for a long time. Equipment health monitoring and fault early warning have become key links to ensure production efficiency and safe production.

[0003] Traditional industrial equipment health assessment methods primarily rely on periodic manual inspections or condition monitoring using vibration and temperature sensors installed on the equipment. However, manual inspections suffer from drawbacks such as high subjectivity, long intervals, and difficulty in detecting early, subtle faults; contact sensors require advance deployment and are sensitive to installation location, resulting in high costs in large-scale equipment clusters. Existing acoustic monitoring technologies mainly collect sound signals from the equipment during operation, extract time-domain, frequency-domain, or cepstral-domain features, and compare them with preset thresholds or historical baselines to determine whether the equipment is abnormal.

[0004] However, due to the complex noise environment in the workshop, the mixed acoustic signals generated by multiple devices operating simultaneously are difficult to separate accurately, resulting in large fault location errors. Furthermore, fixed threshold judgments cannot adapt to fluctuations in normal operating conditions of the equipment. For example, when the equipment undergoes process switching, load changes, or operation by different operators, its normal acoustic characteristics will change, and fixed thresholds are prone to causing a large number of false alarms or missed alarms. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a method and system for health assessment of industrial equipment based on acoustic analysis.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: a method for health assessment of industrial equipment based on acoustic analysis, the method comprising: S1: Set up several excitation reference points and several sets of microphone arrays, apply a reference response, calculate the transfer function for each frequency to generate a transfer function matrix; determine the standard process work set according to the standard operating state of the equipment, and collect the measured time-domain waveform of the microphone array for each action to obtain the reference impulse response; S2: Obtain the execution time and action type number of each process action to obtain the measured pulse response waveform; calculate the deviation index based on the transfer function matrix, the reference pulse response, and the measured pulse response waveform; the deviation index includes the normalized cross-correlation coefficient, pulse energy ratio, pulse width expansion coefficient, and peak attenuation coefficient. S3: For each process action and excitation reference point, the deviation indicators of multiple microphone channels are spatially fused to calculate the joint anomaly degree of a single action component and the comprehensive anomaly of the component. A screening threshold range is set to screen suspicious components and calculate the confidence level for anomaly identification. Let the set of excitation reference points be . The microphone channels are set as Suppose that the standard process motion set includes type A motions, denoted as... .

[0007] Furthermore, the method also includes: S4: Set up an active detection mechanism, which automatically triggers based on pre-set trigger conditions. After triggering, it obtains the acoustic impedance change of the equipment structure surface by emitting ultrasonic pulses and receiving reflected echoes, and calculates the impedance change rate and active detection confidence level. S5: Align the component anomaly degree, anomaly identifier, and impedance change rate in time and space, define an indicator function, calculate the comprehensive anomaly score of a single component and the overall health index of the equipment, and output the warning level and maintenance list according to the screening rules; the indicator function includes a passive self-test effective indicator function and an active detection effective indicator function.

[0008] Furthermore, the triggering condition is that the equipment is in a low spontaneous acoustic energy state or a preset interruptible production period. That is, when the equipment operating parameters are detected to be lower than the minimum dynamic threshold required for health acoustic monitoring or when the production scheduling system confirms that the equipment has entered a non-continuous processing window, the active detection mechanism is automatically triggered.

[0009] Furthermore, the active detection mechanism is configured such that the center frequency of the transmission of the miniature ultrasonic transmitting unit integrated in the microphone array is... ,bandwidth Linear frequency modulated pulse, pulse width Launch interval Continuous launch Next; all microphones simultaneously collect echoes, sampling rate Collection time The acquired echo signal is first matched and filtered with the transmitted coded sequence to obtain a high-resolution echo sequence after pulse compression. Extract each predefined structure reflection surface from the echo sequence. Flight time and echo amplitude By measuring the echo amplitude of M microphone channels, the equivalent structural acoustic impedance of each reflecting surface is inverted using the weighted least squares method. An objective function is defined, the acoustic impedance that satisfies the objective function is solved, and the impedance change rate is calculated.

[0010] An industrial equipment health assessment system based on acoustic analysis, applied to the aforementioned industrial equipment health assessment method based on acoustic analysis, the system comprising: The benchmark modeling module is used to determine the measured time-domain waveform by applying a benchmark response and utilizing the inherent process switching actions of the equipment when the equipment is initially put into operation and confirmed to be healthy and fault-free. It also calculates the transfer function matrix at each frequency and the benchmark impulse response corresponding to each standard process action. The passive monitoring module is used to acquire the execution time and type number of each process action during normal production operation, collect the measured pulse response waveform triggered by the action, and calculate the deviation index based on the transfer function matrix and the reference pulse response. The deviation index includes the normalized cross-correlation coefficient, pulse energy ratio, peak attenuation coefficient and pulse width expansion coefficient deviation index. The anomaly identification module receives various deviation indicators output by the passive monitoring module, performs signal-to-noise ratio weighted spatial fusion on the indicators of multiple microphone channels, calculates the joint anomaly degree of a single action component and the comprehensive anomaly degree of the component within a week; identifies components as anomalies according to a preset threshold range, and calculates the confidence level of passive self-testing. The active detection module is used to set up an active detection mechanism, which is automatically triggered based on pre-set trigger conditions. After triggering, it obtains the acoustic impedance change of the device structure surface by emitting ultrasonic pulses and receiving reflected echoes, and calculates the impedance change rate and active detection confidence level. The fusion early warning module receives the component comprehensive anomaly degree, anomaly identifier, passive confidence degree output by the anomaly identification module, and impedance change rate output by the active detection module; it performs spatiotemporal alignment on the data, calculates the comprehensive anomaly score of a single component using a confidence-weighted fusion formula, and then calculates the overall health index of the equipment, outputting the early warning level and maintenance list according to preset rules.

