High-speed bearing intelligent state monitoring system based on minimal quantity lubrication

Through the intelligent state monitoring system integrating multi-sensors and deep learning algorithms on high-speed bearings, real-time monitoring of bearing operating status and lubricating oil supply are achieved, which solves the problems of high friction and high temperature of high-speed bearings, extends the service life and improves the Reliability of the equipment.

CN118960843AInactive Publication Date: 2024-11-15NANJING UNIV OF AERONAUTICS & ASTRONAUTICS

Patent Information

Application Number
CN202411039657.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

High-speed bearings increase vibration due to high friction and high temperature during operation, which affects the accuracy of fault monitoring and shortens service life. The existing technology fails to effectively detect and regulate lubrication conditions, affecting equipment performance and reliability.

Method used

A high-speed bearing intelligent state monitoring system based on micro lubrication was designed, combining bearing modules, flexible ring circuit modules and intelligent lubrication modules to monitor the bearing status in real time through strain, temperature sensors and vibration sensors, and use the intrinsic time scale decomposition method and The convolutional neural network performs data processing to achieve accurate supply of lubricating oil.

Benefits of technology

Real-time monitoring and diagnosis of bearing operating status is realized, quickly reflecting lubrication, extending the service life of the bearing, reducing friction and wear, and improving the energy efficiency and reliability of the equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118960843A_ABST
    Figure CN118960843A_ABST
Patent Text Reader

Abstract

The invention discloses a high-speed bearing intelligent state monitoring system based on minimal quantity lubrication, and belongs to the field of intelligent bearings. Aiming at the characteristics of weak vibration signal and complex lubrication mechanism of the bearing, multivariate information measuring points are applied to the bearing, an intrinsic time scale decomposition method and a deep learning convolutional neural network algorithm are combined, multiple signal types such as rotating speed, vibration, temperature, stress strain and the like are fused, a multi-source comprehensive diagnosis result is obtained, and the diagnosis accuracy is improved. The operation condition and the lubrication condition of the bearing can be more quickly and fully reflected, the real-time monitoring and diagnosis of the operation state of the bearing can be realized, and meanwhile, the precise lubricating oil supply can be carried out on the bearing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of intelligent bearings, and in particular to a high-speed bearing intelligent state monitoring system based on minimal lubrication. Background Art

[0002] As high-end mechanical equipment, high-speed rolling bearings play an extremely important role in many fields, such as space flywheels and slewing bearings of aircraft engine main shafts. Improving their performance and increasing the reliability of their safe operation are the key to improving the operating efficiency of equipment and extending their service life. For high-speed bearings, when they are running, they are usually accompanied by high friction between the moving pairs, which releases a large amount of heat and causes the bearing components to heat up rapidly. This will not only intensify vibration, produce abnormal noise, and affect the accuracy of fault monitoring and diagnosis, but also cause bearing raceway damage or even raceway peeling, greatly shortening the service life of the bearing, reducing system reliability, and thus affecting the performance of the equipment system. Therefore, in the face of high friction, high temperature and other problems of high-speed bearings, accurate and effective lubrication management and regular maintenance are of great significance to ensure the normal operation of bearings, improve the energy efficiency of equipment, and reduce production losses. This also poses new challenges to the lubrication method and lubrication efficiency of high-speed bearings.

[0003] The lubricating oil micro-supply technology can provide precise lubrication control, ensure that the bearing is always in the best lubrication state, reduce friction and wear, extend the service life of the bearing, and reduce unnecessary lubricating oil consumption, avoid energy waste, and extend endurance. Common bearing lubrication micro-supply technologies now include: lubricating oil atomization technology, lubricating oil closed system, and oil-gas lubrication technology. Lubricating oil atomization technology uses airflow to atomize the lubricating oil into tiny particles, and then delivers them to the bearing surface through a spray device. This technology can achieve very fine lubrication control and can be implemented on running equipment without stopping. The lubricating oil closed system wraps the bearing in a closed system and releases a small amount of lubricating oil only when needed. This system can reduce the pollution of the lubricating oil by the external environment and ensure that the bearing always maintains a good lubrication state during long-term operation. The oil-gas lubrication technology uses the continuous airflow formed by compressed air to form an oil film in the lubrication pipeline with a certain amount of micro-lubricating oil, which is transported to the lubrication point along the pipe wall by the compressed air, which can achieve continuous oil flow, keep the friction point lubricated with new oil at all times, and prevent contaminants from entering the bearing assembly. Compared with oil mist lubrication, its lubricant consumption is reduced by about 10%, achieving energy saving and efficiency improvement. Therefore, the optimal micro-lubricant supply technology solution should be determined first, and then the lubricant control system should be designed based on this.

