A method for predicting the axial compressive bearing capacity of concrete-filled square steel tube short columns

Abstract

The present invention discloses a method for predicting the axial compressive bearing capacity of square steel tube concrete short columns, comprising: 1. performing axial compression tests on N square steel tube concrete short columns, thereby obtaining the axial compressive bearing capacity of the N square steel tube concrete short columns and N sets of axial compressive bearing capacity influencing factor data; 2. obtaining a set of highly correlated influencing factors; 3. obtaining a dimension reduction matrix of influencing factors; 4. training a prediction model for the axial compressive bearing capacity of square steel tube concrete short columns; and 5. predicting the subsequent axial compressive bearing capacity of square steel tube concrete short columns. The method of the present invention has simple steps, utilizes historical data of square steel tube concrete short columns, realizes the acquisition of the axial compressive bearing capacity prediction model of square steel tube concrete short columns, and uses the axial compressive bearing capacity prediction model of square steel tube concrete short columns to predict the axial compressive bearing capacity, with high prediction accuracy, thereby solving the problems of current experimental research, numerical simulation, and theoretical calculation methods.

Description

Technical Field

[0001] The invention belongs to the technical field of architecture and civil engineering, and particularly relates to a method for predicting the axial compression bearing capacity of a square steel tube concrete short column. Background Art

[0002] Concrete-filled steel tube columns are composite components composed of steel tubes and concrete. Developed based on research on spiral reinforced concrete, profiled concrete, and steel tube structures, the steel tube and core concrete interact and complement each other, compensating for their respective shortcomings while also leveraging their respective strengths. The steel tube's cladding effect on the core concrete in concrete-filled steel tube columns transforms the concrete's stress state from uniaxial compression to triaxial compression, enhancing the concrete's compressive strength and plasticity. Furthermore, the presence of the inner concrete significantly improves the lateral and local stability of the steel tube wall, leading to their widespread application in high-rise buildings and long-span bridge structures. Compared to traditional reinforced concrete columns, square concrete-filled steel tube columns offer advantages such as high compressive strength, excellent flexural performance, good seismic resistance, fire resistance, economical efficiency, improved plasticity and ductility, and ease of construction. Therefore, research on the bearing capacity of short square concrete-filled steel tube columns is crucial.

[0003] However, the current methods used to study the influence of various factors on the bearing capacity of short square steel tube concrete-filled columns are experimental research, numerical simulation, and theoretical calculation. However, these methods all have certain limitations. Traditional experimental research is costly, time-consuming, and labor-intensive, making it difficult to conduct large-scale tests, and the test conditions have a significant impact on the test results. Numerical simulations are difficult to fully account for the influence of factors such as residual stress, initial defects, and local buckling in the steel tube. Furthermore, selecting the appropriate regression equation for regression analysis and fitting the simulation results requires skill and experience. Theoretical calculations require a high theoretical foundation for steel tube concrete, and the derivation and calculation of the formulas are difficult and cumbersome.

[0004] Therefore, there is currently a lack of a method for predicting the axial compressive bearing capacity of square steel tube concrete short columns. By utilizing the historical data of square steel tube concrete short columns, a prediction model for the axial compressive bearing capacity of square steel tube concrete short columns can be obtained, and the axial compressive bearing capacity prediction model of square steel tube concrete short columns can be used to predict the axial compressive bearing capacity with high prediction accuracy, thereby solving the problems of current experimental research, numerical simulation, and theoretical calculation methods. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method for predicting the axial compressive bearing capacity of square steel tube concrete short columns in response to the above-mentioned deficiencies in the existing technology. The method has simple steps and a reasonable design. It utilizes historical data of square steel tube concrete short columns to obtain a prediction model for the axial compressive bearing capacity of square steel tube concrete short columns, and uses the prediction model for the axial compressive bearing capacity of square steel tube concrete short columns to predict the axial compressive bearing capacity with high prediction accuracy, thereby solving the problems of current experimental research, numerical simulation, and theoretical calculation methods.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting the axial compressive bearing capacity of square steel tube concrete short columns, characterized in that the method comprises the following steps:

[0007] Step 1: Perform axial compression tests on N square steel tube concrete short columns to obtain the axial compression bearing capacity of the N square steel tube concrete short columns and N groups of axial compression bearing capacity influencing factor data; wherein each group of axial compression bearing capacity influencing factors includes square steel tube length, square steel tube width, square steel tube wall thickness, constraint effect coefficient, steel content ratio, short column height, steel tube yield strength and concrete strength;

[0008] Step 2: Using a computer, perform grey correlation analysis on the axial compressive bearing capacity of N square steel tube concrete short columns and N groups of axial compressive bearing capacity influencing factor data using an initial grey correlation analysis method to obtain a set of highly correlated influencing factors;

[0009] Step 3: Decomposing the highly correlated influencing factor variable matrix corresponding to the highly correlated influencing factor set using a non-negative matrix decomposition method using a computer to obtain a reduced dimension matrix of the influencing factors;

[0010] Step 4: Training of the prediction model for the axial compressive bearing capacity of concrete-filled square steel tube short columns:

[0011] Construct a feedforward neural network model and input the dimension reduction matrix of the influencing factors and their corresponding axial compressive bearing capacity training to obtain a trained feedforward neural network model, and record the trained feedforward neural network model as the axial compressive bearing capacity prediction model of square steel tube concrete short columns;

[0012] Step 5: Prediction of the axial compressive bearing capacity of subsequent square steel tube concrete short columns:

[0013] The influencing factor data of the subsequent square steel tube concrete short column is input into the square steel tube concrete short column axial compressive bearing capacity prediction model in step 4, and the axial compressive bearing capacity prediction value corresponding to the subsequent square steel tube concrete short column is output.

[0014] The above-mentioned method for predicting the axial compressive bearing capacity of a square steel tube concrete short column is characterized in that: Step 2, the specific process is as follows:

[0015] Step 201: Use a computer to calculate the axial compressive bearing capacity of N square steel tube concrete short columns as a reference sequence Y, where Y = [y1, ..., y n ,...,y N ]; where y n represents the nth axial compressive bearing capacity; n and N are positive integers, and 1≤n≤N;

[0016] Take N groups of axial compression bearing capacity influencing factor data as the comparison sequence X, and Among them, x ni represents the value of the i-th influencing factor in the n-th group of axial compressive bearing capacity influencing factor data, i is a positive integer, 1≤i≤I; I represents the total number of influencing factors, and I=8;

[0017] Step 202: Use a computer to perform dimensionless processing on the reference sequence Y to obtain a dimensionless reference sequence Y′, where Y′=[y1′,...,y′ n ,...,y′ N ]; where y′ n represents the nth element in a dimensionless reference sequence, and y1 represents the first axial compressive bearing capacity in the reference sequence Y;

[0018] Step 203: Use a computer to perform dimensionless processing on the comparison sequence X to obtain a dimensionless comparison sequence X′, and Where x′ ni represents the element in row n and column i of the dimensionless comparison sequence X′, and x n1 Indicates the value of the influencing factor in the nth row and the first column of the comparison sequence X;

[0019] Step 204: Use a computer to perform difference processing on the dimensionless reference sequence Y′ and the dimensionless comparison sequence X′ to obtain a difference matrix K, and Among them, the element in the nth row and ith column of the difference matrix K is denoted as k ni , and k ni =|x′ ni -y′ n |, |·| represent absolute values;

[0020] Step 205: Using a computer Get the correlation coefficient R of the element in the nth row and the ith column ni ; where a represents the minimum difference, and b represents the maximum difference, and λ represents the resolution coefficient, and 0<λ<1;

[0021] Step 206: Using a computer Get the grey relational degree r of the i-th influencing factori ;

[0022] Step 207: Use a computer to compare the grey correlation degree of each influencing factor in step 206 with the correlation threshold, and take the influencing factors corresponding to the influencing factors with a correlation degree greater than the correlation threshold as high-correlation influencing factors to obtain a set of high-correlation influencing factors; wherein the number of influencing factors in the set of high-correlation influencing factors is not less than 3.