[0011] The beneficial effects of this invention are: by establishing the transfer function matrix and reference impulse response template under the health state of the device, the acoustic propagation characteristics of each individual device are used as a personalized reference benchmark, thus solving the problem of false alarms due to individual differences caused by the use of a universal threshold in the prior art.

[0012] By utilizing the inherent process switching motions of the equipment as a passive self-excitation source, pulse response waveforms are continuously acquired during normal production, eliminating the need for additional signal injection or shutdown for testing. This solves the problem of existing technologies requiring downtime for maintenance or manual intervention, enabling online health self-inspection without interrupting production. Furthermore, by calculating the overall anomaly degree of components and setting dual thresholds for concern and suspicion, false alarms caused by single disturbances or transient noise from sensors are resolved.

[0013] For suspicious components detected by passive monitoring, active ultrasonic detection is automatically triggered during periods of low-speed equipment operation or planned shutdown. By emitting coded ultrasonic pulses and receiving reflected echoes, the acoustic impedance change rate and geometric anomaly of the equipment's structural surface are independently obtained, solving the blind spot problem of existing technologies that cannot monitor when the equipment's spontaneous sound is weak, and significantly improving the specificity and reliability of anomaly detection. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the industrial equipment health assessment method based on acoustic analysis in an embodiment of the present invention; Figure 2 This is a schematic diagram of the industrial equipment health assessment system architecture based on acoustic analysis in an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0017] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0018] Example 1, as Figure 1 As shown, an industrial equipment health assessment method based on acoustic analysis is described, the method comprising: S1: Set up several excitation reference points and several sets of microphone arrays, apply a reference response, calculate the transfer function for each frequency to generate a transfer function matrix; determine the standard process work set according to the standard operating state of the equipment, and collect the measured time-domain waveform of the microphone array for each action to obtain the reference impulse response.

[0019] After the equipment is newly put into operation or after a complete shutdown and maintenance confirming no faults, original reference data is collected. Specifically, several excitation reference points, denoted as L, are selected at key structural locations on the equipment casing, including the outer surfaces of each bearing housing, the centers of each surface of the gearbox, the front and rear end covers of the motor, and the ends of the guide rail sliders. Each excitation reference point... Each represents a corresponding device or device component. Several microphone arrays, denoted as M (M≥4), are evenly deployed on the workshop roof. Each array contains one main microphone and one auxiliary microphone. Simultaneously, a single-axis accelerometer is attached at each excitation reference point to calibrate the acoustic transmission path. Preferably, L represents 10-30 key component locations, and M represents 4-16 microphone arrays.

[0020] The purpose of applying a reference response is to obtain the acoustic propagation characteristics of the equipment under healthy conditions, serving as a reference benchmark for subsequent monitoring. By applying a pulse excitation of known force, the acoustic response received by the microphone is recorded, thereby calculating the transfer function matrix and the reference pulse response. Specifically, a standard force hammer is used to apply a pulse excitation of a defined force to each excitation reference point, while simultaneously recording the sound pressure signals from all microphone channels and the vibration signals from the accelerometer; alternatively, the inherent process switching actions during the initial power-on of the equipment (such as gripper closure, tool return, and table reversal) are used, and the position and force of these actions are set according to the normal operating state of the equipment, serving as a standard pulse source. The acquired signals are filtered for anti-aliasing and converted to 24-bit analog-to-digital conversion, with the sampling rate set to 200kHz to ensure coverage of audible sound and part of the ultrasonic frequency band.

[0021] Let the set of excitation reference points be... The microphone channels are set as By comparing the cross spectrum and autospectrum of the excitation signal and the response signal, the transfer function is calculated for each frequency, generating a transfer function matrix. The transfer function is defined as follows:

[0022] in, It is a transfer function; For frequency variables, the range of values ​​is ; For the first The excitation signals of the excitation reference points and the response signal of the m-th microphone The cross-power spectral density at time t reflects the degree of correlation between the two signals at frequency f, including amplitude correlation and phase difference information; Let be the self-power spectral density of the excitation signal, representing the distribution of the excitation signal's energy with frequency. When obtaining the transfer function matrix, the Welch method is used to estimate the power spectral density. Specifically, the excitation signal and the microphone response signal are each divided into K segments, each segment having 2048 sampling points, with adjacent segments overlapping by 50%. Performing a Fast Fourier Transform on each segment of data is a well-known technique in the art, and this application does not impose specific limitations on it.

[0023] By eliminating the influence of the excitation signal's own spectral unevenness based on the transfer function, the pure channel transmission characteristics are obtained, thereby accurately identifying the structural characteristics. The complex frequency domain response characteristics between the excitation point and the m-th microphone after the combined effects of solid conduction and air radiation. The modulus characterizes the amplitude frequency attenuation and the phase characterizes the phase frequency delay.

[0024] Based on the equipment's standard operating conditions, a standard procedure set is determined. For each action, a reference impulse response is obtained by acquiring the measured time-domain waveform of the microphone array. Specifically, let the standard procedure set include type A actions, denoted as... Under healthy conditions, for each action 'a', microphone channel 'm', and acquisition time 't', the measured time-domain waveform of the microphone array is acquired and background noise is subtracted to obtain the reference impulse response:

[0025] The reference impulse response covers the arrival times of the pulse main lobe and the reflections of the main structures. The main structures here correspond to the equipment structures at the excitation reference points mentioned above. The specific selection needs to be based on actual requirements or on the standard industry rules used in the current equipment. This application does not impose specific restrictions here.

[0026] S2: Obtain the execution time and action type number of each process action to obtain the measured pulse response waveform; calculate the deviation index according to the transfer function matrix, the reference pulse response and the measured pulse response waveform respectively; the deviation index includes the normalized cross-correlation coefficient, pulse energy ratio, pulse width expansion coefficient and peak attenuation coefficient.

