Real-time lubrication control method and device for gear transmission system

Abstract

The invention provides a real-time lubrication control method and device for a gear transmission system. The method comprises the steps that multi-dimensional collection is conducted on operation state parameters of a gear pair, and a working condition state data set is generated; determining shear rate distribution data and oil film thickness distribution data of the lubrication contact area according to the working condition state data set; inputting the obtained data into a neural network model, and performing joint prediction on the oil film rigidity and the tooth surface friction coefficient of the lubrication contact area to generate a joint prediction result; according to the joint prediction result, whether the gear pair lubrication strategy meets the safe operation requirement or not is judged, and a corresponding lubrication level is generated; and performing differential treatment on the oil film rigidity and the tooth surface friction coefficient according to the lubrication grade, determining a lubrication offset parameter, and performing self-adaptive adjustment on the lubrication control parameter according to the lubrication offset parameter. According to the scheme, the lubrication parameters are adjusted in time according to the joint prediction result, and precise adjustment and optimization of the gear lubrication state under the high-load and high-speed working condition are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transmission gears, in particular to a real-time lubrication control method and device for a gear transmission system. BACKGROUND

[0002] In a high-load and high-speed precision gear transmission system, such as a wind power speed increaser, an aviation power gear box or a new energy vehicle electric drive system, the lubrication state has an important influence on transmission efficiency, temperature rise control and service life. Especially under complex working conditions, the tooth surface contact conditions change frequently, and the lubricating oil film is prone to instantaneous degradation or local rupture, resulting in an increase in friction coefficient and a decrease in oil film stiffness, which in turn causes abnormal vibration, accelerated wear and even gear failure.

[0003] To ensure the long-term reliable operation of the gear pair, current industrial systems rely on external lubrication parameters such as oil temperature, pressure and flow rate for state monitoring, and adjust the lubrication supply through experience strategy. However, this method is usually based on static control logic and experience threshold, and cannot reflect the dynamic evolution process of the lubrication state in the tooth surface contact area, resulting in response lag and insufficient control accuracy. SUMMARY

[0004] The present application provides a real-time lubrication control method and device for a gear transmission system, which solves the problem of insufficient lubrication control accuracy of the gear pair in related technologies.

[0005] The first aspect of the present application provides a real-time lubrication control method for a transmission gear, which comprises: According to the multi-dimensional sensors of the gear transmission system, the operating state parameters of the gear pair are collected in multiple dimensions to generate a working condition state data set; According to the working condition state data set, the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area are determined; The shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model to jointly predict the oil film stiffness and the tooth surface friction coefficient of the lubrication contact area, and generate a joint prediction result; According to the joint prediction result, it is judged whether the current gear pair lubrication strategy meets the safe operation requirement, and the corresponding lubrication level is generated according to the judgment result; According to the lubrication level, the oil film stiffness and the tooth surface friction coefficient are differentially processed to determine the lubrication offset parameter, and the lubrication control parameter is adaptively adjusted according to the lubrication offset parameter.

[0006] Optionally, in the first implementation manner of the first aspect of the present application, the step of collecting the multi-dimensional running state parameters of the gear pair according to the multi-dimensional sensors of the gear transmission system to generate the working condition state data set comprises: acquiring the multi-dimensional running state parameters of the gear pair in the running process according to the multi-dimensional sensors of the gear transmission system to generate an initial running parameter set; constructing a target vector matrix through normalization processing and numerical mapping of the initial running parameter set, and generating a working condition time sequence according to the mutual correlation between the vectors in the target vector matrix; extracting a feature segment within a fixed period through a time window sliding strategy based on the working condition time sequence, and reconstructing the feature segment into a multi-variable input tensor based on a preset dimension; generating the working condition state data set through dynamic mapping of the multi-variable input tensor.

[0007] Optionally, in the second implementation manner of the first aspect of the present application, the step of determining the shear rate distribution data and the oil film thickness distribution data of the lubricated contact area according to the working condition state data set comprises: determining the contact pressure information of the meshing surface of the gear pair according to the working condition state data set, and dividing the pressure information into multi-section contact load data according to the meshing surface distribution area; determining the contact load type and size of each section of the boundary loading unit according to the multi-section contact load data, and generating corresponding load boundary condition data; inputting the load boundary condition data into a lubricated contact state solving unit, spatially distributing the lubricating oil film thickness of each section, and analyzing the velocity relationship of the relative sliding surfaces in the meshing distribution area to generate corresponding shear rate distribution data and oil film thickness distribution data.

[0008] Optionally, in the third implementation manner of the first aspect of the present application, the step of inputting the shear rate distribution data and the oil film thickness distribution data into a preset neural network model to jointly predict the oil film stiffness and the gear surface friction coefficient of the lubricated contact area to generate a joint prediction result comprises: constructing a feature alignment matrix according to the spatial position index of the corresponding nodes in the shear rate distribution data and the oil film thickness distribution data; combining the shear rate, film thickness value and corresponding spatial gradient information of each node in the feature alignment matrix to generate a fusion structure feature input tensor; inputting the fusion structure feature input tensor into a preset neural network model; By establishing a feature representation of the nonlinear coupling relationship between the shear rate and the film thickness distribution in the shared layer of the neural network, an encoding vector is transmitted to the oil film stiffness output channel and the friction coefficient output channel, and the corresponding prediction results are output in the respective channels; The oil film stiffness prediction result and the friction coefficient prediction result are rearranged based on the spatial node sequence to form a two-dimensional joint prediction result matrix, and a joint prediction result is generated.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present application, the step of determining whether the current gear pair lubrication strategy meets the safe operation requirement according to the joint prediction result, and generating a corresponding lubrication level according to the determination result, comprises: According to the joint prediction result, the index values of each spatial node are extracted to construct a feature parameter vector group for lubrication state evaluation; By performing dimension-by-dimension difference calculation on the feature parameter vector group and the safe operation benchmark value corresponding to the safe operation requirement, a corresponding index offset degree is obtained; According to the index offset degree and the preset deviation range, it is determined whether the current gear pair lubrication strategy is in a safe operation determination interval; If the gear pair lubrication strategy is not in the safe operation determination interval, the index offset value is subjected to interval discretization processing, and the lubrication level of the lubrication contact area is determined based on the joint preset oil film stiffness level division rule and friction coefficient classification rule.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present application, the step of performing difference processing on the oil film stiffness and the friction coefficient according to the lubrication level, determining a lubrication offset parameter, and adaptively adjusting the lubrication control parameter according to the lubrication offset parameter, comprises: According to the joint prediction result of the current working period and the historical working period, the oil film stiffness prediction value and the friction coefficient prediction value at the same spatial node are extracted to construct an index difference pair sequence of the corresponding node; By comparing the index difference pair sequence with the parameter tolerance range corresponding to the lubrication level, the index items exceeding the tolerance limit are determined, and node mapping processing is performed based on the index items to generate an abnormal point difference map; According to the difference value amplitude and the node distribution density in the abnormal point difference map, the offset trend of each index in the spatial distribution is weighted to generate a lubrication offset parameter vector; According to the lubrication offset parameter vector, the corresponding adjustment channel is determined through a preset control mapping relationship library combined with the benchmark configuration value of the current lubrication control parameter, and a control parameter adjustment instruction is output.

