Light wall spraying process based on artificial intelligence

Abstract

The invention provides a light wall spraying process based on artificial intelligence, which comprises the following steps of: acquiring key parameters such as spraying thickness, uniformity, material consumption, spraying pressure, environment temperature and humidity and the like in a spraying process in real time through mounted sensor equipment; the data processing module is used for transmitting the data to the data processing module for data cleaning, denoising and format conversion, the spraying process is divided into a plurality of sub-regions, a data set in each sub-region is processed through a self-adaptive distributed generative adversarial network model, the pre-processed data is analyzed by utilizing a self-adaptive distributed generative adversarial network algorithm, and the data is stored in the data processing module. A prediction model of the spraying process is constructed, the prediction model can predict the spraying effect in real time according to the input parameters, feedback signals are generated, and operation parameters of spraying equipment are adjusted through an intelligent optimization algorithm; and closed-loop control is formed, so that the spraying quality is continuously optimized in the construction process, the material waste is reduced, and the construction efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of building construction, and in particular to a lightweight wall spraying process based on artificial intelligence. Background Art

[0002] Lightweight wall spraying technology is an important technology in modern building construction and is widely used in wall decoration and protection. However, the traditional lightweight wall spraying technology mainly relies on manual experience and preset parameters, and it is difficult to optimize and dynamically adjust in real time according to actual conditions during the construction process. The limitations of this process method lead to problems such as unstable spraying quality, material waste and low efficiency, which restricts the overall quality and efficiency of building construction.

[0003] In the prior art, the parameters of the spraying process are usually set based on experimental data before construction and the experience of operators. These parameters may need to be adjusted during the actual construction process due to environmental conditions, equipment status, and changes in wall materials. However, manual adjustment of parameters is not only time-consuming and laborious, but also difficult to ensure consistency and accuracy of the adjustment effect. Especially in large-area spraying or complex structure wall construction, the limitations of manual adjustment are more obvious, and it is difficult to maintain stable spraying quality.

[0004] The traditional spraying process also has the problem of being highly dependent on manual operation. The operation of spraying equipment is usually completed by skilled workers. The workers' operating skills and experience directly affect the spraying quality. However, the operating levels of different workers vary. Even the operations of the same worker at different times may be different, resulting in inconsistent spraying quality. In addition, manual operation is difficult to maintain high efficiency in a complex and changeable construction environment, resulting in low spraying efficiency and high labor costs.

[0005] In addition, the traditional spraying process lacks the ability to monitor and analyze data in real time during the spraying process. During the construction process, it is impossible to obtain and analyze key parameters such as spraying thickness, uniformity, material consumption, spraying pressure, and ambient temperature and humidity in real time, and it is impossible to adjust the operating parameters of the spraying equipment in time. Due to the lack of real-time data support, the spraying quality is difficult to guarantee, material waste is serious, and construction efficiency is low.

[0006] In response to the above problems, there have been some attempts in the prior art to use automated equipment and basic control systems to improve the spraying process. However, these systems can often only achieve simple execution of preset parameters and lack the ability of intelligent optimization and dynamic adjustment. Although the automated system has improved the spraying efficiency to a certain extent, due to the inability to analyze and adjust parameters in real time, its spraying quality and consistency are still difficult to reach the ideal level.

[0007] In addition, most of the spray process optimization methods in the existing technology are based on traditional optimization algorithms, which are inefficient when dealing with complex and multi-variable spray parameters and difficult to achieve real-time optimization. For example, simple rule-based optimization methods cannot cope with the dynamic changes in the spray process, and the optimization effect is not ideal. Although some studies have tried to use machine learning and intelligent algorithms to optimize the spray process, these methods are usually limited to offline data analysis, lack real-time application capabilities, and fail to make full use of real-time data for dynamic optimization.

[0008] In short, the traditional lightweight wall spraying process has obvious defects in real-time optimization, dynamic adjustment, degree of automation and consistency, which seriously restricts the improvement of spraying quality and construction efficiency; for this reason, the applicant proposes an optimization method for lightweight wall spraying process based on artificial intelligence according to the spraying needs. Through intelligent optimization algorithms, real-time data analysis and feedback, integrated automation systems and other means, real-time optimization and dynamic adjustment of the spraying process are achieved, the spraying quality and efficiency are improved, material waste and labor costs are reduced, and it is possible to adapt to complex and changeable construction environments. Summary of the invention

[0009] In order to solve the above technical problems, the present invention proposes a lightweight wall spraying process based on artificial intelligence. Through the installed sensor equipment, key parameters such as spraying thickness, uniformity, material consumption, spraying pressure and ambient temperature and humidity in the spraying process are collected in real time, and these data are transmitted to the data processing module for data cleaning, denoising and format conversion. The spraying process is divided into multiple sub-areas, and the data set in each sub-area is processed by an adaptive distributed generative adversarial network model. The preprocessed data is analyzed using an adaptive distributed generative adversarial network algorithm to construct a prediction model for the spraying process. The prediction model can predict the spraying effect in real time according to the input parameters, generate feedback signals, and adjust the operating parameters of the spraying equipment through an intelligent optimization algorithm; a closed-loop control is formed, so that the spraying quality is continuously optimized during the construction process, material waste is reduced, and construction efficiency is improved.

