Self-adaptive real-time efficiency optimization and predictive maintenance method for electro-hydraulic proportional valve

Abstract

The invention discloses a self-adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve, and relates to the technical field of maintenance control. Processing the acquired real-time operation state data through a data processing mode after the acquisition is completed, inputting the processed real-time operation state data into a long short-term memory network after the processing is completed, and predicting the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time through the long short-term memory network. Meanwhile, based on real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, health state prediction is carried out through a deep learning model, risk assessment is carried out on the electro-hydraulic proportional valve in an uncertainty quantification mode, and finally, risk assessment is carried out on the electro-hydraulic proportional valve based on the risk assessment result of the electro-hydraulic proportional valve. The optimal control instruction of the electro-hydraulic proportional valve is generated through the multi-target robust optimization decision, and the electro-hydraulic proportional valve is driven based on the optimal control instruction, so that the control accuracy of the electro-hydraulic proportional valve is improved.

Description

Technical Field

[0001] This invention relates to the field of rice crust cutting device technology, specifically to an adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve. Background Technology

[0002] The electro-hydraulic proportional valve (EHPV) system is the core actuator for achieving high-precision automatic steering in agricultural tractors. However, during long-term operation, the performance and response characteristics of the EHPV system are susceptible to factors such as oil temperature, oil contamination, valve core wear, and internal leakage, leading to nonlinear parameter drift. This drift directly manifests as an increase in the flow dead zone and a widening of the hysteresis loop, thereby reducing steering accuracy and increasing response time. Summary of the Invention

[0003] The purpose of this invention is to provide an adaptive real-time efficiency optimization and predictive maintenance method for electro-hydraulic proportional valves, which solves the problems existing in the background art.

[0004] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides an adaptive real-time efficiency optimization and predictive maintenance method for electro-hydraulic proportional valves, specifically including the following steps: S1. Install sensors to collect real-time operating status data during the operation of the electro-hydraulic proportional valve; The real-time operating status data includes: current, pressure, and response time; S2. The collected real-time operating status data is processed through data processing methods to obtain processed real-time operating status data; S3. Input the processed real-time operating status data into the long short-term memory network, and predict the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time through the long short-term memory network. S4. Based on the real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, a health status prediction is performed using a deep learning model, and the risk assessment of the electro-hydraulic proportional valve is conducted through uncertainty quantification. S5. Based on the risk assessment results of the electro-hydraulic proportional valve, the optimal control command for the electro-hydraulic proportional valve is generated through multi-objective robust optimization decision-making, and the electro-hydraulic proportional valve is driven based on the optimal control command.

[0005] Preferably, the step of processing the collected real-time operating status data to obtain processed real-time operating status data includes the following steps: A standard set of real-time operating status data must include: current data, pressure data, and response time data; By transforming the dimensions in the real-time operating status data, the current data, pressure data, and response time data in the real-time operating status data are converted into numerical data. Summarize the transformed numerical data and construct triples. The converted numerical data is saved in this way; in, This represents the constructed triple. This represents the current data in the real-time operating status data. This represents the pressure data in the real-time operating status data. This represents the response time data in the real-time operating status data; Filter the stored numerical data using data filtering methods; The collected numerical data is traversed to locate data with missing content, and the missing data is deleted. Set the numerical data after the deletion process is completed as the processed real-time running status data.

[0006] Preferably, the step of inputting the processed real-time operating status data into a long short-term memory network and predicting the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time through the long short-term memory network includes the following steps: S21. Construct an LSTM network model; Historical operating status data of the electro-hydraulic proportional valve are collected and compiled into a historical operating status data time series of length m. The historical operating status data time series is defined. As shown below: ; in, This represents the sequence of driving currents for a proportional electromagnet. This represents the valve core displacement sequence within the time window. This represents the output flow sequence within the time window. This represents the valve orifice pressure differential sequence within a time window; This represents the hydraulic oil temperature sequence within a time window. Indicates the length of the sliding time window; S22. Input the obtained historical operating status data time series as the input threshold into the LSTM network, and input the sensor's operating status data at the corresponding time as candidate values ​​into the LSTM network. LSTM networks include input gates, output gates, and forget gates; S23. Process the current input and the historical running state data from the previous time step using the forget gate, and then update them; A set of observable parameters that directly correspond to the core physics model of EHPV is defined, as shown below: ; in, EHPV flow gain; The electromagnetic time constant; It is the viscous damping coefficient; Coulomb friction; This represents the hysteresis loop coefficient.

