Novel functional coating intelligent control method based on real-time data fusion

Abstract

The invention discloses a novel functional coating intelligent control method based on real-time data fusion, and the method comprises the following steps: S1, collecting multi-source environment data of a functional coating deployment region, and constructing a real-time environment state vector; s2, modeling the adjustable attributes of the functional coating into a control variable set; s3, constructing an initial control strategy set based on a brainstorm optimization algorithm, and generating a new control strategy set through a disturbance generation mechanism; s4, generating a plurality of local disturbance strategies around the reference strategy, and constructing a firework strategy set; s5, combining the new control strategy set with the firework strategy set to form a candidate control strategy complete set; s6, performing fitness calculation on the updated control strategy set; and S7, comparing the actual response vector with the expected response vector. According to the method, multi-source data fusion, control variable modeling and brainstorm optimization algorithm technologies are combined, and a novel functional coating intelligent control method based on real-time data fusion is realized.

Description

Technical Field

[0001] The present invention relates to an intelligent functional coating, and in particular to a novel intelligent control method for functional coatings based on real-time data fusion. Background Art

[0002] With the development of materials science, sensing technology, and intelligent control, functional coatings have gradually been applied to intelligent response scenarios in a variety of complex environments, such as temperature control, corrosion protection, self-cleaning, photothermal conversion, electromagnetic shielding, and other new functional requirements. Such coatings are required not only to have excellent material properties, but also to have a certain degree of environmental adaptability, and to be able to actively adjust their functional output according to changes in the external environment, so as to meet the requirements of reliable operation under multiple working conditions and multi-field coupling states. Such coatings are usually called intelligent functional coatings. Their performance depends on sensitive responses to environmental factors such as external temperature, humidity, light intensity, electric field, gas concentration, etc., and dynamic adjustment of functional behavior is achieved through internal regulation mechanisms.

[0003] However, in the existing technology, the control strategies of functional coatings mostly use fixed models, static rules or artificial experience parameter settings. Their adjustment logic cannot respond to the dynamic changes of environmental conditions in a timely manner, resulting in problems such as response lag, imprecise control, and insufficient adjustment in practical applications. Although some more advanced control methods have introduced simple feedback adjustment mechanisms or control functions, due to the lack of the ability to integrate modeling between complex environmental factors, the strategy generalization ability is weak, the adaptability is poor, and the control dimension is limited. In addition, the existing strategy optimization methods are often based on traditional heuristic algorithms such as genetic algorithms, ant colony optimization, particle swarm optimization, etc., but when faced with complex control parameters in multi-dimensional and highly dynamic spaces, these algorithms are prone to falling into local optimality, low search accuracy, and slow convergence speed.

[0004] For the dynamic control of functional coatings, the existing methods have the following main shortcomings: First, they lack the ability to systematically integrate and model multi-source sensor data, and usually rely only on single or low-dimensional environmental parameters for control strategy design, which makes it difficult to operate stably in a multi-field coupled environment; second, the generation logic of the control strategy does not have the ability of evolutionary learning, and cannot dynamically optimize the control scheme based on historical feedback or multiple rounds of iterations, resulting in strategy aging or failure; third, the strategy optimization algorithms are mostly based on global search, lack global-local collaborative optimization capabilities, and cannot take into account both search breadth and control accuracy; fourth, in terms of feedback update, single-point error correction or simple adjustment based on the current sampling moment are mostly used, which cannot comprehensively model the response deviation of the control behavior in the time domain, and the feedback mechanism lacks consistency and strong robustness.

[0005] Furthermore, traditional optimization control methods often overlook the co-evolutionary relationship between environmental states and control strategies, making it difficult to proactively adjust control behavior in dynamic environments. While some systems have introduced deep learning models or data-driven algorithms, these systems often focus on state prediction or anomaly detection, lacking coupling with the control strategy development process. This makes it difficult to form closed-loop optimization control mechanisms. Furthermore, existing technologies generally fail to utilize swarm intelligence algorithms for modeling structural policy perturbations, and even more so, lack deep integration methods that synergize different swarm intelligence mechanisms for constructing policy spaces.

[0006] Existing control models often employ static structures to integrate strategy updates with environmental response feedback. These models lack systematic feedback channels for response results after strategy execution, making it impossible to establish effective model self-repair and evolution mechanisms. If the coating's actual response after strategy output falls short of expectations, manual resetting of strategy parameters is often necessary. This lack of structured assessment of error sources and model-level correction capabilities severely restricts the practicality and scalability of intelligent control.

[0007] Therefore, how to provide a new functional coating intelligent control method based on real-time data fusion is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0008] One purpose of the present invention is to propose a new intelligent control method for functional coatings based on real-time data fusion. The present invention combines advanced technologies such as multi-source data fusion, control variable modeling, brainstorming optimization algorithm and fireworks search algorithm. By constructing the environmental state vector and the coating control variable set in real time, the heuristic strategy clustering mechanism and the local disturbance generation mechanism are combined to achieve the breadth exploration and local fine-tuning of the control strategy. The system comprehensively evaluates the strategy set based on the multi-objective fitness function, constructs a feedback-driven error correction path, and dynamically adjusts the strategy model parameters to ensure the intelligent response, adaptive regulation and model evolution of the functional coating in a complex environment. This method has high precision, high robustness and closed-loop optimization capabilities, significantly improving the environmental adaptability and control intelligence level of the coating system.

[0009] According to an embodiment of the present invention, a novel intelligent control method for functional coatings based on real-time data fusion includes the following steps:

[0010] S1. Collect multi-source environmental data in the functional coating deployment area, perform normalization processing, and construct a real-time environmental state vector;

[0011] S2, modeling the adjustable properties of the functional coating as a set of control variables;

[0012] S3. Construct an initial control strategy set based on the brainstorming optimization algorithm, cluster and group the control strategies in the set using the real-time environment state vector, and generate a new control strategy set through the disturbance generation mechanism;

[0013] S4. Select the target control strategy from the new control strategy set as the reference strategy, input it into the fireworks search algorithm, generate several local perturbation strategies around the reference strategy, and construct a fireworks strategy set;

[0014] S5. Merge the new control strategy set with the fireworks strategy set to form a complete set of candidate control strategies, sort the complete set of candidate control strategies according to a preset fitness function, and select the top several strategies to form an updated control strategy set;

[0015] S6. Construct a multi-objective fitness function and calculate the fitness of the updated control strategy set;

[0016] S7. Select a target update control strategy to execute a control operation, collect a corresponding actual response vector, and compare the actual response vector with the expected response vector.