[0011] Optionally, in a sixth implementation form of the first aspect of the application, the method further comprises: obtaining the shear rate distribution data and the oil film thickness distribution data adjusted based on the lubrication control parameter adjustment instruction, and generating an input feature data set; outputting an adjusted joint prediction result by inputting the input feature data set into the preset neural network model, and performing node-level comparison between the adjusted joint prediction result and the joint prediction result before adjustment to generate a difference feedback matrix; according to an index difference amplitude of the difference feedback matrix, classifying and labeling the change amplitudes of all nodes, and determining whether the change amplitudes exceed a preset error threshold interval; if the change amplitudes exceed the preset error threshold interval, constructing a training sample expansion set according to the input feature data and the corresponding adjusted joint prediction result; updating the model parameters of the preset neural network model according to the training sample expansion set.

[0012] The second aspect of the application provides a transmission gear real-time lubrication control device, which is used to implement a transmission gear real-time lubrication control method. The collection module is configured to collect multi-dimensional running state parameters of a gear pair according to multi-dimensional sensors of a gear transmission system, and generate a working condition state data set. The determination module is configured to determine shear rate distribution data and oil film thickness distribution data of the lubrication contact area according to the working condition state data set. The prediction module is configured to input the shear rate distribution data and the oil film thickness distribution data into a preset neural network model, jointly predict the oil film stiffness and the friction coefficient of the gear surface of the lubrication contact area, and generate a joint prediction result. The judgment module is configured to determine whether a current gear pair lubrication strategy meets the safety operation requirement according to the joint prediction result, and generate a corresponding lubrication level according to the determination result. The control module is configured to perform differential processing on the oil film stiffness and the friction coefficient of the gear surface according to the lubrication level, determine a lubrication offset parameter, and perform self-adaptive adjustment on a lubrication control parameter according to the lubrication offset parameter.

[0013] The third aspect of the embodiment of the application provides an electronic device including a memory and a processor, wherein the processor is configured to execute a computer program stored in the memory, and the processor implements each step of the transmission gear real-time lubrication control method provided in the first aspect of the embodiment of the application when executing the computer program.

[0014] The fourth aspect of the embodiment of the present application provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, each step in the real-time lubrication control method of the transmission gear provided by the first aspect of the embodiment of the present application is implemented.

[0015] In summary, according to the real-time lubrication control method and device of the gear transmission system provided by the scheme of the present application, the running state parameters of the gear pair are collected in multiple dimensions according to the multi-dimensional sensors of the gear transmission system to generate a working condition state data set; the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area are determined according to the working condition state data set; the shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model to jointly predict the oil film stiffness and the tooth surface friction coefficient of the lubrication contact area, and a joint prediction result is generated; whether the current gear pair lubrication strategy meets the safe operation requirement is judged according to the joint prediction result, and a corresponding lubrication level is generated according to the judgment result; the oil film stiffness and the tooth surface friction coefficient are differentially processed according to the lubrication level, a lubrication offset parameter is determined, and the lubrication control parameter is adaptively adjusted according to the lubrication offset parameter. Through the implementation of the scheme of the present application, the lubrication parameters are adjusted in time according to the joint prediction result, and the precise adjustment and optimization of the gear lubrication state under high load and high speed working conditions are realized. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the real-time lubrication control method of the transmission gear provided by the embodiment of the present application is shown in the figure. Figure 2 The program module schematic diagram of the real-time lubrication control device of the transmission gear provided by the embodiment of the present application is shown in the figure. Figure 3 The structure schematic diagram of the electronic device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0017] In order to make the purposes, characteristics and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0018] In order to solve the problem of insufficient lubrication control precision of the gear pair in the related art, the embodiment of the present application provides a real-time lubrication control method of transmission gear, which comprises the following steps: Figure 1 The flowchart of the real-time lubrication control method of the transmission gear provided by the embodiment is shown in the figure. The real-time lubrication control method of the transmission gear comprises the following steps: Step 110, according to the multi-dimensional sensor of the gear transmission system, the running state parameters of the gear pair are collected in multiple dimensions to generate a working condition state data set.

[0019] Specifically, in the present embodiment, by deploying multiple types of sensor modules in the gear transmission system, key running state parameters are collected in multiple dimensions. These parameters include speed, load, oil temperature, and gear face vibration signals. In order to ensure the usability of the collected data, the system performs real-time filtering and denoising on the original signals, and combines time window strategy to structure the working condition. The processed data not only has good time sequence stability and distinguishability, but also can significantly reduce the influence of environmental noise and running disturbance on system judgment, thereby providing accurate and continuous input conditions for subsequent lubrication performance modeling.

[0020] In an alternative embodiment of the present embodiment, the step of collecting running state parameters of the gear pair in multiple dimensions according to the multi-dimensional sensors of the gear transmission system to generate a working condition state data set includes: acquiring the multi-dimensional running state parameters of the gear pair during operation according to the multi-dimensional sensors of the gear transmission system to generate an initial running parameter set; constructing a target vector matrix through normalization processing and numerical mapping of the initial running parameter set, and generating a working condition time sequence according to the mutual correlation between each vector in the target vector matrix; based on the working condition time sequence, extracting feature segments within a fixed period through time window sliding strategy, and reconstructing the feature segments into a multi-variable input tensor based on a preset dimension; generating a working condition state data set by dynamically mapping the multi-variable input tensor.

