Fabricated anchor cable frame beam reinforced slope design method and system, equipment and medium

Abstract

The invention relates to the technical field of engineering data processing, in particular to an assembly type anchor cable frame beam reinforced slope design method and system, equipment and a medium, a neural network is optimized by improving a butterfly optimization algorithm, improvement is carried out on the basis of the butterfly optimization algorithm, butterfly population distribution is more uniform, and the design efficiency is improved. Selecting factors are introduced to carry out global search, variation is dynamically adjusted, a self-adaptive weight strategy is adopted to optimize butterfly populations, and the optimal butterfly populations are searched as hidden layer parameters and learning rate parameters of the corresponding neural network; and outputting the safest fabricated frame beam slope design scheme through the optimal hidden layer parameters and learning rate parameters of the neural network. And more accurate prediction of the fabricated frame beam slope design scheme is realized. The optimized model can better utilize complex and diverse slope data information, the robustness and accuracy of the prediction model are improved, and important technical support is provided for engineering decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering data processing, and more specifically, to a design method and system, equipment, and medium for reinforcing a slope with an assembled anchor cable frame beam. Background Art

[0002] With the continuous progress and development of social economy, more and more infrastructure construction projects have emerged. The articulated assembled anchor cable frame beam has the characteristics of flexible support. Through the hoop action of the frame beam and the active tension of the anchor cable, the deformation of the slope body can be effectively restricted. The anchor cable frame beam support system can be applied to various rock masses and soil bodies, and can effectively reduce the excavation of the fractured soil body during the construction process, accelerating the construction progress. Compared with traditional passive protection, the anchor cable frame beam structure actively restricts the displacement of the soil body by improving the mechanical properties of the soil body, achieving the overall stability of the slope body. As an important engineering project in the process of infrastructure construction, how to ensure the stability of the slope body during the construction and use process has become an issue that has to be solved in the process of infrastructure construction. Currently, the commonly used support forms can meet the basic requirements of the current slope support project, but there are still problems to varying degrees, such as high technical requirements at the construction site, complex and cumbersome processes, environmental pollution, and impact on the later environmental restoration. Using traditional slope support technologies often causes an extension of the construction period due to the need for a large amount of concrete pouring work on site, and causes damage to the local environment.

[0003] The assembled concrete structure has the advantages of low cost, convenient construction, and shortened construction period. However, due to its relatively short application time, the current theoretical research content is relatively small, the theoretical research lags behind the engineering practice, and the frame design mainly relies on experience at present, which is likely to cause design deficiencies or waste, and may indirectly lead to major engineering accidents such as slope instability.

[0004] Currently, when designing the assembled frame beam slope, it is greatly affected by human factors. Considering the number of parameters and the mutual coupling of various parameters increases the difficulty of manual operation, increasing the failure probability of the structure. Therefore, in order to prevent disasters, manual design will make a conservative estimate of the design, resulting in a great waste of materials and economic costs, and invisibly increasing the construction cost and time cost. At the same time, most of the current frame beams are constructed by on-site casting, which lengthens the construction period, cannot meet the requirements of rapid construction, and there are problems such as poor on-site environment, poor curing effect, and the possibility of landslide accidents during the construction process.

[0005] For the traditional LSTM model, there are many hyperparameters, which are difficult to select manually, and the training process is very slow, having a great impact on the performance loss. Summary of the Invention

[0006] The purpose of the present invention is to provide a design method, system, equipment and medium for strengthening slopes with assembled anchor cable frame beams to solve the problems in the above-mentioned prior art.

[0007] The present invention is realized through the following technical solutions: In the first aspect, the present invention provides a design method for strengthening slopes with assembled anchor cable frame beams, including: Step S1: Obtain the design data and historical monitoring data of the assembled frame beam slope, integrate them into a data sample, and classify and divide the data sample to obtain a training set and a test set; Step S2: Send the training set and the test set to a preset neural network for training and verification, and optimize the preset neural network through a preset improved butterfly optimization algorithm to obtain a prediction model for the optimized design scheme of the assembled frame beam slope; Step S3: Send the historical monitoring data of the assembled frame beam slope to the prediction model for the optimized design scheme of the assembled frame beam slope to predict the design scheme of the assembled frame beam slope.

[0008] Preferably, the step S2 includes: Step S21: Send the training set to a preset neural network for training to obtain a trained neural network; Step S22: Optimize the hidden layer parameters and learning rate parameters of the trained neural network based on the improved butterfly optimization algorithm to obtain optimized hidden layer parameters and learning rate parameters.

[0009] Step S23: Assign the optimized hidden layer parameters and learning rate parameters to the trained neural network to obtain an assigned neural network; Step S24: Send the training set and the test set to the assigned neural network for training and verification to obtain a prediction model for the optimized design scheme of the assembled frame beam slope.

