A big data-driven multi-objective intelligent optimization method for cable forces in the construction phase of cable-stayed bridges

Abstract

A big data-driven, multi-objective intelligent optimization method for cable forces during the construction phase of cable-stayed bridges relates to the technical field of intelligent optimization for bridge construction. The method comprises the following steps: determining the initial cable forces during the construction phase; generating a set of cable forces during the initial construction phase; forming a data set through forward analysis of a finite element model; obtaining a subset of possible bisection points; determining the optimal bisection point; constructing a predictive function model; obtaining a rejection matrix; grading the cable force sets during the construction phase; obtaining a density operator matrix; generating the #imgabs0#th generation of cable force sets during the construction phase; and obtaining the optimal set of cable forces during the construction phase. By combining big data analysis with intelligent optimization algorithms and integrating the multi-objective optimization requirements of the construction phase, the method comprehensively considers the complexity of the construction environment, material properties, and mechanical behavior, effectively improving the construction accuracy and quality of long-span cable-stayed bridges.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent optimization of bridge construction, and in particular to a big data-driven multi-objective intelligent optimization method for cable forces during the construction phase of a cable-stayed bridge. Background Art

[0002] Cable-stayed bridges, a key structural form of modern bridges, are widely used in long-span bridge projects due to their superior load-bearing performance and economic efficiency. However, the construction process of cable-stayed bridges is extremely complex. The distribution and adjustment of cable forces, in particular, directly impact the structural stress distribution, alignment control, and construction safety, and are crucial factors in determining the overall structural integrity of the bridge. Therefore, optimizing cable forces during the construction phase has become a core issue that urgently needs to be addressed in cable-stayed bridge construction.

[0003] Currently, the optimization of cable forces during cable-stayed bridge construction primarily relies on methods such as finite element analysis and iterative calculations. While these methods can meet some engineering requirements, they typically assume deterministic construction conditions and ignore the randomness of the construction environment and structural nonlinearity, which impacts bridge safety and performance. Furthermore, the construction phase involves multiple interrelated objectives, such as stress equilibrium and deformation control. Traditional single-objective optimization methods struggle to achieve global coordination under multi-objective constraints, especially for the collaborative optimization of long-span bridges. With the development of big data technology and intelligent optimization algorithms, bridge construction optimization has ushered in new opportunities. Big data analysis can capture the mechanical behavior characteristics of the construction process, while intelligent optimization algorithms can obtain effective solutions based on multi-objective balance. However, the application of these technologies in cable-stayed bridge construction is still imperfect, and the lack of a systematic approach makes it difficult to achieve optimal results in multi-objective optimization during the construction phase.

[0004] In summary, existing technologies for optimizing cable force during cable-stayed bridge construction have significant shortcomings, including insufficient consideration of the construction environment and the randomness of mechanical behavior, incomplete optimization objectives, and low optimization efficiency. These limitations limit the intelligent and precise nature of cable-stayed bridge construction. Therefore, there is an urgent need to develop an intelligent and reliable construction optimization method to comprehensively improve the scientific nature and reliability of cable force distribution during cable-stayed bridge construction. Summary of the Invention

[0005] To address the shortcomings of the background technology, the present invention provides a big data-driven multi-objective intelligent optimization method for cable forces in the construction phase of cable-stayed bridges. Through big data analysis and intelligent optimization algorithms, combined with the multi-objective optimization requirements of the construction phase, it comprehensively considers the complexity of the construction environment, material properties, and mechanical behavior, and can effectively improve the construction accuracy and bridge quality of large-span cable-stayed bridges.

[0006] To achieve the above objectives, the present invention adopts the following technical solutions: a big data driven cable-stayed bridge construction phase cable force multi-objective intelligent optimization method, including

[0007] Step 1: Determine the cable force during the initial construction phase

[0008] The finite element model of the cable-stayed bridge is established. There are π batches of tensioning during the construction process. The initial cable force S0 of each batch of tensioning is obtained through reverse dismantling simulation. n , n=1,2,...,Π-1,Π, the cable force matrix in the initial construction stage is expressed as follows:

[0009] S0=[S0 1 S0 2 …S0 Π-1 S0 Π ] T ;

[0010] Step 2: Generating cable force sets during the initial construction phase

[0011] For the initial construction stage, the cable force aggregate The generation of cable force is set to [-ΔS0 n ,ΔS0 n ], randomly select the cable force change value Δs0 n(N) , we get the cable force change matrix Δs0 (N) It is expressed as follows:

[0012] Δs0 (N) =[Δs0 1(N) Δs0 2(N) …Δs0 Π-1 ( N) Δs0 Π(N) ] T

[0013] Then the cable force set M0 in the initial construction stage is expressed as follows:

[0014]

[0015] Where, Δs0 n(N) ~U([-ΔS0 n ,ΔS0 n ]), ΔS0 n =0.1×S0 n , N={1,2,...,499,500}, M0 N =S0+Δs0 (N) , C0 n(N) =S0 n +Δs0 n(N) ;

[0016] Step 3: Finite element model analysis to form a data set

[0017] Use each initial construction stage cable force aggregate individual Perform forward analysis to obtain the bridge cable force matrix A (1) N and the bridge anchor point elevation matrix A (2) N ;

[0018] M0 1 ,M0 2 ,...,M0 499 ,M0 500 And the corresponding A (t) 1 ,A (t) 2 ,...,A (t) 499 ,A (t) 500 As the original data set, t = 1, 2, repeatedly randomly extract all individuals from the original data set with replacement Record Composed training set X s , corresponding to A (t) N Denoted as A (t) Ns Form the training subset Y (t) s , suppose that and the corresponding A (t) N There are L pairs and they are represented by M Test φN and A Test(t) φN They form the test set X Test s With the test subset Y Test(t) s , φ=1,2,...L-1,L, which is expressed as follows:

[0019]

[0020] Y (t) Ns =[A (t) 1s A (t) 2s …A (t) 499s A (t) 500s ]

[0021]

[0022] Y Test(t)s =[A Test(t) 1s A Test(t) 2s …A Test(t) L-1s A Test(t) Ls ];

[0023] Step 4: Obtaining a subset of possible binary points

[0024] Set binary features For each training set X s As a binary source node Perform binary division, and each binary node is represented as P represents the level of the node, L P Indicates the sequence number of the node in this layer;

[0025] For a binary node from In the random sampling, 5 different binary features are extracted to obtain the feature subset It is expressed as follows:

[0026]

[0027] Then we get the corresponding characteristic cable force subset It is expressed as follows:

[0028]

[0029] The feature force subset Arrange each row in order of size and calculate the middle value of two adjacent elements in each row as a possible bisection point w={1,2,3,4,5},d={1,2,...,498,499},get the possible bisection point subset It is expressed as follows:

[0030]

[0031] Step 5: Determine the optimal bisection point

[0032] For possible bisection points In the bisection node, if M0 Ns in but If M0 Ns in but Thus we get and Medium M0 Ns Corresponding A (t) NsThe verification matrix is ​​constructed separately and

[0033] Will and A in (t) Ns A (t)left δs With A (t)right ζs , δ=1,2,...m-1,m, ζ=1,2,...n-1,n, then A (t)left δs and A (t)right ζs The average values ​​of and It is expressed as follows:

[0034]

[0035] Then you can get A (t)left δs and and A (t)right ζs and The errors are as well as Since t=1,2, it is divided into two cases as follows:

[0036]

[0037] Possible equinox The corresponding weighted error It is expressed as follows:

[0038]

[0039] Traverse all possible bisection points When the corresponding weighted error When it is the smallest, take the possible bisection point The optimal bisection point for this bisection node;

[0040] Step 6: Prediction function model construction

[0041] by As a prediction node, M0 in each prediction node Ns Corresponding A (t) Ns The average value of the possible prediction value That is to complete the prediction function sub-model PRE (t) s The construction of

[0042] Bring in the test set X Test s The predicted value A can be obtained Test(t) (Pre)φs The prediction subset Y Test(t) (Pre)s It is expressed as follows:

[0043] Y Test(t) (Pre)s =[A Test(t) (Pre)1s A Test(t) (Pre)2s … A Test(t) (Pre)L-1s A Test(t) (Pre)Ls ]

[0044] Find the test subset Y Test(t) s With the prediction subset Y Test(t) (Pre)s Error SE last It is expressed as follows:

[0045]

[0046] If the error SE last If the accuracy requirement is met, it is judged to be an effective prediction function sub-model PRE (t) s , otherwise, repeat step 5;

[0047] Merge all valid prediction function sub-models PRE (t) s The prediction result A (t) (Pre)s And take the average value to get A (t) (Pre) , that is, completing the prediction function model PRE (t) all The construction of

