Multi-site high-fidelity sampling intelligent optimization control system and method

Abstract

The application discloses a multi-site high-fidelity sampling intelligent optimization control system and method, and relates to the technical field of sampling control systems.The system comprises a working condition input and constraint definition module, the output end of the working condition input and constraint definition module is connected with the input end of an RSM agent model module and a multi-objective optimization module, the output end of the RSM agent model module is connected with the input end of the multi-objective optimization module, the data end of the multi-objective optimization module is bidirectionally connected with the data end of a three-domain coordination module, the output end of the three-domain coordination module is connected with the input end of a hierarchical control module, the data end of the hierarchical control module is bidirectionally connected with the data end of a data acquisition module, and the output end of the data acquisition module is connected with the input end of a comprehensive evaluation module.The application aims to solve the technical problems of multi-objective performance imbalance, inter-site coordination conflict, insufficient sampling fidelity, low optimization efficiency and disconnection of parameter design in the existing multi-site high-fidelity sampling technology, and provides a multi-site high-fidelity sampling intelligent optimization control system and method.

Description

Technical Field

[0001] This invention relates to the fields of environmental protection, monitoring, and governance, and particularly to high-fidelity sampling technology in environmental monitoring, specifically to a multi-site high-fidelity sampling intelligent optimization control system and method. Background Technology

[0002] High-fidelity sampling technology has become a core sampling technology in fields such as geological exploration, environmental monitoring, and contaminated site investigation due to its advantages such as high sampling efficiency, minimal disturbance to the strata, and maximum preservation of the original state of soil samples. It has also been widely used in engineering practice. With the increasing demand for engineering and batch sampling, and the increasingly stringent requirements for multi-site collaborative efficiency, high-fidelity sampling accuracy, energy consumption control, and cost control in complex geological environments, the limitations of existing sampling technologies are becoming increasingly apparent. Specifically, these limitations are manifested in the following ways: most existing technologies focus on improving single performance aspects and fail to achieve multi-objective collaborative optimization; there is a lack of inter-site collaborative mechanisms, resulting in weak adaptability to complex scenarios. They generally adopt a "simple parallel control" mode without global conflict arbitration or dynamic task allocation strategies, and the control methods are mostly single feedback control without a hierarchical strategy, making it difficult to cope with complex situations such as differences in geological formations, operational interference, and execution deviations across multiple sites; optimization efficiency is low and parameter design is disconnected, relying on a large number of repetitive simulations or field tests to obtain data, resulting in long calculation cycles. Furthermore, the driving structure parameters and control parameters adopt a "design first, debug later" separation mode, without considering the dynamic coupling relationship between the two, which can easily lead to system performance imbalance and severely limit the adaptability in large-scale, multi-site operation scenarios.

[0003] For example, the patent application CN116084839A discloses a spatial attitude adaptive intelligent targeted drilling system for complex geological environments. While it improves drilling trajectory accuracy and reduces formation disturbance through attitude monitoring and correction mechanisms, it does not address the optimization design of key indicators such as energy consumption, equipment size, and manufacturing cost. Furthermore, it does not mention a multi-site collaborative mechanism, and the control method relies solely on single feedback from attitude correction. Without a hierarchical control strategy, it is difficult to cope with execution deviations and interferences in multi-site operations, resulting in limited overall cost-effectiveness and an inability to adapt to multi-site batch sampling scenarios. Similarly, the patent application CN110080810A discloses a hydraulically driven mine roadway anchor cable installation device. While it optimizes equipment usage costs through a separable mechanism, it does not comprehensively consider core performance aspects such as high-fidelity sampling accuracy, multi-site collaborative efficiency, and operational stability. It also lacks hierarchical control and a global collaborative strategy, failing to meet the needs of multi-objective balance and making it difficult to offset site interference through precise control, thus lacking the ability to adapt to complex scenarios. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problems of existing multi-site high-fidelity sampling technologies, such as multi-objective performance imbalance, inter-site coordination conflict, insufficient sampling fidelity, low optimization efficiency, and disconnection of parameter design. Therefore, this invention proposes a multi-site high-fidelity sampling intelligent optimization control system and method.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multi-site high-fidelity sampling intelligent optimization control system includes a working condition input and constraint definition module for receiving on-site working conditions and engineering constraint parameters for each site; the output of the working condition input and constraint definition module is connected to the input of an RSM surrogate model module and a multi-objective optimization module; the output of the RSM surrogate model module is connected to the input of the multi-objective optimization module; the data terminal of the multi-objective optimization module is bidirectionally connected to the data terminal of a three-domain collaboration module; the output of the three-domain collaboration module is connected to the input of a hierarchical control module; the data terminal of the hierarchical control module is bidirectionally connected to the data terminal of a data acquisition module; and the output of the data acquisition module is connected to the input of a comprehensive evaluation module.

[0006] The comprehensive evaluation module is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

[0007] The output of the comprehensive evaluation module is connected to the input of the rating and compliance judgment unit. The rating and compliance judgment unit compares the evaluation results with the preset compliance threshold based on the comprehensive evaluation index and three-dimensional rating results output by the comprehensive evaluation module. If the standard is met, the working condition label and the optimal parameter set are stored in the multi-parameter mapping database. If the standard is not met, the adjustment instruction is fed back to the RSM proxy model module to trigger iterative optimization.

[0008] The multi-parameter mapping database stores compliance data hierarchically according to the following order: "Number of Sites - Stratigraphic Parameters - Engineering Constraints - 3D Rating - Optimal Parameter Set," providing parameter matching and one-click tuning support for engineering application terminals. The data output end of the multi-parameter mapping database is connected to the data input end of the engineering application terminal, which in turn is connected to the input end of the working condition input and constraint definition module. The engineering application terminal is used to input site working conditions and constraint parameters, receive one-click tuning parameters, and issue configuration commands to achieve rapid system deployment.

[0009] The hierarchical control module includes a global coordinator, local sub-controllers, drive execution units, and a sensor group. The output of the global coordinator is connected to the input of the local sub-controller, the output of the local sub-controller is connected to the input of the drive execution unit, the output of the drive execution unit is connected to the input of the sensor group, and the output of the sensor group is connected to the global coordinator, the local sub-controller, and the data acquisition module, respectively. The global coordinator is responsible for multi-site task allocation and conflict arbitration. The local sub-controller adopts a hierarchical control strategy of "upper-level trajectory planning + lower-level rapid compensation" to achieve multi-site single-machine collaborative operation control, ensuring high-fidelity sampling requirements at each site and improving system fault tolerance. The drive execution units at each site are used to execute core actions such as sampling advancement, stopping, and fine-tuning according to control commands. The sensors at each site are used to collect real-time operating data of the drive execution units and inter-unit collaborative data, providing data support for the feedback control and comprehensive evaluation module. The data acquisition module is used to receive data from the sensor group, determine in real time whether the sampling accuracy meets the standard, and perform data standardization and outlier removal to generate an operating database. The comprehensive evaluation module is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

[0010] The working condition input and constraint definition module receives on-site working conditions and engineering constraint parameters from various sites, processes them through standardization, and outputs them to the RSM surrogate model module and the multi-objective optimization module, providing a unified data foundation for subsequent optimization. The RSM surrogate model module trains a high-precision substitution model of "design variables - multi-objective performance" based on simulation samples, replacing traditional time-consuming simulation or experimental calculations and providing rapid computational support for the multi-objective optimization module. The multi-objective optimization module uses driving structural parameters, control parameters, and distributed parameters as joint design variables, and generates multiple sets of Pareto optimal solutions through the adaptive NSGA-II algorithm to achieve performance trade-offs under multiple constraints such as energy consumption, volume, and cost. The three-domain collaboration module divides the structural design domain, control stability domain, and distributed collaboration domain, obtains a conflict-free feasible domain through the intersection of the three domains, and verifies the engineering feasibility of the Pareto optimal solution, realizing the collaborative adaptation of structural parameters, control parameters, and distributed parameters.

[0011] When the system is working / being used, the following steps are performed: Step 1: Collect on-site working condition parameters and engineering constraint parameters of the target sampling scene, clarify the core objectives of multi-objective optimization, and construct a complete initial input system to provide basic support for subsequent model construction and optimization; Step 2: Based on the initial input data system formed in Step 1, divide the joint design variables into three categories: driving structure, control, and distributed collaboration. Initialize the variable value range, clarify the parameter boundaries, and form the search object and range for multi-objective optimization. Step 3: Based on the variable value range determined in Step 2, generate parameter samples, collect simulation sample data through simulation or field experiments, construct and verify the RSM surrogate model, establish the mapping relationship between parameters and optimization objectives, and improve the efficiency of subsequent multi-objective optimization; Step 4: Based on the RSM proxy model constructed in Step 3, configure the adaptive NSGA-II algorithm parameters, define the fitness function, call the RSM proxy model to calculate the objective function value, solve and generate multiple sets of Pareto optimal solutions, and provide multiple parameter combination schemes with performance trade-offs. Step 5: Based on the Pareto optimal solution set generated in Step 4, divide the boundaries of the three domains: structural design domain, control stability domain, and distributed cooperative domain. Perform intersection of the three domains and engineering feasibility verification on the Pareto optimal solution to determine the cooperative optimal parameter set with engineering feasibility. Step 6: Based on the optimal set of parameters determined in Step 5, deploy the global coordinator and local sub-controllers. Through global coordination and hierarchical control, each site drive execution unit executes sampling actions according to instructions, synchronously collects operating data, and forms a two-way feedback closed loop. Step 7: Based on the operational data collected in Step 6, an operational database is formed, a comprehensive evaluation index set is constructed, and after preprocessing, the index weights are calculated using the entropy weight method to generate a three-dimensional rating and a comprehensive evaluation index to determine whether the system performance meets the standards. Step 8: Based on the performance assessment results in Step 7, if the rating meets the standard, bind the working condition label with the optimal collaborative parameter set and update the multi-parameter mapping database; if it does not meet the standard, update the RSM proxy model until the system performance meets the standard.