[0004] The lubricating oil micro-supply technology can be combined with the intelligent monitoring system to monitor the working status of the bearing in real time through sensors. Use information processing, automatic control and other technologies to detect bearing faults and perform real-time lubricating oil regulation to prevent equipment damage, improve the preventive maintenance capabilities of the equipment, and ensure safe and reliable operation of the system. In the prior art, Liu Hao et al. studied intelligent bearings with embedded structures, and monitored the operating status signals of the bearings by slotting and embedding vibration acceleration sensors in the outer ring of the bearings. Nguyen et al. proposed a new bearing bidirectional load monitoring technology (axial and radial) based on intelligent sensor coatings. The coating is screen-printed on the surface of the cross-shaped steel substrate, and the piezoelectric layer covers most of the surface of the outer ring. The deformation of the outer ring under load is monitored through a surrounding electrode connected to an electrical connector. Related research has not yet considered the detection and feedback regulation of lubrication conditions, and it is very important to ensure that the bearing has an efficient lubrication regulation system during operation. Summary of the invention

[0005] The present invention provides a high-speed bearing intelligent state monitoring system based on minimal lubrication, which can realize real-time monitoring and diagnosis of the bearing operation state and simultaneously accurately supply lubricating oil to the bearing.

[0006] The embodiment of the present invention provides a high-speed bearing intelligent state monitoring system based on minimal lubrication, comprising: a bearing module, a flexible annular circuit module, and an intelligent lubrication module;

[0007] Wherein, the bearing module comprises a bearing inner ring and a bearing outer ring, and the bearing inner ring and the bearing outer ring are provided with a strain sensor and a temperature sensor;

[0008] The flexible annular circuit module includes a vibration sensor, a rotation speed sensor, and a data processing module. The data processing module collects the multi-source information set of the bearing collected by each sensor and converts it into a digital signal for data processing to obtain the bearing status diagnosis result;

[0009] The intelligent lubrication module quantitatively distributes and supplies lubricating oil according to the bearing status diagnosis result, and delivers the lubricating oil to the bearing lubrication point.

[0010] Optionally, in one embodiment of the present invention, the bearing module also includes a bearing seat, bearing rolling elements and a bearing retainer, the bearing outer ring is fixed to the bearing seat, and four evenly distributed grooves are respectively provided on the bearing outer ring and the bearing inner ring, and strain sensors and temperature sensors are cross-installed in the grooves.

[0011] Optionally, in one embodiment of the present invention, the flexible ring circuit module also includes a power management module and a communication module, the power management module supplies power to each sensor, the communication module and the data processing module, and the communication module uses wireless communication to communicate between the data processing module and external devices.

[0012] Optionally, in one embodiment of the present invention, the flexible annular circuit module is fixed to the bearing end cover, the bearing end cover is fixedly mounted on the bearing seat, and the bearing seat is provided with a through hole to allow an oil and gas pipeline to pass through to provide lubricating oil to the bearing.

[0013] Optionally, in one embodiment of the present invention, the intelligent lubrication module includes an oil and gas delivery pipeline, a nozzle, an oil tank, a gear pump, an oil filter, an electromagnetic flowmeter, an air compressor, a gas filter, an air volume flowmeter, an oil and gas mixer, and an oil and gas distributor; the nozzle is located above the level of the inner ring of the rolling bearing, and the air enters the oil and gas mixer from the air compressor through the gas filter and the air volume flowmeter. The lubricating oil is supplied to the oil and gas mixing block by the gear pump through the oil filter and the electromagnetic flowmeter, and the oil and gas mixture is delivered to the oil and gas distributor through the oil and gas mixer. The lubricating oil is quantitatively distributed and supplied to the oil and gas distributor according to the bearing status diagnosis results, and finally the lubricating oil is delivered to the bearing lubrication point by the oil and gas delivery pipeline and the nozzle.