[0027] After the equipment is put into normal production operation, it is connected to the PLC data bus in real time to obtain the execution time and action type number 'a' of each process action. After each action 'a' is triggered... Raw sound pressure data is collected within the time window and denoted as... , where d represents the date of the d-th detection. (Using the action before...) The root mean square value of background noise in channel m is estimated during the pulse-free period. After subtracting background noise from the original sound pressure data, the measured impulse response waveform is obtained as follows: To eliminate the influence of ambient temperature and humidity on the speed of sound, temperature and humidity sensors are placed at fixed locations in the workshop to correct the speed of sound propagation in the air in real time. The time axis of the measured impulse response waveform is also linearly calibrated to the reference temperature conditions (e.g., ...). ), to ensure consistent sound speed.

[0028] The deviation index is calculated based on the transfer function matrix, the reference impulse response, and the measured impulse response waveforms. The deviation index includes the normalized cross-correlation coefficient, the pulse energy ratio, the pulse width expansion coefficient, and the peak attenuation coefficient.

[0029] The normalized cross-correlation coefficient is calculated to reflect the shape similarity between the measured waveform and the reference template under optimal time alignment conditions. The formula is as follows:

[0030] in, It is the normalized cross-correlation coefficient; It is the measured impulse response waveform; yes The reference impulse response at time t, with a range of values. The closer the value is to 1, the higher the consistency of the waveform shape; T represents the fixed integration window length, set to 50ms; τ is the time delay variable, with a value range of [-2, 2], in ms. By taking different values ​​for τ, the reference template is translated relative to the measured waveform on the time axis, and the waveform inner product at different alignment positions is calculated to compensate for the time offset between the measured waveform and the reference template. The numerator is the inner product of the measured waveform and the reference template at the optimal time delay τ, representing the degree of matching between the two waveform shapes; the denominator is the geometric mean of the energy of the two waveforms, used for normalization to eliminate the influence of amplitude differences.

[0031] The pulse energy ratio, reflecting the trend of structural damping or stiffness changes, is calculated using the following formula:

[0032] like This indicates a decrease in structural attenuation, localized loosening, or a decrease in contact stiffness; if This indicates accelerated energy decay, suggesting crack initiation, increased damping, or aging of the connection.

[0033] Based on the transfer function matrix The measured impulse response waveforms of each microphone channel are deconvolved in the frequency domain. First, a Fourier transform is performed on the measured impulse response waveforms to obtain... Then divide by the transfer function of the corresponding channel excitation point combination, and then perform an inverse Fourier transform to obtain the result from the first... The equivalent impact force time series derived from each excitation reference point reflects the magnitude attenuation of the impact response along the structural transmission path and complements the overall energy ratio. The formula is:

[0034] in, It is the equivalent impact force, without distinguishing the m-dimensionality of the microphone, and obtains the comprehensive equivalent impact force sequence of all microphone channels after time alignment and weighted fusion; It is a transfer function; It is the measured impulse response waveform, representing the measured impulse response waveform in the time domain, with time as the independent variable; The measured impulse response waveform in the frequency domain is represented by frequency. and These represent the Fourier transform and inverse transform, respectively. Under healthy conditions, the equivalent impact force time series should be a single narrow pulse; when structural stiffness decreases or gaps exist, the pulse broadens and the peak value decreases. The peak attenuation coefficient and pulse width expansion coefficient are extracted based on the equivalent impact force time series.

[0035] The peak attenuation coefficient is calculated as follows:

[0036] in, The standard inversion peak value under healthy conditions (the largest among the results of each m-channel). It is the reference impulse response.

[0037] The pulse width expansion factor is calculated as follows:

[0038] in, It is the pulse width expansion factor; Indicates half-height and full width; The full width at half maximum (FWHM) is the baseline for a healthy state. Under healthy conditions, the FWHM is the average of measurements taken for all actions 'a' and all channels 'm'. FWHM is the width of the pulse waveform between two time points where the amplitude equals half the peak value. Under healthy conditions, the pulse response should be an approximate Dirac pulse, with a very small FWHM. When there is loosening or cracks in the structure, the impact energy is dispersed, and the pulse broadens. Changes in FWHM directly reflect changes in structural stiffness and contact state, and are not redundant with the peak attenuation coefficient.

[0039] In this embodiment, complex acoustic waveforms are compressed into numerical features that are highly sensitive to device health, thereby enabling precise quantitative analysis.

[0040] S3: For each process action and excitation reference point, spatially fuse the deviation indices of multiple microphone channels, calculate the joint anomaly degree of a single action component and the comprehensive anomaly of the component, set a screening threshold range, screen suspicious components and calculate the confidence level, and mark the anomalies.

[0041] For each process action a and the first Using one excitation reference point, the deviation indices of multiple microphone channels are spatially fused. Since different microphones are at different distances and angles from the excitation point, the signal-to-noise ratio (SNR) of each channel's indices is first calculated as a weighted average. SNR is defined as follows:

[0042] For example, the peak attenuation coefficient fusion value is:

[0043] Similarly, the normalized cross-correlation coefficient fusion value can be calculated. Pulse energy ratio fusion value fusion value of pulse width expansion coefficient .

[0044] Multiple microphone channels are capturing the same physical event (i.e., the first one). (Response signals of each component under action a). Although the amplitude and waveform details of each channel differ, they can all be used to invert the equivalent impact force. To obtain a single index representing the overall state of the component, spatial fusion processing is performed on the inversion results of all microphone channels before calculating the deviation index. The measured waveforms of multiple microphone channels are used as input, and a unique equivalent impact force sequence is obtained by solving a system of simultaneous equations (e.g., using the least squares method). This sequence has already integrated the information from all channels. Therefore, the final calculated peak attenuation coefficient and pulse width expansion coefficient no longer need to depend on m, as they are themselves the results of multi-channel fusion. This avoids subsequent weighted averaging, simplifies the number of indices, and preserves key structural dynamic characteristics.

