Working condition following gear wear monitoring and predicting method based on vibration signals

Abstract

The invention relates to a working condition following gear wear monitoring and predicting method based on a vibration signal, belongs to the technical field of gear transmission, and solves the problem that influence of different working conditions and different wear states of different positions of a gear cannot be considered in gear wear monitoring and prediction in the prior art. Comprising the following steps: S1, establishing a working condition Map graph based on dynamic working condition division of multi-source sensor fusion, and updating the working condition Map graph in real time; s2, establishing an Archard wear correction model, and calculating the wear extent of the gear; s3, establishing a gear nonlinear vibration model considering wear and clearance influence, and solving a vibration signal of a gear system; s4, a CNN-LSTM mixed wear diagnosis model is established, and training is carried out; and S5, acquiring and processing in real time to obtain a vibration signal of the gear system, and inputting the vibration signal into the CNN-LSTM mixed wear diagnosis model to obtain the current wear state probability distribution of the gear and the wear development trend prediction of the gear.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gear transmission, in particular to a working condition following gear wear monitoring and prediction method based on vibration signals. BACKGROUND

[0002] With the increasing requirements of power density and reliability of transmission systems in the fields of new energy vehicles, wind power, aerospace, etc., affected by load mutation, speed fluctuation, temperature change, etc., the wear monitoring of gears under service conditions faces severe challenges. The method of visual detection of gear wear condition after opening the box has high maintenance cost and long maintenance period, and the engineering field urgently needs a method that can efficiently and accurately monitor the wear of gears.

[0003] Traditional monitoring methods based on vibration signals (such as Fourier spectrum analysis) perform well in fixed working conditions, but in variable working conditions, due to the non-stationarity and nonlinearity of the signal, the meshing frequency and load of the gear show strong time-varying characteristics, which leads to the failure of traditional monitoring methods, specifically manifested as feature extraction failure, insufficient model generalization ability, and not considering the coupling effect of multiple physical fields, etc.

[0004] ISO 10816 standard uses vibration signal effective value (RMS) and peak factor as wear evaluation index, and decomposes the vibration signal into different frequency components by Fourier transform, wavelet transform, etc. The main features (such as meshing frequency sideband, energy entropy, etc.) are extracted for diagnosis. This method only relies on fixed diagnostic threshold, does not consider the nonlinear modulation of gear surface micro-wear on vibration signal, and cannot handle nonlinear vibration signal features under variable working conditions, resulting in deviation of vibration signal related feature extraction.

[0005] Gear damage identification models based on neural network (such as CNN, LSTM) self-learning can realize classification of different wear degrees through time and frequency domain features to train the classifier. However, this model relies on a large amount of labeled data, and the time synchronization and space-time alignment algorithm of multi-source data under variable working conditions has high complexity and poor real-time performance. The data distribution difference and threshold deviation lead to poor generalization ability. At the same time, this model does not establish the dynamic mapping relationship between wear and multiple physical fields, resulting in inaccurate diagnosis results.

[0006] Chinese invention patent application, publication number CN109871652A, invention name "a gear pair wear amount prediction method based on dynamic meshing force", discloses using gear rotor system dynamics model to calculate the dynamic meshing force and dynamic load distribution coefficient of gear pair, thereby predicting the wear amount of gear pair. However, this method does not consider the different wear states of different positions of the gear caused by different complex influencing factors under different working conditions, and it is difficult to realize accurate gear wear prediction.

[0007] Chinese invention patent application, publication number CN117634241A, invention name "a gear pair dynamic wear amount prediction method considering tooth surface micro-geometric morphology", discloses that the tooth surface wear amount and the wear amount are calculated by establishing a contact analysis model considering the tooth surface micro-geometric morphology, and a gear pair dynamic load calculation model based on the finite element method. However, this method still does not consider the different wear states of different positions of the gear caused by complex influencing factors under different working conditions, and the prediction result is not accurate enough.

[0008] Therefore, there is a need in the art for improved gear wear monitoring and prediction models to provide more accurate gear wear detection and prediction under complex working conditions. SUMMARY

[0009] The present invention proposes a working condition following gear wear monitoring and prediction method based on vibration signals, considering the influence of lubrication and temperature, based on the mapping relationship between vibration signals and wear degree, realizing dynamic wear prediction.

[0010] The working condition following gear wear monitoring and prediction method based on vibration signals provided by the embodiment of the present invention comprises: S1: dynamic working condition division based on multi-source sensor fusion, establishing a working condition Map graph, and real-time updating the working condition Map graph, the working condition Map graph comprising a plurality of working condition area grids; S2: establishing an Archard wear correction model based on thermal elastohydrodynamic lubrication theory, taking the working condition Map graph as input, and calculating the wear amount of the gear; S3: establishing a gear nonlinear vibration model considering the influence of wear and clearance, taking the working condition Map graph and the wear amount of the gear as input, and calculating the vibration signal of the gear system under different wear states; S4: establishing a CNN-LSTM hybrid wear diagnosis model and training to obtain a trained CNN-LSTM hybrid wear diagnosis model; S5: inputting the vibration signal of the gear system collected and processed in real time into the trained CNN-LSTM hybrid wear diagnosis model, obtaining the current wear state probability distribution of the gear and the wear development trend prediction data of the gear in a future time period, and providing the monitoring and prediction results to the gear system maintenance planning process.