[0014] Optionally, in one embodiment of the present invention, the data processing module is specifically used to preprocess the bearing multi-source information set collected by each sensor, perform feature extraction processing on the preprocessed bearing multi-source information set using the intrinsic time scale decomposition (ITD) method, and process the extracted features using a convolutional neural network to obtain a bearing status diagnosis result.

[0015] Optionally, in one embodiment of the present invention, the preprocessed bearing multi-source information set is subjected to feature extraction processing using an intrinsic time scale decomposition method, including:

[0016] Define a baseline extraction operator L, so that the original signal X in the bearing multi-source information set t After a decomposition, it is as follows:

[0017] X t =LX t +(1-L)X t =L t +H t

[0018] Among them, LX t =L t is the baseline signal, (1-L)X t =H tIt is the remaining signal after the baseline component is extracted from the original signal. After decomposition, it contains a baseline signal and an inherent rotation component.

[0019] Assume that τ k (k=1,2,……) is X t The moment when the extreme point appears, define τ 0 =0, and record X k =X(τ k ), L k =L(τ k );

[0020] Assume that L t , H t In [0,τ k ] is defined, X k In [0,τ k+2 ] is defined, then the value between any two adjacent maximum or minimum points [τ k ,τ k+2 ] Construct a segmented line segment, there must be τ in the interval k+1 The extreme point at the moment, calculate the baseline signal control point L k+1 :

[0021]

[0022] Among them, parameter α is the linear gain, and 0<α<1;

[0023] Between the several intervals generated by the baseline signal control points, each segment of the original signal is linearly transformed to obtain:

[0024]

[0025] Among them, L k t is the baseline signal segment obtained by linear transformation of the original signal in the kth interval segment;

[0026] Calculate the baseline signal segments of all interval segments in turn, and connect them in turn to obtain the complete baseline signal L t , the original signal X t After extracting a baseline signal, the remaining residual signal is the inherent rotation component. At this point, the first decomposition of the original signal is completed.

[0027] L t Continue to decompose as the new original signal, repeat the above decomposition steps until the baseline signal obtained by decomposition is a monotonic function or a constant or a function with less than three extreme points, which is called the trend component. Then the original signal X t The final decomposition of z is:

[0028] X t =L1 t +H 1 t =(L 2 t +H 2 t )+H 1 t =…=H 1 t +H 2 t +…+H m t +L m t

[0029] Each signal of the bearing multi-source information set is decomposed by the intrinsic time scale decomposition method to obtain multiple intrinsic rotation components and a trend component, where each intrinsic rotation component and trend component is represented as a matrix, each row represents a time step, and each column represents a feature.

[0030] Optionally, in one embodiment of the present invention, the extracted features are processed using a convolutional neural network to obtain a bearing state diagnosis result, including:

[0031] The matrix obtained after decomposition by the intrinsic time scale decomposition method is used as input to the convolution layer of the convolutional neural network deep learning model. The convolution process is:

[0032]

[0033] Among them, z k l is the k-th output of the l-th layer, T is the number of convolutional layer kernels, φ is the activation function, W k l is the kth convolution kernel of the lth convolution layer, b k l is the corresponding bias, f k l-1 It is the feature information of the input feature map decomposed by the intrinsic time scale decomposition method;

[0034] Construct the maximum pooling layer, the calculation formula is:

[0035]

[0036] Among them, P j l+1 (i) is the corresponding value of the l+1th layer neuron in the pooling operation, T is the width of the pooling area, and a j l (t) is the activation value of the pooled area;

[0037] Adding a feature concatenation layer after the pooling layer can concatenate the feature maps of multiple intrinsic rotation components in the feature dimension to form a comprehensive feature representation and further complete high-level feature extraction. Assuming there are n feature maps, each with a size of (h, w, c), the feature concatenation representation is:

[0038] concatenated_features=concat(feature_map1,feature_map2,…,feature_mapn,axis=-1)

[0039] Among them: feature_map i represents the i-th feature map; concat represents the concatenation operation, and axis=-1 represents concatenation on the last feature dimension;

[0040] The size of the concatenated feature map is (h,w,n×c). The concatenated feature map is flattened and input into the fully connected layer for classification. The loss function J is set to optimize the model parameters and minimize the expectation of each sample. The convolutional neural network is trained using forward propagation:

[0041]

[0042] Among them, w and b are the weight and bias in the network, x is the input sample, y is the target output, and a l is the output of the lth layer of the model;

[0043] The trained convolutional neural network deep learning model is used to obtain the bearing status diagnosis results. If the bearing is in an oil-deficient state, the gear pump is controlled to increase the oil delivery; otherwise, the oil delivery is reduced.

