Tubular pile defect identification method based on swarm intelligence algorithm

Abstract

The invention relates to the technical field of low-strain method quantitative identification, in particular to a swarm intelligence algorithm-based tubular pile defect identification method, which mainly comprises the following steps of: acquiring measurement speed time-history response curve data V of a tubular pile; constructing a tubular pile defect identification program, and inputting measurement speed time history response curve data, empirical learning parameters, model fixed parameters and model change parameters in the tubular pile defect identification program; and the tubular pile defect identification program carries out iterative calculation to obtain an optimal candidate solution. According to the method, quantitative identification is conducted on the defect position of the prestressed concrete pipe pile under the homogeneous soil covering condition, quantitative identification data are provided and comprise the exciting force amplitude F, the pulse width L, the reference integrity coefficient beta, the pile-soil friction coefficient mu and the pile-soil damping ratio xi, the precision and reliability of a traditional low-strain method are improved, and the method is suitable for large-scale popularization and application. The defect that a traditional method can only qualitatively judge the integrity of the pipe pile is overcome.

Description

Technical Field

[0001] The invention relates to the technical field of low-strain quantitative identification, and in particular to a method for identifying pipe pile defects based on a swarm intelligence algorithm. Background Art

[0002] Pile foundation is an important part of building structure. Pile defects will pose serious safety hazards to the use of structure. Accurately judging the damage of pile foundation is a necessary technical measure to ensure the safe use, maintenance and management of structure. Low strain reflection wave method is to generate stress wave by applying a dynamic load of transient impact force on the top of pile. Stress wave propagates downward along the pile body. When there is obvious wave impedance interface (such as pile bottom, fracture or segregation, mud inclusion, etc.) or pile cross-sectional area change (such as necking or expansion) in the pile body, reflection wave will be generated. By using specific instruments to receive, amplify, filter and process data, reflection information from different parts of pile body can be identified. By analyzing and calculating the reflection information, pile damage and pile body integrity can be identified. However, there are few studies on quantitative analysis and corresponding reliability of pile damage identification by low strain reflection wave method in China, which restricts the further application of this method. How to quickly and accurately identify pile damage by non-destructive dynamic testing method is a research hotspot for many domestic and foreign scholars and engineering technicians.

[0003] In view of the above problems, the present invention combines an advanced nonlinear brute force solution algorithm - a swarm intelligence optimization algorithm to quantitatively identify pipe pile defects, thereby improving the recognition accuracy and reliability of the low strain method. The swarm intelligence algorithm is a type of intelligent optimization method based on group search that is implemented by simulating a certain behavior of a biological group. Its advantages are strong global search capabilities, no need for sensitivity information, and insensitivity to initial values. Common algorithms include particle swarm optimization (PSO), cloud model fruit fly algorithm (CMFOA), JAYA algorithm, squirrel search algorithm, and experience learning algorithm. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a pipe pile defect identification method based on a swarm intelligence optimization algorithm, which can quantitatively identify the defect locations of prestressed concrete pipe piles under homogeneous covering conditions, improve the accuracy and reliability of the traditional low-strain method, and solve the shortcoming that the traditional method can only qualitatively judge the integrity of the pipe piles.

[0005] To achieve the above object, the present invention can be implemented through the following technical solutions:

[0006] A method for identifying pipe pile defects based on a swarm intelligence algorithm, the method comprising:

[0007] Obtaining the measured velocity time history response curve data V of the pipe pile;

[0008] Inputting the measured velocity time history response curve data V, experience learning parameters, model fixed parameters and model variable parameters into the pipe pile defect recognition program;

[0009] The pipe pile defect identification program performs iterative calculations to obtain the optimal candidate solution.

[0010] Furthermore, the experience learning parameters include the maximum number of iterations Iter max , the maximum population size Np, the upper bound of the decision variable X max , the lower bound of the decision variable X min and the maximum number of dimensions n, the upper bound of the decision variable X max and the lower bound of the decision variable X min The setting is performed according to the value range of the model variation parameter.

[0011] Furthermore, the fixed parameters of the model include the radius R of the soil around the pile, the buried depth h of the soil around the pile, the diameter d of the pipe pile, the wall thickness t of the pipe pile, the length l of the pile, the number of parts N of the pile body, the elastic modulus E1 of the pile body material, the elastic modulus E2 of the soil around the pile, the Poisson's ratio v1 of the pile body material, the Poisson's ratio v2 of the soil around the pile, the density ρ1 of the pile body material, the density ρ2 of the soil around the pile, the cohesion c, the friction angle φ, the longitudinal wave velocity C1 of the pile body material and the longitudinal wave velocity C2 of the soil around the pile, and the fixed parameters of the model do not change with the iterative calculation.