[0017] Optionally, the S1 specifically includes:

[0018] S11. Collecting multi-source environmental data in the functional coating deployment area, wherein the multi-source environmental data includes temperature, humidity, light intensity, electric field intensity, and gas concentration;

[0019] S12. Time-align, remove noise, process outliers, and interpolate missing data for the collected multi-source environmental data to construct a standardized observation data structure under the same time base. Specifically, this includes aligning various sensor data according to timestamps to form a unified time series format, using a sliding mean interpolation method to complete missing data, using a median deviation detection method to remove or replace observation outliers, and then using dimensional conversion rules to unify the values ​​of different physical quantities in the multi-source environmental data;

[0020] S13. Calculate the standardized observation data and generate a real-time environment state vector through a weighted statistical mapping method.

[0021] Optionally, the S13 specifically includes:

[0022] S131. Perform multi-dimensional normalization and fusion calculation on the standardized observation data, and generate a real-time environment state vector through a weighted statistical mapping method:

[0023]

[0024] Among them, S(t) is the real-time environmental state vector at time t, Z is the normalization constant, K is the number of types of environmental parameters, Mk is the number of sampling points under the kth type of environmental parameters, k is the category index of environmental data, t is the time index, m is the specific sampling point index in each type of environmental data, km is the index of the mth sampling point in the kth type of environmental data, M is the number of spatial sampling points, w km The fusion weight coefficient set for the m-th sampling point of the k-th class, x km (t) is the original observation value of the mth spatial sampling point in the kth type of multi-source environmental data at time t, σ km is the standard deviation of the m-th sampling point of the k-th category within the preset historical time window, μ km is the mean value of the m-th sampling point of the k-th category within the preset historical time window;

[0025] S132. Using the real-time environment state vector as the environment perception input based on which the control strategy is optimized.

[0026] Optionally, the S2 specifically includes:

[0027] S21. Modeling the adjustable properties of the functional coating as a set of control variables, wherein the control variables include electrical excitation intensity, microstructure adjustment state, thermal response threshold, and surface response pattern, each control variable reflecting the response behavior of the coating under specific environmental conditions;

[0028] S22. Define a reasonable variation range for each control variable, where the variation range is constrained by a preset boundary value;

[0029] S23. Based on the control variables, the control strategy of the functional coating is expressed as a control variable vector:

[0030]

[0031] Where N is the total number of control variables, w i is the weight coefficient of the i-th control variable, i is the index of the control variable, t is the time index, c i (t) is the control value of the i-th control variable at time t, C(t) is the comprehensive control variable vector at time t, S i (t) is the part of the real-time environment state vector at time t that is related to the i-th control variable, is the weighted sum of all control variables from i=1 to i=N.

[0032] Optionally, the S3 specifically includes:

[0033] S31. Construct an initial control strategy set based on a brainstorming optimization algorithm. The control strategy set consists of multiple candidate control strategies. Each candidate control strategy represents the response behavior of the coating under a specific environmental state. The construction formula of the initial control strategy set is:

[0034]

[0035] S32, according to the real-time environment state vector S (t) the initial control strategy set C BSO Perform clustering and grouping to group control strategies with similar response characteristics to reduce the complexity of the strategy space. i is the index of the control variable;

[0036] S33. Use the disturbance generation mechanism to perturb each control strategy and generate a new set of control strategies:

[0037]

[0038] Among them, C BSO' is the set of control strategies after disturbance generation, is the initial control strategy set C BSO The i-th control strategy individual, η is the disturbance gain coefficient, is the starting time point of the disturbance process, is the current time point, Θ i is the disturbance direction selection matrix of the i-th control strategy, and the i-th control variable is The local disturbance adjustment coefficient, is the intermediate time variable in the integration process, is the time from the start of the disturbance To the current moment The time integral of is the i-th control variable at time The disturbance response rate on is the i-th control variable at time The local disturbance adjustment coefficient.

[0039] Optionally, the S4 specifically includes:

[0040] S41, selecting a target control strategy from the new control strategy set as a reference control strategy, recorded as a center strategy;

[0041] S42. Input the center strategy into the fireworks search algorithm. Using this strategy as the explosion center, generate several local perturbation strategies in the local neighborhood to construct a fireworks strategy set. The number and perturbation range of the local perturbation strategies are:

[0042]

[0043] Among them, M spark is the number of spark strategies generated in fireworks search, κ is the upper limit parameter of strategy generation, Fitness(C center ) is the fitness value of the current reference control strategy, is the sum of the fitness of all strategies in the current control strategy set, is the total number of control strategies in the current control strategy set, R i is the local disturbance radius of the i-th control variable, ρ is the disturbance scale adjustment factor, is the maximum fitness value among all strategies, To traverse the index of the current control strategy set, is the kth control strategy in the set, C center is the center control strategy in the current round of fireworks search algorithm, and i is the index of the control variable;

[0044] S43. Based on the local disturbance generation mechanism, a weighted disturbance is applied to each control variable dimension. By introducing the normal distribution disturbance amount and disturbance direction coefficient, combined with the corresponding dimension disturbance radius, the central strategy is disturbed dimension by dimension to construct multiple local spark strategies.

[0045] S44. De-duplicate and correct the boundaries of all local spark strategies to ensure that they fall within the predefined control variable constraints to form a fireworks strategy set.

[0046] Optionally, the S5 specifically includes:

[0047] S51. The new control strategy set generated by the brainstorming optimization algorithm is combined with the fireworks strategy set generated by the fireworks search algorithm to generate a complete set of candidate control strategies:

[0048]

[0049] in, is the first in the fireworks strategy set generated by the fireworks search algorithm Spark strategy, C all is the complete set of candidate control strategies, For all brainstorming strategies With all fireworks perturbation strategies Cross-combination and weighted fusion are performed between The first of the new control strategies generated by the brainstorming optimization algorithm A control strategy, The first Spark strategies, N BSO' The number of new control strategies output by the brainstorming optimization algorithm, M spark The number of spark strategies generated for the fireworks search, To adjust the A brainstorming strategy and The fusion weight between the fireworks strategies, For Complementary weights, To be the first A brainstorming strategy and The new strategy is obtained by weighted fusion of the fireworks strategies in proportion. is the strategy number index in the new control strategy set generated by the brainstorming optimization algorithm, The strategy number index in the fireworks strategy set generated by the fireworks search algorithm;

[0050] S52. Calculate the fitness of each control strategy in the entire set of candidate control strategies. The fitness function is constructed based on multiple performance indicators, including evaluation dimensions such as response speed, energy efficiency, control accuracy, and material fatigue. Each evaluation indicator has a corresponding weight coefficient, which is used to comprehensively calculate the fitness score of each strategy.