[0012] Step 1 includes the following sub-steps: Step 1-1: Collect key information from the target sampling scenario and identify two types of core input parameters: on-site working condition parameters and engineering constraint parameters. All parameters are confirmed through on-site measurements or engineering design documents to ensure data authenticity. Step 1-2: Define and quantify the three core optimization objectives, namely, minimizing the total energy consumption of sampling. E Maximize continuous runtime T Ensure the accuracy of formation parameter acquisition. e , satisfying 0≤ e ≤ e lim ,in e lim To achieve a balanced approach between energy efficiency, reliability, and sampling accuracy, the maximum allowable error is set. Steps 1-3: Integrate the input parameters from Step 1-1 with the optimization objectives from Step 1-2, clarify the influence relationship between each parameter and the objective, define the limiting boundaries of the parameter values ​​under the constraints, and form a complete initial input system to provide a basis for subsequent design variable definition and model training.

[0013] Step 2 includes the following sub-steps: Step 2-1: Divide into two categories of core design variables and clarify the physical meaning of each variable: Traditional variables are the core parameters of the driving structure, which directly affect the mechanical performance of the equipment; Distributed variables are divided into control parameters and distributed parameters, which respectively ensure the accuracy of sampling actions and synchronous operation in multiple sites. Step 2-2: Based on the constraints defined in Step 1-3, and in combination with equipment performance parameters, engineering design specifications and actual site conditions, set reasonable value ranges for each variable to ensure sufficient optimization space without deviating from the actual application scenario, thus providing an effective search range for subsequent sample generation and multi-objective optimization.

[0014] Step 3 includes the following sub-steps: Step 3-1: Generate N sets of sample combinations within the variable value range of Step 2-2. Each set of samples corresponds to a unique "structure-control-distributed" parameter combination to ensure that the value of each variable is evenly distributed within its range, without obvious clustering or omission areas, thereby improving the representativeness of the samples. Step 3-2: Simulate the sampling process under each parameter combination through simulation or field testing, and record the actual values ​​of the three optimization objectives in real time, including the actual energy consumption value collected by the energy consumption monitoring module. E act Continuous operating time of equipment that indirectly characterizes the failure rate T run The acquisition error was calculated by comparing the sampled data with the actual formation parameters. e act , forming a training sample set {( x 1, y 1),( x 2, y 2),…,( x N , y N )},in x For parameter combinations, y =( E act , T run , e act () represents the actual target value; Step 3-3: Construct an RSM proxy model with uncertainty handling.

[0015] This invention also includes a multi-site high-fidelity sampling intelligent optimization control method to solve the technical problems of existing multi-site high-fidelity sampling technologies, such as multi-objective performance imbalance, inter-site coordination conflict, insufficient sampling fidelity, low optimization efficiency, and parameter design disconnect. The method includes the following steps: Step 1: Collect on-site working condition parameters and engineering constraint parameters of the target sampling scene, clarify the core objectives of multi-objective optimization, and construct a complete initial input system to provide basic support for subsequent model construction and optimization; Step 2: Based on the initial input data system formed in Step 1, divide the joint design variables into three categories: driving structure, control, and distributed collaboration. Initialize the variable value range, clarify the parameter boundaries, and form the search object and range for multi-objective optimization. Step 3: Based on the variable value range determined in Step 2, generate N sets of "structure-control-distributed" parameter combination samples. Simulate the sampling process under each parameter combination through simulation or field experiment, collect the actual values ​​of the three optimization objectives, construct a training sample set, construct and verify the RSM surrogate model with uncertainty handling based on the sample set, establish the mapping relationship between parameters and optimization objectives, and improve the efficiency of subsequent multi-objective optimization. Step 4: Based on the RSM proxy model constructed in Step 3, configure the adaptive NSGA-II algorithm parameters, define the fitness function, call the RSM proxy model to calculate the objective function value, solve and generate multiple sets of Pareto optimal solutions, and provide multiple parameter combination schemes with performance trade-offs. Step 5: Based on the Pareto optimal solution set generated in Step 4, divide the boundaries of the three domains: structural design domain, control stability domain, and distributed cooperative domain. Perform intersection of the three domains and engineering feasibility verification on the Pareto optimal solution to determine the cooperative optimal parameter set with engineering feasibility. Step 6: Based on the optimal set of parameters determined in Step 5, formulate multi-site task allocation strategies and conflict arbitration rules, and deploy a hierarchical control strategy of "upper-level trajectory planning + lower-level rapid compensation"; each site executes sampling actions according to control instructions, and simultaneously collects two types of data in real time through sensors, feeds back the site's operation data to the local control terminal for deviation correction, and feeds back the collaborative data of each site to the global coordination terminal for strategy adjustment, forming a closed-loop process of "global coordination - local control - execution feedback"; Step 7: Based on the operational data collected in Step 6, an operational database is formed, a comprehensive evaluation index set is constructed, and after preprocessing, the index weights are calculated using the entropy weight method to generate a three-dimensional rating and a comprehensive evaluation index to determine whether the standard is met. Step 8: Based on the performance assessment results in Step 7, if the rating meets the standard, bind the working condition label with the collaborative optimal parameter set and update the multi-parameter mapping database. Subsequent similar working conditions can quickly match parameters through keyword retrieval. If the rating does not meet the standard, supplement the sample data under the current working condition, update the RSM proxy model, and re-execute Steps 4-7 until the rating meets the standard.

[0016] In step 1, as Figure 3 As shown, it includes the following sub-steps: Step 1-1: Collect key information from the target sampling scenario and identify two types of core input parameters: on-site working condition parameters and engineering constraint parameters. All parameters are confirmed through on-site measurements or engineering design documents to ensure data authenticity. Step 1-2: Define and quantify the three core optimization objectives, namely, minimizing the total energy consumption of sampling. E ( E (Total energy consumed during sampling) and maximize continuous runtime T ( T (For continuous operation of the equipment), and to ensure the accuracy of formation parameter acquisition. e ( e (where is the deviation between the sampled data and the actual formation parameters), satisfying 0 ≤ e ≤ e lim ( e lim (to the maximum allowable error), achieving a synergistic balance between energy efficiency, reliability, and sampling accuracy; Steps 1-3: Integrate the input parameters from Step 1-1 with the optimization objectives from Step 1-2, clarify the influence relationship between each parameter and the objective, define the limiting boundaries of the parameter values ​​under the constraints, and form a complete initial input system to provide a basis for subsequent design variable definition and model training.

[0017] In step 2, as Figure 3 As shown, it includes the following sub-steps: Step 2-1: Divide into two categories of core design variables and clarify the physical meaning of each variable: Traditional variables are the core parameters of the driving structure, which directly affect the mechanical performance of the equipment; Distributed variables are divided into control parameters and distributed parameters, which respectively ensure the accuracy of sampling actions and synchronous operation in multiple sites. Step 2-2: Based on the constraints defined in Step 1-3, and in combination with equipment performance parameters, engineering design specifications and actual site conditions, set reasonable value ranges for each variable to ensure sufficient optimization space without deviating from the actual application scenario, thus providing an effective search range for subsequent sample generation and multi-objective optimization.

[0018] In step 3, as Figure 3 As shown, it includes the following sub-steps: Step 3-1: Generate N sets of sample combinations within the variable value range of Step 2-2. Each set of samples corresponds to a unique "structure-control-distributed" parameter combination to ensure that the value of each variable is evenly distributed within its range, without obvious clustering or omission areas, thereby improving the representativeness of the samples. Step 3-2: Simulate the sampling process under each parameter combination through simulation or field experiments, and record the actual values ​​of the three optimization objectives in real time, including the actual energy consumption values ​​collected through energy consumption monitoring methods. E act Continuous operating time of equipment that indirectly characterizes the failure rate T run The acquisition error was calculated by comparing the sampled data with the actual formation parameters. e act , forming a training sample set {( x 1, y 1),( x 2, y 2),…,( x N , y N )}( x For parameter combinations, y =( E act , T run , e act (The actual value is the target value). Step 3-3: Construct an RSM proxy model with uncertainty handling.

