A combined bearing fault diagnosis method and system

Abstract

The present invention proposes a combined bearing fault diagnosis method and system. The combined bearing is mounted on a bearing test bench. The bearing test bench has a controller and a data acquisition element connected to the controller. The data acquisition element includes a temperature sensor for collecting the combined bearing's oil film temperature distribution, a pressure sensor for collecting the oil film pressure during combined bearing operation, a speed sensor for collecting spindle speed, m acoustic emission sensors for collecting acoustic emission signals generated by the combined bearing during operation, and n vibration sensors for collecting vibration signals generated by the combined bearing during operation. The controller is configured to receive and process the signals collected by each sensor. The combined bearing fault diagnosis method of the present invention utilizes four types of data: vibration data, acoustic emission data, oil film pressure, and oil film temperature, namely, two different types of dynamic response data and two types of lubrication characteristic data. This method obtains comprehensive system fault information, thereby enhancing fault diagnosis accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of combined bearing fault diagnosis, and in particular relates to a combined bearing fault diagnosis method and system. Background Art

[0002] The reliability and stability of gearboxes for high-power nuclear power circulating pumps are closely linked to the performance of nuclear power equipment. As a key component in these gearboxes, the operating status (healthy or faulty) of the combined bearing is a crucial indicator for determining the proper functioning of the nuclear power plant's main seawater circulation pumps. Therefore, the health of the combined bearings must be regularly monitored. However, for nuclear power equipment, shutting down for testing incurs high testing costs. Therefore, simulating the various operating states of the combined bearings in high-power nuclear power gearboxes using a test platform is a practical approach that can significantly reduce the costs of nuclear power equipment downtime.

[0003] CN201910947247.7 discloses a bearing fault detection method, which determines whether a bearing fault occurs by detecting the oil film resistance. This bearing fault detection method relies on single sensor data. When the mechanical system structure is relatively complex, single sensor data or a single type of data is difficult to provide comprehensive information about the system fault, making it difficult for the bearing fault detection method constructed based on these data to achieve high diagnostic accuracy. Summary of the Invention

[0004] The present invention aims to solve the technical problems existing in the prior art. The first purpose of the present invention is to provide a combined bearing fault diagnosis method. The second purpose of the present invention is to provide a combined bearing fault diagnosis system using the aforementioned combined bearing fault diagnosis method.

[0005] To achieve the first objective, the present invention adopts the following technical solution: a method for diagnosing a fault of a combined bearing, wherein the combined bearing is mounted on a bearing test bench, the bearing test bench has a controller and a data acquisition element connected to the controller, the data acquisition element including a temperature sensor for collecting the temperature distribution of the oil film of the combined bearing, a pressure sensor for collecting the oil film pressure during operation of the combined bearing, a speed sensor for collecting the spindle speed, m acoustic emission sensors for collecting acoustic emission signals generated by the combined bearing during operation, and n vibration sensors for collecting vibration signals generated by the combined bearing during operation, the controller being used to receive and process signals collected by each sensor; the fault diagnosis method comprising the following steps:

[0006] 1) After the operating condition of the bearing test bench reaches stability, obtain the signals VB1, VB2, ..., VB from n vibration sensors n , the signals from m acoustic emission sensors are AE1, AE2, ..., AEm , the signals collected by each vibration sensor and acoustic emission sensor are segmented continuously without overlap. After segmentation, the sample length of each vibration signal is i=1,2,…,n, the sample length of each acoustic emission signal is i=1,2,…,m;

[0007] 2) Calculate the power spectrum entropy of the vibration signal according to the following formula: i=1,2,…,n

[0008]

[0009]

[0010]

[0011] in, i=1,2,…,n is the power spectrum density of the vibration signal, j=1,2,…, is the component element of the power spectrum density of the vibration signal of the i-th vibration sensor, for The number of elements of is the total energy of the power spectrum density of the i-th vibration sensor, The ratio of the energy of each frequency point to the total energy;

[0012] 3) Perform Fourier transform on the vibration signal to obtain the frequency domain representation of the vibration signal, and take the amplitude corresponding to 0-1025 Hz, which is recorded as i=1,2,…,n, and perform maximum-minimum normalization processing;