[0023] The above-mentioned method for predicting the axial compressive bearing capacity of a square steel tube concrete short column is characterized in that: Step 3, the specific process is as follows:

[0024] Step 301: Record each highly correlated influencing factor in the highly correlated influencing factor set as the first highly correlated influencing factor, ..., the jth highly correlated influencing factor, ..., the Jth highly correlated influencing factor; wherein j and J are both positive integers, and 1≤j≤J;

[0025] Step 302: Use a computer to create a variable matrix H of highly correlated factors, and Among them, h nj It represents the value of the jth highly correlated influencing factor corresponding to the nth group of axial compressive bearing capacity;

[0026] Step 303: Use a computer to perform standardization on the variable matrix H of the highly correlated factors to generate a standardized matrix H * ,and Among them, h* nj Represents the normalized matrix H * The element in row n and column j in It represents the average value of the elements in the jth column of the variable matrix H of the highly correlated factors, and s j represents the standard deviation of the elements in the jth column of the variable matrix H of the highly correlated factors, and |·| represents the absolute value;

[0027] Step 304: Using a computer Get the correlation coefficient R between the jth column and the j'th column j,j′ ; Wherein, the value range of j and j′ is 1~J, and j≠j′; Represents the normalized matrix H * The average value of the elements in the jth column of Represents the normalized matrix H * The average value of the elements in the j′th column, h* nj′ Represents the normalized matrix H * The element in row n and column j′;

[0028] Step 305: Use a computer to calculate the correlation coefficient R between the jth column and the j′th column. j,j′Perform a judgment. If there is a correlation coefficient greater than 0.5, execute step 306;

[0029] Step 306: Use a computer to use non-negative matrix decomposition to normalize the matrix H * Perform non-negative matrix decomposition to obtain the objective function J(W,M)=||H * -WM|| 2 The first non-negative matrix W and the second non-negative matrix M under minimization; where ||·|| represents the F norm; the size of the first non-negative matrix W is N×R, and the size of the second non-negative matrix M is R×J, where R is a matrix parameter, R is a positive integer greater than or equal to 1, and satisfies (J+N)R<N×J, R<min{J,N};

[0030] Step 307: Use a computer to use the first non-negative matrix W as an influencing factor dimension reduction matrix.

[0031] The above-mentioned method for predicting the axial compressive bearing capacity of a square steel tube concrete short column is characterized in that: Step 4, the specific process is as follows:

[0032] Step 401: Construct a feedforward neural network model; wherein the feedforward neural network model includes an input layer, a first hidden layer, ... a lth hidden layer, ..., an Lth hidden layer, and an output layer, the number of neurons in the input layer is the same as R, the number of neurons in the output layer is 1, l and L are positive integers, and 1≤l≤L;

[0033] Step 402: Using a computer, use the nth row element in the influencing factor dimension reduction matrix as the nth training sample;

[0034] Step 403: Using a computer, the nth training sample is input and the axial compressive bearing capacity of the nth square steel tube concrete short column corresponding to the nth training sample is output to obtain the nth training sample;

[0035] Step 404: Repeat step 403 multiple times until the Nth row element in the influencing factor dimension reduction matrix is used as the Nth training sample to obtain a training set;

[0036] Step 405: Use a computer to input the training set into the feedforward neural network model in step 401 for training to obtain a trained feedforward neural network model.

[0037] Compared with the prior art, the present invention has the following advantages:

[0038] 1. The method of the present invention has simple steps and reasonable design, and solves the problems of current experimental research, numerical simulation and theoretical calculation methods.

[0039] 2. The present invention uses the initial value grey correlation analysis method to perform grey correlation analysis on the axial compressive bearing capacity of N square steel tube concrete short columns and N groups of axial compressive bearing capacity influencing factor data to obtain a set of influencing factors with high correlation. It can consider the degree of correlation between each influencing factor and the prediction parameter and eliminate insignificant influencing factors.

[0040] 3. The present invention uses the non-negative matrix decomposition method to decompose the high-correlation influencing factor variable matrix corresponding to the set of high-correlation influencing factors to obtain the influencing factor dimensionality reduction matrix, so as to reduce the correlation and redundancy between the influencing factors, reduce the input dimension of the feedforward neural network model, and improve the training accuracy and efficiency.