[0011] Optionally, S1 specifically comprises: S11: taking the parameters of rotational speed and torque as clustering core variables, clustering the similar operating states into the same working condition area, and taking the parameter of oil temperature as clustering auxiliary variable, clustering the similar operating states into the same operating state, constructing a working condition Map graph comprising a plurality of working condition area grids, and taking the center point of each area grid as the working condition point of the area; S12: install sensors for real-time collection of gear signals, including installing speed sensors, torque sensors and temperature sensors, real-time collection of gear speed, input torque and oil temperature parameters; S13: obtain the collected signals through the installed sensors, use the improved DBSCAN algorithm, realize the dynamic identification of the area grid boundary of the working condition through density feasibility analysis, realize the automatic identification of real-time working condition data and the update of working condition Map.

[0012] Optionally, in S11: The working condition area divided by the clustering result of the speed and torque parameters as the clustering core variable includes low-speed light load, low-speed heavy load, high-speed light load and high-speed heavy load working condition area; The working condition area divided by the clustering result of the oil temperature parameter as the clustering auxiliary variable includes low temperature, normal temperature and high temperature working condition area; The two working condition areas are combined to obtain a working condition Map including multiple working condition area grids.

[0013] Optionally, S13 specifically includes: The area grid of the working condition Map is the basis, and the dynamically adjusted adaptive neighborhood radius is obtained according to the real-time collected oil temperature parameters of the gear; The improved DBSCAN algorithm is used to realize the dynamic identification of the area grid boundary of the working condition through density feasibility analysis; The minimum point number is adjusted in combination with the wear grade to obtain the dynamically adjusted neighborhood minimum point number; The working condition stay time threshold is set, and for the working condition points with a working condition stay time shorter than the working condition stay time threshold, they are regarded as abnormal working condition switching and are excluded to obtain the dynamically optimized working condition points; Based on the dynamically adjusted adaptive neighborhood radius and the dynamically optimized neighborhood minimum point number, and the dynamically optimized working condition points, the boundary of the area grid is translated and reconstructed to obtain the updated working condition Map.

[0014] Optionally, S2 specifically includes: S21: According to the working condition points determined in the working condition Map, solve the thermal elastohydrodynamic lubrication model to obtain the contact area oil film thickness, oil film pressure and oil film temperature parameters; S22: Combine the oil film thickness parameters to establish a nonlinear mapping of the friction coefficient and the film thickness ratio, realize real-time correction of the friction state, and obtain the real-time corrected film thickness ratio; S23: Combine the oil film pressure and oil film temperature parameters to establish a real-time mapping of the lubricating oil viscosity and the oil film pressure and oil film temperature to obtain the real-time corrected lubricating oil parameters; S24: taking the obtained film thickness ratio as a lubrication state judgment parameter, establishing an Archard wear correction model according to a judgment result, and obtaining a wear amount of the gear.

[0015] Optionally, S24 specifically includes: When the film thickness ratio is less than 0.1, it indicates that the lubrication condition is good, and the wear amount of the gear is calculated according to a classic Archard wear model:

[0016] wherein, V is a wear volume of the gear, is a wear coefficient of the gear, is a normal load of the gear, is a relative sliding distance, is a material hardness; When the film thickness ratio is greater than 0.1, it indicates that the gear is in a boundary lubrication state, and the wear amount of the gear is calculated by multiplying the classic Archard wear model by a correction coefficient: wherein, the correction coefficient

[0017] is obtained by the following formula:

[0018] wherein, f is a wear factor function related to the working condition, , is a wear factor function related to the working condition, indicates a current working condition.

[0019] Optionally, S3 specifically includes: S31: establishing a time-varying meshing stiffness model considering different wear depths along the tooth surface direction of the gear, taking the working condition Map and the wear amount of the gear as inputs, and being used to obtain the time-varying meshing stiffness of the gear; S32: establishing a mapping relationship between the wear amount of the gear and the static transmission error, and being used to obtain the static transmission error; S33: establishing a gear nonlinear dynamics model considering translation-torsion, taking the time-varying meshing stiffness of the gear obtained in S31 as an input, and introducing the static transmission error, and calculating the vibration signal of the gear system under different wear states.

[0020] Optionally, S4 specifically includes: S41: establishing a CNN, stacking convolution layers, including a first layer of convolution, a middle layer of convolution and a deep layer of convolution, obtaining and outputting a 128-dimensional feature vector compressed by global average pooling; ​​​S42: The 128-dimensional feature vector compressed by global average pooling is unfolded in time sequence to obtain an input tensor with a length of T; S43: A bidirectional LSTM architecture is established, each layer containing 128 memory cells, and the input tensor obtained in S42 is taken as input to obtain past and future input tensor time sequence information as LSTM output; S44: The LSTM output is mapped to 6 wear grades through a fully connected layer to output the wear state probability distribution of the gear, and a CNN-LSTM hybrid wear diagnosis model is established; S45: A cross-entropy loss function is used, an Adam optimizer is used, a training strategy is formulated, the vibration signal of the gear system obtained in S3 is classified to obtain training data, and the training data is input into the CNN-LSTM hybrid wear diagnosis model for training to obtain a trained CNN-LSTM hybrid wear diagnosis model.

