Production equipment monitoring method, system and equipment based on industrial internet of things, and medium

Abstract

The invention relates to the field of Internet of Things production, in particular to a production equipment monitoring method, system and equipment based on the industrial Internet of Things and a medium, and the method comprises the following steps: obtaining coupling change information of a monitoring part obtained through combined measurement based on a displacement sensor and an eddy current sensor; the coupling change information comprises a change rate signal and an eddy current energy signal, and the monitoring part is a part needing to be monitored on the target production equipment; acquiring surface feature information of the monitored part obtained based on a phase-controlled eddy current sensor; fusing the coupling change information with the surface feature information to obtain a comprehensive feature signal; inputting the comprehensive characteristic signal into a preset wavelet neural network prediction model to output a risk level of the target production equipment; through dynamic fusion of feature enhancement driven by a physical mechanism and data driving, the dependence of traditional time domain analysis on a single signal amplitude is broken through, and the cross-physics coupling effect of weak damage is amplified from multiple dimensions.

Description

TECHNICAL FIELD

[0001] The present application relates to the production field of the Internet of Things, and in particular to a production equipment monitoring method, system, device and medium based on an industrial Internet of Things. BACKGROUND

[0002] Production equipment plays a crucial role in modern manufacturing, with its functions not only reflected in improving production efficiency, but also covering resource optimization, cost control, and quality assurance.

[0003] Currently, production equipment detection work mainly relies on periodic shutdown detection by manual work, and even in the increasingly mature automatic monitoring system based on the industrial Internet of Things, there are still the following problems: Sensor signal distortion due to electromagnetic interference, mechanical vibration or environmental temperature and humidity changes, making it difficult to accurately capture micron-level damage features; Multi-sensor data fusion mostly uses fixed weight or simple weighted average, which cannot dynamically adapt to the performance differences of sensors, resulting in an increased false detection rate; In addition, the digital phase-locked amplifier has a fixed reference signal frequency, making it difficult to track dynamic signal frequency deviation, limiting the extraction sensitivity of weak damage signals.

[0004] Based on the above problems, there is an urgent need for a method, system, device and medium that can accurately monitor production equipment in real time under multi-sensor monitoring conditions. SUMMARY

[0005] The main purpose of the present application is to provide a production equipment monitoring method based on the industrial Internet of Things, aiming to solve the technical problem of increased false detection rate caused by multi-source sensor data fusion when monitoring production equipment in the prior art.

[0006] To achieve the above purpose, the present application provides a production equipment monitoring method based on the industrial Internet of Things, comprising the following steps: Obtaining coupling change information of a monitoring part obtained by combination measurement based on a displacement sensor and an eddy current sensor; wherein the coupling change information includes a change rate signal and an eddy current energy signal, and the monitoring part is a part of a target production equipment that needs to be monitored; Obtaining surface feature information of the monitoring part based on a phased eddy current sensor; Fusing the coupling change information and the surface feature information to obtain a comprehensive feature signal; Inputting the comprehensive feature signal into a preset wavelet neural network prediction model to output a risk level of the target production equipment.

[0007] Optionally, the obtaining the coupling change information of the monitoring part based on the combination of the displacement sensor and the eddy current sensor comprises the following steps of: The displacement sensor outputs the original displacement of the monitoring part, and the change rate signal is calculated by eliminating high-frequency noise through Kalman filtering, and the expression of the change rate signal is: ; Wherein, is the Kalman gain; is the measured value of the original displacement; is the filtered displacement value at the moment; is the moment; is the filtered displacement value at the moment; The eddy current sensor calculates the eddy current energy signal after the discrete wavelet transform of the collected eddy current signal, and the expression of the eddy current energy signal is: ; Wherein, is the eddy current characteristic signal after wavelet transform; is the number of data points in the sampling window; is the summation index; According to the change rate signal and the eddy current energy signal, the coupling change information of the monitoring part is obtained.

[0008] Optionally, the displacement sensor is also used to calculate the change rate signal by displacement difference calculation of instantaneous speed, and the expression of the displacement change rate at the moment is: ; Wherein, is the sampling time interval; is the displacement change rate at the moment.

[0009] Optionally, the obtaining the coupling change information of the monitoring part based on the change rate signal and the eddy current energy signal comprises the following steps of: Mapping the change rate signal and the eddy current energy signal to a state scoring model, and the state scoring model outputs the coupling change information; The state scoring model satisfies: ; Wherein, is the weight coefficient; is the dynamic coupling state score;​​ is a displacement change rate threshold value; is an eddy current energy threshold value.