[0048] Step 7: Obtaining the rejection matrix

[0049] Known Cable force collection during construction phase It is expressed as follows:

[0050]

[0051] Assume that the prediction function model PRE (t) all The obtained Cable force aggregate individual during construction phase The target prediction value The known target ideal value is A (t)(wanted) ,get Rejection Since t=1,2, it is divided into two cases as follows:

[0052]

[0053] Therefore, the Rejection degree matrix of cable force set during construction phase It is expressed as follows:

[0054]

[0055] Step 8: Cable force assembly and grading during construction phase

[0056] When a Cable force collection during construction phase Rejection and are less than or equal to Target rejection and and exists Less than Then it is called Complete victory Recorded as

[0057] For each Cable force aggregate individual during construction phase exist:

[0058]

[0059] in, for the number of for the number of

[0060] but Collection It is expressed as follows:

[0061]

[0062] For the first During the construction phase, the cable force is graded. The individual is recorded as The quantity is θ F(1) ,χ=1,2,...,θ F(1) -1,θ F(1) , stored in the first-level hierarchical set F(1), expressed as follows:

[0063]

[0064] Assume that there are l levels in total. For the Π-th level hierarchical set F(Π), Π=2,3,...,l-1,l, examine each individual in the hierarchical set F(Π-1) of Statistics Collection middle The number of individuals is θ F(Π) , recorded as Stored in the hierarchical set F(Π), expressed as follows:

[0065]

[0066] Continue grading until the l-th level grading set is obtained;

[0067] Step 9: Obtaining the density operator matrix

[0068] For the H-th level hierarchical set F(H), according to The size of F(H) Sort by, H=1,2...,l-1,l,λ=1,2,...,θ F(H) -1,θ F(H) ,set up Density operator It is expressed as follows:

[0069]

[0070] Get the density operator of the H-th level hierarchical set F(H) It is expressed as follows:

[0071]

[0072] Step 10: Generate the Cable force collection during construction phase

[0073] No. The individual in each level of the cable force set F(H) during the construction phase Pair each pair randomly. If the number of individuals in a certain level of classification set F(H) is odd, the individual with the smallest density operator will be counted into the next level of classification set for pairing, forming 250 pairs. and Each pair and generate and Constructing the offspring matrix It is expressed as follows:

[0074]

[0075] r~U([0,1])

[0076] The offspring matrix With the Generation population The merger constitutes temporary population The temporary population individuals are denoted as L=1,2,...,999,1000, which is expressed as follows:

[0077]

[0078] For the first temporary population Repeat steps 7 to 9 to store each level of collection in the Generation population, until it is stored in a certain level set If the number of individuals in the cable force set of the first construction stage exceeds 500, the hierarchical set of this level is stored according to the size of the density operator, and the ones with larger density operators are given priority. The cable force set during the construction phase is It is expressed as follows:

[0079]

[0080] Step 11: Obtaining the optimal cable force set during construction

[0081] Repeat steps 7 to 10 to get Generation population And satisfy The first Generation population The first-level classification set F(1) is taken as the optimal cable force set in the construction stage.

[0082] Compared with the existing technology, the present invention has the following beneficial effects: it combines big data analysis and intelligent optimization algorithms, comprehensively considers the complexity of cable-stayed bridge construction, can significantly improve the accuracy and efficiency of cable force optimization during the construction phase, and proposes an efficient and intelligent optimization method to meet the multi-objective optimization requirements of cable-stayed bridge construction. It provides scientific support for bridge construction control, cable force and alignment optimization, improves bridge construction accuracy and quality, and ensures the structural safety and performance of the bridge. It also has the following advantages:

[0083] 1. Systematic and efficient optimization method: By comprehensively analyzing large amounts of data from the construction phase and optimizing the cable force distribution and linear control of the completed bridge, the optimal cable force scheme for the construction phase can be quickly obtained without the need for step-by-step debugging, significantly improving the optimization efficiency of cable force during the construction phase.