[0051] S53. Sort all control strategies from high to low according to their fitness scores, select the control strategies with the highest scores, and form an updated control strategy set.

[0052] Optionally, the S6 specifically includes:

[0053] S61. Obtain a current environment state vector according to the updated control strategy set, and match each control strategy in the updated control strategy set with the current environment state vector;

[0054] S62. Set a multi-objective fitness evaluation dimension, including M performance indicators such as response speed, energy efficiency, control accuracy, and material fatigue, and perform indicator mapping and response extraction for each control strategy;

[0055] S63. Calculate the fitness of each control strategy in the entire set of candidate control strategies and construct a multi-objective fitness function:

[0056]

[0057] in, The currently being processed performance evaluation indicators, is the time step for performance evaluation, is the strategy index number in the entire set of candidate control strategies, is the first in the entire set of candidate control strategies A control strategy, For the The comprehensive score value obtained by fitness evaluation of candidate control strategies, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The weight coefficient corresponding to each performance indicator is: For candidate strategies In terms of performance indicators, at time step The actual system response value, For the The performance indicators are at time step The expected target value, is the absolute deviation between the actual response value and the target value, To make a small shift to the target value, ∈ is a very small positive number, T is the time window step for the control performance observation, All sampling moments within the time window for evaluating the effect of the control strategy;

[0058] S64: Record the calculated fitness values ​​of each strategy and form a fitness result set.

[0059] Optionally, the S7 specifically includes:

[0060] S71. Select a target control strategy from the fitness result set, record it as the target strategy, input the target strategy into the functional coating activation unit, and drive the system to perform response regulation:

[0061]

[0062] in, is the strategy index number in the entire set of candidate control strategies, C * For the target control strategy, To find the fitness function Target strategy to achieve maximum value The complete set of candidate control strategies C all The A control strategy, C all is the complete set of candidate control strategies, For control strategy The comprehensive fitness score obtained after multi-objective performance evaluation under the current environment state vector;

[0063] S72. After the target strategy is executed, collect the actual response of the functional coating under the current environmental state vector and construct an actual response vector;

[0064] S73. Compare the expected response vector with the actual response vector;

[0065] Optionally, the S73 specifically includes:

[0066] S731. Extract the expected response vector corresponding to the target strategy, compare it with the actual response vector, and construct a feedback error function:

[0067]

[0068] in, For the current time The total feedback error under The currently being processed performance evaluation indicators, is the current time point, is the starting time point of target strategy execution, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The feedback error weight coefficient of each performance indicator, is the time τ at which the functional coating is applied to the The true response value of the performance index, at the same time τ, is the first The response value of each performance indicator, τ is the integral variable, is the integration operation of the squared error term over the entire strategy execution time interval, To take the square root of the integral result;

[0069] S732. Dynamically adjust the parameter structure of the current control strategy evaluation model based on the total feedback error, including indicator weight correction, error compensation model calibration, and adaptive update of prediction deviation. The corrected model is used as the basic input for control strategy optimization and fitness evaluation.

[0070] S733: Store the execution strategy, environment status, actual response vector, expected response vector and feedback error in the history record library.

[0071] The beneficial effects of the present invention are:

[0072] The present invention proposes a new type of intelligent control method for functional coatings based on real-time data fusion, which overcomes many problems in the existing technology, such as static control strategies, insufficient optimization capabilities, and imperfect feedback mechanisms, and has significant technical advantages and application effects. By fusing multi-source environmental perception data, the system can dynamically construct a high-dimensional environmental state vector, and realize unified modeling and processing of multiple influencing factors such as temperature, humidity, light, electric field, and gas concentration, which greatly improves the response accuracy and control basis of the coating system in complex environments. At the same time, by constructing a set of control variables for the adjustable properties of the coating and introducing a collaborative mechanism of the brainstorming optimization algorithm and the fireworks search algorithm, the system can not only realize a global breadth exploration of the control strategy, but also perform high-resolution local perturbation tuning near the optimal solution of the strategy, thereby effectively avoiding falling into the local optimum and improving the global optimal convergence ability of the control strategy.

[0073] In addition, the multi-objective fitness evaluation function constructed by the present invention can comprehensively consider multiple key indicators such as response speed, energy efficiency, control accuracy and material fatigue, ensuring that the control strategy is comprehensive and balanced in performance. After the strategy is executed, the system collects the deviation between the actual response result and the expected response, constructs a dynamic feedback error function, and introduces it into the subsequent strategy evaluation and model update process, forming a closed-loop intelligent control chain of "perception-optimization-execution-feedback-re-optimization". This feedback mechanism is based on the time series integral norm and multi-indicator weighted difference modeling, which can fully reflect the strategy control effect and drive the adaptive evolution of the model, further enhancing the robustness and continuous optimization capabilities of the system.

[0074] Overall, this invention not only enhances the real-time control capabilities of functional coatings in dynamic, multidimensional environments, but also achieves a technological leap from "fixed rules" to "self-learning evolution" in control strategies. It possesses high engineering practicality and scalability, making it applicable to multiple key technology areas, including smart materials, environmental response devices, and active protection systems. While promoting the intelligent development of functional coatings, this method also provides new insights and pathways for the integrated application of swarm intelligence algorithms in complex control scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0076] Figure 1 This is a flow chart of a novel functional coating intelligent control method based on real-time data fusion proposed by the present invention;

[0077] Figure 2 This is a schematic diagram of a novel functional coating intelligent control method based on real-time data fusion proposed by the present invention;

[0078] Figure 3 This is a data flow diagram of a new functional coating intelligent control method based on real-time data fusion proposed in the present invention. DETAILED DESCRIPTION

[0079] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0080] refer to Figure 1-3 , a new functional coating intelligent control method based on real-time data fusion, including the following steps:

[0081] S1. Collect multi-source environmental data in the functional coating deployment area, perform normalization processing, and construct a real-time environmental state vector;

[0082] S2, modeling the adjustable properties of the functional coating as a set of control variables;

[0083] S3. Construct an initial control strategy set based on the brainstorming optimization algorithm, cluster and group the control strategies in the set using the real-time environment state vector, and generate a new control strategy set through the disturbance generation mechanism;

[0084] S4. Select the target control strategy from the new control strategy set as the reference strategy, input it into the fireworks search algorithm, generate several local perturbation strategies around the reference strategy, and construct a fireworks strategy set;

[0085] S5. Merge the new control strategy set with the fireworks strategy set to form a complete set of candidate control strategies, sort the complete set of candidate control strategies according to a preset fitness function, and select the top several strategies to form an updated control strategy set;

[0086] S6. Construct a multi-objective fitness function and calculate the fitness of the updated control strategy set;

[0087] S7. Select a target update control strategy to execute a control operation, collect a corresponding actual response vector, and compare the actual response vector with the expected response vector.

[0088] The present invention proposes an intelligent control method that integrates brainstorming optimization and a fireworks search mechanism. By constructing a real-time environmental state vector, it accurately captures the dynamic environmental information of the area where the functional coating is located, and models the adjustable properties of the coating as a set of control variables to form a control strategy space. The system uses clustering grouping and heuristic strategy generation mechanisms to conduct in-depth exploration of the strategy space, and combines the optimal reference strategy to trigger a local perturbation search, significantly improving the global coverage and local accuracy of the control strategy. Through multi-objective fitness evaluation and strategy screening and sorting, inefficient strategies are effectively screened out, and ultimately a strategy adaptive system with a feedback closed loop is formed, which improves the response efficiency and environmental adaptability of the intelligent coating.

[0089] In this embodiment, S1 specifically includes:

[0090] S11. Collecting multi-source environmental data in the functional coating deployment area, wherein the multi-source environmental data includes temperature, humidity, light intensity, electric field intensity, and gas concentration;

[0091] S12. Time-align, remove noise, process outliers, and interpolate missing data for the collected multi-source environmental data to construct a standardized observation data structure under the same time base. Specifically, this includes aligning various sensor data according to timestamps to form a unified time series format, using a sliding mean interpolation method to complete missing data, using a median deviation detection method to remove or replace observation outliers, and then using dimensional conversion rules to unify the values ​​of different physical quantities in the multi-source environmental data;

[0092] S13. Calculate the standardized observation data and generate a real-time environment state vector through a weighted statistical mapping method.

[0093] This invention utilizes a multi-source environmental data fusion mechanism to construct a unified, standardized observation structure by time-aligning, eliminating anomalies, and completing missing values ​​for heterogeneous data such as temperature, humidity, illumination, and electric field intensity. It employs sliding mean, weighted median, and substitution interpolation methods to reconstruct the time series of different types of sensory data, effectively addressing the inconsistent spatiotemporal distribution and high missingness rates of multi-source data. The system uses weighted statistical mapping technology to convert multidimensional data into highly reliable real-time environmental state vectors, improving the accuracy of environmental modeling and the continuity and stability of system responses. This method is highly adaptable and suitable for intelligent sensory computing in complex scenarios.

[0094] In this embodiment, the S13 specifically includes:

[0095] S131. Perform multi-dimensional normalization and fusion calculation on the standardized observation data, and generate a real-time environment state vector through a weighted statistical mapping method:

[0096]

[0097] Among them, S(t) is the real-time environmental state vector at time t, Z is the normalization constant, K is the number of types of environmental parameters, M k is the number of sampling points under the kth type of environmental parameters, k is the category index of environmental data, t is the time index, m is the specific sampling point index in each type of environmental data, km is the index of the mth sampling point in the kth type of environmental data, M is the number of spatial sampling points, w km The fusion weight coefficient set for the m-th sampling point of the k-th class, x km (t) is the original observation value of the mth spatial sampling point in the kth type of multi-source environmental data at time t, σ km is the standard deviation of the m-th sampling point of the k-th category within the preset historical time window, μ km is the mean value of the m-th sampling point of the k-th category within the preset historical time window;

[0098] S132. Using the real-time environment state vector as the environment perception input based on which the control strategy is optimized.

[0099] This paper employs a multidimensional weighted normalization fusion calculation method to uniformly map environmental observation data from different categories and spatial sampling points into a moment-by-moment environmental state vector. By introducing historical statistical means and standard deviations to construct a dynamic normalization model, combined with sampling point weights and category weights, it accurately reflects the changing trends of environmental parameters within a specific time window. Using normalized offsets and a category reconciliation mechanism, a high-resolution real-time state representation is constructed, ensuring the stability and distinguishability of environmental feature inputs. This method is widely applicable to intelligent control systems driven by environmental perception, enhancing the real-time and targeted nature of control strategy generation.

[0100] In this embodiment, S2 specifically includes:

[0101] S21. Modeling the adjustable properties of the functional coating as a set of control variables, wherein the control variables include electrical excitation intensity, microstructure adjustment state, thermal response threshold, and surface response pattern, each control variable reflecting the response behavior of the coating under specific environmental conditions;

[0102] S22. Define a reasonable variation range for each control variable, where the variation range is constrained by a preset boundary value;

[0103] S23. Based on the control variables, the control strategy of the functional coating is expressed as a control variable vector:

[0104]

[0105] Where N is the total number of control variables, w i is the weight coefficient of the i-th control variable, i is the index of the control variable, t is the time index, c i (t) is the control value of the i-th control variable at time t, C(t) is the comprehensive control variable vector at time t, S i (t) is the part of the real-time environment state vector at time t that is related to the i-th control variable, is the weighted sum of all control variables from i=1 to i=N.

[0106] The present invention achieves unified modeling of multi-dimensional behaviors such as electrical excitation intensity, structural state, and thermal response by constructing a set of adjustable property control variables for the functional coating, and sets boundary constraints on the control variables to ensure safety and controllability. A weighted mapping method is adopted between the control variables and the environmental state vector, and a fusion formula is used to dynamically generate a control strategy vector to ensure that each control dimension is highly coupled with the actual environmental state. This method supports multivariable collaborative regulation and time series control optimization, enhances the coating system's ability to accurately respond to complex environmental stimuli, and achieves improved flexibility and real-time performance of the intelligent control strategy.

