Intelligent bearing dust cover and state monitoring method based on edge computing method

Abstract

The application relates to the technical field of bearing state monitoring, and provides an intelligent bearing dust cover and a state monitoring method based on an edge computing method. Temperature, rotating speed and vibration acceleration sensors are integrated in a bearing dust cover to collect bearing operation parameters; vibration acceleration signals are processed through an adaptive sampling algorithm based on a short-time RMS energy threshold method, full sampling is performed in a peak value section, and sampling is reduced in a smooth section; temperature compensation calculation is performed on the processed vibration acceleration signals, temperature drift is eliminated through zero offset and sensitivity functions; the compensated signals are analyzed through an AI model, waveform features are extracted and compared with pre-stored health features, and a bearing fault stage is judged. According to the scheme, the sensors are integrated in the dust cover, state monitoring is achieved without changing the bearing structure, and destructive modification of the equipment in the traditional method is avoided. Edge computing and an adaptive sampling strategy are adopted, data transmission amount is reduced, network bandwidth demand is reduced, key data accurate collection is guaranteed, and the bearing state can be determined in real time and accurately.

Description

Technical Field

[0001] This application relates to the field of bearing condition monitoring technology, and in particular to an intelligent bearing dust cover and condition monitoring method based on edge computing. Background Technology

[0002] As a core component in modern industrial equipment, bearings' operating status directly affects the reliability, safety, and service life of the entire mechanical system. In critical fields such as wind power generation, high-speed railways, and precision machine tools, sudden bearing failures can lead to huge economic losses or even safety accidents. Therefore, real-time and accurate monitoring of bearing operating status to achieve predictive maintenance has always been a goal pursued by the industry. Traditional monitoring methods mainly rely on periodic manual inspections or offline analysis. These methods are not only inefficient but also fail to capture transient, early-stage fault characteristics during bearing degradation, resulting in very limited early warning capabilities.

[0003] Existing online bearing condition monitoring technologies typically require the installation of dedicated sensors on the bearing or adjacent components. For example, threaded holes are drilled in the bearing housing to install piezoelectric accelerometers, or eddy current sensor probes are aligned with the bearing outer ring to measure rotational speed. While these methods can acquire monitoring data, the installation process often inevitably requires machining modifications to the existing bearing housing or bearing end cap, such as drilling, tapping, or milling the mounting surface. This modification not only increases process complexity and installation costs, but more importantly, it may alter the mechanical integrity of the bearing support structure, disturbing its original stress distribution, thereby having unpredictable potential impacts on the bearing's load capacity, dynamic characteristics, and even fatigue life.

[0004] Therefore, existing bearing condition monitoring technologies often require the destruction or alteration of the original structure of the bearing and its associated components in order to achieve the condition monitoring function. This not only introduces additional installation complexity and cost, but also brings potential risks that affect the performance and reliability of the bearing itself. Summary of the Invention

[0005] To address the aforementioned issues with existing bearing condition monitoring technologies, achieving the condition monitoring function often requires damaging or altering the original structure of the bearing and its associated components, introducing additional installation complexity and costs, and also bringing potential risks that affect the performance and reliability of the bearing itself.

[0006] This application provides a smart bearing condition monitoring method based on edge computing, the method comprising:

[0007] Temperature sensor, speed sensor and vibration acceleration sensor are integrated into bearing dust cover to collect bearing temperature signal, speed signal and vibration acceleration signal;

[0008] The vibration acceleration signal collected by the vibration acceleration sensor is processed by an adaptive sampling algorithm. The adaptive sampling algorithm is based on the short-time RMS energy threshold method for peak detection. In the identified peak segment, the vibration acceleration signal is maintained at full sampling rate. In the stable segment, the vibration acceleration signal is downsampled by sliding window mean sampling.

[0009] Temperature compensation calculations are performed on the vibration acceleration signal after adaptive sampling processing, and the effects of temperature drift are eliminated by using a zero-bias compensation function and a sensitivity function.

[0010] The vibration acceleration signal after temperature compensation is analyzed by an AI model. Waveform features are extracted from the vibration acceleration signal and compared with pre-stored healthy state waveform features. The bearing failure stage is determined based on the comparison results.

[0011] In one feasible implementation, the peak detection includes calculating the short-time RMS energy threshold of the vibration acceleration signal as a peak judgment benchmark, and setting a dynamic threshold to achieve accurate capture of transient events.