[0010] To achieve the above object, the technical solution adopted by the present invention is:

[0011] A lightweight wall spraying process based on artificial intelligence, characterized in that: the lightweight wall spraying process based on artificial intelligence can be optimized and dynamically adjusted in real time, greatly improving the spraying quality and construction efficiency, and the specific steps are:

[0012] S1. Use the sensor equipment installed on the spraying equipment to collect the spraying thickness, uniformity, material consumption, spraying pressure and ambient temperature and humidity in real time during the spraying process to build a wall spraying parameter data set;

[0013] S2, transmitting the collected wall spraying parameter data set to the data processing module for data cleaning, denoising and format conversion;

[0014] S3. Use an adaptive distributed generative adversarial network algorithm to analyze the preprocessed wall spraying parameter data set and build a prediction model for the spraying process. The prediction model predicts the spraying effect based on the input parameters of the wall spraying parameter data set.

[0015] S4. According to the output results of the prediction model, an intelligent optimization algorithm based on a combination of a genetic algorithm and a particle swarm optimization algorithm is used to adjust the operating parameters of the spraying equipment, wherein the operating parameters include spraying pressure, nozzle speed, and material flow rate;

[0016] S5. Feedback the optimal spraying parameter combination to the spraying equipment, control the operation of the spraying equipment in real time, and dynamically optimize the spraying process according to the actual situation;

[0017] S6. Repeat S1-S5 to form a closed-loop control to ensure continuous optimization of the spraying process;

[0018] S7. Monitor the execution of the spraying process in real time, and record and store key data during the spraying process.

[0019] Furthermore, the specific method of step S1 in the artificial intelligence-based lightweight wall spraying process is as follows:

[0020] S11. Install a laser thickness sensor on the spraying equipment to monitor the spraying thickness in real time, an image processing sensor to monitor the spraying uniformity, a flow sensor to monitor the material consumption, a resistance pressure sensor to monitor the spraying pressure, and a temperature and humidity sensor to monitor the ambient temperature and humidity;

[0021] S12, laser thickness sensor collects spraying thickness data T at different positions on the wall surface during spraying in real time i , where i represents the time point of acquisition;

[0022] S13, image processing sensor real-time acquisition of uniformity data U of the wall surface during spraying i ;

[0023] S14, flow sensor monitors the material consumption in real time during spraying C i ;

[0024] S15, resistance pressure sensor real-time monitoring of the spraying pressure P of the spraying equipment during the spraying process i ;

[0025] S16, temperature and humidity sensor monitors the temperature T of the spraying environment in real time env，i and humidity H env，i ;

[0026] S17, sorting and summarizing all the collected spraying thickness data, uniformity data, material consumption data, spraying pressure data, ambient temperature data and ambient humidity data to construct a wall spraying parameter data set D:

[0027] D={(T i ,U i ,C i ,P i ,T env,i ,H env，i )|i=1,2,...,n}

[0028] Where n represents the total number of time points for data collection.

[0029] Furthermore, the specific method of step S2 in the artificial intelligence-based lightweight wall spraying process is as follows:

[0030] S21, transmitting the wall spraying parameter data set D constructed by step S17 to the data processing module;

[0031] S22, performing data cleaning on the wall spraying parameter data set D transmitted to the data processing module, removing abnormal values ​​and missing values, and obtaining a cleaned wall spraying parameter data set;

[0032] S23, performing denoising on the cleaned wall spraying parameter data set, using a filtering algorithm to eliminate noise interference in the collected data, and obtaining a denoised wall spraying parameter data set;

[0033] S24, converting the format of the wall spraying parameter data set after denoising to the input format required by the prediction model, and obtaining the wall spraying parameter data set D after format conversion formatted .

[0034] Furthermore, the specific method of step S3 in the artificial intelligence-based lightweight wall spraying process is as follows:

[0035] S31, the wall spraying parameter data set D after the format conversion in step S24 formatted Input into the adaptive distributed generative adversarial network model;

[0036] S32, the adaptive distributed generative adversarial network model includes multiple generators G k and the discriminator D k , each generator and discriminator corresponds to a spraying sub-area, and k is the sub-area number;

[0037] S33, dividing the spraying process into multiple sub-areas, and the wall spraying parameter data set D in each sub-area formatted，k ;

[0038]

[0039] Where: m is the number of sub-regions;

[0040] S34, generator G for each sub-region k According to the input wall spraying parameter data set D formatted，k Generate the simulated spraying effect data G of the sub-area k (D foratted，k );

[0041]

[0042] in: Indicates the generated spray thickness data, Represents the generated spray uniformity data, Represents the generated material consumption data, Indicates the generated spray pressure data, Represents the generated ambient temperature data, Represents the generated environmental humidity data, n k represents the total number of data points in sub-region k;

[0043] S35, discriminator D for each sub-region k According to the wall spraying parameter data set D frmatted，k And the simulated data G generated by the generator k (F formatted，k ) to make a distinction and output the distinction result;

[0044] D k (D formatted,，k , G k (D formatted，k ))

[0045] Discriminator D k The output result represents the difference between the generated data and the real data;

[0046] S36, based on each sub-region discriminator D k The feedback result of the corresponding generator G k Perform iterative optimization to update the generator G by minimizing the difference between the generated data and the real data k Parameters of the generator are gradually improved to improve the accuracy of the generated data.