[0019] Compared with the prior art, the present invention has the following technical effects: 1) This invention solves the performance imbalance problem caused by single-objective optimization through multi-objective collaborative optimization, improves the overall engineering practicality of the system, and ensures high-fidelity sampling accuracy while keeping the error within the allowable range; 2) In this invention, the RSM proxy model replaces the traditional time-consuming simulation, which can shorten the optimization calculation cycle by more than 80% and significantly reduce the calculation cost; 3) The “structure-control-distributed three-domain collaborative design” in this invention can eliminate the coupling conflict among the three, and combined with hierarchical control, improve the sampling fidelity and operational stability of multiple sites; 4) The distributed control architecture in this invention ensures that the failure of a single unit does not affect the operation of the overall system, thus ensuring the continuous execution of sampling tasks; 5) The comprehensive evaluation system based on the entropy weight method of this invention realizes a quantitative and unified evaluation of energy saving, reliability, and distributed collaboration levels, avoiding subjective bias; 6) The multi-parameter mapping database in this invention supports rapid matching of operating condition labels and optimal parameters without repeated optimization, providing data support for rapid equipment tuning and batch application. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the high-fidelity sampling system architecture of the present invention; Figure 2 This is a flowchart illustrating the overall technical route of the present invention; Figure 3 This is a schematic diagram of the multi-objective optimization process described in an embodiment of the present invention; Figure 4 This is a schematic diagram of the core process of the adaptive NSGA-II algorithm described in the embodiments of the present invention; Figure 5 This is a schematic diagram of the three-domain collaborative design process according to an embodiment of the present invention; Figure 6 This is a schematic diagram of the three-domain intersection method according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the multi-site hierarchical control process according to an embodiment of the present invention; Figure 8 This is a schematic diagram of the comprehensive evaluation and parameter mapping database process described in the embodiments of the present invention. Detailed Implementation

[0021] like Figure 1 As shown, a multi-site high-fidelity sampling intelligent optimization control system includes a working condition input and constraint definition module for receiving on-site working conditions and engineering constraint parameters for each site; the output of the working condition input and constraint definition module is connected to the input of the RSM surrogate model module and the multi-objective optimization module; the output of the RSM surrogate model module is connected to the input of the multi-objective optimization module; the data terminal of the multi-objective optimization module is bidirectionally connected to the data terminal of the three-domain collaboration module; the output of the three-domain collaboration module is connected to the input of the hierarchical control module; the data terminal of the hierarchical control module is bidirectionally connected to the data terminal of the data acquisition module; and the output of the data acquisition module is connected to the input of the comprehensive evaluation module.

[0022] The comprehensive evaluation module can use an STM32H743VI microcontroller as the core processing unit, and can be paired with an AD7606 16-bit analog-to-digital converter chip to collect sensor data. It can also achieve data interaction through the ESP32-WROOM-32E wireless communication component. It is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

[0023] The output of the comprehensive evaluation module is connected to the input of the rating and compliance judgment unit. The rating and compliance judgment unit compares the evaluation results with the preset compliance threshold based on the comprehensive evaluation index and three-dimensional rating results output by the comprehensive evaluation module. If the standard is met, the working condition label and the optimal parameter set are stored in the multi-parameter mapping database. If the standard is not met, the adjustment instruction is fed back to the RSM proxy model module to trigger iterative optimization.

[0024] The multi-parameter mapping database can be an Alibaba Cloud RDS MySQL cloud database, used to store compliance data hierarchically according to "multiple site quantities - geological parameters - engineering constraints - 3D rating - optimal parameter set", providing parameter matching and one-click tuning support for engineering application terminals; the data output end of the multi-parameter mapping database is connected to the data input end of the engineering application terminal, and the data output end of the engineering application terminal is connected to the input end of the working condition input and constraint definition module; the engineering application terminal uses Advantech industrial tablet computer product series, used to input site working conditions and constraint parameters, receive one-click tuning parameters and issue configuration commands to realize rapid system deployment.

[0025] The hierarchical control module includes a global coordinator, local sub-controllers, drive execution units, and a sensor group. The output of the global coordinator is connected to the input of the local sub-controller, the output of the local sub-controller is connected to the input of the drive execution unit, the output of the drive execution unit is connected to the input of the sensor group, and the output of the sensor group is connected to the global coordinator, the local sub-controller, and the data acquisition module, respectively. The global coordinator is responsible for multi-site task allocation and conflict arbitration. The local sub-controller adopts a hierarchical control strategy of "upper-level trajectory planning + lower-level rapid compensation" to achieve multi-site single-machine collaborative operation control, ensuring high-fidelity sampling requirements at each site and improving system fault tolerance. Each site's drive execution unit is used to execute core actions such as sampling advancement, stopping, and fine-tuning according to control commands. Each site's sensor is used to collect real-time operating data of the drive execution unit and inter-unit collaborative data, providing data support for the feedback control and comprehensive evaluation module. The data acquisition module can use an ADS1256. A 24-bit analog-to-digital converter chip is used to receive data from the sensor group, determine in real time whether the sampling accuracy meets the standard, and perform data standardization and outlier removal to generate a running database. The comprehensive evaluation module is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

[0026] The working condition input and constraint definition module receives on-site working conditions and engineering constraint parameters from various sites, processes them through standardization, and outputs them to the RSM surrogate model module and the multi-objective optimization module, providing a unified data foundation for subsequent optimization. The RSM surrogate model module can use an Intel Core i7 12700H processor with 16GB DDR5 memory to train a high-precision alternative model of "design variables - multi-objective performance" based on simulation samples, replacing traditional time-consuming simulation or experimental calculations and providing fast computational support for the multi-objective optimization module. The multi-objective optimization module uses driving structural parameters, control parameters, and distributed parameters as joint design variables, and generates multiple Pareto optimal solutions through the adaptive NSGA-II algorithm to achieve performance trade-offs under multiple constraints such as energy consumption, volume, and cost. The three-domain collaboration module is used to divide the structural design domain, control stability domain, and distributed collaboration domain, obtains the conflict-free feasible domain through the intersection of the three domains, and verifies the engineering feasibility of the Pareto optimal solution, realizing the collaborative adaptation of structural parameters, control parameters, and distributed parameters.

[0027] like Figure 2 As shown; the system operates / is used in the following steps: Step 1: Collect on-site working condition parameters and engineering constraint parameters of the target sampling scene, clarify the core objectives of multi-objective optimization, and construct a complete initial input system to provide basic support for subsequent model construction and optimization; Step 2: Based on the initial input data system formed in Step 1, divide the joint design variables into three categories: driving structure, control, and distributed collaboration. Initialize the variable value range, clarify the parameter boundaries, and form the search object and range for multi-objective optimization. Step 3: Based on the variable value range determined in Step 2, generate parameter samples, collect simulation sample data through simulation or field experiments, construct and verify the RSM surrogate model, establish the mapping relationship between parameters and optimization objectives, and improve the efficiency of subsequent multi-objective optimization; Step 4: Based on the RSM proxy model constructed in Step 3, configure the adaptive NSGA-II algorithm parameters, define the fitness function, call the RSM proxy model to calculate the objective function value, solve and generate multiple sets of Pareto optimal solutions, and provide multiple parameter combination schemes with performance trade-offs. Step 5: Based on the Pareto optimal solution set generated in Step 4, divide the boundaries of the three domains: structural design domain, control stability domain, and distributed cooperative domain. Perform intersection of the three domains and engineering feasibility verification on the Pareto optimal solution to determine the cooperative optimal parameter set with engineering feasibility. Step 6: Based on the optimal set of parameters determined in Step 5, deploy the global coordinator and local sub-controllers. Through global coordination and hierarchical control, each site drive execution unit executes sampling actions according to instructions, synchronously collects operating data, and forms a two-way feedback closed loop. Step 7: Based on the operational data collected in Step 6, an operational database is formed, a comprehensive evaluation index set is constructed, and after preprocessing, the index weights are calculated using the entropy weight method to generate a three-dimensional rating and a comprehensive evaluation index to determine whether the system performance meets the standards. Step 8: Based on the performance assessment results in Step 7, if the rating meets the standard, bind the working condition label with the optimal collaborative parameter set and update the multi-parameter mapping database; if it does not meet the standard, update the RSM proxy model until the system performance meets the standard.

[0028] In step 1, as Figure 3 As shown, it includes the following sub-steps: Step 1-1: Collect key information from the target sampling scenario and identify two types of core input parameters: on-site working condition parameters and engineering constraint parameters. All parameters are confirmed through on-site measurements or engineering design documents to ensure data authenticity. Step 1-2: Define and quantify the three core optimization objectives, namely, minimizing the total energy consumption of sampling. E ( E (Total energy consumed by the driving execution unit during sampling) and maximize continuous runtime T ( T (For continuous operation of the equipment), and to ensure the accuracy of formation parameter acquisition. e ( e (where is the deviation between the sampled data and the actual formation parameters), satisfying 0 ≤ e ≤ e lim ( e lim (to the maximum allowable error), achieving a synergistic balance between energy efficiency, reliability, and sampling accuracy; Steps 1-3: Integrate the input parameters from Step 1-1 with the optimization objectives from Step 1-2, clarify the influence relationship between each parameter and the objective, define the limiting boundaries of the parameter values ​​under the constraints, and form a complete initial input system to provide a basis for subsequent design variable definition and model training.

[0029] In step 2, as Figure 3 As shown, it includes the following sub-steps: Step 2-1: Divide into two categories of core design variables and clarify the physical meaning of each variable: Traditional variables are the core parameters of the driving structure, which directly affect the mechanical performance of the equipment; Distributed variables are divided into control parameters and distributed parameters, which respectively ensure the accuracy of sampling actions and synchronous operation in multiple sites. Step 2-2: Based on the constraints defined in Step 1-3, and in combination with equipment performance parameters, engineering design specifications and actual site conditions, set reasonable value ranges for each variable to ensure sufficient optimization space without deviating from the actual application scenario, thus providing an effective search range for subsequent sample generation and multi-objective optimization.

