Hydraulic valve flow soft measurement method considering service performance degradation

Abstract

The invention discloses a hydraulic valve flow soft measurement method considering service performance degradation, which comprises the following steps: fitting a valve port flow area change function related to control voltage before performance degradation according to a three-dimensional structure of a hydraulic valve before performance degradation; a valve port flow area change function related to the control voltage at different hydraulic oil temperatures after performance degradation is determined; constructing a valve port flow coefficient prediction model based on the hydraulic oil temperature before performance degradation, the control voltage of the hydraulic valve and the inlet and outlet pressure difference; after performance degradation, predicting a valve port flow coefficient based on a valve port flow coefficient prediction model according to the current hydraulic oil temperature, the control voltage of the hydraulic valve and the inlet and outlet pressure difference; and according to the predicted valve port flow coefficient, the current control voltage of the hydraulic valve, the valve port flow area before and after performance degradation under the hydraulic oil temperature and the current inlet and outlet pressure difference of the hydraulic valve, the actual flow of the hydraulic valve after performance degradation is determined. The flow can be accurately measured after the performance of the hydraulic valve is degraded.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of hydraulic valve flow side face, and particularly relates to a hydraulic valve flow soft measurement method considering service performance degradation. BACKGROUND

[0002] As a core control component of a hydraulic system, the performance of a hydraulic valve directly determines the stability, reliability and control accuracy of the entire system. In a hydraulic system, accurate measurement of the valve port flow is the key to realizing precise control, state monitoring and fault diagnosis. At present, the flow measurement of a hydraulic valve generally relies on traditional physical flow meters. However, such measurement methods have obvious limitations: the installation of a flow meter will increase the complexity and manufacturing cost of the system; its physical volume may cause difficulties in arrangement in a compact system with limited space; in addition, some flow meters also have problems such as slow response speed and limited measurement accuracy.

[0003] The soft measurement models in existing researches are mostly established based on the data of a hydraulic valve in a specific, especially un-worn state. In actual industrial applications, due to oil pollution, mechanical friction and cavitation, etc., the key friction pairs such as the valve core and the valve sleeve of the hydraulic valve will inevitably wear and degrade during long-term service. Wear will directly change the flow-through geometric area and shape of the valve port, causing significant drift of the flow-pressure difference-opening characteristic. At this time, the flow prediction model established before wear will deviate significantly, and even completely fail. If the accuracy of the flow soft measurement after wear is to be improved, the traditional method usually needs to disassemble the valve from the system, and re-collect complete data covering all working conditions on a test bench to rebuild the model. This process is not only time-consuming and laborious, but also has high economic cost, and for many devices that need to be continuously operated or have a closed structure, it is often not operationally feasible. Therefore, when the performance of the hydraulic valve degrades, it is difficult to accurately predict the actual flow output of the hydraulic valve by using the flow prediction model established based on the data before performance degradation. SUMMARY

[0004] The application provides a hydraulic valve flow soft measurement method considering service performance degradation, to solve the problem that when the performance of the hydraulic valve degrades, it is difficult to accurately predict the actual flow output of the hydraulic valve by using the flow prediction model established based on the data before performance degradation.

[0005] According to a first aspect of an embodiment of the application, a hydraulic valve flow soft measurement method considering service performance degradation is provided, comprising:

[0006] According to the three-dimensional structure of the hydraulic valve before performance degradation, a valve port flow area change function g1(u) related to a control voltage u before performance degradation is fitted;

[0007] For each different hydraulic oil temperature, based on the inlet and outlet pressure difference and the actual flow of the hydraulic valve obtained at the corresponding hydraulic oil temperature after performance degradation, a valve port flow area change function g2(u) related to the control voltage u at the hydraulic oil temperature is determined;

[0008] Based on the hydraulic oil temperature, the control voltage and the inlet and outlet pressure difference of the hydraulic valve obtained before performance degradation, a valve port flow coefficient prediction model is constructed;

[0009] According to the current obtained hydraulic oil temperature, the control voltage and the inlet and outlet pressure difference of the hydraulic valve after performance degradation, the valve port flow coefficient is predicted based on the valve port flow coefficient prediction model;

[0010] According to the current obtained control voltage of the hydraulic valve, the valve port flow area before performance degradation is determined, and according to the current obtained hydraulic oil temperature and the control voltage of the hydraulic valve, the valve port flow area after performance degradation is determined;

[0011] According to the predicted valve port flow coefficient, the determined valve port flow area before and after performance degradation, and the current obtained inlet and outlet pressure difference of the hydraulic valve, the actual flow of the hydraulic valve after performance degradation is determined.

[0012] Optionally, the valve port flow area change function g1(u) related to the control voltage u before performance degradation is determined according to the following formula:

[0013]

[0014] Wherein g1(u)=a1+b1u, a1 and b1 are both constants, Q1(T,u,ΔP) is the flow of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, C d1 is the valve port flow coefficient before performance degradation, ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, and ρ is the density of hydraulic oil.

