Method for predicting size precision of section of large prefabricated T-beam
By predicting and monitoring the cross-sectional dimensional accuracy of prefabricated T beams, the problem of lack of in-depth exploration and prevention strategies for influencing factors in the early stage and process of production is solved, and the improvement of manufacturing quality and reduction of production costs is achieved.
Patent Information
- Application Number
- CN202411756913.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-03
- Publication Date
- 2025-02-28
AI Technical Summary
The prior art lacks in-depth exploration and prevention strategies for influencing factors in the prefabricated T-beam section dimensional accuracy control, resulting in unstable manufacturing quality, large production uncertainty and high cost.
A large prefabricated T-beam section dimensional accuracy prediction method is adopted, and the missing values and outliers are processed by obtaining the original data, the feature selection is used using a random forest algorithm, and the key influencing factors are determined in combination with mechanism analysis, and a section dimensional accuracy prediction model is constructed using a backpropagation neural network and genetic algorithm.
Effective prediction and monitoring of the cross-sectional dimensional accuracy of prefabricated T-beams is achieved, reducing production uncertainty and cost, and improving manufacturing quality and economic benefits.
Smart Images
Figure CN119537897A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of new energy planning, and more specifically, to a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam. Background Art
[0002] As a key component of highway transportation, highway bridges carry the growing traffic demand and provide a solid foundation for economic and social development. With the continuous improvement of technical standards and the continuous increase in project scale, bridge construction faces many challenges. The traditional cast-in-place construction method of highway T-beams has problems such as long construction period, harsh on-site working environment, difficult quality control and large labor demand. These problems not only reduce construction efficiency and increase project costs, but also may threaten the durability and safety of bridges. Faced with challenges, the country has actively taken measures to promote scientific and technological innovation and advocated the adoption of prefabricated building concepts. In this context, large prefabricated T-beams have been widely used due to their highly standardized and modular characteristics, as well as the advantages of design following unified standards, efficient production, excellent durability and safety.
[0003] As the core component of bridge engineering, the cross-sectional dimensional accuracy of precast T-beams has a vital impact on the structural stability, safety and overall engineering quality of bridge engineering. In the production process of precast T-beams, the control of cross-sectional dimensional accuracy is a critical and complex task. Although the dimensional detection of precast T-beams has received a certain amount of attention, the current research on the cross-sectional dimensional accuracy of precast T-beams is mostly focused on post-production dimensional detection and evaluation, lacking in-depth exploration of the influencing factors in the early stage and during production and prevention strategies.
[0004] In order to effectively improve the manufacturing quality of prefabricated T-beams in prefabricated beam yards, reduce production uncertainty and costs, optimize engineering efficiency and improve economic benefits. It is necessary to systematically explore the key factors affecting the cross-sectional dimensional accuracy of prefabricated T-beams, quantitatively analyze the importance of each factor, and build a prediction model for cross-sectional dimensional accuracy based on the results of quantitative analysis to achieve effective prediction and monitoring of the cross-sectional dimensional accuracy of prefabricated T-beams. To provide theoretical support for promoting the high-quality development of prefabricated T-beam production and promoting the high-quality transformation of highway transportation construction. In response to the above problems, the present invention proposes a solution. Summary of the invention
[0005] In order to solve the problem that there is a lack of in-depth exploration and prevention strategies for the influencing factors in the early stage and process of production regarding the control of the cross-sectional dimensional accuracy of prefabricated T-beams, the purpose of the present invention is to provide a large-scale prefabricated T-beam cross-sectional dimensional accuracy prediction method for controlling the cross-sectional dimensional accuracy of prefabricated T-beams in the early stage and process of production, which can provide theoretical support for improving the manufacturing quality of prefabricated T-beams, reducing production uncertainty and costs, and improving economic benefits for prefabricated beam yards.