[0041] 4. The present invention constructs a feedforward neural network model and inputs the influencing factor dimensionality reduction matrix and its corresponding axial compressive bearing capacity training to obtain a trained feedforward neural network model, and records it as a square steel tube concrete short column axial compressive bearing capacity prediction model, so as to facilitate the output of the subsequent square steel tube concrete short column corresponding axial compressive bearing capacity prediction value according to the square steel tube concrete short column axial compressive bearing capacity prediction model, thereby facilitating the prediction of the axial compressive bearing capacity of batch square steel tube concrete short columns.

[0042] In summary, the method of the present invention has simple steps and reasonable design. It utilizes historical data of square steel tube concrete short columns to obtain a prediction model for the axial compressive bearing capacity of square steel tube concrete short columns, and uses the prediction model for the axial compressive bearing capacity of square steel tube concrete short columns to predict the axial compressive bearing capacity with high prediction accuracy, thereby solving the problems of current experimental research, numerical simulation, and theoretical calculation methods.

[0043] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION

[0045] like Figure 1 As shown, the method for predicting the axial compressive bearing capacity of a square steel tube concrete short column of the present invention comprises the following steps:

[0046] Step 1: Perform axial compression tests on N square steel tube concrete short columns to obtain the axial compression bearing capacity of the N square steel tube concrete short columns and N groups of axial compression bearing capacity influencing factor data; wherein each group of axial compression bearing capacity influencing factors includes square steel tube length, square steel tube width, square steel tube wall thickness, constraint effect coefficient, steel content ratio, short column height, steel tube yield strength and concrete strength;

[0047] Step 2: Using a computer, perform grey correlation analysis on the axial compressive bearing capacity of N square steel tube concrete short columns and N groups of axial compressive bearing capacity influencing factor data using an initial grey correlation analysis method to obtain a set of highly correlated influencing factors;

[0048] Step 3: Decomposing the highly correlated influencing factor variable matrix corresponding to the highly correlated influencing factor set using a computer and a non-negative matrix decomposition method to obtain a reduced dimension matrix of the influencing factors;

[0049] Step 4: Training of the prediction model for the axial compressive bearing capacity of concrete-filled square steel tube short columns:

[0050] Construct a feedforward neural network model and input the dimension reduction matrix of the influencing factors and their corresponding axial compressive bearing capacity training to obtain a trained feedforward neural network model, and record the trained feedforward neural network model as the axial compressive bearing capacity prediction model of square steel tube concrete short columns;

[0051] Step 5: Prediction of the axial compressive bearing capacity of subsequent square steel tube concrete short columns:

[0052] The influencing factor data of the subsequent square steel tube concrete short column is input into the square steel tube concrete short column axial compressive bearing capacity prediction model in step 4, and the axial compressive bearing capacity prediction value corresponding to the subsequent square steel tube concrete short column is output.

[0053] In this embodiment, the specific process of step 2 is as follows:

[0054] Step 201: Use a computer to calculate the axial compressive bearing capacity of N square steel tube concrete short columns as a reference sequence Y, where Y = [y1, ..., y n ,...,y N ]; where y n represents the nth axial compressive bearing capacity; n and N are positive integers, and 1≤n≤N;

[0055] Take N groups of axial compression bearing capacity influencing factor data as the comparison sequence X, and Among them, x ni represents the value of the i-th influencing factor in the n-th group of axial compressive bearing capacity influencing factor data, i is a positive integer, 1≤i≤I; I represents the total number of influencing factors, and I=8;

[0056] Step 202: Use a computer to perform dimensionless processing on the reference sequence Y to obtain a dimensionless reference sequence Y′, where Y′=[y1′,...,y′ n ,...,y′ N ]; where y′ n represents the nth element in a dimensionless reference sequence, and y1 represents the first axial compressive bearing capacity in the reference sequence Y;

[0057] Step 203: Use a computer to perform dimensionless processing on the comparison sequence X to obtain a dimensionless comparison sequence X′, and Where x′ ni represents the element in row n and column i of the dimensionless comparison sequence X′, and x n1 Indicates the value of the influencing factor in the nth row and the first column of the comparison sequence X;

[0058] Step 204: Use a computer to perform difference processing on the dimensionless reference sequence Y′ and the dimensionless comparison sequence X′ to obtain a difference matrix K, and Among them, the element in the nth row and ith column of the difference matrix K is denoted as k ni , and k ni =|x′ ni -y′ n |, |·| represent absolute values;