[0010] Preferably, the step S22 includes: Step S221: Initialize the hidden layer parameters and learning rate parameters in the preset neural network, and set the parameters of the preset improved butterfly optimization algorithm. The parameters include various parameters of the set butterfly population, including the initial position of each butterfly, the number of the butterfly population, the maximum number of iterations, the dimension of the space where it is located, the sensory modality, the power exponent and the dynamic conversion probability; Step S222: Randomly generate a butterfly population with a preset population size, uniformly distribute the butterfly population based on the ICMIC chaos mapping method, and change the initial position of the butterfly population to obtain an optimized position of the butterfly population; Step S223: Screen the butterfly population by using differential mutation, crossover, and selection strategies to obtain the screened excellent butterfly population; Step S224: Calculate the fitness values of the screened excellent butterfly population, and introduce a selection factor for global search and dynamically adjust mutation for local search based on the calculated fitness values and random numbers and the dynamic conversion probability until the number of iterations reaches the maximum number of iterations, so as to obtain the optimal hidden layer parameters and learning rate parameters.

[0011] Preferably, the step S224 includes: Step S2241: Calculate the fitness values of the screened excellent butterfly population based on a preset fitness function, sort the calculated fitness values of the butterfly population, and select the optimal fitness value of the butterfly population; Step S2242: Generate a random number to judge the relationship with the preset dynamic conversion probability to make a strategy selection, and introduce a selection factor for global search and dynamically adjust mutation for local search to update the butterfly position; Step S2243: Check whether the end condition is satisfied. If it is satisfied, that is, the maximum number of iterations T is reached, the algorithm ends, and the best solution found, that is, the corresponding optimized hidden layer parameters and learning rate parameters, are output; otherwise, return to step S2242 to continue the next round of iteration.

[0012] Preferably, the updating of the butterfly position includes:

[0013]

[0014] where represents the number of butterflies in the population; represents the stimulation factor of the th butterfly in the improved butterfly optimization algorithm, is the probability that the th butterfly moves towards the best butterfly in the previous iteration, , , respectively represent the positions of the th, th, and th butterflies in the butterfly population at the th generation, is the position of the th butterfly in the butterfly population at the th generation; is a random number within [0, 1]; represents the best position of the butterfly population, represents the set dynamic conversion probability; represents the fragrance perception intensity of the th butterfly; represents the current iteration number; represents the preset number of the butterfly population; is a constant, is the th generation and the probability that the th butterfly chooses to move towards the best butterfly.

[0015] Preferably, the step S24 includes: Step S241: Send the hidden layer parameters and the learning rate parameters to the neural network to construct a forward calculation formula, and obtain the forward calculation formula of the neural network after assignment; Step S242: Send the training set to the neural network after assignment for training again, where the output of the neural network after assignment at a preset moment is calculated through the forward calculation formula, and the loss value at the preset moment is calculated based on the output; Step S243: Update the loss value at the preset moment, and predict the data at each moment to obtain the output result corresponding to the historical monitoring data in the neural network.

[0016] Preferably, the construction of the forward calculation formula includes:

[0017] In the formula, is the forget gate of the neural network, is the recurrent weight of the input gate, is the recurrent weight of the forget gate, is the recurrent weight of the output gate, is the recurrent weight of the cell state, is the sigmoid function, is the short-term memory of the neural network at moment, is the short-term memory of the neural network at moment, is the input operation parameter sequence at moment, is the bias term of the input gate, is the bias term of the forget gate, is the bias term of the output gate, is the bias term of the cell state, is the input gate, is the memory cell state at moment, is the intermediate value of the updated memory cell state, and Tanh is the hyperbolic tangent activation function; is the state of memory cells at a moment, is the output gate.

[0018] In a second aspect, the present invention provides an assembled anchor cable frame beam reinforced slope design system, including: A data collection module, configured to obtain assembled frame beam slope design data and historical monitoring data, integrate them into data samples, and classify the data samples; A model construction module, configured to send the training set and test set to a preset neural network for training and verification, and optimize the preset neural network through a preset improved butterfly optimization algorithm to obtain a prediction model of an optimized assembled frame beam slope design scheme; A scheme output module, configured to send the historical monitoring data of the assembled frame beam slope to the prediction model of the optimized assembled frame beam slope design scheme to predict an assembled frame beam slope design scheme; A main control module, connected to the data collection module, the model construction module, and the scheme output module, and used to execute the above-mentioned method for designing an assembled anchor cable frame beam reinforced slope.

[0019] In a third aspect, an electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the above-mentioned method for designing an assembled anchor cable frame beam reinforced slope when executing the computer program.

[0020] In a fourth aspect, a computer-readable storage medium is characterized in that a computer program is stored on the computer-readable storage medium, and the computer program implements the above-mentioned method for designing an assembled anchor cable frame beam reinforced slope when executed by a processor.