[0030] Step 3 includes the following sub-steps: Step 3-1: Generate N sets of sample combinations within the variable value range of Step 2-2. Each set of samples corresponds to a unique "structure-control-distributed" parameter combination to ensure that the value of each variable is evenly distributed within its range, without obvious clustering or omission areas, thereby improving the representativeness of the samples. Step 3-2: Simulate the sampling process under each parameter combination through simulation or field testing, and record the actual values ​​of the three optimization objectives in real time, including the actual energy consumption value collected by the energy consumption monitoring module. E act Continuous operating time of equipment that indirectly characterizes the failure rate T run The acquisition error was calculated by comparing the sampled data with the actual formation parameters. e act , forming a training sample set {( x 1, y 1),( x 2, y 2),…,( x N , y N )}( x For parameter combinations, y =( E act , T run , e act (The actual value is the target value). Step 3-3: Construct an RSM proxy model with uncertainty handling.

[0031] Step 3-3 specifically includes: Step 3-3-1: Using the design variables as independent variables and the three optimization objectives as response variables, the Box-Cox transformation is introduced to address data skewness. The uncertainty of the response variables is quantified using trapezoidal fuzzy logic. The model expression is: ; ; in x i , x j For the first design variable i The, the j 1 variable ( i , j It is a positive integer, 1≤ i ≤ j ≤ k ), kThe total number of variables, αL 0(P) , αU 0(P) For trapezoidal fuzzy numbers P Boundary coefficients of the constant term at the horizontal level, αL i(P) , αUi(P) For trapezoidal fuzzy numbers P Boundary coefficients for the first-order term at the horizontal level. αL ij(P) , αUij(P) For trapezoidal fuzzy numbers P Boundary coefficients of quadratic terms at the horizontal level. y L (P) , y U (P) These are the lower and upper bounds of the response variable, respectively, covering the range of uncertainty of the target value; Step 3-3-2: If the response variable does not satisfy the normality relationship, use the Box-Cox transformation, the formula is: when p ≠0 o'clock, ;when p When =0, ; in y For the original response variable, y(p) For the transformed variables, p For the transformation parameters, t This is a preset constant; Simultaneously, trapezoidal fuzzy quantitative uncertainty is introduced, with its lower bound... α L (P) With the upper realm α U (P) The range of fluctuations in statistical sample data is determined by the following calculation formula: when hour, ;when hour, ;when hour, ;when hour, ; when hour, ;when hour, ;when hour, ;when hour, ; in a , d These are the lower and upper bounds of the trapezoidal fuzzy number, respectively. b The inflection point of the membership function.c This is the inflection point of the membership function. Step 3-3-3: Construct individual conditional expectation curves for each design variable, fixing other variables at their mean or median, and only changing the target variable. The formula is as follows: , i As a parameter vector, it intuitively displays the changing trend of the response target when a single variable changes, improving the interpretability of the model; Steps 3-4: Calculate the model predictions Compared with the actual sample value y act relative error ,like d ≤ d 0 ( d (0 is the error threshold), the model prediction accuracy meets the engineering requirements, and the verification is passed; if d > d 0, add extreme working condition samples and retrain until the error meets the requirements.

[0032] In step 4, as Figure 3 As shown, it includes the following sub-steps: Step 4-1: Based on the variable dimension and the number of optimization targets, set the core parameters of the adaptive NSGA-II algorithm, including population size, number of iterations, adaptive crossover probability, and mutation probability; Step 4-2: Call the RSM proxy model constructed in Step 3-3, using the three optimization objectives as the fitness function and unifying the objective direction, which are: Energy consumption target: , Failure rate target: ( T max (Rated continuous operating time). Sampling accuracy target: ; Step 4-3: The design variables are calculated using the adaptive NSGA-II algorithm. Based on fast non-dominated sorting, crowding calculation, elite retention strategy and adaptive genetic operation, the multi-objective evolution optimization of the design variables is realized, and finally multiple Pareto optimal solutions are generated to form the optimal parameter combination set.

[0033] In step 4-3, as Figure 4 As shown, it includes the following sub-steps: Step 4-3-1: Randomly generate the parameters from Step 4-1, with a scale of... N The initial parent population P 0 and the size after fitness calculation using the RSM surrogate model is N offspring population Q 0 merges to form a mixed populationR t = P t ∪ Q t ( t (current iteration number), population size 2 N This operation expands the search space, prevents the loss of outstanding individuals from the previous generation, and provides a foundation for the preservation of elites; Step 4-3-2: Analysis of mixed populations based on Pareto dominance relationships R t By implementing layering, the computational complexity of traditional sorting can be reduced; (1) For any two individuals in the population x i and x j If the following conditions are met: ; and ; Then it is called x i Dominate x j , recorded as x i x j ,in f k ( The fitness function is defined in step 4-2. (2) For each individual x p ∈ R t Calculate two core attributes: Dominated Count n p ( n p Dominant in the population x p (Number of individuals), dominant solution set S p ( S p For being x p (The set of all individuals under control); (3) Layered iteration, specifically including: ① Initialize the non-dominated layer sequence number rank =1, all n p Individuals with a value of 0 are classified into the first... rank layer F rank ; ② Traversal F rank Each individual x p The solution set of its domination S p Each individual x q ,implement n q = n q 1; if n q =0, then x q Return to the next level F rank+1 ; ③ Order rank = rank +1, repeat the above operation until all individuals are assigned to the corresponding non-dominated layer. F 1, F 2,…, F k ( F 1 is the optimal non-dominated layer). Step 4-3-3: For each non-dominated layer, calculate the crowding distance of individuals within the layer. d p To quantify the sparsity around an individual and avoid clustering of solution sets, the specific calculation is as follows: (1) For each individual x p ∈ F rank ,set up d p =0; (2) For the first k There is one objective function, which normalizes the function values ​​of all individuals within the layer to [0,1]. The formula is as follows: ; in For the first k The objective function is F rank Minimum value of the layer, For the first k The objective function is F rank The maximum value of the layer; (3) For each objective function k ,according to Individuals within the ascending order of the hierarchy are grouped, and the crowding of boundary individuals is set to [value missing]. d p =∞; for sorting positions, iintermediate individuals x p The calculation formula is: ; After traversing all objective functions, d p That is, an individual x p Total congestion distance; Step 4-3-4: Based on the non-dominated layer index and crowding distance, from the mixed population R t Selected from the scale of N The next generation of parent population P t+1 The process is as follows: (1) Individuals are included in the sequence from low to high. P t+1 until it is incorporated into a certain layer F After 1, the population size exceeded N ; (2) To F Individuals in 1 are ranked by crowding distance d p Sort in descending order and select the first few. N -| P t+1 |Individuals included P t+1 To ensure the final population size is N ; Step 4-3-5: [Regarding...] P t+1 Perform selection, crossover, and mutation operations to generate offspring populations. Q t+1 The process is as follows: (1) Using the binary tournament selection method, two individuals are randomly selected, with priority given to individuals with lower non-dominated layer numbers; if the numbers are the same, individuals with larger crowding distances are selected, and this process is repeated. N Next, a selective population is formed. P s ; (2) Simulated binary crossover is used, and the crossover probability is... p c The cross formula adjusts linearly with the number of iterations and is as follows: ; ; in x p , x q ∈ P s For paired individuals,β The cross-distribution index, k To design variable dimensions; (3) Using polynomial mutation, the mutation probability p m The mutation formula is: ; in x k,max , x k,min Design variables x k The boundary values ​​of Δ k For variable asynchronous length, the calculation formula is: when hour, ; when hour, ; or m It is the distribution index of variation. r ∈[0,1] represents a random number; Step 4-3-6: Let the number of iterations be... t = t +1, will P t+1 Substitute the values ​​from step 3-3 into the RSM surrogate model to calculate fitness, and repeat steps 4-3-1 to 4-3-5; stop iterating when the number of iterations reaches the termination condition; finally, multiple Pareto optimal solutions are generated, forming the optimal parameter combination set { X 1, X 2,…, X W}( W (where the number of Pareto optimal solutions is ), and each solution satisfies: ; This means that it is impossible to improve a certain objective without reducing the performance of other objectives, thus failing to achieve optimal multi-objective collaboration.

[0034] In step 5, as Figure 5 As shown, it includes the following sub-steps: Step 5-1: Based on the design specifications of the drive execution device, the stability requirements of control theory, and the distributed coordination mechanism, delineate the boundaries of the three feasible domains; Specifically, the structural design domain is based on strength and spatial constraints, satisfying the strength constraints. ( s The actual stress of the structure, s lim (allowable stress of material) and dimensional constraints ( xVariables for structural design. x min , x max (These represent the minimum and maximum structural dimensions, respectively). The control stability domain is based on Lyapunov stability and phase margin constraints, satisfying the no-overshoot constraint for closed-loop control. ( s p For overshoot, s p,lim To allow overshoot and response time constraints ( t r For response time, t r,lim To allow for response time); the distributed cooperative domain, based on interference and communication delay constraints, satisfies the conflict-free constraint between the driving execution units in each site. ( t com Due to communication delay, t lim (Maximum allowable communication delay) and synchronization response constraints (Δ) t sync For the synchronization response error, Δ t sync,lim To allow for synchronization response error); Step 5-2: For each Pareto optimal solution generated in Step 4-3, perform set operations to find the intersection of the three regions and determine whether the feasible region is non-empty. Figure 6 The schematic diagram of the three-domain intersection method described in this embodiment of the invention is shown below; If there exists at least one solution that simultaneously satisfies the three-domain constraints, proceed to step 5-3; if the feasible domain is empty, return to step 2-2 to adjust the boundary values ​​of the design variables, and re-execute steps 3-4. Step 5-3: Verify the engineering feasibility of the parameter combinations in the intersection of the three domains, including structural verification, control verification, and distributed verification; if all verifications are successful, form a collaborative optimal parameter set after correcting local conflicts. ( X dis For distributed optimal parameters X con To control the optimal parameters, X str (For the optimal parameters of the structure); if the verification fails, the feasible region constraint range is fed back to the multi-objective optimization module, and the Pareto optimal solution is searched again and verified again until the target is met.