[0006] To achieve the above object, the present invention provides the following technical solutions: A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam is characterized in that it comprises the following steps: obtaining the original data of the cross-sectional dimension accuracy of the prefabricated T-beam, performing missing value processing, outlier processing and normalization processing on the original data of the cross-sectional dimension accuracy; listing the characteristic parameters of the prefabricated T-beam, performing feature selection based on a random forest algorithm to obtain a feature ranking that affects the cross-sectional dimension accuracy, and obtaining key influencing factors based on a mechanism analysis combined with the feature ranking; setting the parameters of a back propagation neural network, and optimizing the initial weights of the back propagation neural network using a genetic algorithm; dividing the key influencing factor data set into a training set and a test set, using the training set data to construct a cross-sectional dimension accuracy prediction model based on the optimized initial weights, and using the test set data to verify whether the accuracy of the model meets the requirements.
[0007] In a preferred embodiment, the original data of the cross-sectional dimensional accuracy of the prefabricated T-beam is obtained, and the original data of the cross-sectional dimensional accuracy is processed with missing values, outliers and normalization. The specific steps are as follows: collect process quality data, obtain raw material original data, mold original data and construction process original data through on-site quality inspection reports, obtain dimensional original data from three-dimensional scanning reports, handle missing values of original data with mean filling method, handle outliers with Laida criterion method, and finally use min-max normalization to realize preprocessing of process quality data.
[0008] In a preferred embodiment, the characteristic parameters of the prefabricated T-beam are listed, and feature selection is performed based on the random forest algorithm to obtain a feature ranking that affects the accuracy of the cross-sectional dimensions. The specific steps are as follows: samples are extracted from the original data set based on a random sampling technique with replacement, 70% of the total amount is extracted each time, and this process is repeated k times; in each sample extracted, 2d / 3 features are randomly selected from d features to construct multiple regression tree models; the generated regression tree is allowed to grow naturally until a preset tree number limit is reached, and then the generalization error of the model is calculated; a regression tree is constructed based on the training sample set, and prediction is performed on the corresponding test set, and the importance score of each feature is obtained based on the importance score calculation formula; based on a preset importance score threshold, features with a score exceeding the threshold are screened out.
[0009] In a preferred embodiment, the initial weights of the back propagation neural network are optimized, and the specific steps are as follows: determining the BP network topology, initializing the weight threshold and initializing the genetic algorithm population; the initialization content of the initial weight threshold includes the weight from the hidden layer to the input layer, the weight from the output layer to the hidden layer, and the threshold of the hidden layer and the output layer; the initialization content of the initial genetic algorithm population includes the initial population size, the number of iterations, the crossover rate and the mutation rate, and initial encoding is performed on each individual in the population; the root mean square error is used as the fitness function to calculate the initial fitness value corresponding to each individual in the population: performing a selection operation based on the roulette selection method, dividing the roulette wheel according to the proportion of the fitness value of each individual, taking the reciprocal of the fitness value to obtain a probability formula for the individual to be selected, and according to the probability formula, the larger the reciprocal of the fitness value is, the higher the probability of being selected; performing crossover and mutation operations based on the genetic algorithm; comparing the fitness values of all individuals to determine whether the maximum number of iterations has been reached, if so, stopping the iteration, outputting the optimal weight and optimal threshold parameters, updating the initial parameter settings of the GA-BP neural network, and if not, returning to the selection operation to continue execution.
[0010] In a preferred embodiment, the cross-sectional dimension accuracy prediction model is constructed using training set data, and the test set data is used to verify whether the accuracy of the model meets the requirements. The specific steps are as follows: the network is further trained using the training set, and the mean square error (MSE) of the model is calculated. If the MSE reaches the preset standard, the training is completed; if the MSE does not reach the preset standard, the weight and threshold of the model are adjusted according to the error feedback; it is determined whether the number of model training times reaches the upper limit, and if so, the training is terminated; otherwise, the step of further training the network using the training set is returned to.
[0011] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. The present invention can be used in prefabricated beam yards to effectively predict and monitor the cross-sectional dimensional accuracy of prefabricated T-beams, providing theoretical support for reducing production uncertainty and costs and improving economic benefits.