[0059] Step 205: Using a computer Get the correlation coefficient R of the element in the nth row and the ith column ni ; where a represents the minimum difference, and b represents the maximum difference, and λ represents the resolution coefficient, and 0<λ<1;

[0060] Step 206: Using a computer Get the grey relational degree r of the i-th influencing factor i ;

[0061] Step 207: Use a computer to compare the grey correlation degree of each influencing factor in step 206 with the correlation threshold, and take the influencing factors corresponding to the influencing factors with a correlation degree greater than the correlation threshold as high-correlation influencing factors to obtain a set of high-correlation influencing factors; wherein the number of influencing factors in the set of high-correlation influencing factors is not less than 3.

[0062] In this embodiment, the specific process of step three is as follows:

[0063] Step 301: Record each highly correlated influencing factor in the highly correlated influencing factor set as the first highly correlated influencing factor, ..., the jth highly correlated influencing factor, ..., the Jth highly correlated influencing factor; wherein j and J are both positive integers, and 1≤j≤J;

[0064] Step 302: Use a computer to create a variable matrix H of highly correlated factors, and Among them, h nj It represents the value of the jth highly correlated influencing factor corresponding to the nth group of axial compressive bearing capacity;

[0065] Step 303: Use a computer to perform standardization on the variable matrix H of the highly correlated factors to generate a standardized matrix H * ,and Among them, h* nj Represents the normalized matrix H * The element in row n and column j in It represents the average value of the elements in the jth column of the variable matrix H of the highly correlated factors, and s j represents the standard deviation of the elements in the jth column of the variable matrix H of the highly correlated factors, and |·| represents the absolute value;

[0066] Step 304: Using a computer Get the correlation coefficient R between the jth column and the j'th column j,j′ ; Wherein, the value range of j and j′ is 1~J, and j≠j′; Represents the normalized matrix H * The average value of the elements in the jth column of Represents the normalized matrix H * The average value of the elements in the j′th column, h* nj′ Represents the normalized matrix H * The element in row n and column j′;

[0067] Step 305: Use a computer to calculate the correlation coefficient R between the jth column and the j′th column. j,j′ Perform a judgment. If there is a correlation coefficient greater than 0.5, execute step 306;

[0068] Step 306: Use a computer to use non-negative matrix decomposition to normalize the matrix H * Perform non-negative matrix decomposition to obtain the objective function J(W,M)=||H * -WM|| 2 The first non-negative matrix W and the second non-negative matrix M under minimization; where ||·|| represents the F norm; the size of the first non-negative matrix W is N×R, and the size of the second non-negative matrix M is R×J, where R is a matrix parameter, R is a positive integer greater than or equal to 1, and satisfies (J+N)R<N×J, R<min{J,N};

[0069] Step 307: Use a computer to use the first non-negative matrix W as an influencing factor dimension reduction matrix.

[0070] In this embodiment, the specific process of step 4 is as follows:

[0071] Step 401: Construct a feedforward neural network model; wherein the feedforward neural network model includes an input layer, a first hidden layer, ... a lth hidden layer, ..., an Lth hidden layer, and an output layer, the number of neurons in the input layer is the same as R, the number of neurons in the output layer is 1, l and L are positive integers, and 1≤l≤L;

[0072] Step 402: Using a computer, use the nth row element in the influencing factor dimension reduction matrix as the nth training sample;

[0073] Step 403: Using a computer, the nth training sample is input and the axial compressive bearing capacity of the nth square steel tube concrete short column corresponding to the nth training sample is output to obtain the nth training sample;

[0074] Step 404: Repeat step 403 multiple times until the Nth row element in the influencing factor dimension reduction matrix is used as the Nth training sample to obtain a training set;

[0075] Step 405: Use a computer to input the training set into the feedforward neural network model in step 401 for training to obtain a trained feedforward neural network model.

[0076] In this embodiment, the value of the correlation threshold in step 207 is greater than 0.5, ensuring that the number of influencing factors in the high-correlation influencing factor set is not less than 3, which facilitates subsequent dimensionality reduction processing.

[0077] In this embodiment, in step 401 , L=2.