[0015] Optionally, based on the inlet and outlet pressure difference and the flow of the hydraulic valve obtained at the corresponding hydraulic oil temperature after performance degradation, the valve port flow area change function g2(u) related to the control voltage u at the hydraulic oil temperature is determined according to the following formula:

[0016]

[0017] Wherein g2(u)=a2+b2u, a2 and b2 are both constants, Q2(T,u,ΔP) is the flow of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, C d2 is the valve port flow coefficient after performance degradation, ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, and ρ is the density of hydraulic oil.

[0018] Optionally, the valve port flow coefficient model is a neural network model, the neural network comprising an input layer, a plurality of hidden layers and an output layer, the non-linear mapping from the input layer to the first hidden layer being:

[0019]

[0020] where a n is the nth neuron of the first hidden layer, n is an integer greater than 0 and less than or equal to the number of neurons m in each hidden layer, u is the control voltage of the hydraulic valve, ΔΡ is the pressure difference between the inlet and outlet of the hydraulic valve, T is the temperature, w 1n is the weight of the valve port opening to the nth neuron of the first hidden layer, w 2n is the weight of the pressure difference between the inlet and outlet to the nth neuron of the first hidden layer, w 3n is the weight of the temperature to the nth neuron of the first hidden layer, is the threshold value of the input layer to the nth neuron of the first hidden layer, and f(x) is the activation function corresponding to the input layer to the first hidden layer;

[0021] The non-linear mapping between the adjacent two hidden layers is:

[0022]

[0023] where b n is the nth neuron in the hidden layer b, w mn is the weight of the mth neuron in the hidden layer b-1 to the nth neuron in the hidden layer b, is the threshold value of the nth neuron in the hidden layer b, and f(x) is the activation function between the two hidden layers;

[0024] The non-linear mapping from the last hidden layer to the output layer is:

[0025] C′ d1 = w1b1 + w2b2 + … + w m b m + v

[0026] where C' d1 is the predicted flow coefficient, w1, w2, …, w n is the weight from each neuron in the last hidden layer to the neuron in the output layer, and v is the threshold value of the output neuron.

[0027] Optionally, the valve port flow coefficient prediction model is constructed according to the following steps:

[0028] Step S310, design the number of hidden layers and the number of neurons in each hidden layer in the neural network: the number of hidden layers is taken in a set range of layer numbers, and the number of neurons is taken in a set range of neuron numbers, wherein each layer number in the set range of layer numbers is combined with each number of neurons in the set range of neuron numbers respectively;

[0029] Step S320, for each combination, when the hydraulic valve does not appear service performance degradation, the temperature, the valve opening and the pressure difference between the inlet and outlet of the hydraulic valve under each working condition obtained are taken as a group of samples, each group of samples is input into the neural network corresponding to the combination, the predicted valve flow coefficient of each group of samples is obtained, and the root mean square error and the average absolute error of the valve flow coefficient corresponding to the combination are calculated according to the actual valve flow coefficient and the predicted valve flow coefficient of each group of samples.

[0030] Step S330, determine the combination with the minimum root mean square error and average absolute error in all combinations, and take the number of layers and the number of neurons in the combination as the number of hidden layers and the number of neurons in each hidden layer in the neural network.

[0031] Optionally, when constructing the valve flow coefficient prediction model, it further includes:

[0032] Step S340, when the number of hidden layers and the number of neurons in each hidden layer of the neural network is fixed, each group of samples obtained when the hydraulic valve does not appear service performance degradation is input into the neural network for training to obtain the predicted valve flow coefficient of each group of samples, and the network performance of the neural network is determined according to the actual flow coefficient and the predicted flow coefficient of each group of samples.

[0033] Step S350, judge whether the network performance of the neural network meets the standard, if yes, obtain the optimal weight matrix and the optimal threshold matrix of the neural network, otherwise, adjust the weight matrix and the threshold matrix in the neural network, and return to execute step S340;

[0034] The weight matrix includes the weight of each input parameter in the input layer to each neuron in the first layer of hidden layers, the weight of each neuron in the hidden layer to each neuron in the next layer of hidden layers, and the weight of each neuron in the last layer of hidden layers to the predicted flow in the output layer; the threshold matrix includes the threshold of the input layer to each neuron in the first layer of hidden layers, the threshold of the hidden layer to each neuron in the next layer of hidden layers, and the threshold of the last layer of hidden layers to the predicted flow in the output layer.

[0035] Optionally, the network performance includes the root mean square relative error and the average relative error of the valve flow coefficient.

[0036] Optionally, the method further includes: before constructing the neural network, normalizing the obtained hydraulic oil temperature, hydraulic valve control voltage, inlet and outlet pressure difference and actual flow rate, as well as the output predicted valve flow coefficient, and limiting the values ​​to [0.0001, 0.9999].