[0010] Optionally, the obtaining of the surface feature information of the monitored part based on the phased eddy current sensor comprises the following steps: setting the excitation signal parameters, and collecting the eddy current response signal; selecting a discrete wavelet transform model to decompose and quantify the eddy current response signal, and extracting low-frequency approximate feature signals and high-frequency detail feature signals in layers; performing empirical mode decomposition on the noise-containing signal, performing threshold quantization on the decomposed high-frequency coefficients, then performing signal reconstruction, extracting feature values and outputting the noise-reduced signal; establishing a curve analysis model according to discrete data cubic spline interpolation and two-dimensional curve fitting, and determining the discrete signal feature extraction interval; performing discrete analysis on the quadratic optimization signal, including standard deviation analysis, discrete signal convolution and analysis; establishing a discrete signal feature analysis evaluation model, and using an evaluation algorithm to calculate and output the surface feature information from the discrete analysis numerical value and the feature signal numerical value.

[0011] Optionally, the fusing of the coupling change information and the surface feature information to obtain a comprehensive feature signal comprises the following steps: extracting real and imaginary parts of the eddy current response signal collected by the phased eddy current sensor through a digital lock-in amplifier, and calculating real and imaginary part quotient value features; weighting and fusing the real and imaginary part quotient value features, the displacement change rate signal of the displacement sensor and the eddy current energy signal of the eddy current sensor to generate a comprehensive feature signal; the fusion process of the comprehensive feature signal: ; wherein, is a real part signal component output by the digital lock-in amplifier; is an imaginary part signal component output by the digital lock-in amplifier; is a dynamic weight coefficient, and and are 1.

[0012] Optionally, the input layer of the wavelet neural network prediction model is the comprehensive feature signal, the hidden layer adopts an orthogonal wavelet basis function for multi-resolution analysis, extracts local features through multi-dimensional wavelet decomposition with a scaling factor of 2, and the output layer is a risk level.

[0013] The production equipment monitoring system based on the industrial Internet of Things comprises a management platform, a sensing network platform and an object platform which are sequentially connected in communication, the management platform comprises: The first information acquisition module is configured to acquire coupling change information of a monitoring part obtained based on combined measurement of the displacement sensor and the eddy current sensor, wherein the coupling change information comprises a change rate signal and an eddy current energy signal, and the monitoring part is a part of the target production equipment that needs to be monitored; The second information acquisition module is configured to acquire surface feature information of the monitoring part based on the phase-controlled eddy current sensor. The information fusion module is configured to fuse the coupling change information and the surface feature information to obtain a comprehensive feature signal. The risk prediction module is configured to input the comprehensive feature signal into a preset wavelet neural network prediction model to output a risk level of the target production equipment.

[0014] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program.

[0015] A computer readable storage medium stores a computer program, and a processor executes the computer program.

[0016] The beneficial effects that can be achieved by the present application are as follows: The production equipment monitoring method, system, device and medium based on the industrial Internet of Things provided in the embodiments of the present application break through the dependence of traditional time domain analysis on a single signal amplitude through feature enhancement driven by physical mechanisms and dynamic fusion driven by data, amplify the cross-physical field coupling effect of weak damage from multiple dimensions, specifically, the original displacement signal collected by the displacement sensor is significantly interfered by mechanical vibration and environmental noise, the Kalman filtering algorithm is used to dynamically estimate the real displacement value, when high-frequency noise is dominant, the Kalman gain tends to 0, the filter output depends on historical data, and instantaneous noise is suppressed; when the displacement suddenly changes, the Kalman gain is adaptively increased, and the real displacement change is quickly tracked, in actual production monitoring test, the above method improves the signal-to-noise ratio of the displacement signal by 3-5 times, retains the micron-level displacement fluctuation, solves the problem of high false detection rate caused by data fusion of multiple source sensors in the prior art, and realizes the monitoring of the production equipment. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or the prior art description will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.

[0018] Figure 1 Schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the framework structure of the production equipment monitoring system based on the Industrial Internet of Things of the present invention.

