Intelligent Analysis Method and System for Patrol Inspection Data of Industrial Equipment and Facilities in Multiple Scenarios

Abstract

The present invention discloses an intelligent analysis method and system for multi-scenario industrial equipment and facility inspection data, specifically related to the technical field of equipment status monitoring, and is used to solve the problems of noise interference and false feature misjudgment caused by the physical property differences and non-linear coupling of multi-source sensor data in the prior art; by dynamically dividing the initial frequency band range based on the equipment operation parameters and sensor physical response characteristics, and combining the load rate and historical data to generate frequency band rules matching the working conditions; by quantifying the non-linear coupling strength between multi-frequency band signals to construct a multi-dimensional matrix, stripping the non-linear interference components of high-frequency noise on low-frequency signals; generating dynamic weight coefficients based on the load rate and energy time series synchronization to achieve cross-frequency band energy fusion and equipment status feature enhancement; finally, excluding false feature interference through a multi-level verification mechanism of energy amplitude, time series continuity and frequency band correlation, and outputting a diagnosis result matching the fault feature library.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment status monitoring. More specifically, the present invention relates to an intelligent analysis method and system for multi-scenario industrial equipment and facility inspection data. Background Art

[0002] In the multi-scenario inspection of industrial equipment and facilities, heterogeneous data such as vibration, temperature, and acoustics of equipment are usually collected synchronously by multiple types of sensors, and the comprehensive analysis of equipment status and fault warning are realized through data fusion technology. The existing technology relies on the time series alignment, feature extraction, and pattern recognition of multi-modal data to improve the accuracy of detection results. However, the multi-source sensor data in the complex industrial field often have physical property differences and signal coupling effects, resulting in interference noise during the fusion process, which directly affects the reliability of subsequent analysis.

[0003] In the existing technology, due to the influence of sensor physical property differences and environmental interference during the acquisition and fusion of multi-modal data, the non-linear interaction between different sensor signals will cause the problem of noise coupling amplification. This noise coupling amplification problem will generate pseudo-features highly similar to the true fault features during the data fusion stage, resulting in misjudgment of equipment status and reducing the fault detection accuracy and reliability of the inspection. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the existing technology, the embodiments of the present invention provide an intelligent analysis method and system for multi-scenario industrial equipment and facility inspection data to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An intelligent analysis method for multi-scenario industrial equipment and facility inspection data, comprising the following steps:

[0007] S1. Obtain the operating parameters of industrial equipment and facilities, and determine the initial frequency band division range of multi-source data in combination with the physical response characteristics of multiple types of sensors;

[0008] S2. Adjust the initial frequency band division range to generate a frequency band division rule matching the working conditions;

[0009] S3. Construct a multi-dimensional coupling strength matrix constrained by the frequency band energy distribution;

[0010] S4. Strip the non-linear coupling noise components in the multi-source data according to the coupling strength priority of each frequency band in the multi-dimensional coupling strength matrix to obtain the target frequency band energy characteristics;

[0011] S5. Generate a feature fusion dynamic weight coefficient associated with the load based on the time series synchronization relationship between the equipment load rate and the target frequency band energy characteristics;

[0012] S6. Cross - band energy superposition is performed on the frequency - domain features of multi - source data according to the dynamic weight coefficient to generate a device - state fusion feature vector;

[0013] S7. Based on the key - band energy distribution in the device - state fusion feature vector that matches the fault feature library, pseudo - band features are excluded through multi - level verification, and a fault diagnosis result is output.

[0014] In a preferred embodiment, the construction logic of the multi - dimensional coupling strength matrix is as follows: Based on the frequency - band division rule, the energy accumulation rate of the signals within each frequency band is extracted, and the non - linear coupling strength between each frequency band is calculated through non - linear correlation analysis and noise transfer parameters between multi - band signals to construct a multi - dimensional coupling strength matrix constrained by the frequency - band energy distribution.

[0015] In a preferred embodiment, the operating parameters of industrial equipment and facilities are obtained, and combined with the physical response characteristics of multi - type sensors, the initial frequency - band division range of multi - source data is determined, including:

[0016] Obtain the rotational speed, temperature, and pressure of industrial equipment and facilities, and extract the resonant frequency, sensitivity, and linear response range of multi - type sensors as physical response characteristics;

[0017] According to the resonant frequency of the sensor, the frequency - domain response range is divided, and the effective - band boundaries of each sensor are determined in combination with the sensitivity and linear response range;

[0018] Based on the rotational speed and temperature in the operating parameters, the initial frequency - band division range is dynamically adjusted within the effective - band boundaries to generate an initial frequency - band division result including a low - frequency mechanical vibration band, an intermediate - frequency heat - conduction band, and a high - frequency acoustic - noise band.

[0019] In a preferred embodiment, the initial frequency - band division range is adjusted to generate a frequency - band division rule matching the working conditions, including:

[0020] According to the equipment load rate and the frequency - band energy distribution law under the same load in historical operation data, the energy - proportion thresholds of the low - frequency mechanical vibration band, the intermediate - frequency heat - conduction band, and the high - frequency acoustic - noise band are extracted;

[0021] Based on the current rotational speed and temperature parameters of the equipment, the frequency - boundary offset of the low - frequency mechanical vibration band and the intermediate - frequency heat - conduction band is matched, and the initial frequency - band division range is dynamically expanded or compressed;

[0022] According to the environmental noise intensity, the lower frequency limit of the high - frequency acoustic - noise band is adjusted to generate a frequency - band division rule matching the working conditions including the updated low - frequency, intermediate - frequency, and high - frequency band ranges.

[0023] In a preferred embodiment, the energy accumulation rate of signals within each frequency band is extracted based on the frequency band division rule, and the non-linear coupling strength between each frequency band is calculated through non-linear correlation analysis between multi-frequency band signals and noise transfer parameters to construct a multi-dimensional coupling strength matrix with the frequency band energy distribution as a constraint, including:

[0024] Extract the energy accumulation rate of signals in the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band based on the frequency band division rule, and calculate the energy change gradient of each frequency band per unit time;

[0025] Through the phase synchronization analysis between the low-frequency mechanical vibration band and the high-frequency acoustic noise band, combined with the sensor sensitivity and linear response range in the noise transfer parameters, calculate the non-linear coupling strength between the frequency bands;

[0026] Taking the frequency band energy distribution as a constraint condition, map the energy change gradient and non-linear coupling strength of each frequency band to the matrix row and column elements, and generate a multi-dimensional coupling strength matrix including low-frequency, medium-frequency, and high-frequency coupling relationships.

[0027] In a preferred embodiment, according to the coupling strength priority of each frequency band in the multi-dimensional coupling strength matrix, strip the non-linear coupling noise components in the multi-source data to obtain the target frequency band energy characteristics, including:

[0028] According to the coupling strength values of the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band in the multi-dimensional coupling strength matrix, generate a coupling strength priority sequence in descending order;

[0029] Based on the coupling strength priority sequence, sequentially separate the signals in the overlapping frequency bands of the high-frequency acoustic noise band and the low-frequency mechanical vibration band in the multi-source data, and filter out the non-linear coupling noise components;

[0030] Verify the energy ratio threshold of the remaining frequency bands through the frequency band energy distribution constraint condition, and remove the residual noise components that do not meet the threshold constraint;

[0031] Merge the energy characteristics of the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band after noise stripping to generate the target frequency band energy characteristics.