[0037] Optionally, the actual flow rate Q2(T,u,ΔP) of the hydraulic valve after performance degradation can be determined according to the following formula:

[0038]

[0039] Where C' d1 The valve orifice flow coefficient is the predicted flow coefficient. g1(u) is the valve orifice flow area determined before performance degradation based on the current control voltage of the hydraulic valve. g2(u) represents the valve orifice flow area determined after performance degradation based on the current hydraulic oil temperature and the control voltage of the hydraulic valve. ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained after performance degradation. ρ is the hydraulic oil density.

[0040] Optionally, the control voltage can be replaced by the valve opening.

[0041] The beneficial effects of this invention are:

[0042] 1. This invention does not establish a flow prediction model when measuring hydraulic valve flow rate. Instead, it establishes a flow coefficient prediction model. First, it determines the valve orifice flow area change function related to the control voltage before performance degradation and the valve orifice flow area change function related to the control voltage under different hydraulic oil temperatures after performance degradation. The valve orifice flow coefficient prediction model of this invention is constructed based on the hydraulic oil temperature, hydraulic valve control voltage, and inlet / outlet pressure difference obtained before performance degradation. When determining the actual flow rate of the hydraulic valve after performance degradation, it only needs to rely on a small amount of data obtained after performance degradation and fit the valve orifice flow area change function related to the control voltage under different hydraulic oil temperatures. No further data is required. After the hydraulic valve's performance degrades, a large amount of data is recollected and the model is retrained. This enables rapid updates to the full-condition flow prediction model, greatly saving time and economic costs associated with data acquisition and model training. It also solves the problems of downtime losses and operational feasibility caused by disassembling the valve body. Furthermore, this invention can accurately predict the flow coefficient affected by multiple factors through a flow coefficient prediction model. It then uses a flow area function with clear physical meaning to characterize the structural changes caused by performance degradation. This allows the model to maintain high-precision prediction capabilities even after performance degradation, and the prediction process has a solid physical basis, which can improve the flow prediction accuracy and generalization ability of this invention.

[0043] 2、The valve port flow coefficient prediction model of the present application is a neural network model, which combines the powerful nonlinear mapping capability of the neural network with the classic valve port flow physical formula, and can further improve the flow prediction accuracy;

[0044] 3、The present application designs the number of hidden layers and the number of neurons in the hidden layer of the neural network, which can better learn the characteristics corresponding to different tasks; the design of the neural network structure based on the actual flow coefficient calibration data can make the obtained neural network model more truly reflect the actual flow coefficient characteristics of the hydraulic valve, thereby improving the flow coefficient prediction accuracy and generalization ability of the model;

[0045] 4、The present application uses the root mean square relative error and the average relative error of the valve port flow coefficient to verify the accuracy of the network performance of the neural network, which can improve the accuracy of network performance verification;

[0046] 5、Before constructing the neural network, the obtained hydraulic oil temperature, control voltage of the hydraulic valve, pressure difference between inlet and outlet, actual flow and predicted valve port flow coefficient are normalized, which can prevent different dimensions and data scales from affecting the model performance, is conducive to the gradient stability of the activation function during training, and at the same time, improves the numerical stability and prevents the neurons from being saturated in advance in the neural network. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is an embodiment flow chart of the hydraulic valve flow soft measurement method considering service performance degradation of the present application;

[0048] Figure 2 is a structural schematic diagram of the existing valve control hydraulic system test bench of the present application;

[0049] Figure 3 is a neural network structure schematic diagram of the present application;

[0050] Figure 4 is a root mean square error design table of the number of hidden layers and the number of neurons in each hidden layer in the neural network of the present application under different structures;

[0051] Figure 5 is an average absolute error design table of the number of hidden layers and the number of neurons in each hidden layer in the neural network of the present application under different structures. DETAILED DESCRIPTION

[0052] In order to make the person skilled in the art better understand the technical solutions in the embodiments of the present application, and make the above-mentioned purposes, characteristics and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application are further described in detail below with reference to the drawings.

[0053] In the description of the present application, unless otherwise specified and limited, it is necessary to explain that the term "connection" should be understood broadly, for example, it can be mechanical connection or electrical connection, it can be the communication inside two elements, it can be direct connection or indirect connection through intermediate medium, and the specific meaning of the above-mentioned term can be understood according to the specific circumstances by the ordinary skilled in the art.

