Processing method and device for predicting stress-strain curve of concrete member

Through the curve prediction model of BP network and LSTM stacking network + fully connected layer, the problem of difficulty in collecting stress data is solved, efficient prediction of the stress and strain curve of concrete components is achieved, and design efficiency is improved.

CN119047034APending Publication Date: 2024-11-29LINDU LVJIAN (JIANGSU) ENG DESIGN & RES CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411086032.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

In the design process of reinforced concrete components, the collection of stress data is difficult and requires a lot of engineering calculations, resulting in inefficient design.

Method used

A curve prediction model based on BP network and LSTM stacking network + fully connected layer is used to predict the stress and strain curve of concrete components by training the design parameters and strain data.

Benefits of technology

With only strain data, the stress and strain curve can be effectively constructed, which reduces the calculation amount of design work and improves design efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047034A_ABST
    Figure CN119047034A_ABST
Patent Text Reader

Abstract

The embodiment of the invention relates to a processing method and device for predicting a stress-strain curve of a concrete member. The method comprises the following steps: constructing a curve prediction model; performing model training on the curve prediction model; after the model training is finished, receiving a first design parameter list and a first strain data sequence of the first concrete member; and performing curve prediction by using a curve prediction model according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve. According to the method, the engineering calculation workload of concrete member design work can be reduced, and the design efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of data processing, and in particular to a processing method and device for predicting a stress-strain curve of a concrete member. Background Art

[0002] In the design stage of reinforced concrete components (referred to as concrete components), the compressive performance of the components can be intuitively analyzed through the stress-strain curve. To obtain the stress-strain curve, it is necessary to continuously collect stress-strain data during the compression process in the compression test. We found a problem in the experiment: the collection method of strain data is relatively simple and direct, but stress data cannot be collected directly and requires a lot of engineering calculations to determine. This problem will undoubtedly bring a large amount of calculation workload to the component design work and reduce the design efficiency. Summary of the invention

[0003] The purpose of the present invention is to provide a processing method, device, electronic device and computer-readable storage medium for predicting the stress-strain curve of a concrete component in view of the defects of the prior art. The present invention constructs a curve prediction model based on a BP network and an LSTM stacked network + a fully connected layer; and completes the model training by training the BP network and the LSTM stacked network + a fully connected layer separately; and after the model training is completed, the curve prediction model is used to predict the stress-strain curve according to the design parameter list and strain data sequence of any concrete component. The present invention is applied to the design work of concrete components, and the stress-strain curve can be constructed when there is only a strain data sequence. The present invention can not only reduce the calculation workload of the design work, but also improve the design efficiency.

[0004] To achieve the above object, a first aspect of an embodiment of the present invention provides a method for predicting a stress-strain curve of a concrete member, the method comprising:

[0005] Constructing a curve prediction model; the model input end of the curve prediction model includes a first model input end and a second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with 19 types of normalized design parameters of a concrete component; the second model input end is used to receive a segmented sequence string of a normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output a stress-strain curve of the concrete component;

[0006] Performing model training on the curve prediction model;

[0007] After the model training is completed, a first design parameter list and a first strain data sequence of a first concrete component are received; and the curve prediction model is used to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve.

[0008] Preferably, the curve prediction model includes a BP network, a first denormalization module, an LSTM stacked network, a fully connected layer and a second denormalization module;

[0009] The input end of the BP network is connected to the input end of the first model, and the output end is connected to the input end of the first denormalization module; the BP network includes an input layer, a hidden layer and an output layer; the unit transfer functions used in the hidden layer and the output layer are both sigmoid functions;

[0010] The input layer is composed of 19 parallel input layer units; each of the input layer units is used to receive a type of normalized design parameter in the design parameter vector;

[0011] The hidden layer is composed of a plurality of parallel hidden layer units, each of which is fully connected to all the input layer units; each of the hidden layer units is used to perform a fully connected calculation on the normalized design parameters of all the input layer units connected to itself to generate a corresponding first fully connected vector, and perform an activation process on the first fully connected vector based on a sigmoid function to obtain a corresponding first activation vector;

[0012] The output layer is composed of two parallel output layer units, each of which is fully connected to all the hidden layer units; the first output layer unit is used to perform a fully connected calculation on the first activation vectors of all the hidden layer units connected to itself to generate a corresponding second fully connected vector, and perform activation processing on the second fully connected vector based on the sigmoid function to obtain corresponding first activation data, and send the first activation data as the corresponding normalized peak stress data to the first denormalization module; the second output layer unit is used to perform a fully connected calculation on the first activation vectors of all the hidden layer units connected to itself to generate a corresponding third fully connected vector, and perform activation processing on the third fully connected vector based on the sigmoid function to obtain corresponding second activation data, and send the second activation data as the corresponding normalized peak strain data to the first denormalization module;

[0013] The output end of the first denormalization module is connected to the first input end of the second denormalization module; the first denormalization module is used to perform denormalization processing on the normalized peak stress data output by the BP network according to the preset maximum and minimum peak stress data thresholds to obtain the corresponding peak stress data f cc According to the preset maximum and minimum peak strain data thresholds, the normalized peak strain data output by the BP network is denormalized to obtain the corresponding peak strain data ε cc ; and the peak stress data f cc and the peak strain data ε cc Sending to the second denormalization module;

[0014] The input end of the LSTM stacked network is connected to the input end of the second model, and the output end is connected to the input end of the fully connected layer; the LSTM stacked network is formed by stacking multiple LSTM layers; the end hidden vector h of the LSTM layer of the previous layer in the LSTM stacked network e Passed to the next LSTM layer as the corresponding starting hidden vector h s ;

[0015] The LSTM stacking network dynamically sets the total number of stacked layers of the LSTM layer based on the total number of the segmentation sequences of the segmentation sequence string, and the LSTM layer corresponds to the segmentation sequence one by one;

[0016] Each of the LSTM layers is used to receive the corresponding segmentation sequence; and when it is the first LSTM layer, based on the initial hidden vector h 0 Set its own starting hidden vector h s ; and when it is not the first LSTM layer, the end hidden vector h is based on the LSTM layer of the previous layer e Set its own starting hidden vector h s ; and according to the starting hidden vector h currently set s and extract strain features from the currently received segmentation sequence to obtain a corresponding stress feature vector; and send the stress feature vector to the fully connected layer; and send its own end hidden vector h e The stress feature vector is transmitted to the next LSTM layer; the vector length of the stress feature vector is consistent with the sequence length of the segmentation sequence;

[0017] The output end of the fully connected layer is connected to the second input end of the denormalization module; the fully connected layer is used to sequentially splice all the stress feature vectors output by the LSTM stacked network to obtain the corresponding full-sequence stress feature vector; and perform normalized strain data regression calculation according to the full-sequence stress feature vector to obtain the corresponding normalized stress data sequence; and send the normalized stress data sequence to the second denormalization module; the sequence length of the normalized stress data sequence is consistent with the sequence length of the normalized strain data sequence corresponding to the segmented sequence string;

[0018] The input end of the second denormalization module includes the first input end, the second input end and the third input end, and the third input end is connected to the second model input end; the output end of the second denormalization module is connected to the model output end; the second denormalization module is used to restore the corresponding normalized strain data sequence by sequentially splicing all the segmentation sequences of the segmentation sequence string; and based on the peak strain data ε cc The normalized strain data sequence is denormalized to obtain a corresponding denormalized strain data sequence; and based on the peak stress data f cc The normalized stress data sequence is denormalized to obtain a corresponding denormalized stress data sequence; and stress data and strain data of each pair of sequence index matching in the denormalized stress data sequence and the denormalized strain data sequence are extracted according to the sequence index matching relationship to form a corresponding stress-strain data pair; and all the obtained stress-strain data pairs are sequentially sorted to form a corresponding stress-strain data pair sequence; and two-dimensional curve fitting is performed according to the stress-strain data pair sequence to obtain the corresponding stress-strain curve and output it.

[0019] Preferably, the nineteen categories of normalized design parameters include normalized stirrup minor axis size parameters, normalized stirrup major axis size parameters, normalized stirrup cross-sectional area parameters, normalized stirrup cross-sectional inertia moment parameters, normalized stirrup shape coding parameters, normalized stirrup spacing parameters, normalized stirrup reinforcement ratio parameters, normalized stirrup strength parameters, normalized stirrup limb number parameters, normalized concrete strength parameters, normalized cross-sectional size parameters, normalized concrete cover thickness parameters, normalized component longitudinal size parameters, normalized component aspect ratio parameters, normalized component cross-sectional shape parameters, normalized longitudinal reinforcement strength parameters, normalized longitudinal reinforcement ratio parameters, normalized stirrup eigenvalue parameters and normalized reinforcement eigenvalue parameters.