[0032] In a preferred embodiment, based on the time series synchronization relationship between the equipment load rate and the target frequency band energy characteristics, generate a feature fusion dynamic weight coefficient associated with the load, including:

[0033] According to the change trend of the equipment load rate and the time series fluctuation direction of the target frequency band energy characteristics, calculate the energy-load correlation coefficients of the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band;

[0034] Based on the energy-load correlation coefficient, determine the dynamic adjustment strategy for the weights of the low-frequency mechanical vibration band and the high-frequency acoustic noise band, and generate a weight increment coefficient that is positively or negatively correlated with the load.

[0035] According to the real-time change rate of the equipment load rate, correct the weight increment coefficient, and combine with the energy stability threshold of the intermediate-frequency heat conduction band to generate the dynamic weight coefficients of the low-frequency, intermediate-frequency, and high-frequency bands.

[0036] Normalize the dynamic weight coefficients so that the sum of the weights of the low-frequency, intermediate-frequency, and high-frequency bands is a fixed value, and generate the load-related feature fusion dynamic weight coefficients.

[0037] In a preferred embodiment, perform cross-band energy superposition on the frequency-domain features of the multi-source data according to the dynamic weight coefficients to generate the equipment state fusion feature vector, including:

[0038] Extract the frequency-domain features of the low-frequency mechanical vibration band, the intermediate-frequency heat conduction band, and the high-frequency acoustic noise band in the multi-source data, and convert them into time-frequency energy values corresponding to the respective frequency bands.

[0039] Based on the frequency band division rule, perform frequency band alignment processing on the time-frequency energy values of the low-frequency mechanical vibration band and the high-frequency acoustic noise band to eliminate the energy conflict in the overlapping region between the frequency bands.

[0040] Allocate the dynamic weight coefficients to the corresponding time-frequency energy values according to the frequency band type, and superimpose the cross-band influence factor of the low-frequency mechanical vibration band on the high-frequency acoustic noise band to generate the weighted superposition energy value.

[0041] According to the dynamic correction coefficient of the weighted superposition energy value corrected by the change rate of the equipment load rate, fuse the stability compensation weight of the intermediate-frequency heat conduction band energy mean to generate the equipment state fusion feature vector.

[0042] In a preferred embodiment, based on the key frequency band energy distribution in the equipment state fusion feature vector that matches the fault feature library, exclude the pseudo-frequency band features through multi-level verification, and output the fault diagnosis result, including:

[0043] Based on the energy distribution of the low-frequency mechanical vibration band, the intermediate-frequency heat conduction band, and the high-frequency acoustic noise band in the equipment state fusion feature vector, match the key frequency band energy amplitude threshold and the time sequence continuity condition of the preset fault mode in the fault feature library.

[0044] Eliminate the pseudo-frequency band features that do not reach the energy accumulation rate threshold of the low-frequency mechanical vibration band or the energy peak threshold of the high-frequency acoustic noise band through energy amplitude verification.

[0045] Based on the inter-band correlation of the target frequency band energy characteristics, verify the time sequence synchronization of the energy fluctuation direction of the remaining frequency band features with the fault mode in the fault feature library.

[0046] Modify the dynamic weight coefficient according to the verification result, suppress the weight of the frequency band that does not match the fault feature library, and output the diagnosis result including the fault type and confidence level.

[0047] On the other hand, the present invention provides an intelligent analysis system for multi-scenario industrial equipment and facility inspection data, including:

[0048] Parameter sensing frequency band module: Obtain the operating parameters of industrial equipment and facilities, and determine the initial frequency band division range of multi-source data in combination with the physical response characteristics of multiple types of sensors;

[0049] Operating condition frequency band rule module: Adjust the initial frequency band division range to generate a frequency band division rule matching the operating condition;

[0050] Multi-dimensional coupling strength module: Construct a multi-dimensional coupling strength matrix with the frequency band energy distribution as a constraint;

[0051] Noise stripping feature module: Strip the non-linear coupling noise components in the multi-source data according to the coupling strength priority of each frequency band in the multi-dimensional coupling strength matrix to obtain the target frequency band energy feature;

[0052] Load dynamic weight module: Generate a feature fusion dynamic weight coefficient associated with the load based on the time series synchronization relationship between the equipment load rate and the target frequency band energy feature;

[0053] Cross-frequency energy fusion module: Perform cross-frequency band energy superposition on the frequency domain features of multi-source data according to the dynamic weight coefficient to generate an equipment state fusion feature vector;

[0054] Multi-level diagnosis output module: Based on the key frequency band energy distribution matching the fault feature library in the equipment state fusion feature vector, exclude the pseudo-frequency band features through multi-level verification, and output the fault diagnosis result.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. Through the dynamic frequency band division rule and multi-dimensional coupling strength analysis, it can effectively distinguish the real equipment state features from the noise coupling pseudo-features, solve the misjudgment problem caused by the physical property differences of sensors and signal coupling in the traditional method, and based on the dynamic weight fusion mechanism of the load rate and energy time series synchronization, make the feature fusion process adapt to the actual operating conditions of the equipment in real time, avoid the feature distortion caused by static weight allocation, enhance the data representation ability in complex industrial scenarios. At the same time, the quantitative modeling of the non-linear coupling strength between frequency bands can accurately strip the interference components of high-frequency noise on low-frequency signals, significantly improving the analysis reliability of multi-source data in the time-frequency domain;

[0057] 2. Through a multi-level verification mechanism, a progressive screening of the energy amplitude, time-series continuity, and frequency-band correlation of the fusion features is carried out, excluding the interference of pseudo-features generated by environmental noise and equipment transient fluctuations from multiple dimensions to ensure a high degree of matching between the diagnostic results and the true fault modes. Combining the preset frequency-band energy distribution rules in the fault feature library, the weight coefficients are dynamically corrected to focus on the key fault frequency bands, making the diagnostic logic both adaptive and traceable. It can not only accurately identify the weak features of early faults but also intuitively locate the fault source through the correlation analysis of the frequency-band energy distribution and the weight coefficients, providing a clear decision-making basis for equipment maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a flowchart of the intelligent analysis method for multi-scenario industrial equipment and facility inspection data of the present invention;

[0059] Figure 2 It is a schematic structural diagram of the intelligent analysis system for multi-scenario industrial equipment and facility inspection data of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0060] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0061] Embodiment 1: Figure 1 An intelligent analysis method for multi-scenario industrial equipment and facility inspection data of the present invention is given, which includes the following steps:

[0062] S1. Obtain the operating parameters of industrial equipment and facilities, and determine the initial frequency-band division range of multi-source data in combination with the physical response characteristics of multi-type sensors;

[0063] S2. Adjust the initial frequency-band division range to generate a frequency-band division rule matching the working conditions;

[0064] S3. Construct a multi-dimensional coupling strength matrix constrained by the frequency-band energy distribution;

[0065] S4. Strip the non-linear coupling noise components in the multi-source data according to the coupling strength priority of each frequency band in the multi-dimensional coupling strength matrix to obtain the target frequency-band energy characteristics;

[0066] S5. Generate a feature fusion dynamic weight coefficient associated with the load based on the time-series synchronization relationship between the equipment load rate and the target frequency-band energy characteristics;

[0067] S6. Cross - band energy superposition is performed on the frequency - domain features of multi - source data according to the dynamic weight coefficients to generate a device - state fusion feature vector;

[0068] S7. Based on the key - band energy distribution in the device - state fusion feature vector that matches the fault feature library, pseudo - band features are excluded through multi - level verification, and a fault diagnosis result is output.