[0054] Referring to Figure 1 An embodiment flow chart of the hydraulic valve flow soft measurement method considering service performance degradation of the present application is provided. The method can include:

[0055] Step S100, fitting the valve port flow area change function g1(u) related to the control voltage u before performance degradation according to the three-dimensional structure of the hydraulic valve before performance degradation. In this step, the valve port flow area change function g1(u) related to the control voltage u before performance degradation can be determined according to the following formula:

[0056]

[0057] Wherein g1(u)=a1+b1u, a1 and b1 are both constants, Q1(T,u,ΔP) is the flow of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, C d1 is the valve port flow coefficient before performance degradation, ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, and ρ is the density of the hydraulic oil. When the hydraulic valve does not appear performance degradation, its valve port flow coefficient C d1 can be taken as the engineering experience value, C d1 =0.7, according to the characteristics of the hydraulic oil, ρ can be taken as ρ=900kg / m 3 . In the above formula, C d1 and ρ are constant, and Q1(T,u,ΔP) and ΔP can be measured when u changes, and only a1 and b1 are unknown parameters, so by establishing two equations to solve, the parameters a1 and b1 corresponding to the hydraulic oil temperature can be obtained.

[0058] Step S200, for each different hydraulic oil temperature, based on the inlet and outlet pressure difference and the actual flow of the hydraulic valve obtained at the corresponding hydraulic oil temperature after performance degradation, the valve port flow area change function g2(u) related to the control voltage u at the hydraulic oil temperature is determined. In this step, the valve port flow area change function g2(u) related to the control voltage u at the hydraulic oil temperature can be determined according to the following formula:

[0059]

[0060] wherein g2(u)=a2+b2u, a2 and b2 are constants, Q2(T, u, ΔP) is the flow of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, C d2 is the valve port flow coefficient after performance degradation, ΔP is the pressure difference between the inlet and outlet of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, and σ is the density of the hydraulic oil. In the above formula, σ is constant, u changes, Q2(T, u, ΔP) and ΔP can be measured, and only C d2 , a2 and b2 are unknown parameters. For each hydraulic oil temperature, the parameters a2 and b2 at the hydraulic oil temperature can be fitted based on the least square algorithm.

[0061] The hydraulic oil temperature, the pressure difference between the inlet and outlet of the hydraulic valve, and the actual flow obtained in the present application can be as shown in the figure. Figure 2 wherein 1 is a hydraulic pump for providing hydraulic oil for the test device; 2 is a check valve for ensuring one-way flow of the hydraulic oil and preventing backflow; 3 is a high-pressure filter for filtering impurities and solid particles in the hydraulic oil in the inlet oil path through a filter core or filter medium; 4 is a pressure gauge for roughly indicating the inlet oil pressure controlled by the overflow valve 5, and a pressure sensor 13, 14 needs to be used for accurate measurement; 5 and 8 are both overflow valves, the overflow valve 5 is used to control the inlet oil pressure, the pressure loss caused by the flowmeter and the throttle valve can be ignored, and the pressure is considered as the inlet pressure of the proportional directional valve, the overflow valve 8 is used to control the outlet pressure of the proportional directional valve; 6 is a flowmeter for measuring the flow characteristics under different pressures, temperatures, valve port openings, and before and after the proportional directional valve 7 is worn; 7 is a proportional directional valve, which is a selected hydraulic valve considering service degradation, referred to as a test valve; 9 and 10 are cooling devices for cooling the return oil, 10 specifically is an electric motor connected to a fan, the electric motor drives the fan to send air to the air cooler 9 for cooling the oil to control the hydraulic oil temperature; 11 is a return oil filter for filtering impurities and solid particles in the hydraulic oil in the return oil path through a filter core or filter medium; 12 is a temperature sensor for measuring the oil tank temperature, since the temperature of the hydraulic oil in the entire circuit is balanced, the temperature measured is the temperature of the hydraulic oil passing through the entire hydraulic system, which is also the oil temperature passing through the proportional directional valve 7; 13 and 14 are two pressure sensors installed before and after the proportional directional valve 7 for measuring the absolute pressures before and after the proportional directional valve, and the pressure difference between the two ends of the proportional directional valve can be calculated by subtraction.

[0062] Step S300, based on the hydraulic oil temperature, the control voltage of the hydraulic valve, and the pressure difference between the inlet and outlet obtained before performance degradation, a valve port flow coefficient prediction model is constructed. In this step, the valve port flow coefficient model can be a neural network model, which is combined Figure 3As shown, the neural network can include an input layer, a plurality of hidden layers and an output layer, the input parameters of the input layer of the neural network can be the hydraulic oil temperature obtained before performance degradation, the control voltage of the hydraulic valve and the inlet and outlet pressure difference, the output parameters of the output layer are the valve port flow coefficient, the nonlinear mapping of the input layer of the neural network to the first layer of hidden layers can be:

[0063]

[0064] wherein a n is the nth neuron of the first layer of hidden layers, n is an integer greater than 0 and less than or equal to the number of neurons m in each layer of hidden layers, u is the control voltage of the hydraulic valve, ΔP is the inlet and outlet pressure difference of the hydraulic valve, T is the temperature, w 1n is the weight of the valve port opening to the nth neuron of the first layer of hidden layers, w 2n is the weight of the inlet and outlet pressure difference to the nth neuron of the first layer of hidden layers, w 3n is the weight of the temperature to the nth neuron of the first layer of hidden layers, is the threshold value of the input layer to the nth neuron of the first layer of hidden layers, and f(x) is the corresponding activation function of the input layer to the first layer of hidden layers;

