CN-121993457-A - Electro-hydraulic proportional valve self-adaptive control method, system, equipment and medium for compressor oil cylinder

CN121993457ACN 121993457 ACN121993457 ACN 121993457ACN-121993457-A

Abstract

The invention relates to an electro-hydraulic proportional valve self-adaptive control method, system, equipment and medium of a compressor oil cylinder, wherein the method comprises the steps of collecting parameters of the compressor oil cylinder, preprocessing to obtain a working condition time sequence data set, carrying out reinforcement learning optimization after dividing rule subsets according to working condition clusters according to reference control parameters extracted from the working condition time sequence data set, further synchronously executing fuzzy reasoning and dynamic compensation of the current control parameters of the electro-hydraulic proportional valve, fusing to obtain initial control quantity, calculating working condition urgency to dynamically allocate compensation weights, carrying out graded compensation on the initial control quantity according to the allocated compensation weights to generate final control quantity, inputting the reference control parameters and the final control quantity into a prediction model, predicting valve core abrasion index and abrasion increment, and further executing multistage safety intervention operation. Therefore, the invention integrates multidimensional data and reinforcement learning optimization, collaborative hierarchical compensation and physical neural network abrasion prediction, and improves the self-adaptability and the debugging efficiency of the electro-hydraulic proportional valve.

Inventors

ZHANG SANJUN
LI MENGKANG
LI WUJUN
Zeng can
ZHANG SHENG

Assignees

湖南仁和环保科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251224

Claims (10)