[0069] S1. Obtain the operating parameters of industrial equipment and facilities, and combine the physical response characteristics of multi - type sensors to determine the initial frequency - band division range of multi - source data, including:

[0070] Obtain the rotational speed, temperature, and pressure of industrial equipment and facilities, and extract the resonance frequency, sensitivity, and linear response range of multi - type sensors as physical response characteristics;

[0071] Divide the frequency - domain response range according to the resonance frequency of the sensor, and determine the effective - band boundary of each sensor in combination with the sensitivity and linear response range;

[0072] Based on the rotational speed and temperature in the operating parameters, dynamically adjust the initial frequency - band division range within the effective - band boundary to generate an initial frequency - band division result including a low - frequency mechanical vibration band, a medium - frequency heat - conduction band, and a high - frequency acoustic - noise band.

[0073] The operating parameters of industrial equipment and facilities include rotational speed, temperature, and pressure. The rotational speed is obtained by measuring the rotational speed of the equipment shaft with an optoelectronic encoder. The temperature is collected by a thermocouple sensor for the thermodynamic state on the surface or inside of the equipment. The pressure is monitored by a pressure transmitter for the fluid pressure in the pipeline or cavity. Among the physical response characteristics of multi - type sensors, the resonance frequency is the inherent frequency parameter calibrated at the factory for the sensor, the sensitivity is the proportional coefficient between the sensor output signal and the physical quantity input, and the linear response range is the maximum working interval in which the sensor output signal and the input physical quantity maintain a linear relationship.

[0074] When dividing the frequency - domain response range according to the resonance frequency of the sensor, the frequency - domain response range of the vibration sensor is a preset multiple - frequency interval of the resonance frequency. For example, when the resonance frequency of the vibration sensor is 5 kHz, the frequency - domain response range is divided into 1 kHz to 10 kHz. When determining the effective - band boundary of each sensor in combination with the sensitivity and linear response range, the effective - band boundary is the overlapping area of the frequency - domain response range and the linear response range. For example, when the linear response range of the vibration sensor is 2 kHz to 8 kHz, the effective - band boundary is determined to be 2 kHz to 8 kHz.

[0075] When dynamically adjusting the initial frequency band division range based on the rotational speed and temperature in the operating parameters within the effective frequency band boundary, the frequency range of the low-frequency mechanical vibration band is dynamically adjusted according to the equipment rotational speed. For example, when the equipment rotational speed increases to a preset ratio of the rated speed, the upper limit frequency of the low-frequency mechanical vibration band is correspondingly extended. The center frequency of the intermediate-frequency heat conduction band is adjusted according to the equipment temperature. For example, when the equipment temperature exceeds the preset threshold, the center frequency of the intermediate-frequency heat conduction band shifts towards the low-frequency direction. The lower limit frequency of the high-frequency acoustic noise band is adjusted according to the ambient noise intensity. For example, when the ambient noise intensity increases, the lower limit frequency of the high-frequency acoustic noise band is increased to filter out low-frequency noise interference.

[0076] In the initial frequency band division result, the low-frequency mechanical vibration band is used to monitor the vibration characteristics of the mechanical structure, the intermediate-frequency heat conduction band is used to analyze the thermodynamic state of the equipment, and the high-frequency acoustic noise band is used to capture high-frequency noise signals.

[0077] S2. Adjust the initial frequency band division range to generate a frequency band division rule matching the working conditions, including:

[0078] According to the equipment load rate and the frequency band energy distribution law under the same load in the historical operation data, extract the energy proportion thresholds of the low-frequency mechanical vibration band, the intermediate-frequency heat conduction band, and the high-frequency acoustic noise band;

[0079] Based on the current rotational speed and temperature parameters of the equipment, match the frequency boundary offset of the low-frequency mechanical vibration band and the intermediate-frequency heat conduction band, and dynamically expand or compress the initial frequency band division range;

[0080] Adjust the lower limit frequency of the high-frequency acoustic noise band according to the ambient noise intensity, and generate a frequency band division rule matching the working conditions including the updated low-frequency, intermediate-frequency, and high-frequency band ranges.

[0081] The equipment load rate is the ratio of the current output power of the equipment to the rated power, which is collected in real time through a power sensor. The frequency band energy distribution law under the same load in the historical operation data is obtained by statistically analyzing the energy proportion of multi-source data frequency bands under the same load rate in the historical database. The energy proportion threshold of the low-frequency mechanical vibration band is the minimum proportion of the low-frequency band energy in the overall energy under the same load in the historical data. For example, when the load rate is 70%, the energy proportion threshold of the low-frequency mechanical vibration band is 30%. The energy proportion threshold of the intermediate-frequency heat conduction band is the average proportion of the intermediate-frequency band energy in the overall energy under the same load in the historical data. For example, when the load rate is 70%, the energy proportion threshold of the intermediate-frequency heat conduction band is 45%. The energy proportion threshold of the high-frequency acoustic noise band is the maximum fluctuation proportion of the high-frequency band energy in the overall energy under the same load in the historical data. For example, when the load rate is 70%, the energy proportion threshold of the high-frequency acoustic noise band is 25%.

[0082] The current rotational speed of the device is measured in real time by an optoelectronic encoder, and the temperature parameter is collected in real time by a thermocouple sensor. The frequency boundary offset of the low-frequency mechanical vibration band is determined according to the difference between the rotational speed and the preset rotational speed threshold. For example, when the current rotational speed of the device exceeds 10% of the preset rotational speed threshold, the upper frequency limit of the low-frequency mechanical vibration band expands from 500 Hz to 800 Hz to cover high-frequency vibration harmonics. The frequency boundary offset of the medium-frequency heat conduction band is determined according to the difference between the temperature parameter and the preset temperature threshold. For example, when the device temperature exceeds the preset temperature threshold by 5 °C, the lower frequency limit of the medium-frequency heat conduction band decreases from 30 Hz to 20 Hz to adapt to the change in the heat diffusion rate. When dynamically expanding or compressing the initial frequency band division range, the adjustment amplitude of the frequency boundaries of the low-frequency mechanical vibration band and the medium-frequency heat conduction band has a linear relationship with the rotational speed or temperature deviation. For example, for every 1% that the rotational speed exceeds the preset threshold, the upper frequency limit of the low-frequency mechanical vibration band expands by 10 Hz.

[0083] The ambient noise intensity is measured in real time by a sound level meter, with the unit of decibel. The lower frequency limit of the high-frequency acoustic noise band is dynamically adjusted according to the ambient noise intensity. For example, when the ambient noise intensity is less than 60 decibels, the lower frequency limit of the high-frequency acoustic noise band remains at 2 kHz. When the ambient noise intensity is between 60 decibels and 80 decibels, the lower frequency limit of the high-frequency acoustic noise band is increased to 5 kHz to filter out low-frequency ambient noise. When the ambient noise intensity is greater than 80 decibels, the lower frequency limit of the high-frequency acoustic noise band is increased to 8 kHz. The adjusted lower frequency limit of the high-frequency acoustic noise band has a piecewise linear relationship with the noise intensity. For example, for every 10 decibels increase in the noise intensity, the lower frequency limit increases by 1 kHz.