[0084] 2. The optimization results are closer to reality: By fully considering multiple influencing factors such as the construction environment, material properties, and nonlinear effects, more accurate cable force distribution and linear control are achieved, which can truly and reliably reflect cable force changes and structural responses;

[0085] 3. Intelligent optimization process: It can realize automatic calculation through programming, avoiding tedious manual intervention, improving the intelligence level of the construction process, and ensuring the efficiency and accuracy of the calculation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION

[0087] The technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0088] like Figure 1 As shown in FIG, a big data-driven multi-objective intelligent optimization method for cable forces in the construction phase of a cable-stayed bridge includes the following steps:

[0089] Step 1: Determine the cable force during the initial construction phase

[0090] Establish a finite element model of a cable-stayed bridge. Assume that there are π batches of tensioning during the construction process. Through reverse dismantling simulation, the initial cable force of each batch of tensioning during the construction phase can be obtained as S0. n , n=1,2,...,Π-1,Π, then the cable force matrix in the initial construction stage is expressed as follows:

[0091] S0=[S0 1 S0 2 … S0 Π-1 S0 Π ] T .

[0092] Step 2: Generating cable force sets during the initial construction phase

[0093] For the initial construction stage, the cable force aggregate The generation of cable force is set to [-ΔS0 n ,ΔS0 n ], randomly select the cable force change value Δs0 in each cable force change range n(N) , we get the cable force change matrix Δs0 (N) It is expressed as follows:

[0094] Δs0 (N) =[Δs0 1(N) Δs0 2(N) … Δs0 Π-1(N) Δs0 Π(N) ] T

[0095] Where, Δs0 n(N) ~U([-ΔS0 n ,ΔS0 n ]), ΔS0 n =0.1×S0 n ,U([-ΔS0 n ,ΔS0 n ]) indicates that in the interval [-ΔS0 n ,ΔS0 n ] uniform distribution within, N={1,2,…,499,500}, M0 N =S0+Δs0 (N) .

[0096] Then the cable force set M0 in the initial construction stage is expressed as follows:

[0097]

[0098] Where C0 n(N) =S0 n +Δs0 n(N) .

[0099] Step 3: Finite element model analysis to form a data set

[0100] Use each initial construction stage cable force aggregate individual Perform forward analysis to obtain the bridge cable force matrix A (1) N and the bridge anchor point elevation matrix A (2) N .

[0101] M0 1 ,M0 2 ,…,M0 499 ,M0 500 And the corresponding A (t) 1 ,A (t) 2 ,...,A (t) 499 ,A (t) 500 As the original data set, t = 1, 2, repeat 750 times to randomly extract all 500 initial construction stage cable force set individuals from the original data set with replacement Record Composed training set X s , corresponding to A (t) N Denoted as A (t) Ns Form the training subset Y (t) s , s={1,2,...,749,750}, let and the corresponding A (t) N There are L pairs and they are represented by M Test φN and A Test(t) φN They form the test set X Test s With the test subset Y Test(t) s , φ=1,2,...L-1,L, which is expressed as follows:

[0102]

[0103] Y (t) Ns =[A (t) 1s A (t) 2s … A (t) 499s A (t) 500s ]

[0104]

[0105] Y Test(t) s =[A Test(t) 1s A Test(t) 2s … A Test(t) L-1s A Test(t) Ls ].

[0106] Step 4: Obtaining a subset of possible binary points

[0107] set up is a binary feature, For each training set X s As a binary source node Perform 20 layers of binary splits, and each binary node is represented as P represents the level of the node, L P Indicates the sequence number of the node in this layer, P = 0, 1, 2...8, 9, L P =1,2,...,2P -1,2 P .

[0108] For a binary node from In the random sampling, 5 different binary features are extracted to obtain the feature subset It is expressed as follows:

[0109]

[0110] Then we get the corresponding characteristic cable force subset It is expressed as follows:

[0111]

[0112] The feature force subset Arrange each row in order of size and calculate the middle value of two adjacent elements in each row as a possible bisection point w={1,2,3,4,5},d={1,2,...,498,499},get the possible bisection point subset It is expressed as follows:

[0113]

[0114] Step 5: Determine the optimal bisection point

[0115] For possible bisection points In the bisection node, if M0 Ns in but If M0 Ns in but Thus we get and Medium M0 Ns Corresponding A (t) Ns The verification matrix is ​​constructed separately and

[0116] Will and A in (t) Ns A (t)left δs With A (t)right ζs , δ=1,2,...m-1,m, ζ=1,2,...n-1,n, then A (t)left δs and A (t)right ζs The average values ​​of and It is expressed as follows:

[0117]

[0118] Then you can get A (t)left δs and and A (t)right ζs and The errors are as well as Since t=1,2, it is divided into two cases as follows:

[0119]

[0120] In the formula, ||·|| F is the matrix norm.

