CN116796611A - Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network - Google Patents
Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network Download PDFInfo
- Publication number
- CN116796611A CN116796611A CN202311055386.1A CN202311055386A CN116796611A CN 116796611 A CN116796611 A CN 116796611A CN 202311055386 A CN202311055386 A CN 202311055386A CN 116796611 A CN116796611 A CN 116796611A
- Authority
- CN
- China
- Prior art keywords
- buckling
- force
- cable
- cable force
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 47
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 41
- 238000005457 optimization Methods 0.000 claims abstract description 36
- 238000013136 deep learning model Methods 0.000 claims abstract description 25
- 238000010276 construction Methods 0.000 claims abstract description 21
- 230000008569 process Effects 0.000 claims abstract description 15
- 229910000831 Steel Inorganic materials 0.000 claims abstract description 10
- 239000010959 steel Substances 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 29
- 230000006870 function Effects 0.000 claims description 26
- 239000000463 material Substances 0.000 claims description 25
- 238000009826 distribution Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 241001139947 Mida Species 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 241001125048 Sardina Species 0.000 claims description 4
- 230000006399 behavior Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 235000019512 sardine Nutrition 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 235000019688 fish Nutrition 0.000 claims description 3
- 241000251468 Actinopterygii Species 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 244000062645 predators Species 0.000 claims 2
- 230000000694 effects Effects 0.000 abstract description 4
- 238000012423 maintenance Methods 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 description 6
- 238000012614 Monte-Carlo sampling Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 241001137309 Jordanella floridae Species 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000022542 predatory behavior Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 210000003813 thumb Anatomy 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Bridges Or Land Bridges (AREA)
- Feedback Control In General (AREA)
Abstract
The invention relates to the field of load adjustment of arch bridge buckling ropes and discloses a method for adjusting bridge buckling rope force based on a flagelliform algorithm and an artificial neural network. Combining an SFO algorithm with an artificial neural network model, accurately modeling a nonlinear relation between a cable buckling force and a linear shape in the arch bridge construction process by using an ANN deep learning model, searching an optimal cable buckling force which minimizes the square sum of the deviation of an observation point in a relation model between the cable buckling force predicted by the deep learning model and the linear shape by using the SFO algorithm, and performing global optimization on the deep learning model so as to realize efficient and accurate adjustment of the cable buckling force. The invention can efficiently and accurately adjust the cable buckling force, improves the structural stability and the service life of the steel pipe concrete arch bridge, reduces the engineering cost and has obvious effects on improving the construction and maintenance quality of bridge engineering.
Description
Technical Field
The invention relates to the field of load adjustment of arch bridge buckling ropes, in particular to a method for adjusting bridge buckling rope force based on a flagelliforme algorithm and an artificial neural network.
Background
In bridge engineering, in particular to a steel pipe concrete arch bridge with a complex structure, the adjustment of the buckling cable force is an extremely complex and key link. The size and distribution of the buckling rope force directly influence the stability, safety and service life of the bridge. Improper adjustment of the buckling cable force can cause the bridge structure to bear excessive stress, even cause structural damage, and pose a serious threat to public safety. In practical engineering, the adjustment of the buckling cable force is a tedious and precisely controlled process. Engineers must consider a number of factors, such as the specific structure, material properties, environmental conditions, etc., of the bridge, to determine the appropriate buckling cable force values through complex calculations and multiple rounds of experimentation. However, this method not only consumes a lot of time and human resources, but also makes it difficult to ensure that an optimal adjustment effect is achieved each time due to the numerous and mutual influence of the involved factors. In addition, since each bridge has a unique structure and environmental conditions, the adjustment of the buckle forces is often dependent on the experience of the engineer. However, due to the limitations of experience, such rules of thumb often cannot cope with complex or special situations, resulting in poor tuning results. Therefore, the adjustment of the buckling cable force is a great challenge in bridge engineering, and needs to be solved by adopting a more advanced and accurate method. The method is based on scientific principles and advanced technology, and combines a mathematical model and a calculation method to realize more accurate and efficient adjustment of the buckling cable force, thereby improving the stability, the safety and the service life of the bridge. The method is beneficial to reducing human errors, improving the controllability and predictability of the adjusting effect, and providing more reliable basis for the design and construction of bridge engineering.
In the past, most of the methods for adjusting the buckling cable force are based on a finite element model, and the buckling cable force is adjusted through repeated experiments. Although this method has been widely used in practice, since the relationship between the cable buckling force and the line shape is highly complex in practice, it is often nonlinear and highly dimensional, which makes the conventional error-testing method not only inefficient, costly, but also difficult to guarantee in terms of accuracy. Therefore, a new method capable of adjusting the buckling cable force more accurately and more efficiently is needed to be researched, and the method has profound practical significance for construction and maintenance of bridge engineering.