[0021] Specifically, in the present embodiment, in order to realize the precise control of the lubrication state of the gear transmission system during operation, first, a reliable working condition state data set is constructed based on the operation information of the gear pair. In the specific implementation, the system obtains original operation parameters closely related to the contact state of the gear pair by arranging multi-dimensional sensors inside or outside the gear box, such as strain torque sensors, eddy current speed sensors, thermocouple temperature sensors, and acceleration vibration sensors, etc. The above-mentioned sensors are respectively responsible for collecting torque, speed, lubricating oil temperature and tooth surface vibration acceleration information, and these parameters reflect the key characteristics of the gear pair, such as load level, relative motion state, oil film viscosity and contact surface dynamic response. For example, under high load conditions, torque data can reveal changes in contact stress; when the oil temperature rises, the viscosity of the lubricating oil decreases, which will directly affect the oil film formation and carrying capacity. Therefore, the significance of collecting these data lies in providing a physical basis directly related to the lubrication behavior for subsequent modeling. After collecting the original operation signals output by the above-mentioned sensors, the system forms an initial operation parameter set. However, these original signals inevitably contain noise factors such as electromagnetic interference, mechanical vibration clutter or edge distortion, so the system carries out filtering and denoising processing on the original parameter set. The filtering method can include wavelet threshold denoising or Kalman filtering algorithm, the former extracts the main body of the signal through multi-scale decomposition, and the latter suppresses random interference in time series based on state estimation theory. The denoised signals are also normalized, that is, different dimensional physical quantities are mapped to a unified numerical interval through linear or nonlinear transformation, so that they have equal influence weight in subsequent modeling. After the above processing, each parameter is integrated into a standardized target vector matrix, each column represents the change process of a physical quantity in time series, and each row represents the observation results of multi-dimensional state at a certain time node. After forming the target vector matrix, the system further constructs the working condition time sequence according to the cross-correlation relationship between each vector, the purpose is to describe the cooperative evolution characteristics of the running state variables in the time domain. Cross-correlation analysis can reveal the response delay and intensity relationship between different parameters, such as vibration signals lagging behind load changes, so the time axis is aligned to ensure the consistency of the sequence structure. The generated standardized time sequence has stable rhythm and high correlation, which can be used to describe the time-varying behavior of complex running state. In order to extract features meaningful to lubrication control from it, the system extracts sequence fragments based on time window sliding strategy. Specifically, a fixed length time window is set on the standardized time sequence, and it slides with a fixed step, each time a subsequence with the same length is extracted as a feature fragment. Taking a 50ms window as an example, the system extracts a time period containing all sensor variables every 10ms, which is used to describe the dynamic characteristics of the micro working condition. Through this sliding window mechanism, enough density of time sequence fragments can be obtained without increasing the sampling frequency, so as to capture the rapid evolution characteristics of the lubrication state.Each feature segment is reconstructed into a multivariate input tensor, which is an array structure with multiple dimensions and can be directly recognized and processed by a neural network. In this scenario, the tensor is constructed by arranging the values of each variable at each time point in sequence to form a two-dimensional structure, where one dimension represents the length of time and the other dimension represents the type of parameter. In this way, the input tensor is like a two-dimensional image, with time as the horizontal axis and variables as the vertical axis, and each element in the tensor represents the value of a certain physical parameter at a specific time. By dynamically mapping the input tensor, the final working condition state data set is generated. Dynamic mapping refers to transforming the time series tensor through a certain function or reorganizing the data structure to adapt to subsequent model input requirements while extracting its potential time-space patterns. This data set serves as the basis for subsequent oil film modeling, friction prediction, and control decision-making core functions, and its accuracy and representativeness directly determine the response speed and reliability of the entire lubrication control system.

[0022] Step 120, determining the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area according to the working condition state data set.

[0023] Specifically, in this embodiment, based on the above-mentioned pre-processed working condition data, an analytical calculation model of key physical quantities in the lubrication contact area is established. By fusing and analyzing the contact load conditions and the surface speed relationship of the gear pair, the thickness distribution and shear rate distribution of the lubricating oil film between the gear surfaces are solved. This process utilizes the coupling relationship between fluid dynamics and elastic contact in lubrication theory, combined with computational efficiency optimization methods, to ensure accuracy while meeting real-time requirements. Through this technology, the system can timely capture the change trend of oil film thickness and the local characteristics of shear effect under different working condition changes, thereby more deeply revealing the dynamic evolution behavior of the lubrication layer in the transmission process and providing a physical basis for performance evaluation.

[0024] In an alternative embodiment of the present embodiment, the step of determining the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area according to the working condition state data set includes: determining the contact pressure information of the meshing surface of the gear pair according to the working condition state data set, and dividing the pressure information into multi-section contact load data according to the distribution area of the meshing surface; determining the contact load type and size of each section of the boundary loading unit according to the multi-section contact load data, and generating corresponding load boundary condition data; inputting the load boundary condition data into the lubrication contact state solving unit, spatially distributing the lubricating oil film thickness of each section, and analyzing the speed relationship of the relative sliding surface in the meshing distribution area to generate corresponding shear rate distribution data and oil film thickness distribution data.

[0025] Specifically, in the present embodiment, the contact pressure information of the gear pair meshing surface is calculated based on the constructed operating condition state dataset. The operating condition state dataset contains operating parameters such as torque, speed, temperature, and load variation collected from the sensor system and processed through multiple stages. Based on these parameters, the meshing contact area of the gear pair under a specific operating condition can be reconstructed by combining multi-body dynamics models and finite element contact analysis techniques, and the unit area load at different contact points, i.e., the contact pressure, can be further back-calculated. The contact pressure reflects the local true contact strength between the gear tooth surfaces and is crucial for the formation and maintenance of the lubricating oil film. For example, in the gear meshing area with high load and uneven meshing stiffness, the contact pressure distribution presents a gradient concentration characteristic, and the high pressure area is prone to cause lubricating film rupture or increased boundary friction, so accurate modeling of this area helps to identify potential failure risks in advance. After obtaining the overall contact pressure distribution of the tooth surface, the entire contact area is divided into several functional sections based on the geometric characteristics of the meshing surface. This division can be based on uniform segmentation in the tooth width direction, load density variation trend along the meshing line, or contact heat distribution characteristics. After discretizing the continuous pressure field into multiple section structures, the average pressure, maximum pressure, and contact area in each section can be extracted and summarized as a contact load data set, thereby converting the overall complex distribution into a structured and calculable form. Taking a involute gear pair as an example, a complete meshing period can be divided into the entering zone, stable transmission zone, and disengaging zone, and the load differences in different sections are extracted to facilitate targeted modeling in different lubrication mechanisms. Based on the segmented contact load data, the contact load type and specific size of each section are further determined to generate load boundary condition data that can be input into the lubrication model. In the lubrication state solving, the contact load type determines the lubrication boundary model used. For example, in the light load and high speed section, the contact belongs to the elastohydrodynamic lubrication state, and the main feature is that the oil film thickness is greater than the surface roughness, and the oil film completely isolates the metal surface; while in the low speed starting or high load overlapping area, the boundary lubrication state occurs, and the load is mainly supported by solid contact. Therefore, accurate classification of contact types helps to match different lubrication state solving paths for each section. The corresponding contact load size can be obtained based on the contact area integration or regional average method, and converted into physical quantities such as force load, pressure load, or surface energy density acting on the boundary, thereby forming the initial boundary conditions for the lubrication state solving. After obtaining the load boundary conditions of all sections, they are input into the lubrication contact state solving unit for numerical reconstruction of the oil film state. The solving unit uses finite difference or finite volume method to discretely solve the two-dimensional or three-dimensional contact area based on the EHL calculation model coupled with multiple sections. This model considers factors such as elastic deformation, viscosity-pressure relationship, shear rate dependence, and thermal effect in each calculation unit, and iteratively solves the oil film thickness distribution.The oil film thickness describes the thickness of the liquid layer formed between the contact surfaces by the lubricant, and is directly related to the lubrication effectiveness and the metal surface isolation capacity. At the same time, the solving unit also analyzes the local speed relationship between the relative sliding surfaces to generate shear rate distribution data. The shear rate refers to the rate of relative displacement between the upper and lower surfaces of the oil film layer per unit time, and represents the internal strain rate of the lubricant. High shear rate can cause the viscosity of the lubricating oil to decrease and the temperature to rise, thereby affecting the oil film carrying capacity and the stability of the lubrication state. Therefore, the relative speed information at different positions in the contact area is analyzed in combination with the oil film thickness to construct a two-dimensional shear rate-film thickness distribution map, thereby providing basic data support for the prediction of the oil film stiffness and the friction behavior of the next step.