[0021] The technical solution of the present invention has at least the following advantages and beneficial effects: By using the method provided by the present invention, the neural network is optimized through an improved butterfly optimization algorithm, which is improved on the basis of the butterfly optimization algorithm. First, in the initialization population stage, the ICMIC chaotic mapping method is introduced to make the distribution of the butterfly population more uniform. Then, a selection factor is introduced for global search, dynamic adjustment of mutation, and an adaptive weight strategy is used to optimize the butterfly population, and the optimal butterfly population is found, which is the corresponding hidden layer parameters and learning rate parameters of the neural network. The safest assembled frame beam slope design scheme is output through the optimal hidden layer parameters and learning rate parameters of the neural network. The more accurate prediction of the assembled frame beam slope design scheme is realized. The optimized model can better utilize complex and diverse slope data information, improve the robustness and accuracy of the prediction model, and provide important technical support for engineering decision-making. Brief Description of the Drawings

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other relevant drawings can be obtained based on these drawings.

[0023] Figure 1 It is a schematic diagram of the butterfly optimization algorithm model based on the present invention; Figure 2 It is a schematic diagram of the neural network process of the present invention; Figure 3 It is a schematic diagram of the control process of the present invention. Detailed Embodiments

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0025] The terms "first", "second", etc. in the specification and claims of this application and the above drawings are used to distinguish similar objects and do not necessarily need to be used to describe a specific order or sequence. The naming or numbering of steps that appear in this application does not mean that the steps in the method process must be executed in the time / logical order indicated by the naming or numbering. The named or numbered process steps can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0026] Please refer to Figures 1-3 , a method for designing a reinforced slope of a prefabricated anchor cable frame beam provided by the present invention, includes: Step S1: Obtain the design data and historical monitoring data of the prefabricated frame beam slope, integrate them into data samples, and classify and divide the data samples; Collect and organize the design database of the prefabricated frame beam slope, including the engineering geological conditions of slopes in different regions (including filling information, hydrogeological information, geological conditions, etc.) and the required parameters for the design of the prefabricated frame beam slope (mainly including the cross-sectional form and size of the beam body in the prefabricated frame beam slope, the model of the stress reinforcement bar, the concrete strength grade, the cost of the prefabricated frame beam slope, the displacement monitoring data of the prefabricated frame beam slope, etc.), and incorporate all the design materials of the prefabricated frame beam slope into the database.

[0027] Among them, the fill information includes the cohesion and friction angle of clay, and the cohesion of sand. The hydrogeological information includes rainfall information and groundwater information. The geological conditions include the horizontal acceleration and vertical acceleration considering earthquakes, as well as the situation without considering earthquakes. The slope types include steep slopes, gentle slopes, and retaining walls.

[0028] After classifying and dividing all data samples, and performing standardization and normalization processing, 72% of the samples are selected as the training set and 22% of the samples are selected as the test set.

[0029] S2: Send the training set and the test set to a preset neural network for training and verification, and optimize the preset neural network through a preset improved butterfly optimization algorithm to obtain a prediction model for the optimized prefabricated frame beam slope design scheme; It can be understood that in this step, the neural network is optimized through the improved butterfly optimization algorithm, which is an improvement based on the butterfly optimization algorithm. First, in the initialization population stage, the ICMIC chaos mapping method is introduced to make the distribution of the butterfly population more uniform. Then, a selection factor is introduced for global search, dynamic adjustment of mutation, and an adaptive weight strategy is used to optimize the butterfly population, and the optimal butterfly population is found, which is the corresponding hidden layer parameters and learning rate parameters of the neural network. The output of the optimal hidden layer parameters and learning rate parameters of the neural network is the safest prefabricated frame beam slope design scheme. This realizes a more accurate prediction of the prefabricated frame beam slope design scheme. The optimized model can better utilize complex and diverse slope data information, improve the robustness and accuracy of the prediction model, and provide important technical support for engineering decision-making. Among them, step S2 includes step S21, step S22, step S23, and step S24.

[0030] Step S21: Send the training set to a preset neural network for training to obtain a trained neural network; It can be understood that in this step, by training the training set, the parameters of the neural network are optimized to make it have high prediction ability and generalization ability. The training results can reflect the performance and effect of the model in dealing with the prediction problem of the slope design scheme, and provide a basis for subsequent testing and verification.

[0031] Step S22: Optimize the hidden layer parameters and learning rate parameters of the trained neural network based on the improved butterfly optimization algorithm to obtain optimized hidden layer parameters and learning rate parameters; It can be understood that in this step, the improved butterfly optimization algorithm is used to optimize the hidden layer parameters and learning rate of the neural network, so as to significantly improve the prediction accuracy and generalization ability of the model. The optimized model can more accurately predict the design scheme of the slope, providing more reliable support for engineering practical applications. In this step, step S22 includes step S221, step S222, step S223, and step S224.