[0045] Calculate the joint anomaly degree of a single action component:

[0046] Weights satisfy Based on historical fault data, the initial value is: The joint anomaly degree of a single action component maps the deviations of different dimensions generated by different components in a single action to the interval [0,1]. The closer the value is to 1, the more abnormal it is. and The normal value is 1, and it decreases when it deviates. When calculating the degree of abnormality, the normal value is 0, and the abnormal value is positive. and The normal value is 1, and it usually increases when it deviates. The ideal range for the joint abnormality of a single action component is equal to 0 when all indicators are normal; when extreme abnormalities occur, the maximum is about 1.0.

[0047] For each component location (i.e., the excitation reference point), a set of all actions involving that component within a week is collected. A mapping matrix between process actions and component locations is pre-established. If the impact energy generated by action 'a' is mainly transferred to the component, it is marked as 1; otherwise, it is marked as 0. The number of times marked 1 and the corresponding actions are counted. The number of detection days within a week is denoted as D. The overall anomaly degree of the component is calculated, reflecting the overall anomaly level of the component within a week. The formula is:

[0048] in, It refers to the overall anomaly degree of the component; For all involved components The set of process action numbers, where D is the number of inspection days within a week.

[0049] Set the filtering threshold range [0.25, 0.55]. If the overall anomaly score of a component is not less than 0.25, mark it as a component of interest; if... The components are marked as suspicious; simultaneously, the confidence level of the passive self-inspection of the suspicious components is calculated:

[0050] in, The action 'a' is executed across all excitation channels, and the number of repeated samples is... The arithmetic mean of the above. The closer it is to 1, the higher the consistency of multiple measurements.

[0051] Components are identified based on their overall anomaly and confidence level. The anomaly indicators include 0, 1, and 2, where 0 indicates normal, 1 indicates attention is needed, and 2 indicates suspicious. The confidence level is a secondary screening of components, its purpose being to further screen and identify suspicious components, eliminating randomness.

[0052] S4: Set up an active detection mechanism, which automatically triggers based on pre-set trigger conditions. After triggering, it obtains the acoustic impedance change of the device structure surface by emitting ultrasonic pulses and receiving reflected echoes, and calculates the impedance change rate and active detection confidence level.

[0053] For components with anomaly markers 1 and 2, an active detection mechanism is implemented. This mechanism automatically triggers based on pre-set conditions. Upon triggering, it emits ultrasonic pulses and receives reflected echoes to obtain changes in the acoustic impedance of the equipment's surface structure. The triggering conditions are that the equipment is in a low spontaneous acoustic energy state or within a preset interruptible production period. Specifically, the active detection mechanism is automatically triggered when the equipment's operating parameters are detected to be below the minimum dynamic threshold required for health acoustic monitoring, or when the production scheduling system confirms that the equipment has entered a non-continuous processing window. These conditions include, but are not limited to: the equipment spindle speed being below a preset percentage threshold of the rated speed (e.g., 20%), the equipment being within a planned shutdown window (e.g., lunch break, tooling change, shift handover, equipment standby), or a period of idle operation with no operational actions. The active detection mechanism is configured such that the center frequency of the miniature ultrasonic transmitting unit integrated in the microphone array is [missing information]. ,bandwidth Linear frequency modulated (LFM) pulse, pulse width Launch interval Continuous launch The transmitted signal is encoded using Gray complement sequence pairs. All microphones simultaneously acquire echoes at a sampling rate of [missing information]. Collection time The acquired echo signal is first matched and filtered with the transmitted coded sequence to obtain a high-resolution echo sequence after pulse compression. Extract each predefined structure reflective surface from the sequence. (Corresponding one-to-one with incentive reference points,) Flight time and echo amplitude The time of flight is obtained by locating the peak position after matched filtering. Specifically, for each microphone channel and each predefined structural reflector, a reasonable time search interval is first calculated based on the straight-line distance between the reflector and the microphone, and the possible range of changes in the ambient sound speed. The lower bound of this interval is equal to the distance divided by the maximum possible sound speed, and the upper bound is equal to the distance divided by the minimum possible sound speed. Then, in the high-resolution echo sequence after matched filtering, the peak point with the largest amplitude is searched within this time interval. The time corresponding to this peak point is the flight time of the sound wave from emission to return to the microphone via the reflector. In actual implementation, to improve noise immunity, the weighted average time of several sampling points near the peak is taken as the final flight time. After finding the peak point within the same time search interval, the amplitude of the peak point is directly taken as the echo amplitude.

[0054] Using the transfer function matrix and combining it with the currently measured ambient sound speed (calculated in real time by temperature and humidity sensors), the theoretical flight time for each reflecting surface is constructed:

[0055] Where c is the ambient sound speed; For the m-th microphone to the reflector The spatial straight-line distance (calculated from the equipment's 3D model).

[0056] The deviation between the actual measured flight time and the theoretical flight time is calculated as follows:

[0057] in, This is the deviation in flight time. Calculate the root mean square of the flight time deviation. The geometric anomaly of the structure is calculated as follows:

[0058] in, It is the geometric anomaly. This is an empirical threshold, taken as 0.5 ms, corresponding to an apparent displacement of approximately 0.17 mm. Geometric anomaly reflects changes in the sound wave propagation path caused by structural loosening or displacement.

[0059] Echo amplitude and acoustic impedance of the reflecting surface Related. Based on the formula for the reflection coefficient of a sound wave at perpendicular incidence, the relationship between the reflection amplitude and impedance can be approximated as follows:

[0060] in, The characteristic impedance of air is a constant. For the center frequency, At frequency The modulus of the complex value of the transfer function measured at the location.