1. An electro-hydraulic proportional valve self-adaptive control method of a compressor oil cylinder is characterized by comprising the following steps of: Collecting core working condition parameters and auxiliary operation parameters of a compressor oil cylinder, and preprocessing to obtain a working condition time sequence data set; According to the reference control parameters extracted from the working condition time sequence data set, performing reinforcement learning optimization after dividing rule subsets through working condition clustering, synchronously executing fuzzy reasoning and dynamic compensation of the current control parameters of the electro-hydraulic proportional valve based on the optimized rule subsets, and fusing to obtain initial control quantity; aiming at the working condition category obtained by clustering according with preset intervention conditions, calculating working condition emergency degree to dynamically allocate compensation weights, and carrying out hierarchical compensation on the initial control quantity according to the allocated compensation weights to generate a final control quantity; The reference control parameters and the final control quantity are input into a pre-trained physical enhancement type neural network prediction model, the abrasion index and the abrasion increment of the valve core are predicted through dynamically matching the nonlinear coupling relation between the oil characteristics and the part kinematic characteristics, and one operation among the multistage safety intervention operations is executed based on the combination criteria of the abrasion index and the abrasion increment.
2. The method for adaptively controlling an electro-hydraulic proportional valve of a compressor cylinder as set forth in claim 1, wherein collecting core operating parameters and auxiliary operating parameters of the compressor cylinder and preprocessing to obtain an operating time sequence data set comprises: Acquiring a main shaft rotating speed, an oil cylinder load pressure, an oil temperature, an oil pollution degree and an oil viscosity through a pre-deployed anti-interference sensor network, and solving an oil viscosity change rate based on the oil viscosity to form a core working condition parameter; acquiring a main shaft current change rate, a proportional valve driving current and an oil filter pressure difference through an anti-interference sensor network to form auxiliary operation parameters; and carrying out Kalman filtering on the oil viscosity change rate, and simultaneously carrying out moving average processing on parameters except the oil viscosity change rate of the core working condition parameters and auxiliary operation parameters, and obtaining a working condition time sequence data set by combining GPS clock alignment.
3. The method for adaptively controlling an electro-hydraulic proportional valve of a compressor cylinder according to claim 1, wherein the step of performing reinforcement learning optimization after dividing rule subsets by working condition clusters according to reference control parameters extracted from a working condition time sequence data set, and performing fuzzy reasoning and current control parameter dynamic compensation of the electro-hydraulic proportional valve synchronously based on the optimized rule subsets, and the step of obtaining initial control quantity by fusion comprises the following steps: Extracting multidimensional dynamic characteristics comprising pressure tracking deviation, deviation change rate, load impact strength grade and oil viscosity change rate from a working condition time sequence data set; the method comprises the steps of dividing working condition categories by carrying out density clustering on multidimensional dynamic features, and distributing corresponding rule subsets for each working condition category through multi-objective preferential screening; Migrating an optimal control rule of the same type of working conditions in a preset historical database to a current rule subset by utilizing a constructed Q-Learning model associated with each rule subset and adopting a migration Learning mechanism, guiding online updating of the rule subset by a multi-objective reward function, synchronously executing rule liveness marking and sparsification storage management, and screening and outputting the optimized rule subset according to liveness; Based on the optimized rule subset, fuzzy rule reasoning is executed based on multidimensional dynamic characteristics and membership functions of dynamic telescopic adjustment along with oil viscosity to output preliminary control quantity, and feedforward compensation is synchronously carried out on current control parameters of the electrohydraulic proportional valve driven by load impact grade and oil viscosity change rate; and distributing fusion weights of the preliminary control quantity and the current control parameter compensation quantity output by fuzzy reasoning based on the oil viscosity change rate to generate an initial control quantity resistant to hysteresis.
4. The electro-hydraulic proportional valve adaptive control method of a compressor cylinder as set forth in claim 2, wherein calculating the working condition urgency to dynamically allocate the compensation weight for the working condition category obtained by clustering conforming to the preset intervention condition, performing hierarchical compensation on the initial control amount according to the allocated compensation weight, and generating the final control amount includes: aiming at the working condition category obtained by clustering according with preset intervention conditions, calculating working condition urgency scores generated by coupling load impact strength, oil pollution degree and oil temperature in real time; Dynamically distributing compensation weights of load impact strength, oil pollution degree and oil temperature according to the working condition emergency degree score, wherein the higher the working condition emergency degree score is, the load impact compensation weight is increased, and the compensation weight of the oil pollution degree and the oil temperature is decreased; Performing hierarchical superposition compensation on the basic control quantity according to the determined compensation weight, wherein the load impact compensation quantity with high compensation weight is directly superposed, and the oil pollution degree of the secondary compensation weight and the oil temperature with low compensation weight are sequentially multiplied and corrected through corresponding compensation coefficients; When the control quantity or the state of the actuating mechanism exceeds the limit, the impact strength, the oil pollution degree and the oil temperature correction amplitude are gradually reduced according to the order of the compensation weight from small to large, and the final control quantity is output.