[0035] In step 6, as Figure 7 As shown, it includes the following sub-steps: Step 6-1: Deploy a global coordinator on the system's host computer and formulate two core strategies based on the optimal set of collaborative parameters: Multi-site task allocation assigns sampling priorities according to differences in stratigraphic parameters to avoid excessive concentration during high-fidelity sampling by the driving execution units; Conflict arbitration rules define a conflict determination function. ; in, t com,i For driving execution unit i Communication delay, I ( () is an indicator function that takes the value 1 if the condition is met, and 0 otherwise; when C ( t When )>0, sampling is performed in priority order to avoid stress interference and data transmission conflicts caused by simultaneous sampling; Step 6-2: Each local sub-controller adopts a hierarchical control strategy of "MPC trajectory planning + LQR fast compensation"; Step 6-2 includes the following sub-steps: Step 6-2-1: Upper-layer local MPC trajectory planning, with built-in model predictive control algorithm, establishes a discrete prediction model based on real-time status feedback from sensors: in, For state vectors, h ( t )for t The sampling depth of the driving execution unit at all times. ( t )for h ( t The reciprocal of the time, i.e., the sampling advance speed, u ( t () is the control input. w ( t The disturbance term is represented by ), and A, B, and C are constant matrices. The optimal control trajectory is generated by solving the finite-time optimal control problem. u ( t ); Step 6-2-2: Lower-level local LQR fast compensation for execution deviations in MPC trajectories. ( u act ( t (where is the actual control input), design a linear quadratic regulator with the objective function: ; in, Q The state weight matrix is... RTo control the weight matrix, the optimal control law is obtained by solving the equations, and the deviation is corrected in real time to offset the interference between driving execution units and the influence of sudden geological changes.

[0036] Step 6-3: Each drive execution unit executes sampling actions according to the control instructions output by the local sub-controller. The onboard sensor group collects two types of data in real time. The local feedback of local drive execution unit operation data includes, but is not limited to, displacement, pressure, and sampling accuracy, and is transmitted to the local sub-controller for LQR deviation correction. The global feedback of collaborative data between drive execution units includes, but is not limited to, synchronization error between drive execution units and total energy consumption, and is transmitted to the global coordinator for strategy adjustment, forming a closed-loop process of "global coordination - local control - execution feedback". Step 6-4: Determine whether the sampling accuracy meets the standard based on the raw sensor data. If it does, continue the sampling operation. If it does not meet the standard, optimize the coordination weights, adjust the control parameters, and re-enter the global coordinator.

[0037] In step 7, as Figure 8 As shown, it includes the following sub-steps: Step 7-1: Standardize the data dimensions of the data acquisition module, remove abnormal data to ensure data authenticity, and collect the core data from simulation and field sampling to form an operational database; Step 7-2: Construct a comprehensive evaluation index set using the runtime database: basic indicators include, but are not limited to, energy consumption, stability, and reliability; distributed indicators include, but are not limited to, interference level, fault tolerance, and distributed collaborative efficiency. Step 7-3: Calculate the objective weights of each indicator using the entropy weight method and generate a comprehensive evaluation result to provide a basis for the rating standard judgment.

[0038] Step 7-3 includes the following sub-steps: Step 7-3-1: Calculate the information entropy value ,in y il For the first i The first set of data l The normalized value of each indicator. m The number of data sets, s l Reflects the degree of dispersion of indicator data; Step 7-3-2: Calculate index weights based on information entropy ,in n The total number of indicators, with a weight sum of 1; Step 7-3-3: Calculate the comprehensive evaluation index The three-dimensional rating includes energy efficiency rating, reliability rating, and distributed collaboration rating; the comprehensive evaluation index and the three-dimensional rating results are output to the rating compliance judgment unit, which then completes the compliance judgment.

[0039] Step 8 includes the following sub-steps: Step 8-1: If step 7-3 determines that the standard has been met, change the current working condition label. L With the co-optimal parameter set X Bind and update the multi-parameter mapping database; continue to execute the sampling job according to the current optimal parameter set until all sampling tasks are completed, monitor data in real time during the process, and automatically trigger the re-evaluation process if a sudden change in working conditions occurs. Step 8-2: If step 7-3 fails to meet the standard, supplement the sample data under the current working conditions, update the RSM proxy model, and re-execute steps 4-7.

[0040] This invention also includes a multi-site high-fidelity sampling intelligent optimization control method to solve the technical problems of existing multi-site high-fidelity sampling technologies, such as multi-objective performance imbalance, inter-site coordination conflict, insufficient sampling fidelity, low optimization efficiency, and parameter design disconnect. The method includes the following steps: Step 1: Collect on-site working condition parameters and engineering constraint parameters of the target sampling scene, clarify the core objectives of multi-objective optimization, and construct a complete initial input system to provide basic support for subsequent model construction and optimization; Step 2: Based on the initial input data system formed in Step 1, divide the joint design variables into three categories: driving structure, control, and distributed collaboration. Initialize the variable value range, clarify the parameter boundaries, and form the search object and range for multi-objective optimization. Step 3: Based on the variable value range determined in Step 2, generate N sets of "structure-control-distributed" parameter combination samples. Simulate the sampling process under each parameter combination through simulation or field experiment, collect the actual values ​​of the three optimization objectives, construct a training sample set, construct and verify the RSM surrogate model with uncertainty handling based on the sample set, establish the mapping relationship between parameters and optimization objectives, and improve the efficiency of subsequent multi-objective optimization. Step 4: Based on the RSM proxy model constructed in Step 3, configure the adaptive NSGA-II algorithm parameters, define the fitness function, call the RSM proxy model to calculate the objective function value, solve and generate multiple sets of Pareto optimal solutions, and provide multiple parameter combination schemes with performance trade-offs. Step 5: Based on the Pareto optimal solution set generated in Step 4, divide the boundaries of the three domains: structural design domain, control stability domain, and distributed cooperative domain. Perform intersection of the three domains and engineering feasibility verification on the Pareto optimal solution to determine the cooperative optimal parameter set with engineering feasibility. Step 6: Based on the optimal set of parameters determined in Step 5, formulate multi-site task allocation strategies and conflict arbitration rules, and deploy a hierarchical control strategy of "upper-level trajectory planning + lower-level rapid compensation"; each site executes sampling actions according to control instructions, and simultaneously collects two types of data in real time through sensors, feeds back the site's operation data to the local control terminal for deviation correction, and feeds back the collaborative data of each site to the global coordination terminal for strategy adjustment, forming a closed-loop process of "global coordination - local control - execution feedback"; Step 7: Based on the operational data collected in Step 6, an operational database is formed, a comprehensive evaluation index set is constructed, and after preprocessing, the index weights are calculated using the entropy weight method to generate a three-dimensional rating and a comprehensive evaluation index to determine whether the standard is met. Step 8: Based on the performance assessment results in Step 7, if the rating meets the standard, bind the working condition label with the collaborative optimal parameter set and update the multi-parameter mapping database. Subsequent similar working conditions can quickly match parameters through keyword retrieval. If the rating does not meet the standard, supplement the sample data under the current working condition, update the RSM proxy model, and re-execute Steps 4-7 until the rating meets the standard.

[0041] In step 1, as Figure 3 As shown, it includes the following sub-steps: Step 1-1: Collect key information from the target sampling scenario and identify two types of core input parameters: on-site working condition parameters and engineering constraint parameters. All parameters are confirmed through on-site measurements or engineering design documents to ensure data authenticity. Step 1-2: Define and quantify the three core optimization objectives, namely, minimizing the total energy consumption of sampling. E ( E (Total energy consumed during sampling) and maximize continuous runtime T ( T (For continuous operation of the equipment), and to ensure the accuracy of formation parameter acquisition. e ( e (where is the deviation between the sampled data and the actual formation parameters), satisfying 0 ≤ e ≤ e lim ( e lim (to the maximum allowable error), achieving a synergistic balance between energy efficiency, reliability, and sampling accuracy; Steps 1-3: Integrate the input parameters from Step 1-1 with the optimization objectives from Step 1-2, clarify the influence relationship between each parameter and the objective, define the limiting boundaries of the parameter values ​​under the constraints, and form a complete initial input system to provide a basis for subsequent design variable definition and model training.

[0042] In step 2, as Figure 3 As shown, it includes the following sub-steps: Step 2-1: Divide into two categories of core design variables and clarify the physical meaning of each variable: Traditional variables are the core parameters of the driving structure, which directly affect the mechanical performance of the equipment; Distributed variables are divided into control parameters and distributed parameters, which respectively ensure the accuracy of sampling actions and synchronous operation in multiple sites. Step 2-2: Based on the constraints defined in Step 1-3, and in combination with equipment performance parameters, engineering design specifications and actual site conditions, set reasonable value ranges for each variable to ensure sufficient optimization space without deviating from the actual application scenario, thus providing an effective search range for subsequent sample generation and multi-objective optimization.