[0020] Preferably, the performing model training on the curve prediction model specifically includes:

[0021] The BP network of the curve prediction model is used as the corresponding first training model MA; and the LSTM stacked network and the fully connected layer of the curve prediction model form a corresponding second training model MB;

[0022] The value range of the total number of hidden layer units of the BP network is set to obtain a corresponding first value range; the value range of the segmentation sequence length corresponding to each LSTM layer of the LSTM stacked network is set to a corresponding second value range; the first value range includes multiple first value N i , index i is an integer greater than 1; the second value range includes multiple second values ​​L j , index j is an integer greater than 1;

[0023] And based on each of the first values ​​N i The total number of hidden layer units of the BP network of the first training model MA is set to obtain the corresponding first training model MA i ; and based on each of the second values ​​L j The structure of each LSTM layer of the second training model MB is set to obtain the corresponding second training model MB j ;

[0024] And based on the genetic algorithm, all the first training models MA are trained according to the preset first data set. i Perform optimal model screening to obtain a corresponding optimal model; and optimize and train the optimal model according to the first data set based on a preset first model parameter optimizer to obtain a corresponding optimized model; and solidify the parameters of the BP network based on the model parameters of the optimized model; the first data set includes a plurality of first data records; the first data record includes a first training vector and a first label vector; the first training vector is composed of a group of the nineteen categories of normalized design parameters; the first label vector is composed of two normalized data labels, namely a normalized peak stress data label and a normalized peak strain data label; the first model parameter optimizer includes an SGD optimizer and an ADAM optimizer;

[0025] Based on the preset second model parameter optimizer, each of the second training models MB is optimized according to the preset second data set. j Perform model training; and complete the training of all the second training models MB jPerform model performance evaluation and determine the corresponding optimal model parameters based on the evaluation results; and solidify the parameters of the LSTM stacked network and the fully connected layer based on the optimal model parameters; the second data set includes multiple second data records; the second data record includes a first training strain data sequence and a first label stress data sequence; the first training strain data sequence and the first label stress data sequence are both a normalized sequence; the second model parameter optimizer includes an SGD optimizer and an ADAM optimizer;

[0026] After the parameters of the BP network, the LSTM stacked network and the fully connected layer are solidified, it is confirmed that the model training of the curve prediction model is completed.

[0027] Furthermore, the genetic algorithm is used to train all the first training models MA according to the preset first data set. i Perform optimal model screening to obtain the corresponding optimal model, including:

[0028] Step 51, randomly extracting a plurality of the first data records from the first data set based on a preset first extraction ratio to form a corresponding second data set;

[0029] wherein the first extraction ratio is a percentage less than 1; the total number of records in the first data set matches the product of the first extraction ratio and the total number of records in the second data set;

[0030] Step 52: for each of the first training models MA i Initialize the model parameters of the first training model MA i The current model parameters are used to construct a model parameter vector to obtain a corresponding first model parameter vector; and for each of the first training models MA i Set a corresponding first model error, and initialize each of the first model errors to 1; and initialize a first counter to 1;

[0031] Step 53: each of the first training models MA ias the corresponding current training model; and traverse all the first data records of the second data set; and during the traversal, use the first data record currently traversed as the corresponding current training record; and input the first training vector of the current training record into the current training model for processing to obtain the corresponding normalized peak stress prediction data and normalized peak strain prediction data; and form a corresponding first prediction vector from the normalized peak stress prediction data and the normalized peak strain prediction data; and form a corresponding first prediction vector from the first prediction vector and the first label vector of the current training record; and at the end of this traversal, bring all the first prediction / label vector pairs obtained into a preset first error function for error calculation to obtain the corresponding first prediction error, and reset the first model error corresponding to the current training model to the corresponding first prediction error;

[0032] Wherein, the first error function at least includes an MSE error function;

[0033] Step 54: each of the first training models MA i The inverse of the corresponding first model error is used as the corresponding first model fitness; and all the first model fitnesses are summed up to obtain the corresponding first fitness sum; and the ratio of each first model fitness to the first fitness sum is used as the corresponding first model selection probability; and all the first training model MAs are selected in descending order of the first model selection probabilities. i Sorting to obtain the corresponding first training model sequence;

[0034] Step 55: training the first training model MA i The total number of is counted to obtain the corresponding first total number; and whether the first total number is an odd number is identified; if so, go to step 56; if not, go to step 57;

[0035] Step 56: The first training model MA in the first training model sequence i As the elite model for the time being; and the second and third first training model MA of the first training model sequence i Form a corresponding first model pair, and convert the 4th and 5th first training models MA i A corresponding first model pair is formed, and so on, until the penultimate and the last first training model MA are formed. iThe first training model MA of each first model pair is formed until a corresponding first model pair is formed; and the first model parameter vector of the elite model at that time is directly used as the corresponding first model crossover parameter vector; and based on a preset individual crossover algorithm and a preset first crossover probability, the two first training model MAs of each first model pair are i Perform vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors; and go to step 58;

[0036] The individual crossover algorithm is a type of chromosome crossover algorithm in the GA algorithm, and the chromosome crossover algorithm includes at least a single-point crossover algorithm, a multi-point crossover algorithm, and an arithmetic crossover algorithm;

[0037] Step 57: The first and second first training models MA of the first training model sequence are i Form a corresponding first model pair, and convert the third and fourth first training models MA i A corresponding first model pair is formed, and so on, until the penultimate and the last first training model MA are formed. i until a corresponding first model pair is formed; and based on the individual crossover algorithm and the first crossover probability, the two first training models MA of each first model pair are i Performing vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors;

[0038] Step 58: Based on a preset individual mutation algorithm and a preset first mutation probability, each of the first training models MA i Performing vector data mutation processing on the first model cross parameter vector to obtain a corresponding first model mutation parameter vector;

[0039] The individual mutation algorithm is a type of chromosome mutation algorithm in the GA algorithm, and the chromosome mutation algorithm at least includes a basic position mutation algorithm, a uniform mutation algorithm, a non-uniform mutation algorithm, and a Gaussian approximate mutation algorithm;

[0040] Step 59, identifying whether the first counter exceeds a preset first counter threshold; if the first counter does not exceed the first counter threshold, adding 1 to the first counter, and performing a first training model MA corresponding to the current vector based on each of the first model variation parameter vectors. i The model parameters are reset once, and the process returns to step 53; if the first counter exceeds the first counter threshold, the first first training model MA in the latest first training model sequence is reset. iAs the corresponding optimal model, the model parameters of the optimal model are solidified once, and the optimal model screening is ended.

[0041] Furthermore, the preset first model parameter optimizer optimizes the optimal model according to the first data set to obtain a corresponding optimized model, specifically including:

[0042] Step 61, extracting the first first data record of the first data set as the corresponding current training record;

[0043] Step 62, inputting the first training vector of the current training record into the optimal model for processing to obtain corresponding normalized peak stress prediction data and normalized peak strain prediction data to form a corresponding second prediction vector;

[0044] Step 63, inputting the second prediction vector and the first label vector of the current training record into a preset first model loss function to calculate and obtain a corresponding first loss value;

[0045] Wherein, the first model loss function at least includes an L1 loss function and an L2 loss function;

[0046] Step 64, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first data set, if so, turning to step 65, if not, extracting the next first data record of the first data set as the new current training record and returning to step 62; if the first loss value does not satisfy the first loss value range, performing a round of model parameter modulation on the optimal model in the direction of minimizing the first model loss function based on the first model parameter optimizer, and returning to step 62 at the end of this round of modulation;

[0047] Step 65, solidifying the model parameters of the optimal model, and using the optimal model with solidified parameters as the corresponding optimization model.

[0048] Furthermore, the preset second model parameter optimizer optimizes each of the second training models MB according to the preset second data set. j Perform model training, including:

[0049] Step 71: extract the first second data record of the second data set as the corresponding current training record; and use the second training model MB currently being trained as jas the corresponding current training model; and the second value L corresponding to the current training model j As the corresponding current sequence length;

[0050] Step 72, dividing the first training strain data sequence of the current training record into a plurality of first training segmentation sequences based on the current sequence length; and arranging all the obtained first training segmentation sequences in order to form a corresponding first training segmentation sequence string;

[0051] Step 73, inputting the first training segmentation sequence string into the current training model for processing to obtain a normalized stress data sequence recorded as the corresponding first predicted stress data sequence;

[0052] Step 74, inputting the first predicted stress data sequence and the first label stress data sequence of the current training record into a preset second model loss function to calculate and obtain a corresponding second loss value;

[0053] Wherein, the second model loss function at least includes an L1 loss function and an L2 loss function;

[0054] Step 75, identifying whether the second loss value satisfies a preset second loss value range; if the second loss value satisfies the second loss value range, identifying whether the current training record is the last second data record of the second data set, if so, turning to step 76, if not, extracting the next second data record of the second data set as the new current training record and returning to step 72; if the second loss value does not satisfy the second loss value range, performing a round of model parameter modulation on the current training model in the direction of minimizing the second model loss function based on the second model parameter optimizer, and returning to step 73 at the end of this round of modulation;

[0055] Step 76, solidify the model parameters of the current training model.

