A remote control system for an electromagnetic directional valve and a data processing method

Abstract

The present invention relates to the technical field of electromagnetic reversing valve control, and discloses a remote control system and data processing method for an electromagnetic reversing valve. The system includes modules for receiving, parsing, determining dynamic response parameters of control instructions, collecting, uploading, analyzing status data, and executing control signals. The method is as follows: receiving a reversing control instruction from a remote terminal, parsing it into adjustment parameters such as the number of actions, switching direction, and pressure threshold; determining the dynamic response parameters of the control node based on the parameters, performing adjustments and collecting status data such as current and magnetic flux; uploading the data to an edge computing unit, generating a pulse width modulation correction value for the first control node through processing such as data fusion, frequency domain correlation analysis, and wavelet domain dynamic compensation aggregation; and transmitting it to a fuzzy controller to calibrate the parameters of the drive circuit. The present invention improves the remote control accuracy and stability and is applicable to industrial automation intelligent control.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic directional valve control, and particularly to a remote control system and data processing method for an electromagnetic directional valve. Background Art

[0002] Under the traditional control mode, the parameter adjustment of the electromagnetic directional valve depends on local hardware or on-site manual operation, which cannot meet the requirements of remote real-time control in complex industrial scenarios. For example, in the fields of petrochemical industry, intelligent equipment manufacturing, etc., the equipment is usually distributed in a vast area or a high-risk environment, and it is difficult for operators to obtain the operating status of each node in real time and adjust the parameters, resulting in low efficiency of working condition switching and even production accidents caused by control lag. At the same time, the traditional system lacks a multi-device cooperative control mechanism. When multiple electromagnetic directional valves need to work in cooperation, it is difficult to achieve action synchronization through a unified remote command, affecting the overall coordination of the system.

[0003] The acquisition and analysis of the state data of the electromagnetic directional valve by traditional control technologies stay at a single dimension, and only basic physical quantities such as current and pressure can be obtained, and the data processing method is simple, and the implicit correlations between different parameters cannot be mined. For example, there may be complex relationships between current fluctuations and magnetic flux changes, spool wear degree, etc., but the traditional system only judges the equipment status by setting a single threshold, which is easy to miss early faults or mis-trigger alarms. In addition, the state data is usually only stored or displayed locally, and no closed-loop feedback is formed with the control strategy, and the control parameters cannot be dynamically adjusted according to the real-time data, resulting in a decrease in control accuracy as the equipment operation time increases. Especially in complex working conditions such as high-frequency commutation and variable load, the problem of parameter drift is more significant.

[0004] Traditional control algorithms (such as PID control) are based on fixed parameter logics and are difficult to adapt to the non-linear characteristics of the electromagnetic directional valve at different working stages. For example, the spool is greatly affected by inertial force during the start-up stage, and it needs to cope with interference factors such as hydraulic oil viscosity change and temperature drift during steady-state operation. The traditional algorithm cannot compensate these dynamic errors in real time, which may cause problems such as action delay, excessive impact current or system oscillation. At the same time, when multiple control nodes work together, there is a lack of effective frequency domain feature analysis and interference suppression mechanisms, and signal coupling between nodes is likely to cause resonance or inconsistent responses, further reducing the system stability.

[0005] Under the industrial Internet architecture, due to poor protocol compatibility and insufficient edge computing capabilities, the traditional electromagnetic valve control system is difficult to achieve seamless docking with cloud platforms and upper-level management systems. For example, device status data cannot be efficiently uploaded to the cloud for big data analysis, restricting the implementation of intelligent applications such as predictive maintenance and energy consumption optimization. In addition, the hardware architecture and software modules of traditional systems have poor scalability. When facing the flexible production requirements of multiple varieties and small batches, it is difficult to quickly reconstruct control strategies or adapt to different models of devices, increasing the cost of upgrading and transforming production lines.

[0006] In the prior art, although some research has attempted to introduce Internet of Things technology and machine learning algorithms into electromagnetic valve control, most of them stay at the level of single-point data monitoring or offline model training, and have not constructed a complete closed-loop from data acquisition, multi-dimensional analysis to dynamic control. For example, the remote instruction issuing technology based on wireless transmission fails to solve the computational complexity problem brought by multi-node data fusion, and the fault diagnosis model based on deep learning cannot achieve real-time online inference due to limited edge computing resources. Therefore, there is an urgent need for a new control system and method with the capabilities of remote instruction parsing, multi-dimensional data correlation analysis, and dynamic parameter calibration to improve the adaptability and reliability of electromagnetic valves in industrial intelligent scenarios. Summary of the Invention

[0007] The purpose of the present invention is to provide a remote control system and data processing method for electromagnetic valves to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solutions: A remote control system and data processing method for electromagnetic valves, the method includes:

[0009] Receiving a commutation control instruction sent by a remote terminal;

[0010] Parsing the commutation control instruction to obtain adjustment parameters of the electromagnetic valve, where the adjustment parameters include the number of actions, switching direction, and pressure threshold;

[0011] Based on the adjustment parameters, determining dynamic response parameters of multiple control nodes of the electromagnetic valve;

[0012] Performing real-time adjustment actions based on the dynamic response parameters of the multiple control nodes, and synchronously collecting status data of the multiple control nodes, where the status data includes current data and magnetic flux data;

[0013] Uploading the status data of the multiple control nodes to the edge computing unit;

[0014] In the edge computing unit, multi-dimensional correlation analysis is performed on the status data of the multiple control nodes to generate a control signal, where the control signal is the pulse width modulation correction value of the first control node;

[0015] The control signal is transmitted to a fuzzy controller to calibrate the parameters of the drive circuit corresponding to the first control node.

[0016] Preferably, in the edge computing unit, multi-dimensional correlation analysis is performed on the status data of the multiple control nodes to generate a control signal, where the control signal is the pulse width modulation correction value of the first control node, and it includes:

[0017] Data fusion is performed on the status data of the multiple control nodes to generate a frequency domain sequence of the status data of the multiple control nodes;

[0018] Frequency domain implicit correlation analysis is performed on the frequency domain sequence of the status data of the multiple control nodes to generate frequency domain correlation feature vectors of the multiple control nodes;

[0019] Wavelet domain dynamic compensation aggregation is performed on the frequency domain correlation feature vectors of the multiple control nodes to generate a frequency domain aggregation characterization matrix of the control nodes;

[0020] The first frequency domain correlation feature vector is extracted from the frequency domain correlation feature vectors of the multiple control nodes, and the matching degree of the first frequency domain correlation feature vector relative to the frequency domain aggregation characterization matrix of the control nodes is calculated to generate a first frequency domain matching feature matrix;

[0021] Based on the first frequency domain matching feature matrix, the control signal is generated.