[0007] The system incorporates a parameter observer based on a long short-term memory network to update the key nonlinear parameter vectors of the electro-hydraulic proportional valve. ; The observer takes the valve's control current, valve core displacement, and system pressure as inputs and continuously updates the parameter estimates through online identification technology. S24. The historical running state data processed by the forget gate is output through the output gate to obtain the predicted historical running state data at the corresponding time. The LSTM prediction output is shown below: ; in, Indicates prediction Increment of time parameter drift This represents the trained Long Short-Term Memory (LSTM) network model, outputting its dimensions and parameter vectors. same; Current parameter estimation ; in, for The parameter estimation vector at time step; For the nominal parameter vector; This represents the cumulative sum of drift increments from the initial time to the current time. S25. Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the historical running state data predicted by the LSTM network at the corresponding time and the historical running state data collected at the corresponding time. When the error between the historical running state data predicted by the LSTM network and the historical running state data collected is within the set error threshold, the iteration stops and the trained LSTM network is obtained. Otherwise, the LSTM network parameters are dynamically adjusted until the error converges. S26. Set an LSTM network that meets the error threshold range as a nonlinear parameter drift state prediction model, input the processed real-time running state data into the nonlinear parameter drift state prediction model, and output the predicted nonlinear parameter drift state at the next moment through the nonlinear parameter drift state prediction model. The nonlinear comprehensive drift index is calculated based on the processed real-time operating status data. : ; in, Overall drift index, dimensionless. , Represents the weighting coefficients, satisfying , This is the flow gain drift term, representing the degradation of the valve orifice flow regulation capability; Indicates the nominal flow gain coefficient; This represents the currently estimated flow gain coefficient; This is the electromagnetic time constant drift term, representing the degradation of the electromagnetic response speed; The nominal electromagnetic time constant; This is the currently estimated electromagnetic time constant; The term represents the hysteresis force, indicating the degree of nonlinearity of friction and hysteresis in the valve core motion; This represents the current hysteresis loop area or maximum hysteresis force. The nominal hysteresis force; This is the rise time term for the step response, representing the degradation of the valve's dynamic response speed; This is the current step response rise time; This refers to the nominal rise time.

[0008] Preferably, the step of predicting the health status of the electro-hydraulic proportional valve based on the real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, using a deep learning model, and conducting a risk assessment of the electro-hydraulic proportional valve through uncertainty quantification includes the following steps: S31. Set the nonlinear parameter drift state set ,in This represents the drift state of the first nonlinear parameter of the electro-hydraulic proportional valve, where Indicates the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; control command set in This represents the first control command executed under the first nonlinear parameter drift state of the electro-hydraulic proportional valve, where... This represents the j-th control command under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; S32. Calculate the initial probability based on the set nonlinear parameter drift state set and control command set of the electro-hydraulic proportional valve, and calculate the one-step state transition probability. S33. Calculate multi-step state transition probabilities based on the calculated one-step state transition probability; The possible states at each step are estimated using a probability estimation algorithm; S34. Based on the state set of the electro-hydraulic proportional valve after output estimation, the control command set corresponding to each state, and the probability of executing the corresponding control command, a real-time predictive digital twin model is established. S35. Risk assessment of electro-hydraulic proportional valves is conducted based on the established real-time predictive digital twin model using uncertainty quantification. Collect various indicators of the nonlinear parameter drift state of the electro-hydraulic proportional valve, and calculate the information entropy of each indicator for risk assessment. ; in, , The number of indicators representing the nonlinear drift state of an electro-hydraulic proportional valve. This represents the weight value of the h-th index under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve. This represents the information entropy of the h-th indicator; The information entropy values ​​of the various indicators are set as the risk assessment results of the electro-hydraulic proportional valve. The higher the information entropy value, the higher the risk.

[0009] Preferably, the calculation of the initial probability and the one-step state transition probability based on the set of nonlinear parameters of the electro-hydraulic proportional valve and the set of control commands includes the following steps: The one-step state transition probability is calculated as follows: One-step state transition probability distribution:

[0010] in, This is the probability distribution for a one-step state transition. Representing state After processing by the reward function, the state is reached. The probability, Represents the reward function, , Representing state Execute control commands The reward value at that time.