[0107] In this embodiment, S3 specifically includes:

[0108] S31. Construct an initial control strategy set based on a brainstorming optimization algorithm. The control strategy set consists of multiple candidate control strategies. Each candidate control strategy represents the response behavior of the coating under a specific environmental state. The construction formula of the initial control strategy set is:

[0109]

[0110] S32, according to the real-time environment state vector S (t) the initial control strategy set C BSO Perform clustering and grouping to group control strategies with similar response characteristics to reduce the complexity of the strategy space. i is the index of the control variable;

[0111] S33. Use the disturbance generation mechanism to perturb each control strategy and generate a new set of control strategies:

[0112]

[0113] Among them, C BSO' is the set of control strategies after disturbance generation, is the initial control strategy set C BSO The i-th control strategy individual, η is the disturbance gain coefficient, is the starting time point of the disturbance process, is the current time point, Θ i is the disturbance direction selection matrix of the i-th control strategy, and the i-th control variable is The local disturbance adjustment coefficient, is the intermediate time variable in the integration process, is the time from the start of the disturbance To the current moment The time integral of is the i-th control variable at time The disturbance response rate on is the i-th control variable at time The local disturbance adjustment coefficient.

[0114] The present invention uses a brainstorming optimization algorithm to construct an initial set of control strategies. A policy response structure is formed through a weighted mapping of control variables and environmental states. The system clusters and groups strategies based on the environmental state vector, significantly reducing the dimensionality of the policy space and improving search efficiency. During the policy perturbation generation process, a dynamic integral perturbation term and a time-domain gradient adjustment factor are introduced to construct a local perturbation mechanism that can perceive historical error changes, enabling flexible evolution of candidate strategies. This method effectively improves the diversity and local precision of control strategies, enhancing the system's responsiveness and intelligent control capabilities in high-dimensional dynamic environments.

[0115] In this embodiment, the S4 specifically includes:

[0116] S41, selecting a target control strategy from the new control strategy set as a reference control strategy, recorded as a center strategy;

[0117] S42. Input the center strategy into the fireworks search algorithm. Using this strategy as the explosion center, generate several local perturbation strategies in the local neighborhood to construct a fireworks strategy set. The number and perturbation range of the local perturbation strategies are:

[0118]

[0119] Among them, M spark is the number of spark strategies generated in fireworks search, κ is the upper limit parameter of strategy generation, Fitness(C center ) is the fitness value of the current reference control strategy, is the sum of the fitness of all strategies in the current control strategy set, is the total number of control strategies in the current control strategy set, R i is the local disturbance radius of the i-th control variable, ρ is the disturbance scale adjustment factor, is the maximum fitness value among all strategies, To traverse the index of the current control strategy set, is the kth control strategy in the set, C center is the center control strategy in the current round of fireworks search algorithm, and i is the index of the control variable;

[0120] S43. Based on the local disturbance generation mechanism, a weighted disturbance is applied to each control variable dimension. By introducing the normal distribution disturbance amount and disturbance direction coefficient, combined with the corresponding dimension disturbance radius, the central strategy is disturbed dimension by dimension to construct multiple local spark strategies.

[0121] S44. De-duplicate and correct the boundaries of all local spark strategies to ensure that they fall within the predefined control variable constraints to form a fireworks strategy set.

[0122] This method leverages the explosive perturbation mechanism of the fireworks search algorithm, dynamically generating a set of local perturbation strategies with the current optimal control strategy as the explosion center. By introducing a fitness weighting model and a perturbation radius calculation formula, the number of spark strategies and perturbation intensity are adaptively determined, ensuring more room for fine-tuning around high-quality strategies. This method combines strategy distribution with fitness gradient changes to establish a multidimensional perturbation trajectory, effectively improving strategy accuracy and local search capabilities. A boundary correction mechanism ensures that the generated strategies meet physical constraints, enhancing the algorithm's stability and adaptability in complex control spaces.

[0123] In this embodiment, the S5 specifically includes:

[0124] S51. The new control strategy set generated by the brainstorming optimization algorithm is combined with the fireworks strategy set generated by the fireworks search algorithm to generate a complete set of candidate control strategies:

[0125]

[0126] in, is the first in the fireworks strategy set generated by the fireworks search algorithm Spark strategy, C all is the complete set of candidate control strategies, For all brainstorming strategies With all fireworks perturbation strategies Cross-combination and weighted fusion are performed between The first of the new control strategies generated by the brainstorming optimization algorithm A control strategy, The first Spark strategies, N BSO' The number of new control strategies output by the brainstorming optimization algorithm, M spark The number of spark strategies generated for the fireworks search, To adjust the A brainstorming strategy and The fusion weight between the fireworks strategies, For Complementary weights, To be the first A brainstorming strategy and The new strategy is obtained by weighted fusion of the fireworks strategies in proportion. is the strategy number index in the new control strategy set generated by the brainstorming optimization algorithm, The strategy number index in the fireworks strategy set generated by the fireworks search algorithm;

[0127] S52. Calculate the fitness of each control strategy in the entire set of candidate control strategies. The fitness function is constructed based on multiple performance indicators, including evaluation dimensions such as response speed, energy efficiency, control accuracy, and material fatigue. Each evaluation indicator has a corresponding weight coefficient, which is used to comprehensively calculate the fitness score of each strategy.

[0128] S53. Sort all control strategies from high to low according to their fitness scores, select the control strategies with the highest scores, and form an updated control strategy set.

[0129] This paper combines brainstorming optimization with the fireworks search algorithm to propose a control strategy construction method based on a weighted fusion mechanism. By proportionally blending the two sets of strategies, a complete set of candidate control strategies is constructed, enhancing local precision control capabilities while maintaining strategy diversity. The system introduces a fitness function to comprehensively evaluate fused strategies and sorts and selects them based on multidimensional performance indicators, effectively improving the target matching and convergence efficiency of strategy selection. This method exhibits high reliability and global adaptability in high-dimensional control spaces, significantly enhancing the intelligent evolution and dynamic response capabilities of control systems.