[0121] Possible equinox The corresponding weighted error It is expressed as follows:

[0122]

[0123] Traverse all possible bisection points When the corresponding weighted error When it is the smallest, take the possible bisection point The optimal bisection point for this bisection node.

[0124] Step 6: Prediction function model construction

[0125] by As a prediction node, M0 in each prediction node Ns Corresponding A (t) Ns The average value of the possible prediction value That is to complete the prediction function sub-model PRE (t) s 's construction.

[0126] Bring in the test set X Test s The predicted value A can be obtained Test(t) (Pre)φs The prediction subset Y Test(t) (Pre)s It is expressed as follows:

[0127] Y Test(t) (Pre)s =[A Test(t) (Pre)1s A Test(t) (Pre)2s … A Test(t)(Pre)L-1s A Test(t) (Pre)Ls ]

[0128] Find the test subset Y Test(t) s With the prediction subset Y Test(t) (Pre)s Error SE last It is expressed as follows:

[0129]

[0130] If the error SE last If the accuracy requirement is met, it is judged to be an effective prediction function sub-model PRE (t) s , otherwise, repeat step 5.

[0131] Merge all valid prediction function sub-models PRE (t) s The prediction result A (t) (Pre)s And take the average value to get A (t) (Pre) , that is, completing the prediction function model PRE (t) all 's construction.

[0132] Step 7: Obtaining the rejection matrix

[0133] Known Cable force collection during construction phase It is expressed as follows:

[0134]

[0135] Assume that the prediction function model PRE (t) all The obtained Cable force aggregate individual during construction phase The target prediction value The known target ideal value is A (t) (wanted) , we can get Rejection Since t=1,2, it is divided into two cases as follows:

[0136]

[0137] Therefore, the Rejection degree matrix of cable force set during construction phase It is expressed as follows:

[0138]

[0139] Step 8: Cable force assembly and grading during construction phase

[0140] When a Cable force collection during construction phase Rejection and are less than or equal to Target rejection and and exists Less than Then it is called Complete victory Recorded as

[0141] For each Cable force aggregate individual during construction phase exist:

[0142]

[0143] in, for the number of for The number of

[0144] but Collection It is expressed as follows:

[0145]

[0146] For the first During the construction phase, the cable force is graded. The individual is recorded as The quantity is θ F(1) ,χ=1,2,...,θ F(1) -1,θ F(1) , stored in the first-level hierarchical set F(1), expressed as follows:

[0147]

[0148] Assume that there are l levels in total. For the Π-th level hierarchical set F(Π), Π=2,3,...,l-1,l, examine each individual in the hierarchical set F(Π-1) of Statistics Collection middle The number of individuals is θ F(Π) , recorded as Stored in the hierarchical set F(Π), expressed as follows:

[0149]

[0150] The classification is continued until the l-th classification set is obtained.

[0151] Step 9: Obtaining the density operator matrix

[0152] For the H-th level hierarchical set F(H), according to The size of F(H) Sort by, H=1,2...,l-1,l,λ=1,2,...,θ F(H) -1,θ F(H) ,set up Density operator It is expressed as follows:

[0153]

[0154] Get the density operator of the H-th level hierarchical set F(H) It is expressed as follows:

[0155]

[0156] Step 10: Generate the Cable force collection during construction phase

[0157] No. The individual in each level of the cable force set F(H) during the construction phase Pair each pair randomly. If the number of individuals in a certain level of classification set F(H) is odd, the individual with the smallest density operator will be counted into the next level of classification set for pairing, forming 250 pairs. and Each pair and Generate and Constructing the offspring matrix It is expressed as follows:

[0158]

[0159] Among them, r~U([0,1]).

[0160] The offspring matrix With the Generation population The merger constitutes temporary population The temporary population individuals are denoted as L=1,2,...,999,1000, which is expressed as follows:

[0161]

[0162] For the first temporary population Repeat steps 7 to 9 to store each level of collection in the Generation population, until it is stored in a certain level set If the number of individuals in the cable force set of the first construction stage exceeds 500, the hierarchical set of this level is stored according to the size of the density operator, and the ones with larger density operators are given priority. The cable force set during the construction phase is It is expressed as follows:

[0163]

[0164] Step 11: Obtaining the optimal cable force set during construction

[0165] Repeat steps 7 to 10 to get Generation population And satisfy The first Generation population The first-level classification set F(1) is taken as the optimal cable force set in the construction stage.