[0065] The nonlinear mapping between the adjacent two hidden layers can be:

[0066]

[0067] wherein b n is the nth neuron in the hidden layer b, w mn is the weight of the mth neuron in the hidden layer b-1 to the nth neuron in the hidden layer b, is the threshold value of the nth neuron in the hidden layer b, and f(x) is the activation function between the two hidden layers;

[0068] The nonlinear mapping of the last layer of hidden layers to the output layer (usually using the Purelin linear activation function) can be:

[0069] C′ d1 = w1b1 + w2b2 + … + w m b m + v

[0070] wherein C' d1 is the predicted flow coefficient, w1, w2, …, w nis the weight from the last hidden layer to the output layer, and v is the threshold of the output neuron. The above forward propagation process is from the input layer to the hidden layer, and then to the output layer. The data of the previous layer of neurons is obtained by weighted summation and then adding a bias term to obtain the data of the next layer of neurons. The forward propagation process does not involve the learning process, and the back propagation needs to be used for model learning to update the weights w and the threshold v of each layer.

[0071] In addition, the establishment of the neural network can be divided into designing the number of hidden layers and the number of neurons in each hidden layer in the neural network, and determining the optimal weight matrix and the optimal threshold matrix of the neural network when the hydraulic pump does not appear service performance degradation. The valve port flow coefficient prediction model can be constructed according to the following steps:

[0072] Step S310, designing the number of hidden layers and the number of neurons in each hidden layer in the neural network: taking the number of hidden layers in the set number of layers range, and taking the number of neurons in the set number range, wherein each number of layers in the set number of layers range is combined with each number of neurons in the set number range.

[0073] Step S320, for each combination, taking the temperature, the valve port opening and the inlet and outlet pressure difference of the hydraulic valve under each working condition obtained when the hydraulic valve does not appear service performance degradation as a group of samples, inputting each group of samples into the neural network corresponding to the combination, obtaining the predicted valve port flow coefficient of each group of samples, and calculating the root mean square error and the average absolute error of the valve port flow coefficient corresponding to the combination according to the actual valve port flow coefficient and the predicted valve port flow coefficient of each group of samples. The root mean square error RMSE can be expressed as:

[0074]

[0075] wherein k is the number of samples, C d1i is the actual flow coefficient value of the i-th group of samples, and C d1i is the predicted flow coefficient value of the i-th group of samples.

[0076] The average absolute error MAE can be expressed as:

[0077]

[0078] wherein k is the number of samples, C d1i is the actual flow coefficient value of the i-th group of samples, and C d1i is the predicted flow coefficient value of the i-th group of samples.

[0079] Step S330, determine the combination with the minimum root mean square error and mean absolute error of flow in all combinations, and take the number of layers and the number of neurons in the combination as the number of hidden layers and the number of neurons in each hidden layer of the neural network. Figure 4 and Figure 5 As shown in the figure, when the number of hidden layers is 3 and the number of neurons is 8, the root mean square error RMSE and the mean absolute error MAE of the network are the smallest, and at this time, the structure performance of the neural network is the best.

[0080] The number of hidden layers and the number of neurons in the hidden layer of the neural network are designed, which can make it better to learn the corresponding features of different tasks; the design of the neural network structure based on the actual flow coefficient calibration data can make the obtained neural network model more truly reflect the actual flow coefficient characteristics of the hydraulic valve, thereby improving the flow coefficient prediction accuracy and generalization ability of the model. Therefore, when constructing the valve port flow coefficient prediction model, the present application can also include:

[0081] Step S340, when the number of hidden layers and the number of neurons in each hidden layer of the neural network is fixed, the obtained each group of sample is input into the neural network for training when the hydraulic valve does not appear service performance degradation, and the predicted valve port flow coefficient of each group of sample is obtained, and the network performance of the neural network is determined according to the actual flow coefficient and the predicted flow coefficient of each group of sample.

[0082] Step S350, judge whether the network performance of the neural network meets the standard, if yes, obtain the optimal weight matrix and the optimal threshold matrix of the neural network, otherwise, adjust the weight matrix and the threshold matrix in the neural network, and return to execute step S340.

[0083] The weight matrix includes the weight of each input parameter in the input layer to each neuron in the first layer of hidden layer, the weight of each neuron in the hidden layer to each neuron in the next hidden layer, and the weight of each neuron in the last layer of hidden layer to the predicted flow in the output layer; the threshold matrix includes the threshold of the input layer to each neuron in the first layer of hidden layer, the threshold of the hidden layer to each neuron in the next hidden layer, and the threshold of the last layer of hidden layer to the predicted flow in the output layer.