[0056] Furthermore, all the second training models MB that have completed training j Evaluate model performance and determine the corresponding optimal model parameters based on the evaluation results, including:

[0057] Step 81, randomly extracting a plurality of first data records from the first data set based on a preset second extraction ratio to form a corresponding first evaluation set;

[0058] wherein the second extraction ratio is a percentage less than 1; the total number of records in the first data set matches the product of the first extraction ratio and the total number of records in the first evaluation set;

[0059] Step 82: each of the second training models MB j as the corresponding current evaluation model; and the second value L corresponding to the current evaluation model j as the corresponding current sequence length; and traverse all the first data records of the first evaluation set; and during the traversal, use the currently traversed first data record as the corresponding current evaluation record; and based on the current sequence length, divide the first training strain data sequence of the current evaluation record into multiple second training segmentation sequences; and sequentially sort all the obtained second training segmentation sequences to form a corresponding second training segmentation sequence string; and input the second training segmentation sequence string into the current evaluation model for processing to obtain a normalized stress data sequence recorded as the corresponding second predicted stress data sequence; and form a corresponding first prediction / label sequence pair from the second predicted stress data sequence and the first label stress data sequence of the current evaluation record; and at the end of this traversal, bring all the obtained first prediction / label sequence pairs into a preset second error function for error calculation to obtain the corresponding second prediction error;

[0060] Wherein, the second error function at least includes a MAE error function, an MSE error function, and a RMSE error function;

[0061] Step 83: The second training model MB having the smallest second prediction error is used as j as the corresponding optimal performance model; and the model parameters of the optimal performance model are used as the corresponding optimal model parameters.

[0062] Preferably, the using the curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain the corresponding first stress-strain curve specifically includes:

[0063] Performing normalized parameter tensor preparation according to the first design parameter list to obtain a corresponding first design parameter vector;

[0064] The first design parameter vector is input into the BP network of the curve prediction model for processing to obtain a pair of normalized peak stress / strain data recorded as corresponding normalized stress prediction value and normalized strain prediction value; and the normalized stress prediction value and normalized strain prediction value are input into the first denormalization module of the curve prediction model for processing to obtain a pair of denormalized peak stress / strain data recorded as corresponding first stress prediction value and first strain prediction value;

[0065] and taking 0 as the minimum strain data threshold and the first strain prediction value as the maximum strain data threshold, normalizing the first strain data sequence according to the maximum and minimum strain data thresholds to obtain a normalized data sequence recorded as the corresponding second strain data sequence; and taking the segmentation sequence length corresponding to the LSTM layer of the curve prediction model as the corresponding first sequence length; and dividing the second strain data sequence into a plurality of first segmentation sequences according to the first sequence length; and sorting all the obtained first segmentation sequences in order to form a corresponding first segmentation sequence string;

[0066] and inputting the first segmentation sequence string into the LSTM stacked network of the curve prediction model for processing to obtain corresponding multiple first stress feature vectors;

[0067] and inputting all the obtained first stress feature vectors into the fully connected layer of the curve prediction model for processing to obtain a normalized stress data sequence recorded as the corresponding first stress data sequence;

[0068] The first segmentation sequence string, the first stress data sequence, the first stress prediction value and the first strain prediction value are input into the second denormalization module of the curve prediction model for processing to obtain the corresponding first stress-strain curve.

[0069] Furthermore, the preparing of normalizing the parameter tensor according to the first design parameter list to obtain the corresponding first design parameter vector specifically includes:

[0070] Extracting nineteen types of designated design parameters from the first design parameter list; the nineteen types of designated design parameters include stirrup minor axis dimension parameters, stirrup major axis dimension parameters, stirrup cross-sectional area parameters, stirrup cross-sectional inertia moment parameters, stirrup shape coding parameters, stirrup spacing parameters, stirrup reinforcement ratio parameters, stirrup strength parameters, stirrup limb number parameters, concrete strength parameters, cross-sectional dimension parameters, concrete cover thickness parameters, component longitudinal dimension parameters, component aspect ratio parameters, component cross-sectional shape parameters, longitudinal reinforcement strength parameters, longitudinal reinforcement ratio parameters, stirrup characteristic value parameters, and reinforcement characteristic value parameters;

[0071] Based on the maximum and minimum parameter thresholds corresponding to each type of the nineteen types of specified design parameters, normalization calculation is performed on each type of the specified design parameters, and the nineteen normalized calculation results are used to form the corresponding nineteen types of normalized design parameters; and the nineteen types of normalized design parameters are used to form the corresponding first design parameter vector.

[0072] A second aspect of an embodiment of the present invention provides a device for implementing the processing method for predicting a stress-strain curve of a concrete member as described in the first aspect, the device comprising: a model building module, a model training module and a model application module;

[0073] The model building module is used to build a curve prediction model; the model input end of the curve prediction model includes a first model input end and a second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with 19 types of normalized design parameters of the concrete component; the second model input end is used to receive a segmented sequence string of the normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output the stress-strain curve of the concrete component;

[0074] The model training module is used to perform model training on the curve prediction model;

[0075] The model application module is used to receive a first design parameter list and a first strain data sequence of a first concrete component after model training is completed; and use the curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve.

[0076] A third aspect of an embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;

[0077] The processor is used to be coupled to the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;

[0078] The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

[0079] A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions. When the computer instructions are executed by a computer, the computer executes the instructions of the method described in the first aspect above.

[0080] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for predicting the stress-strain curve of a concrete component. As can be seen from the above content, the embodiment of the present invention constructs a curve prediction model based on a BP network and an LSTM stacked network + a fully connected layer; and completes the model training by training the BP network and the LSTM stacked network + a fully connected layer separately; and after the model training is completed, the curve prediction model is used to predict the stress-strain curve according to the design parameter list and strain data sequence of any concrete component. When the embodiment of the present invention is applied to the design work of concrete components, the stress-strain curve can be constructed when there is only a strain data sequence. The embodiment of the present invention reduces the computational workload while also improving the design efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1 A schematic diagram of a processing method for predicting a stress-strain curve of a concrete member provided in Embodiment 1 of the present invention;

[0082] Figure 2 A schematic diagram of a curve prediction model provided in the first embodiment of the present invention;

[0083] Figure 3 A module structure diagram of a processing device for predicting a stress-strain curve of a concrete member provided in the second embodiment of the present invention;

[0084] Figure 4 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0086] Embodiment 1 of the present invention provides a method for predicting a stress-strain curve of a concrete component. Figure 1 As shown in the schematic diagram of a processing method for predicting a stress-strain curve of a concrete member provided in the first embodiment of the present invention, the method mainly comprises the following steps:

[0087] Step 1: Build a curve prediction model.

[0088] Here, the model input end of the curve prediction model of the embodiment of the present invention includes a first model input end and a second model input end, the model input end of the curve prediction model includes the first model input end and the second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with nineteen categories of normalized design parameters of the concrete component; the second model input end is used to receive a segmented sequence string of the normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output the stress-strain curve of the concrete component. It should be noted that the nineteen categories of normalized design parameters mentioned above include: normalized stirrup minor axis size parameters, normalized stirrup major axis size parameters, normalized stirrup cross-sectional area parameters, normalized stirrup cross-sectional inertia parameters, normalized stirrup shape coding parameters, normalized stirrup spacing parameters, normalized stirrup reinforcement ratio parameters, normalized stirrup strength parameters, normalized stirrup number parameters, normalized concrete strength parameters, normalized cross-sectional size parameters, normalized concrete cover thickness parameters, normalized member longitudinal size parameters, normalized member aspect ratio parameters, normalized member cross-sectional shape parameters, normalized longitudinal reinforcement strength parameters, normalized longitudinal reinforcement ratio parameters, normalized stirrup eigenvalue parameters and normalized reinforcement eigenvalue parameters.