[0043] In step 3, as Figure 3 As shown, it includes the following sub-steps: Step 3-1: Generate N sets of sample combinations within the variable value range of Step 2-2. Each set of samples corresponds to a unique "structure-control-distributed" parameter combination to ensure that the value of each variable is evenly distributed within its range, without obvious clustering or omission areas, thereby improving the representativeness of the samples. Step 3-2: Simulate the sampling process under each parameter combination through simulation or field experiments, and record the actual values ​​of the three optimization objectives in real time, including the actual energy consumption values ​​collected through energy consumption monitoring methods. E act Continuous operating time of equipment that indirectly characterizes the failure rate T run The acquisition error was calculated by comparing the sampled data with the actual formation parameters. e act , forming a training sample set {( x 1, y 1),( x 2, y 2),…,( x N , y N )}( x For parameter combinations, y =( E act , T run , e act (The actual value is the target value). Step 3-3: Construct an RSM proxy model with uncertainty handling; Steps 3-4: Calculate the model predictions Compared with the actual sample value y act relative error ,like d ≤ d 0 ( d(0 is the error threshold), the model prediction accuracy meets the engineering requirements, and the verification is passed; if d > d 0, add extreme working condition samples and retrain until the error meets the requirements.

[0044] Step 3-3 includes the following sub-steps: Step 3-3-1: Using the design variables as independent variables and the three optimization objectives as response variables, the Box-Cox transformation is introduced to address data skewness. The uncertainty of the response variables is quantified using trapezoidal fuzzy logic. The model expression is: ; ; in x i , x j For the first design variable i The, the j 1 variable ( i , j It is a positive integer, 1≤ i ≤ j ≤ k ), k The total number of variables, αL 0(P) , αU 0(P) For trapezoidal fuzzy numbers P Boundary coefficients of the constant term at the horizontal level, αL i(P) , αU i(P) For trapezoidal fuzzy numbers P Boundary coefficients for the first-order term at the horizontal level. αL ij(P) , αU ij(P) For trapezoidal fuzzy numbers P Boundary coefficients of quadratic terms at the horizontal level. y L (P) , y U (P) These are the lower and upper bounds of the response variable, respectively, covering the uncertainty range of the target value; Step 3-3-2: If the response variable does not satisfy the normality relationship, use the Box-Cox transformation, the formula is: when p ≠0 o'clock, ;when p When =0, ; in y For the original response variable, y(p) For the transformed variables, p For the transformation parameters, t This is a preset constant (with a value between 0.1 and 1, to avoid data being 0); Simultaneously, trapezoidal fuzzy quantitative uncertainty is introduced, with its lower bound... α L (P) With the upper realm α U (P) The range of fluctuations in statistical sample data is determined by the following calculation formula: when hour, ;when hour, ;when hour, ;when hour, ; when hour, ;when hour, ;when hour, ;when hour, ; in a , d These are the lower and upper bounds of the trapezoidal fuzzy number, respectively. b The inflection point of the membership function. c This is the inflection point of the membership function. Step 3-3-3: Construct individual conditional expectation curves for each design variable, fixing other variables at their mean or median, and only changing the target variable. The formula is as follows: , i As a parameter vector, it intuitively displays the changing trend of the response target when a single variable changes, improving the interpretability of the model; In step 4, as Figure 3 As shown, it includes the following sub-steps: Step 4-1: Based on the variable dimension and the number of optimization targets, set the core parameters of the adaptive NSGA-II algorithm, including but not limited to population size, number of iterations, adaptive crossover probability, and mutation probability; Step 4-2: Call the RSM proxy model constructed in Step 3-3, using the three optimization objectives as the fitness function and unifying the objective direction, which are: Energy consumption target: , Failure rate target: ( T max (Rated continuous operating time). Sampling accuracy target: ; Step 4-3: The design variables are calculated using the adaptive NSGA-II algorithm. Based on fast non-dominated sorting, crowding calculation, elite retention strategy and adaptive genetic operation, the multi-objective evolution optimization of the design variables is realized, and finally multiple Pareto optimal solutions are generated to form the optimal parameter combination set.

[0045] In step 4-3, as Figure 4 As shown, it includes the following sub-steps: Step 4-3-1: Randomly generate the parameters from Step 4-1, with a scale of... N The initial parent population P 0 and the size after fitness calculation using the RSM surrogate model is N offspring population Q 0 merges to form a mixed population R t = P t ∪ Q t ( t (current iteration number), population size 2 N This operation expands the search space, prevents the loss of outstanding individuals from the previous generation, and provides a foundation for the preservation of elites; Step 4-3-2: Analysis of mixed populations based on Pareto dominance relationships R t By implementing layering, the computational complexity of traditional sorting can be reduced; (1) For any two individuals in the population x i and x j If the following conditions are met: ; and ; Then it is called x i Dominate x j , recorded as x i x j ,in f k ( The fitness function is defined in step 4-2. (2) For each individual x p ∈ R t Calculate two core attributes: Dominated Count n p ( np Dominant in the population x p (Number of individuals), dominant solution set S p ( S p For being x p (The set of all individuals under control); (3) Layered iteration, specifically including: ① Initialize the non-dominated layer sequence number rank =1, all n p Individuals with a value of 0 are classified into the first... rank layer F rank ; ② Traversal F rank Each individual x p The solution set of its domination S p Each individual x q ,implement n q = n q 1; if n q =0, then x q Return to the next level F rank+1 ; ③ Order rank = rank +1, repeat the above operation until all individuals are assigned to the corresponding non-dominated layer. F 1, F 2,…, F k ( F 1 is the optimal non-dominated layer). Step 4-3-3: For each non-dominated layer, calculate the crowding distance of individuals within the layer. d p To quantify the sparsity around an individual and avoid clustering of solution sets, the specific calculation is as follows: (1) For each individual x p ∈ F rank Settings d p =0; (2) For the first k There is one objective function, which normalizes the function values ​​of all individuals within the layer to [0,1]. The formula is as follows: ; in For the first k The objective function is F rank Minimum value of the layer, For the first k The objective function is F rank The maximum value of the layer; (3) For each objective function k ,according to Individuals within the ascending order of the hierarchy are grouped, and the crowding of boundary individuals is set to [value missing]. d p =∞; for sorting positions, i intermediate individuals x p The calculation formula is: ; After traversing all objective functions, d p That is, an individual x p Total congestion distance; Step 4-3-4: Based on the non-dominated layer index and crowding distance, from the mixed population R t Selected from the scale of N The next generation of parent population P t+1 The process is as follows: (1) Individuals are included in the sequence from low to high. P t+1 until it is incorporated into a certain layer F After 1, the population size exceeded N ; (2) To F Individuals in 1 are ranked by crowding distance d p Sort in descending order and select the first few. N -| P t+1 |Individuals included P t+1 To ensure the final population size is N ; Step 4-3-5: [Regarding...] P t+1 Perform selection, crossover, and mutation operations to generate offspring populations. Q t+1 The process is as follows: (1) Using the binary tournament selection method, two individuals are randomly selected, with priority given to individuals with lower non-dominated layer numbers; if the numbers are the same, individuals with larger crowding distances are selected, and this process is repeated. N Next, a selective population is formed. P s ; (2) Simulated binary crossover is used, and the crossover probability is... p c The cross formula adjusts linearly with the number of iterations and is as follows: ; ; in x p , x q ∈ P s For paired individuals, β The cross-distribution index, k To design variable dimensions; (3) Using polynomial mutation, the mutation probability p m The mutation formula is: ; in x k,max , x k,min Design variables x k The boundary values ​​of Δ k For variable asynchronous length, the calculation formula is: when hour, ; when hour, ; or m It is the distribution index of variation. r ∈[0,1] represents a random number; Step 4-3-6: Let the number of iterations be... t = t +1, will P t+1 Substitute the values ​​from step 3-3 into the RSM surrogate model to calculate fitness, and repeat steps 4-3-1 to 4-3-5; stop iterating when the number of iterations reaches the termination condition; finally, multiple Pareto optimal solutions are generated, forming the optimal parameter combination set { X 1, X 2,…, X W}( W(where the number of Pareto optimal solutions is ), and each solution satisfies: ; This means that it is impossible to improve a certain objective without reducing the performance of other objectives, thus failing to achieve optimal multi-objective collaboration.

[0046] In step 5, as Figure 5 As shown, it includes the following sub-steps: Step 5-1: Based on the design specifications of the drive execution device, the stability requirements of control theory, and the distributed coordination mechanism, delineate the boundaries of the three feasible domains; Specifically, the structural design domain is based on strength and spatial constraints, satisfying the strength constraints. ( s The actual stress of the structure, s lim (allowable stress of material) and dimensional constraints ( x Variables for structural design. x min , x max (These represent the minimum and maximum structural dimensions, respectively). The control stability domain is based on Lyapunov stability and phase margin constraints, satisfying the no-overshoot constraint for closed-loop control. ( s p For overshoot, s p,lim To allow overshoot and response time constraints ( t r For response time, t r,lim To allow for response time); the distributed cooperative domain, based on interference and communication delay constraints, satisfies the conflict-free constraint between the execution of drives in each site. ( t com Due to communication delay, t lim (Maximum allowable communication delay) and synchronization response constraints (Δ) t sync For the synchronization response error, Δ t sync,lim To allow for synchronization response error); Step 5-2: For each Pareto optimal solution generated in Step 4-3, perform set operations to find the intersection of the three regions and determine whether the feasible region is non-empty. Figure 6 The schematic diagram of the three-domain intersection method described in this embodiment of the invention is shown below; If there exists at least one solution that simultaneously satisfies the three-domain constraints (the feasible domain is not empty), proceed to step 5-3; if the feasible domain is empty, return to step 2-2 to adjust the boundary values ​​of the design variables, and re-execute steps 3-4. Step 5-3: Verify the engineering feasibility of the parameter combinations in the intersection of the three domains, including structural verification, control verification, and distributed verification; if all verifications are successful, form a collaborative optimal parameter set after correcting local conflicts. ( X dis For distributed optimal parameters X con To control the optimal parameters, X str (For the optimal parameters of the structure); if the verification fails, the feasible region constraint range is fed back to the multi-objective optimization stage, and the Pareto optimal solution is searched again and verified again until the target is met.