[0022] Preferably, data fusion is performed on the status data of the multiple control nodes to generate a frequency domain sequence of the status data of the multiple control nodes, and it includes:

[0023] The status data of the multiple control nodes are subjected to data reduction according to the frequency domain dimension and the adjustment parameter sample dimension to generate a frequency domain sequence of the current data of the multiple control nodes and a frequency domain sequence of the magnetic flux data of the multiple control nodes;

[0024] The corresponding magnetic flux data is added after each current data in the frequency domain sequence of the current data of the multiple control nodes to generate the frequency domain sequence of the status data of the multiple control nodes.

[0025] Preferably, frequency domain implicit correlation analysis is performed on the frequency domain sequence of the status data of the multiple control nodes to generate frequency domain correlation feature vectors of the multiple control nodes, and it includes: inputting the frequency domain sequence of the status data of the multiple control nodes into a frequency domain encoder based on a wavelet transform-Fourier hybrid model to generate the frequency domain correlation feature vectors of the multiple control nodes.

[0026] Preferably, performing wavelet domain dynamic compensation aggregation on the frequency domain correlation feature vectors of the multiple control nodes to generate a frequency domain aggregation representation matrix of the control nodes includes:

[0027] Inputting the frequency domain correlation feature vectors of the multiple control nodes into a wavelet domain feature extraction network to generate a control node frequency domain wavelet feature matrix;

[0028] Based on the control node frequency domain wavelet feature matrix, performing fuzzy membership modulation on the frequency domain correlation feature vectors of each control node among the frequency domain correlation feature vectors of the multiple control nodes to generate a fuzzy frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes;

[0029] Fusing the control node frequency domain wavelet feature matrix and the fuzzy frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes to generate the frequency domain aggregation representation matrix of the control nodes.

[0030] Preferably, based on the control node frequency domain wavelet feature matrix, performing fuzzy membership modulation on the frequency domain correlation feature vectors of each control node among the frequency domain correlation feature vectors of the multiple control nodes to generate a fuzzy frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes includes:

[0031] Calculating the complementary information of the frequency domain correlation feature vectors of each control node among the frequency domain correlation feature vectors of the multiple control nodes relative to the control node frequency domain wavelet feature matrix to generate a frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes;

[0032] Inputting the frequency domain node-wavelet complementary information embedding coding matrix of each control node in the frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes into a complementary information saliency identification module based on fuzzy rules to generate fuzzy weights of the frequency domain complementary information of the multiple control nodes;

[0033] Based on the fuzzy weights of the frequency domain complementary information of the multiple control nodes, performing membership modulation on the frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes to generate the fuzzy frequency domain node-wavelet complementary information embedding coding matrix of the multiple control nodes.

[0034] Preferably, extracting a first frequency domain correlation feature vector from the frequency domain correlation feature vectors of the multiple control nodes and calculating the matching degree of the first frequency domain correlation feature vector relative to the frequency domain aggregation representation matrix of the control nodes to generate a first frequency domain matching feature matrix includes:

[0035] Extract the frequency-domain correlation feature vector corresponding to the first control node from the frequency-domain correlation feature vectors of the multiple control nodes to generate the first frequency-domain correlation feature vector;

[0036] Calculate the covariance matrix between the first frequency-domain correlation feature vector and the frequency-domain aggregation representation matrix of the control nodes to generate the first frequency-domain matching feature matrix.

[0037] Preferably, based on the first frequency-domain matching feature matrix, generating the control signal includes: passing the first frequency-domain matching feature matrix through a control signal generator based on fuzzy inference to generate the control signal.

[0038] Preferably, the present invention further includes an electromagnetic directional valve remote control system, and the system includes:

[0039] A control instruction receiving module, configured to receive a commutation control instruction sent by a remote terminal;

[0040] A control instruction parsing module, configured to parse the commutation control instruction to obtain the adjustment parameters of the electromagnetic directional valve, where the adjustment parameters include the number of actions, the switching direction, and the pressure threshold;

[0041] A dynamic response parameter determination module, configured to determine the dynamic response parameters of multiple control nodes of the electromagnetic directional valve based on the adjustment parameters;

[0042] A state data acquisition module, configured to perform real-time adjustment actions based on the dynamic response parameters of the multiple control nodes, and synchronously acquire the state data of the multiple control nodes, where the state data includes current data and magnetic flux data;

[0043] A state data uploading module, configured to upload the state data of the multiple control nodes to an edge computing unit;

[0044] A state data analysis module, configured to perform multi-dimensional correlation analysis on the state data of the multiple control nodes in the edge computing unit to generate a control signal, where the control signal is the pulse width modulation correction value of the first control node;

[0045] A control signal execution module, configured to transmit the control signal to a fuzzy controller to calibrate the parameters of the drive circuit corresponding to the first control node.

[0046] Preferably, the state data analysis module includes:

[0047] A state data fusion unit, configured to perform data fusion on the state data of the multiple control nodes to generate the state data frequency-domain sequences of the multiple control nodes;

[0048] a frequency domain correlation analysis unit, configured to perform frequency domain implicit correlation analysis on the frequency domain sequences of the status data of the plurality of control nodes to generate frequency domain correlation feature vectors of the plurality of control nodes;

[0049] A wavelet domain compensation aggregation unit, configured to perform wavelet domain dynamic compensation aggregation on the frequency domain correlation feature vectors of the plurality of control nodes to generate a frequency domain aggregation representation matrix of the control nodes;

[0050] a matching feature calculation unit, configured to extract a first frequency domain correlation feature vector from the frequency domain correlation feature vectors of the plurality of control nodes, and calculate a matching degree of the first frequency domain correlation feature vector relative to a frequency domain aggregation representation matrix of the control node to generate a first frequency domain matching feature matrix;

[0051] A control signal generating unit is configured to generate the control signal based on the first frequency domain matching feature matrix.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] By integrating remote command interaction, multi-dimensional data fusion analysis, and dynamic control technologies, a high-precision closed-loop control system has been built, significantly improving the control performance and reliability of solenoid directional valves in industrial automation scenarios. Specific improvements are as follows:

[0054] The system supports receiving reversing control commands sent by remote terminals and parsing them into adjustment parameters such as the number of movements, switching direction, and pressure threshold, enabling remote and dynamic configuration of the electromagnetic reversing valve's operating mode. This feature breaks through the geographical limitations of traditional local control and is particularly suitable for the coordinated control of multiple devices in distributed industrial systems. For example, in smart workshops or energy pipeline networks, operators can issue commands in real time through remote terminals to quickly adjust parameters and switch operating conditions for multi-node devices, reducing manual intervention costs and improving production efficiency and operational flexibility.