[0032] Step S221: Initialize the hidden layer parameters and learning rate parameters in the preset neural network, and set the parameters of the preset improved butterfly optimization algorithm. The parameters include Algorithm Step1: Set various parameters of the butterfly population, including the initial position of each butterfly, the number of butterfly populations, the maximum number of iterations, the dimension of the space where it is located, the sensory modality, the power exponent, and the dynamic conversion probability Step S222: Randomly generate a butterfly population with a preset population size, uniformly distribute the butterfly population based on the ICMIC chaos mapping method, and change the initial position of the butterfly population to obtain the optimized butterfly population position; Specifically, the ICMIC chaos mapping is described as follows:

[0033] In the formula, is the control parameter, and its value range is (0, +∞), and the range of its chaotic orbit state value is (-1, 1). is the initial position of the butterfly population, represents the position of the butterfly population after update; The expression of the changed population distribution is described as:

[0034] In the formula, represents the butterfly population distribution improved by the ICMIC chaos mapping, represents the position of the improved butterfly population.

[0035] Step S223: Screen the butterfly population by using differential mutation, crossover, and selection strategies to obtain the screened excellent butterfly population; It can be understood that when the improved butterfly optimization algorithm runs in this step, in each iteration process of the butterfly population, the position and fitness value of the current butterfly population will be calculated. Although the algorithm itself will update the butterfly population, there will inevitably be poor individuals. If we can make excellent selections for these iterated butterfly populations, theoretically, the speed of searching for the optimal value can be improved. Inspired by the genetic algorithm, it is proposed to introduce differential mutation, crossover, and selection strategies before calculating the position and fitness of the next butterfly population to obtain the screened excellent butterfly population; Optimize the neural network parameters by improving the butterfly optimization algorithm, enabling the neural network to achieve better fitting effects on the training data. By gradually optimizing the butterfly population, the optimization process becomes more concentrated and efficient, thus obtaining the optimal butterfly population information. The neural network parameters corresponding to these optimal position information can significantly enhance the prediction ability and generalization ability of the model, providing reliable optimization parameters for subsequent steps.

[0036] Its three main stages are specifically described as follows: Step1: Mutation Generate a mutation difference vector according to each mutation vector;

[0037] In the formula, is the mutation vector of the th iteration, , , are three random positions in the butterfly population; represents the maximum number of iterations; represents the current iteration number; represents the natural exponential function, that is, the exponential function with the real number e (e≈2.71828) as the base; is an intermediate variable with no practical significance; , , The random index is a randomly selected integer, and ; is a real number scaling factor, takes values in [0,2], and it controls the amplitude of the differential mutation; represents the mutation rate; Step2: Crossover After the previous stage is completed, by performing a crossover operation on the perturbation parameter vector, the perturbation parameters are made diverse, thereby generating a trial vector. Specifically described as:

[0038] In the formula, is a random crossover parameter, with a value range of [0.8,1], is a random number between [0,1]; represents the trial vector; represents the target vector; represents the target vector of the next generation iteration; Step3: Selection After the initial butterfly population goes through the mutation and crossover stages, the test vector is utilized according to the greedy principle and the target vector are compared to determine whether it can become the gene iteration individual of the next generation. If the fitness value of the test vector is less than that of the target vector , then is set as , otherwise, the old value is retained.

[0039] Its expression description is as follows:

[0040] In the formula, represents the test vector; represents the target vector; represents the target vector for the next generation iteration; Step S224: Calculate the fitness value of the screened excellent butterfly population, and based on the calculated fitness value, as well as a random number and a dynamic conversion probability, introduce a selection factor for global search and dynamically adjust mutation for local search until the number of iterations reaches the maximum number of iterations, obtaining the optimal hidden layer parameters and learning rate parameters.

[0041] It can be understood that this step includes Step S2241, Step S2242, and Step S2243.

[0042] Step S2241: Calculate the fitness value of the screened excellent butterfly population based on a preset fitness function, and sort the calculated fitness values of the butterfly population to select the optimal fitness value of the butterfly population; Among them, the preset fitness function is as follows:

[0043] Among them, represents the position of the butterfly population at the th iteration; represents the output result of the neural network model, represents the expected value.

[0044] It can be understood that the corresponding optimal position can be obtained through the optimal fitness value of the butterfly population; Step S2242: Generate a random number to judge the relationship with the preset dynamic conversion probability to make a strategy selection, and introduce a selection factor for global search and dynamically adjust mutation for local search to update the butterfly position; Among them, the formula for updating the position of the butterfly population is as follows;

[0045]