[0012] Furthermore, the model variation parameters include the exciting force amplitude F, the exciting force pulse width L, the pipe pile integrity coefficient set β, the pile-soil friction coefficient μ and the pile-soil damping ratio ξ, and the model variation parameters are optimized with iterative calculation.

[0013] Furthermore, the iterative calculation includes:

[0014] The algorithm population is initialized to obtain the initialized candidate solution X p,q ;

[0015] According to the initial candidate solution X p,q Start iterative calculation;

[0016] Each iteration iterates and calculates the candidate solutions according to the number of individuals p and the number of dimensions q in the population;

[0017] Iterate until the maximum number of iterations Iter is reached max , and obtain the optimal candidate solution.

[0018] Furthermore, each iteration iteratively calculates the candidate solutions according to the number of individuals p in the population and the number of dimensions q, including:

[0019] Let the number of individuals in the population p = 1;

[0020] According to the candidate solution X through matlabp,q Modify the model change parameters in the command stream file;

[0021] Abaqus calls the modified command stream file to generate the pile-soil model and processes the operation to obtain the calculated velocity time-history response curve data.

[0022] According to the measured speed time history response curve data V and the calculated speed time history response curve data Calculate candidate solution X p,q The fitness value f;

[0023] For candidate solution X p,q , randomly select another candidate solution X from the population (p≠1) l,q , used to generate potential search directions, according to the candidate solution X p,q The distance between the old and new positions is used to update the new position of the candidate solution through the model;

[0024] Iterate successively until the number of individuals p in the population reaches the maximum population size Np;

[0025] Entering the spatial enhanced dimension search phase, iteratively calculates the number of dimensions q to obtain better candidate solutions;

[0026] Iterate in sequence until the number of dimensions q reaches the maximum number of dimensions n.

[0027] Further,

[0028] The candidate solution X is initialized according to the following formula: p,q :

[0029] X p,q =X min +rand·(X max -X min ),p=1,2,…,Np,q=1,2,…,n (1)

[0030] X p,q =(F, L, β, μ, ξ); β = (β1, β2,…,β N ) T ;

[0031] Among them, X max is the upper bound of the decision variable, X min is the lower bound of the decision variable, rand is a uniformly distributed random number in the range of [0, 1], p is the number of individuals in the population, q is the number of dimensions, F is the amplitude of the exciting force, L is the pulse width of the exciting force, β is the set of integrity coefficients of the pile, μ is the pile-soil friction coefficient, and ξ is the pile-soil damping ratio.

[0032] Further, the candidate solution X is calculated according to the following formulap,q The fitness value f:

[0033]

[0034] in, is the calculated speed time history response curve data, and V is the measured speed time history response curve data.

[0035] Furthermore, for the candidate solution X p,q , randomly select another candidate solution X from the population (p≠1) l,q , used to generate potential search directions, according to the candidate solution X p,q The distance between the old and new positions is updated using the following two models:

[0036] Mode 1: Candidate solution X p,q The new position directly to the candidate solution X l,q Learning, the exploration radius changes dynamically with iterations:

[0037]

[0038] Mode 2: Based on candidate solution X l,q The experience of generating candidate solutions X near the original position p,q New location:

[0039]

[0040] Among them, rand is a function that takes a random value between [0,1], Iter is the current number of iterations, and Iter max is the maximum number of iterations;

[0041] In each iteration, in order to balance the exploration and exploitation capabilities of the algorithm during the search process, one of the two modes is randomly called;

[0042] If the fitness value of the new position is better than the old position, the position of the candidate solution is updated according to the following formula, otherwise the old position is retained:

[0043]

[0044] Furthermore, entering the spatial enhanced dimension search phase and iteratively calculating the number of dimensions q to obtain a better candidate solution includes the following steps:

[0045] Find the best candidate solution X among all candidate solutions iteratively calculated with the number of individuals in the population p = 1, 2, ..., Np best,q and the worst candidate solution X worst,q ;

[0046] According to the best candidate solution X best,q and the worst candidate solution Xworst,q , and changing the value of one dimension while maintaining the values ​​of other dimensions to generate new candidate solutions