[0021] Optionally, S5 specifically comprises: S51: Collect signals through the installed sensor and process to obtain the vibration signal of the gear system, and use ensemble empirical mode decomposition to perform noise reduction processing on the vibration signal to obtain a denoised vibration signal; S52: Convert the denoised vibration signal into a time-frequency image by using the improved local mean decomposition combined with the S transform, input the trained CNN-LSTM hybrid wear diagnosis model, obtain the wear state probability distribution of the gear, and obtain the wear development trend prediction data of the gear in a future time period, and draw a change curve of the future wear of the predicted gear.

[0022] Compared with the prior art, the working condition following gear wear monitoring and prediction method based on a vibration signal provided by the embodiment of the application has at least the following beneficial effects.

[0023] 1) By establishing a working condition Map and automatically identifying real-time working condition data and updating the working condition Map, the working condition recognition accuracy is improved, the problem of complex working condition switching noise is solved, and by introducing a working condition residence time threshold, short-term abnormal switching working conditions are eliminated, the Map is prevented from frequent oscillation, and the method is especially suitable for industrial equipment with frequent starts and stops.

[0024] 2) By establishing an Archard wear correction model based on the thermal elastohydrodynamic lubrication theory, the dynamic discrimination of the lubrication state is realized, and the wear amount is accurately quantified. Compared with the traditional Archard wear model which does not consider the influence of the lubrication state on the wear, the model distinguishes the lubrication state by the film thickness ratio in real time, according to different judgment results, respectively uses the classical Archard formula to calculate the wear amount, and introduces a correction coefficient to calculate the wear amount, and realizes the quantification of the wear aggravation effect under the boundary lubrication state. 3) Multi-physical field coupling improves calculation accuracy, solves oil film thickness, pressure and temperature parameters through thermal elastohydrodynamic lubrication model, establishes nonlinear mapping of friction coefficient and film thickness ratio, and combines real-time mapping correction of lubricating oil viscosity-pressure-temperature to correct wear coefficient, especially suitable for high-speed heavy load working conditions.

[0025] 4) A nonlinear vibration model of gears considering the influence of wear and clearance is established, the real meshing characteristics are restored through non-uniform wear modeling, a time-varying meshing stiffness model is established from the tooth root to the tooth top, and the stiffness calculation deviation caused by the difference in wear depth is solved through the coupling calculation of five parts such as bending stiffness and Hertz contact stiffness. The wear-clearance-vibration whole link mapping is considered, the long-period error, short-period error and transmission error caused by wear are fused through the static transmission error model, and the Runge-Kutta method is used to solve the nonlinear dynamic equation, which provides high-fidelity data support for subsequent diagnosis and wear prediction.

[0026] 5) Based on the CNN method, the wear impact features are extracted through convolution kernel stacking, the time-frequency features and time series trend are cooperatively captured. Based on the bidirectional LSTM method, the wear evolution trend is captured through the memory unit, the wear amount in the future 100,000 revolutions is predicted, and the wear is divided into multiple grades through the Softmax activation function, the multi-grade wear state output is realized, the wear distribution can be output in real time, which provides quantitative decision basis for gear box operation and maintenance. Compared with the traditional threshold diagnosis, the wear prediction can be advanced by 1-2 operation cycles.

[0027] 6) Through the whole link technical innovation of dynamic working condition division-accurate wear calculation-high-fidelity vibration modeling-intelligent diagnosis and prediction, the invention realizes the accurate judgment of the gear box wear fault, avoids repeated opening and checking, saves time and reduces cost, and can also warn the possible sudden failure caused by wear, thereby reducing the damage risk of the transmission equipment. Through the method, the safety of the transmission system can be effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. The features and advantages of the present application can be more clearly understood by referring to the drawings. The drawings are schematic and should not be understood as any limitation on the present application. For those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0029] Figure 1 It is a flowchart of a working condition following gear wear monitoring and prediction method based on vibration signals according to the embodiments of the present application.

[0030] Figure 2Flow chart of S1 in the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0031] Figure 3 Flow chart of S2 in the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0032] Figure 4 Flow chart of S3 in the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0033] Figure 5 Flow chart of S4 in the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0034] Figure 6 Flow chart of S5 in the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0035] Figure 7 Schematic diagram of the work condition map established in S1 of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0036] Figure 8 Flow chart of the work condition map updating in S1 of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0037] Fig. 9 is a diagram of a healthy gear vibration signal obtained in an embodiment of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0038] Figure 10 Fig. 10 is a diagram of a gear 1st wear vibration signal obtained in an embodiment of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0039] Figure 11 Fig. 11 is a diagram of a gear 2nd wear vibration signal obtained in an embodiment of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0040] Figure 12 Fig. 12 is a diagram of a gear 3rd wear vibration signal obtained in an embodiment of the method for monitoring and predicting the wear of a work condition following gear based on a vibration signal according to an embodiment of the present application.