[0026] In step 130, the shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model to jointly predict the oil film stiffness and the gear surface friction coefficient of the lubrication contact area, and a joint prediction result is generated.

[0027] Specifically, in the embodiment, the obtained oil film thickness and shear rate distribution data are input into a preset neural network model as joint features. The neural network is a multi-output structure and has the ability to simultaneously predict multiple lubrication indicators. By modeling the nonlinear relationship between the shear characteristics and the film thickness change, the system can effectively learn the change law of the oil film stiffness and the gear surface friction coefficient under different input conditions. The network uses a shared feature extraction mechanism to improve modeling efficiency while maintaining the difference between the output channels, so that each output indicator can be optimized. The prediction result is generated in the form of a two-dimensional matrix in the form of a spatial node to express the distribution state of the lubrication characteristics in the gear surface area, so that the control system can have a comprehensive understanding of the overall lubrication condition.

[0028] In an alternative embodiment of the present embodiment, the step of inputting the shear rate distribution data and the oil film thickness distribution data into a preset neural network model to jointly predict the oil film stiffness and the gear surface friction coefficient of the lubrication contact area, and generating a joint prediction result, includes: constructing a feature alignment matrix according to the spatial position index of the corresponding nodes in the shear rate distribution data and the oil film thickness distribution data; combining the shear rate, film thickness value and corresponding spatial gradient information of each node in the feature alignment matrix to generate a fusion type structural feature input tensor; inputting the fusion type structural feature input tensor into the preset neural network model; establishing a feature representation of the nonlinear coupling relationship between the shear rate and the film thickness distribution in the shared layer of the neural network, respectively transmitting the encoding vector to the oil film stiffness output channel and the gear surface friction coefficient output channel, and outputting the corresponding prediction result in the respective channel; and rearranging the oil film stiffness prediction result and the gear surface friction coefficient prediction result based on the spatial node order into a two-dimensional joint prediction result matrix to generate the joint prediction result.

[0029] Specifically, in the present embodiment, the shear rate distribution data and the oil film thickness distribution data are spatially aligned, and a feature alignment matrix is constructed therefrom. In the numerical solution of the lubrication state field, each spatial node represents a small discrete area on the meshing surface of the gear pair, and the node corresponds to a group of shear rate values and oil film thickness values, which are derived from the previous lubrication state solving module. Due to the complex distribution characteristics of the tooth surface contact area in space, a unified spatial position indexing mechanism must be used to accurately match the shear rate data and the film thickness data, ensuring that the physical parameters of the same node can be combined and processed. For example, a two-dimensional position mapping table is constructed, with the transverse and longitudinal coordinates of the meshing surface as the index keys, matching and storing the shear rate and oil film thickness values at each node, thereby forming a spatially consistent feature alignment matrix. On the basis of the constructed feature alignment matrix, physical property enhancement processing is performed on each node. Specifically, the shear rate values, film thickness values, and their gradient information in spatial distribution are fused. Gradient information is an important indicator of the rate of change of a variable in space, reflecting the severity of the change in the contact state, such as the change slope of the shear rate along the meshing line or the volatility of the oil film thickness in the local area. By introducing first or second order spatial derivatives, the representation ability of the input data for boundary mutations, local anomalies, or gradient focusing areas can be enhanced. After the shear rate, film thickness values, and their gradient data are uniformly encoded, they are spliced according to the preset dimensions to construct a multi-channel fusion structure feature tensor. The fusion feature tensor is taken as a complete input and is transmitted into a previously constructed neural network model for nonlinear learning and target parameter prediction. The neural network model uses a shared layer design as the main architecture, aiming to learn the deep nonlinear coupling relationship between the shear rate and the film thickness distribution in a unified data representation space. The core role of the shared layer is to extract cross-channel, high-dimensional composite features through multiple layers of convolution or feedforward structure, thereby capturing the complex dynamic correlations of the oil film state. For example, under high-speed light-load operation, the shear rate is high but the film thickness is stable, which represents a shear-dominated lubrication state, while under low-speed high-load operation, film thickness often drops sharply, causing friction to rise, and such nonlinear relationships can be modeled and extracted through ReLU or Sigmoid activation functions in the deep network. The encoded vectors output from the shared layer are sent to the oil film stiffness channel and the tooth surface friction coefficient channel, respectively. Each channel is based on independent parameter mapping and activation strategies to convert the input vector into the predicted result of the target physical quantity. The oil film stiffness reflects the response ability of the lubricating layer to external load, i.e., the resistance strength of oil film deformation to unit load; the tooth surface friction coefficient is a key indicator of the ratio of frictional resistance to normal pressure between the two contact surfaces. The two output channels obtain parameter-optimized output mapping relationships through independent training, realizing physical prediction modeling of the target state quantity.Finally, in order to correspond to the actual position in space and be used for subsequent lubrication evaluation and regulation, the output oil film stiffness and friction coefficient results at each space node are reordered according to the original space position index to construct a two-dimensional joint prediction result matrix. The matrix is consistent with the topological structure of the meshing surface in space, and the prediction result of each position can directly correspond to the actual lubrication state of the field tooth surface area, so as to realize the visual expression and controllable modeling of the lubrication characteristics in the space layer.

[0030] Step 140, judging whether the current gear pair lubrication strategy meets the safety operation requirement according to the joint prediction result, and generating the corresponding lubrication level according to the judgment result.

[0031] Specifically, in the embodiment, according to the prediction output result of the neural network, it is judged in real time whether the current lubrication strategy meets the safety operation requirement of the gear pair. The system extracts the numerical values of the oil film stiffness and the friction coefficient at different space positions, and compares them with the system set running safety benchmark to analyze the lubrication performance deviation. Combined with the degree and distribution trend of the performance deviation, the system further divides the current state into good, boundary, critical or failure according to the pre-set lubrication level standard. This process introduces dynamic stability constraints and fuzzy rule mechanism, which improves the sensitivity and robustness of the identification of critical lubrication state, and ensures that the system can still make reasonable judgment and response when facing sudden changes in working conditions.

[0032] In an alternative embodiment of the present embodiment, the step of judging whether the current gear pair lubrication strategy meets the safety operation requirement according to the joint prediction result, and generating the corresponding lubrication level according to the judgment result, comprises: extracting the index values of each space node according to the joint prediction result, and constructing a feature parameter vector group for lubrication state evaluation; by calculating the difference value of the feature parameter vector group and the safety operation benchmark value corresponding to the safety operation requirement, the corresponding index deviation is obtained; judging whether the current gear pair lubrication strategy is in the safety operation judgment interval according to the index deviation and the pre-set deviation range; if the gear pair lubrication strategy is not in the safety operation judgment interval, the index deviation value is processed by interval discretization, and the lubrication level of the lubrication contact area is determined based on the pre-set oil film stiffness level division rule and the friction coefficient grading rule.