[0047] Compare the fitness values ​​of the new candidate solutions with the original candidate solutions and retain the better candidate solutions;

[0048] The new candidate solution is generated by the following formula:

[0049]

[0050] Among them, X best,q is the best candidate solution, X worst,q is the worst candidate solution, Iter is the current iteration number, Iter max is the maximum number of iterations, rand is a uniformly distributed random number in the range [0, 1], and q is the number of dimensions.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention quantitatively identifies defect locations of prestressed concrete pipe piles under homogeneous covering soil conditions and provides quantitative identification data, including exciting force amplitude F, pulse width L, benchmark integrity coefficient β, pile-soil friction coefficient μ and pile-soil damping ratio ξ, thereby improving the accuracy and reliability of traditional low-strain methods and overcoming the shortcoming that traditional methods can only qualitatively determine the integrity of pipe piles. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a flow chart of the method of the present invention;

[0054] Figure 2 It is a schematic diagram of establishing a pile-soil model;

[0055] Figure 3 It is a schematic diagram of the defect types of the model pile;

[0056] Figure 4 It is the speed time history curve fitting comparison, algorithm iteration curve, and integrity coefficient curve. DETAILED DESCRIPTION

[0057] The present invention will be further described below in conjunction with the accompanying drawings and specific implementation methods. The terms such as "upper", "inner", "middle", "left", "right" and "one" cited in this specification are only for the convenience of description, and are not used to limit the scope of the present invention. The change or adjustment of their relative relationship shall also be regarded as the scope of the present invention without substantially changing the technical content.

[0058] The method for identifying pipe pile defects based on swarm intelligence algorithm described in the present invention is as follows: Figure 1 As shown, it includes the following stages:

[0059] Phase 1: Experimental data collection

[0060] Obtain the measured velocity time-history response curve data V of the pipe pile.

[0061] Step 1: Install the accelerometer on the top of the pile;

[0062] Step 2: Set the sampling time interval Δt and sampling frequency f s ;

[0063] Step 3: Select a suitable exciting hammer, strike the pile top to measure the velocity time-history response curve data V, and import the data into the pipe pile defect identification program.

[0064] This embodiment uses the OMEGA PIT-QFV low strain pile integrity tester to collect the measured velocity time history response curve data V of the pipe pile. The test system is mainly composed of a signal acquisition instrument, an accelerometer and an exciting hammer. The technical parameters of the instrument are shown in Table 1.

[0065] Table 1 Technical parameters of test equipment

[0066]

[0067] Phase 2: Construct a pipe pile defect recognition program, and input the measured velocity time history response curve data V, experience learning parameters, model fixed parameters and model variable parameters into the pipe pile defect recognition program.

[0068] Use the MATLAB program of the experience learning algorithm to build a pipe pile defect recognition program based on the experience learning algorithm. Enter the experience learning parameters, model fixed parameters and model variable parameters in the Matlab program. Run the pipe pile defect recognition program. During the algorithm iteration calculation process, Abaqus software is called for each calculation. The model variable parameters in the python command stream of Abaqus are modified according to the candidate solution to automate modeling, pre- and post-processing.

[0069] Step 4: Input the experience learning parameters (Iter max , Np, X max , X min , n), the main control parameters of the experience learning algorithm include the maximum number of iterations Iter max , the maximum population size Np, the upper bound of the decision variable X max , the lower bound of the decision variable X minand the maximum number of dimensions n. These parameters are the initialization parameters of the experience learning algorithm. The algorithm needs to pre-set the values ​​and value ranges of the initialization parameters. The upper bound Xmax of the decision variable is the set of the upper limits of the value ranges of all model change parameters; the lower bound Xmin of the decision variable is the set of the lower limits of the value ranges of all model change parameters. max and the lower bound of the decision variable X min Set according to the value range of the model change parameter.

[0070] Step 5: Input the fixed parameters of the Abaqus model (R, h, d, t, l, N, E1, E2, v1, v2, ρ1, ρ2, c, φ, C1, C2). The fixed parameters of the model include the radius of the soil around the pile R, the buried depth of the soil around the pile h, the diameter of the pipe pile d, the wall thickness of the pipe pile t, the length of the pile l, the number of parts of the pile body N, the elastic modulus of the pile body material E1, the elastic modulus of the soil around the pile E2, the Poisson's ratio of the pile body material v1, the Poisson's ratio of the soil around the pile v2, the density of the pile body material ρ1, the density of the soil around the pile ρ2, the cohesion c, the friction angle φ, the longitudinal wave velocity of the pile body material C1, and the longitudinal wave velocity of the soil around the pile C2. The fixed parameters of the model do not change with the iterative calculation. The fixed parameters of the model are set according to the lithology of the soil layer and the size parameters of the pipe pile in the field test, which are consistent with the field test construction information of the first stage. The Poisson's ratio v2 of the soil around the pile, the density of the pile body material ρ1, the density of the soil around the pile ρ2, the cohesion c, and the friction angle φ are set according to the longitudinal wave velocity C1 of the pile body material, and the longitudinal wave velocity C2 of the soil around the pile is set according to experience.