[0012] The formula for calculating the short-time RMS energy threshold is as follows:

[0013] RMS[n]= ;

[0014] In the formula, RMS[n] represents the root mean square value at time point n. The length of the sliding window. This represents the original value of the vibration acceleration signal at time point k;

[0015] A dynamic threshold θ is set based on the short-time RMS energy threshold. The formula for the dynamic threshold θ is:

[0016] ;

[0017] In the formula, This is the gain coefficient. β is the historical RMS mean value during the steady-state period, and β is the safety margin used to compensate for signal baseline drift.

[0018] In one feasible implementation, the peak detection, based on dynamic threshold determination, further includes:

[0019] The rapid change characteristics of the vibration acceleration signal are captured by calculating the absolute rate of change of the vibration acceleration signal;

[0020] An alert is triggered when the absolute rate of change exceeds a preset rate of change threshold.

[0021] The peak threshold condition is that the RMS value continuously exceeds the dynamic threshold and the duration is greater than the preset minimum peak duration.

[0022] The peak exit condition is that the RMS value is continuously lower than the adjusted dynamic threshold and the duration is greater than the preset exit hold time;

[0023] The formula for calculating the absolute rate of change is:

[0024] ΔX[n]=|x[n]-x[n-1]|;

[0025] In the formula, ΔX[n] represents the absolute rate of change at time point n, x[n] is the original value of the vibration acceleration signal at time point n, and x[n-1] is the original value of the vibration acceleration signal at the previous time point.

[0026] In one feasible implementation, the sliding window mean sampling employs an overlapping window method in the stationary segment processing, and its calculation formula is as follows:

[0027] ;

[0028] Step = N / 2;

[0029] In the formula, y[m] is the downsampled output signal, N is the window length, and Step=N / 2 is the sliding step size, achieving a 50% overlap rate to reduce smoothing distortion. The original signal sample, This is the index of the starting point of the m-th window.

[0030] In one feasible implementation, temperature compensation is also performed on the vibration acceleration sensor, and the effective output value after compensation is calculated using the sensor output value, the zero bias compensation function, and the sensitivity function.

[0031] The temperature compensation formula is as follows:

[0032] ;

[0033] In the formula, Oc is the effective output value after compensation, O is the original output value of the sensor, Z(T) is the zero bias compensation function, and S(T) is the sensitivity compensation function.

[0034] Where Z(T) = aT + b; S(T) = cT + d; where T is the temperature value, and a, b, c, and d are constants obtained through sensor calibration.

[0035] In one feasible implementation, the waveform feature extraction in the AI ​​model analysis includes: time-domain features, frequency-domain features, time-frequency features, implicit rotational speed, and implicit temperature;

[0036] The time-domain features identify impact characteristics through kurtosis calculation, the frequency-domain features detect fault frequencies through FFT analysis, the time-frequency features analyze modulation phenomena through wavelet transform, the implicit rotational speed calibrates the rotational frequency through adaptive filtering, and the implicit temperature evaluates thermal deformation through long-term trend analysis.

[0037] In one feasible implementation, the failure phase includes: a health phase, an early failure phase, a mid-stage failure phase, and a late failure phase;

[0038] The criteria for determining the health stage include: the kurtosis value of the time-domain feature is between 2.9 and 3.1, the peak signal-to-noise ratio of the frequency-domain feature is less than 3dB and the energy distribution is uniform, the time-frequency feature has no irregular transient events, and the implicit rotational speed fluctuation is between 0 and 0.5%.

[0039] The criteria for determining the early fault stage include: the time-domain feature exhibits periodic impact pulses with a kurtosis value greater than 5, the frequency-domain feature has a signal-to-noise ratio of less than 5 dB, and the time-frequency feature impact event is synchronized with the rotation period.

[0040] The criteria for determining the intermediate fault stage include: the amplitude of the time-domain characteristic impact increases and a secondary impact occurs, the signal-to-noise ratio of the frequency domain characteristic is between 5 and 15 dB, and the time-frequency characteristic modulation phenomenon is significant and the envelope spectrum energy is concentrated.

[0041] The criteria for determining the late-stage fault include: the time-domain characteristic signal is saturated and periodically disappears, the frequency-domain characteristic broadband energy increases and the characteristic frequency becomes blurred, and the time-frequency characteristics experience a large number of random impacts and the modulation phenomenon disappears.

[0042] Another aspect of this application provides an intelligent bearing dust cover based on an edge computing method, used to implement the intelligent bearing status monitoring method based on the edge computing method described in any of the above claims, comprising: a dust cover body, a circuit board, and a magnet assembly;

[0043] The dust cover body is snapped onto the standard slot on the outer ring of the bearing by a fixing structure on its edge. The dust cover body includes a metal frame and a rubber sealing layer, and is sealed to the bearing.

[0044] The circuit board is fixed to the side of the dust cover body facing the internal space of the bearing, and a Hall sensor, a temperature sensor, a vibration acceleration sensor and a microprocessor are installed on it.