[0047] In step 6, as Figure 7 As shown, it includes the following sub-steps: Step 6-1: Based on the collaborative optimal parameter set, formulate two core strategies: multi-site task allocation allocates sampling priority according to the differences in stratigraphic parameters to avoid excessive concentration during high-fidelity sampling at each site; conflict arbitration rules define the conflict determination function. ; in, t com,i For the venue i Communication delay, I ( ) is an indicator function (it takes the value 1 if the condition is met, and 0 otherwise); when C ( t When )>0, sampling is performed in priority order to avoid stress interference and data transmission conflicts caused by simultaneous sampling; Step 6-2: Each local sub-controller adopts a hierarchical control strategy of "MPC trajectory planning + LQR fast compensation"; Step 6-2 includes the following sub-steps: Step 6-2-1: Upper-layer local MPC trajectory planning, with built-in model predictive control algorithm, establishes a discrete prediction model based on real-time status feedback from sensors: in, For state vectors, h ( t )for t Sampling depth at time, ( t )for h ( t The reciprocal of the time, i.e., the sampling advance speed, u (t () is the control input. w ( t The disturbance term is represented by ), and A, B, and C are constant matrices. The optimal control trajectory is generated by solving the finite-time optimal control problem. u ( t ); Step 6-2-2: Lower-level local LQR fast compensation for execution deviations in MPC trajectories. ( u act ( t (where is the actual control input), design a linear quadratic regulator with the objective function: ; Among them, Q The state weight matrix is... R To control the weight matrix, the optimal control law is obtained by solving the equations, and the deviation is corrected in real time to offset the interference between sites and the impact of sudden geological changes.

[0048] Step 6-3: Each drive actuator performs sampling actions according to the control instructions output by the local sub-controller. The onboard sensor group collects two types of data in real time. The local feedback of site operation data includes, but is not limited to, displacement, pressure, and sampling accuracy, and is transmitted to the local sub-controller for LQR deviation correction. The global feedback of collaborative data between sites includes, but is not limited to, synchronization error between sites and total energy consumption, and is transmitted to the global coordinator for strategy adjustment, forming a closed-loop process of "global coordination - local control - execution feedback". Step 6-4: Determine whether the sampling accuracy meets the standard based on the raw sensor data. If it does, continue the sampling operation. If it does not meet the standard, optimize the coordination weights, adjust the control parameters, and re-enter the global coordination phase.

[0049] In step 7, as Figure 8 As shown, it includes the following sub-steps: Step 7-1: Standardize the dimensions of collected data, remove outlier data to ensure data authenticity, and compile core data from simulation and field sampling to form an operational database; Step 7-2: Construct a comprehensive evaluation index set using the runtime database: basic indicators include, but are not limited to, energy consumption, stability, and reliability; distributed indicators include, but are not limited to, interference level, fault tolerance, and distributed collaborative efficiency. Step 7-3: Calculate the objective weights of each indicator using the entropy weight method and generate a comprehensive evaluation result to provide a basis for the rating standard judgment.

[0050] Step 7-3 includes the following sub-steps: Step 7-3-1: Calculate the information entropy value , among whichy il For the first i The first set of data l The normalized value of each indicator. m The number of data sets, s l Reflects the degree of dispersion of indicator data; Step 7-3-2: Calculate index weights based on information entropy ,in n The total number of indicators, with a weight sum of 1; Step 7-3-3: Calculate the comprehensive evaluation index The three-dimensional rating includes energy efficiency rating, reliability rating, and distributed collaboration rating; the comprehensive evaluation index and the output of the three-dimensional rating are used to determine whether the rating meets the standards, and the standard determination is completed.

[0051] Step 8 includes the following sub-steps: Step 8-1: If step 7-3 determines that the standard has been met, change the current working condition label. L With the co-optimal parameter set X Bind and store data in the database hierarchically; continue to execute the sampling job according to the current optimal parameter set until all sampling tasks are completed, monitor data in real time during the process, and automatically trigger the re-evaluation process if a sudden change in working conditions occurs. Step 8-2: If step 7-3 fails to meet the standard, supplement the sample data under the current working conditions, update the RSM response surface model, and re-execute steps 4-7.

[0052] To verify the effectiveness of the system and method of this invention, a specific implementation verification was carried out using a hydraulic direct-push sampling operation for soil investigation at a contaminated site as an application scenario. It should be noted that, unless otherwise specified, the embodiments and features of this invention can be combined with each other.

[0053] Example 1: The experimental conditions for this embodiment are set as follows: Three sites are selected; the soil density is 1.7 g / cm³, the soil moisture content is 22%, and the target sampling depth is 5 m; engineering constraints include a maximum manufacturing cost of 150,000 yuan and a maximum equipment volume of 0.8 m³. 3 Structural strength allowable stress s lim =300MPa, sampling accuracy error threshold e lim =±1mm, maximum permissible communication delay t lim =0.05s, hydraulic oil density r =870kg / m³, reference bulk modulus of hydraulic oil β0 = 1.4 × 10³ MPa. The joint design variables and their value ranges are shown in the table below:

[0054] The joint design variables shown in the table above cover three core parameters: driving structure, control, and distributed systems. This ensures sufficient optimization space while adhering to actual engineering constraints, providing an effective search range for subsequent sample generation and multi-objective optimization.

[0055] In the process of model construction and multi-objective optimization, Latin hypercube design was first used to generate 150 sets of "structure-control-distributed" parameter combination samples within the above variable value range. The actual values ​​of the objective function (energy consumption) for each set of samples were then collected through joint simulation using AMESim and Simulink. E act Continuous operating time of equipment T run Sampling error e act ), construct training sample set {( x 1, y 1),( x 2, y 2),…,( x 150 , y 150 )}, where y=( E act , T run , e act The RSM surrogate model validation error data is shown in the table below:

[0056] This table shows the prediction accuracy of the RSM surrogate model for the three optimization objectives. The average relative errors for energy consumption, reliability, and sampling accuracy objectives across 150 samples are 2.46%, 1.83%, and 3.52%, respectively, with an overall average relative error... d =3.2%≤5%, which meets the engineering prediction accuracy requirements and does not require additional extreme working condition samples, thus verifying the effectiveness of the model.

[0057] Based on this sample set, a second-order RSM surrogate model with uncertainty handling was constructed. The Box-Cox transform was introduced to handle data skewness, and the uncertainty was quantified through trapezoidal fuzzy modeling. Subsequently, the adaptive NSGA-II algorithm was optimized, with the following parameters: population size 150, number of iterations 250, adaptive crossover probability 0.85 (decreasing linearly with the number of iterations, 0.9 in the early stage and 0.8 in the late stage), mutation probability 0.08 (temporarily increased to 0.12 when population diversity is insufficient), and the fitness function was defined as minimizing.

[0058]

[0059] By comparing the optimization results of the method of this invention with those of the traditional method, it can be seen that the method of this invention achieves significant optimization in core indicators such as total sampling energy consumption, sampling error, and unit collaborative interference degree, with an improvement of more than 23%. The constraints such as equipment volume and manufacturing cost are all controlled within the engineering upper limit and are lower than those of the traditional method, which verifies the effectiveness of multi-objective collaborative optimization.

[0060] Through iterative optimization using fast non-dominated sorting, crowding calculation, and elite retention strategies, a compromise-optimal parameter set was selected, where the driving structural parameter is: r =11mm D =63mm D p =15mL / r, control parameters are: N p =4、 Q =diag([10,5]) R =diag([0.1]), the distributed parameter is T c =0.02s、 W =0.5.

[0061] In the three-domain collaborative verification phase, the feasibility of the parameter combination is verified in the structural design domain, control stability domain, and distributed collaborative domain.

[0062] All verification indicators meet the preset standards, indicating that the intersection of the three domains is non-empty. This parameter combination passes the engineering feasibility verification and forms a collaboratively optimal parameter set. X .

[0063] During the distributed hierarchical control execution and performance testing process, the global coordinator allocates sampling priorities according to the soil compaction of each site, and the conflict determination function... There are no communication conflicts, and sampling is performed in priority order. The local sub-controller adopts a hierarchical control strategy of "MPC trajectory planning + LQR fast compensation". The upper-level MPC trajectory planning prediction step size is 4, and the discrete prediction model response time is 11ms, generating the optimal control trajectory. u ( t =0.05m / s, lower-level LQR fast compensation targets execution deviation Δ u =0.3mm, solve the Riccati equation to obtain the optimal control law, and compensate for the sampling depth error. e =±0.5mm≤ e lim =±1mm; During the fault tolerance test, a displacement sensor failure was simulated in the sampling unit of site #2. The global coordinator reassigned the sampling tasks, and the sampling units of sites #1 and #3 operated normally without interruption of the sampling tasks.