[0041] Figure 13Gear 4 stage wear vibration signal graph obtained in an embodiment of the vibration signal based working condition following gear wear monitoring and prediction method provided according to the embodiment of the present application.

[0042] Figure 14 Gear 5 stage wear vibration signal graph obtained in an embodiment of the vibration signal based working condition following gear wear monitoring and prediction method provided according to the embodiment of the present application.

[0043] Figure 15 Prediction curve graph of wear amount and wear stage obtained in an embodiment of the vibration signal based working condition following gear wear monitoring and prediction method provided according to the embodiment of the present application.

[0044] Figure 16 Prediction time cloud graph of wear stage obtained in an embodiment of the vibration signal based working condition following gear wear monitoring and prediction method provided according to the embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to enable a more clear understanding of the above-mentioned objects, features and advantages of the present application, the present application will be further described below with reference to the drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0046] In the following description, a lot of specific details are set forth in order to facilitate a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and therefore, the protection scope of the present application is not limited by the specific embodiments disclosed below.

[0047] A vibration signal based working condition following gear wear monitoring and prediction method provided according to an embodiment of the present application is described in detail below with reference to the accompanying drawings.

[0048] As shown in Figures 1 to 6 A vibration signal based working condition following gear wear monitoring and prediction method provided according to an embodiment of the present application includes the following steps.

[0049] S1: dynamic working condition division based on multi-source sensor fusion, establishment of a working condition Map graph, and real-time working condition Map graph updating, the working condition Map graph including a plurality of working condition area grids divided, and the center point of each area grid serving as a working condition point representing the area. As shown in Figure 2 S1 specifically includes the following steps.

[0050] S11: Key operating parameters such as rotational speed, torque and oil temperature are taken as clustering core variables, and the clustering results will classify similar operating states into the same working condition area (such as low speed light load, low speed heavy load, high speed light load, high speed heavy load, etc.), and the oil temperature characteristics are taken as clustering auxiliary variables, and the clustering results will classify similar operating states into the same operating state (such as low temperature, normal temperature, high temperature, etc.), and a working condition Map diagram including multiple typical working condition area grids is constructed. The above similar operating states can be determined by the wear rate, which can be calculated by subsequent steps. With the dynamic changes such as aging of the running equipment and changes in the environment, the traditional fixed threshold cannot be updated in real time, which is easy to cause working condition misjudgment (such as when the gear is in the same operating speed and torque, the gear wear rate will also change when the ambient temperature changes), and the working condition Map diagram constructed by the above joint clustering method can overcome the lag of the traditional fixed threshold division, better determine the working condition boundary, and ensure the rationality of the working condition Map diagram division.

[0051] As shown in Figure 7 , as an example, the working condition Map diagram can be set as follows. The horizontal coordinate of the working condition Map diagram is the rotational speed, and the vertical coordinate is the torque. Each 2000 revolutions per minute (rpm) is divided into an interval, and there are 10 intervals from 0 to 20000. Each 50 N / m is divided into an interval, and there are 10 intervals from 0 to 500. The intervals obtained by division are combined, and different colors (represented by the darkness of the color) Figure 7 represent different oil temperatures, and 16 area grids A~L are obtained, and the wear rates in each area grid are similar. The center point of the area grid is taken as the working condition point representing the area to calculate.

[0052] S12: Install sensors to collect signals of the gear in real time, which can include installing rotational speed sensors, torque sensors and temperature sensors to collect parameters of the rotational speed, input torque and oil temperature of the gear in real time. By installing multiple sensors, the traditional single parameter working condition division method can be broken through, and the problem of insufficient model generalization ability under variable working conditions can be solved. The real-time collected parameters of the rotational speed, input torque and oil temperature of the gear can be used for the update of the working condition Map diagram.

[0053] S13: The collected signals are obtained by the installed sensors, and the improved DBSCAN algorithm is adopted to realize dynamic identification of the area grid boundary of the working condition through density feasibility analysis, realize automatic identification of real-time working condition data and update of the working condition Map diagram. Referring to Figure 8 , the specific process of this step is as follows.

[0054] Specifically, based on the area grid of the working condition Map diagram, the dynamically adjusted adaptive neighborhood radius is represented as:

[0055] wherein, is a radius of a region grid unit, is an oil temperature sensitivity coefficient, is a deviation of current oil temperature from historical average oil temperature, is a time. Wherein, the region grid unit is a region grid of the working condition Map graph divided by S11; the current oil temperature is the oil temperature of the gear collected by the temperature sensor in real time.

[0056] By using the improved DBSCAN algorithm, through Adjust the minimum point number according to the wear grade to realize dynamic optimization of the minimum point number of the neighborhood:

[0057] wherein, is a basic point number, is a wear sensitivity coefficient, is a corresponding wear grade.