[0130] In this embodiment, S6 specifically includes:

[0131] S61. Obtain a current environment state vector according to the updated control strategy set, and match each control strategy in the updated control strategy set with the current environment state vector;

[0132] S62. Set a multi-objective fitness evaluation dimension, including M performance indicators such as response speed, energy efficiency, control accuracy, and material fatigue, and perform indicator mapping and response extraction for each control strategy;

[0133] S63. Calculate the fitness of each control strategy in the entire set of candidate control strategies and construct a multi-objective fitness function:

[0134]

[0135] in, The currently being processed performance evaluation indicators, is the time step for performance evaluation, is the strategy index number in the entire set of candidate control strategies, is the first in the entire set of candidate control strategies A control strategy, For the The comprehensive score value obtained by fitness evaluation of candidate control strategies, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The weight coefficient corresponding to each performance indicator is: For candidate strategies In terms of performance indicators, at time step The actual system response value, For the The performance indicators are at time step The expected target value, is the absolute deviation between the actual response value and the target value, To make a small shift to the target value, ∈ is a very small positive number, T is the time window step for the control performance observation, All sampling moments within the time window for evaluating the effect of the control strategy;

[0136] S64: Record the calculated fitness values ​​of each strategy and form a fitness result set.

[0137] This method constructs a multi-objective fitness function to perform a weighted comprehensive evaluation of control strategies across multiple performance dimensions, including response speed, energy efficiency, and control accuracy. The system matches candidate strategies based on the current environmental state and uses the cumulative deviation within a time window as the evaluation basis. Normalization coefficients and perturbation terms are introduced to enhance the stability and resolution of the evaluation. This method effectively characterizes the comprehensive contribution of strategies to target performance, ensuring that strategy ranking is more valuable. The fitness evaluation results can provide a real-time decision-making basis for control optimization, enhancing the scientific nature of strategy screening and the system's adaptability.

[0138] In this embodiment, the S7 specifically includes:

[0139] S71. Select a target control strategy from the fitness result set, record it as the target strategy, input the target strategy into the functional coating activation unit, and drive the system to perform response regulation:

[0140]

[0141] in, is the strategy index number in the entire set of candidate control strategies, C * For the target control strategy, To find the fitness function Target strategy to achieve maximum value The complete set of candidate control strategies C all The A control strategy, C all is the complete set of candidate control strategies, For control strategy The comprehensive fitness score obtained after multi-objective performance evaluation under the current environment state vector;

[0142] S72. After the target strategy is executed, collect the actual response of the functional coating under the current environmental state vector and construct an actual response vector;

[0143] S73. Compare the expected response vector with the actual response vector;

[0144] Based on the results of a multi-objective fitness evaluation, the present invention automatically selects the optimal control strategy from candidate strategies as the system's activation instruction, achieving intelligent response control of the functional coating. By maximizing the fitness function value, the system can accurately identify the control scheme that best suits the current environmental state in the high-dimensional strategy space. The actual response position is further compared with the expected response position, and a closed-loop evaluation mechanism is constructed to ensure that the control effect remains consistent with the system prediction. This method improves the scientific nature and accuracy of strategy execution, significantly enhancing the system's dynamic response efficiency and intelligent judgment capabilities.

[0145] In this embodiment, the S73 specifically includes:

[0146] S731. Extract the expected response vector corresponding to the target strategy, compare it with the actual response vector, and construct a feedback error function:

[0147]

[0148] in, For the current time The total feedback error under The currently being processed performance evaluation indicators, is the current time point, is the starting time point of target strategy execution, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The feedback error weight coefficient of each performance indicator, is the time τ at which the functional coating is applied to the The true response value of the performance index, at the same time τ, is the first The response value of each performance indicator, τ is the integral variable, is the integration operation of the squared error term over the entire strategy execution time interval, To take the square root of the integral result;

[0149] S732. Dynamically adjust the parameter structure of the current control strategy evaluation model based on the total feedback error, including indicator weight correction, error compensation model calibration, and adaptive update of prediction deviation. The corrected model is used as the basic input for control strategy optimization and fitness evaluation.

[0150] S733: Store the execution strategy, environment status, actual response vector, expected response vector and feedback error in the history record library.

[0151] This paper constructs a feedback error function based on the integral norm, comprehensively comparing the control strategy's predicted response with the actual system response over time, and quantifying the dynamic deviations between multidimensional indicators. By introducing a weighted integral error modeling approach, the system can accurately identify the degree of performance deviation of the control strategy during execution and adjust the model parameters and fitness evaluation structure accordingly in real time. Error information serves as an iterative input to achieve self-repair and evolutionary optimization of the control model, effectively enhancing the system's long-term robustness and adaptability, and improving the strategy's generalization and response quality in complex environments.

[0152] Example 1:

[0153] To verify the feasibility of the present invention, it was applied to a new high-end intelligent building exterior wall protection system in a coastal city. This region is subject to complex environmental factors such as high humidity, strong ultraviolet radiation, and salt spray corrosion year-round, placing extremely high demands on the stability, responsiveness, and intelligent adjustment capabilities of the functional coatings used on building exterior walls. Traditionally used nano-thermochromic coatings or waterproof oleophobic coatings often fail under high-temperature exposure or continuous humidity due to their lack of dynamic control capabilities. They also have long response times to environmental changes, high energy consumption, and significant functional degradation after long-term use, seriously affecting the building's appearance and structural safety.

[0154] Using the novel intelligent control method for functional coatings based on real-time data fusion described in this invention, an integrated multifunctional intelligent coating control system was deployed across the entire building facade. During the deployment phase, the system uses an externally integrated environmental monitoring unit to collect real-time data on five key environmental parameters: temperature, relative humidity, ultraviolet radiation intensity, atmospheric electric field strength, and near-surface salt spray concentration. The system updates its state vector every 10 seconds, achieving high-frequency data fusion. The collected data is normalized and mapped into a five-dimensional environmental state space defined within the system to construct the current environmental state vector.

[0155] At the same time, the adjustable properties of the multifunctional composite coating used on the building's exterior walls are modeled as control variables, including five main adjustment dimensions: coating micropore opening rate, molecular orientation structure, electric field response parameters, light reflection factor, and energy conversion threshold. Control variables are represented by candidate strategies in the control strategy set. The system first clusters the initial control strategies based on the current state vector using a brainstorming optimization algorithm, and then expands the strategy search space through a perturbation generation mechanism. Subsequently, the strategy with the highest fitness in the current cluster is used as the central strategy, and a fireworks search algorithm is used to generate several perturbation strategies within its local neighborhood, thereby improving local search accuracy and optimizing parameter matching.