[0044] Optionally, in one embodiment of the present invention, in the bearing module, the temperature sensor adopts SHT20; the stress strain sensor adopts a stress strain gauge.

[0045] Optionally, in one embodiment of the present invention, in the flexible annular circuit module, the speed sensor adopts an eddy current speed sensor, and the measuring point center of the speed sensor is located on the same horizontal line as the bearing roller; the vibration sensor adopts ADX355 type.

[0046] The high-speed bearing intelligent condition monitoring system based on minimal lubrication according to the embodiment of the present invention has the following beneficial effects:

[0047] (1) In view of the characteristics of weak bearing vibration signals and complex lubrication mechanisms, the present invention applies multi-information measurement points to the bearing, integrates multiple signal types such as speed, vibration, temperature, stress and strain, and combines the intrinsic time scale decomposition method with the deep learning convolutional neural network algorithm to obtain multi-source comprehensive diagnosis results, which can more quickly and fully reflect the operation and lubrication conditions of the bearing.

[0048] (2) The present invention adopts a combination of embedded and external methods, and installs the temperature sensor and the stress strain sensor on the inner ring and the outer ring of the bearing respectively to monitor the temperature and oil film pressure of the two respectively. The temperature difference between the inner and outer rings, as well as the oil film pressure of the inner and outer rings can be obtained more accurately, which can reduce the damage to the bearing structure caused by the complete use of embedded methods, and at the same time reduce the error of the inner and outer ring bearing temperature measurement.

[0049] (3) The present invention measures the rotation speed of the bearing cage by installing an eddy current sensor, thereby achieving accurate measurement of the rotation speed of the bearing cage under lubrication conditions.

[0050] (4) The present invention adopts the micro-oil supply technology of oil-gas lubrication, uses compressed air to quantitatively deliver lubricating oil to the bearing cavity, and combines intelligent sensing and intelligent control technology to control the actuator's distribution and supply of oil and gas to meet the needs of precise oil supply.

[0051] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0053] Figure 1 A schematic diagram of the structure of a high-speed bearing intelligent condition monitoring system based on minimal lubrication according to an embodiment of the present invention;

[0054] Figure 2 A schematic diagram of the structure of a smart bearing according to an embodiment of the present invention;

[0055] Figure 3 A schematic diagram of the installation method of the annular circuit board of the intelligent bearing according to an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of the intelligent oil-gas lubrication structure of an embodiment of the present invention;

[0057] Figure 5 This is a schematic diagram of the intelligent oil-gas lubrication control principle of an embodiment of the present invention;

[0058] Figure 6This is a structural diagram of a multi-source signal fusion diagnosis neural network according to an embodiment of the present invention.

[0059] Explanation of the accompanying drawings: 1-bearing outer ring; 2-rolling element; 3-retaining rack; 4-bearing inner ring; 5-bearing outer ring groove; 6-bearing outer ring temperature sensor; 7-bearing outer ring stress and strain sensor; 8-bearing inner ring groove; 9-bearing inner ring temperature sensor; 10-bearing inner ring stress and strain sensor; 11-bearing seat; 12-bearing end cover; 13-annular circuit board; 14-oil and gas transmission pipeline; 15-nozzle; 16-oil tank; 17-gear pump; 18-oil filter; 19-electromagnetic flowmeter; 20 air compressor; 21-gas filter; 22-air volume flowmeter; 23-oil and gas mixer; 24-oil and gas distributor; 25-bearing. DETAILED DESCRIPTION

[0060] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0061] like Figures 1 to 3 As shown, the high-speed bearing intelligent condition monitoring system based on minimal lubrication includes: a bearing module, a flexible ring circuit module, and an intelligent lubrication module.

[0062] The bearing module includes a bearing outer ring 1 and a bearing inner ring 4 , and the bearing inner ring 4 and the bearing outer ring 1 are provided with strain sensors and temperature sensors.

[0063] The flexible annular circuit module includes a vibration sensor, a speed sensor, and a data processing module. The data processing module collects the multi-source information sets of the bearing collected by each sensor and converts them into digital signals for data processing to obtain the bearing status diagnosis results.