[0047] S37. Add the spraying thickness deviation ΔT and the material consumption error ΔC to the loss function of the generator, where:

[0048]

[0049] Where: Dreal，k is the real wall spraying parameter dataset, λ 1 and λ 2 are weight parameters, which respectively adjust the influence of spraying thickness deviation and material consumption error on the generator;

[0050] S38, adding the spraying uniformity error ΔU and the spraying pressure error ΔP to the loss function of the discriminator, where:

[0051]

[0052] Where: 3 and λ 4 are weight parameters, which respectively adjust the influence of spraying uniformity error and spraying pressure error on the discriminator;

[0053] S39. Introduce a multi-scale convolutional layer into the discriminator network structure to extract and fuse features of different scales to optimize the accuracy of spraying effect discrimination. After completing multiple iterations, the optimized generator G is obtained. k , each generator can be based on the input wall spraying parameter dataset D formatted，k Generate spray effect simulation data and set the generator G of all sub-areas k The generated simulation data sets are integrated to construct the prediction model M(D) of the overall spraying process. formatted ).

[0054] Furthermore, the specific method of step S4 in the artificial intelligence-based lightweight wall spraying process is as follows:

[0055] S41, obtain the spraying effect prediction model M (D generated by step S37 formatted ), the output results comprising predicted values ​​of spraying pressure, nozzle speed and material flow rate;

[0056] S42, defining an objective function of the spraying process optimization problem, with minimizing spraying thickness deviation, material consumption error and spraying uniformity error as the optimization goal;

[0057] S43, using a genetic algorithm to initialize a set of candidate solutions, each candidate solution represents a set of spraying operation parameters, including spraying pressure P, nozzle speed V and material flow rate Q;

[0058] S44, in the selection, crossover and mutation operations of the genetic algorithm, each generation selects candidate solutions with higher fitness to perform crossover and mutation to generate new candidate solutions, and through multiple generations of iterations, the spraying operation parameters are gradually optimized;

[0059] S45, further optimizing the candidate solutions generated by the genetic algorithm using a particle swarm optimization algorithm, where each particle represents a set of spraying operation parameters and has a speed and a position, and the optimal solution is found by updating the speed and position;

[0060] S46, searching for the optimal combination of spraying operation parameters through an intelligent optimization algorithm combining a genetic algorithm and a particle swarm optimization algorithm to optimize the spraying process;

[0061]

[0062] S47, combine the optimal spraying operation parameters x * =(P * , V * , Q * ) is fed back to the spraying equipment to adjust the operating parameters of the spraying equipment in real time, so that the spraying process can be dynamically optimized according to the actual situation.

[0063] Furthermore, the specific method of step S39 in step S3 of the artificial intelligence-based lightweight wall spraying process is as follows:

[0064] S391, in the discriminator D k A multi-scale convolution layer is introduced into the network structure, wherein the multi-scale convolution layer includes convolution kernels of different sizes, and the multi-scale features of spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters are extracted and integrated respectively, and the features of spraying thickness and uniformity are processed;

[0065] MSC(x)={Conv 3×3 (x), Conv 5×5 (x), Conv 7×7 (x)}

[0066]

[0067] Where: x is the input spraying parameter data, including spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters;

[0068] Conv k×k (x) represents a convolution operation with a convolution kernel size of k×k;

[0069] and are the weight and bias parameters of the convolution kernel;

[0070] σ is the activation function;

[0071] S392, fusing the features extracted by the multi-scale convolutional layer to form a comprehensive feature map F, so as to optimize the accuracy of the discriminator in judging the spraying effect, and especially enhance the judgment of the uniformity of the spraying thickness and the accuracy of the material consumption;

[0072] F = Concat(MSC 3×3 (x),MSC 5×5 (x), MSC 7×7 (x)

[0073] Among them: Concat represents the concatenation operation of convolution features of different scales;

[0074] MSC 3×3 (x), MSC 5×5 (x), MSC 7×7 (x) represents multi-scale convolution operations using 3x3, 5x5, and 7x7 convolution kernels respectively;

[0075] S393, discriminate the fused feature map and output the discrimination result for evaluating the generator G k The difference between the generated simulated spraying effect data and the real data;

[0076] D k (F) = σ(W·F+b)

[0077] Where: W and b are the weight and bias parameters of the discriminator;

[0078] σ is the activation function;

[0079] S394, introducing specific constraints for the spraying process into the loss function of the discriminator, including spraying thickness deviation, material consumption error, spraying uniformity error and spraying pressure error;

[0080]

[0081] S395, based on discriminator D k The feedback result of the generator G k Perform multiple iterative optimizations to update the generator G by minimizing the difference between the generated data and the real data. k Parameters;

[0082]

[0083] in, To represent the expected value calculation, the generator gradually optimizes the data it generates by minimizing the loss function fed back by the discriminator;

[0084] S396. After completing multiple iterations, the optimized generator G is obtained. k,The generator can generate spraying effect simulation data based on the input spraying parameter data set;

[0085]

[0086] in: To represent the optimized generator;

[0087] S397, using the optimized generator The generated simulation data set integrates the generator simulation data of multiple sub-areas to form a prediction model of the overall spraying process, which is used to predict the spraying effect in real time and guide the optimization of the spraying process.

[0088]

[0089] Where: M(D formatted ) is a prediction model for the overall spraying process. By integrating the generator simulation data of multiple sub-areas, an overall prediction model is formed to predict the spraying effect in real time and optimize the spraying process.