[0089] like Figure 2 As shown in the model schematic diagram of the curve prediction model provided in the first embodiment of the present invention, the curve prediction model of the embodiment of the present invention includes a BP network, a first denormalization module, an LSTM stacked network, a fully connected layer and a second denormalization module; wherein:

[0090] 1) BP network:

[0091] The input end of the BP network of the embodiment of the present invention is connected to the input end of the first model, and the output end is connected to the input end of the first denormalization module; the BP network includes an input layer, a hidden layer and an output layer; the unit transfer functions used in the hidden layer and the output layer are both sigmoid functions;

[0092] The input layer consists of 19 parallel input layer units; each input layer unit is used to receive a class of normalized design parameters in the design parameter vector;

[0093] The hidden layer is composed of multiple parallel hidden layer units, and each hidden layer unit is fully connected to all input layer units; each hidden layer unit is used to perform full connection calculation on the normalized design parameters of all input layer units connected to itself to generate a corresponding first fully connected vector, and activate the first fully connected vector based on the sigmoid function to obtain a corresponding first activation vector;

[0094] The output layer is composed of two parallel output layer units, each of which is fully connected to all hidden layer units; the first output layer unit is used to perform a fully connected calculation on the first activation vectors of all hidden layer units connected to itself to generate a corresponding second fully connected vector, and perform activation processing on the second fully connected vector based on the sigmoid function to obtain corresponding first activation data, and send the first activation data as the corresponding normalized peak stress data to the first anti-normalization module; the second output layer unit is used to perform a fully connected calculation on the first activation vectors of all hidden layer units connected to itself to generate a corresponding third fully connected vector, and perform activation processing on the third fully connected vector based on the sigmoid function to obtain corresponding second activation data, and send the second activation data as the corresponding normalized peak strain data to the first anti-normalization module;

[0095] 2) The first anti-normalization module:

[0096] The output end of the first denormalization module of the embodiment of the present invention is connected to the first input end of the second denormalization module;

[0097] The first denormalization module is used to perform denormalization processing on the normalized peak stress data output by the BP network according to the preset maximum and minimum peak stress data thresholds to obtain the corresponding peak stress data f cc According to the preset maximum and minimum peak strain data thresholds, the normalized peak strain data output by the BP network is denormalized to obtain the corresponding peak strain data ε cc ; and the peak stress data f cc and peak strain data ε cc Sending to the second anti-normalization module;

[0098] Here, the preset maximum and minimum peak stress data thresholds and the preset maximum and minimum peak strain data thresholds are two pairs of preset empirical thresholds;

[0099] 3) LSTM stacked network:

[0100] The input end of the LSTM stacked network of the embodiment of the present invention is connected to the input end of the second model, and the output end is connected to the input end of the fully connected layer; the LSTM stacked network is formed by stacking multiple LSTM layers; the end hidden vector h of the upper LSTM layer in the LSTM stacked network e Passed to the next LSTM layer as the corresponding starting hidden vector h s ;

[0101] The LSTM stacking network dynamically sets the total number of stacked LSTM layers based on the total number of segmentation sequences of the segmentation sequence string, and the LSTM layers correspond to the segmentation sequences one by one;

[0102] Each LSTM layer is used to receive the corresponding segmentation sequence; and when it is the first LSTM layer, it is based on the initial hidden vector h 0 Set its own starting hidden vector h s ; and when it is not the first LSTM layer, it is based on the end hidden vector h passed by the previous LSTM layer e Set its own starting hidden vector h s ; and according to the currently set starting hidden vector h s The strain feature extraction is performed on the currently received segmentation sequence to obtain the corresponding stress feature vector; the stress feature vector is sent to the fully connected layer; and its own end hidden vector h e Pass to the next LSTM layer; the vector length of the stress feature vector is consistent with the sequence length of the segmentation sequence;

[0103] 4) Fully connected layer:

[0104] The output end of the fully connected layer of the embodiment of the present invention is connected to the second input end of the denormalization module; the fully connected layer is used to sequentially splice all stress feature vectors output by the LSTM stacking network to obtain a corresponding full-sequence stress feature vector; and perform normalized strain data regression calculation according to the full-sequence stress feature vector to obtain a corresponding normalized stress data sequence; and send the normalized stress data sequence to the second denormalization module; the sequence length of the normalized stress data sequence is consistent with the sequence length of the normalized strain data sequence corresponding to the segmented sequence string;

[0105] 5) The second anti-normalization module:

[0106] The input end of the second denormalization module of the embodiment of the present invention includes a first input end, a second input end and a third input end, and the third input end is connected to the second model input end; the output end of the second denormalization module is connected to the model output end; the second denormalization module is used to restore the corresponding normalized strain data sequence by sequentially splicing all the segmentation sequences of the segmentation sequence string; and based on the peak strain data ε cc The normalized strain data sequence is denormalized to obtain the corresponding denormalized strain data sequence; and based on the peak stress data f cc The normalized stress data sequence is denormalized to obtain the corresponding denormalized stress data sequence; and the stress data and strain data of each pair of sequence index matching in the denormalized stress data sequence and the denormalized strain data sequence are extracted according to the sequence index matching relationship to form a corresponding stress-strain data pair; and all the obtained stress-strain data pairs are sequentially sorted to form a corresponding stress-strain data pair sequence; and two-dimensional curve fitting is performed on the stress-strain data pair sequence to obtain the corresponding stress-strain curve and output it.

[0107] Step 2, training the curve prediction model;

[0108] Specifically, it includes: step 21, using the BP network of the curve prediction model as the corresponding first training model MA; and forming the corresponding second training model MB by the LSTM stacked network and the fully connected layer of the curve prediction model;

[0109] Step 22, setting the value range of the total number of hidden layer units of the BP network to obtain a corresponding first value range; setting the value range of the segmentation sequence length corresponding to each LSTM layer of the LSTM stack network to obtain a corresponding second value range;

[0110] The first value range includes multiple first values ​​N i , index i is an integer greater than 1; the second value range includes multiple second values ​​L j , index j is an integer greater than 1;

[0111] Here, when setting the two types of value ranges, the setting may be based on two pre-customized value range setting rules, or may be based on a random manner, or may be based on actual application requirements;

[0112] Step 23, and based on each first value N i The total number of hidden layer units of the BP network of the first training model MA is set to obtain the corresponding first training model MA i ; and based on each second value L j The structure of each LSTM layer of the second training model MB is set to obtain the corresponding second training model MB j ;

[0113] Step 24, based on the genetic algorithm, all the first training models MA are trained according to the preset first data set. i Performing optimal model screening to obtain a corresponding optimal model; and optimizing and training the optimal model based on a preset first model parameter optimizer according to a first data set to obtain a corresponding optimized model; and solidifying parameters of the BP network based on the model parameters of the optimized model;

[0114] The first data set includes a plurality of first data records; the first data record includes a first training vector and a first label vector; the first training vector is composed of a set of 19 categories of normalized design parameters; the first label vector is composed of two normalized data labels, namely a normalized peak stress data label and a normalized peak strain data label; the first model parameter optimizer includes an SGD optimizer and an ADAM optimizer;

[0115] Specifically, step 241 includes: based on the genetic algorithm, all first training models MA are trained according to the preset first data set.i Perform optimal model screening to obtain the corresponding optimal model;

[0116] Specifically, it includes: step 2411, randomly extracting a plurality of first data records from the first data set based on a preset first extraction ratio to form a corresponding second data set;

[0117] wherein the first extraction ratio is a percentage less than 1; the total number of records in the first data set multiplied by the first extraction ratio matches the total number of records in the second data set;

[0118] Step 2412: for each first training model MA i Initialize the model parameters of each first training model MA i The current model parameters are used to construct the model parameter vector to obtain the corresponding first model parameter vector; and for each first training model MA i Set a corresponding first model error, and initialize each first model error to 1; and initialize the first counter to 1;

[0119] Step 2413: each first training model MA i as the corresponding current training model; and traverse all the first data records of the second data set; and during the traversal, use the currently traversed first data record as the corresponding current training record; and input the first training vector of the current training record into the current training model for processing to obtain the corresponding normalized peak stress prediction data and normalized peak strain prediction data; and form a corresponding first prediction vector from the normalized peak stress prediction data and the normalized peak strain prediction data; and form a corresponding first prediction / label vector pair from the first prediction vector and the first label vector of the current training record; and at the end of this traversal, bring all the obtained first prediction / label vector pairs into the preset first error function for error calculation to obtain the corresponding first prediction error, and reset the first model error corresponding to the current training model to the corresponding first prediction error;

[0120] Wherein, the first error function at least includes an MSE error function;

[0121] Step 2414: each first training model MA i The inverse of the corresponding first model error is used as the corresponding first model fitness; and the sum of all the first model fitnesses is calculated to obtain the corresponding first fitness sum; and the ratio of each first model fitness to the first fitness sum is used as the corresponding first model selection probability; and all first training model MAs are selected in descending order of the first model selection probability. i Sorting to obtain the corresponding first training model sequence;

[0122] Here, the smaller the first model error is, the greater the corresponding first model fitness is, and the corresponding first model selection probability is also greater, that is, the first training model sequence is sorted in descending order of model performance;

[0123] Step 2415: training the first training model MA i The total number of is counted to obtain the corresponding first total number; and whether the first total number is an odd number is identified; if so, go to step 2416; if not, go to step 2417;