[0045] The magnet assembly is fixed to the inner ring of the bearing;

[0046] The Hall sensor is configured to sense changes in the magnetic field generated when the magnet assembly rotates to generate a rotational speed signal for the bearing;

[0047] The temperature sensor is configured to monitor the temperature signal of the bearing;

[0048] The vibration acceleration sensor is configured to acquire the vibration acceleration signal of the bearing;

[0049] The microprocessor is configured to run an edge computing algorithm to perform real-time processing, AI model analysis, and adaptive sampling of speed signals, temperature signals, and vibration acceleration signals to determine whether there are any abnormalities in the operating state of the bearing.

[0050] In one feasible implementation, the circuit board includes: a micro 4G module and a circuit board body;

[0051] The Hall sensor is located in the central area of ​​the circuit board body;

[0052] The temperature sensor is located on the edge of the circuit board body and near the bearing raceway;

[0053] The vibration acceleration sensor is located on the side of the circuit board body away from the center of the bearing.

[0054] The miniature 4G module is integrated on the circuit board and is used for data transmission between the Hall sensor, temperature sensor, vibration acceleration sensor and the microprocessor.

[0055] In one feasible implementation, the magnet assembly is a plurality of neodymium iron boron patch magnets, which are uniformly fixed to the circumference of the inner ring of the bearing by an adhesive.

[0056] The sensing end of the Hall sensor is aligned radially with the magnetic pole surface of the magnet assembly;

[0057] The microprocessor uses edge computing algorithms to achieve adaptive sampling and temperature compensation, and combines AI models to analyze the characteristics of the vibration acceleration signal to determine the bearing failure stage in real time.

[0058] As described above, this application provides an intelligent bearing dust cover and condition monitoring method based on edge computing. By integrating sensors into the dust cover, condition monitoring can be achieved without altering the bearing structure, avoiding the destructive modifications required by traditional methods. Simultaneously, the use of edge computing and adaptive sampling strategies reduces data transmission volume and network bandwidth requirements, while temperature compensation improves the accuracy of data acquisition. Through high-frequency vibration acquisition and real-time analysis, early bearing fault signals can be captured, providing early warnings of potential problems, reducing the risk of unplanned downtime, and improving equipment operation and maintenance efficiency. Attached Figure Description

[0059] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the implementation of the invention and, together with the description, serve to explain the principles of the embodiments of the invention. It is obvious that the drawings described below are merely some embodiments of the invention, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0060] Figure 1 A flowchart illustrating an intelligent bearing condition monitoring method based on edge computing, which is an exemplary embodiment of this application;

[0061] Figure 2 A schematic diagram of the overall structure of a smart bearing dust cover based on an edge computing method, illustrating an exemplary embodiment of this application;

[0062] Figure 3 A schematic cross-sectional view of an intelligent bearing dust cover based on edge computing method after installation, as shown in an exemplary embodiment of this application;

[0063] Figure 4 This is a schematic diagram of the circuit board structure shown as an exemplary embodiment of this application.

[0064] Explanation of reference numerals in the attached figures:

[0065] 1-Bearing; 2-Dust cover body; 3-Circuit board; 4-Magnet assembly; 301-Hall sensor; 302-Temperature sensor; 303-Vibration acceleration sensor; 304-Miniature 4G module; 305-Circuit board body. Detailed Implementation

[0066] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the embodiments of the invention will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, mechanisms, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of how embodiments of the invention are carried out.

[0067] Existing bearing condition monitoring technologies typically require the installation of additional sensors, which involves damaging the bearing structure and has significant limitations in adaptability, generally requiring customization based on the bearing's condition. Changing the bearing structure can disrupt the stress distribution, potentially affecting its lifespan. To avoid this, a dust cover was designed. Based on the existing structure of a bearing dust cover, sensors for temperature, speed, and vibration acceleration are integrated onto the dust cover without altering the connection method with the bearing, enabling monitoring of the bearing's operating status.

[0068] This application provides an intelligent bearing condition monitoring method based on edge computing, referring to... Figure 1 As shown, the method includes the following steps:

[0069] S100: Integrates temperature sensor, speed sensor and vibration acceleration sensor on bearing dust cover to collect bearing temperature signal, speed signal and vibration acceleration signal.

[0070] First, a specially designed bearing dust cover is used to collect the bearing's temperature, rotational speed, and vibration acceleration signals. A circuit board 3 is fixed inside the dust cover, integrating a temperature sensor 302, a Hall sensor 301 (serving as a rotational speed sensor), and a vibration acceleration sensor 303. The sensors are mounted on the bearing 1's standard slots using the standard dust cover's mounting structure, without altering the bearing's original structure, thus avoiding damage to the bearing's stress distribution. This step enables real-time acquisition of bearing operating parameters while maintaining ease of installation and versatility.