[0012] 2. The present invention can effectively improve the manufacturing quality of prefabricated T-beams and optimize engineering production efficiency.
[0013] 3. The present invention can be used to promote the high-quality development of prefabricated beam production and promote the transformation of highway transportation construction. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 A schematic flow chart of a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application.
[0015] Figure 2A schematic diagram of the particle swarm algorithm flow included in a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application.
[0016] Figure 3 A schematic diagram of a weight threshold updating process in a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application.
[0017] Figure 4 A schematic diagram showing a comparison of prediction accuracy before and after optimization in a method for predicting cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application. DETAILED DESCRIPTION
[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0019] Figure 1 A flow chart of a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application includes the following steps: S1, obtaining original data of the cross-sectional dimension accuracy of the prefabricated T-beam, and performing missing value processing, outlier processing and normalization processing on the original data of the cross-sectional dimension accuracy.
[0020] In this example, obtaining the original data of the cross-sectional dimension accuracy of the prefabricated T-beam includes collecting process quality data, obtaining the original data of raw materials, molds, and construction processes through on-site quality inspection reports, and obtaining the original dimension data from the 3D scanning report. The missing values of the original data are processed by the mean filling method. It can be understood that the mean filling method is a simple and easy method that can quickly fill in missing values to avoid the dataset being unusable due to missing values. By filling in missing values, all original data can be retained to avoid reducing the sample size due to deleting missing values, affecting the representativeness and stability of the analysis. The outlier processing adopts the Laida criterion method, and finally the min-max standardization is used to realize the preprocessing of the process quality data.
[0021] S2, enumerate the characteristic parameters of the prefabricated T-beam, perform feature selection based on the random forest algorithm to obtain the feature ranking that affects the cross-sectional dimensional accuracy, and obtain the key influencing factors based on the mechanism analysis combined with the feature ranking.
[0022] like Figure 2As shown, the characteristic parameters of the prefabricated T-beam are listed, and the feature selection based on the random forest algorithm is used to obtain the feature ranking that affects the cross-sectional size accuracy. The specific steps are as follows: S2-1, samples are extracted from the original data set based on the random sampling technique with replacement, and 70% of the total amount is extracted each time, and this process is repeated k times; S2-2, in each sample extracted, 2d / 3 features are randomly selected from d features to build multiple regression tree models; S2-3, the generated regression tree is allowed to grow naturally until the preset tree number limit is reached, and then the generalization error of the model is calculated. When using an integration method such as random forest, the generalization error can not only help evaluate the performance of a single regression tree, but also evaluate the effect of the integrated model. By comparing the results of multiple sampling and multiple training, and calculating the generalization error of the integrated model, it can be understood whether the integrated strategy improves the generalization ability of the overall model. For example, setting the preset tree number limit, the number of feature selections, and other parameters will affect the generalization error of the model. By calculating the generalization error under different configurations, the best hyperparameter combination can be selected to optimize the performance of the model. S2-4, build a regression tree based on the training sample set, and make predictions on the matching test set, and obtain the importance score of each feature based on the importance score calculation formula; The importance score calculation formula is as follows:
[0023]
[0024] In the formula, k is the number of training sample sets, d is the number of features, MSE is the mean square error, SE is the standard deviation, N is the data sample size, i represents the i-th feature, j represents the j-th training sample set, represents the predicted value of the i-th data, Represents the measured value of the i-th data.
[0025] S2-5, based on a preset importance score threshold, screen out features with scores exceeding the threshold, and use the features exceeding the threshold as important features. After obtaining key features based on mathematical analysis, take the union of the important features obtained by these two methods, and use the obtained results as the key influencing factors of cross-sectional dimensional accuracy.