In recent years, with the rapid development of computer science and artificial intelligence technology, new methods and technologies have begun to be introduced into the engineering technology field. Among these, artificial neural network (Artificial Neural Network, ANN) models, in particular, are beginning to draw attention from the engineering community. ANN is excellent in dealing with complex, nonlinear problems and is therefore increasingly being applied to prediction of complex nonlinear problems.
However, although artificial neural networks have achieved significant results in many areas, the application in arch bridge buckling cable force and linear relationship prediction has not been seen. In addition, the optimization algorithm plays an important role in engineering technology. Traditional optimization algorithms such as gradient descent method, newton method, quasi-Newton method and the like have been successfully applied to various engineering problems including structural optimization, equipment scheduling, production line optimization and the like, but traditional numerical algorithms are prone to be in a local optimal solution in the optimization process, and have the defect of difficult convergence on complex nonlinear problems. In recent years, some heuristic optimization algorithms, such as particle swarm optimization, genetic algorithm, ant colony algorithm and the like, have achieved remarkable results in solving the complex nonlinear optimization problem.
Disclosure of Invention
Aiming at the problems, the invention aims to provide a method for adjusting bridge buckling cable force based on a flagpole algorithm and an artificial neural network, which utilizes the artificial neural network technology to establish a relation model of the buckling cable force and a linear shape and samples a novel optimization flagpole algorithm (Sailed Fish Optimizer, SFO). The cable buckling force is optimized efficiently, the construction and maintenance of bridge engineering are improved, the structural stability and the service life of the bridge are improved, and the engineering cost is reduced. The technical proposal is as follows:
a method for adjusting bridge buckling cable force based on a flagelliforme algorithm and an artificial neural network comprises the following steps:
step 1: establishing an accurate finite element model according to specific parameters of the concrete-filled steel tube arch bridge by Midas finite element software;
step 2: in the finite element model, adding possible factors in the construction process, including construction sequences and working condition changes, so as to enhance the authenticity and accuracy of the model;
step 3: determining that probability distribution characteristics of the cable buckling force, the material characteristics and the geometric parameters are normal distribution, and sampling Monte Carlo;
step 4: taking the sampled cable buckling force, material characteristics and geometric parameter samples as input, and carrying out finite element model analysis for calculating corresponding linear deviation;
step 5: designing and constructing an ANN deep learning model comprising an input layer, two hidden layers and an output layer, wherein the input layer is used for inputting data of a buckling cable force, material characteristics and geometric parameters, and the output layer is used for outputting a linear deviation square sum; the expression is as follows:
wherein ,xa vector representing input data, i.e., buckle cable force, material properties, and geometric parameters;representing information about input dataxModel output of (2); />Is an activation function, W1 and W2 represent weights of the hidden layer and the output layer, respectively, and b1 and b2 represent deviations of the hidden layer and the output layer, respectively;
step 6: inputting the cable buckling force, the material characteristics, the geometric parameters and the linear deviation data into an ANN deep learning model, configuring training parameters, training, and enabling an artificial neural network model to learn and master the mapping relation between the cable buckling force and the linear deviation;
step 7: performing rope buckling force optimization in a trained artificial neural network model by using a flagelliforme algorithm, and generating an initial flagelliforme population;
step 8: simulating the behavior of prehunting, tail-chasing, group-chasing and burst-chasing of the flagelliform fish, searching in a solution space of a value range of the knot rope force, and evaluating the quality of each solution through a fitness function; the excellent solution is reserved through a protrusion mechanism, the solution with poor linear fitting is discarded, convergence is finally achieved, and the buckling cable force capable of minimizing the linear deviation is found.
Further, in the step 6, in the training process, training data and test data are input into the constructed artificial neural network model in batches, and loss function values of each batch are obtained through calculation of the hidden layer and the output layer; meanwhile, calculating the average value of all batch loss functions in each training period, and calculatingε n And calculates the difference value of the loss functions of two adjacent training periods to see whether the difference value is smaller than the set precisionεAs a convergence condition; if the precision is met, finishing training; otherwise, the network parameters are updated according to the loss function values, and then the artificial neural network model is trained until convergence.
Further, the step 7 specifically includes:
step 7.1: parameters defining the optimization algorithm of the flagelliforme
Comprising two populations: a buckling cable force combination 1 corresponding to a prey flagelliforme population, a buckling cable force combination 2 corresponding to a prey sardine population, and the number of individuals N1 of the buckling cable force combination 1, the number of individuals N2 of the buckling cable force combination 2, a preset maximum iteration number T, a dimension D of a problem, a predation coefficient A of the buckling cable force combination 1 and a preset control parameter epsilon;
step 7.2: initializing individual positions of two populations, and calculating the fitness of each individual by using an objective function;
step 7.3: and finding out the individual with the smallest fitness value, namely the smallest sum of squares of the linear deviations, in the two populations as the optimal individual of the two populations.