[0090] The benefits of this application are:

[0091] 1. The lightweight wall spraying process based on artificial intelligence collects key parameters such as spraying thickness, uniformity, material consumption, spraying pressure, and ambient temperature and humidity in real time through the sensor equipment installed on the spraying equipment, and transmits these data to the data processing module for data cleaning, denoising and format conversion. The adaptive distributed generative adversarial network algorithm is used to analyze the pre-processed data and build a prediction model for the spraying process. The prediction model can predict the spraying effect in real time according to the input parameters, generate feedback signals, and adjust the operating parameters of the spraying equipment through intelligent optimization algorithms. A closed-loop control is formed to continuously optimize the spraying quality during the construction process, reduce material waste, and improve construction efficiency;

[0092] 2. The lightweight wall spraying process based on artificial intelligence introduces a multi-scale convolution layer in the network structure of the discriminator, and uses convolution kernels of different sizes to extract and fuse the multi-scale features of spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters. The multi-scale convolution layer can capture the detailed features of the spraying effect more comprehensively, improve the accuracy of the discriminator in judging the spraying effect, especially the judgment of the uniformity of the spraying thickness and material consumption. It further optimizes the performance of the generator and ensures that the generated simulated spraying effect data is highly close to the real data;

[0093] 3. The lightweight wall spraying process based on artificial intelligence divides the spraying process into multiple sub-areas. The data set in each sub-area is processed by an adaptive distributed generative adversarial network model. Each generator and discriminator corresponds to a spraying sub-area. The generator generates simulated spraying effect data for the sub-area based on the input data set, and the discriminator judges the difference between the generated data and the real data. The accuracy of the generated data is gradually improved through iterative optimization of the sub-area generator and discriminator. Finally, the generator simulation data of all sub-areas are integrated to form a prediction model for the overall spraying process. The prediction model can predict the spraying effect in real time, provide more accurate parameter adjustment guidance, effectively adapt to the complex and changeable construction environment, and improve the robustness and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 It is a schematic diagram of the engineering process of the present invention;

[0095] Figure 2 This is a schematic diagram of the intelligent optimization algorithm module of the present invention. DETAILED DESCRIPTION

[0096] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments:

[0097] like Figure 1-2 As shown, a lightweight wall spraying process based on artificial intelligence is shown. The lightweight wall spraying process based on artificial intelligence can be optimized and adjusted dynamically in real time, which greatly improves the spraying quality and construction efficiency. The specific steps are as follows:

[0098] S1. Use the sensor equipment installed on the spraying equipment to collect the spraying thickness, uniformity, material consumption, spraying pressure and ambient temperature and humidity in real time during the spraying process to build a wall spraying parameter data set;

[0099] S2, transmitting the collected wall spraying parameter data set to the data processing module for data cleaning, denoising and format conversion;

[0100] S3. Use an adaptive distributed generative adversarial network algorithm to analyze the preprocessed wall spraying parameter data set and build a prediction model for the spraying process. The prediction model predicts the spraying effect based on the input parameters of the wall spraying parameter data set.

[0101] S4. According to the output results of the prediction model, an intelligent optimization algorithm based on a combination of a genetic algorithm and a particle swarm optimization algorithm is used to adjust the operating parameters of the spraying equipment, wherein the operating parameters include spraying pressure, nozzle speed, and material flow rate;

[0102] S5. Feedback the optimal spraying parameter combination to the spraying equipment, control the operation of the spraying equipment in real time, and dynamically optimize the spraying process according to the actual situation;

[0103] S6. Repeat S1-S5 to form a closed-loop control to ensure continuous optimization of the spraying process;

[0104] S7. Monitor the execution of the spraying process in real time, and record and store key data during the spraying process.

[0105] The specific steps of step S1 in the artificial intelligence-based lightweight wall spraying process are as follows:

[0106] S11. Install a laser thickness sensor on the spraying equipment to monitor the spraying thickness in real time, an image processing sensor to monitor the spraying uniformity, a flow sensor to monitor the material consumption, a resistance pressure sensor to monitor the spraying pressure, and a temperature and humidity sensor to monitor the ambient temperature and humidity;

[0107] S12, laser thickness sensor collects spraying thickness data T at different positions on the wall surface during spraying in real time i , where i represents the time point of acquisition;

[0108] S13, image processing sensor real-time acquisition of uniformity data U of the wall surface during spraying i ;

[0109] S14, flow sensor monitors the material consumption in real time during spraying C i ;

[0110] S15, resistance pressure sensor real-time monitoring of the spraying pressure P of the spraying equipment during the spraying process i ;

[0111] S16, temperature and humidity sensor monitors the temperature T of the spraying environment in real time env，i and humidity H env，i ;

[0112] S17, sorting and summarizing all the collected spraying thickness data, uniformity data, material consumption data, spraying pressure data, ambient temperature data and ambient humidity data to construct a wall spraying parameter data set D:

[0113] D={(T i ,U i ,C i ,P i ,T env，i ,H env，i )|i=1,2,...,n}

[0114] Where n represents the total number of time points for data collection.

[0115] The specific steps of step S2 in the artificial intelligence-based lightweight wall spraying process are as follows:

[0116] S21, transmitting the wall spraying parameter data set D constructed by step S17 to the data processing module;

[0117] S22, performing data cleaning on the wall spraying parameter data set D transmitted to the data processing module, removing abnormal values ​​and missing values, and obtaining a cleaned wall spraying parameter data set;

[0118] S23, performing denoising on the cleaned wall spraying parameter data set, using a filtering algorithm to eliminate noise interference in the collected data, and obtaining a denoised wall spraying parameter data set;

[0119] S24, converting the format of the wall spraying parameter data set after denoising to the input format required by the prediction model, and obtaining the wall spraying parameter data set D after format conversion formatted .