[0055] The system achieves closed-loop feedback from data acquisition to control signal generation by synchronously collecting status information such as current and magnetic flux data from multiple control nodes and combining it with the multi-dimensional correlation analysis capabilities of edge computing units. Specifically, by generating a frequency domain sequence of status data through data fusion and extracting frequency domain correlation feature vectors using a wavelet transform-Fourier hybrid model, the system can deeply explore the implicit correlation between current and magnetic flux changes and accurately identify abnormal trends in equipment operation (such as changes in frequency components during the initial stages of valve core wear). This multi-dimensional analysis mechanism overcomes the limitations of traditional single-point threshold detection and enhances the system's perception of complex operating conditions and its fault warning capabilities.

[0056] Based on the wavelet domain dynamic compensation aggregation algorithm, the system performs fuzzy membership modulation and information fusion on the frequency domain correlation feature vectors of each control node to generate a global frequency domain aggregation representation matrix, realizing the optimization of the collaborative work of multiple control nodes. By calculating the matching degree between the frequency domain features of the target node and the global matrix and combining fuzzy inference to generate a pulse width modulation (PWM) correction value, the parameter deviation of the drive circuit can be compensated in real time, improving the control accuracy. This mechanism can adapt to the non-linear characteristics of the electromagnetic reversing valve in different working stages (such as the inertial force in the starting stage and the temperature drift in the steady state stage), reduce the action delay or system oscillation caused by parameter drift, and enhance the robustness of the control strategy.

[0057] The system realizes the efficient processing of local data and the generation of control signals through the edge computing unit, reducing the dependence on the cloud server, shortening the data processing and response time, and ensuring the real-time nature of control. At the same time, it supports uploading status data, control logs, etc. to the cloud for long-term storage and big data analysis, providing data support for the predictive maintenance of equipment, operation efficiency optimization, etc., and promoting the upgrade of industrial automation systems to intelligent and refined management. The modular system architecture design (such as the control instruction receiving module, status data analysis module) facilitates docking with different industrial protocols, improving the compatibility and expansion ability of the system, and can be flexibly adapted to various models of electromagnetic reversing valves and industrial application scenarios. Brief Description of the Drawings

[0058] Figure 1 It is the working principle diagram of the remote control system and data processing method of the electromagnetic reversing valve described in the present invention;

[0059] Figure 2 It is the working principle diagram of the edge computing unit generating control signals through multi-dimensional correlation analysis;

[0060] Figure 3 It is the working principle diagram of fusing status data to generate the frequency domain sequence of status data;

[0061] Figure 4 It is the working principle diagram of dynamically compensating and aggregating in the wavelet domain to generate a frequency domain aggregation representation matrix;

[0062] Figure 5 It is the working principle diagram of generating a fuzzy frequency domain node-wavelet complementary information embedding coding matrix through fuzzy membership modulation. Detailed Embodiment

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0064] Please refer to Figures 1 - 5 , an electromagnetic directional valve remote control system and data processing method involved in the present invention, and the specific implementation steps are as follows:

[0065] The system, through the control instruction receiving module, monitors and obtains in real time the commutation control instructions sent by the remote terminal. This instruction contains the operation intention for the electromagnetic directional valve and is the starting input of the entire control process.

[0066] The control instruction parsing module parses and processes the received commutation control instructions. Through the preset instruction format and protocol, the adjustment parameters of the electromagnetic directional valve are extracted therefrom, specifically including the number of actions, switching direction, and pressure threshold. These parameters are the core basis for subsequent control operations and determine the working mode and performance requirements of the electromagnetic directional valve.

[0067] Based on the above adjustment parameters, the dynamic response parameter determination module calculates and determines the dynamic response parameters of multiple control nodes of the electromagnetic directional valve in combination with the structural characteristics and working principle of the electromagnetic directional valve. These dynamic response parameters characterize the response characteristics that each control node should have under specific adjustment parameters, such as response time, action amplitude, etc., and provide specific control objectives for subsequent real-time adjustment.

[0068] The state data acquisition module drives the electromagnetic directional valve to perform real-time adjustment actions according to the dynamic response parameters of multiple control nodes. During the adjustment process, the state data of each control node is synchronously acquired, and the state data includes current data and magnetic flux data. These data reflect the working state of the control node in real time and are the basis for subsequent analysis and optimization of control.

[0069] The state data uploading module uploads the state data of multiple control nodes collected through network transmission and other means to the edge computing unit for further analysis and processing.

[0070] In the edge computing unit, the state data analysis module performs multi-dimensional correlation analysis on the state data of multiple control nodes uploaded. Through specific analysis algorithms and models, the internal correlations and laws between the data are mined, and finally a control signal is generated, and this control signal is the pulse width modulation correction value of the first control node.

[0071] The control signal execution module transmits the generated control signal to the fuzzy controller, and the fuzzy controller calibrates the parameters of the drive circuit corresponding to the first control node according to the control signal, so as to optimize and adjust the control performance of the electromagnetic reversing valve.

[0072] Embodiment 1: In the process of the edge computing unit performing multi-dimensional correlation analysis on the status data of multiple control nodes to generate a control signal (the pulse width modulation correction value of the first control node), the following steps are specifically included:

[0073] Perform status data fusion. The status data (current data and magnetic flux data) of multiple control nodes are respectively subjected to data reduction according to the frequency domain dimension and the adjustment parameter sample dimension. In the frequency domain dimension, the time-domain data is converted into a frequency-domain representation through methods such as Fourier transform, and features such as the amplitude and phase of each frequency component are extracted; in the adjustment parameter sample dimension, the data is grouped and screened according to parameters such as the number of actions, switching direction, and pressure threshold, redundant samples are removed, and the valid data matching the current adjustment parameter is retained, so as to generate the current data frequency-domain sequence and the magnetic flux data frequency-domain sequence of multiple control nodes. Subsequently, each data point in the current data frequency-domain sequence is associated with the corresponding magnetic flux data, that is, the magnetic flux data of the same node and the same frequency point is appended after each current data, forming a status data frequency-domain sequence containing dual physical quantity information, realizing the fusion of current and magnetic flux data at the frequency domain level, and providing composite feature data for subsequent analysis.