Further, the step 8 specifically includes:
step 8.1: copying information of two groups of cable buckling force combinations, and then calculating the proportion PD of the current cable buckling force combination 1 to the total population;
step 8.2: updating the individual position of the knot cable force combination 1, comparing the individual fitness value of the new position with the original position, and updating the new position if the fitness value of the new position is more optimal; otherwise, the original position is kept;
step 8.3: calculating a propagation coefficient AP of the cable force combination 2, and determining an updating strategy of the cable force combination 2 according to the magnitude of the propagation coefficient AP;
when the propagation coefficient AP is smaller than 0.5, updating the individual positions of the partial buckling cable force combinations 2 in a mode of adding a random disturbance;
when the propagation coefficient AP is larger than 0.5, updating the individual positions of all the buckling cable force combinations 2 in a way of adding a random disturbance related to the optimal buckling cable force combination position;
step 8.4: after updating the individual position of the buckle cable force combination 2, comparing the fitness value of the new position and the home position, and optimizing;
step 8.5: the individual with better fitness value in the buckling cable force combination 2 is fused into the buckling cable force combination 1, and the optimal individuals of the two populations are updated;
step 8.6: judging whether the current iteration number reaches a preset maximum iteration number T, if so, terminating iteration, and outputting an optimal cable force combination; otherwise, the next iteration is entered.
Compared with the prior art, the invention has the beneficial effects that:
1) The invention provides an innovative technical scheme for adjusting bridge buckling cable tension by using an ANN deep learning model and an SFO algorithm. And carrying out normal distribution random sampling in a given range of the cable buckling force, the material characteristics and the geometric parameters by using a Monte Carlo sampling method. The Monte Carlo sampling method is used as a statistical simulation method widely applied to various fields, and has the main advantages of being capable of effectively and randomly sampling in a large-scale data set, and guaranteeing the universality and representativeness of sampled data. In the invention, the method is used for extracting data according to normal distribution within the upper and lower limits of the cable buckling force, the material characteristics and the geometric parameters, and provides high-quality training data for the following ANN deep learning model.
2) Based on Monte Carlo sampled data, the ANN deep learning model accurately establishes a mapping relationship between the buckling cable force, the material characteristics, the geometric parameters and the linear deviation by learning and simulating the internal relationship between the buckling cable force, the material characteristics and the geometric parameters. The technical scheme fully utilizes the outstanding advantages of deep learning in the aspects of processing big data and constructing a highly complex nonlinear model.
3) The SFO algorithm adopted by the invention plays a key optimization role. The SFO algorithm is optimized based on an ANN deep learning model by taking the characteristic of rapid convergence and efficient optimization, and the minimum square sum of deviation of linear observation points as a target, so that strong support is provided for adjusting the cable buckling force. In the whole, the invention combines the advantages of the Monte Carlo sampling method, the ANN deep learning model and the SFO algorithm, realizes the efficient and accurate adjustment of the bridge buckling cable force, effectively improves the stability and the service life of the bridge structure, reduces the engineering cost, expands the application range, has great beneficial effects on the bridge engineering field, and provides a new thought and research direction for related scientific research.
Drawings
Fig. 1 is a hoisting layout of a large bridge cable.
Fig. 2 is a diagram of a finite element computation model of a bridge.
Fig. 3 is a topological structure diagram of an ANN artificial neural network.
FIG. 4 is a diagram of an SFO-ANN combinatorial optimization model.
Fig. 5 is a graph showing the trend of the loss function value with an ANN training period.
FIG. 6 is a plot of SFO optimal fitness.
Fig. 7 is a comparison of the buckling cable force optimization before and after.
Fig. 8 is a graph comparing linear deviations before and after optimization of the buckle force.
Detailed Description
The invention is described in further detail below with reference to the drawings and the detailed description.
The basic idea of the invention is to combine an SFO algorithm with an artificial neural network (Artificial Neural Network, ANN) model, accurately model the nonlinear relation between the buckling cable force and the linear shape in the arch bridge construction process by using an ANN deep learning model, and then globally optimize the deep learning model by using the SFO algorithm so as to realize efficient and accurate adjustment of the buckling cable force.
In the invention, the ANN deep learning model is mainly used for establishing a relationship model of arch bridge buckling cable force and line shape. The relationship between the arch bridge buckling cable force and the line shape is nonlinear, high-dimensional and complex in actual engineering, and is difficult to describe accurately through a traditional mathematical model. The ANN deep learning model can learn by itself and automatically find rules in data, so that a complex relation model between the cable buckling force and the line shape is effectively established.
The SFO algorithm has the main function of searching the optimal buckling cable force in the arch bridge buckling cable force and linear relation model predicted by the deep learning model. The SFO algorithm is a heuristic optimization algorithm, and has good processing capacity for complex optimization problems by means of observation and simulation of the predatory behavior of the flagfish in nature. In the invention, the arch bridge buckling cable force is used as an optimization variable, the sum of squares of deviation values of linear and predicted linear observation points is designed as an optimization target, and an SFO algorithm is utilized to find the optimal buckling cable force which enables the sum of squares of the deviation of the observation points to be minimum in a relation model of the buckling cable force predicted by a deep learning model and the linear.