[0120] The specific steps of step S3 in the artificial intelligence-based lightweight wall spraying process are as follows:

[0121] S31, the wall spraying parameter data set D after the format conversion in step S24 formatted Input into the adaptive distributed generative adversarial network model;

[0122] S32, the adaptive distributed generation adversarial network model includes multiple generators G k and the discriminator D k , each generator and discriminator corresponds to a spraying sub-area, and k is the sub-area number;

[0123] S33, dividing the spraying process into multiple sub-areas, and the wall spraying parameter data set D in each sub-area formatted，k ;

[0124]

[0125] Where: m is the number of sub-regions;

[0126] S34, generator G for each sub-region k According to the input wall spraying parameter data set D formatted,k Generate the simulated spraying effect data G of the sub-area k (D formatted,k );

[0127]

[0128] in: Indicates the generated spray thickness data, Represents the generated spray uniformity data, Represents the generated material consumption data, Indicates the generated spray pressure data, Represents the generated ambient temperature data, Represents the generated environmental humidity data, n k represents the total number of data points in sub-region k;

[0129] S35, discriminator D for each sub-region k According to the wall spraying parameter data set D formatted，k And the simulated data G generated by the generator k (D formatted,k ) to make a distinction and output the distinction result;

[0130] D k (D formatted，k ,G k (D formattedk ))

[0131] Discriminator D k The output result represents the difference between the generated data and the real data;

[0132] S36, based on each sub-region discriminator G k The feedback result of the corresponding generator G k Perform iterative optimization to update the generator G by minimizing the difference between the generated data and the real data k Parameters of the generator are gradually improved to improve the accuracy of the generated data.

[0133] S37. Add the spraying thickness deviation ΔT and the material consumption error ΔC to the loss function of the generator, where:

[0134]

[0135] Where: D real，k is the real wall spraying parameter dataset, λ 1 and λ 2 are weight parameters, which respectively adjust the influence of spraying thickness deviation and material consumption error on the generator;

[0136] S38, adding the spraying uniformity error ΔU and the spraying pressure error ΔP to the loss function of the discriminator, where:

[0137]

[0138]

[0139] Where: 3 and λ 4 are weight parameters, which respectively adjust the influence of spraying uniformity error and spraying pressure error on the discriminator;

[0140] S39. Introduce a multi-scale convolutional layer into the discriminator network structure to extract and fuse features of different scales to optimize the accuracy of spraying effect discrimination. After completing multiple iterations, the optimized generator G is obtained. k , each generator can be based on the input wall spraying parameter dataset D formatted，k Generate spray effect simulation data and set the generator G of all sub-areas k The generated simulation data sets are integrated to construct the prediction model M(D) of the overall spraying process. formatted ).

[0141] The specific method of step S4 in the artificial intelligence-based lightweight wall spraying process is as follows:

[0142] S41, obtain the spraying effect prediction model M (D generated by step S37 formatted ), the output results shown include predicted values ​​of spray pressure, nozzle speed and material flow rate;

[0143] S42, defining an objective function of the spraying process optimization problem, with minimizing spraying thickness deviation, material consumption error and spraying uniformity error as the optimization goal;

[0144] S43, using a genetic algorithm to initialize a set of candidate solutions, each candidate solution represents a set of spraying operation parameters, including spraying pressure P, nozzle speed V and material flow rate Q;

[0145] S44, in the selection, crossover and mutation operations of the genetic algorithm, each generation selects candidate solutions with higher fitness to perform crossover and mutation to generate new candidate solutions, and through multiple generations of iterations, the spraying operation parameters are gradually optimized;

[0146] S45, further optimizing the candidate solutions generated by the genetic algorithm using a particle swarm optimization algorithm, where each particle represents a set of spraying operation parameters and has a speed and a position, and the optimal solution is found by updating the speed and position;

[0147] S46, searching for the optimal combination of spraying operation parameters through an intelligent optimization algorithm combining a genetic algorithm and a particle swarm optimization algorithm to optimize the spraying process;

[0148]

[0149] S47, combine the optimal spraying operation parameters x * =(P * , V * , Q * ) is fed back to the spraying equipment to adjust the operating parameters of the spraying equipment in real time, so that the spraying process can be dynamically optimized according to the actual situation.

[0150] The specific method of step S39 in step S3 of the lightweight wall spraying process based on artificial intelligence is as follows:

[0151] S391, in the discriminator D k A multi-scale convolution layer is introduced into the network structure. The multi-scale convolution layer includes convolution kernels of different sizes, which respectively extract and fuse the multi-scale features of spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters, and process the features of spraying thickness and uniformity;

[0152] MSC(x)={Conv 3×3 (x),Conv 5×5 (x),Conv 7×7 (x)}

[0153]

[0154] Where: x is the input spraying parameter data, including spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters;

[0155] Conv k×k (x) represents a convolution operation with a convolution kernel size of k×k;

[0156] and are the weight and bias parameters of the convolution kernel;

[0157] σ is the activation function;

[0158] S392, fusing the features extracted by the multi-scale convolutional layer to form a comprehensive feature map F, so as to optimize the accuracy of the discriminator in judging the spraying effect, and especially enhance the judgment of the uniformity of the spraying thickness and the accuracy of the material consumption;

[0159] F = Concat(MSC 3×3 (x),MSC 5×5 (x), MSC 7×7 (x)

[0160] Among them: Concat represents the concatenation operation of convolution features of different scales;

[0161] MSC 3×3 (x), MSC 5×5 (x), MSC 7×7 (x) represents multi-scale convolution operations using 3x3, 5x5, and 7x7 convolution kernels respectively;