[0084] When generating a frequency band division rule that matches the operating conditions and includes the updated low-frequency, medium-frequency, and high-frequency band ranges, the frequency range of the low-frequency mechanical vibration band is updated to 0 Hz to 800 Hz, the frequency range of the medium-frequency heat conduction band is updated to 20 Hz to 100 Hz, and the frequency range of the high-frequency acoustic noise band is updated to 5 kHz to 10 kHz. The updated frequency band division rule is stored in a configuration file or database for subsequent steps to call.

[0085] S3. Construct a multi-dimensional coupling strength matrix with the frequency band energy distribution as a constraint.

[0086] Among them, the construction logic of the multi-dimensional coupling strength matrix is as follows: Based on the frequency band division rule, extract the energy accumulation rate of the signals within each frequency band, and calculate the non-linear coupling strength between each frequency band through non-linear correlation analysis of multi-band signals and noise transfer parameters to construct a multi-dimensional coupling strength matrix with the frequency band energy distribution as a constraint, including:

[0087] Based on the frequency band division rule, extract the energy accumulation rate of the signals within the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band, and calculate the energy change gradient of each frequency band per unit time;

[0088] Through the phase synchronization analysis between the low-frequency mechanical vibration band and the high-frequency acoustic noise band, combined with the sensor sensitivity and the linear response range in the noise transfer parameters, calculate the nonlinear coupling strength between frequency bands;

[0089] Taking the frequency band energy distribution as a constraint condition, map the energy change gradient and the nonlinear coupling strength of each frequency band to the matrix row and column elements, and generate a multi-dimensional coupling strength matrix containing the coupling relationships of low-frequency, intermediate-frequency, and high-frequency.

[0090] The frequency band division rule is the frequency band division rule matching the working conditions generated in step S2, which includes the low-frequency mechanical vibration band, the intermediate-frequency heat conduction band, and the high-frequency acoustic noise band. The energy accumulation rate of the signal in the low-frequency mechanical vibration band is the change rate of the total signal energy per unit time, which is obtained by integrating the energy value of the vibration signal in the low-frequency mechanical vibration band and calculating the energy difference between adjacent time windows. For example, when the total energy in the low-frequency mechanical vibration band increases from 1000 units to 1500 units per second, the energy accumulation rate is 500 units / second. The energy accumulation rate of the signal in the intermediate-frequency heat conduction band is the change rate of the average energy of the thermal imaging signal per unit time, which is obtained by calculating the difference between the average thermal imaging energies of adjacent time windows. For example, when the average energy in the intermediate-frequency heat conduction band decreases from 200 units to 150 units per second, the energy accumulation rate is -50 units / second. The energy accumulation rate of the signal in the high-frequency acoustic noise band is the change rate of the energy peak value of the acoustic signal per unit time, which is obtained by extracting the maximum energy value of the acoustic signal and calculating the peak difference between adjacent time windows. For example, when the energy peak in the high-frequency acoustic noise band fluctuates from 5000 units to 5500 units per second, the energy accumulation rate is 500 units / second.

[0091] The phase synchronization analysis between the low-frequency mechanical vibration band and the high-frequency acoustic noise band is realized by calculating the consistency of the phase differences of the signals in the two frequency bands. The phase information is extracted through the Hilbert transform, and the consistency of the phase differences is measured by statistically analyzing the variance of the phase differences. For example, when the variance of the phase differences between the low-frequency mechanical vibration band and the high-frequency acoustic noise band is less than a preset threshold, it is determined that there is phase synchronization between the two frequency bands. The sensor sensitivity in the noise transfer parameters is the proportional coefficient between the sensor output signal and the input physical quantity defined in step S1, and the linear response range is the effective working frequency band of the sensor defined in step S1. When calculating the nonlinear coupling strength between frequency bands by combining the sensitivity and the linear response range, the sensitivity weight of the high-frequency acoustic noise band is the normalized value of the sensitivity within the linear response range. For example, when the sensitivity of the high-frequency acoustic noise band is 2 mV / Pa and the linear response range is from 5 kHz to 10 kHz, its sensitivity weight is 0.8. When the sensitivity weight of the low-frequency mechanical vibration band is 0.5, the nonlinear coupling strength between the two frequency bands is the product of the phase synchronization coefficient and the sensitivity weight.

[0092] When the frequency band energy distribution is used as a constraint condition, the total energy proportion of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band needs to meet the energy proportion threshold extracted in step S2. For example, the energy proportion of the low-frequency mechanical vibration band needs to be greater than 30%, the energy proportion of the medium-frequency heat conduction band needs to be between 40% and 50%, and the energy proportion of the high-frequency acoustic noise band needs to be less than 25%. When mapping the energy change gradient and the nonlinear coupling strength of each frequency band to the matrix row and column elements, the rows of the matrix correspond to the low-frequency mechanical vibration band and the medium-frequency heat conduction band, and the columns correspond to the high-frequency acoustic noise band. The value of the matrix element is the weighted sum of the energy change gradient of the corresponding row frequency band and the nonlinear coupling strength of the corresponding column frequency band. For example, when the energy change gradient of the low-frequency mechanical vibration band is 500 units / second and its nonlinear coupling strength with the high-frequency acoustic noise band is 0.4, the value of the corresponding matrix element is 500×0.4 = 200. In the generated multi-dimensional coupling strength matrix, when the coupling relationship between the low-frequency mechanical vibration band and the high-frequency acoustic noise band has a significantly higher corresponding element value than other frequency band combinations, it is marked as a strong noise coupling region.

[0093] S4. According to the coupling strength priority of each frequency band in the multi-dimensional coupling strength matrix, strip the non-linear coupling noise components in the multi-source data to obtain the energy characteristics of the target frequency band, including:

[0094] Generate a coupling strength priority sequence in descending order according to the coupling strength values of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band in the multi-dimensional coupling strength matrix;

[0095] Based on the coupling strength priority sequence, sequentially perform signal separation on the overlapping frequency bands between the high-frequency acoustic noise band and the low-frequency mechanical vibration band in the multi-source data to filter out the non-linear coupling noise components;

[0096] Verify the energy proportion threshold of the remaining frequency bands through the frequency band energy distribution constraint condition, and remove the residual noise components that do not meet the threshold constraint;

[0097] Merge the energy characteristics of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band after noise stripping to generate the energy characteristics of the target frequency band.

[0098] The multi-dimensional coupling strength matrix is the matrix generated in step S3 that includes the coupling relationships of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band. The coupling strength value of the low-frequency mechanical vibration band is the statistical mean of the row elements corresponding to the low-frequency mechanical vibration band in the matrix. The coupling strength value of the medium-frequency heat conduction band is the statistical median of the row elements corresponding to the medium-frequency heat conduction band. The coupling strength value of the high-frequency acoustic noise band is the maximum value of the column elements corresponding to the high-frequency acoustic noise band. When generating the coupling strength priority sequence in descending order, if the coupling strength value of the high-frequency acoustic noise band is 0.9, the low-frequency mechanical vibration band is 0.8, and the medium-frequency heat conduction band is 0.5, then the priority sequence is the high-frequency acoustic noise band, the low-frequency mechanical vibration band, and the medium-frequency heat conduction band.