The main application scene of the invention is the construction of a steel tube concrete arch bridge. Through the combined application of the ANN deep learning model and the SFO algorithm, the invention can efficiently and accurately adjust the buckling cable force, improve the structural stability and the service life of the steel tube concrete arch bridge, reduce the engineering cost and have remarkable effects on improving the construction and maintenance quality of bridge engineering.
1) The technical scheme adopted by the invention comprises the following steps:
step 1: and establishing an accurate finite element model according to specific parameters of the concrete filled steel tube arch bridge by Midas finite element software.
Taking a certain large-span steel tube concrete arch bridge as a research object, a large-span steel tube concrete arch bridge cable hoisting layout is shown in fig. 1, and an accurate model is built in Midas finite element software. The first step of model establishment is to describe the concrete-filled steel tube arch bridge in detail and accurately according to the actual structural parameters, such as the size, the shape, the material properties and the like. Including but not limited to bridge structural design, material mechanical properties, and related engineering practices. On this basis, accuracy and fidelity of the model can be ensured. The bridge finite element computation model diagram is shown in fig. 2.
Step 2: in the finite element model, factors which may occur in the construction process, including construction sequences and working condition changes, are added to enhance the authenticity and accuracy of the model.
After the finite element model is constructed, the factors possibly occurring in the construction process are increased, and the authenticity and accuracy of the model are further enhanced. Taking the construction process of a bridge as an example, the installation of the main arch ring sections adopts symmetrical suspension splicing of two banks, the main arch ring sections are divided into 4 hoisting sections, and the construction of each section involves a buckling and hanging cable hoisting system. Therefore, it is necessary to activate the corresponding buckling cables in the finite element model according to such construction sequences and operating condition changes to simulate the entire construction process. The simulation process can help us understand and predict the behavior and response of the bridge structure under different construction stages and working conditions, and provides an important reference basis for actual construction.
Step 3: and determining that the probability distribution characteristics of the cable buckling force, the material characteristics and the geometric parameters are normal distribution, and carrying out Monte Carlo sampling.
(1)
wherein ,xin order to input the variable(s),uis the mean value of the variables,is the mean square error of the variable;
within the definite range of the cable force, the material property and the geometric parameter, the invention generates according to the Monte Carlo sampling method principlenThe cable force, material property and geometric parameter are random values. These randomly generated halyard force values are fed into the midas finite element model as input parameters, and the corresponding output values are the sum of squares of the linear deviations. These datasets, which consist of the cable buckling force and the linear deviation values, are used as training data for the subsequent deep learning model, so as to achieve the goal of model learning of complex mapping relation between cable buckling force and linear deviation.
Step 4: and taking the sampled cable buckling force, material characteristics and geometric parameter samples as input, carrying out finite element model analysis, and calculating corresponding linear deviation.
And comparing the finite element model with actual observation data to ensure the accuracy and the credibility of the finite element model. Through the steps, a series of linear deviation square sum data corresponding to the cable buckling force samples can be obtained, and a foundation is provided for subsequent model training and optimization.
Step 5: an artificial neural network model comprising an input layer, two hidden layers and an output layer is designed and constructed, wherein the input layer is used for inputting data of buckling cable force, material characteristics and geometric parameters, and the output layer is used for outputting linear deviation square sums.
In the invention, an Artificial Neural Network (ANN) model is constructed so as to solve the problem of complex mapping between the cable buckling force and the linear deviation. The depth network consists of an input layer, two hidden layers and an output layer, and has four layers. The input layer is the data of the cable buckling force, the material characteristics and the geometric parameters, and the output layer is the square sum of linear deviation. The hidden layer uses a tanh activation function and the output layer uses a linear activation function. The topology structure of the ANN artificial neural network is shown in figure 3.
The model learning process is performed by minimizing prediction errors and using Adam optimizers for parameter optimization, and then using back propagation algorithms to update the weights. To describe this model graphically, we can represent it with such a mathematical expression:
(2)
wherein ,xis the vector of input data, namely the buckling cable force, the material characteristics and the geometric parameters,is an activation function, W1 and W2 represent weights of the hidden layer and the output layer, respectively, and b1 and b2 represent deviations of the hidden layer and the output layer, respectively. After a preset training period, the network can learn and master the mapping relation between the cable buckling force and the linear deviation.
Step 6: and inputting the data of the cable buckling force, the material characteristics, the geometric parameters and the linear deviation into a deep learning model, configuring proper training parameters, training, and enabling the artificial neural network model to learn and master the mapping relation between the cable buckling force and the linear deviation.