[0074] Perform frequency-domain implicit correlation analysis. The above-mentioned status data frequency-domain sequence is input into the frequency-domain encoder based on the wavelet transform-Fourier hybrid model. The encoder first performs global frequency-domain decomposition on the fused frequency-domain sequence through Fourier transform to obtain the energy distribution characteristics of each frequency component; then uses the multi-resolution analysis ability of wavelet transform to perform hierarchical processing on the frequency-domain signal, extracts the detailed characteristics in different frequency sub-bands, and captures the dynamic change law of the data in the local frequency band. Through the collaborative processing of the hybrid model, the correlation characteristics between current and magnetic flux in the frequency-domain sequence are converted into the frequency-domain correlation feature vectors of multiple control nodes, and each vector represents the coupling relationship between current and magnetic flux and the potential correlation between nodes in the frequency-domain space corresponding to the control node.

[0075] Perform wavelet-domain dynamic compensation aggregation. In the first step, input the frequency-domain correlation feature vector into the wavelet-domain feature extraction network. Through the scale decomposition of wavelet transform, extract the frequency-domain wavelet feature matrix of control nodes. This matrix contains the frequency-domain detail features of each node at different wavelet scales, reflecting the local characteristics of the signal at multiple scales. In the second step, based on the frequency-domain wavelet feature matrix of control nodes, calculate the complementary information between each frequency-domain correlation feature vector and this matrix, and generate the frequency-domain node-wavelet complementary information embedding coding matrix. The complementary information is obtained by comparing the differences between the feature vector and the wavelet feature matrix, representing the frequency-domain information unique to each node that is not covered by the wavelet features. In the third step, input the complementary information embedding coding matrix into the complementary information saliency identification module based on fuzzy rules. The module performs fuzzy quantization on the complementary information in each matrix according to the preset fuzzy rules (such as setting the weight intervals corresponding to different complementary information values), and generates the fuzzy weight of the frequency-domain complementary information. The weight value reflects the importance of the complementary information for analyzing the node state. In the fourth step, use the fuzzy weight to perform membership modulation on the complementary information embedding coding matrix, that is, multiply each element in the matrix by the corresponding weight, highlight the key complementary information, suppress the secondary information, and generate the fuzzy frequency-domain node-wavelet complementary information embedding coding matrix. Finally, fuse the frequency-domain wavelet feature matrix of control nodes and the fuzzy frequency-domain node-wavelet complementary information embedding coding matrix, and generate the frequency-domain aggregation representation matrix of control nodes through matrix splicing or weighted summation, etc. This matrix synthesizes the global frequency-domain features, multi-scale detail features, and personalized complementary information of the nodes, forming a comprehensive representation of the node state.

[0076] Then perform matching feature calculation. Extract the frequency-domain correlation feature vector corresponding to the first control node from the frequency-domain correlation feature vectors of multiple control nodes, that is, the first frequency-domain correlation feature vector. This vector represents the frequency-domain state feature of the target node. Then calculate the covariance matrix between the first frequency-domain correlation feature vector and the frequency-domain aggregation representation matrix. The elements of the covariance matrix reflect the linear correlation between each dimension of the first frequency-domain correlation feature vector and the corresponding dimensions of the node feature vectors in the frequency-domain aggregation representation matrix, thereby generating the first frequency-domain matching feature matrix. This matrix quantifies the matching degree and correlation strength between the target node and other nodes in the frequency-domain feature space.

[0077] Finally, control signal generation is performed. The first frequency-domain matching feature matrix is input into the control signal generator based on fuzzy inference. There is a preset fuzzy rule base inside the generator. The rule base is established based on the control logic of the electromagnetic directional valve and engineering experience. For example, the pulse width modulation correction direction and amplitude corresponding to different covariance value intervals are set. Through fuzzy processing, the numerical values in the matrix are converted into fuzzy linguistic variables (such as "high matching degree", "medium matching degree", "low matching degree"). Reasoning is carried out according to the fuzzy rules, and then a specific control signal, that is, the pulse width modulation correction value of the first control node, is output through defuzzification calculation. This correction value is used to adjust the pulse width parameter of the target node drive circuit to achieve real-time calibration of the drive circuit, so as to optimize the control accuracy and dynamic response characteristics of the electromagnetic directional valve.

[0078] Embodiment 2: In the process of performing data fusion on the state data of multiple control nodes to generate a state data frequency-domain sequence, first, data reduction in the frequency domain dimension is carried out. For the current data of each control node, Fourier transform is used to convert it from the time domain to the frequency domain. Fourier transform decomposes the time-domain signal into the superposition of sine waves of different frequencies, and obtains the amplitude spectrum and phase spectrum of the current data in the frequency domain, thereby generating a current data frequency-domain sequence. In this process, according to the working frequency range of the electromagnetic directional valve (such as the preset effective frequency band of 10Hz - 1000Hz), high-frequency noise and low-frequency interference outside this range are filtered out, and only the frequency components related to the dynamic characteristics of the valve parts are retained. For example, for the current time-domain data with a sampling frequency of 2000Hz, the amplitudes of each frequency point are calculated through discrete Fourier transform (DFT), the low-frequency interference of mechanical vibration below 10Hz and the electronic noise frequency band above 1000Hz are removed, and only the frequency-domain data points within 10Hz - 1000Hz are retained to form a refined current data frequency-domain sequence.

[0079] Data reduction in the frequency domain dimension is performed on the magnetic flux data. The magnetic flux data reflects the change of the internal magnetic field of the electromagnetic directional valve. The magnetic flux time-domain signal collected by devices such as Hall sensors is converted into a frequency-domain representation after Fourier transform. Since the magnetic field change is directly related to the current drive and its effective frequency range is the same as that of the current data, the same frequency interval filtering rule is adopted to generate a magnetic flux data frequency-domain sequence. The magnetic flux data frequency-domain sequence of each control node corresponds one-to-one with the current data frequency-domain sequence at the frequency points to ensure the dimensional consistency during subsequent fusion.