[0071] For example: the pipe pile is concrete C80, the diameter of the pipe pile is 500mm, the pile length is 10m, and the soil around the pile is clay.

[0072] The initial fixed parameters of Abaqus are R = 5m, h = 10m, d = 0.5m, t = 0.125m, l = 10m, N = 20. Other model material property parameters are shown in Table 2.

[0073] Table 2 Model material property parameters

[0074]

[0075] Step 6: Input Abaqus model variation parameters (F, L, β, μ, ξ), which include exciting force amplitude F, exciting force pulse width L, pipe pile integrity coefficient set β, pile-soil friction coefficient μ. Pile-soil damping ratio ξ. The model variation parameters are set to an initial value based on the experience of field tests and soil conditions. They will be continuously optimized during the iterative calculation process to find the optimal value within the value range. The optimal candidate solution required by the present invention is the model variation parameters (F, L, β, μ, ξ).

[0076] Table 3 Algorithm parameter value settings

[0077]

[0078] Phase 3: Iterative Calculation

[0079] The algorithm population candidate solution X sought by the present invention p,q That is, the model variation parameters (F, L, β, μ, ξ), the candidate solution X p,q is a set of numbers, the number of individuals in the population is p, and the number of dimensions is q;

[0080]

[0081] Among them, p is the number of individuals in the population, q is the number of dimensions, F is the amplitude of the exciting force, L is the pulse width of the exciting force, β is the set of integrity coefficients of the pipe pile, μ is the pile-soil friction coefficient, and ξ is the pile-soil damping ratio.

[0082] Step 7: Initialize the algorithm population. The initial population position is randomly generated. The initialized candidate solution X is obtained according to the following formula: p,q :

[0083] X p,q =X min +rand·(X max -X min ),p=1,2,…,Np,q=1,2,…,n (1)

[0084] X p,q =(F, L, β, μ, ξ); β = (β1, β2,…,β N ) T ;

[0085] Among them, X max is the upper bound of the decision variable, X min is the lower bound of the decision variable, rand is a uniformly distributed random number in the range [0, 1], p is the number of individuals in the population, q is the number of dimensions, F is the amplitude of the exciting force, L is the pulse width of the exciting force, β is the set of pile integrity coefficients, μ is the pile-soil friction coefficient, and ξ is the pile-soil damping ratio. In the first iterative calculation, the candidate solution X is initialized. p,q to start the calculation.

[0086] The fitness value of the candidate solution f=(f1,f2,…,f Np ) is calculated according to the following objective function:

[0087] f p = obj(X p,1 ,X p,2 ,…,X p,n ),p=1,2,…,Np (2)

[0088] The fitness value represents the degree of fit between the calculated candidate solution and the actual situation. The smaller the fitness value, the more accurate the fit, and the more consistent the calculated candidate solution is with the actual situation.

[0089] Step 8: Start iterative calculation, the number of iterations iter = 1.

[0090] Step 9: Let the number of individuals in the population p = 1;

[0091] Step 10: Call the Python interpreter built into the Abaqus program and calculate the candidate solution X. p,q Modify the modeling command flow (modify the model change parameters, and keep other parameters unchanged), automatically generate the pile-soil model through the command flow, and perform processing operations to obtain the calculation speed time history response curve data

[0092] Abaqus software is used for finite element modeling. The pre- and post-modeling processing can be realized by running command streams through the built-in Python interpreter. The model soil (foundation soil) is assumed to be a viscoelastic Voigt body. The constitutive model of the pile concrete is a linear elastic model.

[0093] Use Abaqus finite element software to build a three-dimensional pile-soil model, and show the main modeling process Figure 2 . Establish a three-dimensional pile-soil model and perform processing operations to obtain the calculated velocity time history response curve data. The steps are as follows:

[0094] (1) Establishing the model entity components: The unit grid is divided using eight-node reduced-integral linear hexahedral units. The model entity components are established based on seven parameters: the radius of the pile R, the depth of the pile h, the diameter of the pile d, the wall thickness of the pile t, the length of the pile l, and the number of pile body components N.