[0084] The network performance can include the root mean square relative error and the average relative error of the valve port flow coefficient. In order to verify the accuracy of the network performance of the neural network, it is not easy to clearly explain the prediction effect using the specific root mean square error and the average absolute error, and in order to represent the percentage of error, the root mean square relative error and the average relative error are introduced.

[0085] The root mean square relative error RRMSE can be expressed as:

[0086]

[0087] wherein is the average value of the actual flow coefficient.

[0088] The average relative error MRE can be expressed as:

[0089]

[0090] In this step, in order to prevent different dimensions and data scales from affecting the performance of the model, the method further comprises: before constructing the neural network, normalizing the obtained hydraulic oil temperature, control voltage of the hydraulic valve, pressure difference between the inlet and outlet, actual flow, and output predicted valve port flow coefficient. In order to accelerate the convergence speed of the model, ensure its stability and generalization ability. Normalization usually limits data in the interval [0, 1], and the processing method of the present application is to limit data in [0.0001, 0.9999] instead of the traditional [0, 1]. If the data after normalization is exactly 0 or 1, it will cause the derivative of the activation function to be too small, and the gradient cannot be effectively transmitted during back propagation. The advantage of this is to avoid the exact 0 or 1 after normalization, which is conducive to the gradient of the activation function during training, and at the same time, improves the numerical stability and prevents neurons from saturating in advance in the neural network.

[0091] The expression is:

[0092]

[0093] is the expression for normalizing each input parameter respectively; is the expression for normalizing the output layer predicted flow coefficient.

[0094] Step S400, according to the current obtained hydraulic oil temperature, control voltage of the hydraulic valve and pressure difference between the inlet and outlet, the valve port flow coefficient prediction model is used to predict the current valve port flow coefficient; according to the current obtained control voltage of the hydraulic valve, the valve port flow area before performance degradation is determined, and according to the current obtained hydraulic oil temperature and control voltage of the hydraulic valve, the valve port flow area after performance degradation is determined. Wherein, the current obtained hydraulic oil temperature, control voltage of the hydraulic valve and pressure difference between the inlet and outlet are input into the valve port flow coefficient prediction model, so as to predict the current valve port flow coefficient.

[0095] Step S500, according to the predicted valve port flow coefficient, the determined valve port flow area before and after performance degradation, and the current obtained pressure difference between the inlet and outlet of the hydraulic valve, the actual flow of the hydraulic valve after performance degradation is determined.

[0096] According to the valve flow equation, the output flow formula of the hydraulic valve under different working conditions before performance degradation (including wear, etc.) can be expressed as:

[0097]

[0098] After the performance degradation of the hydraulic valve, the port flow area and the flow coefficient corresponding to the control signal will change, and the output flow formula of the hydraulic valve under different working conditions can be expressed as:

[0099]

[0100] The relationship between the flow coefficient and the change function before and after the performance degradation of the hydraulic valve satisfies:

[0101] C d2 g1(u)=C d1 g2(u)

[0102] Thus, we can get:

[0103]

[0104] Substituting the above formula into the output flow formula of the hydraulic valve under different working conditions after performance degradation, we can get:

[0105]

[0106] Since C d1 is the flow coefficient obtained under narrow conditions, this flow formula can only predict the output flow of the hydraulic valve under partial working conditions after wear.

[0107] Since the neural network output is the predicted valve port flow coefficient C' d1 under a wide temperature range, the predicted flow coefficient value C' d1 is replaced with the flow coefficient value C d1 in the formula, and based on the improved flow formula, the actual flow Q2(T, u, ΔP) of the hydraulic valve under full working conditions after performance degradation can be predicted:

[0108]

[0109] where C' d1For the predicted valve flow coefficient, g1(u) is the valve flow area determined according to the current obtained control voltage of the hydraulic valve before performance degradation (wherein the current obtained control voltage of the hydraulic valve is input into the change function g1(u), so that the valve flow area corresponding to the current obtained control voltage of the hydraulic valve before performance degradation is obtained), g2(u) represents the valve flow area determined according to the current obtained hydraulic oil temperature and the control voltage of the hydraulic valve after performance degradation (wherein the change function g2(u) at the current obtained hydraulic oil temperature after performance degradation is first determined, and then the current obtained control voltage of the hydraulic valve is input into the change function g2(u), so that the valve flow area corresponding to the current obtained hydraulic oil temperature and the control voltage of the hydraulic valve after performance degradation is obtained), ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained after performance degradation, and ρ is the hydraulic oil density.

[0110] It should be noted that the execution sequence between the above steps S100 to S300 can be exchanged with each other, and the execution sequence is not limited by the step number. In addition, since the control voltage and the valve opening degree are in a proportional relationship, the control voltage can also be replaced by the valve opening degree when the valve flow coefficient prediction model is constructed and the actual flow of the hydraulic valve after performance degradation is determined.