[0071] S200: The vibration acceleration signal collected by the vibration acceleration sensor is processed by an adaptive sampling algorithm. The adaptive sampling algorithm is based on the short-time RMS energy threshold method for peak detection. In the identified peak segment, the vibration acceleration signal is maintained at the full sampling rate. In the stable segment, the vibration acceleration signal is downsampled by sliding window mean sampling.

[0072] Specifically, the microprocessor calculates the short-time RMS value of the vibration signal in real time and dynamically identifies peak events; the peak segment stores data at the original sampling rate (e.g., 10kHz), and the mean value is calculated and output through a sliding window for the stable segment, reducing the amount of data, effectively reducing redundant data generated by high-frequency sampling, and alleviating bandwidth pressure.

[0073] S300: Performs temperature compensation calculations on the vibration acceleration signal after adaptive sampling processing, and uses a zero-bias compensation function and a sensitivity function to eliminate the influence of temperature drift.

[0074] In this step, temperature compensation calculation is performed on the downsampled signal. The feedback value of temperature sensor 302 is used to perform zero bias and sensitivity compensation on the vibration signal, eliminating the influence of temperature drift of the piezoelectric accelerometer and improving the accuracy of vibration measurement.

[0075] S400: Analyzes the vibration acceleration signal after temperature compensation processing using an AI model, extracts waveform features from the vibration acceleration signal, compares the waveform features with pre-stored healthy state waveform features, and determines the bearing failure stage based on the comparison results.

[0076] This step uses an AI model running on an edge computing device to analyze waveforms in real time and quickly detect anomalies.

[0077] This embodiment combines sensor integration with edge computing intelligent processing, solving problems such as complex installation, large amounts of redundant data transmission, and poor versatility in traditional monitoring. Furthermore, it does not require changes to the bearing structure, thus protecting its mechanical integrity. The method significantly reduces data transmission volume and improves monitoring real-time performance through local adaptive sampling and AI analysis; it is also applicable to standard bearings, demonstrating strong versatility.

[0078] In some embodiments, peak detection is achieved through the following steps: calculating the short-time energy of the vibration acceleration signal as a criterion, and setting a dynamic threshold accordingly to accurately capture transient events.

[0079] First, the short-time RMS energy of the signal is calculated. The microprocessor acquires the vibration signal x[n] in real time and sets a sliding window of a fixed length (e.g., corresponding to a duration of 10ms). (Number of data points). For each time point n, its short-time RMS value is calculated using the following formula:

[0080] RMS[n]= .

[0081] In the formula, RMS[n] represents the root mean square value at time point n, which is used to quantify the local energy level of the signal within the window and provide a benchmark for subsequent peak detection. The length of the sliding window. This represents the original value of the vibration acceleration signal at time point k.

[0082] Then, a dynamic threshold θ is set based on the short-time RMS energy threshold. The formula for the dynamic threshold θ is:

[0083] .

[0084] In the formula, This represents the historical RMS mean over a stable period. β is a gain factor set between 2.5 and 3.5 to effectively cover background noise; β is a small safety margin to compensate for possible baseline drift of the signal.

[0085] During the detection process, the microprocessor continuously compares the real-time calculated RMS[n] with the dynamic threshold. Once RMS[n]> If so, it is determined that a peak event has been triggered.

[0086] This embodiment employs a dynamic threshold mechanism to reliably distinguish genuine abnormal transient events from background noise, effectively reducing false alarms. This ensures that subsequent adaptive sampling, while significantly compressing the data volume, does not lose key characteristic information of bearing faults, enhancing the robustness and accuracy of the entire monitoring system.

[0087] In some embodiments, to further improve the accuracy and reliability of transient event capture, the peak detection mechanism introduces signal absolute rate of change analysis in addition to dynamic threshold judgment, and clarifies the entry and exit conditions for peak values.

[0088] First, based on the rapid early warning of the rate of change, the microprocessor calculates the absolute rate of change ΔX[n] of the vibration signal in real time and sets a threshold for the rate of change. (For example, based on signal characteristics), when An early warning is triggered, and a comprehensive judgment is made in conjunction with the RMS threshold.

[0089] The formula for calculating the absolute rate of change is:

[0090] ΔX[n]=∣x[n]-x[n-1]∣.

[0091] In the formula, ΔX[n] represents the absolute rate of change at time point n, x[n] is the original value of the vibration acceleration signal at time point n, and x[n-1] is the original value of the vibration acceleration signal at the previous time point.