[0162] S393, discriminate the fused feature map and output the discrimination result for evaluating the generator G kThe difference between the generated simulated spraying effect data and the real data;

[0163] D k (F) = σ(W·F+b)

[0164] Where: W and b are the weight and bias parameters of the discriminator;

[0165] σ is the activation function;

[0166] S394, introducing specific constraints for the spraying process into the loss function of the discriminator, including spraying thickness deviation, material consumption error, spraying uniformity error and spraying pressure error;

[0167]

[0168] S395, based on discriminator D k The feedback result of the generator G k Perform multiple iterative optimizations to update the generator G by minimizing the difference between the generated data and the real data. k Parameters;

[0169]

[0170] in, To represent the expected value calculation, the generator gradually optimizes the data it generates by minimizing the loss function fed back by the discriminator;

[0171] S396. After completing multiple iterations, the optimized generator G is obtained. k ,The generator can generate spraying effect simulation data based on the input spraying parameter data set;

[0172]

[0173] in: To represent the optimized generator;

[0174] S397, using the optimized generator The generated simulation data set integrates the generator simulation data of multiple sub-areas to form a prediction model of the overall spraying process, which is used to predict the spraying effect in real time and guide the optimization of the spraying process.

[0175]

[0176] Where: M(D formatted ) is a prediction model for the overall spraying process. By integrating the generator simulation data of multiple sub-areas, an overall prediction model is formed to predict the spraying effect in real time and optimize the spraying process.

[0177] The following are some specific parameter data examples:

[0178] S1, data collection and preprocessing;

[0179] On-site construction began on August 1, 2024, starting with the installation of sensors on the spray equipment;

[0180] The laser thickness sensor monitors the spray thickness on the wall surface in real time, and records data such as: at 10:00 am on August 2, 2024, the thickness data is 8.2mm, 10:05 is 8.1mm, and 10:10 is 8.3mm;

[0181] Image processing sensors monitor spray uniformity, such as: at 10:00 a.m. on August 2, 2024, the uniformity score was 95%, at 10:05 it was 94%, and at 10:10 it was 96%;

[0182] The flow sensor monitors material consumption. For example, at 10:00 am on August 2, 2024, the consumption is 1.1 kg / m 2 , 10:05 is 1.0kg / m 2 , 10:10 is 1.2kg / m 2 ;

[0183] The resistance pressure sensor monitors the spraying pressure, and the data is as follows: at 10:00 am on August 2, 2024, the pressure is 120 psi, at 10:05 it is 115 psi, and at 10:10 it is 125 psi;

[0184] The temperature and humidity sensor monitors environmental parameters, such as: at 10:00 am on August 2, 2024, the temperature is 25°C and the humidity is 60%, and at 10:05 the temperature is 26°C and the humidity is 58%;

[0185] S2, model training and optimization;

[0186] The collected data is transmitted to the data processing module for cleaning, denoising and format conversion to form a data set D;

[0187] Use the adaptive distributed generative adversarial network model to train the data set and generate a spraying effect prediction model;

[0188] This is optimized during the second use. For example, on August 5, 2024, the prediction model generated spraying effect simulation data based on the input data set. The spraying thickness prediction value was 8.2mm, the uniformity prediction value was 95%, and the material consumption prediction value was 1.1kg / m 2 , the spraying pressure prediction value is 120psi;

[0189] S3, intelligent optimization and real-time adjustment;

[0190] According to the output results of the prediction model, the operating parameters of the spraying equipment are adjusted using an intelligent optimization algorithm based on a combination of genetic algorithm and particle swarm optimization algorithm;

[0191] In the third spraying, such as on August 10, 2024, the prediction model output shows that the current spray thickness deviation is large, and the intelligent optimization algorithm recommends adjusting the spray pressure to 118psi, the nozzle speed to 20m / s, and the material flow rate to 1.05kg / m 2 ; Input the optimized parameters into the spraying equipment and adjust the spraying operation parameters in real time, so that the spraying process can be dynamically optimized according to the actual situation;

[0192] S4, real-time data analysis and feedback;

[0193] The system monitors data changes during the spraying process in real time and makes dynamic adjustments based on the prediction model and optimization algorithm;

[0194] During the fourth construction, such as on August 15, 2024, during the spraying process, the sensor data showed that the spraying thickness began to deviate from the target value, and the system automatically adjusted the spraying pressure and material flow to ensure the spraying quality;

[0195] In order to verify the effectiveness of the method of the present invention, we conducted several comparative experiments. The following Table 1 is the experimental data, showing the specific differences between the traditional method and the method of the present invention:

[0196] Table 1 Comparison of experimental data between traditional method and the method of the present invention

[0197]

[0198] In the specific training sample, we selected a specific wall area for detailed data recording and analysis. The area of ​​the wall area is 50 square meters, the ambient temperature is between 25℃ and 30℃, and the humidity is between 40% and 60%;

[0199] Traditional method:

[0200] 1. Spraying thickness data: average thickness is 8mm, deviation range is ±1.5mm;

[0201] 2. Material consumption data: total consumption is 60kg, with an average consumption of 1.2kg per square meter;

[0202] 3. Spraying time: total time is 2.5 hours, average speed is 20 square meters / hour;

[0203] 4. Spraying quality: 42.5 square meters are qualified, and the qualified rate is 85%;

[0204] 5. Material waste: 6kg, waste rate is 10%;

[0205] The method of the present invention:

[0206] 1. Spraying thickness data: average thickness is 8mm, deviation range is ±0.5mm;

[0207] 2. Material consumption data: total consumption is 50kg, with an average consumption of 1.0kg per square meter;

[0208] 3. Spraying time: total time is 2 hours, average speed is 25 square meters / hour;

[0209] 4. Spraying quality: 47.5 square meters are qualified, and the qualified rate is 95%;

[0210] 5. Material waste: 2.5kg, waste rate is 5%;

[0211] It can be seen from the above comparative data that the method of the present invention is superior to the traditional method in terms of uniformity of spraying thickness, material consumption, spraying time efficiency, spraying quality consistency and material waste rate, etc. These data fully prove the feasibility and effectiveness of the method of the present invention.