[0080] Perform data reduction on the sample dimension of the adjustment parameters. The adjustment parameters include the number of actions, the switching direction, and the pressure threshold, and each adjustment parameter corresponds to a different working scenario. For example, when the number of actions in the commutation control instruction sent by the remote terminal is "100 times / minute", the switching direction is "forward", and the pressure threshold is "5 MPa", the system forms a unique sample identifier according to the combination of these three parameters. During data processing, traverse the original state data of all control nodes, filter out the data samples that match the current sample identifier, and eliminate the irrelevant data under other adjustment parameters. For example, the current and magnetic flux data collected by a control node when the number of actions is "80 times / minute" will be filtered, and only the corresponding data under the "100 times / minute" sample will be retained, so as to generate a subset of the current data frequency domain sequence and the magnetic flux data frequency domain sequence that match the current adjustment parameters.

[0081] After completing the double reduction of the frequency domain dimension and the adjustment parameter sample dimension, enter the data fusion stage. For each control node, associate each data point in its current data frequency domain sequence with the corresponding magnetic flux data. The specific operation is as follows: Under the same adjustment parameter sample, for each frequency point (such as f = 50 Hz) in the current data frequency domain sequence, find the magnetic flux value at the same frequency point in the magnetic flux data frequency domain sequence, and use this magnetic flux value as an additional dimension to splice it after the current data to form a composite data point , where is the current frequency domain amplitude, is the magnetic flux frequency domain amplitude. For example, if the current frequency domain amplitude of control node A at f = 50 Hz is 0.8 A and the corresponding magnetic flux frequency domain amplitude is 0.05 Wb, the composite data at this point after fusion is represented as .

[0082] Through the above steps, the state data frequency domain sequence of each control node is extended from a single current frequency domain sequence to a composite frequency domain sequence containing both current and magnetic flux physical quantities. This fusion method not only retains the independent characteristics of each physical quantity in the frequency domain (such as the fundamental wave component and harmonic distribution of the current, and the frequency response characteristics of the magnetic flux), but also establishes the corresponding relationship between the two at the same frequency point, providing a basis for subsequent analysis of the current-magnetic flux coupling characteristics (such as electromagnetic conversion efficiency and hysteresis loss characteristics). For example, at a specific frequency point, if the current amplitude increases while the magnetic flux amplitude does not increase synchronously, it may indicate the presence of magnetic saturation. This composite data can be directly used for frequency domain implicit correlation analysis without additional data alignment processing.

[0083] The frequency-domain sequences of the state data of the multiple generated control nodes are stored in matrix form, with each row corresponding to a control node, each column corresponding to a frequency point, and each element being the current and magnetic flux composite value of the node at the corresponding frequency point. For example, the matrix dimension of the frequency-domain sequence of state data containing 3 control nodes and 100 frequency points is 3×100×2, where the third dimension is the physical quantity dimension (current, magnetic flux). This structured data format facilitates subsequent input to the frequency-domain encoder for feature extraction and meets the computational requirements of multi-node parallel analysis.

[0084] The entire data fusion process ensures the effectiveness and pertinence of the input data through data reduction in dual dimensions, enhances the features of multi-source data through physical quantity association, provides a standardized composite feature input for subsequent frequency-domain implicit correlation analysis, and lays the foundation for multi-dimensional association analysis.

[0085] Embodiment 3: The process of performing frequency-domain implicit correlation analysis on the frequency-domain sequences of the state data of multiple control nodes to generate frequency-domain correlation feature vectors is specifically implemented as follows:

[0086] Input the frequency-domain sequences of the state data of multiple control nodes into a frequency-domain encoder based on a wavelet transform-Fourier hybrid model. The processing flow of this encoder is divided into two stages: global frequency-domain decomposition by Fourier transform and local detail extraction by wavelet transform.

[0087] In the Fourier transform stage, the encoder performs global frequency-domain analysis on the frequency-domain sequences of the state data of each control node (including the composite frequency-domain data of current and magnetic flux). Through the mathematical principle of Fourier transform, the current amplitude, phase, magnetic flux amplitude, and phase at each frequency point in the sequence are decomposed into the superposition of different frequency components, forming a frequency-domain energy distribution map. For example, for the composite frequency-domain data of a certain control node in the frequency band of 10Hz - 1000Hz, Fourier transform can identify the energy ratios of the fundamental frequency (such as 50Hz), harmonics (such as 100Hz, 150Hz, etc.), and the phase differences between current and magnetic flux at each frequency point (reflecting the lag characteristics of electromagnetic coupling). The processing in this stage can capture the distribution characteristics of the state data in the overall frequency domain, such as the position of the main energy frequency band and the richness of harmonic components, providing a frequency-domain feature basis from a global perspective for subsequent analysis.

[0088] In the wavelet transform stage, the encoder utilizes the multi-resolution analysis feature of wavelet transform to perform hierarchical processing on the frequency-domain signal after Fourier transform. Specifically, an appropriate wavelet basis function (such as Daubechies wavelet or Symlet wavelet) is selected to perform multi-level decomposition on the frequency-domain signal (for example, decomposed into 3 levels), dividing the entire frequency-domain range into different sub-bands (such as low-frequency sub-band, medium-frequency sub-band, high-frequency sub-band). Each sub-band corresponds to a different frequency interval and can capture the detailed changes of the signal within a specific frequency band. For example, the low-frequency sub-band (10Hz - 125Hz) can reflect the mechanical vibration fundamental frequency characteristics of the electromagnetic directional valve, the medium-frequency sub-band (125Hz - 500Hz) corresponds to the dynamic response characteristics of electromagnetic drive, and the high-frequency sub-band (500Hz - 1000Hz) may contain high-frequency components such as electronic noise or local magnetic domain changes. Through this hierarchical processing, the encoder can extract detailed features at different scales from the frequency-domain signal, such as the stable trend in the low-frequency band and the transient fluctuations in the medium- and high-frequency bands.

[0089] Under the synergistic effect of the hybrid model, the frequency-domain encoder performs cross-physical quantity correlation analysis on the current and magnetic flux data in the state data frequency-domain sequence. For example, at the fundamental frequency point, analyze the proportional relationship between the current amplitude and the magnetic flux amplitude (reflecting the electromagnetic conversion efficiency); at the harmonic point, analyze the phase difference change between the current and the magnetic flux (reflecting the hysteresis loss characteristics). Through this analysis, the encoder can identify the coupling mode of current and magnetic flux in the frequency-domain space for different control nodes, as well as the potential correlation between nodes (such as whether there is co-frequency interference or synchronous response between adjacent nodes).