[0061] By measuring the echo amplitudes of M microphone channels, the equivalent structural acoustic impedance of each reflecting surface is inverted using the weighted least squares method. The objective function is defined as follows:

[0062] Where k>0 is the gain constant. In the active detection echo amplitude model, the gain constant is used to quantitatively correlate the theoretical reflection coefficient with the measured echo amplitude. Its theoretical basis is the plane wave perpendicular incidence reflection theorem: when ultrasound is incident perpendicularly from air onto the surface of a solid structure, the amplitude of the reflected echo is proportional to the difference between the acoustic impedance of the solid surface and the air impedance, specifically depending on the absolute value of the reflection coefficient. The actual measured echo amplitude is also affected by various fixed or quasi-fixed factors, such as the gain of the transmitting circuit, the conversion efficiency of the electroacoustic transducer, the attenuation of sound waves propagating in air, the amplification factor at the receiving end, and multipath propagation. The combined effect of interference factors is characterized by a single overall gain constant k. All gains and losses in the entire link from transmission to reception, excluding the reflection coefficient, are combined into a single scalar coefficient. Under healthy equipment conditions, the value of k is calculated through a calibration measurement with known structural impedance. Specifically, when the equipment is first put into operation or confirmed to be healthy and fault-free, active detection is performed to measure the echo amplitude of each microphone channel. Simultaneously, using the obtained transfer function and health reference impedance, the theoretical value of the reflection coefficient is calculated. The ratio of the measured amplitude to the theoretical reflection coefficient (divided by the square of the transfer function amplitude) is taken as k. Since k remains stable during normal equipment use, this k value can be used to back-calculate the current reflection coefficient from the measured echo amplitude during subsequent detections, thereby calculating the real-time acoustic impedance of the equipment's structural surface.

[0063] By solving the optimization problem, we obtain the result that... Minimum acoustic impedance Then, the rate of change of impedance is calculated:

[0064] in, It is the rate of change of impedance; The health reference impedance is obtained by pre-calibration using the same active probing procedure under the initial health condition of the equipment. Simultaneously, the active probing confidence level is calculated.

[0065] in, This represents the standard deviation of the time-of-flight measurements for each microphone channel; a larger standard deviation indicates a lower echo signal-to-noise ratio and a lower confidence level. According to general confidence level guidelines, only impedance change rates with a confidence level greater than 0.8 are retained.

[0066] The rate of change of impedance is defined as the absolute value of the difference between the current impedance and the healthy reference impedance, divided by the healthy reference impedance. The healthy reference impedance is pre-measured under the same active probing procedure when the equipment is initially in good health. The rate of change of impedance reflects the degree of change in structural stiffness, connection tightness, or material integrity. An increase in impedance usually indicates surface hardening or reduced gaps, while a decrease in impedance indicates loosening or material softening.

[0067] S5: Align the component anomaly degree, anomaly identifier, and impedance change rate in time and space, define an indicator function, calculate the comprehensive anomaly score of a single component and the overall health index of the equipment, and output the warning level and maintenance list according to the screening rules; the indicator function includes a passive self-test effective indicator function and an active detection effective indicator function.

[0068] The obtained component anomaly degree S( ) and anomaly indicators , and the obtained rate of change of impedance Perform spatiotemporal alignment and excitation reference point With reflective surface The numbering must be strictly consistent, that is Define the characteristic function:

[0069]

[0070] This is a passive self-check validity indicator function, used to determine whether to include the passive self-check results in the fusion formula. When the overall anomaly degree of a component reaches the attention threshold, If the passive self-test detects an abnormality, it is set to 1; otherwise, it is set to 0.

[0071] It is an effective indication function for active detection. When the impedance change rate... If the change exceeds 10%, it is considered that the active detection has detected a significant structural change, and is set to 1; otherwise, it is set to 0. The threshold of 0.1 is set according to the experiment; values ​​below this value are considered to be measurement noise or normal fluctuations.

[0072] A confidence-weighted fusion formula is introduced to calculate the comprehensive anomaly score for a single component:

[0073] in, To prevent small regularization constants with a denominator of zero; a normalization factor of 0.3 is used, which is an empirically set maximum reference impedance change rate, to ensure that the impedance change rate is normalized to... , and when The time-forced truncation is 1. When (i.e., when neither passive nor active exception detection is triggered) define .

[0074] Calculate the overall health index of the equipment:

[0075] in, It is the overall health index of the equipment. The exponential function has non-linear decay characteristics, which significantly amplifies the cumulative effect of multiple minor anomalies, consistent with the physical evolution law of equipment fatigue damage.

[0076] Estimated remaining useful life:

[0077] in, The rated service life of the equipment at the time of manufacture. The calibration coefficient represents the average comprehensive anomaly score when it reaches a certain level. At that point, the remaining lifespan theoretically becomes zero.

[0078] Output warning level, the filtering rule is: when If a yellow alert is triggered, it is recommended to schedule planned maintenance. when Or there exists any component that satisfies If a red alert is triggered, immediate shutdown and maintenance are recommended. A list of maintenance recommendations will also be output, sorted by priority as follows: High priority: and (Dual confirmation of abnormality, both passive and active); Medium priority: but (Passive detection only, active detection not confirmed), it is recommended to re-inspect or add sensor points to improve spatial coverage and signal-to-noise ratio; Low priority: but (Active detection only, passive no response) It is necessary to check whether the vibration excitation source is in a dormant state or the coupling has failed.

[0079] In this embodiment, by establishing the transfer function matrix and reference impulse response template under the health state of the device, the acoustic propagation characteristics of each individual device are used as a personalized reference benchmark, thus solving the problem of false alarms due to individual differences caused by the use of a general threshold in the prior art.

[0080] By utilizing the inherent process switching motions of the equipment as a passive self-excitation source, pulse response waveforms are continuously acquired during normal production, eliminating the need for additional signal injection or shutdown for testing. This solves the problem of existing technologies requiring downtime for maintenance or manual intervention, enabling online health self-inspection without interrupting production. Furthermore, by calculating the overall anomaly degree of components and setting dual thresholds for concern and suspicion, false alarms caused by single disturbances or transient noise from sensors are resolved.