[0033] Specifically, in the present embodiment, the key physical indicator values corresponding to each spatial node are extracted from the joint prediction results, including oil film stiffness and friction coefficient of tooth surface. The oil film stiffness refers to the deformation resistance generated by the oil film under unit load, and the higher the value, the better the load-carrying capacity of the lubricating layer; the friction coefficient of tooth surface describes the ratio of frictional resistance to normal load between two meshing surfaces, and its change is directly related to energy consumption, heat accumulation and wear rate. Therefore, in the two-dimensional joint prediction result matrix, the two physical parameters of each spatial node are extracted point by point, and the parameter vector group containing the lubrication state characteristics of the whole space is constructed, thereby forming a multi-dimensional data set representing the lubrication health state of the whole tooth surface. In order to determine whether the current lubrication strategy meets the safe operation requirements of the gear pair, the extracted characteristic parameter vector group is compared with the established safe operation benchmark value dimension by dimension. The safe operation benchmark value is a set of reference indicator values representing the ideal lubrication state determined by experimental data, historical operation records or design requirements, which represents the optimal lubrication interval of the gear pair under different loads, speeds and temperatures. The deviation of each physical parameter at the current position can be obtained by calculating the difference between the predicted value and the corresponding benchmark value, which represents the degree and direction of the actual lubrication state deviating from the ideal state. For example, under high load and high speed, if the oil film stiffness value of a certain node is lower than the benchmark level and the friction coefficient is too high, it indicates that the lubrication degradation exists in this area, and the offset value will present a combination of negative offset and positive offset. According to the pre-set deviation threshold range, the above offset is judged to determine whether the current lubrication strategy is in the safe operation judgment interval. The deviation range is used to define the fluctuation limit of the physical indicator, which is set according to the system design margin and material performance. If the offset falls within the range, it is considered that the current lubrication state is acceptable, otherwise it is considered that there is a risk. Taking the case of 15% lower oil film stiffness and 20% higher friction coefficient of a certain node as an example, if the tolerance threshold is set to ±10%, the lubrication state of the node can be considered to be out of the safe interval. When it is confirmed that there is a safety risk, the index offset is discretized, that is, the continuous offset value is mapped to a discrete level label, which is convenient for subsequent comprehensive evaluation of the lubrication level. This discretization process divides multiple intervals according to the offset degree of oil film stiffness and friction coefficient, and assigns a level label to each interval. For example, oil film stiffness offset less than -20% can be defined as "low stiffness level three", and friction coefficient offset greater than +25% can be defined as "high friction level two". Finally, combined with the level labels of the two indicators, the lubrication level of the tooth surface contact area is comprehensively evaluated using the pre-set joint judgment rule. For example, if the oil film stiffness of a certain node is in the second low interval and the friction coefficient is in the third high interval, it can be jointly judged as "lubrication level IV", which means that the lubrication strategy needs to be adjusted to prevent further deterioration.The lubrication level can be used for real-time display of system state, and also used as an input control signal for subsequent regulation process, so as to realize closed-loop management of the running state of the gear pair.

[0034] In step 150, the oil film stiffness and the friction coefficient of the tooth surface are differentially processed according to the lubrication level, the lubrication deviation parameter is determined, and the lubrication control parameter is adaptively adjusted according to the lubrication deviation parameter.

[0035] Specifically, in the embodiment, according to the generated lubrication level in the evaluation, the system performs differential analysis on the prediction index, extracts the deviation trend of the oil film stiffness and the friction coefficient at the spatial node, and constructs a lubrication deviation parameter vector. Combined with the preset mapping relationship of the current control parameter configuration, the adjustment direction and amplitude are determined, and the corresponding control instruction is output, so as to realize adaptive adjustment of the lubrication control parameter. The technical process realizes system-level closed-loop optimization, so that the lubrication supply can be matched with the actual working condition in real time, and the running efficiency and stability of the transmission system are effectively improved.

[0036] In an optional embodiment of the present embodiment, the step of differentially processing the oil film stiffness and the friction coefficient of the tooth surface according to the lubrication level, determining the lubrication deviation parameter, and adaptively adjusting the lubrication control parameter according to the lubrication deviation parameter, comprises: extracting the oil film stiffness prediction value and the friction coefficient prediction value of the same spatial node according to the joint prediction results of the current working period and the historical working period, constructing the index difference value pair sequence of the corresponding node; by comparing the index difference value pair sequence with the parameter tolerance range corresponding to the lubrication level, determining the index item exceeding the tolerance limit, and performing node mapping processing based on the index item, generating an abnormal point difference atlas; according to the difference value amplitude and the node distribution density in the abnormal point difference atlas, the deviation trend of each index in the spatial distribution is weighted processed, and a lubrication deviation parameter vector is generated; according to the lubrication deviation parameter vector, combined with the reference configuration value of the current lubrication control parameter, the corresponding adjustment channel is determined through the preset control mapping relationship library, and the control parameter adjustment instruction is output.

[0037] Specifically, in the present embodiment, the nodes refer to spatially discrete sampling points or calculation unit positions on the meshing surface of the gear pair, which are used to represent the spatial distribution of the oil film parameters and friction performance. The meshing surface is a complex three-dimensional curved surface, and the force, relative sliding speed, contact stress and lubrication state at different positions during the meshing process of the gear pair can be different, so it must be divided into a limited number of spatial sub-regions for analysis and calculation. The historical working period can be the current working period or the last working period. In order to realize the dynamic adaptive adjustment of the lubrication state of the gear pair, the continuity of the evolution trend of the lubrication state in the time dimension is established, so the predicted values of the oil film stiffness and the friction coefficient of the same spatial node should be extracted based on the joint prediction results of the current working period and the last working period. The oil film stiffness reflects the anti-compressive deformation ability of the lubricating layer under load, and the friction coefficient of the tooth surface represents the sliding resistance characteristics of the contact surface. By extracting the prediction results of the two parameters of each node in two or more consecutive working periods and arranging them in chronological order to construct a difference pair sequence, the evolution direction and severity of the lubrication state in the local spatial region can be accurately reflected. For example, when the load changes from high to low, if the oil film stiffness of a node continuously decreases while the friction coefficient continuously increases, the difference pair sequence will present a negative-positive combination, indicating that the lubrication performance is deteriorating. After constructing the index difference pair sequence, the sequence is compared with the parameter tolerance range corresponding to different lubrication levels point by point. The parameter tolerance range is the allowable fluctuation interval set according to the lubrication level, which is used to determine whether the change of a physical index exceeds the system tolerance range. If the change value of the oil film stiffness of a node is greater than -15%, and the maximum downward tolerance of the current lubrication level specified by the system is -10%, the index of the node is identified as abnormal. In this way, the out-of-limit index items can be screened from all spatial nodes, and then the node mapping process is performed combined with the spatial coordinate information, so as to construct a graphical data structure containing the distribution information of abnormal points on the two-dimensional or three-dimensional gear meshing surface, i.e. the abnormal point difference map, which is used to intuitively express the spatial aggregation characteristics of the unbalanced lubrication performance. Then, based on the difference amplitude and node distribution density in the abnormal point difference map, the change trend of the oil film stiffness and the friction coefficient on the entire contact surface is weighted calculated, and the lubrication offset parameter vector is generated. The larger the difference amplitude, the more severe the change of the lubrication state, and the higher the node density, the more concentrated the problem area in space. By fusing and weighting these factors, the comprehensive offset direction and severity of each parameter in the meshing area can be quantitatively obtained, thereby forming a set of lubrication offset parameters with spatial distribution significance. After obtaining the lubrication offset parameter vector, it is further fused with the baseline configuration value of the current lubrication control parameter, and the corresponding control adjustment channel is determined based on the pre-set control mapping relationship library. The lubrication control parameters include oil supply pressure, oil injection rate, oil temperature, oil viscosity adjustment strategy, etc., and the baseline configuration value is the initial or running set value of the system.The control mapping relationship library defines corresponding adjustment mechanisms under different lubrication offset modes by pre-establishing a function mapping between offset parameters and control parameters, for example, a low stiffness offset triggers a pressure compensation channel, and an increased friction coefficient triggers an oil temperature adjustment channel. Finally, the mapping result is converted into a control parameter adjustment instruction and output to the lubrication control system, which automatically modifies the lubrication strategy, thereby realizing dynamic adaptive control of the system and ensuring that the gear pair maintains an efficient and low-wear operating state under varying operating conditions.