5. The method for adaptively controlling an electro-hydraulic proportional valve of a compressor cylinder according to claim 3, wherein inputting the reference control parameter and the final control quantity into a pre-trained physical enhanced neural network prediction model, predicting a valve element wear index and a wear increment by dynamically matching a nonlinear coupling relationship between an oil characteristic and a part kinematic characteristic, and performing one of a plurality of safety intervention operations based on a combination criterion of the wear index and the wear increment comprises: Performing time sequence alignment treatment on the reference control parameter and the final control quantity, extracting characteristics comprising oil viscosity change rate, oil temperature gradient characteristics and valve core displacement frequency response characteristics, and generating a structural characteristic matrix; Inputting the feature matrix into a physical enhanced physical information neural network model, forward calculating a theoretical viscosity value through a viscosity-temperature relation equation of an embedded model, generating an absolute residual error term corresponding to an actual measured value, inverting an equivalent friction coefficient by a combined embedded valve core wear equation by adopting an accompanying method, generating a dynamic residual error term corresponding to a preset working condition calibration value, and fusing the absolute residual error term and the dynamic residual error term to obtain an enhanced feature vector; A cross-domain attention weight distribution mechanism is adopted, the viscosity attenuation characteristic generated by the oil viscosity change rate and the displacement energy obtained by vibration energy integration of the valve core displacement frequency response characteristic after band-pass filtering are subjected to weighted fusion, and a joint characterization vector with a cross-domain coupling relation is output; based on the joint characterization vector, predicting and outputting the wear index and the wear increment of the valve core in parallel by utilizing the double-branch output layer of the model; when the abrasion index of the valve core is in a preset first early warning interval and the abrasion increment exceeds a first threshold value, performing self-adaptive relaxation adjustment of the final control quantity, extracting an absolute residual error item and a dynamic residual error item in the model, and adjusting at least part of weights of the reinforcement-learned reward value function in real time according to the absolute residual error item and the dynamic residual error item; outputting a prompt containing oil filter replacement early warning when the valve core wear index is in a preset second early warning interval and the wear increment exceeds a second threshold value, and adjusting at least part of weights of the multi-objective rewarding functions to be above a preset weight value; When the valve core wear index is larger than the preset upper limit value of the second early warning interval and the wear increment exceeds a third threshold value, the redundant valve group switching operation is forcedly executed, and the current working condition characteristics are written into the abnormal case database.
6. The electro-hydraulic proportional valve adaptive control method of the compressor cylinder as set forth in claim 5, wherein the physical information neural network model includes: the system comprises a two-channel physical embedding layer, a theoretical residual error generating module of a viscosity-temperature relation equation embedded in a forward channel integration manner, and an accompanying differential inversion module of a valve core abrasion equation embedded in a reverse channel integration manner; The self-adaptive residual fusion layer is used for dynamically mixing the absolute residual items and the dynamic residual items through the working condition sensing weight distributor and forming an enhanced feature vector through splicing; The cross-domain attention interaction layer comprises a physical guiding multi-head attention unit and is used for fusing viscosity attenuation trend characteristics and displacement high-frequency energy distribution characteristics by adopting a weight distribution mechanism guided by a physical priori; The dual-branch prediction output layer comprises a wear index branch and a wear increment branch and is used for outputting the wear index and the wear increment of the valve core in parallel.
7. The method for adaptively controlling an electro-hydraulic proportional valve of a compressor cylinder according to claim 5 or 6, The viscosity-temperature relation equation is configured as an exponential function of the oil viscosity associated temperature gradient at the reference temperature and is used for describing an exponential decay law of the oil viscosity along with the temperature change; the valve core wear equation is configured as a differential equation of the coupling effect of the contact pressure and the vibration energy of the preset frequency band, and is used for dynamically quantifying the wear rate increment under the friction coefficient of the material.
8. An electro-hydraulic proportional valve self-adaptive control system of a compressor oil cylinder, which is characterized by comprising: The data processing module is used for acquiring core working condition parameters and auxiliary operation parameters of the compressor oil cylinder and obtaining a working condition time sequence data set through pretreatment; The initial analysis module is used for performing reinforcement learning optimization after dividing rule subsets according to the reference control parameters extracted by the working condition time sequence data sets and working condition clustering, synchronously performing fuzzy reasoning and dynamic compensation of the current control parameters of the electro-hydraulic proportional valve based on the optimized rule subsets, and fusing to obtain initial control quantity; The final analysis module is used for calculating the working condition emergency degree to dynamically allocate the compensation weight according to the working condition category obtained by clustering according with the preset intervention condition, and carrying out hierarchical compensation on the initial control quantity according to the allocated compensation weight to generate a final control quantity; And the prediction feedback module is used for inputting the reference control parameters and the final control quantity into a pre-trained physical enhancement type neural network prediction model, predicting the wear index and the wear increment of the valve core by dynamically matching the nonlinear coupling relation between the oil characteristics and the part kinematic characteristics, and executing one operation of the multi-stage safety intervention operation based on the combination criterion of the wear index and the wear increment.
9. An electro-hydraulic proportional valve self-adaptive control device of a compressor oil cylinder is characterized by comprising: at least one processor; and a memory communicatively coupled to the at least one processor; Wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the electro-hydraulic proportional valve adaptive control method of the compressor ram of any one of claims 1-7.
10. A computer readable storage medium having stored thereon computer executable instructions, wherein the executable instructions when executed by a processor implement the electro-hydraulic proportional valve adaptive control method of a compressor ram as claimed in any one of claims 1 to 7.