[0124] Step 2416: The first training model MA in the first training model sequence i As the elite model of the time; and the second and third first training model MA of the first training model sequence i Form a corresponding first model pair, and convert the 4th and 5th first training models MA i Form a corresponding first model pair, and so on, until the penultimate and last first training model MA i until a corresponding first model pair is formed; and the first model parameter vector of the elite model of the time is directly used as the corresponding first model crossover parameter vector; and based on the preset individual crossover algorithm and the preset first crossover probability, the two first training model MAs of each first model pair are i Perform vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors; and go to step 2418;

[0125] The individual crossover algorithm is a type of chromosome crossover algorithm in the GA algorithm, and the chromosome crossover algorithm includes at least a single-point crossover algorithm, a multi-point crossover algorithm, and an arithmetic crossover algorithm; the first crossover probability is a preset probability value, such as 0.4;

[0126] Step 2417: The first and second first training models MA of the first training model sequence are i Form a corresponding first model pair, and combine the 3rd and 4th first training models MA i Form a corresponding first model pair, and so on, until the penultimate and last first training model MA i Until a corresponding first model pair is formed; and based on the individual crossover algorithm and the first crossover probability, the two first training models MA of each first model pair i Performing vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors;

[0127] Step 2418: Based on the preset individual mutation algorithm and the preset first mutation probability, each first training model MA iPerforming vector data mutation processing on the first model cross parameter vector to obtain a corresponding first model mutation parameter vector;

[0128] The individual mutation algorithm is a type of chromosome mutation algorithm in the GA algorithm, and the chromosome mutation algorithm includes at least a basic position mutation algorithm, a uniform mutation algorithm, a non-uniform mutation algorithm, and a Gaussian approximate mutation algorithm; the first mutation probability is a preset probability value, such as 0.01;

[0129] Step 2419, identifying whether the first counter exceeds the preset first counter threshold; if the first counter does not exceed the first counter threshold, the first counter is increased by 1, and the first training model MA corresponding to the current vector is modified based on each first model variation parameter vector. i The model parameters are reset once, and the process returns to step 2413; if the first counter exceeds the first counter threshold, the first first training model MA in the latest first training model sequence is reset. i As the corresponding optimal model, the model parameters of the optimal model are solidified once, and the optimal model screening is ended;

[0130] Here, the first counter threshold is a preset integer;

[0131] Step 242, optimizing and training the optimal model based on the first data set using a preset first model parameter optimizer to obtain a corresponding optimized model;

[0132] Specifically, it includes: step 2421, extracting the first first data record of the first data set as the corresponding current training record;

[0133] Step 2422, input the first training vector of the current training record into the optimal model for processing to obtain corresponding normalized peak stress prediction data and normalized peak strain prediction data to form a corresponding second prediction vector;

[0134] Step 2423, inputting the second prediction vector and the first label vector of the current training record into a preset first model loss function to calculate and obtain a corresponding first loss value;

[0135] Wherein, the first model loss function includes at least an L1 loss function and an L2 loss function;

[0136] Step 2424, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first data set, if so, turning to step 2425, if not, extracting the next first data record of the first data set as a new current training record and returning to step 2422; if the first loss value does not satisfy the first loss value range, performing a round of model parameter modulation on the optimal model in the direction of minimizing the first model loss function based on the first model parameter optimizer, and returning to step 2422 at the end of this round of modulation;

[0137] Here, the first loss value range is a preset loss value range;

[0138] Step 2425, solidify the model parameters of the optimal model, and use the optimal model with solidified parameters as the corresponding optimization model;

[0139] Step 243, solidifying the parameters of the BP network based on the model parameters of the optimization model;

[0140] Step 25, based on the preset second model parameter optimizer, each second training model MB is optimized according to the preset second data set. j Perform model training; and complete the training of all second training models MB j Evaluate model performance and determine the corresponding optimal model parameters based on the evaluation results; and solidify the parameters of the LSTM stacked network and the fully connected layer based on the optimal model parameters;

[0141] Wherein, the second data set includes a plurality of second data records; the second data record includes a first training strain data sequence and a first label stress data sequence; the first training strain data sequence and the first label stress data sequence are both a normalized sequence; the second model parameter optimizer includes an SGD optimizer and an ADAM optimizer;

[0142] Specifically, step 251 includes: based on a preset second model parameter optimizer, each second training model MB is optimized according to a preset second data set. j Conduct model training;

[0143] Specifically, step 2511 extracts the first second data record of the second data set as the corresponding current training record; and converts the second training model MB of the current training into j as the corresponding current training model; and the second value L corresponding to the current training model j As the corresponding current sequence length;

[0144] Step 2512, dividing the first training strain data sequence of the current training record into a plurality of first training segmentation sequences based on the current sequence length; and arranging all the obtained first training segmentation sequences in order to form a corresponding first training segmentation sequence string;

[0145] Step 2513, inputting the first training segmentation sequence string into the current training model for processing to obtain a normalized stress data sequence recorded as the corresponding first predicted stress data sequence;

[0146] Step 2514, inputting the first predicted stress data sequence and the first label stress data sequence of the current training record into a preset second model loss function to calculate and obtain a corresponding second loss value;

[0147] Wherein, the second model loss function includes at least an L1 loss function and an L2 loss function;

[0148] Step 2515, identifying whether the second loss value satisfies a preset second loss value range; if the second loss value satisfies the second loss value range, identifying whether the current training record is the last second data record of the second data set, if so, turning to step 2516, if not, extracting the next second data record of the second data set as a new current training record and returning to step 2512; if the second loss value does not satisfy the second loss value range, performing a round of model parameter modulation on the current training model in the direction of minimizing the second model loss function based on the second model parameter optimizer, and returning to step 2513 at the end of this round of modulation;

[0149] Here, the second loss value range is a preset loss value range;

[0150] Step 2516, solidify the model parameters of the current training model;

[0151] Step 252, all second training models MB that have completed training are j Conduct model performance evaluation and determine the corresponding optimal model parameters based on the evaluation results;

[0152] Specifically, it includes: step 2521, randomly extracting a plurality of first data records from the first data set based on a preset second extraction ratio to form a corresponding first evaluation set;

[0153] wherein the second extraction ratio is a percentage less than 1; the total number of records in the first data set matches the product of the first extraction ratio and the total number of records in the first evaluation set;

[0154] Step 2522: each second training model MB j as the corresponding current evaluation model; and the second value L corresponding to the current evaluation modelj as the corresponding current sequence length; and traverse all the first data records of the first evaluation set; and during the traversal, use the currently traversed first data record as the corresponding current evaluation record; and based on the current sequence length, divide the first training strain data sequence of the current evaluation record into multiple second training segmentation sequences; and sort all the obtained second training segmentation sequences in order to form a corresponding second training segmentation sequence string; and input the second training segmentation sequence string into the current evaluation model for processing to obtain a normalized stress data sequence recorded as the corresponding second predicted stress data sequence; and form a corresponding first prediction / label sequence pair from the second predicted stress data sequence and the first label stress data sequence of the current evaluation record; and at the end of this traversal, bring all the obtained first prediction / label sequence pairs into the preset second error function for error calculation to obtain the corresponding second prediction error;

[0155] Wherein, the second error function at least includes a MAE error function, an MSE error function, and a RMSE error function;

[0156] Step 2523: The second training model MB with the smallest second prediction error is j as the corresponding optimal performance model; and taking the model parameters of the optimal performance model as the corresponding optimal model parameters;

[0157] Step 253, and solidify the parameters of the LSTM stacked network and the fully connected layer based on the optimal model parameters;

[0158] Step 26: After the parameters of the BP network, the LSTM stacked network and the fully connected layer are solidified, confirm that the model training of the curve prediction model is completed.