[0046] It is understandable that when each butterfly emits a certain amount of fragrance, the fragrance will spread within the area where it is located and be perceived by other butterflies. The fragrance released by a butterfly is related to its fitness, which means that when the position of a butterfly changes, its fitness will also change accordingly. When a butterfly detects that another butterfly in the area releases a stronger fragrance, it will move in that direction. This stage is called global search. In addition, there is a situation where when a butterfly cannot perceive a fragrance stronger than its own, it will move freely, and this stage is called the local search stage. The update formula is based on the generated random number and the set dynamic transition probability to make a strategy choice. If , the global search strategy is executed; otherwise, the local search strategy is executed, and a selection factor is introduced for global search and dynamic adjustment of mutation for local search; during the global search process of the butterfly optimization algorithm. In order to improve the convergence speed of the algorithm while ensuring the diversity of the population, the selection factor in the genetic algorithm is introduced to replace the random number . Using dynamic adjustment of mutation for local search can avoid premature convergence. In the early stage of local search, it can jump out of the local optimum and find the optimal solution; in the later stage of iteration, when the convergence requirement is met, the mutation factor approaches 0, making the search stable; In the formula, represents the number of butterflies in the population; represents the stimulation factor of the -th butterfly in the improved butterfly optimization algorithm, represents the selection factor, which is the probability that the -th butterfly moves towards the best butterfly at the current iteration, , , respectively represent the positions of the -th, -th, -th butterflies in the butterfly population at the -th generation, is the position of the -th butterfly in the butterfly population at the -th generation; is a random number within [0, 1]; represents the best position of the butterfly population, represents the set dynamic transition probability; represents the -th butterfly's fragrance perception intensity; represents the current iteration number; Represents the quantity of a preset butterfly population; is a constant, is the probability that the th butterfly in the generation moves towards the best butterfly;

[0047] Step S2243, check whether the end condition is satisfied. If it is satisfied, that is, the maximum number of iterations T is reached, then the algorithm ends and outputs the best solution found, namely the corresponding optimized hidden layer parameters and learning rate parameters; otherwise, return to Step S2242 to continue the next round of iteration. It can be understood that in this step, the neural network after training is updated with the parameters optimized by the improved butterfly optimization algorithm, so that the model parameters can better adapt to the characteristics of the data and the prediction task. The optimized hidden layer parameters and learning rate parameters can improve the prediction accuracy and generalization ability of the model, so as to more accurately predict the design scheme of the prefabricated frame beam slope.

[0048] Step S24, send the training set and the test set to the neural network after assignment for training and verification to obtain an optimized prediction model for the design scheme of the prefabricated frame beam slope.

[0049] In this step, by assigning the optimized parameters to the neural network, the neural network model is in an optimal state, so that it can be trained and predicted more accurately. When the optimized neural network processes the prediction of the design scheme of the prefabricated frame beam slope, it will have higher accuracy and generalization ability, and can provide more reliable support for engineering practical applications. In this step, Step S24 includes Step S241, Step S242, and Step S243.

[0050] Step S241, send the hidden layer parameters and the learning rate parameters to the neural network for forward calculation formula construction to obtain the forward calculation formula of the neural network after assignment; The construction of the forward calculation formula includes:

[0051] In the formula, is the forget gate of the neural network, is the recurrent weight of the input gate, is the recurrent weight of the forget gate, is the recurrent weight of the output gate, is the recurrent weight of the cell state, is the sigmoid function, is the neural network at time short-term memory, is the short-term memory of the neural network at moment, is the input operation parameter sequence at the bias term of the input gate, the bias term of the forget gate, the bias term of the output gate, the bias term of the cell state, is the input gate, is the memory cell state at is the intermediate value of the updated memory cell state, Tanh is the hyperbolic tangent activation function; is the memory cell state at is the output gate.

[0052] Step S242: Send the training set to the neural network after assignment for retraining. Among them, the output of the neural network after assignment at a preset moment is calculated through the forward calculation formula, and the loss value at the preset moment is calculated based on the output; It can be understood that in this step, the loss value at the preset moment is calculated through a preset loss value calculation formula, and the loss value calculation formula is as follows:

[0053] Among them, represents the loss value at represents the observed value of the training sample at moment, represents the output of the neural network model at moment.

[0054] Step S243: Update the loss value at the preset moment and predict the data at each moment to obtain the output result corresponding to the historical monitoring data in the neural network.

[0055] In this step, by updating the loss value and predicting the data at each moment, the parameters of the model can be gradually optimized, so that the model can more accurately predict the design scheme of the precast frame beam slope. By predicting the historical monitoring data, the performance of the model can be verified, and valuable prediction results can be provided for actual engineering applications.

[0056] Step S3: Send the historical monitoring data of the precast frame beam slope to the optimized precast frame beam slope design scheme prediction model to predict the precast frame beam slope design scheme.

[0057] It can be understood that the prefabricated frame beam slope design scheme prediction system can consider multiple parameters, has strong adaptability, fault tolerance ability and self-improving ability, and improves the calculation accuracy.