[0058] When the boundary points of the region grid of a certain working condition are continuously broken through by the new data points for a set number of times threshold (for example, N times), the boundary translation of the region grid of the working condition is triggered and reconstruction is carried out. The judgment method of whether the boundary points of the region grid are broken through by the new data points is that if the new data points exceed the boundary points of the region grid, it is judged that the boundary points of the region grid are broken through by the new data points. After the boundary reconstruction of the region grid of the working condition Map graph, the optimization parameter can also be fed back to the real-time data acquisition step through the density feasibility analysis.

[0059] In addition, for complex working conditions, a working condition stay time threshold can be set, and if the working condition stay time is shorter than the working condition stay time threshold, it is considered as abnormal working condition switching, and the working condition (working condition point) is excluded and not calculated, so as to solve the problem of repeated switching of the working condition Map graph under complex working conditions. According to the needs, the above working condition stay time threshold can be set to, for example but not limited to, 5s, 10s, 12s, etc.

[0060] Based on the above dynamically adjusted adaptive neighborhood radius and minimum point number, and the dynamic optimization of the working condition point, the working condition dynamic reconstruction is realized, and then the update of the working condition Map graph is realized.

[0061] Thus, the real-time updated working condition Map graph is obtained and used as the input of the Archard wear correction model based on the thermal elastohydrodynamic lubrication (EHL) theory and the gear nonlinear vibration model considering the influence of wear and gap.

[0062] S2: Establish an Archard wear correction model based on the thermal elastohydrodynamic lubrication (EHL) theory, and use the working condition Map graph as the input to calculate the wear amount of the gear. For example, Figure 3As shown, S2 specifically includes the following steps.

[0063] S21: Based on the operating points determined in the operating condition map, solve the thermo-elastohydrodynamic (TEHL) model to obtain the oil film thickness, oil film pressure and oil film temperature parameters in the contact area.

[0064] S22: Based on the oil film thickness parameter, establish the friction coefficient versus film thickness ratio. The nonlinear mapping enables real-time correction of the friction state, resulting in a real-time corrected film thickness ratio. (Film thickness ratio) It can be obtained from the oil film thickness and the coefficient of friction.

[0065] S23: By combining the parameters of oil film pressure and oil film temperature, a real-time mapping between lubricating oil viscosity and oil film pressure and oil film temperature is established to realize the real-time correction of lubricating oil parameters and obtain the real-time corrected lubricating oil parameters.

[0066] S24: The obtained film thickness ratio As a parameter for judging lubrication status, an Archard wear correction model is established based on the judgment result to obtain the wear amount of the gear.

[0067] Specifically, when the film thickness ratio When the lubrication conditions are good, the wear of the gear is calculated according to the classic Archard wear model:

[0068] in, This represents the wear volume of the gear. The wear coefficient of the gear. For the normal load on the gear, This is the relative sliding distance. The material hardness of the gear. The wear coefficient of the gear. The wear coefficient of the gear can be obtained based on the friction coefficient and the lubricating oil parameters obtained from S23, which are related to the friction coefficient and the lubricating oil parameters.

[0069] When film thickness ratio When the condition is defined as boundary lubrication, the lubrication effect is poor. The wear of the gear is calculated by multiplying the classic Archard wear model by a correction factor.

[0070] The correction factor can be obtained from the following formula:

[0071] in, , The wear factor function, which is related to operating conditions, can be calibrated experimentally. The current working condition state is represented.

[0072] The above obtains an Archard wear correction model based on thermal elastohydrodynamic lubrication (EHL) theory.

[0073] S3: A nonlinear vibration model of the gear considering the influence of wear and gap is established, and the working condition map and the wear amount of the gear are used as inputs to obtain the vibration signal of the gear system under different wear states. As shown in FIG. 8, S3 specifically includes the following steps. Figure 4

[0074] S31: A time-varying meshing stiffness model considering different wear depths along the tooth surface direction of the gear is established, and the working condition map and the wear amount of the gear are used as inputs to obtain the time-varying meshing stiffness of the gear.

[0075] The wear amount at each position along the tooth root to the tooth top of the gear is different, and the time-varying meshing stiffness of the gear under different wear degrees at each position of the gear is obtained by calculating the bending stiffness, Hertz contact stiffness, compression stiffness, shear stiffness, and matrix deformation stiffness under the corresponding wear amount at each position of the gear. The time-varying meshing stiffness model established in the above manner solves the problem that the traditional model assumes uniform wear, resulting in inaccurate results.

[0076] S32: A mapping relationship between the wear amount of the gear and the static transmission error is established to obtain the static transmission error. Gear wear will cause changes in the gear meshing gap and changes in the static transmission error. Along the tooth root to the tooth top, a mapping between different wear depths and the static transmission error is established to obtain the static transmission error variation function. The calculation process of the static transmission error (USTE) is as follows.

[0077] Establish the long-period error:

[0078] In the formula, is the driving gear rotation frequency, is the tangential comprehensive total deviation, is the time. The tangential comprehensive total deviation is the sum of the tangential deviations of all gears in the gear system.