[0156] During the strategy fusion process, the system combines the two sets of strategies using a weighted combination to construct a complete set of candidate control strategies. After fusion, the system evaluates each strategy based on the constructed multi-objective fitness function, scoring it based on four dimensions: response speed, energy efficiency, control accuracy, and material fatigue. The optimal strategy is then selected based on fitness ranking. The selected strategy is executed by an embedded drive unit, activating the coating material for an actual response. The system then collects the actual coating response data and compares it with the strategy's predicted data. The feedback error vector is constructed through an integral form to further optimize the control model and achieve closed-loop self-learning.

[0157] The actual deployment took place from June to August 2024, on the south side of the Binhai Experimental Building in Siming District, Xiamen. The test target was a three-story exterior wall on the east facade, covering approximately 1,600 square meters. Environmental data was collected 24 hours a day for 60 consecutive days, and comparative analysis was conducted with traditional static control methods before, during, and after the system's operation.

[0158] Table 1 Comparison of optimization effects of new functional coating intelligent control methods based on real-time data fusion

[0159]

[0160] Table 1 shows that under various typical climate conditions, the method described in this invention demonstrates a response speed approximately 60% to 70% faster than conventional methods, reduces unit energy consumption by over 50%, improves response accuracy by an average of over 10%, and significantly enhances material performance stability, with the 30-day fatigue decay rate reduced to less than half. These data demonstrate that the method described in this invention offers significant advantages in response control capabilities, energy efficiency management, and long-term durability in dynamic environments, fully meeting the high-performance control requirements for intelligent functional coatings in complex urban environments.

[0161] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A novel intelligent control method for functional coatings based on real-time data fusion, characterized in that: The steps include: S1. Collect multi-source environmental data in the functional coating deployment area, perform normalization processing, and construct a real-time environmental state vector; S2, modeling the adjustable properties of the functional coating as a set of control variables; S3. Construct an initial control strategy set based on the brainstorming optimization algorithm, cluster and group the control strategies in the set using the real-time environment state vector, and generate a new control strategy set through the disturbance generation mechanism; S4. Select the target control strategy from the new control strategy set as the reference strategy, input it into the fireworks search algorithm, generate several local perturbation strategies around the reference strategy, and construct a fireworks strategy set; S5. Merge the new control strategy set with the fireworks strategy set to form a complete set of candidate control strategies, sort the complete set of candidate control strategies according to a preset fitness function, and select the top several strategies to form an updated control strategy set; S6. Construct a multi-objective fitness function and calculate the fitness of the updated control strategy set; S7. Select a target update control strategy to execute a control operation, collect a corresponding actual response vector, and compare the actual response vector with the expected response vector.

2. A novel functional coating intelligent control method based on real-time data fusion according to claim 1, characterized in that: Said S1 specifically includes: S11. Collecting multi-source environmental data in the functional coating deployment area, wherein the multi-source environmental data includes temperature, humidity, light intensity, electric field intensity, and gas concentration; S12. Time-align, remove noise, process outliers, and interpolate missing data for the collected multi-source environmental data to construct a standardized observation data structure under the same time base. Specifically, this includes aligning various sensor data according to timestamps to form a unified time series format, using a sliding mean interpolation method to complete missing data, using a median deviation detection method to remove or replace observation outliers, and then using dimensional conversion rules to unify the values ​​of different physical quantities in the multi-source environmental data; S13. Calculate the standardized observation data and generate a real-time environment state vector through a weighted statistical mapping method.

3. A novel functional coating intelligent control method based on real-time data fusion according to claim 2, characterized in that: The S13 specifically includes: S131. Perform multi-dimensional normalization and fusion calculation on the standardized observation data, and generate a real-time environment state vector through a weighted statistical mapping method: Among them, S(t) is the real-time environmental state vector at time t, Z is the normalization constant, K is the number of types of environmental parameters, M k is the number of sampling points under the kth type of environmental parameters, k is the category index of environmental data, t is the time index, m is the specific sampling point index in each type of environmental data, km is the index of the mth sampling point in the kth type of environmental data, M is the number of spatial sampling points, w km The fusion weight coefficient set for the m-th sampling point of the k-th class, x km (t) is the original observation value of the mth spatial sampling point in the kth type of multi-source environmental data at time t, σ km is the standard deviation of the m-th sampling point of the k-th category within the preset historical time window, μ km is the mean value of the m-th sampling point of the k-th category within the preset historical time window; S132. Using the real-time environment state vector as the environment perception input based on which the control strategy is optimized.

4. A novel functional coating intelligent control method based on real-time data fusion according to claim 1, characterized in that: The S2 specifically includes: S21. Modeling the adjustable properties of the functional coating as a set of control variables, wherein the control variables include electrical excitation intensity, microstructure adjustment state, thermal response threshold, and surface response pattern, each control variable reflecting the response behavior of the coating under specific environmental conditions; S22. Define a reasonable variation range for each control variable, where the variation range is constrained by a preset boundary value; S23. Based on the control variables, the control strategy of the functional coating is expressed as a control variable vector: Where N is the total number of control variables, w i is the weight coefficient of the i-th control variable, i is the index of the control variable, t is the time index, c i (t) is the control value of the i-th control variable at time t, C(t) is the comprehensive control variable vector at time t, S i (t) is the part of the real-time environment state vector at time t that is related to the i-th control variable, is the weighted sum of all control variables from i=1 to i=N.

5. The novel functional coating intelligent control method based on real-time data fusion according to claim 1 is characterized in that: The S3 specifically includes: S31. Construct an initial control strategy set based on a brainstorming optimization algorithm. The control strategy set consists of multiple candidate control strategies. Each candidate control strategy represents the response behavior of the coating under a specific environmental state. The construction formula of the initial control strategy set is: S32, according to the real-time environment state vector S (t) the initial control strategy set C BSO Perform clustering and grouping to group control strategies with similar response characteristics to reduce the complexity of the strategy space. i is the index of the control variable; S33. Use the disturbance generation mechanism to perturb each control strategy and generate a new set of control strategies: Among them, C BSO' is the set of control strategies after disturbance generation, is the initial control strategy set C BSO The i-th control strategy individual, η is the disturbance gain coefficient, is the starting time point of the disturbance process, is the current time point, Θ i is the disturbance direction selection matrix of the i-th control strategy, and the i-th control variable is The local disturbance adjustment coefficient, is the intermediate time variable in the integration process, is the time from the start of the disturbance To the current moment The time integral of is the i-th control variable at time The disturbance response rate on is the i-th control variable at time The local disturbance adjustment coefficient.