[0038] In an optional implementation of the embodiment, the shear rate distribution data and the oil film thickness distribution data adjusted based on the lubrication control parameter adjustment instruction are obtained to generate an input feature data set; the input feature data set is input into a preset neural network model to output an adjusted joint prediction result, and the adjusted joint prediction result is compared with the joint prediction result before adjustment at a node level to generate a difference feedback matrix; according to an index difference amplitude of the difference feedback matrix, the change amplitudes of all nodes are classified and labeled, and it is determined whether the change amplitudes exceed a preset error threshold interval; if the change amplitudes exceed the preset error threshold interval, a training sample expansion set is constructed according to the input feature data and the corresponding adjusted joint prediction result; and the model parameters of the preset neural network model are updated according to the training sample expansion set.

[0039] Specifically, in the present embodiment, after the lubrication control parameter adjustment instruction is issued and executed by the system, the control variables of the lubrication supply of the gear pair, the oil injection angle, the flow rate, or the viscosity grade, etc. change, and the changes in these variables directly affect the flow behavior and pressure response of the lubricating oil in the contact area. At this time, the system re-collects the operating state parameters of the gear pair through the multi-dimensional sensor and inputs them into the lubrication state analysis module to construct the corresponding contact pressure field and velocity field. Combined with the geometric meshing model and material parameters, the system re-applies the load boundary conditions to the contact area and re-calculates the local shear rate and film thickness values of each node in the oil film area, thereby obtaining the adjusted shear rate distribution data and oil film thickness distribution data. By structurally integrating the adjusted shear rate distribution data and oil film thickness distribution data, the input feature data set under the current lubrication state is generated. The generated input feature data set is sent to the trained neural network model, and the model performs forward propagation according to the internal learned weight structure to output the joint prediction result in the current period. Subsequently, the joint prediction result in the period before the lubrication control parameter adjustment is called, and the new and old prediction results are compared item by item with the node number as the index to form a difference feedback matrix. By classifying and counting the numerical change amplitudes in the difference feedback matrix, the nodes are classified and labeled according to the offset strength. This classification and labeling is used to distinguish between small fluctuations and significant changes in the prediction values, thereby filtering out nodes that may pose a threat to lubrication safety. For example, if the friction coefficient of a node changes by more than ±0.15 and it is in a high shear region, the system can label it as a "high-risk fluctuation node". At the same time, the system compares the change amplitudes of all nodes with the preset error tolerance interval to determine whether there is a structural deviation between the current neural network model output and the actual working condition. The error tolerance interval is set by empirical data, experimental statistical values, or stable prediction error distribution under standard working conditions, representing the acceptable error range of the model when it is running stably. When the prediction change amplitude of some nodes exceeds the above tolerance range, it indicates that the current model structure has lost its stable prediction ability in that region, and the system will trigger the model self-updating logic. The system pairs the input-output sample groups according to the corresponding input feature data in the current period and the joint prediction result output by the model, and then constructs an extended training sample set. This extended set is used as new samples to supplement the original training data and cover new areas that the original model distribution cannot effectively fit. The extended training sample set is input into the online updating module of the neural network model to perform weight correction based on the incremental learning rule. Incremental learning is different from complete retraining, as it only adjusts the local parameters of the network structure affected by the new samples, thereby preserving the original learning ability while enhancing the model's response to new working conditions. For example, the gradient rollback protection mechanism can be used to avoid catastrophic forgetting, and regularization constraints can be introduced to prevent the model from overfitting to new samples.The updated neural network will continue to be used for lubrication state prediction in subsequent cycles, forming a closed-loop self-updating control system. In this way, the system not only dynamically tracks changes in lubrication performance, but also continuously optimizes model prediction accuracy, improving the running robustness of complex gear systems under non-steady-state conditions.

[0040] According to the transmission gear real-time lubrication control method provided by the scheme, the running state parameters of the gear pair are collected in multiple dimensions according to the multi-dimensional sensors of the gear transmission system, and a working condition state data set is generated; the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area are determined according to the working condition state data set; the shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model, the oil film stiffness and the friction coefficient of the gear surface of the lubrication contact area are jointly predicted, and a joint prediction result is generated; whether the current gear pair lubrication strategy meets the safe operation requirement is judged according to the joint prediction result, and the corresponding lubrication level is generated according to the judgment result; the oil film stiffness and the friction coefficient of the gear surface are differentially processed according to the lubrication level, the lubrication offset parameter is determined, and the lubrication control parameter is adaptively adjusted according to the lubrication offset parameter. Through the implementation of the scheme, the lubrication parameters are adjusted in time according to the joint prediction result, and the precise adjustment and optimization of the gear lubrication state under high load and high speed working conditions are realized.

[0041] Figure 2 A transmission gear real-time lubrication control device is provided for the embodiments of the present application. The transmission gear real-time lubrication control device can be used to implement the transmission gear real-time lubrication control method in the foregoing embodiments. As shown in the figure, the transmission gear real-time lubrication control device mainly includes: Figure 2 The acquisition module 10 is configured to collect running state parameters of the gear pair in multiple dimensions according to multi-dimensional sensors of the gear transmission system, and generate a working condition state data set. The determination module 20 is configured to determine shear rate distribution data and oil film thickness distribution data of the lubrication contact area according to the working condition state data set. The prediction module 30 is configured to input the shear rate distribution data and the oil film thickness distribution data into a preset neural network model, jointly predict the oil film stiffness and the friction coefficient of the gear surface of the lubrication contact area, and generate a joint prediction result. The judgment module 40 is configured to judge whether the current gear pair lubrication strategy meets the safe operation requirement according to the joint prediction result, and generate the corresponding lubrication level according to the judgment result. The control module 50 is configured to differentially process the oil film stiffness and the friction coefficient of the gear surface according to the lubrication level, determine the lubrication offset parameter, and adaptively adjust the lubrication control parameter according to the lubrication offset parameter.