[0159] Step 3, after the model training is completed, a first design parameter list and a first strain data sequence of the first concrete component are received; and a curve prediction model is used to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve;

[0160] Specifically, it includes: step 31, after the model training is completed, receiving a first design parameter list and a first strain data sequence of a first concrete component;

[0161] Step 32, using a curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve;

[0162] Specifically, it includes: step 321, performing normalized parameter tensor preparation according to the first design parameter list to obtain a corresponding first design parameter vector;

[0163] Specifically, it includes: step 3211, extracting nineteen categories of designated design parameters from the first design parameter list;

[0164] Among them, the 19 types of specified design parameters include stirrup minor axis size parameters, stirrup major axis size parameters, stirrup cross-sectional area parameters, stirrup cross-sectional inertia parameters, stirrup shape coding parameters, stirrup spacing parameters, stirrup reinforcement ratio parameters, stirrup strength parameters, stirrup leg number parameters, concrete strength parameters, cross-sectional size parameters, concrete cover thickness parameters, component longitudinal size parameters, component aspect ratio parameters, component cross-sectional shape parameters, longitudinal reinforcement strength parameters, longitudinal reinforcement ratio parameters, stirrup characteristic value parameters and reinforcement characteristic value parameters;

[0165] Step 3212, based on the maximum and minimum parameter thresholds corresponding to each of the 19 types of specified design parameters, normalization calculation is performed on each type of specified design parameters, and the obtained 19 normalization calculation results form corresponding 19 types of normalized design parameters; and the obtained 19 types of normalized design parameters form a corresponding first design parameter vector;

[0166] Here, the embodiment of the present invention pre-sets the maximum and minimum parameter thresholds corresponding to each type of specified design parameter;

[0167] Step 322, and input the first design parameter vector into the BP network of the curve prediction model for processing to obtain a pair of normalized peak stress / strain data recorded as corresponding normalized stress prediction values ​​and normalized strain prediction values; and input the normalized stress prediction values ​​and normalized strain prediction values ​​into the first denormalization module of the curve prediction model for processing to obtain a pair of denormalized peak stress / strain data recorded as corresponding first stress prediction values ​​and first strain prediction values;

[0168] Step 323, with 0 as the minimum strain data threshold and the first strain prediction value as the maximum strain data threshold, the first strain data sequence is normalized according to the maximum and minimum strain data thresholds to obtain a normalized data sequence recorded as the corresponding second strain data sequence; and the segmentation sequence length corresponding to the LSTM layer of the curve prediction model is used as the corresponding first sequence length; and the second strain data sequence is equally divided into a plurality of first segmentation sequences according to the first sequence length; and all the obtained first segmentation sequences are sequentially sorted to form a corresponding first segmentation sequence string;

[0169] Step 324, input the first segmented sequence string into the LSTM stacked network of the curve prediction model for processing to obtain corresponding multiple first stress feature vectors;

[0170] Step 325, input all the obtained first stress characteristic vectors into the fully connected layer of the curve prediction model for processing to obtain a normalized stress data sequence recorded as the corresponding first stress data sequence;

[0171] In step 326, the first segmentation sequence string, the first stress data sequence, the first stress prediction value and the first strain prediction value are input into a second denormalization module of the curve prediction model for processing to obtain a corresponding first stress-strain curve.

[0172] Figure 3 This is a module structure diagram of a processing device for predicting stress-strain curves of concrete components provided in Embodiment 2 of the present invention. The device is a terminal device or server that implements the aforementioned method embodiment, or can be a device that enables the aforementioned terminal device or server to implement the aforementioned method embodiment. For example, the device can be a device or chip system of the aforementioned terminal device or server. Figure 3 As shown, the device includes: a model building module 201, a model training module 202 and a model application module 203.

[0173] The model building module 201 is used to build a curve prediction model; the model input end of the curve prediction model includes a first model input end and a second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with nineteen categories of normalized design parameters of the concrete component; the second model input end is used to receive a segmented sequence string of the normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output the stress-strain curve of the concrete component.

[0174] The model training module 202 is used to perform model training on the curve prediction model.

[0175] The model application module 203 is used to receive the first design parameter list and the first strain data sequence of the first concrete component after the model training is completed; and use the curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain the corresponding first stress-strain curve.

[0176] An embodiment of the present invention provides a processing device for predicting a stress-strain curve of a concrete member, which can execute the method steps in the above method embodiment. Its implementation principle and technical effect are similar and will not be repeated here.

[0177] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the model building module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above-mentioned module is determined. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0178] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital signal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0179] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the above method embodiments are generated. The above computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the above-mentioned computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, Bluetooth, microwave, etc.) methods. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The above-mentioned available medium can be a magnetic medium (such as a floppy disk, a hard disk, a tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)), etc.

[0180] Figure 4 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. The electronic device may be a terminal device or a server for implementing the method of the aforementioned embodiment, or may be a terminal device or a server for implementing the method of the aforementioned embodiment connected to the aforementioned terminal device or server. Figure 4 As shown, the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; the transceiver 303 is coupled to the processor 301, and the processor 301 controls the transceiver 303. Various instructions may be stored in the memory 302 to complete various processing functions and implement the processing steps described in the aforementioned embodiment method. Preferably, the electronic device involved in the embodiment of the present invention also includes: a power supply 304, a system bus 305 and a communication port 306. The system bus 305 is used to realize the communication connection between components. The above-mentioned communication port 306 is used for connecting and communicating between the electronic device and other peripherals.

[0181] exist Figure 4The system bus 305 mentioned in the figure can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM), and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk storage.

[0182] The above-mentioned processor can be a general-purpose processor, including a central processing unit CPU, a network processor (Network Processor, NP), a graphics processing unit (Graphics Processing Unit, GPU), etc.; it can also be a digital signal processor DSP, an application-specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0183] It should be noted that an embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the method and processing process provided in the above embodiments.

[0184] An embodiment of the present invention further provides a chip for executing instructions, wherein the chip is used to execute the processing steps described in the aforementioned method embodiment.

[0185] The embodiment of the present invention provides a processing method, device, electronic device and computer-readable storage medium for predicting the stress-strain curve of a concrete component. As can be seen from the above content, the embodiment of the present invention constructs a curve prediction model based on a BP network and an LSTM stacked network + a fully connected layer; and completes the model training by training the BP network and the LSTM stacked network + a fully connected layer separately; and after the model training is completed, the curve prediction model is used to predict the stress-strain curve according to the design parameter list and strain data sequence of any concrete component. When the embodiment of the present invention is applied to the design work of concrete components, the stress-strain curve can be constructed when there is only a strain data sequence. The embodiment of the present invention reduces the computational workload while also improving the design efficiency.

[0186] The professionals should further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0187] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0188] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting stress-strain curve of concrete components, characterized in that: The method comprises: Constructing a curve prediction model; the model input end of the curve prediction model includes a first model input end and a second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with 19 types of normalized design parameters of a concrete component; the second model input end is used to receive a segmented sequence string of a normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output a stress-strain curve of the concrete component; Performing model training on the curve prediction model; After the model training is completed, a first design parameter list and a first strain data sequence of a first concrete component are received; and the curve prediction model is used to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve.

2. The method for predicting the stress-strain curve of a concrete member according to claim 1, characterized in that: The curve prediction model includes a BP network, a first anti-normalization module, an LSTM stacked network, a fully connected layer and a second anti-normalization module; The input end of the BP network is connected to the input end of the first model, and the output end is connected to the input end of the first denormalization module; the BP network includes an input layer, a hidden layer and an output layer; the unit transfer functions used in the hidden layer and the output layer are both sigmoid functions; The input layer is composed of 19 parallel input layer units; each of the input layer units is used to receive a type of normalized design parameter in the design parameter vector; The hidden layer is composed of a plurality of parallel hidden layer units, and each of the hidden layer units is fully connected to all the input layer units; Each of the hidden layer units is used to perform full connection calculation on the normalized design parameters of all the input layer units connected to itself to generate a corresponding first fully connected vector, and perform activation processing on the first fully connected vector based on a sigmode function to obtain a corresponding first activation vector; The output layer is composed of two parallel output layer units, and each of the output layer units is fully connected to all the hidden layer units; The first output layer unit is used to perform full connection calculation on the first activation vectors of all the hidden layer units connected to itself to generate a corresponding second fully connected vector, and perform activation processing on the second fully connected vector based on a sigmo id function to obtain corresponding first activation data, and send the first activation data as corresponding normalized peak stress data to the first denormalization module; The second output layer unit is used to perform full connection calculation on the first activation vectors of all the hidden layer units connected to itself to generate a corresponding third fully connected vector, and perform activation processing on the third fully connected vector based on the sigmo id function to obtain corresponding second activation data, and send the second activation data as the corresponding normalized peak strain data to the first denormalization module; The output end of the first denormalization module is connected to the first input end of the second denormalization module; The first denormalization module is used to perform denormalization processing on the normalized peak stress data output by the BP network according to the preset maximum and minimum peak stress data thresholds to obtain the corresponding peak stress data f cc ; According to the preset maximum and minimum peak strain data thresholds, the normalized peak strain data output by the BP network is denormalized to obtain the corresponding peak strain data ε cc ; and the peak stress data f cc and the peak strain data ε cc Sending to the second denormalization module; The input end of the LSTM stacked network is connected to the input end of the second model, and the output end is connected to the input end of the fully connected layer; The LSTM stacking network is formed by stacking multiple LSTM layers; the end hidden vector h of the LSTM layer of the previous layer in the LSTM stacking network e Passed to the next LSTM layer as the corresponding starting hidden vector h s ; The LSTM stacking network dynamically sets the total number of stacked layers of the LSTM layer based on the total number of the segmentation sequences of the segmentation sequence string, and the LSTM layer corresponds to the segmentation sequence one by one; Each of the LSTM layers is used to receive the corresponding segmentation sequence; And when it is the first LSTM layer, it sets its own starting hidden vector h based on the initial hidden vector h0 s ; and when it is not the first LSTM layer, the end hidden vector h is based on the LSTM layer of the previous layer e Set its own starting hidden vector h s ; and according to the starting hidden vector h currently set s and extract strain features from the currently received segmentation sequence to obtain a corresponding stress feature vector; and send the stress feature vector to the fully connected layer; and send its own end hidden vector h e The stress feature vector is transmitted to the next LSTM layer; the vector length of the stress feature vector is consistent with the sequence length of the segmentation sequence; The output end of the fully connected layer is connected to the second input end of the denormalization module; the fully connected layer is used to sequentially concatenate all the stress feature vectors output by the LSTM stacking network to obtain a corresponding full-sequence stress feature vector; and performing normalized strain data regression calculation according to the full sequence stress characteristic vector to obtain a corresponding normalized stress data sequence; and sending the normalized stress data sequence to the second denormalization module; the sequence length of the normalized stress data sequence is consistent with the sequence length of the normalized strain data sequence corresponding to the segmented sequence string; The input end of the second denormalization module includes the first input end, the second input end and the third input end, and the third input end is connected to the second model input end; the output end of the second denormalization module is connected to the model output end; the second denormalization module is used to restore the corresponding normalized strain data sequence by sequentially splicing all the segmentation sequences of the segmentation sequence string; Based on the peak strain data ε cc Performing a denormalization process on the normalized strain data sequence to obtain a corresponding denormalized strain data sequence; Based on the peak stress data f cc Performing a denormalization process on the normalized stress data sequence to obtain a corresponding denormalized stress data sequence; And according to the sequence index matching relationship, each pair of stress data and strain data matching the sequence index in the denormalized stress data sequence and the denormalized strain data sequence is extracted to form a corresponding stress-strain data pair; and all the obtained stress-strain data pairs are sequentially sorted to form a corresponding stress-strain data pair sequence; and two-dimensional curve fitting is performed according to the stress-strain data pair sequence to obtain the corresponding stress-strain curve and output it.