[0058] In another embodiment, the sparrow optimization algorithm can also be adopted to initialize the position information of the sparrow population, set the maximum number of iterations, construct the fitness function, and calculate the fitness value of the sparrow through the fitness function, and select the smallest fitness value as the initial fitness of the sparrow; Construct the sparrow discoverer function model, sparrow follower function model and sparrow warning function model respectively, and calculate the first new position of the sparrow discoverer, the second new position of the sparrow follower and the third new position of the sparrow warning function respectively; Update the initial fitness according to whether the fitness of the first new position is better than the initial fitness, update the fitness of the first new position according to whether the fitness of the second new position is better than the fitness of the first new position, and update the initial fitness of the second new position according to whether the fitness of the third new position is better than the initial fitness of the second new position; Obtain the best position information of the sparrow through the updated fitness, construct a neural network, and assign values to the hidden layer parameters and learning rate parameters in the neural network based on the best sparrow position information to obtain the optimized hidden layer parameters and learning rate parameters; It can be understood that in this step, the neural network is updated with the parameters optimized by the moth-flame optimization algorithm, so that the model parameters can better adapt to the characteristics of the data and the prediction task. The optimized hidden layer parameters and learning rate parameters can improve the prediction accuracy and generalization ability of the model, so as to more accurately predict the prefabricated frame beam slope design scheme.

[0059] Send the hidden layer parameters and learning rate parameters to the neural network to construct the forward calculation formula, and obtain the LSTM model; Send the training set to the neural network after assignment for training again, calculate the output of the neural network at the preset time through the forward calculation formula of the neural network, and calculate the loss value at the preset time based on the output of the neural network at the preset time; It can be understood that in this step, the loss value at the preset time is calculated through the preset loss value calculation formula, and the loss value calculation formula is as follows:

[0060] where, represents the loss value at the time, represents the observed value of the training sample at the time, Output at a moment.

[0061] Update the loss value at the preset moment, and predict the data at each moment to obtain the output result corresponding to the historical monitoring data in the neural network; In this step, by updating the loss value and predicting the data at each moment, the parameters of the model can be gradually optimized, enabling the model to more accurately predict the design scheme of the prefabricated and assembled frame beam slope. By predicting the historical monitoring data, the performance of the model can be verified, and valuable prediction results can be provided for practical engineering applications.

[0062] Send the training set and the test set to the assigned neural network for training and verification to obtain an optimized design scheme for the prefabricated frame beam slope.

[0063] It can be understood that in this step, through re-training and verification, it is ensured that the optimized neural network model not only performs well on the training set but also has good generalization ability on the test set. This step makes the model have practical application value, can effectively predict the situation of the prefabricated frame beam slope design scheme, and provides reliable data support for engineering decision-making.

[0064] Among many deep learning models, RNN introduces the concept of time series. Relying on its cyclic network structure, the ability to maintain memory, and the ability to process non-linear data, it shows stronger adaptability in solving the long-term dependence problem of sequence data. However, when the number of network layers increases, RNN has problems such as gradient vanishing and gradient explosion in information feedback, and it is difficult for RNN to model large-span time series. LSTM solves the problem that RNN cannot establish a prediction model for large-span time series. The LSTM model replaces the neurons in the hidden layer of the RNN model with memory units with a "gating" mechanism, which consists of a storage unit and three logic gates: forget, input, and output. The "gating" mechanism controls the writing of new information and the forgetting of previously accumulated information, thus solving the long-term dependence problem in the cyclic network structure and avoiding gradient explosion and gradient vanishing caused by the multiplication effect in gradient backpropagation in the cyclic network structure. And the number of hidden layers HN and the learning rate directly affect the output of the LSTM prediction model. SSA can be used to find the optimal values of HN and . In the LSTM-SSA model, the root mean square error of the training data set is used as the fitness of the LSTM model for optimal search, and finally the optimal parameter values are found to improve the optimization of the prefabricated frame beam design scheme, reduce the error rate in complex designs, reduce the disaster incidence rate, ensure the life and property safety of relevant personnel, and improve the design accuracy and speed.

[0065] Therefore, by using the method provided by the present invention and the long short-term memory network model (LSTM), relevant parameters such as different geological environments, construction conditions, and concrete materials are used as input feature values of the neural network model. All the collected and sorted design materials and data are incorporated into the established neural network. Its samples expand as the data after subsequent construction such as monitoring increases and the subsequent construction plans are continuously supplemented. When it is necessary to treat and protect the slope, by inputting the engineering area, strength parameters of rock and soil masses, other external load strength parameters, etc. into the constructed neural network, the neural network first predicts all possible precast frame beam design plans according to the content of the input layer, and searches for all design plans through the sparrow search algorithm (SSA) to obtain the optimal design plan. This design system can automatically compare the plans, select the optimal plan that meets all requirements, save costs, make more efficient use of the economy, and thus minimize the time loss caused by factors such as complex regional geological conditions as much as possible, and improve design safety.

[0066] In an exemplary embodiment of the present invention, the initialization of the sparrow population position information includes:

[0067] In the formula, is the initial position information of the sparrow population, is the th hidden layer parameter of the th iteration, is the learning rate parameter of the

[0068] th hidden layer of the th iteration. Wherein, the maximum number of iterations , the current number of iterations , the constant and the dimension are set. In the LSTM-SSA model, the sparrow position is composed of the parameters to be optimized in the LSTM model (the hidden layer parameter is

[0069] and the learning rate parameter is

[0070] In the formula, is the fitness value of sparrows, is the number of training sets, is the The output result of a training set in a neural network model, is the observed value of the training set sample at time. In an exemplary embodiment of the present invention, the sparrow finder function model includes:

[0071] In the formula, is the sparrow position information calculated by the sparrow finder function model at the th iteration, is the maximum number of iterations, is the current number of iterations, is the learning rate, is a random number obeying the normal distribution, is a unit vector, is the warning value of the sparrow population position, is the safety threshold of the sparrow population position.