[0019] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] If there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of such features. In addition, the meaning of "and / or" appearing throughout the text includes three parallel schemes. Taking "A and / or B" as an example, it includes scheme A, or scheme B, or a scheme in which A and B are satisfied at the same time. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in this field to implement. When the combination of technical solutions is mutually contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0022] Example 1: As attached Figure 1 As shown, this embodiment provides a production equipment monitoring method based on the industrial Internet of Things, including the following steps: Acquiring coupling change information of a monitored portion obtained through combined measurement of a displacement sensor and an eddy current sensor; wherein the coupling change information includes a rate of change signal and an eddy current energy signal, and the monitored portion is a portion on the target production equipment to be monitored; Acquiring surface feature information of the monitored part based on a phased eddy current sensor; fusing the coupling change information with the surface feature information to obtain a comprehensive feature signal; The comprehensive characteristic signal is input into a preset wavelet neural network prediction model to output the risk level of the target production equipment.

[0023] It should be noted that in the existing production equipment monitoring method, the sensor signal is distorted due to electromagnetic interference, mechanical vibration or environmental temperature and humidity changes, making it difficult to accurately capture micron-level damage characteristics. In addition, for the fusion process of multi-sensor data, fixed weight or simple weighted average is often used, which cannot dynamically adapt to the performance difference of the sensor, resulting in an increase in false detection rate. Finally, the digital phase-locked amplifier has a fixed reference signal frequency, making it difficult to track dynamic signal frequency deviation, limiting the extraction sensitivity of weak damage signals.

[0024] Based on the above problems, the production equipment monitoring method based on industrial Internet of Things is proposed in the embodiment. Through feature enhancement driven by physical mechanism and dynamic fusion driven by data, the cross-physical field coupling effect of weak damage is amplified from multiple dimensions, breaking the dependence of traditional time domain analysis on single signal amplitude. Specifically, the original displacement signal collected by the displacement sensor is significantly affected by mechanical vibration and environmental noise. The Kalman filter algorithm is used to dynamically estimate the true displacement value. When high-frequency noise dominates, the Kalman gain tends to 0, and the filter output depends on historical data to suppress transient noise. When the displacement changes suddenly (such as equipment loosening), the Kalman gain is adaptively increased to quickly track the true displacement change. In actual production monitoring tests, the above method improves the signal-to-noise ratio of the displacement signal by 3-5 times, and retains the micron-level displacement fluctuation.

[0025] In addition, the filtered displacement signal is subjected to a difference operation to convert the static displacement into an instantaneous speed signal. The periodic vibration caused by weak damage may only have a micron-level amplitude in the displacement time domain, but it can be amplified to the order of millimeters per second through the speed signal. It can be understood that the slow displacement drift caused by equipment thermal expansion will be automatically eliminated in the difference operation. It can also be understood that the original eddy current signal collected by the eddy current sensor is decomposed into multiple frequency bands by DWT. The first and second layer detail coefficients are retained, and the fourth and fifth layer approximation coefficients are subjected to soft threshold processing to eliminate 50Hz power frequency interference and low-frequency electromagnetic noise. The eddy current energy signal is calculated in the selected frequency band, which has the physical meaning of electromagnetic energy dissipation caused by damage. The square operation of the eddy current energy signal further amplifies the contribution of small damage.

[0026] In some embodiments, the displacement mutation can be the loosening of the transmission components or poor contact of the bearing wear in the production equipment.

[0027] In some embodiments, the physical mechanism driven feature enhancement is the switching response between mechanical motion and electromagnetic response.

[0028] In some embodiments, the original eddy current signal is decomposed into 5 layers of db4 wavelet basis by DWT.

[0029] In some embodiments, the eddy current energy signal is characterized by a distortion of the eddy current distribution in the crack region, resulting in a 30-50% increase in high-frequency energy, when a fatigue crack of 5 μm depth appears on the metal surface. In the embodiment, the coupling change information of the monitoring part is obtained based on the combined measurement of the displacement sensor and the eddy current sensor, including the following steps: The original displacement of the monitoring part is output based on the displacement sensor, and the change rate signal is calculated by eliminating high-frequency noise through Kalman filtering, and the expression of the change rate signal is: ; Wherein, is the Kalman gain; is the measured value of the original displacement; is the filtered displacement value at the moment; is the moment; is the filtered displacement value at the moment; is the filtered displacement value at the moment; The eddy current energy signal is calculated based on the eddy current sensor after the eddy current signal collected is decomposed by discrete wavelet transform, and the expression of the eddy current energy signal is: ; Wherein, is the eddy current characteristic signal after wavelet transform; is the number of data points in the sampling window; is the summation index; The coupling change information of the monitoring part is obtained according to the change rate signal and the eddy current energy signal.