In the training process, training data and test data are input into a constructed network model in batches, and loss function values of each batch are obtained through calculation of a hidden layer and an output layer. Meanwhile, calculating the average value of all batch loss functions in each training period, and calculatingε n And calculates the difference value of the loss functions of two adjacent training periods to see whether the difference value is smaller than the set precisionε=0.0001 as a convergence condition. If the precision is met, finishing training; otherwise, updating network parameters according to the loss function value, and then performing ANN network training until convergence. The trend of the loss function value with the ANN training period is shown in fig. 5.
To evaluate the performance of the model, we use mean square error (Mean Squared Error, MSE) to measure the accuracy of training and testing. After model convergence, we predict the input data.
By constructing an accurate depth network, complex relationships between the cable buckling force and the linear deviation are successfully predicted and optimized, and a new method and a new view angle are provided for solving similar problems.
Step 7: and carrying out the optimization of the rope buckling force in the trained artificial neural network model by using a flagelliforme algorithm, and generating an initial flagelliforme population.
The SFO-ANN combined optimization model diagram is shown in FIG. 4. In the SFO algorithm, first, main parameters of the SFO optimization algorithm need to be defined, including decision variable matrix size, population size, preset iteration times and the like. Wherein N1 and N2 represent the number of individuals of two populations, the halyard force combination 1 (corresponding to the predated population of flag fish) and the halyard force combination 2 (corresponding to the predated population of sardine), respectively, T is the preset maximum number of iterations, D is the dimension of the problem (i.e., the number of decision variables), a is the predation coefficient of the halyard force combination 1, epsilon is the preset control parameter. Thereafter, the individual positions of the two populations are initialized and fitness of each individual is calculated using an objective function. Then, the individual with the smallest fitness value (i.e. the least sum of squares of linear deviations) in the two populations is found as the optimal individual for the two populations.
Step 8: the behavior (prehunting, tail hunting, group hunting and kick girth) of the simulated flagelliform is searched in a solution space of the value range of the buckle rope force, and the advantages and disadvantages of each solution are evaluated through the fitness function.
In the main cycle of SFO, the information of two groups of buckling cable force combinations (two populations) is copied first, and then the proportion PD of the current buckling cable force combination 1 to the total population is calculated. Then updating the individual position of the buckling cable force combination 1, comparing the individual fitness value of the new position with the original position, and updating the new position if the fitness value of the new position is more optimal; otherwise, the original position is maintained. Then, the propagation coefficient AP of the cable force combination 2 is calculated, and the updating strategy of the cable force combination 2 is determined according to the size of the AP. When the AP is smaller than 0.5, only the individual positions of part of the buckling cable force combinations 2 are updated in a way of adding a random disturbance; when the AP is greater than 0.5, the individual positions of all of the lanyard force combinations 2 are updated by adding a random disturbance associated with the optimal lanyard force combination (individual flagelliform) position. After updating the individual position of the buckling cable force combination 2, the adaptation values of the new position and the original position are compared and optimized similarly to the buckling cable force combination 1.
And finally, merging the individual with the better fitness value in the buckling cable force combination 2 into the buckling cable force combination 1, and updating the optimal individual of the two populations. Then judging whether the current iteration number reaches the preset maximum iteration number, if so, stopping iteration, and outputting an optimal cable force combination; otherwise, the next iteration is entered.
Step 9: and respectively inputting the optimal buckling cable force and the optimal pre-optimization cable force which are found by using an SFO algorithm into a midas model as inputs, and comparing the output linear deviation. By means of the comparison, the superiority of the SFO algorithm in the aspect of cable buckling force optimization can be intuitively verified.
2) Rope buckling force optimization result
Fig. 6 shows a relationship curve between fitness and iteration number in the process of carrying out the cable buckling force optimization of the SFO-ANN hybrid algorithm, and it can be known from the graph that the convergence rate is obviously reduced when the flagelliforme population evolves to about 18 th generation, and the convergence rate is basically converged to the vicinity of the optimal solution when the flagelliforme population evolves to about 30 th generation.
Fig. 7 shows the cable buckling force optimization result, and it can be seen from the graph that, compared with the cable buckling force of the original design, the permanent cable buckling force after being optimized based on the SFO-ANN is greatly reduced from 108kN to 71kN, from the 108kN to 34.2%, from the 221kN to 189kN, and from the 221kN to 14.5%. The cable buckling force of the No. 3 cable is reduced to 291kN from 303kN to about 4 percent. The optimized permanent buckling cable force is lowered to a certain extent, and the integral stress of the structure is more reasonable on the premise of ensuring that the calculated linear approximation target linear shape.
Fig. 8 shows the calculation results of the linear deviation before and after the optimization of the buckling cable force, and the comparison of the calculation results of the designed cable force and the optimized cable force shows that the linear deviation between each control section of the arch rib and the theoretical elevation is greatly reduced based on the combination of the cable forces after the optimization of the SFO-ANN, the relative deviation between the calculated linear deviation of each control section and the target linear deviation is averagely reduced by 16.78 and mm, wherein the average reduction amplitude of the linear deviation at the arch top is about 85.3%, and the effectiveness of the SFO-ANN combination optimization model is verified by the linear forms of the arch sections 1 to 7.