[0072] Among them, when it means that the warning value is less than the safety value, there is no predator in the foraging environment at this time, and the finder can perform extensive search operations; when it means that some sparrows in the population have found a predator and issued a warning to other sparrows in the population, and all sparrows need to fly to a safe area to forage Secondly, the sparrow follower function model includes:

[0073] In the formula, is the sparrow position information calculated by the sparrow follower function model at the th iteration, is the optimal position found by the finder at the th iteration, is a 1*d matrix whose elements are randomly assigned 1 or -1, where d is a natural number, is a natural number.

[0074] Among them, when is satisfied, when it indicates that the th joiner has not obtained food and is in a hungry state. At this time, it needs to fly to other places to forage to obtain more energy.

[0075] Secondly, the sparrow early warning function model includes:

[0076] In the formula, The position information of the sparrow at the -th iteration calculated by the sparrow early warning function model, is the global optimal position at the -th iteration, is a random number obeying the standard normal distribution and used as a step size control parameter, is the position information of the sparrow at the -th iteration, is the fitness value of the current sparrow individual, is the global best fitness value, is a random number in [0, 1], is the global worst position at the -th iteration, is the global worst fitness value, is the tolerance value, is the number of iterations, is the initialization parameter after the Logistic chaotic map iteration.

[0077] Among them, when it means that the sparrow is at the edge of the population and is extremely vulnerable to predators. When it means that the sparrows in the middle of the population are also in danger, and at this time they need to approach other sparrows to reduce the risk of being preyed on. At the same time, to improve the convergence speed and accuracy of SSA, and are corrected.

[0078] An exemplary implementation manner of the present invention for constructing a forward calculation formula includes:

[0079] In the formula, is the forget gate of the neural network, is the recurrent weight of the input gate, is the recurrent weight of the forget gate, is the recurrent weight of the output gate, is the recurrent weight of the cell state, is the sigmoid function, is the short-term memory of the neural network at time, is the short-term memory of the neural network at time, is the input operation parameter sequence at time, is the bias term of the input gate, is the bias term of the forget gate, is the bias term of the output gate, is the bias term of the cell state, is the input gate, is the state of memory cells at time is the intermediate value of the updated memory cell state, and Tanh is the hyperbolic tangent activation function; is the state of memory cells at time is the output gate.

[0080] In this embodiment, the current design calculation and scheme comparison of prefabricated frame beam slopes are still in the stage of manual operation. From design to construction completion, the time cost will greatly increase the possibility of disasters. The prefabricated frame beam slope design based on the improved SSA-LSTM neural network has a high operation speed, reduces the time cost, and at the same time reduces the performance loss.

[0081] For mountainous and complex areas, the large number of parameters considered in the prefabricated frame beam slope design greatly increases the error rate of manual operation. The prefabricated frame beam slope design based on the SSA-LSTM neural network can consider multiple parameters, has strong adaptability, fault tolerance and self-improving capabilities, and improves the design accuracy; Due to the difficulty in obtaining relevant engineering data, it is impossible to provide accurate references for the project. To prevent disasters, the design of artificial prefabricated frame beams is too conservative, and it is difficult to design a very optimized design scheme, wasting a lot of manpower and material resources. The prefabricated frame beam slope design based on the SSA-LSTM neural network can automatically search for prefabricated frame beam slope design schemes that meet any requirements according to user needs, reducing the labor costs of design and construction.

[0082] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0083] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for designing a slope reinforcement using an assembled anchor cable frame beam, characterized in that: include: Step S1: Obtain the prefabricated frame beam slope design data and historical monitoring data, integrate them into data samples, and classify the data samples to obtain training sets and test sets; Step S2: sending the training set and the test set to a preset neural network for training and verification, and optimizing the preset neural network by a preset improved butterfly optimization algorithm to obtain an optimized prediction model of the prefabricated frame beam slope design scheme; Step S3: sending the historical monitoring data of the prefabricated frame beam slope to the optimized prefabricated frame beam slope design scheme prediction model to predict the prefabricated frame beam slope design scheme.

2. The method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 1, characterized in that: The step S2 comprises: Step S21: sending the training set to a preset neural network for training to obtain a trained neural network; Step S22: optimizing the hidden layer parameters and learning rate parameters of the trained neural network based on the improved butterfly optimization algorithm to obtain optimized hidden layer parameters and learning rate parameters; Step S23: assigning the optimized hidden layer parameters and learning rate parameters to the trained neural network to obtain the assigned neural network; Step S24: sending the training set and the test set to the assigned neural network for training and verification, to obtain an optimized prediction model for the slope design scheme of the prefabricated frame beam.