3. The method for predicting the stress-strain curve of a concrete member according to claim 2, characterized in that: The nineteen categories of normalized design parameters include normalized stirrup minor axis size parameters, normalized stirrup major axis size parameters, normalized stirrup cross-sectional area parameters, normalized stirrup cross-sectional inertia moment parameters, normalized stirrup shape coding parameters, normalized stirrup spacing parameters, normalized stirrup reinforcement ratio parameters, normalized stirrup strength parameters, normalized stirrup limb number parameters, normalized concrete strength parameters, normalized cross-sectional size parameters, normalized concrete cover thickness parameters, normalized component longitudinal size parameters, normalized component aspect ratio parameters, normalized component cross-sectional shape parameters, normalized longitudinal reinforcement strength parameters, normalized longitudinal reinforcement ratio parameters, normalized stirrup characteristic value parameters and normalized reinforcement characteristic value parameters.

4. The method for predicting the stress-strain curve of a concrete member according to claim 2, characterized in that: The performing model training on the curve prediction model specifically includes: The BP network of the curve prediction model is used as the corresponding first training model MA; and the LSTM stacked network and the fully connected layer of the curve prediction model form a corresponding second training model MB; The value range of the total number of hidden layer units of the BP network is set to obtain a corresponding first value range; the value range of the segmentation sequence length corresponding to each LSTM layer of the LSTM stacked network is set to a corresponding second value range; the first value range includes multiple first value N i , index i is an integer greater than 1; the second value range includes multiple second values ​​L j , index j is an integer greater than 1; And based on each of the first values ​​N i The total number of hidden layer units of the BP network of the first training model MA is set to obtain the corresponding first training model MA i ; and based on each of the second values ​​L j The structure of each LSTM layer of the second training model MB is set to obtain the corresponding second training model MB j ; Based on the genetic algorithm, all the first training models MA are trained according to the preset first data set. i Perform optimal model screening to obtain a corresponding optimal model; and optimize and train the optimal model according to the first data set based on a preset first model parameter optimizer to obtain a corresponding optimized model; and solidify the parameters of the BP network based on the model parameters of the optimized model; the first data set includes a plurality of first data records; the first data record includes a first training vector and a first label vector; the first training vector is composed of a group of the nineteen categories of normalized design parameters; the first label vector is composed of two normalized data labels, namely a normalized peak stress data label and a normalized peak strain data label; the first model parameter optimizer includes an SGD optimizer and an ADAM optimizer; Based on the preset second model parameter optimizer, each of the second training models MB is optimized according to the preset second data set. j Perform model training; and complete the training of all the second training models MB j Perform model performance evaluation and determine the corresponding optimal model parameters based on the evaluation results; and solidify the parameters of the LSTM stacked network and the fully connected layer based on the optimal model parameters; the second data set includes multiple second data records; the second data record includes a first training strain data sequence and a first label stress data sequence; the first training strain data sequence and the first label stress data sequence are both a normalized sequence; the second model parameter optimizer includes an SGD optimizer and an ADAM optimizer; After the parameters of the BP network, the LSTM stacked network and the fully connected layer are solidified, it is confirmed that the model training of the curve prediction model is completed.

5. The method for predicting the stress-strain curve of a concrete member according to claim 4, characterized in that: The first training model MA is trained based on a genetic algorithm according to a preset first data set. i Perform optimal model screening to obtain the corresponding optimal model, including: Step 51, randomly extracting a plurality of the first data records from the first data set based on a preset first extraction ratio to form a corresponding second data set; wherein the first extraction ratio is a percentage less than 1; the total number of records in the first data set matches the product of the first extraction ratio and the total number of records in the second data set; Step 52: for each of the first training models MA i Initialize the model parameters of the first training model MA i The current model parameters are used to construct a model parameter vector to obtain a corresponding first model parameter vector; and for each of the first training models MA i Set a corresponding first model error, and initialize each of the first model errors to 1; and initialize a first counter to 1; Step 53: each of the first training models MA i as the corresponding current training model; and traverse all the first data records of the second data set; and during the traversal, use the first data record currently traversed as the corresponding current training record; and input the first training vector of the current training record into the current training model for processing to obtain the corresponding normalized peak stress prediction data and normalized peak strain prediction data; and form a corresponding first prediction vector from the normalized peak stress prediction data and the normalized peak strain prediction data; and form a corresponding first prediction vector from the first prediction vector and the first label vector of the current training record; and at the end of this traversal, bring all the first prediction / label vector pairs obtained into a preset first error function for error calculation to obtain the corresponding first prediction error, and reset the first model error corresponding to the current training model to the corresponding first prediction error; Wherein, the first error function at least includes an MSE error function; Step 54: each of the first training models MA i The inverse of the corresponding first model error is used as the corresponding first model fitness; and all the first model fitnesses are summed up to obtain the corresponding first fitness sum; and the ratio of each first model fitness to the first fitness sum is used as the corresponding first model selection probability; and all the first training model MAs are selected in descending order of the first model selection probabilities. i Sorting to obtain the corresponding first training model sequence; Step 55: training the first training model MA i The total number of is counted to obtain the corresponding first total number; and whether the first total number is an odd number is identified; if so, go to step 56; if not, go to step 57; Step 56: The first training model MA in the first training model sequence i As the elite model for the time being; and the second and third first training model MA of the first training model sequence i Form a corresponding first model pair, and convert the 4th and 5th first training models MA i A corresponding first model pair is formed, and so on, until the penultimate and the last first training model MA are formed. i The first training model MA of each first model pair is formed until a corresponding first model pair is formed; and the first model parameter vector of the elite model at that time is directly used as the corresponding first model crossover parameter vector; and based on a preset individual crossover algorithm and a preset first crossover probability, the two first training model MAs of each first model pair are i Perform vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors; and go to step 58; The individual crossover algorithm is a type of chromosome crossover algorithm in the GA algorithm, and the chromosome crossover algorithm includes at least a single-point crossover algorithm, a multi-point crossover algorithm, and an arithmetic crossover algorithm; Step 57: The first and second first training models MA of the first training model sequence are i Form a corresponding first model pair, and convert the third and fourth first training models MA i A corresponding first model pair is formed, and so on, until the penultimate and the last first training model MA are formed. i until a corresponding first model pair is formed; and based on the individual crossover algorithm and the first crossover probability, the two first training models MA of each first model pair are i Performing vector data cross processing on the first model parameter vector to obtain two corresponding first model cross parameter vectors; Step 58: Based on a preset individual mutation algorithm and a preset first mutation probability, each of the first training models MA i Performing vector data mutation processing on the first model cross parameter vector to obtain a corresponding first model mutation parameter vector; The individual mutation algorithm is a type of chromosome mutation algorithm in the GA algorithm, and the chromosome mutation algorithm at least includes a basic position mutation algorithm, a uniform mutation algorithm, a non-uniform mutation algorithm, and a Gaussian approximate mutation algorithm; Step 59, identifying whether the first counter exceeds a preset first counter threshold; if the first counter does not exceed the first counter threshold, adding 1 to the first counter, and performing a first training model MA corresponding to the current vector based on each of the first model variation parameter vectors. i The model parameters are reset once, and the process returns to step 53; if the first counter exceeds the first counter threshold, the first first training model MA in the latest first training model sequence is reset. i As the corresponding optimal model, the model parameters of the optimal model are solidified once, and the optimal model screening is ended.