Claims (4)
1. The method for adjusting the bridge buckling cable tension based on the flagelliforme algorithm and the artificial neural network is characterized by comprising the following steps of:
step 1: establishing an accurate finite element model according to specific parameters of the concrete-filled steel tube arch bridge by Midas finite element software;
step 2: in the finite element model, adding possible factors in the construction process, including construction sequences and working condition changes, so as to enhance the authenticity and accuracy of the model;
step 3: determining that probability distribution characteristics of the cable buckling force, the material characteristics and the geometric parameters are normal distribution, and sampling Monte Carlo;
step 4: taking the sampled cable buckling force, material characteristics and geometric parameter samples as input, and carrying out finite element model analysis for calculating corresponding linear deviation;
step 5: designing and constructing an ANN deep learning model comprising an input layer, two hidden layers and an output layer, wherein the input layer is used for inputting data of a buckling cable force, material characteristics and geometric parameters, and the output layer is used for outputting a linear deviation square sum; the expression is as follows:
wherein ,xa vector representing input data, i.e., buckle cable force, material properties, and geometric parameters; />Representing information about input dataxModel output of (2); />Is an activation function, W1 and W2 represent weights of the hidden layer and the output layer, respectively, and b1 and b2 represent deviations of the hidden layer and the output layer, respectively;
step 6: inputting the cable buckling force, the material characteristics, the geometric parameters and the linear deviation data into an ANN deep learning model, configuring training parameters, and training to enable the ANN deep learning model to learn and master the mapping relation between the cable buckling force and the linear deviation;
step 7: performing rope buckling force optimization in a trained ANN deep learning model by using a flagelliforme algorithm, and generating an initial flagelliforme population;
step 8: simulating the behavior of prehunting, tail-chasing, group-chasing and burst-chasing of the flagelliform fish, searching in a solution space of a value range of the knot rope force, and evaluating the quality of each solution through a fitness function; the excellent solution is reserved through a protrusion mechanism, the solution with poor linear fitting is discarded, convergence is finally achieved, and the buckling cable force capable of minimizing the linear deviation is found.
2. The method for adjusting the bridge buckling and rope force based on the flagelliform algorithm and the artificial neural network according to claim 1, wherein in the step 6, training data and test data are input into a constructed ANN deep learning model in batches, and loss function values of each batch are obtained through calculation of a hidden layer and an output layer; meanwhile, calculating the average value of all batch loss functions in each training period, and calculatingε n And calculates the difference value of the loss functions of two adjacent training periods to see whether the difference value is smaller than the set precisionεAs a convergence condition; if the precision is met, finishing training; otherwise, updating network parameters according to the loss function value, and then training the ANN deep learning model until convergence.
3. The method for adjusting the bridge buckling cable force based on the flagelliforme algorithm and the artificial neural network according to claim 1, wherein the step 7 specifically comprises:
step 7.1: parameters defining the optimization algorithm of the flagelliforme
Comprising two populations: a combination of lasso forces 1 corresponding to the population of predators of the group of sardine, a combination of lasso forces 2 corresponding to the population of predators of the group of sardine; the method comprises the steps of setting the number N1 of the individual buckling cable force combinations 1, the number N2 of the individual buckling cable force combinations 2, the preset maximum iteration number T, the dimension D of the problem, the predation coefficient A of the buckling cable force combinations 1 and the preset control parameter epsilon;
step 7.2: initializing individual positions of two populations, and calculating the fitness of each individual by using an objective function;
step 7.3: and finding out the individual with the smallest fitness value, namely the smallest sum of squares of the linear deviations, in the two populations as the optimal individual of the two populations.