[0081] For suspicious components detected by passive monitoring, active ultrasonic detection is automatically triggered during low-speed equipment operation or planned shutdown periods. By emitting coded ultrasonic pulses and receiving reflected echoes, the acoustic impedance change rate and geometric anomaly of the equipment's structural surface are independently acquired. This overcomes the blind spot problem of existing technologies being unable to monitor when the equipment's spontaneous sound is weak, significantly improving the specificity and reliability of anomaly detection. It represents a shift from passive monitoring to active detection.

[0082] By integrating the overall health index and remaining service life of the equipment from the early warning module, the multi-dimensional anomaly score is mapped to an intuitive percentage health value. Based on a dual-confirmation rule, tiered early warnings and priority maintenance recommendations are output. This allows for planned maintenance before actual equipment failure, avoiding significant economic losses caused by sudden production stoppages.

[0083] Example 2: The present invention also provides an industrial equipment health assessment system based on acoustic analysis, such as... Figure 2 As shown, the system includes: The benchmark modeling module is used to determine the measured time-domain waveform by applying a benchmark response (known pulse excitation) and utilizing the inherent process switching actions of the equipment when the equipment is initially put into operation and confirmed to be healthy and fault-free. It also uses the inherent process switching actions of the equipment to collect the acoustic response data of the microphone array, calculate the transfer function matrix at each frequency, and the benchmark pulse response corresponding to each standard process action.

[0084] The passive monitoring module is used to access the PLC data bus in real time during normal equipment operation to obtain the execution time and type number of each process action, collect the measured pulse response waveform triggered by the action, and calculate deviation indicators based on the transfer function matrix and reference pulse response provided by the benchmark modeling module. These deviation indicators include normalized cross-correlation coefficient, pulse energy ratio, peak attenuation coefficient, and pulse width expansion coefficient deviation indicators. This module enables passive acoustic health self-checking during continuous equipment production.

[0085] The anomaly identification module receives various deviation indicators output by the passive monitoring module, performs signal-to-noise ratio weighted spatial fusion on the indicators from multiple microphone channels, calculates the joint anomaly degree of a single action and the comprehensive anomaly degree of the components over a week; identifies components as anomalies based on preset threshold ranges, and calculates the confidence level of passive self-testing. It then filters out suspicious components that may have health problems.

[0086] The active detection module is used to set up an active detection mechanism. Based on pre-set trigger conditions, it automatically triggers the detection. After triggering, it emits ultrasonic pulses and receives reflected echoes to obtain changes in acoustic impedance on the equipment's structural surface, calculating the impedance change rate and active detection confidence level. Specifically, for components of interest and suspected components output by the anomaly identification module, the active detection mode is automatically triggered when the equipment is in a low spontaneous acoustic energy state or within a planned shutdown window. It controls the ultrasonic transmitting unit integrated in the microphone array to emit coded ultrasonic pulses and receive echoes from the equipment's structural reflective surfaces. Through matched filtering and inversion calculations, it obtains the time-of-flight deviation, geometric anomaly, acoustic impedance change rate, and active detection confidence level for each reflective surface. This enables independent active verification and quantitative assessment of the structural condition of passively detected components.

[0087] The fusion early warning module receives the component comprehensive anomaly degree, anomaly identifier, passive confidence degree output by the anomaly identification module, and impedance change rate output by the active detection module; it performs spatiotemporal alignment on the data, calculates the comprehensive anomaly score of a single component using a confidence-weighted fusion formula, and then calculates the overall health index of the equipment, outputting the early warning level and maintenance list according to preset rules.

[0088] In some embodiments, the benchmark modeling module establishes an acoustic propagation fingerprint of the equipment under health conditions, including a transfer function matrix and a benchmark impulse response template. The transfer function matrix records the frequency domain transmission characteristics from each structural excitation point to each microphone, eliminating the influence of the excitation source's own spectral unevenness; the benchmark impulse response template records the time domain waveform generated by each standard process action under health conditions. This provides all subsequent monitoring data with an accurate physical reference benchmark, rather than relying on a statistically significant baseline. Since this benchmark is acquired when the equipment is healthy, any measured response deviating from this benchmark can be quantified as a deviation index, thereby avoiding false alarms caused by individual equipment differences or differences in the workshop acoustic environment.

[0089] The passive monitoring module operates in real-time during normal equipment production, utilizing existing process switching actions as a natural self-excitation source, requiring no additional signal injection or shutdown. By calculating four deviation indicators, it converts complex acoustic waveforms into numerical characteristics highly sensitive to equipment health. The normalized cross-correlation coefficient reflects the degree of waveform distortion, capturing waveform changes caused by minute structural variations; the pulse energy ratio reflects macroscopic changes in structural damping or stiffness—increased energy indicates loosening, while decreased energy indicates increased damping or cracks; the peak attenuation coefficient and pulse width expansion coefficient are calculated based on the equivalent impact force after deconvolution, directly characterizing the amplitude attenuation and pulse broadening of the structural impact response, complementing the energy ratio. This enables continuous health monitoring without interrupting production, with each indicator having a clear physical meaning. Compared to existing technologies that only collect sound amplitude or spectrum, this module effectively eliminates background noise and aliasing interference from other equipment by locking and deconvolutioning process pulses, significantly improving the signal-to-noise ratio and early fault detection capabilities.

[0090] The anomaly identification module statistically fuses a large amount of single-time monitoring data generated by the passive monitoring module. First, it uses signal-to-noise ratio weighted fusion to combine the indicators from multiple microphone channels into a single component-level indicator, eliminating the impact of microphone position differences. Then, it calculates the joint anomaly degree of a single action component, mapping four different dimensions of deviation indicators to a unified [0,1] interval. Finally, it accumulates all actions involving the component within a week to obtain the component's comprehensive anomaly degree. Components are categorized into three types: normal, of concern, and suspicious, and confidence levels are calculated to exclude random fluctuations. This quickly identifies components with genuinely abnormal trends, and through statistical averaging of multiple days and actions, it eliminates false alarms caused by single disturbances or transient sensor noise, improving the reliability of early warnings.