[0090] Finally, the frequency-domain encoder converts the state data frequency-domain sequence of each control node into a frequency-domain correlation feature vector. Each element of this vector corresponds to a specific frequency-domain feature dimension, such as fundamental frequency current amplitude, fundamental frequency magnetic flux amplitude, fundamental frequency phase difference, proportion of harmonic energy of each order, wavelet sub-band energy distribution, etc. These feature dimensions numerically characterize the correlation characteristics of current and magnetic flux in the frequency domain for the control node and the implicit correlation between nodes. For example, the frequency-domain correlation feature vector of a certain control node may contain the following dimensions: 50Hz current amplitude, 50Hz magnetic flux amplitude, 50Hz current-magnetic flux phase difference, proportion of 100Hz current harmonic energy, proportion of 100Hz magnetic flux harmonic energy, proportion of low-frequency wavelet sub-band energy, proportion of high-frequency wavelet sub-band energy, etc.

[0091] It should be noted that the parameters of the frequency-domain encoder (such as the number of wavelet decomposition levels and the frequency resolution of the Fourier transform) can be adjusted according to the specific model of the electromagnetic directional valve, the working frequency range, and the control accuracy requirements. For example, for valve components with high requirements for high-frequency response, the frequency resolution of the Fourier transform can be increased and the number of wavelet decomposition levels can be increased to capture more subtle high-frequency features; for valve components mainly characterized by low-frequency mechanical properties, the analysis of the high-frequency band can be simplified and the feature extraction of the low-frequency sub-bands can be focused on.

[0092] Through the above processing based on the wavelet transform-Fourier hybrid model, the frequency-domain encoder realizes multi-level feature extraction of the frequency-domain sequence of state data, retains both the macroscopic features of the global frequency-domain distribution and explores the microscopic dynamics of local frequency bands. The generated frequency-domain correlation feature vector can comprehensively and accurately reflect the electromagnetic coupling state of the control nodes and the implicit correlation between nodes, providing key feature inputs for subsequent wavelet-domain dynamic compensation aggregation and control signal generation.

[0093] Example 4: The process of performing wavelet-domain dynamic compensation aggregation on the frequency-domain correlation feature vectors of multiple control nodes to generate the frequency-domain aggregation characterization matrix of the control nodes is specifically implemented as follows:

[0094] Input the frequency-domain correlation feature vectors of multiple control nodes into the wavelet-domain feature extraction network. Based on the multi-scale analysis principle of wavelet transform, this network decomposes each frequency-domain correlation feature vector. By selecting a specific wavelet basis function (such as a wavelet basis with compact support properties), the frequency-domain correlation feature vector is transformed from the original feature space to the wavelet domain, generating a frequency-domain wavelet feature matrix of the control nodes. Each row of this matrix corresponds to a control node, and each column corresponds to a scale or frequency sub-band in the wavelet domain. The matrix element value reflects the feature intensity of the node at the corresponding wavelet scale. For example, if the wavelet domain is divided into 3 scales (low frequency, medium frequency, high frequency), then the frequency-domain wavelet feature matrix of each node contains 3 eigenvalues, corresponding to the energy distribution or detail features at different scales.

[0095] Based on the frequency-domain wavelet feature matrix of the control nodes, perform fuzzy membership modulation on the frequency-domain correlation feature vectors of each control node, which is specifically divided into the following steps:

[0096] Complementary information calculation: For the frequency-domain correlation eigenvector of each control node, calculate its complementary information with the frequency-domain wavelet feature matrix of the control node. The complementary information is obtained by comparing the original eigenvalues of this vector with the eigenvalues of the corresponding nodes in the wavelet feature matrix, reflecting the personalized features not included in the wavelet feature matrix. For example, if the value of the frequency-domain correlation eigenvector of a certain node in the dimension of "fundamental frequency current phase difference" is 0.8, and the value of the "low-frequency scale feature" of the corresponding node in the frequency-domain wavelet feature matrix is 0.6, then the difference value of 0.2 is regarded as the complementary information of this dimension, indicating that this node has uniqueness in the fundamental frequency phase difference beyond the low-frequency features in the wavelet domain.

[0097] Generate the frequency-domain node-wavelet complementary information embedding coding matrix: Organize the complementary information of each node according to the feature dimensions to form the complementary information embedding coding matrix of this node. The dimension of this matrix is the same as that of the frequency-domain correlation eigenvector, and each element corresponds to the complementary information value of a feature dimension, thus structurally embedding the complementary information into the feature representation.

[0098] Fuzzy saliency identification: Input the frequency-domain node-wavelet complementary information embedding coding matrix of each node into the complementary information saliency identification module based on fuzzy rules. There is a preset fuzzy rule set inside the module. For example, "high saliency", "medium saliency", "low saliency" and other fuzzy linguistic variables are set according to the size of the complementary information value. Through fuzzy processing (such as using a triangular membership function), the complementary information value is converted into the corresponding fuzzy membership value to generate the frequency-domain complementary information fuzzy weight. For example, if the complementary information value of a certain node in "fundamental frequency current phase difference" is 0.2, according to the preset rules, this value may be mapped to "medium saliency", corresponding to a fuzzy weight of 0.6; if the complementary information value is 0.5, it may be mapped to "high saliency", corresponding to a weight of 0.9.

[0099] Membership modulation: Use the frequency-domain complementary information fuzzy weight to perform weighted processing on the frequency-domain node-wavelet complementary information embedding coding matrix, that is, multiply each matrix element by the corresponding fuzzy weight to generate the fuzzy frequency-domain node-wavelet complementary information embedding coding matrix. Through this operation, the complementary information with high saliency is enhanced (weight close to 1), and the one with low saliency is suppressed (weight close to 0), thus highlighting the personalized features more critical for node state analysis.

[0100] After the above processing, it enters the feature fusion stage: fuse the frequency-domain wavelet feature matrix of the control nodes with the fuzzy frequency-domain node-wavelet complementary information embedding and encoding matrix of multiple control nodes. The fusion method can be matrix splicing or weighted summation. For example, take the frequency-domain wavelet feature matrix as the global feature basis, and the fuzzy complementary information matrix of each node as the local feature supplement, and splice them column by column to form a new matrix. The finally generated frequency-domain aggregation representation matrix of the control nodes contains two parts of features:

[0101] Global frequency-domain wavelet features: reflecting the common features of all control nodes at multiple scales in the wavelet domain, such as the overall energy distribution trend, scale characteristics of the main frequency components, etc.;

[0102] Personalized fuzzy complementary features: the complementary information modulated by the fuzzy weight, representing the unique frequency-domain feature differences of each node, such as abnormal phase differences at specific frequency points, atypical harmonic components, etc.