[0013] 4) The acoustic emission signal is denoised by wavelet denoising, and the improved adaptive noise complete set empirical mode decomposition method is used to decompose the denoised acoustic emission signal, and the first five intrinsic modal components IMF1 to IMF5 are extracted. The energy entropy of the acoustic emission signal of the i-th acoustic emission sensor is calculated according to the following formula: i=1,2,…,n

[0014]

[0015]

[0016] in, is the energy of the jth IMF of the i-th acoustic emission sensor, is the ratio of each IMF energy to the total energy;

[0017] 5) Perform Fourier transform on the acoustic emission signal to obtain the frequency domain representation of the acoustic emission signal, and take the amplitude corresponding to 0-50kHz, which is recorded as i=1,2,…,m, and perform maximum-minimum normalization processing;

[0018] 6) Collect the oil film temperature T and oil film pressure P after the bearing test bench runs smoothly, and perform maximum-minimum normalization processing on the oil film temperature T and oil film pressure P respectively;

[0019] 7) The frequency domain amplitude of the vibration signal Input the feature extraction network, and assume that the output of the data from the i-th vibration sensor after passing through the feature extraction network is

[0020] 8) Obtain vibration characteristics F VB , vibration characteristics F VB It is the weighted fusion of the output of all vibration sensor data after passing through the feature extraction network;

[0021] 9) The frequency domain amplitude of the acoustic emission signal Input the feature extraction network, and assume that the output of the data from the i-th acoustic emission sensor after passing through the feature extraction network is

[0022] 10) Obtain acoustic emission characteristics F AE , acoustic emission characteristics F AE It is the weighted fusion of the output of all acoustic emission sensor data after passing through the feature extraction network;

[0023] 11) Feature fusion: Flatten the vibration feature F into a one-dimensional vector VB , acoustic emission characteristics F flattened into a one-dimensional vector AE , and the oil film temperature data T and the oil film pressure data P processed in step 6) are subjected to feature fusion to obtain the input feature F of the classifier network = [F VB ,F AE ,T,P];

[0024] 12) Fault diagnosis: Input the feature F in step 11) into the classifier network to obtain the fault diagnosis result.

[0025] The above technical solution, the combined bearing fault diagnosis method of the present invention utilizes four types of data, namely vibration data, acoustic emission data, oil film pressure and oil film temperature, namely two different dynamic response data and two lubrication characteristic data, to obtain comprehensive system fault information, so that the fault diagnosis accuracy is higher; and the number of vibration sensors and acoustic emission sensors of the present invention is multiple, and weighted fusion is used to obtain vibration feature information and acoustic emission feature information, thereby improving data utilization efficiency.

[0026] In a preferred embodiment of the present invention, in step 7), the feature extraction network of the vibration signal consists of three convolution blocks, all of which are composed of a one-dimensional convolutional neural network layer, a batch normalization layer, a Gaussian error linear unit activation layer and a one-dimensional maximum pooling layer stacked together, assuming that the output of convolution block 1 is c1, the output of convolution block 2 is c2, and the input of convolution block 3 is c1+c2; in step 9), the feature extraction network of the acoustic emission signal has the same structure as the feature extraction network of the vibration signal.

[0027] In a preferred embodiment of the present invention, in step 8), the vibration characteristic F VB Perform weighted fusion according to the following formula:

[0028]

[0029] in, is assigned to The weight of .

[0030] In a preferred embodiment of the present invention, in step 8), the following weight conversion function is used to obtain

[0031]

[0032] in, is the power spectrum entropy of the vibration signal from the i-th vibration sensor, mean VB is the mean value of the power spectrum entropy of the vibration signals of n vibration sensors.

[0033] The above technical solution uses a weight conversion function to achieve weighted fusion of information between n vibration sensors, which can effectively enhance the robustness of the fault diagnosis method and improve data utilization efficiency.

[0034] In another preferred embodiment of the present invention, in step 8), the vibration characteristic F is obtained. VB The specific process is to obtain Afterwards, and Perform the dot product and get and combine it with Add together to obtain the weighted fusion vibration feature F VB .

[0035] In another preferred embodiment of the present invention, in step 10), the acoustic emission feature F AE Perform weighted fusion according to the following formula:

[0036]

[0037] in, is assigned to The weight of .

[0038] In another preferred embodiment of the present invention, in step 10), the following weight conversion function is used to obtain

[0039]

[0040] in, is the energy entropy of the acoustic emission signal from the i-th acoustic emission sensor, mean AE is the mean energy entropy of the acoustic emission signals of m acoustic emission sensors.