[0030] It should be noted that for the displacement sensor, the Kalman filtering algorithm is used to dynamically balance the weight of real-time measurement value and historical data, which not only suppresses the high-frequency noise introduced by mechanical vibration, but also retains the true trend of micron-level displacement change of the equipment. The filtered displacement signal is converted into instantaneous speed characteristics through difference operation, which converts the slow displacement fluctuation that is difficult to detect into a dynamic change rate with clear physical meaning, so that the early features of mechanical faults such as equipment loosening and shaft misalignment are amplified by 10-20 times.

[0031] ​Meanwhile, the eddy current sensor adopts discrete wavelet transform to carry out multi-scale decomposition on the original eddy current signal, capture the electromagnetic field distortion effect caused by the surface micro-cracks by extracting high-frequency detail components in layers, and quantize and reconstruct the low-frequency background noise by threshold, and finally aggregate the time-domain dispersed weak damage characteristics into the eddy current energy index with statistical significance through energy integration calculation. The above combined processing makes the micron-level cracks easily submerged in noise in the traditional time-domain analysis present a 3-5 times characteristic enhancement in the frequency domain energy dimension.

[0032] It also needs to be explained that the mechanical state representation of the displacement signal and the electromagnetic characteristic analysis of the eddy current signal form a complement: when the equipment vibrates intensively due to poor lubrication, the displacement change rate rises significantly while the eddy current energy remains stable, and the system can automatically exclude such interference; on the contrary, when the surface appears hidden cracks, the eddy current energy increases sharply while the displacement change rate does not exceed the threshold, and the early warning can still be triggered.

[0033] In the embodiment, the displacement sensor is also used to obtain the change rate signal by displacement difference calculation of the instantaneous speed, and the expression of the displacement change rate at the moment is: ; Among them, is the sampling time interval; is the displacement change rate at the moment.

[0034] The displacement sensor first carries out adaptive noise reduction processing on the original displacement data by Kalman filtering algorithm, dynamically balances the weight relationship between real-time measurement value and historical data, effectively suppresses the pollution of high-frequency electromagnetic interference and random mechanical vibration to the real trend of displacement, and improves the signal-to-noise ratio of the displacement signal.

[0035] On this basis, the time window difference algorithm is used to convert the filtered displacement sequence into an instantaneous speed signal, and this conversion process has a double enhancement effect: Firstly, the micron-level slowly changing displacement (converted into a speed mutation feature with a significant magnitude difference) that is originally submerged in noise is enhanced, and the detection sensitivity of early wear is improved; Secondly, through the inherent characteristics of difference operation, the low-frequency drift interference such as equipment thermal expansion is automatically eliminated, and the misjudgment caused by traditional absolute value monitoring is avoided. This displacement-speed cooperative analysis mechanism successfully captures the 0.8 μm level axial displacement fluctuation that cannot be identified by traditional methods in the main shaft monitoring case of the screw compressor, and early warning of bearing raceway spalling fault for 35 days.

[0036] In the embodiment, the coupling change information of the monitoring part is obtained according to the change rate signal and the eddy current energy signal, and the method comprises the following steps: mapping the rate of change signal and the eddy current energy signal to a state score model, the state score model outputting coupling change information; The state score model satisfies: ; wherein, is a weight coefficient; is a dynamic coupling state score; is a displacement rate of change threshold value; is an eddy current energy threshold value.

[0037] The mechanical vibration features extracted by the displacement sensor and the electromagnetic response features captured by the eddy current sensor are subjected to threshold normalization processing, eliminating the dimensional differences of the displacement dimension and the energy dimension; then the contribution weights of the two types of features are dynamically adjusted by a dynamic weight coefficient, In some embodiments, the weight of the displacement rate of change is increased to capture mechanical imbalance features under high-speed rotating conditions, and the weight of the eddy current energy is strengthened to sensitively identify surface cracks in the static load stage. By introducing a damage coupling enhancement criterion in the model, when both types of features exceed the threshold at the same time, a nonlinear score growth is triggered, so that the composite damage that is easily missed in traditional methods presents an exponential response in the score space.