[0095] (2) Define material properties: Assign material properties to different components of the single-layer soil and pipe pile based on the elastic modulus of the pile material E1, the elastic modulus of the soil around the pile E2, the Poisson's ratio of the pile material v1, the Poisson's ratio of the soil around the pile v2, the density of the pile material ρ1, the density of the soil around the pile ρ2, the cohesion c, the friction angle φ, the longitudinal wave velocity of the pile material C1, the longitudinal wave velocity of the soil around the pile C2, and the pile-soil damping ratio ξ.

[0096] (3) Modify the material density properties of different units of the pile according to the pile integrity coefficient set β. Simulating pile body defects will cause impedance changes. Pile body defects will cause the cross-sectional wave impedance to decrease. The wave impedance calculation formula is Z = ρcA. In order to simulate the changes in pile body defects, numerical modeling defects are used. According to formula (3), the pile defects are equivalent to the reduction of material density:

[0097]

[0098] Where β is the pile integrity coefficient, Z d is the wave impedance of the medium after the stress wave passes through the defect interface, ρ d is the density of the medium after the stress wave passes through the defect interface, Z0 is the wave impedance of the medium before the stress wave passes through the defect interface, and ρ0 is the density of the medium after the stress wave passes through the defect interface. The model of this embodiment uses density reduction as the damage model.

[0099] When modeling, the pile foundation is divided into N = 20 parts in equal proportion, and the initial integrity coefficient β of each part is set. N =1, N = 1, 2…20, input the complete density ρ, and modify the component unit material mass density ρ N =ρ×β N Thus changing the cross-sectional wave impedance Z N =ρ N cA.

[0100] (4) The pile-soil interaction is defined using the surface-to-surface contact of Abaqus, with the tangential contact model selected as penalty friction and the normal contact model selected as hard contact (the pile bottom is considered as a free end without contact with the soil);

[0101] (5) Apply an exciting force load p(t) to the top of the pile. The exciting force load p(t) can be obtained according to the formula corresponding to F and L in the model variation parameters. The boundary conditions of the soil around the pile constrain the lateral displacement on the two sides and the vertical displacement on the bottom. The vibration source of the low-strain reflection wave method can be equivalent to a transient pulse signal, and its characteristic indicators mainly include parameters such as the exciting force amplitude and pulse width. The waveform of the pulse load excited by the exciting hammer is close to the shape of the raised cosine pulse wave function and is represented by the raised cosine pulse function. The calculation method of the exciting force load p(t) is as follows:

[0102]

[0103] Where t is the time variable, p(t) is the pulse load, F is the exciting force amplitude, and L is the pulse width. The two parameters F and L can be calculated through the characteristic points of the measured vibration time domain curve.

[0104] (6) Dynamic analysis was used, with the analysis time set to 13 ms and the integration step set to 0.01 ms.

[0105] (7) After the calculation is completed, the time-history response curve data of the pile top calculation velocity of the pipe pile model is extracted. The calculation results include time-history data such as velocity, acceleration, stress and strain of all nodes, which can be output as needed and set through command streams.

[0106] Step 11: Combine and calculate the speed time history response curve data and the measured speed time history response curve data V, we can construct the following objective function, that is, calculate the standard deviation of the measured speed time history response curve data, and calculate the candidate solution X through the following objective function p,q The fitness value f:

[0107]

[0108] in, is the calculated speed time history response curve data, and V is the measured speed time history response curve data.

[0109] Candidate Solution X p,q is a set of numbers, which contains F, L, β, u and ξ. In each iteration, the fitness value of the candidate solution must be calculated, that is, the fitness value is calculated by formula (5).

[0110] Where: f is the candidate solution X of the experience learning algorithm p,q The fitness value of X p,q represents the candidate solution in the empirical learning algorithm; F is the exciting force amplitude, L is the exciting force pulse width, β is the set of pile integrity coefficients, μ is the pile-soil friction coefficient, and ξ is the pile-soil damping ratio.