[0041] The above technical solution uses a weight conversion function to achieve weighted fusion of information between m acoustic emission sensors, which can effectively enhance the robustness of the fault diagnosis method and improve data utilization efficiency.

[0042] In another preferred embodiment of the present invention, in step 10), the acoustic emission characteristics F are obtained. AE The specific process is to obtain Afterwards, and Perform the dot product and get and Add and obtain the weighted fusion acoustic emission feature F AE .

[0043] To achieve the above-mentioned second purpose, the present invention adopts the following technical solution: a combined bearing fault diagnosis system, including a data acquisition unit and a diagnosis unit, the data acquisition unit is connected to the vibration sensor and the acoustic emission sensor respectively, the data acquisition unit acquires signals from n vibration sensors and acquires signals from m acoustic emission sensors, the diagnosis unit receives the signals acquired by the data acquisition unit and uses the combined bearing fault diagnosis method provided in this application to perform diagnosis to obtain a combined bearing fault diagnosis result.

[0044] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments with reference to the following drawings, in which:

[0046] Figure 1 This is a flow chart of a combined bearing fault diagnosis method according to an embodiment of the present application.

[0047] Figure 2It is a flowchart of the feature extraction network in an embodiment of the present application. DETAILED DESCRIPTION

[0048] The following describes embodiments of the present invention in detail. Examples of the embodiments 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 only to explain the present invention and are not to be construed as limiting the present invention.

[0049] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "vertical", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0050] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0051] The present invention provides a method for diagnosing combined bearing faults. In the present invention, the combined bearing is the tested combination, which is mounted on a bearing test bench. The bearing test bench has a controller and a data acquisition element connected to the controller. The data acquisition element includes a temperature sensor for collecting the combined bearing oil film temperature distribution, a pressure sensor for collecting the oil film pressure during operation of the combined bearing, a speed sensor for collecting the spindle speed, m (e.g., m = 4) acoustic emission sensors for collecting acoustic emission signals generated by the combined bearing during operation, and n (e.g., n = 4) vibration sensors for collecting vibration signals generated by the combined bearing during operation. The controller is configured to receive and process the signals collected by each sensor. The specific structure and operating principle of the bearing test bench are prior art and will not be described in detail herein.

[0052] This invention utilizes a bearing test bench to conduct simulated tests of normal operating conditions and typical faults such as bearing wear, oil starvation, oil contamination, and rotor eccentricity for gearbox assembly bearings used in high-power nuclear power circulating pumps (see below for the detailed process). The signals collected by the data acquisition component include data from one normal operating condition and four fault conditions. Vibration data is collected from four vibration sensors, acoustic emission data from four acoustic emission sensors, and oil film temperature and pressure data from the temperature and pressure sensors after the bearing test bench has been operating smoothly.

[0053] like Figure 1 As shown, the fault diagnosis method of the present invention includes the following steps:

[0054] 1) Install the tested combined bearing on the bearing test bench for testing. After the operating conditions of the bearing test bench reach stability (speed stability and temperature no longer fluctuate widely), obtain the signals VB1, VB2, ..., VB from n vibration sensors. n , the signals from m acoustic emission sensors are AE1, AE2, ..., AE m , the signals collected by each vibration sensor and acoustic emission sensor are segmented continuously without overlap. After segmentation, the sample length of each vibration signal is i=1,2,…,n, the sample length of each acoustic emission signal is i=1,2,…,m.

[0055] 2) Calculate the power spectrum entropy of the vibration signal according to the following formula: i=1,2,…,n

[0056]

[0057]

[0058]

[0059] in, i=1,2,…,n is the power spectrum density of the vibration signal, j=1,2,…, is the component element of the power spectrum density of the vibration signal of the i-th vibration sensor, for The number of elements of is the total energy of the power spectrum density of the i-th vibration sensor, It is the ratio of the energy of each frequency point to the total energy.

[0060] 3) Perform Fourier transform on the vibration signal to obtain the frequency domain representation of the vibration signal, and take the amplitude corresponding to 0-1025 Hz, which is recorded as i=1, 2, ..., n, and perform maximum-minimum normalization processing on it. The specific maximum-minimum normalization processing is a common method in this field and will not be described in detail here.