6. A novel functional coating intelligent control method based on real-time data fusion according to claim 1, characterized in that: The S4 specifically includes: S41, selecting a target control strategy from the new control strategy set as a reference control strategy, recorded as a center strategy; S42. Input the center strategy into the fireworks search algorithm. Using this strategy as the explosion center, generate several local perturbation strategies in the local neighborhood to construct a fireworks strategy set. The number and perturbation range of the local perturbation strategies are: Among them, M spark is the number of spark strategies generated in fireworks search, κ is the upper limit parameter of strategy generation, Fitness(C center ) is the fitness value of the current reference control strategy, is the sum of the fitness of all strategies in the current control strategy set, is the total number of control strategies in the current control strategy set, R i is the local disturbance radius of the i-th control variable, ρ is the disturbance scale adjustment factor, is the maximum fitness value among all strategies, To traverse the index of the current control strategy set, is the kth control strategy in the set, C center is the center control strategy in the current round of fireworks search algorithm, and i is the index of the control variable; S43. Based on the local disturbance generation mechanism, a weighted disturbance is applied to each control variable dimension. By introducing the normal distribution disturbance amount and disturbance direction coefficient, combined with the corresponding dimension disturbance radius, the central strategy is disturbed dimension by dimension to construct multiple local spark strategies. S44. De-duplicate and correct the boundaries of all local spark strategies to ensure that they fall within the predefined control variable constraints to form a fireworks strategy set.

7. The novel functional coating intelligent control method based on real-time data fusion according to claim 1 is characterized in that: The S5 specifically includes: S51. The new control strategy set generated by the brainstorming optimization algorithm is combined with the fireworks strategy set generated by the fireworks search algorithm to generate a complete set of candidate control strategies: in, is the first in the fireworks strategy set generated by the fireworks search algorithm Spark strategy, C all is the complete set of candidate control strategies, For all brainstorming strategies With all fireworks perturbation strategies Cross-combination and weighted fusion are performed between The first of the new control strategies generated by the brainstorming optimization algorithm A control strategy, The first Spark strategies, N BSO' The number of new control strategies output by the brainstorming optimization algorithm, M spark The number of spark strategies generated for the fireworks search, To adjust the A brainstorming strategy and The fusion weight between the fireworks strategies, For Complementary weights, To be the first A brainstorming strategy and The new strategy is obtained by weighted fusion of the fireworks strategies in proportion. is the strategy number index in the new control strategy set generated by the brainstorming optimization algorithm, The strategy number index in the fireworks strategy set generated by the fireworks search algorithm; S52. Calculate the fitness of each control strategy in the entire set of candidate control strategies. The fitness function is constructed based on multiple performance indicators, including evaluation dimensions such as response speed, energy efficiency, control accuracy, and material fatigue. Each evaluation indicator has a corresponding weight coefficient, which is used to comprehensively calculate the fitness score of each strategy. S53. Sort all control strategies from high to low according to their fitness scores, select the control strategies with the highest scores, and form an updated control strategy set.

8. The novel functional coating intelligent control method based on real-time data fusion according to claim 1 is characterized in that: The S6 specifically includes: S61. Obtain a current environment state vector according to the updated control strategy set, and match each control strategy in the updated control strategy set with the current environment state vector; S62. Set a multi-objective fitness evaluation dimension, including M performance indicators such as response speed, energy efficiency, control accuracy, and material fatigue, and perform indicator mapping and response extraction for each control strategy; S63. Calculate the fitness of each control strategy in the entire set of candidate control strategies and construct a multi-objective fitness function: in, The currently being processed performance evaluation indicators, t is the time step of performance evaluation, is the strategy index number in the entire set of candidate control strategies, is the first in the entire set of candidate control strategies A control strategy, For the The comprehensive score value obtained by fitness evaluation of candidate control strategies, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The weight coefficient corresponding to each performance indicator is: For candidate strategies In terms of performance indicators, the actual system response value at time step t, Z k (t) is the The expected target value of the performance indicator at time step t, is the absolute deviation between the actual response value and the target value, Z k (t)+∈ is a small offset to the target value, ∈ is a very small positive number, T is the time window step for control performance observation, and t is all sampling moments in the time window for evaluating the effect of the control strategy; S64: Record the calculated fitness values ​​of each strategy and form a fitness result set.

9. The novel functional coating intelligent control method based on real-time data fusion according to claim 1, characterized in that: The S7 specifically includes: S71. Select a target control strategy from the fitness result set, record it as the target strategy, input the target strategy into the functional coating activation unit, and drive the system to perform response regulation: in, is the strategy index number in the entire set of candidate control strategies, C * For the target control strategy, To find the fitness function Target strategy to achieve maximum value The complete set of candidate control strategies C all The A control strategy, C all is the complete set of candidate control strategies, For control strategy The comprehensive fitness score obtained after multi-objective performance evaluation under the current environment state vector; S72. After the target strategy is executed, collect the actual response of the functional coating under the current environmental state vector and construct an actual response vector; S73. Compare the expected response vector with the actual response vector.

10. A novel functional coating intelligent control method based on real-time data fusion according to claim 9, characterized in that: The S73 specifically includes: S731. Extract the expected response vector corresponding to the target strategy, compare it with the actual response vector, and construct a feedback error function: in, For the current time The total feedback error under The currently being processed performance evaluation indicators, is the current time point, is the starting time point of target strategy execution, M is the total number of performance indicators considered in the multi-objective fitness evaluation, For the The feedback error weight coefficient of each performance indicator, is the time τ at which the functional coating is applied to the The true response value of the performance index, at the same time τ, is the first The response value of each performance indicator, τ is the integral variable, is the integration operation of the squared error term over the entire strategy execution time interval, To take the square root of the integral result; S732. Dynamically adjust the parameter structure of the current control strategy evaluation model based on the total feedback error, including indicator weight correction, error compensation model calibration, and adaptive update of prediction deviation. The corrected model is used as the basic input for control strategy optimization and fitness evaluation. S733: Store the execution strategy, environment status, actual response vector, expected response vector and feedback error in the history record library.

Images