[0099] When processing the overlapping frequency bands in the multi-source data based on the coupling strength priority sequence, the overlapping frequency band between the high-frequency acoustic noise band and the low-frequency mechanical vibration band is the intersection region of their frequency ranges. For example, if the lower limit frequency of the high-frequency acoustic noise band after adjustment is 2 kHz and the upper limit frequency of the low-frequency mechanical vibration band after adjustment is 500 Hz, there is no overlapping frequency band at this time, and the separation process is skipped. If the lower limit frequency of the high-frequency acoustic noise band is adjusted to 500 Hz and the upper limit frequency of the low-frequency mechanical vibration band is extended to 800 Hz, then the overlapping frequency band is from 500 Hz to 800 Hz. When separating the signals in this overlapping frequency band, the signal stripping is achieved by retaining the characteristics of the high-frequency acoustic noise band and suppressing the coupling noise components of the low-frequency mechanical vibration band, specifically by dynamically adjusting the passband range of the filter.

[0100] When verifying the energy ratio threshold of the remaining frequency bands through the frequency band energy distribution constraint conditions, the energy ratio of the remaining low-frequency mechanical vibration band needs to meet the low-frequency energy ratio threshold extracted in step S2, for example, not less than 30%. If the remaining energy ratio is 25%, then the secondary noise filtering is performed by suppressing the energy values of the abnormal frequency points. The energy ratio of the remaining medium-frequency heat conduction band needs to be within the medium-frequency energy ratio interval defined in step S2, for example, 40% to 50%. The energy ratio of the remaining high-frequency acoustic noise band needs to meet the high-frequency energy ratio threshold defined in step S2, for example, not higher than 25%. After verification, the energy characteristics of the remaining low-frequency mechanical vibration band are the effective energy accumulation rate of the vibration signal, the energy characteristics of the medium-frequency heat conduction band are the effective energy mean of the thermal imaging signal, and the energy characteristics of the high-frequency acoustic noise band are the effective energy peak of the acoustic signal.

[0101] When merging the energy characteristics of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band after noise stripping, the effective energy accumulation rate of the low-frequency mechanical vibration band is the energy change rate of the vibration signal after denoising in the frequency band from 0 Hz to 800 Hz, the effective energy mean value of the medium-frequency heat conduction band is the energy mean value of the thermal imaging signal after denoising in the frequency band from 20 Hz to 100 Hz, and the effective energy peak value of the high-frequency acoustic noise band is the energy peak value of the acoustic signal after denoising in the frequency band from 5 kHz to 10 kHz. The energy characteristics of the merged target frequency band are stored indexed by time stamps, including the effective energy values of each frequency band and the corresponding frequency band ranges.

[0102] S5. Generate a dynamic weight coefficient for feature fusion related to the load based on the time-series synchronization relationship between the device load rate and the energy characteristics of the target frequency band, including:

[0103] Calculate the energy-load correlation coefficients of the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band according to the change trend of the device load rate and the time-series fluctuation direction of the energy characteristics of the target frequency band;

[0104] Based on the energy-load correlation coefficients, determine the dynamic weight adjustment strategy for the low-frequency mechanical vibration band and the high-frequency acoustic noise band, and generate a weight increment coefficient that is positively or negatively correlated with the load;

[0105] Modify the weight increment coefficient according to the real-time change rate of the device load rate, and combine with the energy stability threshold of the medium-frequency heat conduction band to generate the dynamic weight coefficients of the low frequency, medium frequency, and high frequency;

[0106] Normalize the dynamic weight coefficients so that the sum of the weights of the low frequency, medium frequency, and high frequency is a fixed value, and generate a dynamic weight coefficient for feature fusion related to the load.

[0107] The device load rate is the ratio of the current output power of the device to the rated power, which is collected in real time by a power sensor and converted into a percentage value. The energy characteristics of the target frequency band are the energy accumulation rate of the low-frequency mechanical vibration band, the energy mean value of the medium-frequency heat conduction band, and the energy peak value of the high-frequency acoustic noise band generated after denoising in step S4. The change trend of the device load rate is the change direction of the load rate over time. For example, when the load rate continuously rises from 60% to 80%, it is a positive change trend, and when it drops from 80% to 70%, it is a negative change trend. The time-series fluctuation direction of the energy characteristics of the target frequency band is the change direction of the energy value over time. For example, when the energy accumulation rate of the low-frequency mechanical vibration band increases from 500 units / second to 800 units / second, it is a positive fluctuation, and when the energy peak value of the high-frequency acoustic noise band drops from 5500 units to 5000 units, it is a negative fluctuation.

[0108] When calculating the energy-load correlation coefficients for the low-frequency mechanical vibration band, the medium-frequency heat conduction band, and the high-frequency acoustic noise band, it is achieved by statistically analyzing the consistency between the changing trend of the load rate and the direction of energy fluctuation. For example, when the load rate changes positively and the energy accumulation rate in the low-frequency mechanical vibration band fluctuates positively synchronously, the energy-load correlation coefficient of the low-frequency mechanical vibration band is +1. When the load rate changes positively but the energy peak value in the high-frequency acoustic noise band fluctuates negatively, the energy-load correlation coefficient of the high-frequency acoustic noise band is -1. The energy-load correlation coefficient of the medium-frequency heat conduction band is calculated through the covariance between the changing trend of the load rate and the direction of the mean energy fluctuation. For example, when the mean energy increases synchronously as the load rate rises, the covariance is positive.

[0109] When determining the weight dynamic adjustment strategy based on the energy-load correlation coefficients, the weight increment coefficients for the low-frequency mechanical vibration band and the high-frequency acoustic noise band are set according to the sign of the correlation coefficient. For example, when the energy-load correlation coefficient of the low-frequency mechanical vibration band is +1, its weight increment coefficient is +0.1; when the correlation coefficient is -1, the weight increment coefficient is -0.1. The adjustment direction of the weight increment coefficient for the high-frequency acoustic noise band is opposite to that of the low-frequency mechanical vibration band. For example, when the correlation coefficient is +1, the weight increment coefficient is -0.05. The weight increment coefficient for the medium-frequency heat conduction band is fixed at 0, unless its mean energy exceeds the energy stability threshold defined in step S2. For example, when the mean energy of the medium-frequency heat conduction band exceeds the 50% threshold, the weight increment coefficient is adjusted to -0.05 to suppress overload interference.

[0110] When correcting the weight increment coefficients according to the real-time change rate of the device load rate, the change rate of the load rate is the change amplitude of the load rate per unit time. For example, if the load rate rises from 70% to 80% within 10 seconds, the change rate is 1% / second. When the change rate exceeds the preset rate threshold (such as 0.5% / second), the weight increment coefficient of the low-frequency mechanical vibration band is adjusted by doubling. For example, the original increment coefficient +0.1 is corrected to +0.2. The weight increment coefficient of the high-frequency acoustic noise band is scaled proportionally according to the rate threshold. For example, for every 0.1% / second exceeding the threshold, the increment coefficient increases by 0.01. The energy stability threshold for the medium-frequency heat conduction band is in the range of 40% to 50% defined in step S2. When the mean energy exceeds this range, the weight increment coefficient is further reduced by 0.02.