[0042] ​In an optional implementation of the embodiment, the collection module is specifically configured to: acquire multi-dimensional running state parameters of the gear pair during running according to the multi-dimensional sensors of the gear transmission system, to generate an initial running parameter set; construct a target vector matrix through normalization processing and numerical mapping of the initial running parameter set, and generate a working condition time sequence according to the mutual correlation between vectors in the target vector matrix; based on the working condition time sequence, extract a feature segment within a fixed period through a time window sliding strategy, and reconstruct the feature segment into a multi-variable input tensor based on a preset dimension; generate a working condition state data set through dynamic mapping of the multi-variable input tensor.

[0043] In an optional implementation of the embodiment, the determination module is specifically configured to: determine contact pressure information of a meshing surface of the gear pair according to the working condition state data set, and divide the pressure information into multi-section contact load data according to a distribution area of the meshing surface; determine the contact load type and size of each section of the boundary loading unit according to the multi-section contact load data, to generate corresponding load boundary condition data; input the load boundary condition data into the lubrication contact state solving unit, to perform spatial distribution deconstruction on the lubricating oil film thickness of each section, and analyze the velocity relationship of the relative sliding surfaces in the meshing distribution area, to generate corresponding shear rate distribution data and oil film thickness distribution data.

[0044] In an optional implementation of the embodiment, the prediction module is specifically configured to: construct a feature alignment matrix according to the spatial position index of the corresponding nodes in the shear rate distribution data and the oil film thickness distribution data; perform combined processing on the shear rate, film thickness value and corresponding spatial gradient information of each node in the feature alignment matrix, to generate a fusion type structure feature input tensor; input the fusion type structure feature input tensor into a preset neural network model; establish a feature representation of the nonlinear coupling relationship between the shear rate and the film thickness distribution in the shared layer of the neural network, respectively transmit the encoding vectors to the oil film stiffness output channel and the tooth surface friction coefficient output channel, and output the corresponding prediction results in the respective channels; rearrange the oil film stiffness prediction result and the tooth surface friction coefficient prediction result into a two-dimensional joint prediction result matrix based on the spatial node order, to generate a joint prediction result.

[0045] In an optional implementation of the embodiment, the judging module is specifically configured to: extract index values of each spatial node according to the joint prediction result, and construct a feature parameter vector group for lubrication state evaluation; obtain a corresponding index offset degree by performing dimension-by-dimension difference calculation on the feature parameter vector group and a safe operation benchmark value corresponding to a safe operation requirement; determine whether the current gear pair lubrication strategy is in a safe operation determination interval according to the index offset degree and a preset deviation range; if the gear pair lubrication strategy is not in the safe operation determination interval, perform interval discretization processing on the index offset value, and jointly determine a lubrication level of the lubrication contact area based on a preset oil film stiffness grading rule and a friction coefficient grading rule.

[0046] In an optional implementation of the embodiment, the control module is specifically configured to: extract an oil film stiffness prediction value and a gear face friction coefficient prediction value at the same spatial node according to joint prediction results of the current working period and the historical working period, and construct an index difference pair sequence of the corresponding node; determine an index item exceeding a tolerance limit by comparing the index difference pair sequence with a parameter tolerance range corresponding to the lubrication level, perform node mapping processing based on the index item, and generate an abnormal point difference atlas; perform weighted processing on an offset trend of each index in the spatial distribution according to a difference value amplitude and a node distribution density in the abnormal point difference atlas, and generate a lubrication offset parameter vector; determine a corresponding adjustment channel by a preset control mapping relationship library according to the lubrication offset parameter vector and in combination with a benchmark configuration value of the current lubrication control parameter, and output a control parameter adjustment instruction.

[0047] In an optional implementation of the embodiment, the control module is further configured to: obtain shear rate distribution data and oil film thickness distribution data adjusted based on the lubrication control parameter adjustment instruction, generate an input feature data set; output an adjusted joint prediction result by inputting the input feature data set into a preset neural network model, and perform node-level comparison between the adjusted joint prediction result and the joint prediction result before adjustment, to generate a difference value feedback matrix; classify and label a change amplitude of all nodes according to an index difference value amplitude of the difference value feedback matrix, and determine whether the change amplitude exceeds a preset error threshold interval; if the change amplitude exceeds the preset error threshold interval, construct a training sample expansion set according to the input feature data and the corresponding adjusted joint prediction result; update model parameters of the preset neural network model according to the training sample expansion set.

[0048] According to the transmission gear real-time lubrication control device provided in the scheme, the operation state parameters of the gear pair are collected in multiple dimensions according to the multi-dimensional sensors of the gear transmission system, and a working condition state data set is generated; the shear rate distribution data and the oil film thickness distribution data of the lubrication contact area are determined according to the working condition state data set; the shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model, the oil film stiffness and the tooth surface friction coefficient of the lubrication contact area are jointly predicted, and a joint prediction result is generated; whether the current gear pair lubrication strategy meets the safe operation requirement is judged according to the joint prediction result, and the corresponding lubrication level is generated according to the judgment result; the oil film stiffness and the tooth surface friction coefficient are differentially processed according to the lubrication level, the lubrication offset parameter is determined, and the lubrication control parameter is adaptively adjusted according to the lubrication offset parameter. Through the implementation of the scheme, the lubrication parameters are adjusted in time according to the joint prediction result, and the precise adjustment and optimization of the gear lubrication state under high load and high speed working conditions are realized.

[0049] According to the scheme provided in the application Figure 3 An electronic device is provided in the embodiments of the application. The electronic device can be used to implement the transmission gear real-time lubrication control method in the foregoing embodiments, and mainly includes: The memory 301, the processor 302, and the computer program 303 stored in the memory 301 and executable on the processor 302 are communicatively connected. When the processor 302 executes the computer program 303, the transmission gear real-time lubrication control method in the foregoing embodiments is implemented. The number of processors can be one or more.

[0050] The memory 301 can be a high-speed random access memory (RAM) or a non-volatile memory such as a disk memory. The memory 301 is used to store executable program codes, and the processor 302 is coupled with the memory 301.

[0051] Further, the embodiments of the application also provide a computer readable storage medium, which can be arranged in the electronic device in the above embodiments. The computer readable storage medium can be the memory in the above embodiments. Figure 3

[0052] ​The computer readable storage medium stores a computer program, which is executed by the processor to implement the real-time lubrication control method of the transmission gear in the foregoing embodiments. Further, the computer readable storage medium can also be a U disk, a mobile hard disk, a read-only memory (ROM), a RAM, a magnetic disk or an optical disk, and various storage medium capable of storing program codes.

[0053] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the system, device and unit described above can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0054] The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage medium capable of storing program codes.

[0055] The foregoing and the foregoing embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for real-time lubrication control of transmission gears, characterized in that, include: Based on the multi-dimensional sensors of the gear transmission system, the operating status parameters of the gear pair are collected in multiple dimensions to generate a working condition data set. Based on the operating condition dataset, determine the shear rate distribution data and oil film thickness distribution data of the lubrication contact area; The shear rate distribution data and the oil film thickness distribution data are input into a preset neural network model to jointly predict the oil film stiffness and tooth surface friction coefficient of the lubrication contact area, and generate a joint prediction result. Based on the joint prediction results, it is determined whether the current gear pair lubrication strategy meets the requirements for safe operation, and a corresponding lubrication level is generated based on the determination results; Based on the lubrication grade, the oil film stiffness and the tooth surface friction coefficient are differentially processed to determine the lubrication offset parameter, and the lubrication control parameter is adaptively adjusted based on the lubrication offset parameter.