[0079] Establish the short-period error:

[0080] In the formula, is the meshing frequency, is the tangential comprehensive deviation of a single gear.

[0081] Establish the random error: ​

[0082] wherein, is a random number uniformly distributed between [0, 1].

[0083] The transmission error caused by gear wear is established as and the driven gear wear amount Projecting in the direction of the meshing line:

[0084] wherein, denotes the pressure angle of the gear.

[0085] Based on the long-period error , the short-period error , the random error and the transmission error caused by gear wear , the static transmission error (USTE) is obtained as .

[0086] The static transmission error variation function is thus established, and based on the static transmission error variation function, the static transmission error variation can be obtained.

[0087] S33: A gear nonlinear dynamics model considering translation-torsion is established, taking the time-varying meshing stiffness of the gear obtained in S31 as input, and introducing the static transmission error obtained from the static transmission error variation function, the differential equations of the gear nonlinear dynamics model are solved by the Runge-Kutta method to obtain the vibration signal of the gear system under different wear states.

[0088] The Runge-Kutta method for solving the differential equations of the gear nonlinear dynamics model is as follows:

[0089] wherein, is the mass matrix, is the damping matrix, is the stiffness matrix, is the generalized coordinate, is the gap function, is the generalized force vector matrix.

[0090] wherein, the above stiffness matrix items are obtained based on the time-varying meshing stiffness of the gear in S31, and the above generalized coordinate items are obtained based on the static transmission error and the meshing line direction displacement in S32, and the second derivative of the meshing line direction displacement (generalized coordinate) is obtained, i.e. the vibration acceleration.

[0091] S4: a CNN (Convolutional Neural Network)-LSTM (Long Short-Term Memory) hybrid wear diagnosis model is established to obtain the wear state probability distribution of the gear system, and the CNN-LSTM hybrid wear diagnosis model is trained based on the vibration signal obtained in S3 to obtain a trained CNN-LSTM hybrid wear diagnosis model. As shown in FIG. 8, S4 specifically includes the following steps. Figure 5

[0092] S41: a CNN (Convolutional Neural Network) is established, and a convolutional layer stack is performed, including a first layer convolution, a middle layer convolution, and a deep layer convolution, to obtain and output a 128-dimensional feature vector compressed by global average pooling, wherein the key information in the time-frequency domain is retained. The specific process of establishing the CNN by the convolutional layer stack is as follows.

[0093] The first layer convolution uses a 3x3 convolution kernel (step 1, padding 1) to extract the weak impact vibration signal in the early stage of wear, sets the number of convolution kernels to 64, and uses ReLU as the activation function. The first layer convolution can extract the vibration signal of the gear system obtained in S3.

[0094] The middle layer convolution uses a 5x5 convolution kernel (step 2, padding 2) to capture more macroscopic structural features, such as changes in harmonic distribution caused by wear intensification, and the number of convolution kernels is increased to 128, combined with Batch Normalization (Batch Normalization) to accelerate convergence. The input of the middle layer convolution is the output of the first layer convolution.

[0095] The deep layer convolution uses a 7x7 convolution kernel (step 2, padding 3) to extract global features, and the number of convolution kernels is 256. The dimension is reduced by maximum pooling (pooling kernel 2x2), and the convolution layer outputs a 128-dimensional feature vector compressed by global average pooling, which retains the key information in the time-frequency domain and avoids the overfitting risk brought by the fully connected layer. The input of the deep layer convolution is the output of the middle layer convolution.

[0096] S42: The 128-dimensional feature vector compressed by global average pooling extracted by the CNN is unfolded in time sequence to form an input tensor with a shape of [time step, feature dimension] with a length of T.

[0097] S43: a bidirectional LSTM architecture is established, each layer containing 128 memory cells, and the input tensor obtained in S42 is used as input to obtain past and future input tensor time sequence information as LSTM output. Through the LSTM architecture, the wear evolution trend of the gear under different working conditions can be captured, the past and future input tensor time sequence information can be obtained, and the dropout layer (dropout rate = 0.2) is used to prevent overfitting.

[0098] ​S44: Map the LSTM output to 6 wear grades (e.g., including 0-5 grades) through a fully connected layer, the activation function adopts Softmax, and output the wear state probability distribution of the gear. The wear state probability distribution reflects the current wear condition of the gear. Among them, the 6 wear grades are as follows: 0 grade represents that the gear is in normal and non-wear state; 1-2 grades correspond to initial wear, which is manifested as slight scratches or slight material loss on the tooth surface; 3-4 grades are medium wear, and the tooth profile shape changes significantly; and 5 grade represents severe wear, which may cause tooth surface spalling, tooth breakage and other major failure risks. Through the above explicit grade division, the operation and maintenance personnel can quickly judge the wear degree of the gear. Among them, the wear state probability distribution of the gear system can be obtained based on the wear state probability distribution of a single gear.

[0099] Thus, the CNN-LSTM hybrid wear diagnosis model is obtained, which is used to obtain the wear state probability distribution of the gear system based on the vibration signal of the gear system.