[0061] 4) The AE signal is denoised by wavelet denoising, where the wavelet selected is Daubechies (5) wavelet; then the AE signal is decomposed by the improved adaptive noise complete set empirical mode decomposition method, and the first five intrinsic mode components IMF1 to IMF5 are extracted. The energy entropy of the AE signal of the ith AE sensor is calculated according to the following formula: i=1,2,…,n,

[0062]

[0063]

[0064] in, is the energy of the jth IMF of the i-th acoustic emission sensor, is the ratio of each IMF energy to the total energy.

[0065] 5) Perform Fourier transform on the acoustic emission signal to obtain the frequency domain representation of the acoustic emission signal, and take the amplitude corresponding to 0-50kHz, which is recorded as i=1,2,…,m, and perform maximum-minimum normalization processing on it.

[0066] 6) Collect the oil film temperature T and oil film pressure P after the bearing test bench runs smoothly, and perform maximum-minimum normalization processing on the oil film temperature T and oil film pressure P respectively to obtain and

[0067] 7) The frequency domain amplitude of the vibration signal Input feature extraction network, such as Figure 2 As shown in the figure, the feature extraction network of the vibration signal consists of three convolution blocks. All convolution blocks are composed of a one-dimensional convolutional neural network layer 1D-CNN, a batch normalization layer Batch Normalization, a Gaussian error linear unit activation layer CELU and a one-dimensional maximum pooling layer MaxPool stacked together. Let the output of convolution block 1 be c1, the output of convolution block 2 be c2, and the input of convolution block 3 be c1+c2; let the output of the data from the i-th vibration sensor after passing through the feature extraction network be

[0068] 8) Obtain vibration characteristics F VB , vibration characteristics F VBIt is the weighted fusion of the output of all vibration sensor data after passing through the feature extraction network. The closer the vibration signal is to white noise, the greater its power spectrum entropy is. Therefore, the power spectrum entropy can be used to analyze the vibration signals from different vibration sensors. Weighted, that is, the signal with larger power spectrum entropy corresponds to A smaller weight should be assigned to reduce its interference with the final diagnosis result. The signal with smaller power spectrum entropy corresponds to A larger weight should be assigned to increase its contribution to the final diagnosis result. VB Perform weighted fusion according to the following formula:

[0069]

[0070] in, is assigned to The weight is obtained by the following weight conversion function

[0071]

[0072] in, is the power spectrum entropy of the vibration signal from the i-th vibration sensor, mean VB is the mean value of the power spectrum entropy of the vibration signals of n vibration sensors, that is, The weight conversion function can be used to select the value of the power spectrum entropy that is smaller. Assigning a larger weight corresponds to a larger power spectrum entropy value Assign smaller weights. Since each time the weighted n They correspond to different Therefore, the mean is introduced into the weight conversion function VB Realize Dynamic adjustment.

[0073] In obtaining Afterwards, and Perform the dot product and get and combine it with Add together to obtain the weighted fusion vibration feature F VB .

[0074] 9) The frequency domain amplitude of the acoustic emission signal Input feature extraction network, refer to Figure 2 The feature extraction network of acoustic emission signals has the same structure as that of vibration signals. Suppose the output of the data from the i-th acoustic emission sensor after passing through the feature extraction network is

[0075] 10) Obtain acoustic emission characteristics FAE , acoustic emission characteristics F AE It is the weighted fusion of the output of all acoustic emission sensor data after passing through the feature extraction network. The acoustic emission feature weighted fusion is similar to the vibration feature weighted fusion step. It uses energy entropy to fusion the data from different acoustic emission sensors. Weighted, that is, the acoustic emission signal with larger energy entropy corresponds to A smaller weight should be assigned to reduce its interference with the final diagnosis result; the signal with smaller energy entropy corresponds to A larger weight should be assigned to increase its contribution to the final diagnosis result. AE Perform weighted fusion according to the following formula:

[0076]

[0077] in, is assigned to The weight is obtained by the following weight conversion function

[0078]

[0079] in, is the energy entropy of the acoustic emission signal from the i-th acoustic emission sensor, mean AE is the mean value of the energy entropy of the acoustic emission signals of m acoustic emission sensors, that is, The weight conversion function can be used to convert the corresponding energy entropy value to a smaller value. Assigning a larger weight corresponds to a larger energy entropy value Assign smaller weights. Since each time the weighted fusion is performed, They correspond to different Therefore, the mean is introduced into the weight conversion function AE Realize Dynamic adjustment.