[0026] Figure 3 A schematic diagram of a weight threshold updating process in a method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam provided in an embodiment of the present application is specifically as follows: S3. Set the parameters of the back propagation neural network and optimize the initial weights of the back propagation neural network using a genetic algorithm. The specific steps are as follows: S3-1, determine the BP network topology, initialize the weight threshold and initialize the genetic algorithm population; S3-2, the initialization content of the initialization weight threshold includes the weight from the hidden layer to the input layer, the weight from the output layer to the hidden layer, and the threshold of the hidden layer and the output layer; S3-3, the initialization content of the initialization genetic algorithm population includes the initial population size, the number of iterations, the crossover rate and the mutation rate, and perform initial encoding on each individual in the population; S3-4, taking the root mean square error as the fitness function, calculate the initial fitness value corresponding to each individual in the population; The root mean square error is used as the fitness function, and the specific calculation formula is as follows:
[0027] In the formula, is the root mean square error, i.e. the initial fitness value, represents the predicted value of the mth data, Represents the measured value of the mth data, where m=1,2,3,...,R, R is an integer and n is the number of individuals in the population.
[0028] S3-5, performing a selection operation based on the roulette wheel selection method, dividing the roulette wheel according to the proportion of the fitness value of each individual, taking the reciprocal of the fitness value to obtain the probability formula of the individual being selected, and according to the probability formula, the larger the reciprocal of the fitness value is, the higher the probability of being selected; The specific calculation formula of the probability formula is as follows:
[0029]
[0030] In the formula, Px is the probability of an individual being selected, and Fx is the fitness value corresponding to the individual. is the inverse of the xth fitness value, and n is the number of individuals in the population S3-6 performs crossover and mutation operations based on a genetic algorithm. The crossover and mutation operations are specifically performed in the following steps: Based on real crossover, chromosome and chromosomes A crossover operation is performed at position t, and the specific calculation formula of the crossover operation is as follows:
[0031] The specific calculation formula for the mutation operation is as follows:
[0032]
[0033] In the formula, and Gene The upper and lower bounds of , g is the number of iterations, is the maximum number of evolutions, and b is a random number between [0,1].
[0034] It should be noted that the crossover operation simulates the process of gene recombination in nature, and its purpose is to generate diversity by combining the genomes of two "parent" individuals into a new "offspring" individual. The role of the crossover operation includes generating new solutions by crossing two individuals, and sometimes finding better solutions. The individuals produced by the crossover are usually a "hybrid" of the two parent individuals, combining the advantages of each. Crossover allows high-quality gene combinations to be passed on to the next generation faster, thereby accelerating the convergence of the genetic algorithm. The mutation operation can prevent the population from falling into the local optimal solution during the search process, maintain the diversity of the population, and prevent the genetic algorithm from converging prematurely.
[0035] S3-7, compare the fitness values of all individuals to determine whether the maximum number of iterations has been reached. If so, stop the iteration, output the optimal weight and optimal threshold parameters, and update the initial parameter settings of the GA-BP neural network. If not, return to the selection operation and continue execution.
[0036] S4. Divide the key influencing factor data set into a training set and a test set, use the training set data to build a cross-sectional dimension accuracy prediction model based on the optimized initial weights, and use the test set data to verify whether the accuracy of the model meets the requirements. The specific steps are as follows: The network is further trained using the training set to calculate the mean square error (MSE) of the model. If the MSE reaches the preset standard, the training is completed. If the MSE does not reach the preset standard, the weight and threshold of the model are adjusted according to the error feedback. Determine whether the number of model training times reaches the upper limit, if so, terminate the training, if not, return to the step of further training the network using the training set.
[0037] Embodiment 1 The raw material data, mold data and construction process data were obtained from the on-site quality inspection report of a prefabricated beam yard, and the size data was obtained from the 3D scanning report. The missing values of the original data were processed by the mean filling method, the outliers were processed by the Laida criterion method, and finally the min-max standardization was adopted to obtain the data of 198 prefabricated T beams. Each data has 24 characteristic parameters, including 8 characteristics of raw materials and mix ratio, 6 characteristics of construction process, 5 characteristics of mold, and 6 characteristics of size.