[0078] In this embodiment, the number of neurons in the first hidden layer in step 401 is c represents the adjustment coefficient, and its value ranges from 1 to 10. When l is greater than or equal to 2, the number of neurons in the two adjacent hidden layers satisfies θ l represents the number of neurons in the lth hidden layer, θ l-1 represents the number of neurons in the l-1th hidden layer.

[0079] In this embodiment, the feedforward neural network model is a Feedforward Neural Network (FNN) model.

[0080] In this embodiment, it should be noted that in step 305 , there may be a case where the correlation coefficient is greater than 0.5.

[0081] In this embodiment, the resolution coefficient λ in step 205 is set to 0.5.

[0082] In this embodiment, it should be noted that step five is to process the influencing factor data of the subsequent N' groups of square steel tube concrete short columns through step three, and input them into the square steel tube concrete short column axial compressive bearing capacity prediction model to output N' axial compressive bearing capacity prediction values corresponding to the N' groups; wherein N' is a positive integer not less than 5.

[0083] In summary, the method of the present invention has simple steps and reasonable design. It utilizes historical data of square steel tube concrete short columns to obtain a prediction model for the axial compressive bearing capacity of square steel tube concrete short columns, and uses the prediction model for the axial compressive bearing capacity of square steel tube concrete short columns to predict the axial compressive bearing capacity with high prediction accuracy, thereby solving the problems of current experimental research, numerical simulation, and theoretical calculation methods.

[0084] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for predicting the axial compressive bearing capacity of square steel tube concrete short columns, characterized in that: The method comprises the following steps: Step 1: Perform axial compression tests on N square steel tube concrete short columns to obtain the axial compression bearing capacity of the N square steel tube concrete short columns and N groups of axial compression bearing capacity influencing factor data; wherein each group of axial compression bearing capacity influencing factors includes square steel tube length, square steel tube width, square steel tube wall thickness, constraint effect coefficient, steel content ratio, short column height, steel tube yield strength and concrete strength; Step 2: Using a computer, perform grey correlation analysis on the axial compressive bearing capacity of N square steel tube concrete short columns and N groups of axial compressive bearing capacity influencing factor data using an initial grey correlation analysis method to obtain a set of highly correlated influencing factors; Step 3: Decomposing the highly correlated influencing factor variable matrix corresponding to the highly correlated influencing factor set using a computer and a non-negative matrix decomposition method to obtain a reduced dimension matrix of the influencing factors; Step 4: Training of the prediction model for the axial compressive bearing capacity of concrete-filled square steel tube short columns: Construct a feedforward neural network model and input the dimension reduction matrix of the influencing factors and their corresponding axial compressive bearing capacity training to obtain a trained feedforward neural network model, and record the trained feedforward neural network model as the axial compressive bearing capacity prediction model of square steel tube concrete short columns; Step 5: Prediction of the axial compressive bearing capacity of subsequent square steel tube concrete short columns: Input the influencing factor data of the subsequent square steel tube concrete short column into the square steel tube concrete short column axial compressive bearing capacity prediction model in step 4, and output the predicted axial compressive bearing capacity value corresponding to the subsequent square steel tube concrete short column; Step 3: The specific process is as follows: Step 301: Record each highly correlated influencing factor in the highly correlated influencing factor set as the first highly correlated influencing factor, ..., the jth highly correlated influencing factor, ..., the Jth highly correlated influencing factor; wherein j and J are both positive integers, and 1≤j≤J; Step 302: Use a computer to create a variable matrix H of highly correlated factors, and Among them, h nj It represents the value of the jth highly correlated influencing factor corresponding to the nth group of axial compressive bearing capacity; Step 303: Use a computer to perform standardization on the variable matrix H of the highly correlated factors to generate a standardized matrix H * ,and Among them, h* nj Represents the normalized matrix H * The element in row n and column j in It represents the average value of the elements in the jth column of the variable matrix H of the highly correlated factors, and s j represents the standard deviation of the elements in the jth column of the variable matrix H of the highly correlated factors, and |·| represents the absolute value; Step 304: Using a computer Get the correlation coefficient R between the jth column and the j'th column j,j′ ; Wherein, the value range of j and j′ is 1~J, and j≠j′; Represents the normalized matrix H * The average value of the elements in the jth column of Represents the normalized matrix H * The average value of the elements in the j′th column, h* nj′ Represents the normalized matrix H * The element in row n and column j′; Step 305: Use a computer to calculate the correlation coefficient R between the jth column and the j′th column. j,j′ Perform a judgment. If there is a correlation coefficient greater than 0.5, execute step 306; Step 306: Use a computer to use non-negative matrix decomposition to normalize the matrix H * Perform non-negative matrix decomposition to obtain the objective function J(W,M)=||H * -WM|| 2 The first non-negative matrix W and the second non-negative matrix M under minimization; where ||·|| represents the F norm; the size of the first non-negative matrix W is N×R, and the size of the second non-negative matrix M is R×J, where R is a matrix parameter, R is a positive integer greater than or equal to 1, and satisfies (J+N)R<N×J, R<min{J,N}; Step 307: Using a computer, use the first non-negative matrix W as a dimension reduction matrix of influencing factors; Step 4: The specific process is as follows: Step 401: Construct a feedforward neural network model; wherein the feedforward neural network model includes an input layer, a first hidden layer, ... a lth hidden layer, ..., an Lth hidden layer, and an output layer, the number of neurons in the input layer is the same as R, the number of neurons in the output layer is 1, l and L are positive integers, and 1≤l≤L; Step 402: Using a computer, use the nth row element in the influencing factor dimension reduction matrix as the nth training sample; Step 403: Using a computer, the nth training sample is input and the axial compressive bearing capacity of the nth square steel tube concrete short column corresponding to the nth training sample is output to obtain the nth training sample; Step 404: Repeat step 403 multiple times until the Nth row element in the influencing factor dimension reduction matrix is used as the Nth training sample to obtain a training set; Step 405: Use a computer to input the training set into the feedforward neural network model in step 401 for training to obtain a trained feedforward neural network model.