6. The method for predicting the stress-strain curve of a concrete member according to claim 4, characterized in that: The first model parameter optimizer based on the preset first model performs optimization training on the optimal model according to the first data set to obtain a corresponding optimization model, specifically including: Step 61, extracting the first first data record of the first data set as the corresponding current training record; Step 62, inputting the first training vector of the current training record into the optimal model for processing to obtain corresponding normalized peak stress prediction data and normalized peak strain prediction data to form a corresponding second prediction vector; Step 63, inputting the second prediction vector and the first label vector of the current training record into a preset first model loss function to calculate and obtain a corresponding first loss value; Wherein, the first model loss function at least includes an L1 loss function and an L2 loss function; Step 64, identifying whether the first loss value satisfies a preset first loss value range; if the first loss value satisfies the first loss value range, identifying whether the current training record is the last first data record of the first data set, if so, turning to step 65, if not, extracting the next first data record of the first data set as the new current training record and returning to step 62; if the first loss value does not satisfy the first loss value range, performing a round of model parameter modulation on the optimal model in the direction of minimizing the first model loss function based on the first model parameter optimizer, and returning to step 62 at the end of this round of modulation; Step 65, solidifying the model parameters of the optimal model, and using the optimal model with solidified parameters as the corresponding optimization model.

7. The method for predicting the stress-strain curve of a concrete member according to claim 4, characterized in that: The preset second model parameter optimizer optimizes each of the second training models MB according to the preset second data set. j Perform model training, including: Step 71: extract the first second data record of the second data set as the corresponding current training record; and use the second training model MB currently being trained as j as the corresponding current training model; and the second value L corresponding to the current training model j As the corresponding current sequence length; Step 72, dividing the first training strain data sequence of the current training record into a plurality of first training segmentation sequences based on the current sequence length; and arranging all the obtained first training segmentation sequences in order to form a corresponding first training segmentation sequence string; Step 73, inputting the first training segmentation sequence string into the current training model for processing to obtain a normalized stress data sequence recorded as the corresponding first predicted stress data sequence; Step 74, inputting the first predicted stress data sequence and the first label stress data sequence of the current training record into a preset second model loss function to calculate and obtain a corresponding second loss value; Wherein, the second model loss function at least includes an L1 loss function and an L2 loss function; Step 75, identifying whether the second loss value satisfies a preset second loss value range; if the second loss value satisfies the second loss value range, identifying whether the current training record is the last second data record of the second data set, if so, turning to step 76, if not, extracting the next second data record of the second data set as the new current training record and returning to step 72; if the second loss value does not satisfy the second loss value range, performing a round of model parameter modulation on the current training model in the direction of minimizing the second model loss function based on the second model parameter optimizer, and returning to step 73 at the end of this round of modulation; Step 76, solidify the model parameters of the current training model.

8. The method for predicting the stress-strain curve of a concrete member according to claim 4, characterized in that: The pair completes the training of all the second training models MB j Evaluate model performance and determine the corresponding optimal model parameters based on the evaluation results, including: Step 81, randomly extracting a plurality of first data records from the first data set based on a preset second extraction ratio to form a corresponding first evaluation set; wherein the second extraction ratio is a percentage less than 1; the total number of records in the first data set matches the product of the first extraction ratio and the total number of records in the first evaluation set; Step 82: each of the second training models MB j as the corresponding current evaluation model; and the second value L corresponding to the current evaluation model j as the corresponding current sequence length; and traverse all the first data records of the first evaluation set; and during the traversal, use the currently traversed first data record as the corresponding current evaluation record; and based on the current sequence length, divide the first training strain data sequence of the current evaluation record into multiple second training segmentation sequences; and sequentially sort all the obtained second training segmentation sequences to form a corresponding second training segmentation sequence string; and input the second training segmentation sequence string into the current evaluation model for processing to obtain a normalized stress data sequence recorded as the corresponding second predicted stress data sequence; and form a corresponding first prediction / label sequence pair from the second predicted stress data sequence and the first label stress data sequence of the current evaluation record; and at the end of this traversal, bring all the obtained first prediction / label sequence pairs into a preset second error function for error calculation to obtain the corresponding second prediction error; Wherein, the second error function at least includes a MAE error function, an MSE error function, and a RMSE error function; Step 83: The second training model MB having the smallest second prediction error is used as j as the corresponding optimal performance model; and the model parameters of the optimal performance model are used as the corresponding optimal model parameters.

9. The method for predicting the stress-strain curve of a concrete member according to claim 3, characterized in that: The using the curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve specifically includes: Performing normalized parameter tensor preparation according to the first design parameter list to obtain a corresponding first design parameter vector; The first design parameter vector is input into the BP network of the curve prediction model for processing to obtain a pair of normalized peak stress / strain data recorded as corresponding normalized stress prediction value and normalized strain prediction value; and the normalized stress prediction value and normalized strain prediction value are input into the first denormalization module of the curve prediction model for processing to obtain a pair of denormalized peak stress / strain data recorded as corresponding first stress prediction value and first strain prediction value; and taking 0 as the minimum strain data threshold and the first strain prediction value as the maximum strain data threshold, normalizing the first strain data sequence according to the maximum and minimum strain data thresholds to obtain a normalized data sequence recorded as the corresponding second strain data sequence; and taking the segmentation sequence length corresponding to the LSTM layer of the curve prediction model as the corresponding first sequence length; and dividing the second strain data sequence into a plurality of first segmentation sequences according to the first sequence length; and sorting all the obtained first segmentation sequences in order to form a corresponding first segmentation sequence string; and inputting the first segmentation sequence string into the LSTM stacked network of the curve prediction model for processing to obtain corresponding multiple first stress feature vectors; and inputting all the first stress feature vectors obtained into the fully connected layer of the curve prediction model for processing to obtain a normalized stress data sequence recorded as the corresponding first stress data sequence; The first segmentation sequence string, the first stress data sequence, the first stress prediction value and the first strain prediction value are input into the second denormalization module of the curve prediction model for processing to obtain the corresponding first stress-strain curve.

10. The method for predicting the stress-strain curve of a concrete member according to claim 9, characterized in that: The preparing of normalizing the parameter tensor according to the first design parameter list to obtain the corresponding first design parameter vector specifically includes: Extracting nineteen types of designated design parameters from the first design parameter list; the nineteen types of designated design parameters include stirrup minor axis dimension parameters, stirrup major axis dimension parameters, stirrup cross-sectional area parameters, stirrup cross-sectional inertia moment parameters, stirrup shape coding parameters, stirrup spacing parameters, stirrup reinforcement ratio parameters, stirrup strength parameters, stirrup limb number parameters, concrete strength parameters, cross-sectional dimension parameters, concrete cover thickness parameters, component longitudinal dimension parameters, component aspect ratio parameters, component cross-sectional shape parameters, longitudinal reinforcement strength parameters, longitudinal reinforcement ratio parameters, stirrup characteristic value parameters, and reinforcement characteristic value parameters; Based on the maximum and minimum parameter thresholds corresponding to each type of the nineteen types of specified design parameters, normalization calculation is performed on each type of the specified design parameters, and the nineteen normalized calculation results are used to form the corresponding nineteen types of normalized design parameters; and the nineteen types of normalized design parameters are used to form the corresponding first design parameter vector.

11. A device for executing the method for predicting stress-strain curve of a concrete member according to any one of claims 1 to 10, characterized in that: The device comprises: a model building module, a model training module and a model application module; The model building module is used to build a curve prediction model; the model input end of the curve prediction model includes a first model input end and a second model input end, and the curve prediction model has only one model output end; the first model input end is used to receive a design parameter vector with 19 types of normalized design parameters of the concrete component; the second model input end is used to receive a segmented sequence string of the normalized strain data sequence of the concrete component, and the segmented sequence string is formed by sequentially sorting a plurality of segmented sequences; the model output end is used to output the stress-strain curve of the concrete component; The model training module is used to perform model training on the curve prediction model; The model application module is used to receive a first design parameter list and a first strain data sequence of a first concrete component after model training is completed; and use the curve prediction model to perform curve prediction according to the first design parameter list and the first strain data sequence to obtain a corresponding first stress-strain curve.

12. An electronic device, characterized in that: include: memory, processors, and transceivers; The processor is used to couple with the memory, read and execute instructions in the memory, so as to implement the method according to any one of claims 1 to 10; The transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer is enabled to execute the method according to any one of claims 1 to 10.