[0064] The intelligent lubrication module quantitatively distributes and supplies lubricating oil according to the bearing status diagnosis results and delivers the lubricating oil to the bearing lubrication points.

[0065] Optionally, in one embodiment of the present invention, the bearing module also includes a bearing seat 11, bearing rolling elements 2 and a bearing retainer 3, the bearing outer ring 1 is fixed to the bearing seat 11, and four evenly distributed grooves (5 and 8) are respectively provided on the bearing outer ring 1 and the bearing inner ring 4, and strain sensors (7 and 10) and temperature sensors (6 and 9) are cross-installed in the grooves.

[0066] Optionally, the temperature sensor adopts SHT20, the stress strain sensor adopts a stress strain gauge, the vibration signal sensor adopts ADX355, the speed sensor adopts an eddy current speed sensor, and the measuring point center of the eddy current speed sensor is on the same horizontal line as the bearing roller.

[0067] Optionally, in one embodiment of the present invention, the flexible ring circuit module also includes a power management module and a communication module, the power management module supplies power to each sensor, communication module and data processing module, and the communication module uses wireless communication to communicate between the data processing module and external devices.

[0068] Specifically, the power management module may include a power supply, a voltage stabilizing circuit, a noise filter, a protection circuit, etc. The data processing module may include a data acquisition module and a microcontroller, etc. The data acquisition module uses wireless connection technology to collect multi-source information of the bearing collected by each sensor and transmit it to the microcontroller for data processing and analysis. The communication module uses a wireless communication module, which is mainly responsible for data transmission and communication with external devices.

[0069] Optionally, in one embodiment of the present invention, the flexible annular circuit module is fixed to the bearing end cover, and the bearing end cover is fixedly mounted on the bearing seat 11. The bearing seat 11 is provided with a through hole to allow the oil and gas delivery pipeline 14 to pass through to provide lubricating oil to the bearing.

[0070] Optionally, in one embodiment of the present invention, the intelligent lubrication module includes an oil and gas delivery pipeline 14, a nozzle 15, an oil tank 16, a gear pump 17, an oil filter 18, an electromagnetic flowmeter 19, an air compressor 20, a gas filter 21, an air volume flowmeter 22, an oil and gas mixer 23, and an oil and gas distributor 24; the nozzle 15 is located above the level of the inner ring 4 of the rolling bearing, and the air enters the oil and gas mixer 23 from the air compressor 20 through the gas filter 21 and the air volume flowmeter 22, and the lubricating oil is supplied to the oil and gas mixer 23 from the gear pump 17 through the oil filter 18 and the electromagnetic flowmeter 19, and the oil and gas mixture is transported to the oil and gas distributor 24 through the oil and gas mixer 23. The nozzle 15 is located above the level of the inner ring of the rolling bearing, and the lubricating oil is quantitatively distributed and supplied to the oil and gas distributor 24 according to the bearing status diagnosis result, and finally the lubricating oil is delivered to the bearing lubrication point by the oil and gas delivery pipeline 14 and the nozzle 15.

[0071] like Figure 4 and Figure 5 As shown, various sensors collect multi-source signals of the bearing, transmit them to the data processing module for processing and analysis, judge the lubrication status of the bearing, and control the amount of lubricating oil according to the lubrication status. If it is in an oil-deficient state, the gear pump is controlled to increase the oil output; otherwise, the oil output is reduced.

[0072] Optionally, in one embodiment of the present invention, the data processing module is specifically used to preprocess the bearing multi-source information set collected by each sensor, perform feature extraction processing on the preprocessed bearing multi-source information set using the intrinsic time scale decomposition (ITD) method, and process the extracted features using a convolutional neural network to obtain a bearing status diagnosis result.

[0073] Specifically, the data acquisition module in the data processing module collects the multi-source sensor information set output by the bearing intelligent sensor system, performs data preprocessing, fills missing values, processes outliers, and removes noise for each sensor data, and then performs data alignment processing in sequence to ensure that different sensor data are aligned in time. The collected multi-source sensor information set is transmitted to the microcontroller, and the collected information is subjected to denoising and feature extraction processing using the intrinsic time scale decomposition method. By constructing a convolutional neural network model, the features are processed and fused, and training and optimization are performed to complete fault diagnosis.