[0038] In some embodiments, the surface feature information of the monitored part obtained based on the phased eddy current sensor includes the following steps: setting the excitation signal parameters to collect the eddy current response signal; selecting a discrete wavelet transform model to decompose and quantify the eddy current response signal, and extracting low-frequency approximate feature signals and high-frequency detail feature signals in layers; performing empirical mode decomposition on the noise-containing signal, threshold quantizing the high-frequency coefficients after decomposition, and then performing signal reconstruction to extract feature values and output the noise-reduced signal; establishing a curve analysis model according to discrete data cubic spline interpolation and two-dimensional curve fitting to determine the discrete signal feature extraction interval; performing discrete analysis on the quadratic optimization signal, including standard deviation analysis, discrete signal convolution and analysis; establishing a discrete signal feature analysis evaluation model, and using an evaluation algorithm to calculate and output surface feature information from the discrete analysis numerical value and the feature signal numerical value results.

[0039] By setting the excitation signal parameters and collecting the eddy current response signals, high-fidelity original data basis is provided for subsequent analysis. On this basis, the signal is decomposed by discrete wavelet transform model, and the low-frequency approximate feature signal and high-frequency detail feature signal are extracted layer by layer, which effectively distinguishes the background noise and damage features, especially for the suppression of high-frequency noise, while retaining the high-frequency detail information reflecting the surface micro-cracks. Further combined with empirical mode decomposition, the noisy signal is adaptively decomposed, and through threshold quantization of high-frequency coefficients and signal reconstruction, the influence of random noise on feature extraction is dynamically eliminated, which significantly improves the signal-to-noise ratio. Subsequently, through cubic spline interpolation and two-dimensional curve fitting technology, the discrete data is optimized to construct a curve analysis model to accurately determine the feature extraction interval, solving the problem of fuzzy feature positioning caused by data dispersion in traditional methods. Finally, through standard deviation analysis, discrete signal convolution and analysis, etc. Mathematical means for multi-dimensional quantification of the second optimization signal, and the establishment of a feature analysis evaluation model, the abstract discrete signal is converted into quantifiable surface feature information. This series of technical means are closely linked, which not only overcomes the defects of single noise reduction method that cannot balance the suppression of high-frequency noise and the elimination of low-frequency drift, but also realizes the accurate capture of the cross-physical field coupling effect of weak damage through dynamic optimization of the feature extraction process.

[0040] In the embodiment, the fusion of the coupling change information and the surface feature information to obtain a comprehensive feature signal comprises the following steps: The real and imaginary part components of the eddy current response signal collected by the phase-controlled eddy current sensor are extracted by a digital lock-in amplifier, and a real and imaginary part quotient characteristic is calculated; The real and imaginary part quotient characteristic is weighted and fused with the displacement change rate signal of the displacement sensor and the eddy current energy signal of the eddy current sensor to generate a comprehensive feature signal; The fusion process of the comprehensive feature signal: ; Wherein, is the real part signal component output by the digital lock-in amplifier; is the imaginary part signal component output by the digital lock-in amplifier; is a dynamic weight coefficient, and and are 1.

[0041] It can be understood that the sine wave digital lock-in amplifier (DLA) is a powerful tool for extracting defect signal features from a noisy environment, and its basic principle involves digitizing the input signal and the known reference signal, and then performing phase-sensitive detection in the digital domain. The system locks the frequency of the reference signal, multiplies it with the input signal to extract the required signal component; then remove the unwanted high-frequency noise through digital low-pass filtering, leaving clear signals for further analysis.

[0042] It can be understood that the real and imaginary part quantities of the eddy current response signal of the phase-controlled eddy current sensor are extracted by using a digital phase-locked amplifier, and the real and imaginary part quotient value characteristics are calculated, through digital phase-sensitive detection and dynamic frequency offset tracking, the system can adaptively adjust the phase matching of the reference signal and the input signal, which significantly improves the capture ability of the weak damage signal under complex working conditions such as electromagnetic interference and mechanical vibration.

[0043] In some embodiments, when micron-level cracks appear on the surface of the production equipment, the distortion of the eddy current distribution in the crack area will cause nonlinear changes in the real and imaginary part quantities, and the real and imaginary part quotient value characteristics can effectively suppress the amplitude drift caused by changes in environmental temperature and humidity, making the damage characteristics have higher discrimination in the frequency domain and the phase domain.