[0111] Step 12: Generate a new position of the population. According to the learning strategy based on the experience of other solutions, the population candidate solution is considered to move to a better position around its initial position. p,q , randomly select another candidate solution X from the population (p≠1) l,q , X l,q is the candidate solution of the qth dimension of the lth individual in the population obtained in the previous iteration, l is a number randomly selected from 2, 3, ..., Np to generate potential search directions. In the first iteration, select X in the initialized population. l,q According to the candidate solution X p,q The distance between the old and new positions is updated using the following two models:

[0112] Mode 1: Candidate solution X p,q The new position directly to the candidate solution X l,q Learning, the exploration radius changes dynamically with iterations:

[0113]

[0114] Mode 2: Based on candidate solution X l,q The experience of generating candidate solutions X near the original position p,q New location:

[0115]

[0116] Among them, rand is a function that takes a random value between [0,1], Iter is the current number of iterations, and Iter max is the maximum number of iterations.

[0117] In these two modes, the purpose of mode 1 is to explore candidate solutions in the feasible space, thereby improving the algorithm's exploration ability and avoiding falling into the local optimal solution; mode 2 is to improve the algorithm's convergence performance, thereby enhancing the algorithm's development ability. In each iteration, in order to balance the algorithm's exploration and development capabilities during the search process, one of the two modes is randomly called.

[0118] If the fitness value of the new position is better than the old position, the position of the candidate solution is updated, otherwise the old position is retained:

[0119]

[0120] Step 13: Determine whether the number of individuals in the conditional population p < Np. If so, p = p + 1, and return to step 10. If p = Np, enter the spatial search enhancement phase.

[0121] Spatial search enhancement phase: During the search process of the empirical learning algorithm, all dimensions of each candidate solution are updated simultaneously. However, changes in one-dimensional variables may have a negative impact on variables in other dimensions, resulting in poor convergence performance in each dimension. In order to further enhance the in-depth search in each dimension, the following steps are taken in each iteration:

[0122] (1) Find the best candidate solution X among all candidate solutions iteratively calculated in a population with p = 1, 2, ..., Np. best,q and the worst candidate solution X worst,q ;

[0123] (2) Based on the best candidate solution X best,q and the worst candidate solution X worst,q , and changing the value of one dimension while maintaining the values ​​of other dimensions to generate new candidate solutions

[0124] The new candidate solution is generated by the following formula:

[0125]

[0126] Among them, rand is a function that takes random values ​​between [0,1], X best,q is the best candidate solution, X worst,q is the worst candidate solution, Iter is the current iteration number, Iter max is the maximum number of iterations and q is the number of dimensions.

[0127] (3) Compare the fitness value of the new candidate solution with that of the original candidate solution and retain the better candidate solution, that is, the candidate solution with the smaller fitness value;

[0128] (4) Repeat steps (1) and (3) for other dimensions.

[0129] Step 14: Determine the condition Iter <Iter max , if it is established, return to step 9. If Iter=Iter max Output current It is the optimal candidate solution and the iteration ends.

[0130] Step 15: Draw the fitting comparison curve, integrity coefficient set β curve and iteration curve.

[0131] The present invention is based on the method for identifying pipe pile defects of the swarm intelligence algorithm, and the identification results include the excitation force amplitude F, the pulse width L, the integrity coefficient set β, the pile-soil friction coefficient μ, and the pile-soil damping ratio ξ. The method for identifying pipe pile defects based on the empirical learning algorithm can obtain quantitative identification data, with high inversion accuracy and fast convergence speed. It can accurately invert the position range and degree of model pile defects, realize the quantitative identification of foundation pile defects under a single homogeneous covering soil, improve the accuracy and reliability of the low strain method, and provide a technical reference for further research on the quantitative identification of foundation pile defects under complex geological conditions in actual engineering. According to the identification results, a comparative fitting curve, an integrity coefficient β curve and an iteration curve can also be drawn. From the fitting curve, it can be seen whether the calculation result is close to the actual situation and the degree of fitting. From the integrity system β curve, it can be seen that the specific defect degree of the pipe pile (i.e., quantitative identification) can be seen. From the iteration curve, it can be seen how the convergence is and whether the iterative calculation has been completed. At the same time, it can also provide relevant parameters such as the maximum number of iterations for the next identification inversion calculation.