[0166] Example

[0167] This example is implemented for the Yongjiang Bridge section of the Ningbo-Wenzhou Expressway. The specific data are as follows:

[0168] S1. Establish a finite element model of a cable-stayed bridge. Through reverse dismantling simulation, the cable force matrix S0 in the initial construction stage is obtained as follows:

[0169]

[0170] S2, give the cable force variation range [-ΔS0 n ,ΔS0 n ]as follows:

[0171]

[0172] Randomly select a cable force change value Δs0 in each cable force change range n(N) Get the cable force change matrix Δs0 (N) , some data are given here as follows:

[0173]

[0174] Then we get the cable force set M0 in the initial construction stage. Some data are given here as follows:

[0175]

[0176] S3, using each initial construction stage cable force set individual M0 NPerform forward analysis to obtain the bridge cable force matrix A (1) N and the bridge anchor point elevation matrix A (2) N , some data are given here as follows:

[0177]

[0178] M0 1 ,M0 2 ,...,M0 499 ,M0 500 And the corresponding A (t) 1 ,A (t) 2 ,...,A (t) 499 ,A (t) 500 As the original data set, t = 1, 2, repeat 750 times to randomly extract all 500 initial construction stage cable force set individuals from the original data set with replacement Record Composed training set X s , where the training set X is given 1 Some of the data are as follows:

[0179]

[0180] The corresponding A (t) N , denoted as A (t) Ns , forming the training subset Y (t) s , where the training subset Y is given (1) 1 and Y (2) 1 Some of the data are as follows:

[0181]

[0182] Suppose that the number of and the corresponding A (t) N There are L pairs, represented by M Test φN and A Test(t) φN ,φ=1,2,...L-1,L, forming the test set X Test s With the test subset Y Test(t) s , where the test set X is given Test 1Some of the data are as follows:

[0183]

[0184] And the test subset Y Test ( 1) 1 and Y Test ( 2) 1 Some of the data are as follows:

[0185]

[0186] S4, set is a binary feature, For each training set X s As a binary source node Perform 20 layers of binary splits, and each binary node is represented as

[0187] For example, for a binary node from In the random sampling, 5 different binary features are extracted to obtain the feature subset Then we get the corresponding characteristic cable force subset Some of the data are given here as follows:

[0188]

[0189] The feature force subset Arrange each row in order of size and calculate the middle value of two adjacent elements in each row as a possible bisection point Get a subset of possible bisection points Some of the data are given here as follows:

[0190]

[0191] S5. For the possible bisection point 3082.91, in the following bisection nodes:

[0192]

[0193] Available and as follows:

[0194]

[0195] Get the verification matrix and as follows:

[0196]

[0197] Validation Matrix and as follows:

[0198]

[0199] average value and as follows:

[0200]

[0201] average value and as follows:

[0202]

[0203] The available errors are as follows:

[0204]

[0205] Possible equinox The corresponding weighted error as follows:

[0206]

[0207] Traverse all possible bisection points When the corresponding weighted error When it is the smallest, take the possible bisection point The optimal bisection point for this bisection node.

[0208] S6, As a prediction node, M0 in each prediction node Ns Corresponding A (t) Ns The average value of the possible prediction value That is to complete the prediction function sub-model PRE (t) s 's construction.

[0209] Bring in the test set X given in step 3 Test 1 Part of the data gets the predicted subset Y Test(1) (Pre)1 and Y Test(2) (Pre)1 , some data are given here as follows:

[0210]

[0211] Find the test subset Y Test(t) s With the prediction subset Y Test(t) (Pre)s Error SElast as follows:

[0212]

[0213] Complete the prediction function model PRE (t) all 's construction.