[0011] Preferably, the process of generating optimal control commands for the electro-hydraulic proportional valve based on the risk assessment results of the electro-hydraulic proportional valve through multi-objective robust optimization decision-making, and driving the electro-hydraulic proportional valve based on the optimal control commands, includes the following steps: S51. Initialize multi-objective robust optimization decision; The multi-objective robust optimization decision includes: accuracy objective decision, energy efficiency objective decision, and robustness objective decision; The accuracy target decision function is shown below: ; in, Represents the desired state trajectory. ; express The predicted state vector at time step; To minimize the deviation between the state and the expected value in the prediction time domain; The energy efficiency target decision function is shown below: ; Hydraulic power item: ; in, The unit of flow rate is (L / min); Indicates pressure difference (bar); This represents the volumetric efficiency (0~1), with 600 being the unit conversion factor (1 bar·L / min = 1 / 600 kW). Penalty for controlling the rate of change: ; in, Indicates the control variable in Time-based control input; Indicates the rate of change of the control input; The robust objective decision function is shown below: ; Part 1: Weighted State Error ; Among them, the robustness weight matrix It increases with increasing parameter uncertainty; It is a robust regulator; The parameter uncertainty covariance matrix, ; Part Two: Control Deviation Penalties ; in, This represents the nominal control quantity under the current operating conditions. ; To control the deviation penalty coefficient, usually ; Comprehensive optimization problem ; Solving for the optimal control sequence under dynamic model constraints. Weight according to Adaptive adjustment; S52. Based on multi-objective robust optimization decision-making, the optimal control command for the electro-hydraulic proportional valve is generated through a genetic algorithm.

[0012] Preferably, the step of generating the optimal control command for the electro-hydraulic proportional valve using a genetic algorithm based on multi-objective robust optimization decision-making includes the following steps: Summarize the index data of the electro-hydraulic proportional valve under various states to construct a population, and set the population size and number of iterations for genetics; In constructing the initial population, y sets of indicator data are randomly selected from the population construction process. The risk assessment results of the electro-hydraulic proportional valve and the multi-objective robust optimization decision are used as the fitness function of the genetic algorithm, and the fitness of each individual in the population is calculated. Each individual is set to represent a set of index data for an electro-hydraulic proportional valve; The fitness formula for each individual is as follows: ; Where Fit(k) represents the fitness of the k-th individual. Represents the fitness function; The best individuals are selected from all individuals based on their fitness values ​​and using a roulette wheel selection method. After selection, the selected superior individuals are crossbred using a sequential crossbringing method, resulting in a new population. ; A random individual is selected from the population and mutated with a set probability to produce a mutated population. ; Compare the fitness difference between the initial population and the population after crossover and mutation using a genetic algorithm. ; when <0 indicates that the fitness of the mutated population is higher than that of the initial population. Accept this population. ≥0 indicates that the fitness of the mutated population is lower than that of the initial population, and the population should be rejected. Furthermore, the algorithm determines whether the maximum number of iterations has been reached based on the number of iterations. If the maximum number of iterations has been reached, the optimal solution is output; otherwise, the iteration continues. The optimal output solution is set as the optimal control command for the electro-hydraulic proportional valve.

[0013] The beneficial effects of this invention are as follows: (1) Predictive maintenance capability: Real-time monitoring of EHPV health status and early warning of faults are achieved through digital twin technology. (2) Energy efficiency adaptive optimization: dynamically balance control accuracy and energy efficiency under the multi-objective optimization framework.

[0014] (3) Strong robustness guarantee: Risk assessment based on uncertainty quantification ensures the reliability of the system under parameter drift conditions.

[0015] (4) Full life cycle management: realize a complete health management closed loop from fault early warning to adaptive compensation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a diagram of the overall architecture of the two-layer machine learning-driven EHPV adaptive control system of the present invention.

[0018] Figure 2 This is a schematic diagram of the EHPV digital twin and drift prediction principle of the present invention.