[0044] In some embodiments, the real and imaginary part quotient value characteristics are dynamically integrated with the displacement change rate signal of the displacement sensor and the eddy current energy signal of the eddy current sensor through a weighted fusion mechanism, wherein the dynamic weight coefficient is adaptively adjusted according to the real-time performance of the sensor and the intensity of environmental interference. For example, in high-speed rotating working conditions, the displacement change rate signal is more sensitive to mechanical imbalance, and the system automatically increases its weight to capture instantaneous abnormalities; while in the static load stage, the eddy current energy signal has a more significant response to surface cracks, and the weight coefficient is dynamically optimized accordingly. This fusion strategy based on physical mechanisms and data-driven not only overcomes the defect that fixed weight fusion is not sensitive to changes in working conditions, but also reduces the false detection rate of a single sensor through multi-dimensional feature complementation. Tests show that this scheme reduces the false alarm rate of composite damage by 40%-60%, and improves the recognition sensitivity of micron-level cracks to the order of 0.5 μm.

[0045] In addition, the dynamic weight mechanism realizes the optimal allocation of resources by evaluating the signal-to-noise ratio and stability of each sensor signal in real time. When a sensor is temporarily disabled due to electromagnetic interference, the system automatically reduces its weight and enhances the contribution of other reliable signal sources to ensure the continuity of the monitoring process. The introduction of the real and imaginary part quotient value characteristics provides a new analysis dimension for damage diagnosis. Traditional methods rely on amplitude or energy indicators, which are easily affected by environmental noise. However, phase information and real and imaginary part ratios are more sensitive to changes in the electromagnetic characteristics of damage, especially in the early stages of hidden damage. The quotient value characteristics can trigger an early warning 10%-20% in advance. Finally, the fusion framework provides high-discrimination input features for subsequent wavelet neural network prediction models. Through the synergistic enhancement of multi-dimensional features, the neural network can more accurately learn the damage evolution law, and together promote the transformation of industrial equipment monitoring from passive response to active prediction, providing core technical support for the reliability of intelligent manufacturing systems.

[0046] In this embodiment, the input layer of the wavelet neural network prediction model is a comprehensive feature signal, the hidden layer uses an orthogonal wavelet basis function for multi-resolution analysis, local features are extracted through multi-dimensional wavelet decomposition with a scaling factor of 2, and the output layer is a risk level.

[0047] It should be noted that the multi-resolution analysis of the hidden layer by the orthogonal wavelet basis function breaks through the limitation of traditional neural networks that only rely on single-dimensional modeling in the time domain or the frequency domain. The orthogonal wavelet basis function has time-frequency localization characteristics and can decompose the comprehensive feature signal at different scales, such as separating high-frequency transient impact and low-frequency periodic wear characteristics in mechanical vibration signals, and distinguishing high-frequency micro-crack distortion and low-frequency background noise in electromagnetic response signals.

[0048] This multi-scale analysis capability enables the model to accurately capture weak damage features in cross-physical field coupling effects, and to model the micron-level displacement fluctuations of displacement sensors and high-frequency energy mutations of eddy current sensors in the time-frequency joint domain, significantly improving the recognition sensitivity of composite damage.

[0049] It should also be noted that through multi-dimensional wavelet decomposition with a scaling factor of 2, the model can adaptively adjust the granularity of feature extraction: in the early stage of device operation, a larger scaling factor is used to extract macro trend features to capture overall performance degradation; in the fault development stage, the scaling factor is reduced to focus on local mutation features to identify specific damage types.

[0050] In some embodiments, wavelet decomposition is used to automatically generate feature expressions with physical meaning, and the self-learning ability of the neural network is used to dynamically optimize feature weights, such as in gearbox monitoring, the model autonomously discovers the nonlinear correlation between gear wear and vibration signal harmonic components, and the fault monitoring accuracy is improved to 99.2%.

[0051] In some embodiments, taking actual engineering verification in a certain type of production equipment as an example:

[0052] Among them, the existing technology relies on a single type of sensor for monitoring, specifically, surface adhesive wear detection uses a contact thickness gauge for monthly sampling, or a fixed laser triangulation sensor; internal wear monitoring uses magnetic flux leakage detection technology, and a permanent magnet is configured to magnetize the steel wire rope to measure the magnetic flux leakage; crack detection is based on impedance plane analysis of single-frequency eddy current method, and phase angle change is manually interpreted; dynamic frequency deviation is suppressed by a hardware filtering circuit to suppress environmental interference, without active frequency deviation tracking capability. The data in the table verifies the effectiveness of the core technologies such as multi-sensor fusion, dynamic frequency tracking, and gradient feature analysis, providing new technical possibilities for industrial equipment intelligent monitoring.