[0214] S7, known Cable force collection during construction phase Some of the data are given here as follows:

[0215]

[0216] Assume that the prediction function model PRE (t) all The obtained Cable force aggregate individual during construction phase The target prediction value and as follows:

[0217]

[0218] Known target ideal value A (t) (wanted) as follows:

[0219]

[0220] Available Rejection Therefore, the Rejection degree matrix of cable force set during construction phase as follows:

[0221]

[0222] S8, take the part given in step 7 As an example, the cable force data of the first construction stage is Cable force collection during construction phase Rejection and are less than or equal to Target rejection and and exists Less than Then it is called Complete victory Recorded as

[0223] For the first During the construction phase, the cable force is graded. The individual is recorded as The quantity is θ F(1) ,χ=1,2,...,θ F(1) -1,θ F(1) , stored in the first-level hierarchical set F(1) as follows:

[0224]

[0225] There are 5 levels in total, and the 2nd, 3rd, 4th and 5th level classifications are as follows:

[0226]

[0227] S9. For the H-th level hierarchical set F(H), according to The size of F(H) Sort and get the density operator of the H-th level hierarchical set F(H) as follows:

[0228]

[0229] Here, since there is only one Cable force aggregate individual during construction phase No sorting is required, and the density operators are all ∞.

[0230] S10, take the part given in step 7 As an example, the cable force data of the construction stage is used as an example, and the child matrix as follows:

[0231]

[0232] The offspring matrix With the Generation population The merger constitutes temporary population Some of the data are given here as follows:

[0233]

[0234] For the first temporary population Repeat steps 7 to 9 to get Cable force collection during construction phase Some of the data are as follows:

[0235]

[0236] S11, repeat steps 7 to 10 to obtain Generation population Some of the data are as follows:

[0237]

[0238] satisfy The first Generation population The first-level classification set F(1) is taken as the optimal cable force set during the construction phase. Some data are given here as follows:

[0239]