[0092] Secondly, the system's determination of peak events is not an instantaneous jump, but rather based on a defined duration condition. The entry condition is that the RMS value must continuously exceed a dynamic threshold. And the duration is greater than the minimum peak duration. (e.g., 5ms).

[0093] The exit condition is: the RMS value remains below the adjusted threshold (e.g., / , =1.2-1.5) and the duration is greater than the exit hold time. (e.g., 20ms).

[0094] This embodiment introduces a rate of change-assisted judgment and a strict duration determination. This design effectively filters out instantaneous interference, ensures that the identified peak events have clear start and end boundaries and physical meaning, and enhances the robustness of the adaptive sampling algorithm.

[0095] In some embodiments, when the signal is identified as a stationary segment, the microprocessor uses a sliding window averaging method with a 50% overlap to downsample the signal in order to preserve its low-frequency trend while reducing the amount of data.

[0096] Specifically, the algorithm sets the window length to N and the sliding step size to N / 2, thereby achieving a 50% overlap rate. For each output point m, its value... It is calculated using the following formula:

[0097] .

[0098] In the formula, y[m] is the downsampled output signal, and N is the window length. The original signal sample, This is the index of the starting point of the m-th window.

[0099] The calculation in this embodiment involves one mean filtering and downsampling. The main purpose of using an overlapping window design is to avoid signal distortion that may be caused by downsampling and smoothing, and to ensure the representativeness of the data in the stationary segment. This method is activated only in the non-peak segment, thereby achieving effective data compression while preserving the macroscopic trend information of the signal to the maximum extent.

[0100] In some embodiments, to eliminate the impact of sensor temperature drift on measurement accuracy, the system performs compensation calculations on the vibration signal based on real-time temperature readings after completing adaptive sampling.

[0101] Specifically, the microprocessor reads the real-time value T from the temperature sensor 302 and combines it with the original output value O from the vibration acceleration sensor 303 to calculate the corrected effective output Oc using the following compensation formula:

[0102] .

[0103] In the formula, Oc is the effective output value after compensation, and O is the original output value of the sensor.

[0104] Where: Z(T) = aT + b is the compensation function for zero bias varying with temperature; S(T) = cT + d is the compensation function for sensitivity varying with temperature. Parameters a, b, c, and d are constant coefficients obtained through pre-calibration of the sensor.

[0105] The compensation process in this embodiment can effectively suppress the zero bias and sensitivity drift of the piezoelectric accelerometer caused by changes in ambient temperature, thereby significantly improving the measurement accuracy of vibration signals and providing a high-quality data foundation for the reliable analysis of subsequent AI models.

[0106] In some embodiments, the AI ​​model analysis stage diagnoses the bearing condition through multi-dimensional waveform feature extraction. Waveform feature extraction encompasses time-domain features, frequency-domain features, time-frequency features, implicit rotational speed, and implicit temperature; each dimension employs a specialized algorithm to comprehensively extract fault information from the vibration signal.

[0107] During the healthy phase, the bearing operates smoothly, exhibiting a waveform characteristic of an approximately Gaussian distribution in the time domain with a kurtosis value stable at around 3. The frequency domain energy is uniformly distributed without significant peaks, and the time-frequency characteristics show no irregular transient events, implying minimal speed fluctuations, typically below 0.5%. These characteristics collectively indicate that the bearing is in a damage-free state, providing a benchmark for subsequent fault diagnosis.

[0108] As bearings begin to show early failures, such as damage less than 0.5 mm, the time-domain characteristics first show periodic impact pulses, with the kurtosis value significantly increasing to above 5. In the frequency-domain characteristics, frequency energy initially appears, but the signal-to-noise ratio is below 5 dB. The time-frequency characteristics show that the impact event is synchronized with the rotation period, implying that although the speed is basically stable, there are already slight fluctuations, indicating the occurrence of a failure.

[0109] In the mid-stage of the fault, bearing damage expands, time-domain characteristics deteriorate further, impact amplitude increases and secondary impacts occur, frequency domain characteristic frequencies and their sidebands become more pronounced, signal-to-noise ratio rises to 5-15 dB, time-frequency characteristic modulation is significant, envelope spectrum energy is concentrated, implicit speed fluctuations increase, and the influence of load changes intensifies. The multi-dimensional characteristic changes in this stage provide crucial information for fault location.

[0110] When the fault progresses to the late stage, the damage exceeds the threshold, the time domain signal saturates and periodically disappears, the frequency domain broadband energy increases, the characteristic frequency becomes blurred, a large number of random impacts appear in the time and frequency characteristics, the modulation phenomenon disappears, and the implied speed fluctuations are severe or even lost lock. These characteristics all point to the severe degradation of the bearing.