3. The method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 2, characterized in that: The step S22 comprises: Step S221: Initializing the hidden layer parameters and learning rate parameters in the preset neural network, and setting the parameters of the preset improved butterfly optimization algorithm, the parameters including the initial position of each butterfly, the number of butterfly populations, the maximum number of iterations, the dimension of the space, the sensory modality, the power index and the dynamic conversion probability; Step S222: randomly generating a butterfly group with a preset population size, uniformly distributing the butterfly group based on the ICMIC chaotic mapping method, and changing the initial position of the butterfly group to obtain an optimized butterfly group position; Step S223: using differential mutation, crossover, and selection strategies to screen the butterfly population, and obtaining a screened excellent butterfly population; Step S224: Calculate the fitness value of the selected excellent butterfly population, and introduce a selection factor based on the calculated fitness value and random numbers and dynamic conversion probabilities to perform global search and dynamically adjust mutations to perform local search until the number of iterations reaches the maximum number of iterations, thereby obtaining the optimal hidden layer parameters and learning rate parameters.

4. The method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 3 is characterized in that: The step S224 includes: Step S2241: Calculate the fitness value of the selected excellent butterfly population based on a preset fitness function, sort the calculated fitness values ​​of the butterfly population, and select the optimal fitness value of the butterfly population; Step S2242: Generate a random number to judge the relationship between it and the preset dynamic conversion probability and make a strategy selection, introduce a selection factor to perform a global search, dynamically adjust the variation to perform a local search and update the butterfly position; Step S2243: Check whether the end condition is met. If it is met, that is, the maximum number of iterations T is reached, the algorithm ends and outputs the best solution found, which is the optimized hidden layer parameters and learning rate parameters; otherwise, return to step S2242 and continue to the next round of iteration.

5. The method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 4, characterized in that: The updating of butterfly position includes: In the formula, represents the number of butterflies in the population; Indicates the first The stimulus of a butterfly, Represents the selection factor, which is the number of iterations under the current number of iterations. The probability that the choice of a butterfly moves to the best butterfly, , , They represent the butterfly population in The first Only Only The butterfly corresponds to the position, For butterfly populations in The first The butterfly corresponds to the position; is a random number in [0,1]; represents the best position for the butterfly population, Indicates the set dynamic conversion probability; Indicates The intensity of scent perception of a butterfly; Indicates the current iteration number; Indicates the number of preset butterfly populations; is a constant, For the Next The probability that the choice of a butterfly moves to the best butterfly.

6. A method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 5, characterized in that: The step S24 comprises: Step S241, sending the hidden layer parameters and the learning rate parameters to the neural network to construct a forward calculation formula, and obtaining a forward calculation formula of the neural network after assignment; Step S242, sending the training set to the neural network after the assignment for further training, wherein the output of the neural network after the assignment at a preset time is calculated by a forward calculation formula, and the loss value at the preset time is calculated based on the output; Step S243: update the loss value at the preset moment, and predict the data at each moment to obtain the output result corresponding to the historical monitoring data in the neural network.

7. The method for designing a slope reinforcement using an assembled anchor cable frame beam according to claim 6, characterized in that: The construction of the forward calculation formula includes: In the formula, is the forget gate of the neural network, is the recursive weight of the input gate, is the recursive weight of the forget gate, is the recursive weight of the output gate, is the recursive weight of the unit state, is the sigmoid function, For neural networks Short-term memory of the moment, For neural networks Short-term memory of the moment, for Enter the operating parameter sequence at any time. is the bias term of the input gate, is the bias term of the forget gate, is the bias term of the output gate, is the bias term of the cell state, is the input gate, for The state of memory cells at each moment, is the intermediate value of the updated memory cell state, Tanh is the hyperbolic tangent activation function; for The state of memory cells at each moment, is the output gate.

8. A prefabricated anchor cable frame beam slope reinforcement design system, characterized in that: include: The data collection module is configured to obtain the prefabricated frame beam slope design data and historical monitoring data, integrate them into data samples, and classify the data samples to obtain training sets and test sets; A model building module is configured to send the training set and the test set to a preset neural network for training and verification, and optimize the preset neural network through a preset improved butterfly optimization algorithm to obtain an optimized prediction model of the prefabricated frame beam slope design scheme; A solution output module is configured to send the historical monitoring data of the prefabricated frame beam slope to the optimized prefabricated frame beam slope design solution prediction model to predict the prefabricated frame beam slope design solution; The main control module is connected to the data collection module, the model building module and the solution output module, and is used to execute the design method for reinforcing the slope with an assembled anchor frame beam as described in any one of claims 1 to 7.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for designing a slope reinforced with an assembled anchor frame beam as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, it implements the method for designing slope reinforcement of an assembled anchor frame beam as described in any one of claims 1-7.

Images