[0240] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other configurations without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations coming within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0241] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A big data-driven multi-objective intelligent optimization method for cable force in the construction phase of a cable-stayed bridge, characterized by: The following steps are involved: Step 1: Determine the cable force during the initial construction phase The finite element model of the cable-stayed bridge is established. There are π batches of tensioning during the construction process. The initial cable force S0 of each batch of tensioning is obtained through reverse dismantling simulation. n , n=1,2,...,Π-1,Π, the cable force matrix in the initial construction stage is expressed as follows: S0=[S0 1 S0 2 … S0 Π-1 S0 Π ] T ; Step 2: Generating cable force sets during the initial construction phase For the initial construction stage, the cable force aggregate The generation of cable force is set to [-ΔS0 n ,ΔS0 n ], randomly select the cable force change value Δs0 n(N) , we get the cable force change matrix Δs0 (N) It is expressed as follows: Δs0 (N) =[Δs0 1(N) Δs0 2(N) … Δs0 Π-1(N) Δs0 Π(N) ] T Then the cable force set M0 in the initial construction stage is expressed as follows: where, Δs0 n(N) ~U([-ΔS0 n ,ΔS0 n ), ΔS0 n =0.1×S0 n , N={1,2,...,499,500}, M0 N =S0+Δs0 (N) , C0 n(N) =S0 n +Δs0 n(N) ; Step 3: Finite element model analysis to form a data set Use each initial construction stage cable force aggregate individual Perform forward analysis to obtain the bridge cable force matrix A (1) N and the bridge anchor point elevation matrix A (2) N ; M0 1 ,M0 2 ,...,M0 499 ,M0 500 And the corresponding A (t) 1 ,A (t) 2 ,...,A (t) 499 ,A (t) 500 As the original dataset, t=1,2, repeat randomly extracting all individuals from the original data set with replacement Record Composed training set X s , corresponding to A (t) N Denoted as A (t) Ns Form the training subset Y (t) s , suppose that and the corresponding A (t) N There are L pairs and they are represented by M Test øN and A Test(t) øN They form the test set X Test s With the test subset Y Test(t) s , ø=1,2,...L-1,L, which is expressed as follows: AND (t) Ns =[A (t) 1s TO (t) 2s …TO (t) 499s TO (t) 500s ] AND Test(t) s =[A Test(t) 1s TO Test(t) 2s …TO Test(t) L-1s TO Test(t) Ls ]; Step 4: Obtaining a subset of possible binary points Set binary features For each training set X s As a binary source node Perform binary division, and each binary node is represented as P represents the level of the node, L P Indicates the sequence number of the node in this layer; For a binary node from In the random sampling, 5 different binary features are extracted to obtain the feature subset It is expressed as follows: Then we get the corresponding characteristic cable force subset It is expressed as follows: The feature force subset Arrange each row in order of size and calculate the middle value of two adjacent elements in each row as a possible bisection point w={1,2,3,4,5},d={1,2,...,498,499},get the possible bisection point subset It is expressed as follows: Step 5: Determine the optimal bisection point For possible bisection points In the bisection node, if M0 Ns in but If M0 Ns in but Thus we get and Medium M0 Ns Corresponding A (t) Ns The verification matrix is ​​constructed separately and Will and A in (t) Ns A (t)left δs With A (t)right ζs , δ=1,2,...m-1,m, ζ=1,2,...n-1,n, then A (t)left δs and A (t)right ζs The average values ​​of and It is expressed as follows: Then you can get A (t)left δs and and A (t)right ζs and The errors are as well as Since t=1,2, it is divided into two cases as follows: Possible equinox The corresponding weighted error It is expressed as follows: Traverse all possible bisection points When the corresponding weighted error When it is the smallest, take the possible bisection point The optimal bisection point for this bisection node; Step 6: Prediction function model construction by As a prediction node, M0 in each prediction node Ns Corresponding A (t) Ns The average value of the possible prediction value That is to complete the prediction function sub-model PRE (t) s The construction of Bring in the test set X Test s The predicted value A can be obtained Test(t) (Pre)øs The prediction subset Y Test(t) (Pre)s It is expressed as follows: AND Test(t) (Pre)s =[A Test(t) (Pre)1s TO Test(t) (Pre)2s … TO Test(t) (Pre)L-1s TO Test(t) (Pre)Ls ] Find the test subset Y Test(t) s With the prediction subset Y Test(t) (Pre)s Error SE last It is expressed as follows: If the error SE last If the accuracy requirement is met, it is judged to be an effective prediction function sub-model PRE (t) s , otherwise, repeat step 5; Merge all valid prediction function sub-models PRE (t) s The prediction result A (t) (Pre)s And take the average value to get A (t) (Pre) , that is, completing the prediction function model PRE (t) all The construction of Step 7: Obtaining the rejection matrix Known Cable force collection during construction phase It is expressed as follows: Assume that the prediction function model PRE (t) all The obtained Cable force aggregate individual during construction phase The target prediction value The known target ideal value is A (t) (wanted) ,get Rejection Since t=1,2, it is divided into two cases as follows: Therefore, the Rejection degree matrix of cable force set during construction phase It is expressed as follows: Step 8: Cable force assembly and grading during construction phase When a Cable force collection during construction phase Rejection and are less than or equal to Target rejection and and exists Less than Then it is called Complete victory Recorded as A, B∈N; For each Cable force aggregate individual during construction phase exist: in, for the number of for the number of but Collection It is expressed as follows: For the first During the construction phase, the cable force is graded. The individual is recorded as The quantity is θ F(1) ,χ=1,2,...,θ F(1) -1,θ F(1) , stored in the first-level hierarchical set F(1), expressed as follows: Assume that there are l levels in total. For the Π-th level hierarchical set F(Π), Π=2,3,...,l-1,l, examine each individual in the hierarchical set F(Π-1) of Statistics Collection middle The number of individuals is θ F(Π) , recorded as Stored in the hierarchical set F(Π), expressed as follows: Continue grading until the l-th level grading set is obtained; Step 9: Obtaining the density operator matrix For the H-th level hierarchical set F(H), according to The size of F(H) Sort by, H=1,2...,l-1,l,λ=1,2,...,θ F(H) -1,θ F(H) ,set up Density operator It is expressed as follows: Get the density operator of the H-th level hierarchical set F(H) It is expressed as follows: Step 10: Generate the Cable force collection during construction phase No. The individual in each level of the cable force set F(H) during the construction phase Pair each pair randomly. If the number of individuals in a certain level of classification set F(H) is odd, the individual with the smallest density operator will be counted into the next level of classification set for pairing, forming 250 pairs. and Each pair and generate and Constructing the offspring matrix It is expressed as follows: r～U([0,1]) The offspring matrix With the Generation population The merger constitutes temporary population The temporary population individuals are denoted as It is expressed as follows: For the first temporary population Repeat steps 7 to 9 to store each level of collection in the Generation population, until it is stored in a certain level set If the number of individuals in the cable force set of the first construction stage exceeds 500, the hierarchical set of this level is stored according to the size of the density operator, and the ones with larger density operators are given priority. The cable force set during the construction phase is It is expressed as follows: Step 11: Obtaining the optimal cable force set during construction Repeat steps 7 to 10 to get Generation population And satisfy The first Generation population The first-level classification set F(1) is taken as the optimal cable force set in the construction stage.

Images