[0111] As can be seen from the above embodiment, when the hydraulic valve flow is measured, the flow prediction model is not established, but the flow coefficient prediction model is established, and the valve flow area change function related to the control voltage before performance degradation and the valve flow area change function related to the control voltage at each different hydraulic oil temperature after performance degradation are first determined. The valve flow coefficient prediction model of the present application is constructed based on the hydraulic oil temperature obtained before performance degradation, the control voltage of the hydraulic valve and the inlet and outlet pressure difference. When the actual flow of the hydraulic valve after performance degradation is determined, only a small amount of data obtained after performance degradation is needed, and the valve flow area change function related to the control voltage at different hydraulic oil temperatures is fitted. It is not necessary to collect a large amount of data after performance degradation of the hydraulic valve and retrain the model. Therefore, the full working condition flow prediction model can be quickly updated, the time and economic cost consumed by data collection and model training are greatly saved, and the downtime loss and operation feasibility problem caused by disassembly of the valve body are solved. In addition, the flow coefficient affected by the coupling of multiple factors can be accurately predicted through the flow coefficient prediction model, and the structural changes caused by performance degradation are represented through the flow area function with clear physical meaning. The model can still maintain high prediction accuracy after performance degradation, and the prediction process has a solid physical basis, which can improve the flow prediction accuracy and generalization ability of the present application. The valve flow coefficient prediction model of the present application is a neural network model, which combines the powerful nonlinear mapping ability of neural network with the classical valve flow physical formula, and can further improve the flow prediction accuracy.

[0112] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0113] It is to be understood that the application is not limited to the precise construction herein disclosed and shown in the drawings, and that various modifications and changes can be effected therein by those skilled in the art without departing from the scope of the application. The scope of the application is to be limited only by the appended claims.

Claims

1. A soft measurement method for the flow rate of a hydraulic valve considering service performance degradation, characterized in that, include: Based on the three-dimensional structure of the hydraulic valve before performance degradation, fit the valve orifice flow area change function g1(u) related to the control voltage u before performance degradation; For each different hydraulic oil temperature, based on the inlet and outlet pressure difference and actual flow rate of the hydraulic valve obtained at the corresponding hydraulic oil temperature after performance degradation, the valve port flow area change function g2(u) related to the control voltage u at that hydraulic oil temperature is determined; Based on the hydraulic oil temperature, hydraulic valve control voltage, and inlet / outlet pressure difference obtained before performance degradation, a valve orifice flow coefficient prediction model is constructed. After performance degradation, based on the current hydraulic oil temperature, hydraulic valve control voltage, and inlet / outlet pressure difference, the current valve orifice flow coefficient is predicted using the valve orifice flow coefficient prediction model. Based on the current hydraulic valve control voltage, the valve orifice flow area before performance degradation is determined, and based on the current hydraulic oil temperature and hydraulic valve control voltage, the valve orifice flow area after performance degradation is determined. Based on the predicted valve orifice flow coefficient, the determined valve orifice flow area before and after performance degradation, and the currently obtained inlet and outlet pressure difference of the hydraulic valve, the actual flow rate of the hydraulic valve after performance degradation is determined.

2. The soft flow measurement method for hydraulic valves considering service performance degradation according to claim 1, characterized in that, The valve orifice flow area change function g1(u) related to the control voltage u before performance degradation is determined by the following formula: Where g1(u) = a1 + b1u, a1 and b1 are constants, Q1(T,u,ΔP) is the flow rate of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, and C d1 ΔP is the valve orifice flow coefficient before performance degradation, ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained at the same hydraulic oil temperature before performance degradation, and ρ is the hydraulic oil density.

3. The soft measurement method for hydraulic valve flow rate considering service performance degradation according to claim 1, characterized in that, Based on the inlet and outlet pressure difference and flow rate of the hydraulic valve obtained at the corresponding hydraulic oil temperature after performance degradation, the valve orifice flow area change function g2(u) related to the control voltage u at that hydraulic oil temperature is determined according to the following formula: Where g2(u) = a2 + b2u, a2 and b2 are constants, Q2(T,u,ΔP) is the flow rate of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, and C d2 ΔP is the valve orifice flow coefficient after performance degradation, ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained at the same hydraulic oil temperature after performance degradation, and ρ is the hydraulic oil density.