[0074] A multi-source information fusion diagnosis algorithm based on ITD and neural network deep learning method is used to process and diagnose the collected bearing speed, load, temperature and other information to obtain the lubrication condition of the bearing. In one embodiment of the present invention, the pre-processed bearing multi-source information set is subjected to feature extraction processing using the intrinsic time scale decomposition method, including:

[0075] Define a baseline extraction operator L, so that the original signal X in the bearing multi-source information set t After a decomposition, it is as follows:

[0076] X t =LX t +(1-L)X t =L t +H t

[0077] Among them, LX t =L t is the baseline signal, (1-L)X t =H t The signal remaining after the baseline component is extracted from the original signal, namely the intrinsic rotation component PRC, which contains a baseline signal and an intrinsic rotation component after decomposition;

[0078] Assume that τ k (k=1,2,……) is X t The moment when the extreme point appears, define τ 0 =0, and record X k =X(τ k ), L k =L(τ k );

[0079] Assume that Lt , H t In [0,τ k ] is defined, X k In [0,τ k+2 ] is defined, then the value between any two adjacent maximum or minimum points [τ k ,τ k+2 ] Construct a segmented line segment, there must be τ in the interval k+1 The extreme point at the moment, calculate the baseline signal control point L k+1 :

[0080]

[0081] Among them, parameter α is the linear gain, and 0<α<1;

[0082] Between the several intervals generated by the baseline signal control points, each segment of the original signal is linearly transformed to obtain:

[0083]

[0084] Among them, L k t is the baseline signal segment obtained by linear transformation of the original signal in the kth interval segment;

[0085] Calculate the baseline signal segments of all interval segments in turn, and connect them in turn to obtain the complete baseline signal L t , the original signal X t After extracting a baseline signal, the remaining residual signal is the inherent rotation component. At this point, the first decomposition of the original signal is completed.

[0086] L t Continue to decompose as the new original signal, repeat the above decomposition steps until the baseline signal obtained by decomposition is a monotonic function or a constant or a function with less than three extreme points, which is called the trend component. Then the original signal X t The final decomposition of z is:

[0087] X t =L 1 t +H 1 t =(L 2 t +H 2 t )+H 1 t =…=H 1 t +H 2 t +…+H m t +L mt

[0088] Each signal of the bearing multi-source information set is decomposed by the intrinsic time scale decomposition method to obtain multiple intrinsic rotation components and a trend component, where each intrinsic rotation component and trend component is represented as a matrix, each row represents a time step, and each column represents a feature.

[0089] Optionally, in one embodiment of the present invention, Figure 6 As shown in the figure, the extracted features are processed using a convolutional neural network to obtain the bearing status diagnosis results, including:

[0090] The matrix obtained after decomposition by the intrinsic time scale decomposition method is used as input to the convolution layer of the convolutional neural network deep learning model. The convolution process is:

[0091]

[0092] Among them, z k l is the k-th output of the l-th layer, T is the number of convolutional layer kernels, φ is the activation function, W k l is the kth convolution kernel of the lth convolution layer, b k l is the corresponding bias, f k l-1 It is the feature information of the input feature map decomposed by the intrinsic time scale decomposition method;

[0093] Construct the maximum pooling layer, the calculation formula is:

[0094]

[0095] Among them, P j l+1 (i) is the corresponding value of the l+1th layer neuron in the pooling operation, T is the width of the pooling area, and a j l (t) is the activation value of the pooled area;

[0096] Adding a feature concatenation layer after the pooling layer can concatenate the feature maps of multiple intrinsic rotation components in the feature dimension to form a comprehensive feature representation and further complete high-level feature extraction. Assuming there are n feature maps, each with a size of (h, w, c), the feature concatenation representation is:

[0097] concatenated_features=concat(feature_map1,feature_map2,…,feature_mapn,axis=-1)

[0098] Among them: feature_map i represents the i-th feature map; concat represents the concatenation operation, and axis=-1 represents concatenation on the last feature dimension;

[0099] The size of the concatenated feature map is (h,w,n×c). The concatenated feature map is flattened and input into the fully connected layer for classification. The loss function J is set to optimize the model parameters and minimize the expectation of each sample. The convolutional neural network is trained using forward propagation:

[0100]

[0101] Among them, w and b are the weight and bias in the network, x is the input sample, y is the target output, and a l is the output of the lth layer of the model;

[0102] The trained convolutional neural network deep learning model is used to obtain the bearing status diagnosis results. If the bearing is in an oil-deficient state, the gear pump is controlled to increase the oil delivery; otherwise, the oil delivery is reduced.