[0103] The structure of this frequency-domain aggregation representation matrix is: the number of rows is equal to the number of control nodes, and the number of columns is equal to the sum of the global wavelet feature dimension and the personalized complementary feature dimension. For example, if the global wavelet feature matrix contains 3 dimensions (low frequency, medium frequency, high frequency), and the fuzzy complementary information matrix of each node contains 5 dimensions (complementary information corresponding to 5 feature dimensions), then each row of the fused matrix contains 3 + 5 = 8 dimensions, comprehensively covering the common and individual features of the nodes.

[0104] Through wavelet-domain dynamic compensation aggregation, the system can adaptively enhance the personalized difference features while retaining the overall frequency-domain characteristics of the control nodes, avoiding information loss caused by single feature space analysis. This processing method not only improves the richness of feature representation, but also realizes the dynamic evaluation of feature importance through the fuzzy weight mechanism, laying a foundation for accurately extracting the matching features of the target node from multiple nodes subsequently.

[0105] Embodiment 5: The specific implementation manner of extracting the first frequency-domain correlation feature vector from the frequency-domain correlation feature vectors of multiple control nodes and generating a control signal is as follows:

[0106] Extract the first frequency-domain correlation feature vector. According to the node identifier preset in the system (such as the control node number or physical location index), accurately screen out the vector corresponding to the first control node from the set of frequency-domain correlation feature vectors of multiple control nodes, denoted as . This vector contains the current working state of the first control node, such as the numerical values of the fundamental frequency current amplitude, harmonic magnetic flux phase difference, wavelet sub-band energy ratio, etc. in the frequency domain space, fully characterizing the current working state of this node.

[0107] Calculate the first frequency-domain matching feature matrix. Take Frequency domain aggregation representation matrix of the control node Perform correlation analysis. Frequency domain aggregation representation matrix The composite matrix generated by Example 4 contains the global wavelet features and personalized complementary features of all control nodes, and its dimension is ( is the total number of control nodes, is the total number of feature dimensions). By calculating and The covariance matrix between , generate the first frequency domain matching feature matrix. The calculation formula of the covariance matrix is: ;

[0108] in: is the frequency domain aggregation representation matrix Middle The feature vector (row vector) of the control nodes; for The row vector mean of ; is the first frequency domain correlation eigenvector The mean of Represents vector transpose.

[0109] Covariance matrix Each element of Reflects No. Feature dimensions and Middle The node feature vector The degree of linear correlation between the feature dimensions. A positive value indicates a positive correlation (e.g., the first node is Nodes have consistent trends along this dimension. Negative values indicate negative correlation (opposite trends), and the absolute value reflects the strength of the correlation. This matrix quantifies the degree of matching between the first control node and other nodes in the frequency domain feature space, forming a structured matching feature representation.

[0110] Generate a control signal. The first frequency-domain matching feature matrix is input into a control signal generator based on fuzzy inference. The generator contains a fuzzy rule base built based on the control logic of the electromagnetic reversing valve. For example, if the covariance value of a certain dimension is "significantly positive," the duty cycle of the pulse width modulation (PWM) signal of the first control node drive circuit is adjusted to enhance the characteristics of that dimension. If the covariance value is "significantly negative," the duty cycle is adjusted in the opposite direction to suppress the difference.

[0111] The generator processes the matrix data in the following steps:

[0112] Fuzzification: Map the values in the covariance matrix to a preset fuzzy language variable interval (such as "extremely high", "high", "medium", "low", "extremely low"), and each interval corresponds to a specific membership function (such as a trapezoidal function).

[0113] Fuzzy inference: According to the fuzzy rule base, logically combine the fuzzy inputs of each dimension to derive a fuzzy output for the PWM correction value (such as "increase", "slightly increase", "remain", "slightly decrease", "decrease").

[0114] Defuzzification: Convert the fuzzy output into an exact value through methods such as the centroid method, that is, the pulse width modulation correction value of the first control node . This correction value is used for parameter calibration of the drive circuit. For example, adjust the high-level duration of the PWM signal, thereby changing the drive energy of the control node, and realizing real-time optimization of the response characteristics of the electromagnetic directional valve.

[0115] The above process quantifies the frequency-domain feature matching degree between nodes through covariance analysis, and combines fuzzy inference to achieve a non-linear mapping from feature differences to control signals, ensuring precise calibration of the drive circuit of the first control node, and forming a complete link from state data acquisition to closed-loop control.

[0116] The above process realizes the closed-loop analysis from the original state data to the control signal through multi-layer processing of data fusion, feature extraction, fuzzy modulation, matching calculation and fuzzy inference, ensuring the scientificity and accuracy of the parameter calibration of the drive circuit of the first control node, and forming a complete multi-dimensional correlation analysis and control signal generation link.

[0117] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0118] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A remote control data processing method for an electromagnetic reversing valve, characterized in that, Including: Receiving a commutation control instruction sent by a remote terminal; Parsing the commutation control instruction to obtain adjustment parameters of an electromagnetic commutation valve, where the adjustment parameters include the number of action times, switching direction, and pressure threshold; Determining dynamic response parameters of multiple control nodes of the electromagnetic commutation valve based on the adjustment parameters; Performing real-time adjustment actions based on the dynamic response parameters of the multiple control nodes, and synchronously collecting status data of the multiple control nodes, where the status data includes current data and magnetic flux data; Uploading the status data of the multiple control nodes to an edge computing unit; In the edge computing unit, performing multi-dimensional correlation analysis on the status data of the multiple control nodes to generate a control signal, where the control signal is a pulse width modulation correction value of a first control node; including appending magnetic flux data of the same node and the same frequency point after each current data to form a status data frequency domain sequence containing dual physical quantity information; Performing frequency domain implicit correlation analysis on the status data frequency domain sequence of the multiple control nodes to generate frequency domain correlation feature vectors of the multiple control nodes; Performing wavelet domain dynamic compensation aggregation on the frequency domain correlation feature vectors of the multiple control nodes to generate a frequency domain aggregation characterization matrix of the control nodes; Extracting a first frequency domain correlation feature vector from the frequency domain correlation feature vectors of the multiple control nodes, and calculating the matching degree of the first frequency domain correlation feature vector with respect to the frequency domain aggregation characterization matrix of the control nodes to generate a first frequency domain matching feature matrix; Generating the control signal based on the first frequency domain matching feature matrix; Transmitting the control signal to a fuzzy controller to calibrate parameters of a drive circuit corresponding to the first control node.