Description

Electro-hydraulic proportional valve self-adaptive control method, system, equipment and medium for compressor oil cylinder Technical Field The invention relates to the technical field of intelligent control of hydraulic transmission systems, in particular to an electro-hydraulic proportional valve self-adaptive control method, system, equipment and medium of a compressor oil cylinder. Background In the field of control of electro-hydraulic proportional valves of compressor cylinders, a control method based on a fixed fuzzy rule and a temperature compensation mechanism is mostly adopted in the prior art. Such methods typically build a fuzzy rule base through preset input parameter dimensions and rely on initial expert experience to set membership function parameter ranges. However, as the input dimension setting fails to cover dynamic influencing factors such as oil state change, when the load is changed rapidly or the oil pollution degree is aggravated in the operation process of the compressor, the system is difficult to accurately identify parameter association characteristics, so that the suitability of a fuzzy rule is reduced, and further the phenomena of regulation hysteresis and pressure overshoot of the proportional valve are caused to be remarkable. In the parameter optimization process, the existing control method is usually subjected to self-adaptive adjustment based on a limited optimization target, and long-term operation indexes of valve core movement characteristics are not fully considered. The method needs to go through a longer parameter self-adaptive period in a new equipment debugging stage, and when the compressor switches different working modes, the parameter optimization efficiency is limited due to the fact that a working condition characteristic distinguishing mechanism is lost. In addition, in a multi-rule matching decision link, the existing scoring model mainly expands priority judgment around a single performance index, and the working condition characteristic differences corresponding to different compressor types cannot be effectively fused, so that rule selection accuracy is reduced, and fluctuation amplitude of control parameters is increased. Especially when the oil viscosity is in nonlinear change, the membership function of the fixed parameter structure is difficult to adapt to the dynamic change trend, and nonlinear distortion in the proportional valve adjusting process is further aggravated. Disclosure of Invention First, the technical problem to be solved In view of the defects and shortcomings of the prior art, the invention provides a self-adaptive control method, a system, equipment and a medium for an electrohydraulic proportional valve of a compressor oil cylinder, which solve the technical problems of poor working condition adaptability, control parameter mismatch caused by long debugging period and control fluctuation aggravated on one side of a rule scoring model due to insufficient dimension of a fuzzy rule in the control technology of the electrohydraulic proportional valve of the compressor oil cylinder. (II) technical scheme In order to achieve the above purpose, the main technical scheme adopted by the invention comprises the following steps: In a first aspect, an embodiment of the present invention provides a method for adaptively controlling an electrohydraulic proportional valve of a compressor cylinder, including: Collecting core working condition parameters and auxiliary operation parameters of a compressor oil cylinder, and preprocessing to obtain a working condition time sequence data set; According to the reference control parameters extracted from the working condition time sequence data set, performing reinforcement learning optimization after dividing rule subsets through working condition clustering, synchronously executing fuzzy reasoning and dynamic compensation of the current control parameters of the electro-hydraulic proportional valve based on the optimized rule subsets, and fusing to obtain initial control quantity; aiming at the working condition category obtained by clustering according with preset intervention conditions, calculating working condition emergency degree to dynamically allocate compensation weights, and carrying out hierarchical compensation on the initial control quantity according to the allocated compensation weights to generate a final control quantity; The reference control parameters and the final control quantity are input into a pre-trained physical enhancement type neural network prediction model, the abrasion index and the abrasion increment of the valve core are predicted through dynamically matching the nonlinear coupling relation between the oil characteristics and the part kinematic characteristics, and one operation among the multistage safety intervention operations is executed based on the combination criteria of the abrasion index and the abrasion increment. Optionally, collecting core working conditio