[0103] According to the high-speed bearing intelligent condition monitoring system based on minimal lubrication proposed in an embodiment of the present invention, multi-information measurement points are applied to the bearing in view of the characteristics of weak bearing vibration signals and complex lubrication mechanisms. The system combines ITD with a deep learning convolutional neural network algorithm, integrates multiple signal types such as rotation speed, vibration, temperature, stress and strain, and obtains multi-source comprehensive diagnosis results. It can more quickly and fully reflect the operating and lubrication conditions of the bearing, realize real-time monitoring and diagnosis of the bearing operating status, and accurately supply lubricating oil to the bearing.

[0104] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0105] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying 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 the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

Claims

1. An intelligent condition monitoring system for high-speed bearings based on minimal lubrication, characterized in that: include: Bearing module, flexible ring circuit module, intelligent lubrication module; Wherein, the bearing module comprises a bearing inner ring and a bearing outer ring, and the bearing inner ring and the bearing outer ring are provided with a strain sensor and a temperature sensor; The flexible annular circuit module includes a vibration sensor, a rotation speed sensor, and a data processing module. The data processing module collects the multi-source information set of the bearing collected by each sensor and converts it into a digital signal for data processing to obtain the bearing status diagnosis result; The intelligent lubrication module quantitatively distributes and supplies lubricating oil according to the bearing status diagnosis result, and delivers the lubricating oil to the bearing lubrication point.

2. The system according to claim 1, characterized in that The bearing module also includes a bearing seat, bearing rolling elements and a bearing retainer. The bearing outer ring is fixed to the bearing seat. The bearing outer ring and the bearing inner ring are respectively provided with four evenly distributed grooves, and strain sensors and temperature sensors are cross-mounted in the grooves.

3. The system according to claim 1, characterized in that The flexible ring circuit module also includes a power management module and a communication module. The power management module supplies power to each sensor, the communication module and the data processing module. The communication module uses wireless communication to communicate between the data processing module and external devices.

4. The system according to claim 2, characterized in that The flexible annular circuit module is fixed to the bearing end cover, and the bearing end cover is fixedly installed on the bearing seat. The bearing seat is provided with a through hole so that the oil and gas transmission pipeline can pass through to provide lubricating oil for the bearing.

5. The system according to claim 1, characterized in that The intelligent lubrication module includes an oil and gas delivery pipeline, a nozzle, an oil tank, a gear pump, an oil filter, an electromagnetic flowmeter, an air compressor, a gas filter, an air volume flowmeter, an oil and gas mixer, and an oil and gas distributor; the nozzle is located above the level of the inner ring of the rolling bearing, the air enters the oil and gas mixer from the air compressor through the gas filter and the air volume flowmeter, the lubricating oil is supplied to the oil and gas mixer from the gear pump through the oil filter and the electromagnetic flowmeter, and the oil and gas mixture is delivered to the oil and gas distributor through the oil and gas mixer, the lubricating oil is quantitatively distributed and supplied to the oil and gas distributor according to the bearing status diagnosis result, and finally the lubricating oil is delivered to the bearing lubrication point through the oil and gas delivery pipeline and the nozzle.

6. The system according to claim 1, characterized in that The data processing module is specifically used to preprocess the bearing multi-source information set collected by each sensor, perform feature extraction processing on the preprocessed bearing multi-source information set using the intrinsic time scale decomposition method, and process the extracted features using a convolutional neural network to obtain a bearing status diagnosis result.