2. A method for predicting the axial compressive bearing capacity of a square concrete-filled steel tube short column according to claim 1, characterized in that: Step 2: The specific process is as follows: Step 201: Use a computer to calculate the axial compressive bearing capacity of N square steel tube concrete short columns as a reference sequence Y, where Y = [y1, ..., y n ,...,y N ]; where y n represents the nth axial compressive bearing capacity; n and N are positive integers, and 1≤n≤N; Take N groups of axial compression bearing capacity influencing factor data as the comparison sequence X, and Among them, x ni represents the value of the i-th influencing factor in the n-th group of axial compressive bearing capacity influencing factor data, i is a positive integer, 1≤i≤I; I represents the total number of influencing factors, and I=8; Step 202: Use a computer to perform dimensionless processing on the reference sequence Y to obtain a dimensionless reference sequence Y′, where Y′=[y1′,...,y′ n ,...,y′ N ]; where y′ n represents the nth element in a dimensionless reference sequence, and y1 represents the first axial compressive bearing capacity in the reference sequence Y; Step 203: Use a computer to perform dimensionless processing on the comparison sequence X to obtain a dimensionless comparison sequence X′, and Where x′ ni represents the element in row n and column i of the dimensionless comparison sequence X′, and x n1 Indicates the value of the influencing factor in the nth row and the first column of the comparison sequence X; Step 204: Use a computer to perform difference processing on the dimensionless reference sequence Y′ and the dimensionless comparison sequence X′ to obtain a difference matrix K, and Among them, the element in the nth row and ith column of the difference matrix K is denoted as k ni , and k ni =|x′ ni -y′ n |, |·| represent absolute values; Step 205: Using a computer Get the correlation coefficient R of the element in the nth row and the ith column ni ; where a represents the minimum difference, and b represents the maximum difference, and λ represents the resolution coefficient, and 0<λ<1; Step 206: Using a computer Get the grey relational degree r of the i-th influencing factor i ; Step 207: Use a computer to compare the grey correlation degree of each influencing factor in step 206 with the correlation threshold, and take the influencing factors corresponding to the influencing factors with a correlation degree greater than the correlation threshold as high-correlation influencing factors to obtain a set of high-correlation influencing factors; wherein the number of influencing factors in the set of high-correlation influencing factors is not less than 3.

Images