2. The remote control data processing method of the electromagnetic directional valve according to claim 1, wherein Performing data fusion on the status data of the multiple control nodes to generate a status data frequency domain sequence of the multiple control nodes, including: Performing data reduction on the status data of the multiple control nodes according to the frequency domain dimension and adjustment parameter sample dimension to generate a current data frequency domain sequence of the multiple control nodes and a magnetic flux data frequency domain sequence of the multiple control nodes; Adding corresponding magnetic flux data after each current data in the current data frequency domain sequence of the multiple control nodes to generate the status data frequency domain sequence of the multiple control nodes.

3. The electromagnetic directional valve remote control data processing method according to claim 2, wherein Performing frequency domain implicit correlation analysis on the status data frequency domain sequence of the multiple control nodes to generate frequency domain correlation feature vectors of the multiple control nodes, including: inputting the status data frequency domain sequence of the multiple control nodes into a frequency domain encoder based on a wavelet transform-Fourier hybrid model to generate the frequency domain correlation feature vectors of the multiple control nodes.

4. The electromagnetic directional valve remote control data processing method according to claim 3, wherein Performing wavelet domain dynamic compensation aggregation on the frequency domain correlation feature vectors of the multiple control nodes to generate a frequency domain aggregation characterization matrix of the control nodes, including: Inputting the frequency domain correlation feature vectors of the multiple control nodes into a wavelet domain feature extraction network to generate a control node frequency domain wavelet feature matrix; Based on the frequency-domain wavelet feature matrix of the control nodes, perform fuzzy membership modulation on the frequency-domain correlation feature vectors of each control node among the frequency-domain correlation feature vectors of the multiple control nodes to generate a fuzzy frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes; Fuse the frequency-domain wavelet feature matrix of the control nodes and the fuzzy frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes to generate the frequency-domain aggregation representation matrix of the control nodes.

5. The electromagnetic directional valve remote control data processing method according to claim 4, characterized in that Based on the frequency-domain wavelet feature matrix of the control nodes, performing fuzzy membership modulation on the frequency-domain correlation feature vectors of each control node among the frequency-domain correlation feature vectors of the multiple control nodes to generate a fuzzy frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes includes: Calculate the complementary information of the frequency-domain correlation feature vector of each control node among the frequency-domain correlation feature vectors of the multiple control nodes relative to the frequency-domain wavelet feature matrix of the control nodes to generate a frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes; Input the frequency-domain node-wavelet complementary information embedding coding matrix of each control node in the frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes into a complementary information saliency identification module based on fuzzy rules to generate the frequency-domain complementary information fuzzy weights of the multiple control nodes; Based on the frequency-domain complementary information fuzzy weights of the multiple control nodes, perform membership modulation on the frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes to generate the fuzzy frequency-domain node-wavelet complementary information embedding coding matrix of the multiple control nodes.

6. The electromagnetic directional valve remote control data processing method according to claim 5, wherein Extract a first frequency-domain correlation feature vector from the frequency-domain correlation feature vectors of the multiple control nodes, and calculate the matching degree of the first frequency-domain correlation feature vector relative to the frequency-domain aggregation representation matrix of the control nodes to generate a first frequency-domain matching feature matrix, including: Extract the frequency-domain correlation feature vector corresponding to the first control node from the frequency-domain correlation feature vectors of the multiple control nodes to generate the first frequency-domain correlation feature vector; Calculate the covariance matrix between the first frequency-domain correlation feature vector and the frequency-domain aggregation representation matrix of the control nodes to generate the first frequency-domain matching feature matrix.

7. The electromagnetic directional valve remote control data processing method according to claim 6, wherein Based on the first frequency-domain matching feature matrix, generate the control signal, including: passing the first frequency-domain matching feature matrix through a control signal generator based on fuzzy inference to generate the control signal.

8. A remote control system for an electromagnetic directional valve, characterized in that, Including: A control instruction receiving module, configured to receive a commutation control instruction sent by a remote terminal; A control instruction parsing module, configured to parse the commutation control instruction to obtain the adjustment parameters of the electromagnetic commutation valve, where the adjustment parameters include the number of actions, the switching direction, and the pressure threshold; A dynamic response parameter determination module, configured to determine the dynamic response parameters of multiple control nodes of the electromagnetic commutation valve based on the adjustment parameters; A state data acquisition module, configured to perform real-time adjustment actions based on the dynamic response parameters of the multiple control nodes and synchronously acquire the state data of the multiple control nodes, where the state data includes current data and magnetic flux data; A status data uploading module, configured to upload the status data of the multiple control nodes to an edge computing unit; A status data analysis module, configured to perform multi-dimensional correlation analysis on the status data of the multiple control nodes in the edge computing unit to generate a control signal, where the control signal is a pulse width modulation correction value of a first control node; A control signal execution module, configured to transmit the control signal to a fuzzy controller to calibrate parameters of a drive circuit corresponding to the first control node; The status data analysis module includes: A status data fusion unit, configured to append magnetic flux data of the same node and the same frequency point after each current data to form a status data frequency domain sequence containing dual physical quantity information; A frequency domain correlation analysis unit, configured to perform frequency domain implicit correlation analysis on the status data frequency domain sequences of the multiple control nodes to generate frequency domain correlation feature vectors of the multiple control nodes; A wavelet domain compensation aggregation unit, configured to perform wavelet domain dynamic compensation aggregation on the frequency domain correlation feature vectors of the multiple control nodes to generate a frequency domain aggregation characterization matrix of the control nodes; A matching feature calculation unit, configured to extract a first frequency domain correlation feature vector from the frequency domain correlation feature vectors of the multiple control nodes, and calculate a matching degree of the first frequency domain correlation feature vector with respect to the frequency domain aggregation characterization matrix of the control nodes to generate a first frequency domain matching feature matrix; A control signal generation unit, configured to generate the control signal based on the first frequency domain matching feature matrix.

Images