[0019] Figure 3 This is a diagram showing the optimization structure of the RMPC multi-objective cost function J of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In a specific embodiment of the present invention, Reference Figure 1 As shown, this invention provides a method for adaptive real-time efficiency optimization and predictive maintenance of an electro-hydraulic proportional valve, comprising: S1. Install sensors to collect real-time operating status data during the operation of the electro-hydraulic proportional valve; The real-time operating status data includes: current, pressure, and response time; S2. The collected real-time operating status data is processed through data processing methods to obtain processed real-time operating status data; S3. Input the processed real-time operating status data into the long short-term memory network, and predict the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time through the long short-term memory network. S4. Based on the real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, a health status prediction is performed using a deep learning model, and the risk assessment of the electro-hydraulic proportional valve is conducted through uncertainty quantification. S5. Based on the risk assessment results of the electro-hydraulic proportional valve, the optimal control command for the electro-hydraulic proportional valve is generated through multi-objective robust optimization decision-making, and the electro-hydraulic proportional valve is driven based on the optimal control command. Furthermore, referring to Figure 1 As shown, the real-time operating status data collected is processed using data processing methods to obtain the processed real-time operating status data, including the following steps: A standard set of real-time operating status data must include: current data, pressure data, and response time data; By transforming the dimensions in the real-time operating status data, the current data, pressure data, and response time data in the real-time operating status data are converted into numerical data. Furthermore, the transformed numerical data is aggregated and triples are constructed. The converted numerical data is saved in this way; in, This represents the constructed triple. This represents the current data in the real-time operating status data. This represents the pressure data in the real-time operating status data. This represents the response time data in the real-time operating status data; Furthermore, the stored numerical data is filtered using data filtering methods; The collected numerical data is traversed to locate data with missing content, and the missing data is deleted. Set the numeric data after the deletion process is completed as the processed real-time running status data; Furthermore, referring to Figure 1 As shown, the processed real-time operating status data is input into a long short-term memory network (LSTM) to predict the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time. This includes the following steps: S21. Construct an LSTM network model; Historical operating status data of the electro-hydraulic proportional valve are collected and compiled into a historical operating status data time series of length m. The historical operating status data time series is defined. As shown below: ; in, This represents the sequence of driving currents for a proportional electromagnet. This represents the valve core displacement sequence within the time window. This represents the output flow sequence within the time window. This represents the valve orifice pressure differential sequence within a time window; This represents the hydraulic oil temperature sequence within a time window. Indicates the length of the sliding time window; S22. Input the obtained historical operating status data time series as the input threshold into the LSTM network, and input the sensor's operating status data at the corresponding time as candidate values ​​into the LSTM network. LSTM networks include input gates, output gates, and forget gates; S23. Process the current input and the historical running state data from the previous time step using the forget gate, and then update them; A set of observable parameters that directly correspond to the core physics model of EHPV is defined, as shown below: ; in, EHPV flow gain; The electromagnetic time constant; It is the viscous damping coefficient; Coulomb friction; This represents the hysteresis loop coefficient.

[0022] The system incorporates a parameter observer based on a long short-term memory network to update the key nonlinear parameter vectors of the electro-hydraulic proportional valve. ; The observer takes the valve's control current, valve core displacement, and system pressure as inputs and continuously updates the parameter estimates through online identification technology. S24. The historical running state data processed by the forget gate is output through the output gate to obtain the predicted historical running state data at the corresponding time. The LSTM prediction output is shown below: ; in, Indicates prediction Increment of time parameter drift This represents the trained Long Short-Term Memory (LSTM) network model, outputting its dimensions and parameter vectors. same; Current parameter estimation ; in, for The parameter estimation vector at time step; For the nominal parameter vector; This represents the cumulative sum of drift increments from the initial time to the current time. S25. Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the historical running state data predicted by the LSTM network at the corresponding time and the historical running state data collected at the corresponding time. When the error between the historical running state data predicted by the LSTM network and the historical running state data collected is within the set error threshold, the iteration stops and the trained LSTM network is obtained. Otherwise, the LSTM network parameters are dynamically adjusted until the error converges. S26. The aforementioned LSTM network predicts and outputs the key nonlinear parameter vector of the electro-hydraulic proportional valve (EHPV). The estimated value reflects the independent, numerical drift of each parameter. However, for higher-level control systems (such as adaptive optimization controllers), a comprehensive, normalized, and physically meaningful single health index is needed to concisely represent the overall performance degradation of the valve and serve as the basis for adjusting the control strategy.