2. The real-time lubrication control method for transmission gears according to claim 1, characterized in that, The step of collecting multi-dimensional operating state parameters of the gear pair from multiple dimensions using multi-dimensional sensors of the gear transmission system to generate a working condition data set includes: The multi-dimensional operating status parameters of the gear pair during operation are obtained from the multi-dimensional sensors of the gear transmission system, and an initial operating parameter set is generated. A target vector matrix is ​​constructed by normalizing and numerically mapping the initial set of operating parameters, and a working condition time series is generated based on the interrelationships between the vectors in the target vector matrix. Based on the operating condition time series, feature segments within a fixed period are extracted using a time window sliding strategy, and the feature segments are reconstructed into multivariable input tensors based on a preset dimension. By dynamically mapping the multivariate input tensor, a working condition data set is generated.

3. The real-time lubrication control method for transmission gears according to claim 1, characterized in that, The step of determining the shear rate distribution data and oil film thickness distribution data of the lubrication contact area based on the operating condition dataset includes: The contact pressure information of the gear pair meshing surface is determined based on the working condition data set, and the pressure information is divided into multi-segment contact load data according to the meshing surface distribution area. Based on the multi-segment contact load data, determine the contact load type and magnitude of each segment of the boundary loading unit, and generate corresponding load boundary condition data; The load boundary condition data is input into the lubrication contact state calculation unit to deconstruct the spatial distribution of the lubricating oil film thickness in each section, and to analyze the velocity relationship between the relative sliding surfaces in the meshing distribution area, thereby generating corresponding shear rate distribution data and oil film thickness distribution data.

4. The real-time lubrication control method for transmission gears according to claim 3, characterized in that, The step of inputting the shear rate distribution data and the oil film thickness distribution data into a preset neural network model to jointly predict the oil film stiffness and tooth surface friction coefficient in the lubrication contact area and generate a joint prediction result includes: A feature alignment matrix is ​​constructed based on the spatial position indices of corresponding nodes in the shear rate distribution data and the oil film thickness distribution data. The shear rate, membrane thickness, and corresponding spatial gradient information of each node are combined in the feature alignment matrix to generate a fused structural feature input tensor. The fusion-type structural feature input tensor is input into a preset neural network model; By establishing a feature representation of the nonlinear coupling relationship between shear rate and film thickness distribution in the shared layer of the neural network, the encoding vector is transmitted to the oil film stiffness output channel and the tooth surface friction coefficient output channel respectively, and the corresponding prediction results are output in their respective channels. The oil film stiffness prediction results and the tooth surface friction coefficient prediction results are rearranged into a two-dimensional joint prediction result matrix based on the spatial node order to generate the joint prediction result.

5. The real-time lubrication control method for transmission gears according to claim 1, characterized in that, The step of determining whether the current gear pair lubrication strategy meets the safe operation requirements based on the joint prediction results, and generating the corresponding lubrication level based on the determination results, includes: Based on the joint prediction results, the index values ​​of each spatial node are extracted to construct a feature parameter vector group for lubrication status assessment. By performing a dimension-by-dimensional difference calculation between the feature parameter vector group and the safety operation benchmark value corresponding to the safety operation requirements, the corresponding index offset is obtained; Based on the deviation of the index from the preset deviation range, it is determined whether the current gear pair lubrication strategy is within the safe operation judgment range; If the gear pair lubrication strategy is not within the safe operation judgment range, the index offset value is discretized into a range, and the lubrication level of the lubrication contact area is determined jointly based on the preset oil film stiffness level classification rule and friction coefficient classification rule.

6. The real-time lubrication control method for transmission gears according to claim 1, characterized in that, The step of performing differential processing on the oil film stiffness and the tooth surface friction coefficient according to the lubrication grade to determine the lubrication offset parameter, and adaptively adjusting the lubrication control parameters according to the lubrication offset parameter, includes: Based on the joint prediction results of the current working cycle and the historical working cycle, the predicted values ​​of oil film stiffness and tooth surface friction coefficient at the same spatial node are extracted to construct the index difference pair sequence of the corresponding node. By comparing the index difference pair sequence with the parameter tolerance range corresponding to the lubrication level, the index items that exceed the tolerance limit are identified, and node mapping processing is performed based on the index items to generate an anomaly point difference map. Based on the difference magnitude and node distribution density in the anomaly point difference map, the offset trend of each indicator in the spatial distribution is weighted to generate a lubrication offset parameter vector. Based on the lubrication offset parameter vector and the baseline configuration value of the current lubrication control parameters, the corresponding adjustment channel is determined through a preset control mapping relationship library, and the control parameter adjustment command is output.

7. The real-time lubrication control method for transmission gears according to claim 6, characterized in that, The method further includes: Obtain shear rate distribution data and oil film thickness distribution data after adjustment based on the lubrication control parameter adjustment command, and generate input feature dataset; By inputting the input feature dataset into the preset neural network model, the adjusted joint prediction result is output, and the adjusted joint prediction result is compared with the original joint prediction result at the node level to generate a difference feedback matrix. Based on the magnitude of the index difference in the difference feedback matrix, the magnitude of change of all nodes is classified and labeled, and it is determined whether the magnitude of change exceeds the preset error threshold range. If the change exceeds the preset error threshold range, then an expanded training sample set is constructed based on the input feature data and the corresponding adjusted joint prediction results; The model parameters of the preset neural network model are updated based on the expanded set of training samples.

8. A real-time lubrication control device for transmission gears, characterized in that, The real-time lubrication control device for transmission gears is used to implement the real-time lubrication control method for transmission gears as described in claim 1, and the real-time lubrication control device for transmission gears includes: The data acquisition module is used to collect the operating status parameters of the gear pair from multiple dimensions based on the multi-dimensional sensors of the gear transmission system, and generate a working condition data set. The determination module is used to determine the shear rate distribution data and oil film thickness distribution data of the lubrication contact area based on the operating condition dataset. The prediction module is used to input the shear rate distribution data and the oil film thickness distribution data into a preset neural network model to jointly predict the oil film stiffness and tooth surface friction coefficient of the lubrication contact area and generate a joint prediction result. The judgment module is used to determine whether the current gear pair lubrication strategy meets the safe operation requirements based on the joint prediction results, and to generate the corresponding lubrication level based on the judgment results; The control module is used to perform differential processing on the oil film stiffness and the tooth surface friction coefficient according to the lubrication level, determine the lubrication offset parameter, and adaptively adjust the lubrication control parameter according to the lubrication offset parameter.

9. An electronic device, characterized in that, Includes memory and processor, of which: The processor is used to execute computer programs stored in the memory; When the processor executes the computer program, it implements the steps in the real-time lubrication control method for transmission gears according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the real-time lubrication control method for transmission gears according to any one of claims 1 to 7.

Images