7. The system according to claim 6, characterized in that The preprocessed bearing multi-source information set is processed using the intrinsic time scale decomposition method for feature extraction, including: Define a baseline extraction operator L, so that the original signal X in the bearing multi-source information set t After a decomposition, it is as follows: X t =LX t +(1-L)X t =L t +H t Among them, LX t =L t is the baseline signal, (1-L)X t =H t It is the remaining signal after the baseline component is extracted from the original signal. After decomposition, it contains a baseline signal and an inherent rotation component. Assume that τ k (k=1,2,……) is X t When the extreme point appears, define τ0 = 0, and record X k =X(τ k ), L k =L(τ k ); Assume that L t , H t In [0,τ k ] is defined, X k In [0,τ k+2 ] is defined, then the value between any two adjacent maximum or minimum points [τ k ,τ k+2 ] Construct a segmented line segment, there must be τ in the interval k+1 The extreme point at the moment, calculate the baseline signal control point L k+1 : Among them, parameter α is the linear gain, and 0<α<1; Between the several intervals generated by the baseline signal control points, each segment of the original signal is linearly transformed to obtain: Among them, L k t is the baseline signal segment obtained by linear transformation of the original signal in the kth interval segment; Calculate the baseline signal segments of all interval segments in turn, and connect them in turn to obtain the complete baseline signal L t , the original signal X t After extracting a baseline signal, the remaining residual signal is the inherent rotation component. At this point, the first decomposition of the original signal is completed. L t Continue to decompose as the new original signal, repeat the above decomposition steps until the baseline signal obtained by decomposition is a monotonic function or a constant or a function with less than three extreme points, which is called the trend component. Then the original signal X t The final decomposition of z is: X t =L 1 t +H 1 t =(L 2 t +H 2 t )+H 1 t =…=H 1 t +H 2 t +…+H m t +L m t Each signal of the bearing multi-source information set is decomposed by the intrinsic time scale decomposition method to obtain multiple intrinsic rotation components and a trend component, where each intrinsic rotation component and trend component is represented as a matrix, each row represents a time step, and each column represents a feature.

8. The system according to claim 7, characterized in that The extracted features are processed using a convolutional neural network to obtain the bearing status diagnosis results, including: The matrix obtained after decomposition by the intrinsic time scale decomposition method is used as input to the convolution layer of the convolutional neural network deep learning model. The convolution process is: Among them, z k l is the k-th output of the l-th layer, T is the number of convolutional layer kernels, φ is the activation function, W k l is the kth convolution kernel of the lth convolution layer, b k l is the corresponding bias, f k l-1 It is the feature information of the input feature map decomposed by the intrinsic time scale decomposition method; Construct the maximum pooling layer, the calculation formula is: in, is the corresponding value of the l+1th layer neuron of the pooling operation, T is the width of the pooling area, and a j l (t) is the activation value of the pooled area; Adding a feature concatenation layer after the pooling layer can concatenate the feature maps of multiple intrinsic rotation components in the feature dimension to form a comprehensive feature representation and further complete high-level feature extraction. Assuming there are n feature maps, each with a size of (h, w, c), the feature concatenation representation is: concatenated_features=concat(feature_map1,feature_map2,…,feature_mapn,axis=-1) Among them: feature_map i represents the i-th feature map; concat represents the concatenation operation, and axis=-1 represents concatenation on the last feature dimension; The size of the concatenated feature map is (h,w,n×c). The concatenated feature map is flattened and input into the fully connected layer for classification. The loss function J is set to optimize the model parameters and minimize the expectation of each sample. The convolutional neural network is trained using forward propagation: Among them, w and b are the weight and bias in the network, x is the input sample, y is the target output, and a l is the output of the lth layer of the model; The trained convolutional neural network deep learning model is used to obtain the bearing status diagnosis results. If the bearing is in an oil-deficient state, the gear pump is controlled to increase the oil delivery; otherwise, the oil delivery is reduced.

9. The system according to claim 1, characterized in that In the bearing module, the temperature sensor adopts SHT20; the stress strain sensor adopts stress strain gauge.

10. The system according to claim 1, characterized in that In the flexible annular circuit module, the speed sensor adopts an eddy current speed sensor, and the measuring point center of the speed sensor and the bearing roller are located on the same horizontal line; the vibration sensor adopts the ADX355 type.

Citation Information

Patent Citations

  • ITD, spectral kurtosis, smooth iteration envelope analysis method of rolling bearing

    CN106168538A

  • Bearing device

    CN108397480A

  • Rolling bearing device and method of detecting abnormality of bearing portion

    CN109668731A

  • Rolling bearing fault diagnosis method based on multi-scale convolutional neural network

    CN113269221A

  • Intelligent bearing

    CN113483016A

Cited By

  • Mechanical seal lubricating method and lubricating system

    CN119848635A

  • Self-adaptive oil-gas lubrication control method and device for wind driven generator bearing

    CN121382563A