4. The method for adjusting the bridge buckling cable force based on the flagelliforme algorithm and the artificial neural network according to claim 3, wherein the step 8 specifically comprises:
step 8.1: copying information of two groups of cable buckling force combinations, and then calculating the proportion PD of the current cable buckling force combination 1 to the total population;
step 8.2: updating the individual position of the knot cable force combination 1, comparing the individual fitness value of the new position with the original position, and updating the new position if the fitness value of the new position is more optimal; otherwise, the original position is kept;
step 8.3: calculating a propagation coefficient AP of the cable force combination 2, and determining an updating strategy of the cable force combination 2 according to the magnitude of the propagation coefficient AP;
when the propagation coefficient AP is smaller than 0.5, updating the individual positions of the partial buckling cable force combinations 2 in a mode of adding a random disturbance;
when the propagation coefficient AP is larger than 0.5, updating the individual positions of all the buckling cable force combinations 2 in a way of adding a random disturbance related to the optimal buckling cable force combination position;
step 8.4: after updating the individual position of the buckle cable force combination 2, comparing the fitness value of the new position and the home position, and optimizing;
step 8.5: the individual with better fitness value in the buckling cable force combination 2 is fused into the buckling cable force combination 1, and the optimal individuals of the two populations are updated;
step 8.6: judging whether the current iteration number reaches a preset maximum iteration number T, if so, terminating iteration, and outputting an optimal cable force combination; otherwise, the next iteration is entered.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311055386.1A CN116796611B (en) | 2023-08-22 | 2023-08-22 | Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311055386.1A CN116796611B (en) | 2023-08-22 | 2023-08-22 | Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116796611A true CN116796611A (en) | 2023-09-22 |
CN116796611B CN116796611B (en) | 2023-10-31 |
Family
ID=88039863
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311055386.1A Active CN116796611B (en) | 2023-08-22 | 2023-08-22 | Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116796611B (en) |
Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107884113A (en) * | 2017-10-19 | 2018-04-06 | 哈尔滨工业大学 | A kind of thrust method of testing for underwater propeller propeller |
US20190259182A1 (en) * | 2018-02-20 | 2019-08-22 | Lyft, Inc. | Deep Direct Localization from Ground Imagery and Location Readings |
CN113435625A (en) * | 2021-05-25 | 2021-09-24 | 国网河北省电力有限公司衡水供电分公司 | Dynamic economic dispatching optimization method and device for power system |
CN113613170A (en) * | 2021-08-16 | 2021-11-05 | 苏州云享阁智能科技有限公司 | High-precision indoor positioning system and method based on low-cost antenna extension |
CN113627606A (en) * | 2020-05-06 | 2021-11-09 | 江苏科技大学 | RBF neural network optimization method based on improved particle swarm optimization |
WO2022031038A1 (en) * | 2020-08-04 | 2022-02-10 | 주식회사 에이슬립 | Computing device for predicting sleep state on basis of data measured in sleep environment of user |
CN114330194A (en) * | 2021-12-31 | 2022-04-12 | 桂林电子科技大学 | Analog circuit fault diagnosis method based on SFO (Small form-factor pluggable) optimization depth limit learning machine |
CN114912673A (en) * | 2022-04-29 | 2022-08-16 | 河海大学 | Water level prediction method based on whale optimization algorithm and long-term and short-term memory network |
CN115062750A (en) * | 2022-06-16 | 2022-09-16 | 合肥学院 | Compound water solubility prediction method of dynamic evolution whale optimization algorithm |
KR20220159762A (en) * | 2021-05-26 | 2022-12-05 | 인하대학교 산학협력단 | Construction method of artificial neural network model using genetic algorithm and variable optimization method using the same |
CN115510549A (en) * | 2022-10-22 | 2022-12-23 | 广西大学 | CFST arch bridge cable-stayed buckling and hanging cable force calculation method based on improved PSO optimization algorithm |
CN115526086A (en) * | 2022-10-27 | 2022-12-27 | 中国石油大学(华东) | Carbonate reservoir water drive breakthrough time prediction method based on proxy model |
CN115600468A (en) * | 2022-11-09 | 2023-01-13 | 广西路桥工程集团有限公司(Cn) | Optimization calculation method for skew-pulling buckling of arch bridge by considering nonlinear effect |
CN115630431A (en) * | 2022-11-08 | 2023-01-20 | 广西路桥工程集团有限公司 | Steel pipe concrete arch bridge buckling cable force prediction method and device based on MLP |
CN115828759A (en) * | 2022-12-14 | 2023-03-21 | 广西路桥工程集团有限公司 | Method and device for quickly determining optimal cable buckling force in construction period of multi-constraint arch bridge |
CN115983077A (en) * | 2023-01-30 | 2023-04-18 | 中铁二十局集团第五工程有限公司 | Method, equipment and medium for acquiring cable buckling force of Y-shaped arch bridge by inclined pulling buckling method |
-
2023
- 2023-08-22 CN CN202311055386.1A patent/CN116796611B/en active Active
Patent Citations (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107884113A (en) * | 2017-10-19 | 2018-04-06 | 哈尔滨工业大学 | A kind of thrust method of testing for underwater propeller propeller |
