Heavy machinery-oriented ultrahigh pressure axial plunger pump bearing pressure optimization method

Abstract

The invention provides a heavy machinery-oriented ultrahigh pressure axial plunger pump bearing pressure optimization method, which relates to the field of axial plunger pump bearing design, and comprises the following steps of: acquiring a bearing characteristic data set of a key bearing part of an ultrahigh pressure axial plunger pump in heavy machinery; processing the obtained bearing characteristic data set and dividing the bearing characteristic data set into a training set and a test set; according to the characteristics of the bearing characteristic data set, a bearing characteristic evaluation model of the ultrahigh-pressure axial plunger pump is obtained; performing evaluation and generalization ability test on the ultrahigh pressure axial plunger pump bearing characteristic evaluation model by using the verification set; the design parameters are input into the ultrahigh-pressure axial plunger pump bearing characteristic evaluation model and optimized, ultrahigh-pressure axial plunger pump bearing result data are obtained, and the minimum ultimate bearing pressure value in all the components is taken as the maximum ultimate bearing pressure of the ultrahigh-pressure axial plunger pump. According to the method, data of key bearing components are fused, the model is built by using the neural network, and the accuracy, the calculation efficiency and the generalization ability of bearing capacity evaluation are improved.

Description

Technical Field

[0001] The present invention relates to the field of load-bearing design of axial piston pumps, and particularly to an optimization method for the load-bearing pressure of an ultra-high pressure axial piston pump for heavy machinery. Background Art

[0002] Ultra-high pressure axial piston pumps are the core hydraulic components of heavy machinery and are widely used in high-end manufacturing fields such as construction machinery, aerospace, and ocean engineering. With the continuous increase in the demand for high-performance hydraulic systems in high-end manufacturing, ultra-high pressure axial piston pumps are developing towards higher pressures and higher speeds to achieve higher power densities, while realizing compact and lightweight designs.

[0003] However, in the face of the harsh operating conditions of heavy machinery, the load-bearing design of ultra-high pressure axial piston pumps faces many technical challenges. Ultra-high pressure piston pumps generally refer to piston pumps with a rated operating pressure of 70 MPa or higher in the industry. As the working pressure increases, the friction pairs inside the ultra-high pressure axial piston pump, such as the piston pair and the slipper pair, are long-term subjected to contact stresses of 70 MPa or higher. Under intense friction, the oil temperature rises sharply and remains at 80 - 90 °C for a long time, resulting in increased friction and wear, and a decline in the stability of the oil film, seriously affecting the reliability and service life of the pump. In addition, the ultra-high pressure check valve is subjected to high-pressure shocks between 70 MPa and 2 MPa. Under the action of long-term fatigue loads, it is prone to seal failure and hysteresis phenomena, affecting the dynamic response and efficiency of the system. The load-bearing capacity of the main shaft and bearing system under ultra-high loads within a limited volume also becomes a design difficulty. The current design methods mainly target conventional medium and high pressure piston pumps and rely on empirical formulas and experimental optimization, lacking a systematic optimization design method, resulting in a long design cycle, limited accuracy, and difficulty in achieving efficient parameter matching and structural optimization.

[0004] Therefore, the present invention proposes an optimization method for the load-bearing pressure of an ultra-high pressure axial piston pump for heavy machinery, which can achieve efficient and precise design optimization while ensuring ultra-high load-bearing capacity, and improve the load-bearing capacity and reliability of the ultra-high pressure axial piston pump. Summary of the Invention

[0005] To address the deficiencies of the above-mentioned existing technologies, the objective of the present invention is to provide an optimized method for the bearing pressure of an ultra-high pressure axial piston pump for heavy machinery. By collecting the bearing characteristic data sets of the piston pair, slipper pair, one-way flow valve, and main shaft in the ultra-high pressure axial piston pump of heavy machinery, the bearing characteristic data sets are divided into a training set and a test set through outlier processing, data normalization, and data augmentation; based on the characteristics of the bearing characteristic data sets of the ultra-high pressure axial piston pump, an evaluation model for the bearing characteristics of the ultra-high pressure axial piston pump is obtained. Through the bearing characteristic evaluation model, new design parameters can be predicted, not only improving the accuracy of bearing capacity evaluation, but also enhancing the calculation efficiency and generalization ability of the model, thereby improving the design efficiency of the ultra-high pressure axial piston pump and shortening the design cycle.

[0006] Specifically, the present invention provides an optimized method for the bearing pressure of an ultra-high pressure axial piston pump for heavy machinery, and the specific steps are as follows:

[0007] S1. According to the key bearing components of the ultra-high pressure axial piston pump in heavy machinery, collect the bearing characteristic data sets of the piston pair, slipper pair, one-way flow valve, and main shaft respectively and verify their accuracy;

[0008] S2. Perform outlier processing, data normalization, and data augmentation on the bearing characteristic data sets obtained in step S1, and divide the bearing characteristic data sets into a training set and a test set;

[0009] S3. According to the characteristics of the bearing characteristic data sets of the ultra-high pressure axial piston pump, construct a multi-layer perceptron model to obtain an evaluation model for the bearing characteristics of the ultra-high pressure axial piston pump;

[0010] According to the dimensions of the independent variable parameters and target variable parameters of the key bearing components, obtain the number of neurons in the input layer and output layer respectively. Construct a hidden layer based on the non-linearity between the bearing characteristics and bearing capacity of the ultra-high pressure axial piston pump, and train the model using relevant functions;

[0011] S4. Use the validation set to evaluate the bearing characteristic evaluation model and test its generalization ability;

[0012] S5. Input the design parameters into the bearing characteristic evaluation model and optimize them to obtain the bearing result data of the ultra-high pressure axial piston pump. Take the minimum value of the ultimate bearing pressure among the key bearing components as the maximum ultimate bearing pressure of the ultra-high pressure axial piston pump.

[0013] Preferably, step S1 includes the following sub-steps:

[0014] Respectively obtain the oil film thickness, bearing pressure distribution, ultimate bearing pressure, and / or friction coefficient under ultra-high pressure conditions in the target variable parameters according to the independent variable parameters of the bearing characteristics of the piston pair, slipper pair, and main shaft;

[0015] According to the independent variable parameters of the load-bearing characteristics of the one-way flow control valve, simulate the flow field state of the one-way flow control valve under ultra-high pressure load and the opening and closing time of the spool under pressure switching to obtain the response speed in the target variable parameters; simulate the structural stress and deformation of the one-way flow control valve under ultra-high pressure load to obtain the ultimate load-bearing pressure in the target variable parameters.

[0016] Verify the obtained load-bearing characteristic data set through on-site tests.

[0017] Preferably, step S3 includes a sub-step of constructing a hidden layer, using ReLU as the non-linear activation function, and adding a Dropout layer after ReLU in the first hidden layer, the second hidden layer, the third hidden layer, and the fourth hidden layer to prevent overfitting of the ultra-high pressure axial piston pump load-bearing characteristic evaluation model.

[0018] Preferably, the independent variable parameters of the load-bearing characteristics of the plunger pair include working pressure, plunger speed, hydraulic oil viscosity, hydraulic oil temperature, plunger diameter, plunger head curvature radius, cylinder bore fit clearance, surface coating thickness, cylinder block elastic modulus, plunger elastic modulus, cylinder block hardness, and plunger hardness.

[0019] Preferably, the independent variable parameters of the load-bearing characteristics of the one-way flow control valve include inlet pressure, outlet pressure, flow rate, temperature, spool opening, spool diameter, valve seat diameter, spool-valve seat clearance, spool mass, spool elastic modulus, valve seat elastic modulus, spool hardness, and valve seat hardness.

[0020] Preferably, in step S2, the outlier processing is obtained by calculating the mean μ and standard deviation σ of each parameter in the ultra-high pressure axial piston pump load-bearing characteristic data set, specifically:

[0021]

[0022] Among them, μ is the data mean in the ultra-high pressure axial piston pump load-bearing characteristic data set, N is the number of data in the ultra-high pressure axial piston pump load-bearing characteristic data set, x i is the i-th data in the ultra-high pressure axial piston pump load-bearing characteristic data set, and σ is the data standard deviation in the ultra-high pressure axial piston pump load-bearing characteristic data set.

[0023] Preferably, in step S2, the ultra-high pressure axial piston pump load-bearing characteristic data set after outlier processing is normalized, specifically:

[0024]

[0025] Among them, x is the original data in the ultra-high pressure axial piston pump load-bearing characteristic data set, x min and x maxare the minimum and maximum values in the bearing characteristic dataset of the ultra-high pressure axial piston pump, respectively, and x ′ is the data after normalizing the bearing characteristic dataset of the ultra-high pressure axial piston pump.

[0026] Preferably, step S3 further includes using the root mean square error as a loss function to optimize the prediction accuracy of the bearing characteristic evaluation model of the ultra-high pressure axial piston pump, and using an adaptive learning rate mechanism to optimize the non-linear relationship between the bearing parameters of the ultra-high pressure axial piston pump.

[0027] Preferably, in step S5, the design parameters include the geometric structures, material properties, and operating condition parameters of the plunger pair, slipper pair, one-way flow control valve, and main shaft.

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] 1. Through finite element analysis, computational fluid dynamics, and friction and wear tests, and verified by on-site performance tests and life tests, the present invention collects and constructs a bearing characteristic dataset of the ultra-high pressure axial piston pump, and realizes the prediction of the bearing capacity of the ultra-high pressure axial piston pump through training with a neural network framework, achieving accurate prediction of the bearing capacity of the ultra-high pressure axial piston pump, improving the design efficiency of the ultra-high pressure axial piston pump, and shortening the design cycle. At the same time, this method can be applied to different working conditions and structural parameters, providing a basis for the optimal design and performance improvement of the ultra-high pressure axial piston pump, and helping to improve the reliability, stability, and service life of the system.

[0030] 2. Based on neural network, finite element analysis, and computational fluid dynamics methods, the present invention constructs bearing characteristic evaluation models for key components in the ultra-high pressure axial piston pump, such as plunger pairs, slipper pairs, one-way flow control valves, and main shafts, to achieve accurate prediction of the ultimate bearing pressure, oil film thickness, friction coefficient, and bearing pressure distribution. Compared with the prior art, this method fully integrates multi-source data, combines CFD and FEA simulation analysis, and uses a neural network model for efficient prediction, not only improving the accuracy of bearing capacity evaluation, but also enhancing the computational efficiency and generalization ability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flowchart of the bearing pressure optimization method for the ultra-high pressure axial piston pump for heavy machinery according to the present invention;

[0032] Figure 2 is a diagram of the ultimate bearing pressure of the plunger pair in the bearing pressure optimization method for the ultra-high pressure axial piston pump for heavy machinery according to the present invention;

[0033] Figure 3 is a diagram of the ultimate bearing pressure of the main shaft in the bearing pressure optimization method for the ultra-high pressure axial piston pump for heavy machinery according to the present invention;

[0034] Figure 4 This is the service life diagram of the ultra-high pressure axial piston pump under different pressure conditions in the ultra-high pressure axial piston pump bearing pressure optimization method for heavy machinery of the present invention. Detailed implementation manners

[0035] Hereinafter, the implementation manners of the present invention will be described with reference to the drawings.

[0036] The ultra-high pressure axial piston pump bearing pressure optimization method for heavy machinery of the present invention, as a plunger pump bearing capacity prediction method, as Figure 1 shown, the specific steps are as follows:

[0037] S1. According to the key bearing components of the ultra-high pressure axial piston pump in heavy machinery, including the plunger pair, the slipper pair, the one-way flow distribution valve and the main shaft, use finite element analysis (FEA), computational fluid dynamics (CFD) and friction and wear tests to respectively collect the bearing characteristic data sets of the plunger pair, the slipper pair, the one-way flow distribution valve and the main shaft, and use the on-site test method to collect the performance evolution and life data of the ultra-high pressure axial piston pump to form a bearing result data set, and the bearing result data set is used to verify the accuracy of the bearing characteristic data set, which specifically includes the following sub-steps:

[0038] S11. Establish a computational fluid dynamics simulation model of the plunger pair based on the geometric structure and operating condition parameters in the independent variable parameters of the plunger pair bearing characteristics, and simulate the flow field state of the plunger pair under ultra-high pressure bearing based on CFD simulation to obtain the oil film thickness and bearing pressure distribution in the target variable parameters.

[0039] Establish a structural field calculation model of the plunger pair based on the geometric structure and operating condition parameters in the independent variable parameters of the plunger pair bearing characteristics, apply the obtained bearing pressure distribution of the plunger pair to the structural field calculation model and simulate the structural stress and deformation of the plunger pair under ultra-high pressure bearing based on FEA simulation to obtain the ultimate bearing pressure in the target variable parameters, as Figure 2 shown, the figure shows the bearing state and pressure distribution of the plunger pair in the ultra-high pressure axial piston pump, and the bearing characteristic results of the plunger pair are obtained according to the bearing state and pressure distribution of the plunger pair, so as to provide a basis for realizing the training of the bearing design model of the ultra-high pressure axial piston pump.

[0040] Build a friction and wear test bench for the plunger pair, simulate the use state of the plunger pair under ultra-high pressure conditions, and obtain the friction coefficient and wear rate of the plunger pair under ultra-high pressure state through tribological experiments.

[0041] Specifically, the load-bearing characteristics of the plunger pair are mainly affected by the operating conditions, geometric structure, and material properties of the plunger pair. Therefore, the independent variable parameters of the load-bearing characteristics of the plunger pair include working pressure, plunger speed, hydraulic oil viscosity, hydraulic oil temperature, plunger diameter, plunger head curvature radius, cylinder bore clearance, surface coating thickness, cylinder block elastic modulus, plunger elastic modulus, cylinder block hardness, and plunger hardness, a total of 12-dimensional independent variable parameters; the target variables of the load-bearing characteristics of the plunger pair include ultimate load-bearing pressure, friction coefficient, oil film thickness, and load-bearing pressure distribution, a total of 4-dimensional target variable parameters. The ultimate load-bearing pressure of the plunger pair refers to the maximum pressure that the plunger pair can withstand, which affects the stability and reliability of the system; the oil film thickness of the plunger pair refers to the thickness of the lubricating oil film, which determines the friction coefficient and load-bearing capacity; the friction coefficient of the plunger pair is an important indicator of the friction characteristics of the plunger pair, which affects friction loss and service life; the load-bearing pressure distribution refers to the load-bearing capacity distribution of the contact surface of the plunger pair, which reflects the relationship between friction and load-bearing characteristics.

[0042] S12. Establish a computational fluid dynamics simulation model of the slipper pair based on the geometric structure and operating condition parameters among the independent variable parameters of the load-bearing characteristics of the slipper pair. Based on CFD simulation, simulate the flow field state of the slipper pair under ultra-high pressure load, and obtain the oil film thickness and load-bearing pressure distribution among the target variable parameters.

[0043] Establish a structural field calculation model of the slipper pair based on the geometric structure and operating condition parameters among the independent variable parameters of the load-bearing characteristics of the slipper pair. Apply the obtained load-bearing pressure distribution of the slipper pair to the structural field calculation model and simulate the structural stress and deformation of the slipper pair under ultra-high pressure load based on FEA to obtain the ultimate load-bearing pressure among the target variable parameters.

[0044] Build a friction and wear test bench for the slipper pair, simulate the use state of the slipper pair under ultra-high pressure conditions, and obtain the friction coefficient and wear rate of the slipper pair under ultra-high pressure through tribological experiments.

[0045] Specifically, the load-bearing characteristics of the slipper pair are mainly affected by the operating conditions, geometric structure, and material properties of the slipper pair. Therefore, the independent variable parameters of the load-bearing characteristics of the slipper pair include working pressure, slipper linear velocity, hydraulic oil viscosity, hydraulic oil temperature, swash plate angle, slipper diameter, spherical curvature radius, clearance between the slipper and the swash plate, surface roughness of the swash plate, spherical contact area of the slipper, elastic modulus of the slipper, elastic modulus of the swash plate, hardness of the slipper, and hardness of the swash plate, totaling 14-dimensional independent variable parameters. The target variables of the load-bearing characteristics of the slipper pair include ultimate load-bearing pressure, friction coefficient, oil film thickness, and load-bearing pressure distribution, totaling 4-dimensional target variable parameters. Among them, the ultimate load-bearing pressure of the slipper pair refers to the maximum pressure that the slipper pair can withstand, which affects the stability and reliability of the entire ultra-high pressure axial piston pump system. The oil film thickness of the slipper pair refers to the thickness of the lubricating oil film, which has an important impact on friction and wear resistance. The friction coefficient of the slipper pair is an important indicator of the friction characteristics of the slipper pair, which affects the friction loss and service life of the slipper pair. The load-bearing pressure distribution refers to the load-bearing capacity distribution of the contact surface of the slipper pair, reflecting the relationship between friction and load-bearing characteristics.

[0046] S13. Establish a computational fluid dynamics simulation model of the one-way flow valve based on the geometric structure and operating condition parameters among the independent variable parameters of the load-bearing characteristics of the one-way flow valve. Based on CFD simulation, simulate the flow field state of the one-way flow valve under ultra-high pressure load and the opening and closing time of the one-way valve spool under pressure switching to obtain the pressure distribution of the one-way valve under ultra-high pressure conditions and the response speed among the target variable parameters.

[0047] Establish a structural field calculation model of the one-way flow valve based on the geometric structure and operating condition parameters among the independent variable parameters of the load-bearing characteristics of the one-way flow valve. Apply the load-bearing pressure distribution of the one-way flow valve obtained above to the structural field calculation model and simulate the structural stress and deformation of the one-way flow valve under ultra-high pressure load based on FEA to obtain the ultimate load-bearing pressure among the target variable parameters.

[0048] Specifically, the load-bearing characteristics of the one-way flow valve are mainly affected by the operating conditions, geometric structure, and material properties of the one-way valve. Therefore, the independent variable parameters of the load-bearing characteristics of the one-way flow valve include inlet pressure, outlet pressure, flow rate, temperature, spool opening, spool diameter, valve seat diameter, clearance between the spool and the valve seat, spool mass, elastic modulus of the spool, elastic modulus of the valve seat, hardness of the spool, and hardness of the valve seat, totaling 13-dimensional independent variable parameters. The target variables of the load-bearing characteristics of the one-way flow valve include ultimate load-bearing pressure and response speed, totaling 2-dimensional target variable parameters. The ultimate load-bearing pressure of the one-way flow valve refers to the maximum load-bearing capacity of the spool and the valve body, which affects the operating pressure and durability of the valve. The response speed of the one-way flow valve refers to the opening and closing time of the one-way valve, which affects the flow distribution performance.

[0049] S14. Establish a calculation model for the spindle structure field based on the geometric structure and operating condition parameters among the independent variable parameters of the main bearing load characteristics. Apply the bearing pressure to the structure field calculation model according to the working load of the spindle, and simulate the structural stress and deformation of the spindle under ultra-high pressure based on FEA to obtain the ultimate bearing pressure and bearing pressure distribution in the target variable parameters. As Figure 3 shown, the figure shows the bearing state and pressure distribution of the main bearing in the ultra-high pressure axial piston pump. The bearing characteristics results of the spindle are obtained according to the bearing state and pressure distribution of the main bearing, thus providing a basis for the training of the bearing design model of the ultra-high pressure axial piston pump.

[0050] Build an ultra-high pressure pump spindle friction and wear test bench to simulate the usage state of the spindle under ultra-high pressure conditions, and obtain the friction coefficient and wear rate of the spindle under ultra-high pressure conditions through tribological experiments.

[0051] Specifically, the main bearing load characteristics are mainly affected by the operating conditions, geometric structure and material characteristics of the spindle. Therefore, the independent variable parameters of the main bearing load characteristics include working load, spindle speed, working temperature, hydraulic oil viscosity, spindle diameter, bearing inner diameter, spindle length, spindle elastic modulus, bearing elastic modulus, spindle hardness and bearing hardness, a total of 11-dimensional independent variable parameters. The target variables of the main bearing load characteristics include ultimate bearing pressure, friction coefficient and bearing pressure distribution, a total of 3-dimensional target variable parameters. The ultimate bearing pressure of the spindle refers to the maximum bearing capacity of the main bearing surface, which affects the stability of the spindle during operation. The friction coefficient refers to the friction loss and energy efficiency loss between the spindle and the bearing. The bearing pressure distribution refers to the pressure distribution on the contact surface between the spindle and the bearing, which determines the operating pressure and service life of the contact surface.

[0052] S15. To verify the reliability and accuracy of the bearing characteristic data set of the ultra-high pressure axial piston pump, collect the performance evolution curve during the operation of the ultra-high pressure axial piston pump corresponding to the working conditions of the bearing characteristic data set and the service life under the rated working conditions in the form of on-site tests of the ultra-high pressure axial piston pump. When the operating time of the ultra-high pressure axial piston pump is not less than 8000 hours under the same working conditions, it is considered that the ultra-high pressure bearing characteristic data set is reliable and accurate. The results are as Figure 4 shown. Figure 4 For the service life of the ultra-high pressure axial piston pump under different pressure conditions, it can be obtained from the data shown in the figure that the service life of the ultra-high pressure axial piston pump corresponding to the bearing characteristic data set of the ultra-high pressure axial piston pump exceeds 8000 hours, thus verifying that the bearing characteristic data set of the ultra-high pressure axial piston pump obtained by CFD simulation, FEA simulation and friction and wear tests in the present invention is accurate and reliable, thereby laying a data foundation for training an accurate bearing characteristic evaluation model of the ultra-high pressure axial piston pump.

[0053] S2. Perform outlier handling, data normalization, and data augmentation on the bearing characteristic dataset obtained in step S1. Divide the ultra-high pressure axial piston pump bearing characteristic dataset into a training set and a test set in a ratio of 8:2 to avoid overfitting of the trained model and ensure the accuracy and stability of the model. Use data augmentation methods to expand the training set and improve the accuracy and generalization ability of the model. The ultra-high pressure axial piston pump bearing characteristic dataset mainly consists of numerical data. Increase data samples by means of data interpolation to make the model perform more stably in different scenarios. The specific processing process is as follows:

[0054] Outlier handling is to remove outliers in the bearing characteristic dataset using the 3σ principle to ensure the accuracy of the bearing characteristic dataset. It is obtained by calculating the mean μ and standard deviation σ of each parameter in the ultra-high pressure axial piston pump bearing characteristic dataset. The specific expression is:

[0055]

[0056] where μ is the data mean in the ultra-high pressure axial piston pump bearing characteristic dataset, N is the number of data in the ultra-high pressure axial piston pump bearing characteristic dataset, x i is the i-th data in the ultra-high pressure axial piston pump bearing characteristic dataset, and σ is the data standard deviation in the ultra-high pressure axial piston pump bearing characteristic dataset.

[0057] Perform normalization processing on the ultra-high pressure axial piston pump bearing characteristic dataset after outlier handling to avoid the influence of dimensional differences on the model. Since the numerical ranges of different input variables may vary greatly, directly inputting the ultra-high pressure axial piston pump bearing characteristic dataset into the neural network will cause some features to have too much influence on the training. Therefore, the dataset is scaled to between [0,1]. The specific expression is:

[0058]

[0059] where x is the original data in the ultra-high pressure axial piston pump bearing characteristic dataset, x min and x max are the minimum and maximum values in the ultra-high pressure axial piston pump bearing characteristic dataset respectively, and x ′ is the data after normalization of the ultra-high pressure axial piston pump bearing characteristic dataset.

[0060] S3. According to the characteristics of the ultra-high pressure axial piston pump bearing characteristic dataset, in order to model complex non-linear relationships through machine learning methods and improve the prediction accuracy of the ultra-high pressure axial piston pump performance, construct a multi-layer perceptron (MLP) model to predict the ultra-high pressure axial piston pump bearing characteristics and obtain an ultra-high pressure axial piston pump bearing characteristic evaluation model. The specific steps include:

[0061] S31. Construct the input layer: Determine the number of neurons in the input layer based on the dimensions of the independent variable parameters of the key load-bearing components of the ultra-high pressure axial piston pump. The input layer receives the operating conditions parameters, geometric parameters, and material parameters of each load-bearing component in the ultra-high pressure axial piston pump load characteristic dataset as the independent variable parameters of the ultra-high pressure axial piston pump load characteristic evaluation model.

[0062] S32. Construct four hidden layers: The first hidden layer has 128 neurons, which is responsible for learning the complex associations between the data in the ultra-high pressure axial piston pump load characteristic dataset and extracting high-dimensional features, such as non-linear relationships and local patterns; the second hidden layer has 64 neurons, which is responsible for further screening and extracting the key features in the ultra-high pressure axial piston pump load characteristics and reducing the impact of redundant information. The third hidden layer has 32 neurons, which is responsible for further reducing the data dimension, removing noise, and retaining the most critical information for the target variable, i.e., the load-bearing capacity of each component. The fourth hidden layer provides a compact feature representation for the final output layer to ensure that the model can accurately predict.

[0063] The load characteristics of the ultra-high pressure axial piston pump involve multi-physics field coupling such as fluid mechanics, contact mechanics, and tribology. The relationships between its input parameters such as pressure, velocity, and temperature and the load-bearing capacity are highly non-linear. Using ReLU as the non-linear activation function can help the neural network learn the complex input-output mapping relationship and improve the learning ability of the ultra-high pressure axial piston pump load characteristic evaluation model for non-linear features. Therefore, ReLU rectified linear unit operations are performed after each hidden layer. When x > 0, where x is the output of the neurons in the previous layer, the gradient of ReLU is always 1, and there will be no gradient vanishing problem, ensuring that the hidden layer can be effectively trained and enhancing the learning ability of the ultra-high pressure axial piston pump load characteristic evaluation model.

[0064] To prevent the ultra-high pressure axial piston pump load characteristic evaluation model from overfitting, Dropout random inactivation layers are added after ReLU in the first, second, third, and fourth hidden layers. The Dropout(0.3) layers after the first and second hidden layers are used to reduce the overfitting of the complex features of the ultra-high pressure axial piston pump load characteristic evaluation model. The Dropout(0.2) layer after the third hidden layer allows information to flow during the training process, and no Dropout is added to the last layer to ensure stable prediction.

[0065] When Dropout acts on neurons:

[0066] h′ = h·M

[0067] Among them, h is the output of the original neuron, M is a random mask matrix that follows the Bernoulli distribution Bernoulli(1 - p), each neuron is discarded with probability p, and p is the Dropout rate. For example, 0.3 means that 30% of the neurons are randomly ignored in each training iteration.

[0068] S33. Construct the output layer: Obtain the number of neurons in the output layer according to the dimension of the target variable parameters of the key load-bearing components of the ultra-high pressure axial piston pump.

[0069] S34. Use the root mean square error RMSE as the loss function to optimize the prediction accuracy of the ultra-high pressure axial piston pump load-bearing characteristic evaluation model. The evaluation of the load-bearing capacity of the ultra-high pressure axial piston pump belongs to a regression problem, and its goal is to predict a continuous variable such as the ultimate load-bearing pressure, oil film thickness, etc. RMSE is a commonly used loss function suitable for regression tasks, which can effectively measure the deviation between the predicted value and the true value, making the optimization direction of the model clear. The specific expression of the root mean square error RMSE is:

[0070]

[0071] Among them, n is the number of sample data in the ultra-high pressure axial piston pump load-bearing characteristic data set, and y i is the actual value of the ultra-high pressure axial piston pump load-bearing characteristic data set (load-bearing capacity parameters in experimental data or simulation data, such as ultimate load-bearing pressure, oil film thickness, etc.), is the predicted value of the model, represents the squared error of a single sample.

[0072] For the highly non-linear ultra-high pressure axial piston pump load-bearing characteristic evaluation model, use the adaptive learning rate mechanism of the Adam optimizer, with a learning rate of 0.001, to optimize the non-linear relationship between the ultra-high pressure axial piston pump load-bearing parameters, and adopt a step-by-step decay strategy to improve the stability of the ultra-high pressure axial piston pump load-bearing characteristic evaluation model training. The parameter update rule of the Adam optimizer is:

[0073] m t = β1m t-1 +(1 - β1)g t

[0074]

[0075] Among them, g t is the current gradient during the training process of the ultra-high pressure axial piston pump load-bearing characteristic evaluation model, and m t is the first-order moment gradient mean, used to simulate the effect of the momentum method, and v tis the mean of the squared second-moment gradient, used to adaptively adjust the learning rate. β1 and β2 are exponential decay coefficients, usually β1 = 0.9, β2 = 0.999. η is the learning rate, such as 0.001, and ∈ is a small value to prevent division-by-zero errors, generally taking 10^-8.

[0076] Batch training is an optimization method between full-gradient descent and stochastic gradient descent. Since the dataset of the load-bearing characteristics of ultra-high-pressure axial piston pumps is usually not particularly large, it is suitable for mini-batch training. Using batch training, when updating the parameters of the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model each time, a small batch of samples is used to improve the training efficiency of the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model. The specific expression of batch training is:

[0077]

[0078] where θ t is the parameter of the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model after the t-th iteration, m is the batch size (such as 32), is the gradient of each sample, and η is the learning rate.

[0079] Using the Early Stopping mechanism to prevent overfitting and ensure that the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model can generalize well under actual working conditions. Further, when there is no obvious improvement in the validation set for 10 consecutive rounds, the training stops to prevent overfitting. Early Stopping is a regularization strategy used to prevent the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model from overfitting on the training set. When the loss function of the validation set has not improved significantly for N consecutive rounds (such as 10 rounds), the training stops. The ultra-high-pressure axial piston pump load-bearing characteristics evaluation model usually relies on finite element simulation and experimental data, and the dataset size is limited, so overfitting is likely to occur. Therefore, using Early Stopping can prevent the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model from overfitting the simulation data and ensure that the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model can generalize well under actual working conditions.

[0080] K-fold cross-validation is a technique used to evaluate the generalization ability of a model. The dataset is split into K parts. Each time, K - 1 parts are selected as the training set, and the remaining 1 part is used as the test set. The training is repeated K times, and finally the average value is taken. Using K-fold cross-validation to improve the reliability of the ultra-high-pressure axial piston pump load-bearing characteristics evaluation model. The specific expression of K-fold cross-validation is:

[0081]

[0082] [[ID=2--5]]where K is the number of folds, which takes the value of 5 in a preferred embodiment of the present invention, and Test Score i is the model evaluation score for the i-th fold.

[0083] Due to the relatively limited data in the load-bearing characteristic dataset of the ultra-high pressure axial piston pump, K-fold cross-validation can make full use of all the data to improve the reliability of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump.

[0084] S4. Evaluate and test the generalization ability of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump using the validation set. The specific operation steps are as follows:

[0085] Validation set evaluation: Evaluate the performance of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump on unseen data by using the independent validation set in the load-bearing characteristic dataset of the ultra-high pressure axial piston pump. The main evaluation indicators include: Coefficient of determination: Measure the fitting degree of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump to the actual data, and evaluate whether the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump can accurately describe the load-bearing capacity characteristics. Root mean square error: As the main error evaluation indicator, it reflects the average error level of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump. Mean absolute error: As an auxiliary indicator, it measures the stability of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump and avoids the influence of outliers.

[0086] Generalization ability test: To evaluate the stability of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump under different working conditions, use different working condition combinations such as high temperature, high pressure, low flow rate, etc. to test the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump, and analyze the changes of the prediction error under different working conditions. Comparison of the output results with the actual situation: Compare the prediction results of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump with the actual experimental or simulation results, and analyze the feasibility of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump in practical applications.

[0087] By testing under extreme working conditions such as high temperature, high pressure, and low flow rate, it can be ensured that the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump is not only effective under ideal conditions but also can adapt to the changing actual working environment. The generalization ability test helps to verify the sensitivity of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump to input changes, ensure that reliable prediction results can be provided under various working conditions, and thus improve the application performance of the ultra-high pressure axial piston pump under complex working conditions. By testing the performance of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump under different working conditions, overfitting problems can be effectively discovered and avoided, ensuring that the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump has good generalization ability and avoiding the situation of being only applicable to the training data.

[0088] S5. Input the design parameters into the evaluation model of the load-bearing characteristics of the ultra-high pressure axial piston pump and optimize them to obtain the load-bearing result data of the ultra-high pressure axial piston pump. Take the minimum value of the ultimate load-bearing pressure among all components as the maximum ultimate load-bearing pressure of the ultra-high pressure axial piston pump. Specifically, the design parameters include the geometric structures, material properties, and operating condition parameters of the piston pair, slipper pair, one-way flow valve, and main shaft. The specific implementation steps are as follows:

[0089] Input the geometric structures, material properties, and operating condition parameters of the piston pair, slipper pair, one-way flow valve, and main shaft into the evaluation model of the load-bearing characteristics of the ultra-high pressure axial piston pump respectively to obtain the ultimate load-bearing pressure P1 of the piston pair, which represents the maximum pressure that the piston pair can bear, the oil film thickness of the piston pair, the friction coefficient of the piston pair, and the load-bearing pressure distribution of the piston pair; the ultimate load-bearing pressure P2 of the slipper pair, which represents the maximum pressure that the slipper pair can bear, the oil film thickness of the slipper pair, the friction coefficient of the slipper pair, and the load-bearing pressure distribution of the slipper pair; the ultimate load-bearing pressure P3 of the one-way flow valve, which represents the maximum pressure that the one-way flow valve can bear and the response speed of the one-way flow valve; the ultimate load-bearing pressure P4 of the main shaft, which represents the maximum pressure that the main shaft can bear, the friction coefficient of the main shaft, and the load-bearing pressure distribution of the main shaft.

[0090] Based on the load-bearing characteristics of the piston pair, slipper pair, one-way flow valve, and main shaft in the obtained ultra-high pressure axial piston pump, analyze the load-bearing capacity and defect positions of the ultra-high pressure axial piston pump in the current design. If it does not meet the design requirements, re-input the design parameters into the evaluation model of the load-bearing characteristics of the ultra-high pressure axial piston pump for calculation. If it meets the design requirements, compare the magnitudes of the ultimate load-bearing pressures of the piston pair, slipper pair, one-way flow valve, and main shaft, and take the minimum value P5 of the ultimate load-bearing pressures among all components as the final load-bearing capacity of the ultra-high pressure axial piston pump to ensure that the design meets the safety and stability under ultra-high pressure conditions.

[0091] Starting from the load-bearing characteristics of the piston pair, slipper pair, one-way flow valve, and main shaft through the above steps, collect load-bearing data based on computational fluid dynamics technology, finite element simulation technology, and friction and wear, etc. Build an evaluation model for the load-bearing capacity of each component through neural network technology to comprehensively evaluate the load-bearing capacity of the piston pump. Through the calculation of the above process, obtain the ultimate load-bearing pressure of the ultra-high pressure axial piston pump with a shorter design cycle, and improve the safety, stability, and service life of the fluid transmission system. At the same time, the steps of the present invention can be adjusted, combined, and deleted according to actual needs.

[0092] In a specific embodiment of the present invention, evaluation models for the load-bearing characteristics of the piston pair, slipper pair, one-way flow valve, and main shaft in the ultra-high pressure axial piston pump are built respectively.

[0093] The input layer of the bearing characteristic evaluation model of the plunger pair is provided with 13 neurons. The hidden layer is divided into three layers. The first hidden layer is provided with 64 neurons, the second hidden layer is provided with 32 neurons, the third hidden layer is provided with 16 neurons, and the output layer is provided with 4 neurons. The parameters of the plunger pair in the design parameters are input into the bearing characteristic evaluation model of the plunger pair to obtain the ultimate bearing pressure P1 of the plunger pair.

[0094] The input layer of the bearing characteristic evaluation model of the slipper pair is provided with 14 neurons. The hidden layer is divided into three layers. The first hidden layer is provided with 64 neurons, the second hidden layer is provided with 32 neurons, the third hidden layer is provided with 16 neurons, and the output layer is provided with 4 neurons. The parameters of the slipper pair in the design parameters are input into the bearing characteristic evaluation model of the slipper pair to obtain the ultimate bearing pressure P2 of the slipper pair.

[0095] The input layer of the bearing characteristic evaluation model of the unidirectional flow valve is provided with 13 neurons. The hidden layer is divided into four layers. The first hidden layer is provided with 64 neurons, the second hidden layer is provided with 32 neurons, the third hidden layer is provided with 16 neurons, and the output layer is provided with 2 neurons. The parameters of the unidirectional flow valve in the design parameters are input into the bearing characteristic evaluation model of the unidirectional flow valve to obtain the ultimate bearing pressure P3 of the unidirectional flow valve.

[0096] The input layer of the bearing characteristic evaluation model of the main shaft is provided with 11 neurons. The hidden layer is divided into four layers. The first hidden layer is provided with 128 neurons, the second hidden layer is provided with 64 neurons, the third hidden layer is provided with 32 neurons, the fourth hidden layer is provided with 16 neurons, and the output layer is provided with 3 neurons. The parameters of the main shaft in the design parameters are input into the bearing characteristic evaluation model of the main shaft to obtain the ultimate bearing pressure P4 of the main shaft.

[0097] The minimum value among the ultimate bearing pressures P1 of the plunger pair, P2 of the slipper pair, P3 of the unidirectional flow valve, and P4 of the main shaft obtained above in the ultra-high pressure axial piston pump is used as the final bearing capacity of the ultra-high pressure axial piston pump in this specific embodiment to ensure that the designed ultra-high pressure axial piston pump meets the safety and stability under ultra-high pressure conditions.

[0098] The embodiments described above are only used to describe the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An optimization method for the bearing pressure of an ultra-high pressure axial piston pump for heavy machinery, characterized in that, It includes the following steps: S1. According to the key load-bearing components of the ultra-high pressure axial piston pump in heavy machinery, collect the load-bearing characteristic data sets of the piston pair, slipper pair, one-way flow valve, and main shaft respectively, and verify their accuracy; S2. Perform outlier processing, data normalization, and data augmentation on the load-bearing characteristic data sets obtained in step S1, and divide the load-bearing characteristic data sets into a training set and a test set; S3. According to the characteristics of the load-bearing characteristic data sets of the ultra-high pressure axial piston pump, construct a multi-layer perceptron model to obtain the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump; According to the dimensions of the independent variable parameters and target variable parameters of the key load-bearing components, obtain the number of neurons in the input layer and output layer respectively. Construct a hidden layer according to the non-linearity between the load-bearing characteristics and load-bearing capacity of the ultra-high pressure axial piston pump, and train the model using relevant functions; S4. Use the validation set to evaluate the load-bearing characteristic evaluation model and test its generalization ability; S5. Input the design parameters into the load-bearing characteristic evaluation model and optimize them to obtain the load-bearing result data of the ultra-high pressure axial piston pump. Take the minimum value of the ultimate load-bearing pressure among the key load-bearing components as the maximum ultimate load-bearing pressure of the ultra-high pressure axial piston pump.

2. The method for optimizing the bearing pressure of an ultra-high pressure axial piston pump for heavy machinery according to claim 1, wherein: Step S1 includes the following sub-steps: According to the independent variable parameters of the load-bearing characteristics of the piston pair, slipper pair, and main bearing respectively, obtain the oil film thickness, load-bearing pressure distribution, ultimate load-bearing pressure, and / or friction coefficient under ultra-high pressure conditions in the target variable parameters; According to the independent variable parameters of the load-bearing characteristics of the one-way flow valve, simulate the flow field state of the one-way flow valve under ultra-high pressure load and the opening and closing time of the valve core under pressure switching to obtain the response speed in the target variable parameters; Simulate the structural stress and deformation of the one-way flow valve under ultra-high pressure load to obtain the ultimate load-bearing pressure in the target variable parameters; Use on-site tests to verify the obtained load-bearing characteristic data sets.

3. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 1, characterized in that: Step S3 includes the sub-step of constructing a hidden layer. Use ReLU as the non-linear activation function, and add a Dropout layer after ReLU in the first hidden layer, second hidden layer, third hidden layer, and fourth hidden layer to prevent overfitting of the load-bearing characteristic evaluation model of the ultra-high pressure axial piston pump.

4. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 1, wherein: The independent variable parameters of the load-bearing characteristics of the piston pair include working pressure, piston speed, hydraulic oil viscosity, hydraulic oil temperature, piston diameter, piston head curvature radius, cylinder bore fit clearance, surface coating thickness, cylinder block elastic modulus, piston elastic modulus, cylinder block hardness, and piston hardness.

5. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 1, wherein: The independent variable parameters of the load-bearing characteristics of the one-way flow valve include inlet pressure, outlet pressure, flow rate, temperature, valve core opening, valve core diameter, valve seat diameter, valve core-valve seat clearance, valve core mass, valve core elastic modulus, valve seat elastic modulus, valve core hardness, and valve seat hardness.

6. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 1, wherein: In step S2, the outlier processing is obtained by calculating the mean μ and standard deviation σ of each parameter in the load-bearing characteristic data set of the ultra-high pressure axial piston pump. Specifically: Among them, μ is the mean value of the data in the load characteristic dataset of the ultra-high pressure axial piston pump, N is the number of data in the load characteristic dataset of the ultra-high pressure axial piston pump, and x i is the i-th data in the load characteristic dataset of the ultra-high pressure axial piston pump, and σ is the standard deviation of the data in the load characteristic dataset of the ultra-high pressure axial piston pump.

7. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 6, wherein: In step S2, perform normalization processing on the load-bearing characteristic data set of the ultra-high pressure axial piston pump after outlier processing. Specifically: Among them, x is the original data in the load-bearing characteristic dataset of the ultra-high pressure axial piston pump, x min and x max are respectively the minimum value and the maximum value in the load-bearing characteristic dataset of the ultra-high pressure axial piston pump, x ′ is the data after normalization of the load-bearing characteristic dataset of the ultra-high pressure axial piston pump.

8. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 6, characterized in that: Step S3 further includes using the root mean square error as a loss function to optimize the prediction accuracy of the bearing characteristic evaluation model of the ultra-high pressure axial piston pump, and using an adaptive learning rate mechanism to optimize the non-linear relationship between the bearing parameters of the ultra-high pressure axial piston pump.

9. The optimized method for bearing pressure of the ultra-high pressure axial piston pump for heavy machinery according to claim 1, wherein: In step S5, the design parameters include the geometric structures, material properties, and operating condition parameters of the plunger pair, slipper pair, unidirectional flow valve, and main shaft.

Images