[0111] In this embodiment, the AI ​​model compares the above features with the pre-stored health status waveforms in real time to comprehensively judge the fault stage, enabling rapid response and effective prevention of sudden faults.

[0112] Another aspect of this application provides a smart bearing dust cover based on an edge computing method, used to implement the aforementioned smart bearing condition monitoring method based on an edge computing method, with reference to... Figures 2-4 As shown, it includes: a dust cover body 2, a circuit board 3, and a magnet assembly 4.

[0113] The dust cover body 2 is snapped into the standard groove of the outer ring of the bearing 1 by an edge fixing structure, forming a non-intrusive installation structure.

[0114] Furthermore, the dust cover body 2 includes a metal frame and a rubber sealing layer. The metal frame is interference-fitted with the outer ring groove of the bearing 1, and the rubber sealing layer covers the outer surface of the metal frame. This structure forms a double sealing protection: the metal frame provides mechanical support, and the rubber sealing layer prevents contaminants from entering the bearing.

[0115] During installation, the metal frame and the outer ring groove are press-fitted together to form a mechanical fixation, and the rubber sealing layer is compressed to form a sealed interface. This double sealing improves the dustproof and waterproof rating and prevents insufficient sealing performance of the dust cover. At the same time, the mechanical strength of the metal frame ensures installation stability and avoids loosening caused by vibration.

[0116] The circuit board 3 is fixed to the side of the dust cover body 2 facing the internal space of the bearing 1, and integrates a Hall sensor 301, a temperature sensor 302, a vibration acceleration sensor 303 and a microprocessor.

[0117] Among them, Hall sensor 301 is configured to sense the change in magnetic field generated when magnet assembly 4 rotates to generate rotation speed signal of bearing 1; temperature sensor 302 is configured to monitor temperature signal of bearing 1; and vibration acceleration sensor 303 is configured to collect vibration acceleration signal of bearing 1.

[0118] In some embodiments, the circuit board 3 further includes a circuit board body 305, a Hall sensor 301 disposed in the central region, and a temperature sensor 302 located at the edge near the raceway of the bearing 1, ensuring that the temperature sensor 302 can accurately monitor the raceway temperature signal of the bearing 1. A vibration acceleration sensor 303 is located on the side away from the center of the bearing 1, enabling the vibration acceleration sensor 303 to efficiently capture radial vibration acceleration signals. Each sensor is connected to a microprocessor via printed circuitry to form an integrated data acquisition module.

[0119] Furthermore, the microprocessor is configured to run edge computing algorithms to perform real-time processing, AI model analysis, and adaptive sampling of speed signals, temperature signals, and vibration acceleration signals in order to determine whether there are any abnormalities in the operating status of bearing 1.

[0120] It is understandable that the circuit board 3 in this embodiment improves data accuracy by optimizing the sensor position, eliminates environmental influences by using a temperature compensation algorithm, and reduces the size of the circuit board through integrated design, thereby reducing the space occupied by the circuit board and improving the system integration to adapt to the limited space of the dust cover.

[0121] The magnet assembly 4 is fixed to the inner ring of the bearing 1. In some embodiments, the magnet assembly 4 consists of multiple neodymium iron boron (NdFeB) patch magnets, uniformly bonded to the circumferential surface of the inner ring of the bearing 1, and aligned radially with the Hall sensor 301. Surface treatment before bonding ensures bonding strength, forming a stable magnetic field signal source. The structure ensures a linear relationship between magnetic field changes and bearing rotation speed, achieving accurate rotation speed measurement. Non-contact measurement also extends bearing life. Simultaneously, the use of NdFeB material ensures stable magnetic field strength, solving the measurement error problem caused by low magnetic field strength.

[0122] Specifically, the bearing condition detection process of the intelligent bearing dust cover based on edge computing in this application is as follows: When bearing 1 rotates, magnet assembly 4 rotates with the inner ring. Hall sensor 301 senses the change in magnetic field and generates a pulse signal, which is processed by the microprocessor and converted into rotational speed data. Temperature sensor 302 monitors the temperature of bearing 1 in real time, and vibration acceleration sensor 303 collects vibration acceleration signals. The microprocessor processes the rotational speed signal, temperature signal, and vibration acceleration signal in real time: correcting the temperature drift of the vibration acceleration sensor through the temperature signal, and analyzing the vibration characteristics by combining the rotational speed signal.

[0123] The structure of the intelligent bearing dust cover provided in this embodiment solves the problem that traditional technology requires damaging the bearing structure to install sensors, and avoids the impact of stress distribution changes on bearing life; it also solves the technical problem of untimely response caused by cloud transmission delay by realizing local data processing through edge computing.