[0080] In obtaining Afterwards, and Perform the dot product and get and Add and obtain the weighted fusion acoustic emission feature F AE .

[0081] 11) Feature fusion: Flatten the vibration feature F into a one-dimensional vector VB , acoustic emission characteristics F flattened into a one-dimensional vector AE , and the oil film temperature data T and the oil film pressure data P processed in step 6) are subjected to feature fusion to obtain the input feature F of the classifier network = [F VB ,FAE ,T,P].

[0082] 12) Fault diagnosis: Figure 1 As shown in Figure 1, the classifier network consists of a random dropout layer, two fully connected layers with Gaussian error linear activation functions, and a softmax classification layer. The feature F in step 11) is input into the classifier network to obtain the fault diagnosis result.

[0083] It should be noted that the classifier network model is trained based on the input dynamic response data and lubrication characteristic data (specifically, the dynamic response data and lubrication characteristic data of the combined bearing under normal operating conditions and typical fault conditions can be obtained through the simulation tests of normal operating conditions, simulation tests of combined bearing wear conditions, simulation tests of oil-deficient conditions, simulation tests of oil contamination conditions, and simulation tests of rotor eccentricity conditions below, so as to obtain the original sample data as training data), and the patterns of normal operating conditions and typical fault conditions are learned. After the classifier network model training is completed (the specific model training method is the existing technology and is not described in detail here), it can automatically determine whether the data represents a normal operating condition or a fault condition based on the input data feature F.

[0084] The combined bearings in gearboxes for high-power nuclear power circulating pumps exhibit different performance under normal operating conditions and typical fault conditions. These differences are reflected in the bearings' dynamic response data and lubrication characteristics. Therefore, we collected dynamic response and lubrication characteristic data for the combined bearings under normal operating conditions and typical fault conditions to obtain raw sample data for training.

[0085] 1. Simulation test of normal working conditions (healthy state)

[0086] A fault-free combined bearing (i.e., a healthy combined bearing) is installed on a bearing test bench for testing. After the operating conditions reach stability (the speed is stable and the temperature no longer fluctuates over a large range), a data acquisition component is used to collect various dynamic response data (including vibration data and acoustic emission data) and lubrication performance data (including oil film pressure data and oil film temperature data) of the combined bearing under this operating condition. Data collection is performed at least three times, with the same interval between each collection.

[0087] 2. Simulation test of combined bearing wear conditions

[0088] The artificially processed worn combined bearing is installed on the bearing test bench for testing. After the operating conditions reach stability, the dynamic response data and lubrication performance data of the combined bearing under the operating conditions are collected using a data acquisition component. The data collection is performed at least three times, with the same interval between each collection.

[0089] 3. Simulation test of oil-starved conditions

[0090] Install a fault-free combined bearing on a bearing test bench and conduct an oil-starved test. After the operating conditions reach stability, maintain this state for 15 minutes before controlling the oil pump's lubricating oil output flow (or shutting down the oil pump). Reduce the lubricating oil output flow to a certain level (or stop the oil supply) so that the residual lubricating oil inside the bearing is insufficient to form a complete dynamic oil film. While controlling the lubricating oil output flow (or shutting down the oil pump), use the test system to begin collecting various dynamic response data and lubrication characteristic data of the bearing under oil-starved conditions. Data collection is performed at least three times, with the same interval between each collection.

[0091] 4. Simulation test of oil contamination conditions

[0092] A fault-free combined bearing is installed on a bearing test bench for an oil starvation test. Before the test, a certain mass of iron powder particles (or other dust) is added to the lubricating oil used in the test. By adjusting the mass of the added iron powder particles (or other dust) and the amount of lubricating oil, a contaminated oil with a certain mass concentration is prepared. An ultrasonic bath is used to evenly distribute the particles within the contaminated oil. At the beginning of the test, the oil circulation and cooling device are first started. After the combined bearing is soaked with the contaminated oil, the test begins. After the operating conditions reach stability, a data acquisition component is used to collect various dynamic response data and lubrication performance data of the combined bearing under these operating conditions. Data collection is performed at least three times, with the same interval between each collection.