[0023] To address this, this invention proposes a "nonlinear composite drift index." This index, through a carefully designed weighted aggregation model, integrates the drift of multiple key parameters, as well as directly observable performance degradation phenomena (such as increased hysteresis and slower response), into a scalar value varying between 0 and 1. This definition serves as a crucial bridge between parameter prediction and control applications. An LSTM network that meets the error threshold range is set as the nonlinear parameter drift state prediction model. The processed real-time running state data is input into the nonlinear parameter drift state prediction model, and the nonlinear parameter drift state predicted for the next time step is output through the nonlinear parameter drift state prediction model. The nonlinear comprehensive drift index is calculated based on the processed real-time operating status data. : in, Overall drift index, dimensionless. , Represents the weighting coefficients, satisfying , This is the flow gain drift term, representing the degradation of the valve orifice flow regulation capability; Indicates the nominal flow gain coefficient; This represents the currently estimated flow gain coefficient; This is the electromagnetic time constant drift term, representing the degradation of the electromagnetic response speed; The nominal electromagnetic time constant; This is the currently estimated electromagnetic time constant; The term represents the hysteresis force, indicating the degree of nonlinearity of friction and hysteresis in the valve core motion; This represents the current hysteresis loop area or maximum hysteresis force. This refers to the nominal hysteresis force (value under healthy conditions). This is the rise time term for the step response, representing the degradation of the valve's dynamic response speed; The current step response rise time (10%~90%). This is the nominal rise time (value under healthy conditions).

[0024] By introducing a nonlinear comprehensive drift index This approach combines the drift of multiple abstract parameters predicted by LSTM with observable system performance degradation, transforming them into a stable, reliable, and easy-to-use system-level health status feature. This allows upper-level controllers to easily... The value (e.g., setting a threshold) , This allows for real-time adjustment of the target weights, relaxation of constraints, or triggering of early warnings, thereby achieving a seamless transition from "parameter prediction" to "intelligent control and maintenance decision-making." Furthermore, referring to Figure 1 As shown, based on real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, a health status prediction is performed using a deep learning model, and a risk assessment of the electro-hydraulic proportional valve is conducted through uncertainty quantification, including the following steps: S31. Set the nonlinear parameter drift state set ,in This represents the drift state of the first nonlinear parameter of the electro-hydraulic proportional valve, where Indicates the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; control command set in This represents the first control command executed under the first nonlinear parameter drift state of the electro-hydraulic proportional valve, where... This represents the j-th control command under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; S32. Calculate the initial probability based on the set nonlinear parameter drift state set and control command set of the electro-hydraulic proportional valve, and calculate the one-step state transition probability. The one-step state transition probability is calculated as follows: One-step state transition probability distribution:

[0025] in, This is the probability distribution for a one-step state transition. Representing state After processing by the reward function, the state is reached. The probability, Represents the reward function, , Representing state Execute control commands The reward value at that time; S33. Calculate multi-step state transition probabilities based on the calculated one-step state transition probability; The possible states at each step are estimated using a probability estimation algorithm; S34. Based on the state set of the electro-hydraulic proportional valve after output estimation, the control command set corresponding to each state, and the probability of executing the corresponding control command, a real-time predictive digital twin model is established. S35. Risk assessment of electro-hydraulic proportional valves is conducted based on the established real-time predictive digital twin model using uncertainty quantification. Collect various indicators of the nonlinear parameter drift state of the electro-hydraulic proportional valve, and calculate the information entropy of each indicator for risk assessment. ; in, , The number of indicators representing the nonlinear drift state of an electro-hydraulic proportional valve. This represents the weight value of the h-th index under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve. This represents the information entropy of the h-th indicator; Furthermore, the information entropy values ​​of the calculated indicators are set as the risk assessment results of the electro-hydraulic proportional valve; the higher the information entropy value, the higher the risk. Furthermore, referring to Figure 1 As shown, based on the risk assessment results of the electro-hydraulic proportional valve, the optimal control command for the electro-hydraulic proportional valve is generated through multi-objective robust optimization decision-making, and the electro-hydraulic proportional valve is driven based on the optimal control command, including the following steps: S51. Initialize multi-objective robust optimization decision; The multi-objective robust optimization decision includes: accuracy objective decision, energy efficiency objective decision, and robustness objective decision; The accuracy target decision function is shown below: ; in, Represents the desired state trajectory. ; express The predicted state vector at time step; To minimize the deviation between the state and the expected value in the prediction time domain; The energy efficiency target decision function is shown below: ; Hydraulic power item: ; in, The unit of flow rate is (L / min); Indicates pressure difference (bar); This represents the volumetric efficiency (0~1), with 600 being the unit conversion factor (1 bar·L / min = 1 / 600 kW). Penalty for controlling the rate of change: ; in, Indicates the control variable in Time-based control input; Indicates the rate of change of the control input; The robust objective decision function is shown below: ; Part 1: Weighted State Error ; Among them, the robustness weight matrix It increases with increasing parameter uncertainty; It is a robust regulator; The parameter uncertainty covariance matrix, ; Part Two: Control Deviation Penalties ; in, This represents the nominal control quantity under the current operating conditions. ; To control the deviation penalty coefficient, usually ; Comprehensive optimization problem ; Solving for the optimal control sequence under dynamic model constraints. Weight according to Adaptive adjustment; S52. Based on multi-objective robust optimization decision-making, the optimal control command for the electro-hydraulic proportional valve is generated through a genetic algorithm. Summarize the index data of the electro-hydraulic proportional valve under various states to construct a population, and set the population size and number of iterations for genetics; Furthermore, y sets of indicator data are randomly selected to construct the initial population during the population construction process; The risk assessment results of the electro-hydraulic proportional valve and the multi-objective robust optimization decision are used as the fitness function of the genetic algorithm, and the fitness of each individual in the population is calculated. Each individual is set to represent a set of index data for an electro-hydraulic proportional valve; The fitness formula for each individual is as follows: ; Where Fit(k) represents the fitness of the k-th individual. Represents the fitness function; Furthermore, based on the fitness value of each individual and a roulette wheel selection method, the best individuals among all individuals are selected; Furthermore, after selection, the selected superior individuals are crossbred using a sequential crossbringing method, resulting in a new population. ; Furthermore, an individual is randomly selected from the population and mutated with a set probability to produce a mutated population. ; Furthermore, the fitness difference between the initial population and the population after crossover and mutation using the genetic algorithm is compared. ; when <0 indicates that the fitness of the mutated population is higher than that of the initial population. Accept this population. ≥0 indicates that the fitness of the mutated population is lower than that of the initial population, and the population should be rejected. Furthermore, the algorithm determines whether the maximum number of iterations has been reached based on the number of iterations. If the maximum number of iterations has been reached, the optimal solution is output; otherwise, the iteration continues. The optimal solution for the output is set as the optimal control command for the electro-hydraulic proportional valve; It should be noted that The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A method for adaptive real-time efficiency optimization and predictive maintenance of an electro-hydraulic proportional valve, characterized in that, Includes the following steps: S1. Install sensors to collect real-time operating status data during the operation of the electro-hydraulic proportional valve; The real-time operating status data includes: current, pressure, and response time; S2. The collected real-time operating status data is processed through data processing methods to obtain processed real-time operating status data; S3. Input the processed real-time operating status data into the long short-term memory network, and predict the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time through the long short-term memory network. S4. Based on the real-time prediction of the nonlinear parameter drift state of the electro-hydraulic proportional valve, a health status prediction is performed using a deep learning model, and the risk assessment of the electro-hydraulic proportional valve is conducted through uncertainty quantification. S5. Based on the risk assessment results of the electro-hydraulic proportional valve, the optimal control command for the electro-hydraulic proportional valve is generated through multi-objective robust optimization decision-making, and the electro-hydraulic proportional valve is driven based on the optimal control command.

2. The method for adaptive real-time efficiency optimization and predictive maintenance of an electro-hydraulic proportional valve according to claim 1, characterized in that, The process of processing the collected real-time operating status data to obtain the processed real-time operating status data includes the following steps: A standard set of real-time operating status data must include: current data, pressure data, and response time data; By transforming the dimensions in the real-time operating status data, the current data, pressure data, and response time data in the real-time operating status data are converted into numerical data. Summarize the transformed numerical data and construct triples. The converted numerical data is saved in this way; in, This represents the constructed triple. This represents the current data in the real-time operating status data. This represents the pressure data in the real-time operating status data. This represents the response time data in the real-time operating status data; Filter the stored numerical data using data filtering methods; The collected numerical data is traversed to locate data with missing content, and the missing data is deleted. Set the numerical data after the deletion process is completed as the processed real-time running status data.