[0064] In the comprehensive evaluation phase, a comprehensive evaluation index set is constructed based on the preprocessed operational data from the data acquisition module. The entropy weight method is used to calculate the index weights, ultimately yielding the comprehensive evaluation index. S =0.86≥0.8, Energy efficiency level 0.82 (Good), Reliability level 0.85 (Good), Distributed collaboration level 0.88 (Good), meeting the compliance threshold requirements. In the multi-parameter mapping database application verification, the working condition labels "3 sites + density 1.7g / cm³ + moisture content 22% + sampling depth 5m + cost 142,000 yuan" were matched with the optimal collaborative parameter set. X The system binds and updates the multi-parameter mapping database, allowing for rapid parameter tuning for subsequent similar operating conditions through database keyword searches, significantly improving efficiency compared to traditional re-optimization.

Claims

1. A multi-site high-fidelity sampling intelligent optimization control system, characterized in that, It includes a working condition input and constraint definition module, used to receive on-site working conditions and engineering constraint parameters for each site; the output of the working condition input and constraint definition module is connected to the input of the RSM surrogate model module and the multi-objective optimization module; the output of the RSM surrogate model module is connected to the input of the multi-objective optimization module; the data end of the multi-objective optimization module is bidirectionally connected to the data end of the three-domain collaboration module; the output of the three-domain collaboration module is connected to the input of the hierarchical control module; the data end of the hierarchical control module is bidirectionally connected to the data end of the data acquisition module; and the output of the data acquisition module is connected to the input of the comprehensive evaluation module.

2. The system according to claim 1, characterized in that, The comprehensive evaluation module is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

3. The system according to claim 1, characterized in that, The output of the comprehensive evaluation module is connected to the input of the rating and compliance judgment unit. The rating and compliance judgment unit compares the evaluation results with the preset compliance threshold based on the comprehensive evaluation index and three-dimensional rating results output by the comprehensive evaluation module. If the standard is met, the working condition label and the optimal parameter set are stored in the multi-parameter mapping database. If the standard is not met, the adjustment instruction is fed back to the RSM proxy model module to trigger iterative optimization.

4. The system according to claim 3, characterized in that, The multi-parameter mapping database is used to store compliance data in the hierarchy of "multiple site quantities - stratigraphic parameters - engineering constraints - three-dimensional rating - optimal parameter set", providing parameter matching and one-click tuning support for engineering application terminals; the data output end of the multi-parameter mapping database is connected to the data input end of the engineering application terminal, and the data output end of the engineering application terminal is connected to the input end of the working condition input and constraint definition module. The engineering application terminal is used to input on-site working conditions and constraint parameters, receive one-click parameter setting and issue configuration commands to achieve rapid system deployment.

5. The system according to claim 1, characterized in that, The hierarchical control module includes a global coordinator, local sub-controllers, drive execution units, and a sensor group. The output of the global coordinator is connected to the input of the local sub-controller, the output of the local sub-controller is connected to the input of the drive execution unit, the output of the drive execution unit is connected to the input of the sensor group, and the output of the sensor group is connected to the global coordinator, the local sub-controller, and the data acquisition module, respectively. The global coordinator is responsible for task allocation and conflict arbitration in multiple locations. The local sub-controller adopts a hierarchical control strategy of "upper-level trajectory planning + lower-level fast compensation" to realize multi-location single-machine collaborative operation control, ensure the high-fidelity sampling requirements of each location, and improve the system's fault tolerance. Each site drive execution unit is used to execute core actions such as sampling propulsion, stopping, and fine-tuning according to control commands; each site sensor is used to collect real-time operating data of the drive execution unit and inter-unit collaborative data, providing data support for the feedback control and comprehensive evaluation module; The data acquisition module is used to receive data from the sensor group, determine in real time whether the sampling accuracy meets the standard, and perform data standardization and outlier removal to generate an operating database. The comprehensive evaluation module is used to construct a comprehensive evaluation index system covering energy saving, reliability, and distributed collaboration based on the entropy weight method, calculate the comprehensive evaluation index and three-dimensional rating results, and realize quantitative and unified evaluation.

6. The system according to claim 1, characterized in that, The working condition input and constraint definition module receives on-site working conditions and engineering constraint parameters from various sites, processes them through standardization, and outputs them to the RSM surrogate model module and the multi-objective optimization module, providing a unified data foundation for subsequent optimization. The RSM surrogate model module trains a high-precision substitution model of "design variables - multi-objective performance" based on simulation samples, replacing traditional time-consuming simulation or experimental calculations and providing rapid computational support for the multi-objective optimization module. The multi-objective optimization module uses driving structural parameters, control parameters, and distributed parameters as joint design variables, and generates multiple sets of Pareto optimal solutions through the adaptive NSGA-II algorithm to achieve performance trade-offs under multiple constraints such as energy consumption, volume, and cost. The three-domain collaboration module divides the structural design domain, control stability domain, and distributed collaboration domain, obtains a conflict-free feasible domain through the intersection of the three domains, and verifies the engineering feasibility of the Pareto optimal solution, realizing the collaborative adaptation of structural parameters, control parameters, and distributed parameters.

7. The system according to any one of claims 1 to 6, characterized in that, When the system is working / being used, the following steps are taken: Step 1: Collect on-site working condition parameters and engineering constraint parameters of the target sampling scene, clarify the core objectives of multi-objective optimization, and construct a complete initial input system to provide basic support for subsequent model construction and optimization; Step 2: Based on the initial input data system formed in Step 1, divide the joint design variables into three categories: driving structure, control, and distributed collaboration. Initialize the variable value range, clarify the parameter boundaries, and form the search object and range for multi-objective optimization. Step 3: Based on the variable value range determined in Step 2, generate parameter samples, collect simulation sample data through simulation or field experiments, construct and verify the RSM surrogate model, establish the mapping relationship between parameters and optimization objectives, and improve the efficiency of subsequent multi-objective optimization; Step 4: Based on the RSM proxy model constructed in Step 3, configure the adaptive NSGA-II algorithm parameters, define the fitness function, call the RSM proxy model to calculate the objective function value, solve and generate multiple sets of Pareto optimal solutions, and provide multiple parameter combination schemes with performance trade-offs. Step 5: Based on the Pareto optimal solution set generated in Step 4, divide the boundaries of the three domains: structural design domain, control stability domain, and distributed cooperative domain. Perform intersection of the three domains and engineering feasibility verification on the Pareto optimal solution to determine the cooperative optimal parameter set with engineering feasibility. Step 6: Based on the optimal set of parameters determined in Step 5, deploy the global coordinator and local sub-controllers. Through global coordination and hierarchical control, each site drive execution unit executes sampling actions according to instructions, synchronously collects operating data, and forms a two-way feedback closed loop. Step 7: Based on the operational data collected in Step 6, an operational database is formed, a comprehensive evaluation index set is constructed, and after preprocessing, the index weights are calculated using the entropy weight method to generate a three-dimensional rating and a comprehensive evaluation index to determine whether the system performance meets the standards. Step 8: Based on the performance assessment results in Step 7, if the rating meets the standard, bind the working condition label with the optimal collaborative parameter set and update the multi-parameter mapping database; if it does not meet the standard, update the RSM proxy model until the system performance meets the standard.

8. The system according to claim 7, characterized in that, Step 1 includes the following sub-steps: Step 1-1: Collect key information from the target sampling scenario and identify two types of core input parameters: on-site working condition parameters and engineering constraint parameters. All parameters are confirmed through on-site measurements or engineering design documents to ensure data authenticity. Step 1-2: Define and quantify the three core optimization objectives, namely, minimizing the total energy consumption of sampling. E Maximize continuous runtime T Ensure the accuracy of formation parameter acquisition. ε , satisfying 0≤ ε ≤ ε lim ,in ε lim To achieve a balanced approach between energy efficiency, reliability, and sampling accuracy, the maximum allowable error is set. Steps 1-3: Integrate the input parameters from Step 1-1 with the optimization objectives from Step 1-2, clarify the influence relationship between each parameter and the objective, define the limiting boundaries of the parameter values ​​under the constraints, and form a complete initial input system to provide a basis for subsequent design variable definition and model training.

9. The system according to claim 8, characterized in that, Step 2 includes the following sub-steps: Step 2-1: Divide into two categories of core design variables and clarify the physical meaning of each variable: Traditional variables are the core parameters of the driving structure, which directly affect the mechanical performance of the equipment; Distributed variables are divided into control parameters and distributed parameters, which respectively ensure the accuracy of sampling actions and synchronous operation in multiple sites. Step 2-2: Based on the constraints defined in Step 1-3, and in combination with equipment performance parameters, engineering design specifications and actual site conditions, set reasonable value ranges for each variable to ensure sufficient optimization space without deviating from the actual application scenario, thus providing an effective search range for subsequent sample generation and multi-objective optimization.

10. The system according to claim 9, characterized in that, Step 3 includes the following sub-steps: Step 3-1: Generate N sets of sample combinations within the variable value range of Step 2-2. Each set of samples corresponds to a unique "structure-control-distributed" parameter combination to ensure that the value of each variable is evenly distributed within its range, without obvious clustering or omission areas, thereby improving the representativeness of the samples. Step 3-2: Simulate the sampling process under each parameter combination through simulation or field testing, and record the actual values ​​of the three optimization objectives in real time, including the actual energy consumption value collected by the energy consumption monitoring module. E act Continuous operating time of equipment that indirectly characterizes the failure rate T run The acquisition error was calculated by comparing the sampled data with the actual formation parameters. ε act , forming a training sample set {( x 1, y 1),( x 2, y 2),…,( x N , y N )},in x For parameter combinations, y =( E act , T run , ε act () represents the actual target value; Step 3-3: Construct an RSM proxy model with uncertainty handling.

Images