[0093] 5. Simulation test of rotor eccentricity

[0094] A combined bearing used for rotor eccentricity simulation testing was installed on a bearing test bench for testing. The three bearing pads on one side of the combined bearing were artificially machined to be taller than the three bearing pads on the other side. After the operating conditions reached stability, a data acquisition component was used to collect dynamic response data and lubrication performance data for the combined bearing under these operating conditions. Data collection was performed at least three times, with equal intervals between each collection.

[0095] Example 2

[0096] This embodiment provides a combined bearing fault diagnosis system, including a data acquisition unit and a diagnosis unit. The data acquisition unit is connected to a vibration sensor and an acoustic emission sensor, respectively. The data acquisition unit acquires signals from n (for example, four) vibration sensors and acquires signals from m (for example, four) acoustic emission sensors. The diagnosis unit receives the signals acquired by the data acquisition unit and performs diagnosis using the combined bearing fault diagnosis method of embodiment one to obtain a combined bearing fault diagnosis result.

[0097] The present invention measures multiple physical quantities: two different dynamic response data sets (vibration data and acoustic emission data) and two lubrication characteristic data sets (oil film pressure data and oil film temperature data). This diverse set of measurement data helps provide information about the fault from multiple perspectives, preventing fault diagnosis systems from misjudging the fault mechanism due to insufficient information. Furthermore, this rich information helps further understand and explain the failure mechanism of the combined bearing. Since vibration signals are the most commonly used dynamic response data for validating dynamic models and fault diagnosis, four vibration sensors are used to collect vibration signals of the combined bearing under normal conditions and under various typical fault conditions. Acoustic emission (AE) is the phenomenon in which elastic stress waves are generated by the rapid release of energy from a measured object due to structural changes. Therefore, AE signals have important applications in mechanical fault diagnosis. AE is dynamic and real-time, enabling continuous measurement and monitoring of combined bearings. Furthermore, AE signals exhibit excellent anti-interference and anti-noise capabilities, helping to improve the accuracy of fault diagnosis. Therefore, four AE signal sensors are used to collect AE response signals of the combined bearing under normal and fault conditions. The fault state of the combined bearing has a direct impact on the internal dynamic pressure oil film. Therefore, temperature sensors and pressure sensors are equipped to measure the changes in the internal oil film temperature and pressure of the bearing under normal conditions and various typical fault conditions.

[0098] Throughout this specification, reference to terms such as "preferred embodiment," "one embodiment," "some embodiments," "example," "specific example," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A combined bearing fault diagnosis method, characterized in that: The combined bearing is mounted on a bearing test bench, which has a controller and a data acquisition element connected to the controller. The data acquisition element includes a temperature sensor for collecting the temperature distribution of the oil film of the combined bearing, a pressure sensor for collecting the oil film pressure during operation of the combined bearing, a speed sensor for collecting the spindle speed, m acoustic emission sensors for collecting acoustic emission signals generated by the combined bearing during operation, and n vibration sensors for collecting vibration signals generated by the combined bearing during operation. The controller is used to receive and process the signals collected by each sensor. The fault diagnosis method comprises the following steps: 1) After the operating condition of the bearing test bench reaches stability, obtain the signals VB1, VB2, ..., VB from n vibration sensors n , the signals from m acoustic emission sensors are AE1, AE2, ..., AE m , the signals collected by each vibration sensor and acoustic emission sensor are segmented continuously without overlap. After segmentation, the sample length of each vibration signal is The sample length of each acoustic emission signal is 2) Calculate the power spectrum entropy of the vibration signal according to the following formula: in, is the power spectrum density of the vibration signal, is the component element of the power spectrum density of the vibration signal of the i-th vibration sensor, for The number of elements of is the total energy of the power spectrum density of the i-th vibration sensor, The ratio of the energy of each frequency point to the total energy; 3) Perform Fourier transform on the vibration signal to obtain the frequency domain representation of the vibration signal, and take the amplitude corresponding to 0-1025 Hz, which is recorded as And perform maximum-minimum normalization processing on it; 4) The acoustic emission signal is denoised by wavelet denoising, and the improved adaptive noise complete set empirical mode decomposition method is used to decompose the denoised acoustic emission signal, and the first five intrinsic modal components IMF1 to IMF5 are extracted. The energy entropy of the acoustic emission signal of the i-th acoustic emission sensor is calculated according to the following formula: in, is the energy of the jth IMF of the i-th acoustic emission sensor, is the ratio of each IMF energy to the total energy; 5) Perform Fourier transform on the acoustic emission signal to obtain the frequency domain representation of the acoustic emission signal, and take the amplitude corresponding to 0-50kHz, which is recorded as And perform maximum-minimum normalization processing on it; 6) Collect the oil film temperature T and oil film pressure P after the bearing test bench runs smoothly, and perform maximum-minimum normalization processing on the oil film temperature T and oil film pressure P respectively; 7) The frequency domain amplitude of the vibration signal Input the feature extraction network, and assume that the output of the data from the i-th vibration sensor after passing through the feature extraction network is 8) Obtain vibration characteristics F VB , vibration characteristics F VB It is the weighted fusion of the output of all vibration sensor data after passing through the feature extraction network; 9) The frequency domain amplitude of the acoustic emission signal Input the feature extraction network, and assume that the output of the data from the i-th acoustic emission sensor after passing through the feature extraction network is 10) Obtain acoustic emission characteristics F AE , acoustic emission characteristics F AE It is the weighted fusion of the output of all acoustic emission sensor data after passing through the feature extraction network; 11) Feature fusion: Flatten the vibration feature F into a one-dimensional vector VB , acoustic emission characteristics F flattened into a one-dimensional vector AE , and the oil film temperature data T and the oil film pressure data P processed in step 6) are subjected to feature fusion to obtain the input feature F of the classifier network = [F VB ,F AE ,T,P]; 12) Fault diagnosis: Input the feature F in step 11) into the classifier network to obtain the fault diagnosis result.