[0038] The random forest algorithm was used for feature selection. The samples were randomly drawn with replacement 80 times. 16 features were randomly selected from all 18 feature attributes to construct a regression tree model for cross-sectional dimension accuracy. The termination condition was set to 80 decision trees. After the model was built, the mean square error was calculated and the score of each feature was calculated. In the feature engineering based on random forest, the features with importance score σ>0.05 were selected as important features, and the key features obtained by mechanism analysis were shown in Table 1. The results obtained by taking the union of the important features obtained by the two methods were used as the key influencing factors of cross-sectional dimension accuracy. There were 14 key features in total, including 6 raw material and mix ratio features, including water-binder ratio, cement, yellow sand, 1# crushed stone, 2# crushed stone and admixture. There were 3 mold features, including inner membrane size (inner membrane cross-sectional width, height and thickness), mold stiffness and mold surface treatment. There were 5 construction process features, including vibration frequency, vibration time, steam curing temperature, humidity and time.
[0039] Table 1
[0040] The trainglm function is selected as the BP neural network model training function, where the learning rate is 0.1, the training cycle is 2000, the expected error is 0, and the allowed number of failures is 2000.
[0041] The BP neural network model constructed in this example has a topological structure of 14-17-1, so there are 14×17+17×1=255 weights and 17+1=18 thresholds, that is, each individual in the population must be composed of 273 gene fragments. The population size of the model is set to 50 individuals, the genetic generation is set to 600, the crossover rate is set to 0.4, and the mutation rate is set to 0.1. The optimal weights and thresholds of the BP neural network are obtained by genetic algorithm, and the initial weights and thresholds are substituted to update. 80% of the data in the cross-sectional dimension accuracy data set is used for training, and 20% is used for testing. There are 40 sets of data used for testing, with width (top width, bottom width), height, and cross-sectional dimensions (top plate thickness, web thickness, bottom plate thickness) as target outputs.
[0042] Table 2
[0043] Depend on Figure 4It can be seen that the GA-BP model optimized by genetic algorithm has achieved significant improvement in prediction accuracy. In six different output dimensions, the average relative error of the BP model is 29.34%, while the GA-BP model optimized by genetic algorithm reduces this indicator to 21.33%, achieving a performance improvement of 27.3%. In addition, it can be seen from Table 2 that compared with the original BP model, the GA-BP model has achieved an improvement from 1.68 to 1.21 in MAE and an improvement from 2.03 to 1.49 in RMSE. Therefore, the present invention adjusts the neural network parameters through genetic algorithms, which can effectively improve the prediction performance of the model. Different from the traditional research focusing on finished product detection, this preventive and process control method can effectively improve the manufacturing quality of prefabricated T-beams and optimize engineering efficiency.
[0044] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.
[0045] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.
[0046] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0047] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.
[0048] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.
[0049] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for predicting the cross-sectional dimensional accuracy of a large prefabricated T-beam, characterized in that: The steps include: Acquire original data of cross-sectional dimension accuracy of prefabricated T-beams, and perform missing value processing, outlier processing and normalization processing on the original data of cross-sectional dimension accuracy; The characteristic parameters of prefabricated T-beams are listed, and feature selection based on the random forest algorithm is performed to obtain the feature ranking that affects the cross-sectional dimensional accuracy, and the key influencing factors are obtained based on the mechanism analysis combined with the feature ranking; Set the parameters of the back-propagation neural network and use genetic algorithm to optimize the initial weights of the back-propagation neural network; The key influencing factor data set is divided into a training set and a test set. The training set data is used to build a cross-sectional dimension accuracy prediction model based on the optimized initial weights, and the test set data is used to verify whether the accuracy of the model meets the requirements.
2. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 1, characterized in that: The original data of the cross-sectional dimension accuracy of the prefabricated T-beam is obtained, and the original data of the cross-sectional dimension accuracy is processed with missing values, outliers and normalization. The specific steps are as follows: Collect process quality data, obtain original material data, mold original data and construction process original data through on-site quality inspection report, obtain original size data from 3D scanning report, use mean filling method to handle missing values of original data, use Laida criterion method to handle outliers, and finally use min-max standardization to realize preprocessing of process quality data.
3. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 1, characterized in that: The characteristic parameters of the prefabricated T-beam are listed, and feature selection is performed based on the random forest algorithm to obtain the feature ranking that affects the cross-sectional dimensional accuracy. The specific steps are as follows: Draw samples from the original data set based on random sampling with replacement, taking 70% of the total each time, and repeat this process k times; In each sample drawn, 2d / 3 features are randomly selected from d features to build multiple regression tree models; The generated regression tree is allowed to grow naturally until the preset limit on the number of trees is reached, and then the generalization error of the model is calculated; A regression tree is constructed based on the training sample set, and predictions are made on the matching test set. The importance score of each feature is calculated based on the importance score calculation formula. Based on a preset importance score threshold, features with scores exceeding the threshold are filtered out.
4. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 1, characterized in that: The specific steps of optimizing the initial weights of the back propagation neural network are as follows: Determine the BP network topology, initialize the weight threshold and initialize the genetic algorithm population; The initialization content of the initialization weight threshold includes the weight from the hidden layer to the input layer, the weight from the output layer to the hidden layer, and the thresholds of the hidden layer and the output layer; The initialization content of the genetic algorithm population includes the initial population size, the number of iterations, the crossover rate and the mutation rate, and each individual in the population is initially encoded; Using the root mean square error as the fitness function, calculate the initial fitness value corresponding to each individual in the population: The selection operation is performed based on the roulette selection method. The roulette wheel is divided according to the proportion of the fitness value of each individual. The reciprocal of the fitness value is taken to obtain the probability formula of the individual being selected. According to the probability formula, the larger the reciprocal of the fitness value is, the higher the probability of being selected is. Perform crossover and mutation operations based on genetic algorithms; Compare the fitness values of all individuals to determine whether the maximum number of iterations has been reached. If so, stop the iteration, output the optimal weights and threshold parameters, and update the initial parameter settings of the GA-BP neural network. If not, return to the selection operation and continue execution.
5. The method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 1, characterized in that: The cross-sectional dimension accuracy prediction model is constructed using the training set data, and the test set data is used to verify whether the accuracy of the model meets the requirements. The specific steps are as follows: The network is further trained using the training set to calculate the mean square error (MSE) of the model. If the MSE reaches the preset standard, the training is completed. If the MSE does not reach the preset standard, the weight and threshold of the model are adjusted according to the error feedback. Determine whether the number of model training times reaches the upper limit, if so, terminate the training, if not, return to the step of further training the network using the training set.
6. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 3, characterized in that: The importance score calculation formula is as follows: In the formula, k is the number of training sample sets, d is the number of features, MSE is the mean square error, SE is the standard deviation, N is the data sample size, i represents the i-th feature, j represents the j-th training sample set, represents the predicted value of the i-th data, Represents the measured value of the i-th data.
7. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 4, characterized in that: The root mean square error is used as the fitness function, and the specific calculation formula is as follows: In the formula, is the root mean square error, i.e. the initial fitness value, represents the predicted value of the mth data, Represents the measured value of the mth data, where m=1,2,3,...,R, R is an integer and n is the number of individuals in the population.
8. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 4, characterized in that: The specific calculation formula of the probability formula is as follows: In the formula, Px is the probability of an individual being selected, and Fx is the fitness value corresponding to the individual. is the reciprocal of the x-th fitness value, and n is the number of individuals in the population.
9. A method for predicting the cross-sectional dimension accuracy of a large prefabricated T-beam according to claim 4, characterized in that: The crossover operation and mutation operation have the following specific steps: Based on real crossover, chromosome and chromosomes A crossover operation is performed at position t, and the specific calculation formula of the crossover operation is as follows: The specific calculation formula for the mutation operation is as follows: In the formula, and Gene The upper and lower bounds of , g is the number of iterations, is the maximum number of evolutions, and b is a random number between [0,1].