4. The soft flow measurement method for hydraulic valves considering service performance degradation according to claim 1, characterized in that, The valve orifice flow coefficient model is a neural network model, which includes an input layer, multiple hidden layers, and an output layer. The nonlinear mapping from the input layer to the first hidden layer is as follows: Where a n Let n be the nth neuron in the first hidden layer, where n is an integer greater than 0 and less than or equal to the number of neurons m in each hidden layer; u is the control voltage of the hydraulic valve; ΔP is the pressure difference between the inlet and outlet of the hydraulic valve; T is the temperature; and w is the voltage. 1n The weight w represents the distance from the valve opening of the hydraulic valve to the nth neuron in the first hidden layer. 2n The weights w represent the pressure difference between the inlet and outlet layers, applied to the n neurons of the first hidden layer. 3n Let be the weights from temperature to the nth neuron in the first hidden layer. f(x) is the threshold from the input layer to the nth neuron of the first hidden layer, and f(x) is the activation function corresponding to the input layer to the first hidden layer. The nonlinear mapping between two adjacent hidden layers is: Where b n It is the nth neuron in hidden layer b, w mn Let be the weights from the m-th neuron in hidden layer b-1 to the n-th neuron in hidden layer b. f(x) is the threshold of the nth neuron in hidden layer b, and f(x) is the activation function between two hidden layers. The nonlinear mapping from the last hidden layer to the output layer is: C' d1 =w1b1+w2b2+…+w m b m +v Where C' d1 Let w1, w2, ..., w be the predicted flow coefficients. n It represents the weights from each neuron in the last hidden layer to the neurons in the output layer, and v is the threshold of the output neuron.

5. The soft flow measurement method for hydraulic valves considering service performance degradation according to claim 4, characterized in that, Construct a valve orifice flow coefficient prediction model according to the following steps: Step S310: Design the number of hidden layers and the number of neurons in each hidden layer in the neural network: take the number of hidden layers within a set range and take the number of neurons within a set range, wherein each layer within the set range is combined with each neuron within the set range. Step S320: For each combination, take the temperature, valve opening and inlet / outlet pressure difference of the hydraulic valve under each working condition when the hydraulic valve has not experienced service performance degradation as a set of samples, input each set of samples into the neural network corresponding to the combination, obtain the predicted valve flow coefficient of each set of samples, and calculate the root mean square error and mean absolute error of the valve flow coefficient corresponding to the combination based on the actual valve flow coefficient and the predicted valve flow coefficient of each set of samples. Step S330: Determine the combination with the smallest root mean square error and mean absolute error among all combinations, and use the number of layers and quantity in the combination as the number of hidden layers and the number of neurons in each hidden layer of the neural network.

6. The soft measurement method for hydraulic pump output flow rate considering service performance degradation according to claim 4 or 5, characterized in that, The construction of the valve orifice flow coefficient prediction model also includes: Step S340: When the number of hidden layers and the number of neurons in each hidden layer of the neural network are fixed, the samples obtained when the hydraulic valve has not experienced service performance degradation are input into the neural network for training to obtain the predicted valve flow coefficient of each sample. The neural network determines the network performance based on the actual flow coefficient and the predicted flow coefficient of each sample. Step S350: Determine whether the network performance of the neural network meets the standard. If yes, obtain the optimal weight matrix and the optimal threshold matrix of the neural network. Otherwise, adjust the weight matrix and threshold matrix in the neural network and return to step S340. The weight matrix includes the weights of each input parameter in the input layer to each neuron in the first hidden layer, the weights of each neuron in the hidden layer to each neuron in the next hidden layer, and the weights of each neuron in the last hidden layer to the predicted flow in the output layer; the threshold matrix includes the thresholds of each neuron in the input layer to each neuron in the first hidden layer, the thresholds of each neuron in the hidden layer to each neuron in the next hidden layer, and the thresholds of the predicted flow in the last hidden layer to the output layer.

7. The soft measurement method for hydraulic pump output flow rate considering service performance degradation according to claim 6, characterized in that, The network performance includes the root mean square relative error and the average relative error of the valve orifice flow coefficient.

8. The soft measurement method for hydraulic pump output flow rate considering service performance degradation according to claim 4, characterized in that, The method further includes: before constructing the neural network, normalizing the obtained hydraulic oil temperature, hydraulic valve control voltage, inlet and outlet pressure difference and actual flow rate, as well as the output predicted valve flow coefficient, and limiting the values ​​to [0.0001, 0.9999].

9. The soft measurement method for hydraulic pump output flow considering service performance degradation according to claim 1, characterized in that, The actual flow rate Q2(T,u,ΔP) of the hydraulic valve after performance degradation is determined using the following formula: Where C' d1 The valve orifice flow coefficient is the predicted flow coefficient. g1(u) is the valve orifice flow area determined before performance degradation based on the current control voltage of the hydraulic valve. g2(u) represents the valve orifice flow area determined after performance degradation based on the current hydraulic oil temperature and the control voltage of the hydraulic valve. ΔP is the inlet and outlet pressure difference of the hydraulic valve obtained after performance degradation. ρ is the hydraulic oil density.

10. The soft measurement method for hydraulic pump output flow rate considering service performance degradation according to claim 1, characterized in that, The control voltage is replaced by the valve opening degree.

Images