[0132] In order to study the influence and effect of the low-strain method on the integrity judgment of a single homogeneous covered pipe pile, and to compare the recognition effect of the traditional method and the quantitative identification program of pile defects based on the swarm intelligence algorithm, the present invention designs an indoor comparison test of the low-strain method for pipe piles. The nylon hand hammer is used to test the time-history curve data V of the pile top velocity, and the curve is analyzed to summarize the law of the low-strain method for single and double defect integrity judgment. According to the test requirements, 10 model piles (C80 hollow prestressed concrete pipe piles) were designed and manufactured, and the design parameters of the model piles are shown in Table 4. The diameter is 500mm (wall thickness 125mm), of which 2 piles are 8.0m long and 8 piles are 10.0m long. The defect types are divided into 2, 3, and 4 types of defects, and the integrity coefficients are β, respectively. 2 =93.6%, β 3 =66.9%, β 4 =50.0%, the pile impedance is changed by cutting the defect, and the model pile defect is shown in Figure 2 Figure 3The soil around the pile is clay soil, and the pile bottom is tested in a free pile state.

[0133] Table 4 Model pile design parameters

[0134]

[0135]

[0136] The pipe pile defect recognition program based on the experience learning algorithm was used to identify and invert 10 model defect piles. The parameter settings are shown in Table 3. The initial parameters of the experience learning algorithm are set to Iter max =100, Np=500, X max and X min The Abaqus model variable parameters are set according to the value range; the Abaqus model variable parameters are initially F = 2.6 kN and L = 1.7 ms, β n =1, μ=0.5, ξ=0.1; the initial fixed parameters of Abaqus are R=5m, h=10m, d=0.5m, t=0.125m, l=10m, N=20. Other model material property parameters are shown in Table 2.

[0137] Finally, the inversion curve fitting comparison, iteration curve, and unit integrity coefficient curve of 10 model piles obtained by iterative calculation of the pipe pile defect identification program are shown in Figure 4 , the algorithm identification parameters are shown in Table 5. As can be seen from the figure, the measured curves of the 10 model piles and the calculated curves of the finite element model inversion can basically fit, and the algorithm iteration curves basically converge at the 60th iteration, the fitness value is close to 0, and the algorithm converges quickly. The unit integrity coefficient β curve basically corresponds to the position of the design defect of the foundation pile, and the values ​​are close, with an error of no more than 10%. Whether it is a complete pile, a single or double defect pile, the defect position and degree can be accurately identified, and accurate quantitative analysis results can be provided. The inversion method has high accuracy and fast convergence speed. Considering the influence of measurement error, noise, model making error and pile heterogeneity, the inversion results basically reflect the real situation of the indoor model. The quantitative identification of pile defects under single homogeneous covering soil is basically achieved, which provides a reference for further research on the quantitative identification of pile defects under complex geological conditions in actual engineering.

[0138] Table 5 Summary of algorithm recognition parameter results

[0139]

[0140]

[0141] The implementation methods of the present invention are not limited to this. According to the above content of the present invention, using common technical knowledge and customary means in the field, without departing from the above basic technical ideas of the present invention, the present invention can also make other various forms of modification, replacement or combination, all of which fall within the scope of protection of the present invention.

Claims

1. A method for identifying pipe pile defects based on swarm intelligence algorithm, characterized in that: The method comprises: Obtaining the measured velocity time history response curve data V of the pipe pile; Inputting the measured velocity time history response curve data V, experience learning parameters, model fixed parameters and model variable parameters into the pipe pile defect recognition program; The pipe pile defect identification program performs iterative calculations to obtain the optimal candidate solution.

2. The method for identifying pipe pile defects according to claim 1, characterized in that: The experience learning parameters include the maximum number of iterations Iter max , the maximum population size Np, the upper bound of the decision variable X max , the lower bound of the decision variable X min and the maximum number of dimensions n, the upper bound of the decision variable X max and the lower bound of the decision variable X min The setting is performed according to the value range of the model variation parameter.

3. The method for identifying pipe pile defects according to claim 1, characterized in that: The fixed parameters of the model include the radius R of the soil around the pile, the buried depth h of the soil around the pile, the diameter d of the pipe pile, the wall thickness t of the pipe pile, the length l of the pile, the number of parts N of the pile body, the elastic modulus E1 of the pile body material, the elastic modulus E2 of the soil around the pile, the Poisson's ratio v1 of the pile body material, the Poisson's ratio v2 of the soil around the pile, the density ρ1 of the pile body material, the density ρ2 of the soil around the pile, the cohesion c, the friction angle φ, the longitudinal wave velocity C1 of the pile body material and the longitudinal wave velocity C2 of the soil around the pile. The fixed parameters of the model do not change with the iterative calculation.