3. The adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve according to claim 1, characterized in that, The process of inputting the processed real-time operating status data into a long short-term memory network (LSTM) and using the LSTM to predict the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time includes the following steps: Historical operating status data of the electro-hydraulic proportional valve are collected and compiled into a historical operating status data time series of length m. The historical operating status data time series is defined. As shown below: ； in, This represents the sequence of driving currents for a proportional electromagnet. This represents the valve core displacement sequence within the time window. This represents the output flow sequence within the time window. This represents the valve orifice pressure differential sequence within a time window; This represents the hydraulic oil temperature sequence within a time window. Indicates the length of the sliding time window; S22. Input the obtained historical operating status data time series as the input threshold into the LSTM network, and input the sensor's operating status data at the corresponding time as candidate values ​​into the LSTM network. LSTM networks include input gates, output gates, and forget gates; S23. Process the current input and the historical running state data from the previous time step using the forget gate, and then update them; A set of observable parameters that directly correspond to the core physics model of EHPV is defined, as shown below: ； in, EHPV flow gain; The electromagnetic time constant; It is the viscous damping coefficient; Coulomb friction; This represents the hysteresis loop coefficient.

4. The system incorporates a parameter observer based on a long short-term memory network to update the key nonlinear parameter vectors of the electro-hydraulic proportional valve. ; The observer takes the valve's control current, valve core displacement, and system pressure as inputs and continuously updates the parameter estimates through online identification technology. S24. The historical running state data processed by the forget gate is output through the output gate to obtain the predicted historical running state data at the corresponding time. The LSTM prediction output is shown below: ； in, Indicates prediction Increment of time parameter drift This represents the trained Long Short-Term Memory (LSTM) network model, outputting its dimensions and parameter vectors. same; Current parameter estimation ； in, for The parameter estimation vector at time step; For the nominal parameter vector; This represents the cumulative sum of drift increments from the initial time to the current time. S25. Iteratively use the LSTM network to determine and obtain the trained LSTM network; Set an error threshold between the historical running state data predicted by the LSTM network at the corresponding time and the historical running state data collected at the corresponding time. When the error between the historical running state data predicted by the LSTM network and the historical running state data collected is within the set error threshold, the iteration stops and the trained LSTM network is obtained. Otherwise, the LSTM network parameters are dynamically adjusted until the error converges. S26. Set an LSTM network that meets the error threshold range as a nonlinear parameter drift state prediction model, input the processed real-time running state data into the nonlinear parameter drift state prediction model, and output the predicted nonlinear parameter drift state at the next moment through the nonlinear parameter drift state prediction model. The nonlinear comprehensive drift index is calculated based on the processed real-time operating status data. : ； in, Overall drift index, dimensionless. , Represents the weighting coefficients, satisfying , This is the flow gain drift term, representing the degradation of the valve orifice flow regulation capability; Indicates the nominal flow gain coefficient; This represents the currently estimated flow gain coefficient; This is the electromagnetic time constant drift term, representing the degradation of the electromagnetic response speed; The nominal electromagnetic time constant; This is the currently estimated electromagnetic time constant; The term represents the hysteresis force, indicating the degree of nonlinearity of friction and hysteresis in the valve core motion; This represents the current hysteresis loop area or maximum hysteresis force. The nominal hysteresis force; This is the rise time term for the step response, representing the degradation of the valve's dynamic response speed; This is the current step response rise time; This refers to the nominal rise time.

5. The adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve according to claim 1, characterized in that, The process of predicting the nonlinear parameter drift state of the electro-hydraulic proportional valve in real time, predicting its health status using a deep learning model, and assessing its risk through uncertainty quantification includes the following steps: S31. Set the nonlinear parameter drift state set ,in This represents the drift state of the first nonlinear parameter of the electro-hydraulic proportional valve, where Indicates the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; control command set in This represents the first control command executed under the first nonlinear parameter drift state of the electro-hydraulic proportional valve, where... This represents the j-th control command under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve; S32. Calculate the initial probability based on the set nonlinear parameter drift state set and control command set of the electro-hydraulic proportional valve, and calculate the one-step state transition probability. S33. Calculate multi-step state transition probabilities based on the calculated one-step state transition probability; The possible states at each step are estimated using a probability estimation algorithm; S34. Based on the state set of the electro-hydraulic proportional valve after output estimation, the control command set corresponding to each state, and the probability of executing the corresponding control command, a real-time predictive digital twin model is established. S35. Risk assessment of electro-hydraulic proportional valves is conducted based on the established real-time predictive digital twin model using uncertainty quantification. Collect various indicators of the nonlinear parameter drift state of the electro-hydraulic proportional valve, and calculate the information entropy of each indicator for risk assessment. ； in, , The number of indicators representing the nonlinear drift state of an electro-hydraulic proportional valve. This represents the weight value of the h-th index under the drift state of the s-th nonlinear parameter of the electro-hydraulic proportional valve. This represents the information entropy of the h-th indicator; The information entropy values ​​of the various indicators are set as the risk assessment results of the electro-hydraulic proportional valve. The higher the information entropy value, the higher the risk.