[0053] Embodiment 2 As shown in the accompanying drawings Figure 2 Based on the same inventive concept as the foregoing embodiments, the present embodiment provides an industrial internet of things-based production equipment monitoring system, which comprises a management platform, a sensing network platform and an object platform connected in sequence, and the management platform comprises: a first information acquisition module configured to acquire coupling change information of a monitoring part obtained based on combined measurement of a displacement sensor and an eddy current sensor; wherein the coupling change information comprises a change rate signal and an eddy current energy signal, and the monitoring part is a part of a target production equipment that needs to be monitored; a second information acquisition module configured to acquire surface feature information of the monitoring part based on a phase-controlled eddy current sensor; an information fusion module configured to fuse the coupling change information and the surface feature information to obtain a comprehensive feature signal; a risk prediction module configured to input the comprehensive feature signal into a preset wavelet neural network prediction model to output a risk level of the target production equipment.

[0054] The related explanations and examples of the modules in the system of the present embodiment can refer to the methods of the foregoing embodiments, which will not be described here again.

[0055] In addition, it should be noted that the system of the present embodiment further comprises a user platform and a service platform connected in communication with each other, the service platform is connected in communication with the management platform, thereby forming a standard internet of things five-platform structure.

[0056] The physical entity of the user platform comprises various user terminals such as mobile phones, computers, special terminals and the like, and the user terminal can realize service through combination with a user information system software.

[0057] The service platform is a functional platform for realizing service communication.

[0058] The management platform is a functional platform for realizing operation management of the internet of things system, and in some embodiments, the management platform can comprise a plurality of management sub-platforms, the management sub-platforms are respectively connected with the sensing network platform, and each management sub-platform respectively comprises the first information acquisition module, the second information acquisition module, the information fusion module and the risk prediction module.

[0059] In the present embodiment, through integration of multi-sensor collaborative monitoring, dynamic data fusion and intelligent prediction model, the core problems of distortion of multi-source sensor data, rigid fusion strategy and difficulty in capturing weak damage signals in the prior art are systematically solved.

[0060] The system first deploys displacement sensors, eddy current sensors and phased eddy current sensors through a first information acquisition module and a second information acquisition module in the management platform to form a multi-physical field collaborative monitoring network. The displacement sensors capture mechanical vibration and displacement changes, the eddy current sensors analyze electromagnetic field distortion, and the phased eddy current sensors focus on surface micro-crack features. The combination of multiple types of sensors breaks through the limitation of single sensors being easily disturbed by the environment, comprehensively obtains equipment state information from three dimensions of mechanical motion, electromagnetic response and surface topography, and significantly improves the robustness of data acquisition.

[0061] The information fusion module optimizes and integrates the multi-source data in real time through a dynamic weight mechanism: in the high-speed rotating working condition, the displacement change rate is preferentially weighted to capture mechanical imbalance characteristics; in the static load stage, the eddy current energy signal is strengthened to sensitively identify surface damage. This dynamic adaptation strategy overcomes the defect that the traditional fixed weight fusion is not sensitive to working condition changes, and by introducing phase information through real and imaginary part quotient value characteristics, it further suppresses the interference of environmental drift on amplitude signals, so that the complementarity of multi-source data is maximally utilized.

[0062] Embodiment 3 Based on the same inventive idea as the foregoing embodiments, this embodiment provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the method described above.

[0063] Embodiment 4 Based on the same inventive idea as the foregoing embodiments, this embodiment provides a computer readable storage medium, which stores a computer program, and a processor executes the computer program to realize the method described above.

[0064] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent flow transformation obtained by using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A production equipment monitoring method based on industrial Internet of Things, characterized in that: The following steps are involved: Acquiring coupling change information of a monitored portion obtained through combined measurement of a displacement sensor and an eddy current sensor; wherein the coupling change information includes a rate of change signal and an eddy current energy signal, and the monitored portion is a portion on the target production equipment to be monitored; Acquiring surface feature information of the monitored part based on a phased eddy current sensor; fusing the coupling change information with the surface feature information to obtain a comprehensive feature signal; The comprehensive characteristic signal is input into a preset wavelet neural network prediction model to output the risk level of the target production equipment.

2. The production equipment monitoring method based on industrial Internet of Things according to claim 1, characterized in that: The method of obtaining coupling change information of the monitoring part obtained by combined measurement of the displacement sensor and the eddy current sensor comprises the following steps: Based on the original displacement of the monitoring part output by the displacement sensor, the rate of change signal is calculated by eliminating high-frequency noise through Kalman filtering. The expression for calculating the rate of change signal is: ； in, is the Kalman gain; is the measured value of the original displacement; for The displacement value after filtering at the moment; For the moment; for The filtered displacement value at the moment; Based on the eddy current sensor, the collected eddy current signal is decomposed by discrete wavelet transform and then the eddy current energy signal is calculated. The expression for calculating the eddy current energy signal is: ； in, is the eddy current characteristic signal after wavelet transformation; is the number of data points in the sampling window; To sum the index; The coupling change information of the monitored part is obtained based on the change rate signal and the eddy current energy signal.

3. The production equipment monitoring method based on industrial Internet of Things according to claim 2, characterized in that: The displacement sensor is also used to calculate the instantaneous velocity by displacement differential to obtain the rate of change signal. The expression for the displacement change rate at the moment of calculation is: ； in, is the sampling time interval; is the displacement change rate at a given moment.

4. The production equipment monitoring method based on industrial Internet of Things according to claim 3, characterized in that: The method of obtaining coupling change information of the monitored part according to the change rate signal and the eddy current energy signal comprises the following steps: Mapping the rate of change signal and the eddy current energy signal to a state scoring model, which outputs coupling change information; The state scoring model satisfies: ； in, is the weight coefficient; Score the dynamic coupling status; is the displacement change rate threshold; is the eddy current energy threshold.

5. The production equipment monitoring method based on industrial Internet of Things according to claim 3, characterized in that: The obtaining of surface feature information of the monitored part based on the phased eddy current sensor comprises the following steps: Set the excitation signal parameters and collect the eddy current response signal; The discrete wavelet transform model is selected to decompose and quantify the eddy current response signal, and the low-frequency approximate characteristic signal and the high-frequency detail characteristic signal are extracted hierarchically. Perform empirical mode decomposition on the noisy signal, perform threshold quantization on the decomposed high-frequency coefficients, then reconstruct the signal, extract the eigenvalues ​​and output the noise-reduced signal; Based on discrete data cubic spline interpolation and two-dimensional curve fitting, a curve analysis model is established to determine the discrete signal feature extraction interval; Perform discrete analysis on quadratic optimization signals, including standard deviation analysis, discrete signal convolution and analysis; A discrete signal feature analysis and evaluation model is established, and an evaluation algorithm is used to calculate the discrete analysis values ​​and feature signal numerical results to output surface feature information.

6. The production equipment monitoring method based on industrial Internet of Things according to claim 5, characterized in that: The step of fusing the coupling change information with the surface feature information to obtain a comprehensive feature signal comprises the following steps: The real and imaginary parts of the eddy current response signal collected by the phase-controlled eddy current sensor are extracted through a digital lock-in amplifier, and the quotient characteristics of the real and imaginary parts are calculated; Perform weighted fusion of the real and imaginary part quotient feature with the displacement change rate signal of the displacement sensor and the eddy current energy signal of the eddy current sensor to generate a comprehensive feature signal; The fusion process of comprehensive feature signals: ； in, is the real signal component output by the digital lock-in amplifier; is the imaginary signal component output by the digital lock-in amplifier; is the dynamic weight coefficient, and the sum is 1.

7. The production equipment monitoring method based on industrial Internet of Things according to claim 1 or 6, characterized in that: The input layer of the wavelet neural network prediction model is the comprehensive feature signal, the hidden layer uses orthogonal wavelet basis functions for multi-resolution analysis, and extracts local features through multidimensional wavelet decomposition with a scaling factor of 2. The output layer is the risk level.

8. The production equipment monitoring system based on industrial Internet of Things is characterized by: It includes a management platform, a sensor network platform and an object platform that are connected in sequence. The management platform includes: A first information acquisition module is configured to acquire coupling change information of a monitored portion obtained through combined measurement of a displacement sensor and an eddy current sensor; wherein the coupling change information includes a rate of change signal and an eddy current energy signal, and the monitored portion is a portion of the target production equipment to be monitored; A second information acquisition module is used to obtain surface feature information of the monitored part based on a phased eddy current sensor; an information fusion module, configured to fuse the coupling change information with the surface feature information to obtain a comprehensive feature signal; The risk prediction module is used to input the comprehensive characteristic signal into a preset wavelet neural network prediction model to output the risk level of the target production equipment.

9. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and the processor executes the computer program to implement the method according to any one of claims 1 to 7.

Images