2. A combined bearing fault diagnosis method according to claim 1, characterized in that: In step 7), the feature extraction network of the vibration signal consists of three convolutional blocks. All convolutional blocks are composed of a one-dimensional convolutional neural network layer, a batch normalization layer, a Gaussian error linear unit activation layer, and a one-dimensional maximum pooling layer. Let the output of convolution block 1 be c1, the output of convolution block 2 be c2, and the input of convolution block 3 be c1+c2; In step 9), the feature extraction network of the acoustic emission signal and the feature extraction network of the vibration signal have the same structure.

3. A combined bearing fault diagnosis method according to claim 1, characterized in that: In step 12), the classifier network consists of a random dropout layer, two fully connected layers with Gaussian error linear activation function, and a softmax classification layer.

4. A combined bearing fault diagnosis method according to claim 1, characterized in that: In step 8), the vibration feature F VB Perform weighted fusion according to the following formula: in, is assigned to The weight of .

5. A combined bearing fault diagnosis method according to claim 4, characterized in that: In step 8), the following weight conversion function is used to obtain in, is the power spectrum entropy of the vibration signal from the i-th vibration sensor, mean VB is the mean value of the power spectrum entropy of the vibration signals of n vibration sensors.

6. A combined bearing fault diagnosis method according to claim 5, characterized in that: In step 8), obtain the vibration characteristics F VB The specific process is to obtain Afterwards, and Perform the dot product and get and combine it with Add together to obtain the weighted fusion vibration feature F VB .

7. A combined bearing fault diagnosis method according to any one of claims 1 to 6, characterized in that: In step 10), the acoustic emission feature F AE Perform weighted fusion according to the following formula: in, is assigned to The weight of .

8. A combined bearing fault diagnosis method according to claim 7, characterized in that: In step 10), the following weight conversion function is used to obtain in, is the energy entropy of the acoustic emission signal from the i-th acoustic emission sensor, mean AE is the mean energy entropy of the acoustic emission signals of m acoustic emission sensors.

9. A combined bearing fault diagnosis method according to claim 7, characterized in that: In step 10), the acoustic emission characteristics F are obtained AE The specific process is to obtain Afterwards, and Perform the dot product and get and Add and obtain the weighted fusion acoustic emission feature F AE .

10. A combined bearing fault diagnosis system, characterized in that: The invention comprises a data acquisition unit and a diagnosis unit, wherein the data acquisition unit is connected to the vibration sensor and the acoustic emission sensor respectively, the data acquisition unit acquires signals from n vibration sensors and acquires signals from m acoustic emission sensors, the diagnosis unit receives the signals acquired by the data acquisition unit and performs diagnosis using the method described in one of claims 1 to 9 to obtain a diagnosis result of the combined bearing fault.

Images