[0091] The active detection module is automatically triggered during low-speed equipment operation or planned shutdown periods, actively detecting suspicious components marked by the anomaly identification module. Unlike passive monitoring, which relies on the equipment's own emitted sounds, active detection acquires changes in acoustic impedance and geometric displacement on the equipment's structural surface by emitting coded ultrasonic pulses and receiving echoes. The impedance change rate reflects quantitative changes in structural stiffness, connection tightness, or material integrity; the geometric anomaly is calculated using the root mean square of the time-of-flight deviation, reflecting structural loosening or relative displacement. The active detection confidence level is calculated, automatically lowering the confidence level when the time-of-flight measurement standard deviation is too large, avoiding unreliable results under low signal-to-noise ratio conditions. In situations where the equipment's own emitted sounds are weak or the equipment is shut down, it provides a quantitative assessment method independent of passive monitoring; cross-validation between the two significantly improves the specificity and accuracy of the diagnosis.

[0092] The fusion early warning module receives all output data from the passive monitoring and active detection links and calculates a comprehensive anomaly score for each component using a confidence-weighted fusion formula. The fusion formula is designed to fully consider the confidence levels of both detection methods: when passive self-inspection has high confidence and the anomaly is clear, its contribution weight increases accordingly; when active detection has high confidence and significant impedance or geometric changes, its contribution weight increases. The highest score is achieved when both methods simultaneously confirm an anomaly. Based on the comprehensive component anomaly score, the overall equipment health index and remaining service life are further calculated. The health index uses an exponential function, non-linearly amplifying the cumulative effect of multiple minor anomalies, which better reflects the physical laws of equipment fatigue damage; the remaining service life is estimated based on the factory rated life and the linear decay of the average anomaly score. Finally, an early warning is output along with a priority-ordered list of maintenance recommendations.

[0093] In summary, this system achieves a complete workflow from benchmark establishment, passive monitoring, anomaly screening, proactive verification to integrated early warning through the collaborative work of five modules. Each module's function and methodological steps correspond one-to-one, and the data flow between modules is clear and unambiguous. It solves key technical problems in traditional acoustic monitoring methods, such as high false alarm rates due to reliance on fixed thresholds, inability to distinguish between normal operating condition fluctuations and early faults, monitoring gaps during low-speed or shutdown periods, and lack of cross-validation methods.

[0094] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0095] To make the technical solution of the present invention clearer, the above specific embodiments may involve several specific numerical values, thresholds, parameters, formulas, and examples. Those skilled in the art should understand that these specific contents are merely exemplary descriptions given for ease of implementation and are not essential technical features necessary for realizing the present invention, nor do they constitute a limitation on the scope of protection of the present invention. The mathematical formulas, parameter values, model structures, calculation rules, constraints, and data acquisition methods described in the specific embodiments section of this application are preferred examples for realizing the technical concept of the present invention, not the only implementation methods. Based on the core logic and inventive concept disclosed in this application, those skilled in the art can make reasonable equivalent substitutions, parameter adjustments, or logical optimizations to the parameters, formulas, and algorithm models according to actual conditions, industry standards, or conventional experimental methods, without producing substantial differences. Any changes that do not depart from the inventive concept of this application fall within the scope of this application. All mathematical expressions in this application should ensure that their physical meaning is clear when applied; parameters such as weights and thresholds can be calibrated using statistical or machine learning methods known in the art.

[0096] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0098] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0099] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0100] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0101] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for health assessment of industrial equipment based on acoustic analysis, characterized in that, include: S1: Set up several excitation reference points and several microphone arrays, apply a reference response, and calculate the transfer function for each frequency to generate a transfer function matrix; Based on the standard operating status of the equipment, the standard process work set is determined, and the reference impulse response is obtained by collecting the measured time-domain waveform of the microphone array for each action; S2: Obtain the execution time and action type number of each process action to obtain the measured pulse response waveform; The deviation index is calculated based on the transfer function matrix, the reference impulse response, and the measured impulse response waveforms; the deviation index includes the normalized cross-correlation coefficient, the pulse energy ratio, the pulse width expansion coefficient, and the peak attenuation coefficient. S3: For each process action and excitation reference point, the deviation indicators of multiple microphone channels are spatially fused to calculate the joint anomaly degree of a single action component and the comprehensive anomaly of the component. A screening threshold range is set to screen suspicious components and calculate the confidence level for anomaly identification. Let the set of excitation reference points be . The microphone channels are set as Suppose that the standard process motion set includes type A motions, denoted as... .

2. The industrial equipment health assessment method based on acoustic analysis according to claim 1, characterized in that, The method further includes: S4: Set up an active detection mechanism, which automatically triggers based on pre-set trigger conditions. After triggering, it obtains the acoustic impedance change of the equipment structure surface by emitting ultrasonic pulses and receiving reflected echoes, and calculates the impedance change rate and active detection confidence level. S5: Align the component anomaly degree, anomaly identifier, and impedance change rate in time and space, define an indicator function, calculate the comprehensive anomaly score of a single component and the overall health index of the equipment, and output the warning level and maintenance list according to the screening rules; the indicator function includes a passive self-test effective indicator function and an active detection effective indicator function.

3. The industrial equipment health assessment method based on acoustic analysis according to claim 1, characterized in that, In step S1, the transfer function is calculated for each frequency by the ratio of the cross spectrum to the self spectrum of the excitation signal and the response signal, generating a transfer function matrix. The transfer function is defined as follows: in, It is a transfer function; For frequency variables, the range of values ​​is ; For the first The excitation signals of the excitation reference points and the response signal of the m-th microphone The cross-power spectral density at time t reflects the degree of correlation between the two signals at frequency f, including amplitude correlation and phase difference information; Let be the self-power spectral density of the excitation signal, representing the distribution of the excitation signal's energy with frequency; Under healthy conditions, for each action 'a', microphone channel 'm', and acquisition time 't', the measured time-domain waveform of the microphone array is acquired and background noise is subtracted to obtain the reference impulse response: .