[0124] In some embodiments of this application, the circuit board 3 further includes a miniature 4G module 304, which is soldered to the upper surface of the circuit board body 305. The miniature 4G module 304 is connected to the microprocessor via printed circuitry to achieve data transmission functionality. The size of the miniature 4G module 304 is adapted to the internal space of the dust cover, ensuring that it does not affect the layout of other sensors.

[0125] Understandably, the data processed by the microprocessor is uploaded to the cloud system via the micro 4G module 304. Wireless transmission enhances system flexibility, while the low-power design of the micro 4G module extends battery life, ensuring continuous monitoring and extending system lifespan.

[0126] As can be seen from the above embodiments, the intelligent bearing dust cover and condition monitoring method of this application achieves real-time monitoring of bearing condition through a non-invasive design, avoiding the destructive modification of the bearing structure required by traditional methods, thereby maintaining the mechanical integrity and service life of the bearing. The use of edge computing effectively reduces data transmission volume and lowers network bandwidth requirements, while adaptive sampling and temperature compensation improve the accuracy of data acquisition. The solution also enhances the system's environmental adaptability and reliability, supports remote monitoring and intelligent decision-making, facilitates early fault warning, reduces the risk of unplanned downtime, and improves equipment operation and maintenance efficiency.

[0127] Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the disclosure in the specification and examples. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

Claims

1. A smart bearing condition monitoring method based on edge computing, characterized in that, The method includes: Temperature sensor, speed sensor and vibration acceleration sensor are integrated into bearing dust cover to collect bearing temperature signal, speed signal and vibration acceleration signal; The vibration acceleration signal collected by the vibration acceleration sensor is processed by an adaptive sampling algorithm. The adaptive sampling algorithm is based on the short-time RMS energy threshold method for peak detection. In the identified peak segment, the vibration acceleration signal is maintained at full sampling rate. In the stable segment, the vibration acceleration signal is downsampled by sliding window mean sampling. Temperature compensation calculations are performed on the vibration acceleration signal after adaptive sampling processing, and the effects of temperature drift are eliminated by using a zero-bias compensation function and a sensitivity function. The vibration acceleration signal after temperature compensation is analyzed by an AI model. Waveform features are extracted from the vibration acceleration signal and compared with pre-stored healthy state waveform features. The bearing failure stage is determined based on the comparison results. The peak detection includes calculating the short-time RMS energy threshold of the vibration acceleration signal as the peak judgment benchmark, and setting a dynamic threshold to achieve accurate capture of transient events. The formula for calculating the short-time RMS energy threshold is as follows: RMS[n]= ; In the formula, RMS[n] represents the root mean square value at time point n. The length of the sliding window. This represents the original value of the vibration acceleration signal at time point k; A dynamic threshold θ is set based on the short-time RMS energy threshold. The formula for the dynamic threshold θ is: ； In the formula, This is the gain coefficient. β is the historical RMS mean of the steady-state segment, and β is the safety margin used to compensate for signal baseline drift. The peak detection, based on dynamic threshold judgment, also includes: The rapid change characteristics of the vibration acceleration signal are captured by calculating the absolute rate of change of the vibration acceleration signal; An alert is triggered when the absolute rate of change exceeds a preset rate of change threshold. The peak threshold condition is that the RMS value continuously exceeds the dynamic threshold and the duration is greater than the preset minimum peak duration. The peak exit condition is that the RMS value is continuously lower than the adjusted dynamic threshold and the duration is greater than the preset exit hold time; The formula for calculating the absolute rate of change is: ΔX[n]=|x[n]-x[n-1]|; In the formula, ΔX[n] represents the absolute rate of change at time point n, x[n] is the original value of the vibration acceleration signal at time point n, and x[n-1] is the original value of the vibration acceleration signal at the previous time point.

2. The intelligent bearing condition monitoring method based on edge computing as described in claim 1, characterized in that, The sliding window mean sampling employs an overlapping window method in the processing of the stationary segment, and its calculation formula is as follows: ； Step = N / 2; In the formula, y[m] is the downsampled output signal, N is the window length, and Step=N / 2 is the sliding step size, achieving a 50% overlap rate to reduce smoothing distortion. The original signal sample, This is the index of the starting point of the m-th window.

3. The intelligent bearing condition monitoring method based on edge computing as described in claim 1, characterized in that, It also includes temperature compensation for the vibration acceleration sensor, and calculating the effective output value after compensation using the sensor output value, zero bias compensation function and sensitivity function; The temperature compensation formula is as follows: ； In the formula, Oc is the effective output value after compensation, O is the original output value of the sensor, Z(T) is the zero bias compensation function, and S(T) is the sensitivity compensation function. Where Z(T) = aT + b; S(T) = cT + d; where T is the temperature value, and a, b, c, and d are constants obtained through sensor calibration.