[0100] S45: Adopting cross-entropy loss function, using Adam optimizer, formulating training strategy, classifying the vibration signal of the gear system obtained from the gear nonlinear dynamics model established in S3, and inputting the vibration signal as training data into the established CNN-LSTM hybrid wear diagnosis model for training, to obtain the trained CNN-LSTM hybrid wear diagnosis model.

[0101] S5: Input the vibration signal of the gear system obtained by real-time acquisition and processing into the trained CNN-LSTM hybrid wear diagnosis model, to obtain the current wear state probability distribution of the gear and the wear development trend prediction data of the gear in a future time period, which are used to provide to the gear system maintenance planning process as the basis for the operation and maintenance personnel to formulate maintenance strategy. The vibration signal of the gear system is obtained by executing the above S1 to S3. As shown in Figure 6 S5 specifically includes the following steps.

[0102] S51: Acquire information through the installed sensor and obtain the vibration signal of the gear system, and adopt ensemble empirical mode decomposition (EEMD) for noise reduction processing to obtain the noise-reduced vibration signal. The noise reduction processing includes adaptively decomposing the vibration signal into multiple intrinsic mode functions (IMF), eliminating noise-dominant IMF components, and suppressing the interference of background noise on gear meshing features and wear impact features.

[0103] S52: The improved local mean decomposition (LMD) combined with S transform (LMD-S) is used to convert the denoised vibration signal into a time-frequency image. Furthermore, the converted time-frequency image is taken as the input of the trained CNN-LSTM hybrid wear diagnosis model to obtain the wear state probability distribution of the gear. In addition, the LSTM is combined to capture the long-term dependence of the time sequence characteristics, and the wear development trend prediction data of the gear in the future time period is output. Moreover, the future wear amount change curve of the predicted gear is drawn, which reflects the wear amount change of the gear in the future, for example but not limited to, one running cycle (100,000 revolutions as one running cycle). Thus, the maintenance strategy can be planned in advance to avoid sudden failure. The LMD decomposes the multi-component signal adaptively, and the S transform maintains the phase consistency in the time-frequency domain.

[0104] Embodiment 1 In order to better understand the present application, the following describes an embodiment of a working condition following gear wear monitoring and prediction method based on a vibration signal provided by an embodiment according to the present application. In this embodiment, a group of gear systems are monitored and predicted for wear.

[0105] Figures 9 to 14 For this embodiment, after the real-time sensor information of a gear system is collected and processed by the method provided in the above embodiments, the vibration signals of the gear in six wear levels from normal to severe wear are obtained, Figure 15 the prediction curve of the wear amount and the wear level with the running cycle, Figure 16 the prediction time cloud map of the wear level with the running cycle.

[0106] As can be seen from the figure, Figures 9 to 14 It can be seen that the vibration peak reflects the change of the wear degree, and the time domain vibration signal peak gradually increases with the increase of the wear level from health to 5-level wear, which intuitively verifies the positive correlation between wear and vibration response, and reflects the time domain feature capturing ability of the model for different wear states.

[0107] As can be seen from the figure, Figure 15 It can be seen that the model can realize the time sequence fitting of the wear prediction curve, and the prediction curve of the wear amount with the running cycle shows the quantitative prediction ability of the model for the wear development trend, which provides a visual reference for the future wear growth of the gear and supports the early formulation of the maintenance strategy.

[0108] As can be seen from the figure, Figure 16 The visualization of the wear state time cloud map is realized, the wear level distribution of different running cycles is intuitively presented through the color gradient, and the evolution process of the wear from the initial stage to the severe stage is clearly shown, which facilitates the quick identification of the current wear stage and the future trend and improves the intuitiveness of the operation and maintenance decision.

[0109] All the optional technical solutions above can be combined to form optional embodiments of the present application, and will not be described one by one here.

[0110] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0111] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A vibration signal based operating condition following gear wear monitoring and prediction method, characterized in that, Comprise: S1: dynamic working condition division based on multi-source sensor fusion, establish working condition Map figure, and real-time working condition Map figure update, working condition Map figure includes multiple working condition area grid; S2: establish Archard wear correction model based on thermal elastohydrodynamic lubrication theory, take working condition Map figure as input, calculate gear wear; S3: establish gear nonlinear vibration model considering wear and gap influence, take working condition Map figure and gear wear as input, solve gear system vibration signal under different wear state; S4: establish CNN-LSTM hybrid wear diagnosis model, and train, obtain trained CNN-LSTM hybrid wear diagnosis model; S5: the vibration signal of gear system collected and processed in real time is input into the trained CNN-LSTM hybrid wear diagnosis model, the current wear state probability distribution of gear and the wear development trend prediction data of gear in a future time period are obtained as monitoring and prediction results and are provided to gear system maintenance planning process.

2. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 1, wherein, S1 specifically comprises: S11: take the parameters of rotating speed and torque as clustering core variables, and the clustering results will be the same running state into the same working condition area, and take the parameter of oil temperature as clustering auxiliary variable, and the clustering results will be the same running state into the same running state, construct working condition Map figure including multiple working condition area grid, and the center point of each area grid is taken as the working condition point of the area; S12: install sensors for real-time acquisition of gear signals, including installation of rotating speed sensor, torque sensor and temperature sensor, real-time acquisition of rotating speed, input torque and oil temperature parameters of gear; S13: the collected signals are obtained through the installed sensors, the improved DBSCAN algorithm is adopted, the dynamic identification of working condition area grid boundary is realized through density feasibility analysis, and the automatic identification of real-time working condition data and working condition Map figure update are realized.

3. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 2, wherein, S13 specifically comprises: The area grid of working condition Map figure is used as the basis, the dynamically adjusted adaptive neighborhood radius is obtained according to the real-time collected oil temperature parameters of gear; The improved DBSCAN algorithm is adopted to obtain the minimum point number of the neighborhood by The minimum point number of the neighborhood is dynamically adjusted by combining the wear grade adjustment. Set working condition residence time threshold, for the working condition points with working condition residence time less than working condition residence time threshold, regard as abnormal working condition switching, eliminate the working condition points, and obtain dynamically optimized working condition points; Based on dynamically adjusted adaptive neighborhood radius, dynamically optimized neighborhood minimum point number and dynamically optimized working condition points, the boundary translation and reconstruction of area grid are carried out, and the updated working condition Map figure is obtained.

4. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 2, wherein, S2 specifically comprises: S21: according to the working condition points determined in working condition Map figure, solve the thermal elastohydrodynamic lubrication model, obtain the parameters of oil film thickness, oil film pressure and oil film temperature; S22: combine the parameters of oil film thickness, establish the nonlinear mapping of friction coefficient and film thickness ratio, realize the real-time correction of friction state, and obtain the real-time corrected film thickness ratio; S23: combine the parameters of oil film pressure and oil film temperature, establish the real-time mapping of lubricating oil viscosity, oil film pressure and oil film temperature, and obtain the real-time corrected lubricating oil parameters; S24: Taking the obtained film thickness ratio as a lubrication state judgment parameter, an Archard wear correction model is established according to a judgment result, and a wear amount of the gear is obtained.

5. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 4, wherein, S24 specifically comprises: When the film thickness ratio is greater than 3, it indicates a good lubrication condition and the wear of the gear is calculated according to the classical Archard wear model: wherein, is the wear volume of the gear, is the wear coefficient of the gear, is the normal load of the gear, is the relative sliding distance, is the material hardness; When the film thickness ratio When the film thickness ratio When the film thickness ratio When the film thickness ratio When the film thickness ratio When the film thickness ratio When the film thickness ratio When the film thickness ratio When the wherein the correction factor is obtained from the equation: wherein , is a wear factor function related to the operating condition, denotes the current operating condition.

6. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 1, wherein, S3 specifically comprises: S31: A time-varying meshing stiffness model considering different wear depths along the tooth surface direction of the gear is established, and the working condition Map and the wear amount of the gear are taken as inputs to obtain the time-varying meshing stiffness of the gear; S32: A mapping relationship between the wear amount of the gear and the static transmission error is established to obtain the static transmission error; S33: A gear nonlinear dynamics model considering translation-torsion is established, the time-varying meshing stiffness of the gear obtained in S31 is taken as an input, and the static transmission error is introduced to obtain the vibration signal of the gear system under different wear states.

7. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 1, wherein, S4 specifically comprises: S41: A CNN is established, and a convolution layer stack is performed, including a first layer convolution, a middle layer convolution and a deep layer convolution, to obtain and output a 128-dimensional feature vector compressed by global average pooling; S42: The 128-dimensional feature vector compressed by global average pooling is unfolded in time sequence to obtain a sequence with a length of T to obtain an input tensor; S43: A bidirectional LSTM architecture is established, each layer contains 128 memory cells, the input tensor obtained in S42 is taken as an input, and past and future input tensor time sequence information is obtained as an LSTM output; S44: The LSTM output is mapped to 6 wear grades through a full connection layer, and a wear state probability distribution of the gear is output to establish a CNN-LSTM hybrid wear diagnosis model; S45: A cross-entropy loss function is adopted, an Adam optimizer is used, a training strategy is formulated, the vibration signal of the gear system obtained in S3 is classified to obtain training data, and the training data is input into the CNN-LSTM hybrid wear diagnosis model for training to obtain a trained CNN-LSTM hybrid wear diagnosis model.

8. The vibration signal based operating condition following gear wear monitoring and prediction method of claim 2, wherein, S5 specifically comprises: S51: The vibration signal of the gear system is obtained by collecting and processing signals through the installed sensor, and the ensemble empirical mode decomposition is adopted to perform noise reduction processing on the vibration signal to obtain a denoised vibration signal; S52: The denoised vibration signal is converted into a time-frequency image by using the improved local mean decomposition combined with the S transform, and is input into the trained CNN-LSTM hybrid wear diagnosis model to obtain the wear state probability distribution of the gear and the wear development trend prediction data of the gear in a future time period, and a change curve of the future wear amount of the predicted gear is drawn.

Images