[0212] The above description is only a preferred embodiment of the present invention and does not constitute any other form of limitation to the present invention. Any modification or equivalent change made based on the technical essence of the present invention still falls within the scope of protection required by the present invention.

Claims

1. A lightweight wall spraying process based on artificial intelligence, characterized by: The specific steps are: S1. Use the sensor equipment installed on the spraying equipment to collect the spraying thickness, uniformity, material consumption, spraying pressure and ambient temperature and humidity in real time during the spraying process to build a wall spraying parameter data set; S2, transmitting the collected wall spraying parameter data set to the data processing module for data cleaning, denoising and format conversion; S3. Use an adaptive distributed generative adversarial network algorithm to analyze the preprocessed wall spraying parameter data set and build a prediction model for the spraying process. The prediction model predicts the spraying effect based on the input parameters of the wall spraying parameter data set. S4. According to the output results of the prediction model, an intelligent optimization algorithm based on a combination of a genetic algorithm and a particle swarm optimization algorithm is used to adjust the operating parameters of the spraying equipment, wherein the operating parameters include spraying pressure, nozzle speed, and material flow rate; S5. Feedback the optimal spraying parameter combination to the spraying equipment, control the operation of the spraying equipment in real time, and dynamically optimize the spraying process according to the actual situation; S6. Repeat S1-S5 to form a closed-loop control to ensure continuous optimization of the spraying process; S7. Monitor the execution of the spraying process in real time, and record and store key data during the spraying process.

2. According to claim 1, a lightweight wall spraying process based on artificial intelligence is characterized in that: The specific method of step S1 in the artificial intelligence-based lightweight wall spraying process is as follows: S11. Install a laser thickness sensor on the spraying equipment to monitor the spraying thickness in real time, an image processing sensor to monitor the spraying uniformity, a flow sensor to monitor the material consumption, a resistance pressure sensor to monitor the spraying pressure, and a temperature and humidity sensor to monitor the ambient temperature and humidity; S12, laser thickness sensor collects spraying thickness data T at different positions on the wall surface during spraying in real time i , where i represents the time point of acquisition; S13, image processing sensor real-time acquisition of uniformity data U of the wall surface during spraying i ; S14, flow sensor monitors the material consumption in real time during spraying C i ; S15, resistance pressure sensor real-time monitoring of the spraying pressure P of the spraying equipment during the spraying process i ; S16, temperature and humidity sensor monitors the temperature T of the spraying environment in real time env，i and humidity H env，i ; S17, sorting and summarizing all the collected spraying thickness data, uniformity data, material consumption data, spraying pressure data, ambient temperature data and ambient humidity data to construct a wall spraying parameter data set D: D={(T i ,U i ,C i ,P i ,T env,i ,H env，i )|i=1,2,...,n} Where n represents the total number of time points for data collection.

3. The artificial intelligence-based lightweight wall spraying process according to claim 1 is characterized in that: The specific method of step S2 in the artificial intelligence-based lightweight wall spraying process is as follows: S21, transmitting the wall spraying parameter data set D constructed by step S17 to the data processing module; S22, performing data cleaning on the wall spraying parameter data set D transmitted to the data processing module, removing abnormal values ​​and missing values, and obtaining a cleaned wall spraying parameter data set; S23, performing denoising on the cleaned wall spraying parameter data set, using a filtering algorithm to eliminate noise interference in the collected data, and obtaining a denoised wall spraying parameter data set; S24, converting the format of the wall spraying parameter data set after denoising to the input format required by the prediction model, and obtaining the wall spraying parameter data set D after format conversion formatted .