[0111] When normalizing the dynamic weight coefficients, the initial weight of the low-frequency mechanical vibration band is 0.4, the weight of the intermediate-frequency heat conduction band is 0.3, and the weight of the high-frequency acoustic noise band is 0.3. After adjustment according to the weight increment coefficient, for example, the low-frequency weight increases by 0.2 to 0.6, the high-frequency weight decreases by 0.05 to 0.25, and the intermediate-frequency weight remains 0.3. At this time, the total weight is 0.6 + 0.3 + 0.25 = 1.15. The normalization process scales each weight proportionally to a sum of 1. For example, the low-frequency weight is adjusted to 0.6 / 1.15 ≈ 0.52, the intermediate-frequency is adjusted to 0.3 / 1.15 ≈ 0.26, and the high-frequency is adjusted to 0.25 / 1.15 ≈ 0.22. In the finally generated feature fusion dynamic weight coefficients associated with the load, the weight of the low-frequency mechanical vibration band is 0.52, the weight of the intermediate-frequency heat conduction band is 0.26, and the weight of the high-frequency acoustic noise band is 0.22.

[0112] S6. Superimpose the frequency-domain features of the multi-source data across frequency bands according to the dynamic weight coefficients to generate a device status fusion feature vector, including:

[0113] Extract the frequency-domain features of the low-frequency mechanical vibration band, intermediate-frequency heat conduction band, and high-frequency acoustic noise band in the multi-source data, and convert them into time-frequency energy values corresponding to the respective frequency bands;

[0114] Based on the frequency band division rule, perform frequency band alignment processing on the time-frequency energy values of the low-frequency mechanical vibration band and the high-frequency acoustic noise band to eliminate energy conflicts in the overlapping regions between frequency bands;

[0115] Allocate the dynamic weight coefficients to the corresponding time-frequency energy values according to the frequency band type, and superimpose the cross-frequency band influence factors of the low-frequency mechanical vibration band on the high-frequency acoustic noise band to generate a weighted superimposed energy value;

[0116] According to the change rate of the device load rate, correct the dynamic correction coefficient of the weighted superimposed energy value, and fuse the stability compensation weight of the intermediate-frequency heat conduction band energy mean to generate a device status fusion feature vector.

[0117] The frequency-domain features of the multi-source data are the target frequency band energy features of the low-frequency mechanical vibration band, intermediate-frequency heat conduction band, and high-frequency acoustic noise band generated in step S4. The frequency-domain features of the low-frequency mechanical vibration band are converted into the time-frequency energy values of the vibration signal within the frequency band of 0 Hz to 500 Hz through Fourier transform, and the time-frequency energy value is the square root of the energy accumulation rate of the vibration signal per unit time. The frequency-domain features of the intermediate-frequency heat conduction band are converted into the time-frequency energy values of the thermal imaging signal within the frequency band of 20 Hz to 100 Hz through short-time Fourier transform, and the time-frequency energy value is the logarithmic transformation value of the energy mean of the thermal imaging signal. The frequency-domain features of the high-frequency acoustic noise band are converted into the time-frequency energy values of the acoustic signal within the frequency band of 5 kHz to 10 kHz through wavelet transform, and the time-frequency energy value is the normalized value of the energy peak of the acoustic signal.

[0118] When performing frequency band alignment processing on the time-frequency energy values of the low-frequency mechanical vibration band and the high-frequency acoustic noise band based on the frequency band division rule generated in step S2, if the upper frequency limit of the low-frequency mechanical vibration band is 500 Hz and the lower frequency limit of the high-frequency acoustic noise band is 5 kHz, there is no overlapping area and the original time-frequency energy values are directly retained. If the low-frequency mechanical vibration band extends to 800 Hz and the lower limit of the high-frequency acoustic noise band is adjusted to 2 kHz, the overlapping area is from 500 Hz to 800 Hz. When performing alignment processing on the time-frequency energy values in this overlapping area, the lower limit frequency of the high-frequency acoustic noise band is reset to 800 Hz, and the energy values within 500 Hz to 800 Hz are reallocated according to the dynamic weight coefficient of the high-frequency acoustic noise band to eliminate energy conflicts.

[0119] When distributing the dynamic weight coefficients associated with the load generated in step S5 to the corresponding time-frequency energy values according to the frequency band type, the weight coefficient of the low-frequency mechanical vibration band is 0.52, the intermediate-frequency heat conduction band is 0.26, and the high-frequency acoustic noise band is 0.22. When superimposing the cross-frequency band influence factor of the low-frequency mechanical vibration band on the high-frequency acoustic noise band, the influence factor is 10% of the product of the time-frequency energy value of the low-frequency mechanical vibration band and the time-frequency energy value of the high-frequency acoustic noise band. For example, if the low-frequency time-frequency energy value is 1000 units and the high-frequency is 500 units, the cross-frequency band influence factor is 1000×500×10% = 50000 units. The weighted superimposed energy value is the low-frequency time-frequency energy value × 0.52 + the high-frequency time-frequency energy value × 0.22 + the cross-frequency band influence factor.

[0120] When correcting the dynamic correction coefficient of the weighted superimposed energy value according to the change rate of the equipment load rate, if the change rate of the load rate is 1% / second and the preset rate threshold is 0.5% / second, the dynamic correction coefficient is 1+(1% - 0.5%) / 0.5% = 2. The corrected weighted superimposed energy value is multiplied by the dynamic correction coefficient. For example, if the original value is 10000 units, it is corrected to 20000 units. When fusing the stability compensation weight of the intermediate-frequency heat conduction band energy mean, if the intermediate-frequency energy mean is within the range of 40% to 50% defined in step S2, the compensation weight is 0.1; if it exceeds the range, the compensation weight is -0.1.

[0121] The final device state fusion feature vector is the corrected weighted superimposed energy value × (1 + compensation weight). For example, if the corrected value is 20000 units and the compensation weight is 0.1, the feature vector is 20000×1.1 = 22000 units.

[0122] S7. Based on the key frequency band energy distribution in the device state fusion feature vector that matches the fault feature library, eliminate the pseudo-frequency band features through multi-level verification, and output the fault diagnosis result, including:

[0123] Based on the energy distribution of the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band in the device status fusion feature vector, match the energy amplitude threshold and timing continuity condition of the key frequency bands of the preset fault modes in the fault feature library;

[0124] Verify and eliminate the pseudo-frequency band features that do not reach the energy accumulation rate threshold of the low-frequency mechanical vibration band or the energy peak threshold of the high-frequency acoustic noise band through energy amplitude verification;

[0125] Based on the inter-band correlation of the energy characteristics of the target frequency band, verify the timing synchronization of the energy fluctuation direction of the remaining frequency band features with the fault modes in the fault feature library;

[0126] According to the verification results, correct the dynamic weight coefficient, suppress the weights of the frequency bands that do not match the fault feature library, and output the diagnostic results including the fault type and confidence level.