4. The method for identifying pipe pile defects according to claim 1, characterized in that: The model variation parameters include the exciting force amplitude F, the exciting force pulse width L, the pipe pile integrity coefficient set β, the pile-soil friction coefficient μ and the pile-soil damping ratio ξ, and the model variation parameters are optimized with iterative calculation.

5. The method for identifying pipe pile defects according to claim 1, characterized in that: The iterative calculation includes: The algorithm population is initialized to obtain the initialized candidate solution X p,q ; According to the initial candidate solution X p,q Start iterative calculation; Each iteration iterates and calculates the candidate solutions according to the number of individuals p and the number of dimensions q in the population; Iterate until the maximum number of iterations Iter is reached max , and obtain the optimal candidate solution.

6. The method for identifying pipe pile defects according to claim 5, characterized in that: Each iteration is iteratively calculated according to the number of individuals p in the population and the number of dimensions q to obtain candidate solutions, including: Let the number of individuals in the population p = 1; Through matlab according to the candidate solution X p,q Modify the model change parameters in the command stream file; Abaqus calls the modified command stream file to generate the pile-soil model and processes the operation to obtain the calculated velocity time-history response curve data. According to the measured speed time history response curve data V and the calculated speed time history response curve data Calculate candidate solution X p,q The fitness value f; For candidate solution X p,q , randomly select another candidate solution X from the population (p≠1) l,q , used to generate potential search directions, according to the candidate solution X p,q The distance between the old and new positions is used to update the new position of the candidate solution through the model; Iterate successively until the number of individuals p in the population reaches the maximum population size Np; Entering the spatial enhanced dimension search phase, iteratively calculates the number of dimensions q to obtain better candidate solutions; Iterate in sequence until the number of dimensions q reaches the maximum number of dimensions n.

7. The method for identifying pipe pile defects according to claim 5, characterized in that: The candidate solution X is initialized according to the following formula: p,q : X p,q =X min +rand·(X max -X min ),p=1,2,…,Np,q=1,2,…,n (1) X p,q =(F,L,β,μ,ξ);β=(β1,β2,…,β N ) T ; Among them, X max is the upper bound of the decision variable, X min is the lower bound of the decision variable, rand is a uniformly distributed random number in the range of [0, 1], p is the number of individuals in the population, q is the number of dimensions, F is the amplitude of the exciting force, L is the pulse width of the exciting force, β is the set of integrity coefficients of the pile, μ is the pile-soil friction coefficient, and ξ is the pile-soil damping ratio.

8. The method for identifying pipe pile defects according to claim 6, characterized in that: The candidate solution X is calculated according to the following formula p,q The fitness value f: in, is the calculated speed time history response curve data, and V is the measured speed time history response curve data.

9. The method for identifying pipe pile defects according to claim 6, characterized in that: For candidate solution X p,q , randomly select another candidate solution X from the population (p≠1) l,q , used to generate potential search directions, according to the candidate solution X p,q The distance between the old and new positions is updated using the following two models: Mode 1: Candidate solution X p,q The new position directly to the candidate solution X l,q Learning, the exploration radius changes dynamically with iterations: Mode 2: Based on candidate solution X l,q The experience of generating candidate solutions X near the original position p,q New location: Among them, rand is a function that takes a random value between [0,1], Iter is the current number of iterations, and Iter max is the maximum number of iterations; In each iteration, in order to balance the exploration and exploitation capabilities of the algorithm during the search process, one of the two modes is randomly called; If the fitness value of the new position is better than the old position, the position of the candidate solution is updated according to the following formula, otherwise the old position is retained:

10. The method for identifying pipe pile defects according to claim 6, characterized in that: Entering the spatial enhanced dimension search phase, iteratively calculating the number of dimensions q to obtain a better candidate solution includes the following steps: Find the best candidate solution X among all candidate solutions iteratively calculated with the number of individuals in the population p = 1, 2, ..., Np best,q and the worst candidate solution X worst,q ; According to the best candidate solution X best,q and the worst candidate solution X worst,q , and changing the value of one dimension while maintaining the values ​​of other dimensions to generate new candidate solutions Compare the fitness values ​​of the new candidate solutions with the original candidate solutions and retain the better candidate solutions; The new candidate solution is generated by the following formula: Among them, X best,q is the best candidate solution, X worst,q is the worst candidate solution, Iter is the current iteration number, Iter max is the maximum number of iterations, rand is a uniformly distributed random number in the range [0, 1], and q is the number of dimensions.

Images