4. The industrial equipment health assessment method based on acoustic analysis according to claim 1, characterized in that, The measured impulse response waveform was obtained after each action a was triggered. Raw sound pressure data is collected within the time window and denoted as... Where d represents the date of the d-th detection; using the action before The root mean square value of background noise in channel m is estimated during the pulse-free period. After subtracting background noise from the original sound pressure data, the measured impulse response waveform is obtained as follows: .

5. The industrial equipment health assessment method based on acoustic analysis according to claim 1, characterized in that, In step S2, the normalized cross-correlation coefficient is calculated using the following formula: in, It is the normalized cross-correlation coefficient; It is the measured impulse response waveform; yes The reference impulse response at time t, with a range of values. T represents the fixed integration window length; τ is the time delay variable, with a value range of [-2, 2] and a unit of ms. Based on the transfer function matrix For each microphone channel, the measured impulse response waveform is deconvolved in the frequency domain. First, a Fourier transform is performed on the measured impulse response waveform to obtain... Then divide by the transfer function of the corresponding channel excitation point combination, perform an inverse Fourier transform, and obtain the result from the first... The equivalent impact force time series is derived from each excitation reference point, and the peak attenuation coefficient and pulse width expansion coefficient are extracted based on the equivalent impact force time series.

6. The industrial equipment health assessment method based on acoustic analysis according to claim 1, characterized in that, In step S3, the signal-to-noise ratio is defined: in, It's the signal-to-noise ratio; It is the measured impulse response waveform; The root mean square value of the background noise; the peak attenuation coefficient fused value is: Similarly, the normalized cross-correlation coefficient fusion value can be calculated. Pulse energy ratio fusion value Fusion value with pulse width expansion coefficient ; Calculate the joint anomaly degree of a single action component: in, It is the joint anomaly degree of a single action component; the weights satisfy... Based on historical fault data, the initial value is: The deviations of different dimensions generated by different components in a single action are uniformly mapped to the interval [0,1], and the closer the value is to 1, the more abnormal it is. The formula for calculating the overall anomaly degree of a component is: in, For all involved components The set of process action numbers, where D is the number of inspection days within a week.

7. The industrial equipment health assessment method based on acoustic analysis according to claim 2, characterized in that, The triggering condition is that the equipment is in a low spontaneous acoustic energy state or a preset interruptible production period. That is, when the equipment operating parameters are detected to be lower than the minimum dynamic threshold required for health acoustic monitoring or when the production scheduling system confirms that the equipment has entered a non-continuous processing window, the active detection mechanism is automatically triggered.

8. The industrial equipment health assessment method based on acoustic analysis according to claim 2, characterized in that, The active detection mechanism is configured such that the center frequency of the transmission of the miniature ultrasonic transmitting unit integrated in the microphone array is [missing information]. ,bandwidth Linear frequency modulated pulse, pulse width Launch interval Continuous launch Next; all microphones simultaneously collect echoes, sampling rate Collection time The acquired echo signal is first matched and filtered with the transmitted coded sequence to obtain a high-resolution echo sequence after pulse compression. ; Extract each predefined structure reflector from the echo sequence. Flight time and echo amplitude By measuring the echo amplitude of M microphone channels, the equivalent structural acoustic impedance of each reflecting surface is inverted using the weighted least squares method. An objective function is defined, the acoustic impedance that satisfies the objective function is solved, and the impedance change rate is calculated.

9. The industrial equipment health assessment method based on acoustic analysis according to claim 2, characterized in that, In step S5, define the characteristic function: It is a passive self-check validity indicator function that determines whether to include the passive self-check results in the fusion formula; When the overall anomaly level of a component reaches the attention threshold, an anomaly is identified. If the passive self-test detects an abnormality, set it to 1; otherwise, set it to 0. It is an effective indication function for active detection, when the impedance change rate If the active detection detects a structural change, it is set to 1; otherwise, it is set to 0. A confidence-weighted fusion formula is introduced to calculate the comprehensive anomaly score of a single component and the overall health index of the equipment.

10. An industrial equipment health assessment system based on acoustic analysis, applied to the industrial equipment health assessment method based on acoustic analysis as described in any one of claims 1 to 9, characterized in that, The system includes: The benchmark modeling module is used to determine the measured time-domain waveform by applying a benchmark response and utilizing the inherent process switching actions of the equipment when the equipment is initially put into operation and confirmed to be healthy and fault-free. It also calculates the transfer function matrix at each frequency and the benchmark impulse response corresponding to each standard process action. The passive monitoring module is used to acquire the execution time and type number of each process action during normal production operation, collect the measured pulse response waveform triggered by the action, and calculate the deviation index based on the transfer function matrix and the reference pulse response. The deviation index includes the normalized cross-correlation coefficient, pulse energy ratio, peak attenuation coefficient and pulse width expansion coefficient deviation index. The anomaly identification module receives various deviation indicators output by the passive monitoring module, performs signal-to-noise ratio weighted spatial fusion on the indicators of multiple microphone channels, calculates the joint anomaly degree of a single action component and the comprehensive anomaly degree of the component within a week; identifies components as anomalies according to a preset threshold range, and calculates the confidence level of passive self-testing. The active detection module is used to set up an active detection mechanism, which is automatically triggered based on pre-set trigger conditions. After triggering, it obtains the acoustic impedance change of the device structure surface by emitting ultrasonic pulses and receiving reflected echoes, and calculates the impedance change rate and active detection confidence level. The fusion early warning module receives the component comprehensive anomaly degree, anomaly identifier, passive confidence degree output by the anomaly identification module, and impedance change rate output by the active detection module; it performs spatiotemporal alignment on the data, calculates the comprehensive anomaly score of a single component using a confidence-weighted fusion formula, and then calculates the overall health index of the equipment, outputting the early warning level and maintenance list according to preset rules.

Images