[0127] The device status fusion feature vector is the feature vector generated in step S6, which includes the weighted superposition energy values of the low-frequency mechanical vibration band, medium-frequency heat conduction band, and high-frequency acoustic noise band. The energy amplitude threshold of the key frequency bands of the preset fault modes in the fault feature library is obtained through the statistics of historical fault data. For example, the energy accumulation rate threshold of the low-frequency mechanical vibration band corresponding to the bearing wear fault is 500 units / second, and the energy peak threshold of the high-frequency acoustic noise band corresponding to the cavitation fault is 6000 units. The timing continuity condition is the continuous consistency of the energy fluctuation directions of each frequency band under the fault mode. For example, the bearing wear fault requires that the energy accumulation rate of the low-frequency mechanical vibration band continuously fluctuates positively for more than 10 seconds.

[0128] When verifying and eliminating pseudo-frequency band features through energy amplitude verification, if the energy accumulation rate of the low-frequency mechanical vibration band is 400 units / second and lower than the threshold of 500 units / second, it is determined as a pseudo-feature and eliminated. If the energy peak of the high-frequency acoustic noise band is 5500 units and lower than the threshold of 6000 units, it is determined as a pseudo-feature and eliminated. The energy mean of the medium-frequency heat conduction band needs to be in the range of 40% to 50% defined in step S2, otherwise it is regarded as abnormal and eliminated. After eliminating the pseudo-features, the remaining energy accumulation rate of the low-frequency mechanical vibration band is 600 units / second, the energy peak of the high-frequency acoustic noise band is 6500 units, and the energy mean of the medium-frequency heat conduction band is 45%.

[0129] When verifying the timing synchronization based on the energy characteristics of the target frequency band and the inter - band correlation, the phase synchronization between the low - frequency mechanical vibration band and the high - frequency acoustic noise band is measured by the phase synchronization coefficient generated in step S3. For example, for a bearing wear fault, the phase synchronization coefficient between the low - frequency and high - frequency should be greater than 0.8 and last for more than 5 seconds. If the current phase synchronization coefficient is 0.7 and the duration is 3 seconds, it is determined that the timing synchronization condition is not met. For a cavitation fault, the fluctuation directions of the energy peak of the high - frequency acoustic noise band and the energy mean of the medium - frequency heat conduction band should be opposite. For example, when the high - frequency peak rises, the medium - frequency mean should synchronously decrease; otherwise, it is regarded as out - of - sync timing.

[0130] When correcting the dynamic weight coefficient according to the verification result, if the energy characteristics of the low - frequency mechanical vibration band match the bearing wear mode in the fault feature library, its dynamic weight coefficient is increased from 0.52 to 0.6 to enhance the diagnostic weight. If the energy characteristics of the high - frequency acoustic noise band do not match the cavitation fault mode, its weight coefficient is decreased from 0.22 to 0.15 to suppress interference. When the energy mean of the medium - frequency heat conduction band is in the stable range, the weight coefficient remains 0.25; when it exceeds the range, it is adjusted down to 0.1. The corrected dynamic weight coefficients are normalized to a total of 1. For example, the low - frequency is 0.6, the medium - frequency is 0.25, and the high - frequency is 0.15. The final output of the fault diagnosis result includes the fault type (such as bearing wear) and the confidence level (such as 85%), and the confidence level is calculated by weighting according to the number of matching frequency - band characteristics and the weight coefficients.

[0131] Embodiment 2: Figure 2 The structure diagram of the intelligent analysis system for multi - scenario industrial equipment and facility inspection data of the present invention is given. The intelligent analysis system for multi - scenario industrial equipment and facility inspection data includes:

[0132] Parameter sensing frequency - band module: Obtain the operating parameters of industrial equipment and facilities, and determine the initial frequency - band division range of multi - source data in combination with the physical response characteristics of multi - type sensors;

[0133] Operating condition frequency - band rule module: Adjust the initial frequency - band division range to generate a frequency - band division rule matching the operating condition;

[0134] Multi - dimensional coupling strength module: Construct a multi - dimensional coupling strength matrix with the frequency - band energy distribution as a constraint;

[0135] Noise stripping feature module: Strip the non - linear coupling noise components in the multi - source data according to the coupling strength priority of each frequency band in the multi - dimensional coupling strength matrix to obtain the target frequency - band energy characteristics;

[0136] Load dynamic weight module: Generate a feature - fusion dynamic weight coefficient associated with the load based on the timing synchronization relationship between the equipment load rate and the target frequency - band energy characteristics;

[0137] Cross-frequency energy fusion module: Superimpose the frequency-domain features of multi-source data across frequency bands according to dynamic weight coefficients to generate a device status fusion feature vector;

[0138] Multi-level diagnosis output module: Based on the key frequency band energy distribution in the device status fusion feature vector that matches the fault feature library, exclude pseudo-frequency band features through multi-level verification, and output the fault diagnosis result.

[0139] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0140] It should be noted that the present invention can be deployed on the device itself to achieve embedded applications, or can also run on a PC or other terminal with a user interface, so as to meet various hardware environments and usage requirements.

[0141] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access, or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0142] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices, and modules described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0143] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of devices or modules can be in electrical, mechanical, or other forms.

[0144] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0145] In addition, in each embodiment of the present application, the functional modules can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0146] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0147] As described above, this is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0148] Finally, the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-scenario industrial equipment and facility inspection data intelligent analysis method, characterized in that: The steps include: S1. Obtain the operating parameters of industrial equipment and facilities, combine the physical response characteristics of multiple types of sensors, and determine the initial frequency band division range of multi-source data, including: Obtain the speed, temperature and pressure of industrial equipment and facilities, and extract the resonant frequency, sensitivity and linear response range of multiple types of sensors as physical response characteristics; The frequency domain response range is divided according to the resonant frequency of the sensor, and the effective frequency band boundary of each sensor is determined by combining the sensitivity and linear response range; Based on the rotation speed and temperature in the operating parameters, the initial frequency band division range is dynamically adjusted within the effective frequency band boundary to generate an initial frequency band division result including a low-frequency mechanical vibration band, a medium-frequency heat conduction band and a high-frequency acoustic noise band; S2. Adjust the initial frequency band division range and generate a frequency band division rule that matches the working conditions; S3, construct a multidimensional coupling strength matrix with frequency band energy distribution as constraint; S4, stripping nonlinear coupling noise components in multi-source data according to the coupling intensity priority of each frequency band in the multi-dimensional coupling intensity matrix to obtain energy characteristics of the target frequency band; S5. Based on the timing synchronization relationship between the equipment load rate and the target frequency band energy characteristics, a load-related feature fusion dynamic weight coefficient is generated, including: According to the change trend of equipment load rate and the time series fluctuation direction of energy characteristics of target frequency band, the energy-load correlation coefficient of low-frequency mechanical vibration band, medium-frequency heat conduction band and high-frequency acoustic noise band is calculated; Determine the dynamic adjustment strategy of the weights of the low-frequency mechanical vibration band and the high-frequency acoustic noise band based on the energy-load correlation coefficient, and generate a weight increment coefficient that is positively or negatively correlated with the load; The weight increment coefficient is corrected according to the real-time change rate of the equipment load rate, and the dynamic weight coefficients of low frequency, medium frequency and high frequency are generated in combination with the energy stability threshold of the medium frequency heat conduction belt; The dynamic weight coefficient is normalized so that the sum of the weights of low frequency, medium frequency and high frequency is a fixed value, and the load-related feature fusion dynamic weight coefficient is generated; S6, performing cross-band energy superposition on the frequency domain features of the multi-source data according to the dynamic weight coefficient to generate a device state fusion feature vector; S7. Based on the key frequency band energy distribution in the equipment state fusion feature vector that matches the fault feature library, the pseudo-frequency band features are eliminated through multi-level verification, and the fault diagnosis results are output.

2. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 1 is characterized in that: The construction logic of the multidimensional coupling strength matrix is ​​as follows: based on the frequency band division rules, the energy accumulation rate of the signal in each frequency band is extracted, and the nonlinear coupling strength between the frequency bands is calculated through the nonlinear correlation analysis between the multi-band signals and the noise transfer parameters to construct a multidimensional coupling strength matrix with the frequency band energy distribution as the constraint.

3. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 1 is characterized in that: Adjust the initial frequency band division range and generate frequency band division rules that match the working conditions, including: According to the frequency band energy distribution law under the same load in the equipment load rate and historical operation data, the energy proportion thresholds of the low-frequency mechanical vibration band, the medium-frequency heat conduction band and the high-frequency acoustic noise band are extracted; Based on the current speed and temperature parameters of the equipment, the frequency boundary offset between the low-frequency mechanical vibration band and the medium-frequency heat conduction band is matched, and the initial frequency band division range is dynamically expanded or compressed; The lower frequency limit of the high-frequency acoustic noise band is adjusted according to the ambient noise intensity, and a frequency band division rule matching the working conditions including the updated low-frequency, medium-frequency and high-frequency band ranges is generated.

4. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 2 is characterized in that: Based on the frequency band division rule, the energy accumulation rate of the signal in each frequency band is extracted. The nonlinear coupling strength between the frequency bands is calculated through the nonlinear correlation analysis between the multi-band signals and the noise transfer parameter to construct a multidimensional coupling strength matrix constrained by the frequency band energy distribution, including: Based on the frequency band division rule, the energy accumulation rate of the signals in the low-frequency mechanical vibration band, the medium-frequency heat conduction band and the high-frequency acoustic noise band is extracted, and the energy change gradient of each frequency band per unit time is calculated; By analyzing the phase synchronization between the low-frequency mechanical vibration band and the high-frequency acoustic noise band, combined with the sensor sensitivity and linear response range in the noise transfer parameters, the nonlinear coupling strength between the frequency bands is calculated; Taking the frequency band energy distribution as a constraint, the energy change gradient and nonlinear coupling strength of each frequency band are mapped into matrix row and column elements to generate a multidimensional coupling strength matrix containing low-frequency, medium-frequency and high-frequency coupling relationships.

5. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 1 is characterized in that: According to the coupling intensity priority of each frequency band in the multi-dimensional coupling intensity matrix, the nonlinear coupling noise components in the multi-source data are stripped off to obtain the energy characteristics of the target frequency band, including: According to the coupling strength values ​​of the low-frequency mechanical vibration band, the medium-frequency heat conduction band and the high-frequency acoustic noise band in the multi-dimensional coupling strength matrix, a coupling strength priority sequence is generated in descending order; Based on the coupling strength priority sequence, the overlapping frequency bands of high-frequency acoustic noise bands and low-frequency mechanical vibration bands in multi-source data are sequentially separated to filter out non-linear coupling noise components. The energy proportion threshold of the remaining frequency bands is verified through the frequency band energy distribution constraint condition, and the residual noise components that do not meet the threshold constraint are removed; The energy characteristics of the noise-stripped low-frequency mechanical vibration band, mid-frequency thermal conduction band, and high-frequency acoustic noise band are combined to generate the energy characteristics of the target frequency band.

6. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 1 is characterized in that: The frequency domain features of multi-source data are superimposed across frequency bands according to dynamic weight coefficients to generate a device state fusion feature vector, including: Extract the frequency domain features of low-frequency mechanical vibration band, medium-frequency heat conduction band and high-frequency acoustic noise band in multi-source data, and convert them into time-frequency energy values ​​of corresponding frequency bands respectively; Based on the frequency band division rules, the time-frequency energy values ​​of the low-frequency mechanical vibration band and the high-frequency acoustic noise band are aligned to eliminate the energy conflict in the overlapping areas between the frequency bands. The dynamic weight coefficient is assigned to the corresponding time-frequency energy value according to the frequency band type, and the cross-band influence factor of the low-frequency mechanical vibration band on the high-frequency acoustic noise band is superimposed to generate a weighted superposition energy value; The dynamic correction coefficient of the weighted superposition energy value is corrected according to the rate of change of the equipment load rate, and the stability compensation weight of the mean energy of the medium frequency heat conduction band is integrated to generate the equipment state fusion feature vector.

7. The multi-scenario industrial equipment and facility inspection data intelligent analysis method according to claim 1 is characterized in that: Based on the key frequency band energy distribution in the equipment status fusion feature vector that matches the fault feature library, the pseudo-frequency band features are eliminated through multi-level verification, and the fault diagnosis results are output, including: Based on the energy distribution of low-frequency mechanical vibration band, medium-frequency heat conduction band and high-frequency acoustic noise band in the fusion feature vector of equipment status, the key frequency band energy amplitude threshold and time series continuity condition of the preset fault mode in the fault feature library are matched; Through energy amplitude verification, pseudo-frequency band features that do not reach the energy accumulation rate threshold of the low-frequency mechanical vibration band or the energy peak threshold of the high-frequency acoustic noise band are eliminated; Based on the inter-band correlation of the target frequency band energy feature, verify the time synchronization between the energy fluctuation direction of the remaining frequency band features and the fault mode in the fault feature library; The dynamic weight coefficient is corrected according to the verification results, the frequency band weights that do not match the fault feature library are suppressed, and the diagnosis results including the fault type and confidence level are output.

8. A multi-scenario industrial equipment and facility inspection data intelligent analysis system, used to implement the multi-scenario industrial equipment and facility inspection data intelligent analysis method according to any one of claims 1 to 7, characterized in that: include: Parameter sensing frequency band module: obtains the operating parameters of industrial equipment and facilities, combines the physical response characteristics of multiple types of sensors, and determines the initial frequency band division range of multi-source data; Working condition frequency band rule module: adjusts the initial frequency band division range and generates frequency band division rules that match the working conditions; Multidimensional coupling strength module: constructs a multidimensional coupling strength matrix constrained by frequency band energy distribution; Noise stripping feature module: strips the nonlinear coupling noise components in the multi-source data according to the coupling intensity priority of each frequency band in the multi-dimensional coupling intensity matrix to obtain the energy characteristics of the target frequency band; Load dynamic weight module: Generates load-related feature fusion dynamic weight coefficients based on the timing synchronization relationship between the equipment load rate and the target frequency band energy characteristics; Cross-frequency energy fusion module: The frequency domain features of multi-source data are superimposed across frequency bands according to dynamic weight coefficients to generate a device state fusion feature vector; Multi-level diagnosis output module: Based on the key frequency band energy distribution in the equipment status fusion feature vector that matches the fault feature library, it eliminates pseudo-frequency band features through multi-level verification and outputs the fault diagnosis results.

Images