4. The artificial intelligence-based lightweight wall spraying process according to claim 1 is characterized in that: The specific method of step S3 in the artificial intelligence-based lightweight wall spraying process is as follows: S31, the wall spraying parameter data set D after the format conversion in step S24 formatted Input into the adaptive distributed generative adversarial network model; S32, the adaptive distributed generative adversarial network model includes multiple generators G k and the discriminator D k , each generator and discriminator corresponds to a spraying sub-area, and k is the sub-area number; S33, dividing the spraying process into multiple sub-areas, and the wall spraying parameter data set D in each sub-area formatted，k ; Where: m is the number of sub-regions; S34, generator G for each sub-region k According to the input wall spraying parameter data set D formatted，k Generate the simulated spraying effect data G of the sub-area k (D formatted，k ); in: Indicates the generated spray thickness data, Represents the generated spray uniformity data, Represents the generated material consumption data, Indicates the generated spray pressure data, Represents the generated ambient temperature data, Represents the generated environmental humidity data, n k represents the total number of data points in sub-region k; S35, discriminator D for each sub-region k According to the wall spraying parameter data set D formatted，k And the simulated data G generated by the generator k (D formatted，k ) to make a distinction and output the distinction result; D k (D formatted,k ,G k (D formatted，k )) Discriminator D k The output result represents the difference between the generated data and the real data; S36, based on each sub-region discriminator D k The feedback result of the corresponding generator G k Perform iterative optimization to update the generator G by minimizing the difference between the generated data and the real data k Parameters of the generator are gradually improved to improve the accuracy of the generated data. S37. Add the spraying thickness deviation ΔT and the material consumption error ΔC to the loss function of the generator, where: Where: D real，k is a real wall spraying parameter data set, λ1 and λ2 are weight parameters, which respectively adjust the influence of spraying thickness deviation and material consumption error on the generator; S38, adding the spraying uniformity error ΔU and the spraying pressure error ΔP to the loss function of the discriminator, where: Among them: λ3 and λ4 are weight parameters, which respectively adjust the influence of spraying uniformity error and spraying pressure error on the discriminator; S39. Introduce a multi-scale convolutional layer into the discriminator network structure to extract and fuse features of different scales to optimize the accuracy of spraying effect discrimination. After completing multiple iterations, the optimized generator G is obtained. k , each generator can be based on the input wall spraying parameter dataset D formatted,k Generate spray effect simulation data and set the generator G of all sub-areas k The generated simulation data sets are integrated to construct the prediction model M(D) of the overall spraying process. formatted ).

5. The artificial intelligence-based lightweight wall spraying process according to claim 1, characterized in that: The specific method of step S4 in the artificial intelligence-based lightweight wall spraying process is as follows: S41, obtain the spraying effect prediction model M (D generated by step S37 formatted ), the output results comprising predicted values ​​of spraying pressure, nozzle speed and material flow rate; S42, defining an objective function of the spraying process optimization problem, with minimizing spraying thickness deviation, material consumption error and spraying uniformity error as the optimization goal; S43, using a genetic algorithm to initialize a set of candidate solutions, each candidate solution represents a set of spraying operation parameters, including spraying pressure P, nozzle speed V and material flow rate Q; S44, in the selection, crossover and mutation operations of the genetic algorithm, each generation selects candidate solutions with higher fitness to perform crossover and mutation to generate new candidate solutions, and through multiple generations of iterations, the spraying operation parameters are gradually optimized; S45, further optimizing the candidate solutions generated by the genetic algorithm using a particle swarm optimization algorithm, where each particle represents a set of spraying operation parameters and has a speed and a position, and the optimal solution is found by updating the speed and position; S46, searching for the optimal combination of spraying operation parameters through an intelligent optimization algorithm combining a genetic algorithm and a particle swarm optimization algorithm to optimize the spraying process; S47, combine the optimal spraying operation parameters x * =(P * , V * , Q * ) is fed back to the spraying equipment to adjust the operating parameters of the spraying equipment in real time, so that the spraying process can be dynamically optimized according to the actual situation.

6. The artificial intelligence-based lightweight wall spraying process according to claim 4 is characterized in that: The specific method of step S39 in step S3 in the artificial intelligence-based lightweight wall spraying process is as follows: S391, in the discriminator D k A multi-scale convolution layer is introduced into the network structure, wherein the multi-scale convolution layer includes convolution kernels of different sizes, and the multi-scale features of spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters are extracted and integrated respectively, and the features of spraying thickness and uniformity are processed; MSC(x)={Conv 3×3 (x),Conv 5×5 (x),Conv 7×7 (x)} Where: x is the input spraying parameter data, including spraying thickness, uniformity, material consumption, spraying pressure and environmental parameters; Conv k×k (x) represents a convolution operation with a convolution kernel size of k×k; and are the weight and bias parameters of the convolution kernel; σ is the activation function; S392, fusing the features extracted by the multi-scale convolutional layer to form a comprehensive feature map F, so as to optimize the accuracy of the discriminator in judging the spraying effect, and especially enhance the judgment of the uniformity of the spraying thickness and the accuracy of the material consumption; F=Concat(MSC 3×3 (x),MSC 5×5 (x),MSC 7×7 (x)) Among them: Concat represents the concatenation operation of convolution features of different scales; MSC 3×3 (x), MSC 5×5 (x), MSC 7×7 (x) represents multi-scale convolution operations using 3x3, 5x5, and 7x7 convolution kernels respectively; S393, discriminate the fused feature map and output the discrimination result for evaluating the generator G k The difference between the generated simulated spraying effect data and the real data; D k (F)=σ(W·F+b) Where: W and b are the weight and bias parameters of the discriminator; σ is the activation function; S394, introducing specific constraints for the spraying process into the loss function of the discriminator, including spraying thickness deviation, material consumption error, spraying uniformity error and spraying pressure error; S395, based on discriminator D k The feedback result of the generator G k Perform multiple iterative optimizations to update the generator G by minimizing the difference between the generated data and the real data. k Parameters; in, To represent the expected value calculation, the generator gradually optimizes the data it generates by minimizing the loss function fed back by the discriminator; S396. After completing multiple iterations, the optimized generator G is obtained. k ,The generator can generate spraying effect simulation data based on the input spraying parameter data set; in: To represent the optimized generator; S397, using the optimized generator The generated simulation data set integrates the generator simulation data of multiple sub-areas to form a prediction model of the overall spraying process, which is used to predict the spraying effect in real time and guide the optimization of the spraying process. Where: M(D formatted ) is a prediction model for the overall spraying process. By integrating the generator simulation data of multiple sub-areas, an overall prediction model is formed to predict the spraying effect in real time and optimize the spraying process.

Images