6. The adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve according to claim 4, characterized in that, The calculation of the initial probability and the one-step state transition probability based on the set of nonlinear parameters of the electro-hydraulic proportional valve and the control command set includes the following steps: The one-step state transition probability is calculated as follows: One-step state transition probability distribution:

7. Among them, This is the probability distribution for a one-step state transition. Representing state After processing by the reward function, the state is reached. The probability, Represents the reward function, , Representing state Execute control commands The reward value at that time.

8. The adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve according to claim 1, characterized in that, The process of generating optimal control commands for the electro-hydraulic proportional valve based on the risk assessment results of the electro-hydraulic proportional valve through multi-objective robust optimization decision-making, and driving the electro-hydraulic proportional valve based on the optimal control commands includes the following steps: S51. Initialize multi-objective robust optimization decision; The multi-objective robust optimization decision includes: accuracy objective decision, energy efficiency objective decision, and robustness objective decision; The accuracy target decision function is shown below: ； in, Represents the desired state trajectory. ; express The predicted state vector at time step; To minimize the deviation between the state and the expected value in the prediction time domain; The energy efficiency target decision function is shown below: ； Hydraulic power item: ； in, The unit of flow rate is (L / min); Indicates pressure difference (bar); This represents the volumetric efficiency (0~1), with 600 being the unit conversion factor (1 bar·L / min = 1 / 600 kW). Penalty for controlling the rate of change: ； in, Indicates the control variable in Time-based control input; Indicates the rate of change of the control input; The robust objective decision function is shown below: ； Part 1: Weighted State Error ； Among them, the robustness weight matrix It increases with increasing parameter uncertainty; It is a robust regulator; The parameter uncertainty covariance matrix, ; Part Two: Control Deviation Penalties ； in, This represents the nominal control quantity under the current operating conditions. ; To control the deviation penalty coefficient, usually ; Comprehensive optimization problem ； Solving for the optimal control sequence under dynamic model constraints. Weight according to Adaptive adjustment; S52. Based on multi-objective robust optimization decision-making, the optimal control command for the electro-hydraulic proportional valve is generated through a genetic algorithm.

9. The adaptive real-time efficiency optimization and predictive maintenance method for an electro-hydraulic proportional valve according to claim 6, characterized in that, The method of generating the optimal control command for the electro-hydraulic proportional valve based on multi-objective robust optimization decision-making and using a genetic algorithm includes the following steps: Summarize the index data of the electro-hydraulic proportional valve under various states to construct a population, and set the population size and number of iterations for genetics; In constructing the initial population, y sets of indicator data are randomly selected from the population construction process. The risk assessment results of the electro-hydraulic proportional valve and the multi-objective robust optimization decision are used as the fitness function of the genetic algorithm, and the fitness of each individual in the population is calculated. Each individual is set to represent a set of index data for an electro-hydraulic proportional valve; The fitness formula for each individual is as follows: ； Where Fit(k) represents the fitness of the k-th individual. Represents the fitness function; The best individuals are selected from all individuals based on their fitness values ​​and using a roulette wheel selection method. After selection, the selected superior individuals are crossbred using a sequential crossbringing method, resulting in a new population. ; A random individual is selected from the population and mutated with a set probability to produce a mutated population. ; Compare the fitness difference between the initial population and the population after crossover and mutation using a genetic algorithm. ; when <0 indicates that the fitness of the mutated population is higher than that of the initial population. Accept this population. ≥0 indicates that the fitness of the mutated population is lower than that of the initial population, and the population should be rejected. Furthermore, the algorithm determines whether the maximum number of iterations has been reached based on the number of iterations. If the maximum number of iterations has been reached, the optimal solution is output; otherwise, the iteration continues. The optimal output solution is set as the optimal control command for the electro-hydraulic proportional valve.

Images