4. The intelligent bearing condition monitoring method based on edge computing as described in claim 1, characterized in that, The waveform feature extraction in the AI ​​model analysis includes: time-domain features, frequency-domain features, time-frequency features, implicit rotational speed, and implicit temperature; The time-domain features identify impact characteristics through kurtosis calculation, the frequency-domain features detect fault frequencies through FFT analysis, the time-frequency features analyze modulation phenomena through wavelet transform, the implicit rotational speed calibrates the rotational frequency through adaptive filtering, and the implicit temperature evaluates thermal deformation through long-term trend analysis.

5. The intelligent bearing condition monitoring method based on edge computing as described in claim 4, characterized in that, The failure phases include: health phase, early failure phase, mid-term failure phase, and late failure phase; The criteria for determining the health stage include: the kurtosis value of the time-domain feature is between 2.9 and 3.1, the peak signal-to-noise ratio of the frequency-domain feature is less than 3dB and the energy distribution is uniform, the time-frequency feature has no irregular transient events, and the implicit rotational speed fluctuation is between 0 and 0.5%. The criteria for determining the early fault stage include: the time-domain feature exhibits periodic impact pulses with a kurtosis value greater than 5, the frequency-domain feature has a signal-to-noise ratio of less than 5 dB, and the time-frequency feature impact event is synchronized with the rotation period. The criteria for determining the intermediate fault stage include: the amplitude of the time-domain characteristic impact increases and a secondary impact occurs, the signal-to-noise ratio of the frequency domain characteristic is between 5 and 15 dB, and the time-frequency characteristic modulation phenomenon is significant and the envelope spectrum energy is concentrated. The criteria for determining the late-stage fault include: the time-domain characteristic signal is saturated and periodically disappears, the frequency-domain characteristic broadband energy increases and the characteristic frequency becomes blurred, and the time-frequency characteristics experience a large number of random impacts and the modulation phenomenon disappears.

6. A smart bearing dust cover based on edge computing, used to implement the smart bearing condition monitoring method based on edge computing as described in any one of claims 1-5, characterized in that, include: Dust cover body (2), circuit board (3) and magnet assembly (4); The dust cover body (2) is snapped onto the standard slot of the outer ring of the bearing (1) by the fixing structure of its edge. The dust cover body (2) includes a metal skeleton and a rubber sealing layer and is sealed to the bearing (1). The circuit board (3) is fixed to the side of the dust cover body (2) facing the internal space of the bearing (1), and a Hall sensor (301), a temperature sensor (302), a vibration acceleration sensor (303) and a microprocessor are provided on it; The magnet assembly (4) is fixed to the inner ring of the bearing (1); The Hall sensor (301) is configured to sense the change in magnetic field generated when the magnet assembly (4) rotates to generate a rotational speed signal of the bearing (1); The temperature sensor (302) is configured to monitor the temperature signal of the bearing (1); The vibration acceleration sensor (303) is configured to acquire the vibration acceleration signal of the bearing (1); The microprocessor is configured to run an edge computing algorithm to perform real-time processing, AI model analysis and adaptive sampling processing on the rotation speed signal, temperature signal and vibration acceleration signal, in order to determine whether there is an abnormality in the operating state of the bearing (1).

7. The intelligent bearing dust cover based on edge computing method according to claim 6, characterized in that, The circuit board (3) includes: a micro 4G module (304) and a circuit board body (305). The Hall sensor (301) is disposed in the central area of ​​the circuit board body (305); The temperature sensor (302) is located on the edge of the circuit board body (305) and close to the raceway of the bearing (1); The vibration acceleration sensor (303) is disposed on the side of the circuit board body (305) away from the center of the bearing; The micro 4G module (304) is integrated on the circuit board body (305) and is used for data transmission between the Hall sensor (301), temperature sensor (302), vibration acceleration sensor (303) and the microprocessor.

8. The intelligent bearing dust cover based on edge computing method according to claim 7, characterized in that, The magnet assembly (4) consists of multiple neodymium iron boron patch magnets, which are uniformly fixed to the circumference of the inner ring of the bearing (1) by adhesive. The sensing end of the Hall sensor (301) is aligned radially with the magnetic pole surface of the magnet assembly (4); The microprocessor uses edge computing algorithms to achieve adaptive sampling and temperature compensation, and combines AI models to analyze the characteristics of the vibration acceleration signal to determine the bearing failure stage in real time.

Images