US20190259182A1 (en) * | 2018-02-20 | 2019-08-22 | Lyft, Inc. | Deep Direct Localization from Ground Imagery and Location Readings |
CN113627606A (en) * | 2020-05-06 | 2021-11-09 | 江苏科技大学 | RBF neural network optimization method based on improved particle swarm optimization |
WO2022031038A1 (en) * | 2020-08-04 | 2022-02-10 | 주식회사 에이슬립 | Computing device for predicting sleep state on basis of data measured in sleep environment of user |
CN113435625A (en) * | 2021-05-25 | 2021-09-24 | 国网河北省电力有限公司衡水供电分公司 | Dynamic economic dispatching optimization method and device for power system |
KR20220159762A (en) * | 2021-05-26 | 2022-12-05 | 인하대학교 산학협력단 | Construction method of artificial neural network model using genetic algorithm and variable optimization method using the same |
CN113613170A (en) * | 2021-08-16 | 2021-11-05 | 苏州云享阁智能科技有限公司 | High-precision indoor positioning system and method based on low-cost antenna extension |
CN114330194A (en) * | 2021-12-31 | 2022-04-12 | 桂林电子科技大学 | Analog circuit fault diagnosis method based on SFO (Small form-factor pluggable) optimization depth limit learning machine |
CN114912673A (en) * | 2022-04-29 | 2022-08-16 | 河海大学 | Water level prediction method based on whale optimization algorithm and long-term and short-term memory network |
CN115062750A (en) * | 2022-06-16 | 2022-09-16 | 合肥学院 | Compound water solubility prediction method of dynamic evolution whale optimization algorithm |
CN115510549A (en) * | 2022-10-22 | 2022-12-23 | 广西大学 | CFST arch bridge cable-stayed buckling and hanging cable force calculation method based on improved PSO optimization algorithm |
CN115526086A (en) * | 2022-10-27 | 2022-12-27 | 中国石油大学(华东) | Carbonate reservoir water drive breakthrough time prediction method based on proxy model |
CN115630431A (en) * | 2022-11-08 | 2023-01-20 | 广西路桥工程集团有限公司 | Steel pipe concrete arch bridge buckling cable force prediction method and device based on MLP |
CN115600468A (en) * | 2022-11-09 | 2023-01-13 | 广西路桥工程集团有限公司(Cn) | Optimization calculation method for skew-pulling buckling of arch bridge by considering nonlinear effect |
CN115828759A (en) * | 2022-12-14 | 2023-03-21 | 广西路桥工程集团有限公司 | Method and device for quickly determining optimal cable buckling force in construction period of multi-constraint arch bridge |
CN115983077A (en) * | 2023-01-30 | 2023-04-18 | 中铁二十局集团第五工程有限公司 | Method, equipment and medium for acquiring cable buckling force of Y-shaped arch bridge by inclined pulling buckling method |
Non-Patent Citations (1)
Title |
---|
凌中正: "基于旗鱼优化算法的桥梁支座损伤识别", 中国优秀硕士学位论文全文数据库工程科技Ⅱ辑(月刊), no. 02, pages 034 - 661 * |
Also Published As
Publication number | Publication date |
---|---|
CN116796611B (en) | 2023-10-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhao et al. | A hybrid algorithm based on self-adaptive gravitational search algorithm and differential evolution | |
CN111262858A (en) | Network security situation prediction method based on SA _ SOA _ BP neural network | |
Chu et al. | Fuzzy hierarchical surrogate assists probabilistic particle swarm optimization for expensive high dimensional problem | |
CN107316099A (en) | Ammunition Storage Reliability Forecasting Methodology based on particle group optimizing BP neural network | |
CN111461457A (en) | Foundation pit displacement prediction method based on particle swarm optimization BP neural network | |
Zhou et al. | Boosted local dimensional mutation and all-dimensional neighborhood slime mould algorithm for feature selection | |
CN107506865A (en) | A kind of load forecasting method and system based on LSSVM optimizations | |
CN108805346A (en) | A kind of hot continuous rolling force forecasting method based on more hidden layer extreme learning machines | |
CN111709511A (en) | Harris eagle optimization algorithm based on random unscented Sigma point variation | |
CN104732067A (en) | Industrial process modeling forecasting method oriented at flow object | |
Kaveh et al. | Feasibility of PSO-ANFIS-PSO and GA-ANFIS-GA models in prediction of peak ground acceleration | |
CN108830035A (en) | A kind of novel water process coagulant dosage control method, computer, computer program | |
Su et al. | Analysis and improvement of GSA’s optimization process | |
CN110852435A (en) | Neural evolution calculation model | |
CN116796611B (en) | Method for adjusting bridge buckling cable force based on flagelliforme algorithm and artificial neural network | |
CN110210072B (en) | Method for solving high-dimensional optimization problem based on approximate model and differential evolution algorithm | |
CN116522747A (en) | Two-stage optimized extrusion casting process parameter optimization design method | |
Liu et al. | Crowd intelligence evolution based on complex network | |
El-Dahshan et al. | Artificial neural network and genetic algorithm hybrid technique for nucleus–nucleus collisions | |
Hua et al. | New knowledge-based genetic algorithm for excavator boom structural optimization | |
Lin et al. | A hybrid neural network model for sea ice thickness forecasting | |
Teymourifar | A comparison between two approaches to optimize weights of connections in artificial neural networks | |
CN105426960A (en) | BP neural network and MBFO algorithm-based aluminum electrolysis energy conservation and emission reduction control method | |
Li et al. | Research of steel plate temperature prediction based on the improved PSO-ANN algorithm for roller hearth normalizing furnace | |
Wang et al. | Double elite co-evolutionary genetic algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |