A method and system for generating a concrete parameter optimization model

Abstract

The present application relates to the technical field of computer model generation, and is a concrete parameter optimization model generation method and system, specifically comprising: establishing a digital twin evolution model fusing prestressed vector field and heat conduction gradient; extracting effective axial force prediction value and interface debonding displacement increment changing with fire duration based on the digital twin evolution model, calculating and obtaining dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient; generating structure topological robustness maintenance rate; online evaluating deviation tolerance of maintenance rate generated by the current model and target maintenance rate, and adjusting concrete aggregate gradation parameter combination. The present application solves the problem in the prior art that the dynamic collapse risk of the effective contact network of aggregate cannot be characterized, it is difficult to give a gradation optimization boundary with thermal and mechanical robustness, and a technical blind area exists in the safe operation and maintenance of the structure.

Description

Technical Field

[0001] This invention relates to the field of computer model generation technology, and is a method and system for generating concrete parameter optimization models. Background Technology

[0002] Yellow River sand, as an ultrafine sand (fineness modulus 0.5~1.5), is abundant, but its poor particle size distribution, high mud content, and complex interfacial transition zone directly affect its application in high-performance concrete. Meanwhile, precast prestressed concrete continuous box girders are widely used in highway bridges due to their advantages such as high construction efficiency and high torsional stiffness. In the micromechanical behavior analysis of such structures, the aggregate gradation parameters of concrete not only determine the dense packing strength of the material but are also the core boundary elements defining the damage evolution process of large-span components under complex stress fields.

[0003] However, existing concrete aggregate gradation optimization techniques typically follow static packing theory, focusing primarily on static load-bearing stability at room temperature or uniform heating under standard temperature rise curves. They rarely consider the interfacial weakening and thermal expansion incompatibility issues of ultra-fine sands like Yellow River sand at high temperatures. In real-world fire scenarios, the localized extreme thermal loads generated by vehicle fires on high-speed bridges can lead to extremely large longitudinal temperature gradients between the bottom slab and web of box girders. This non-uniform thermal expansion effect induces severe constraint incompatibility in the wet joints and prestressed anchorage concentration areas characteristic of prefabricated components. Existing technologies neglect the initiation of microcracks in such web-bottom slab coupling zones under strong constraint conditions, and fail to reveal the debonding-skeleton reconstruction chain reaction exacerbated by differences in aggregate-paste interfacial bonding properties in Yellow River sand concrete. This causes the apparent geometric gradation preset in traditional models to degenerate into an inefficient load-bearing gradation during macroscopic damage processes.

[0004] In summary, existing methods cannot characterize the dynamic collapse risk of the effective contact network of aggregates. In particular, for concrete made with ultra-fine sand such as Yellow River sand, the interfacial debonding behavior under fire is significantly different from that of ordinary aggregates. This makes it difficult to provide thermally robust gradation optimization boundaries during the design stage when facing atypical structural fractures and ductility deterioration caused by localized abnormal fires, thus creating a technical blind spot for structural safety operation and maintenance. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0006] The technical problem to be solved by the present invention is that the existing technology cannot characterize the dynamic collapse risk of the effective contact network of aggregates, and it is difficult to give thermally robust gradation optimization boundaries in the design stage, resulting in technical blind spots in the safe operation and maintenance of structures. The present invention proposes a method and system for generating concrete parameter optimization models.

[0007] To achieve the above objectives, the technical solution of the concrete parameter optimization model generation method of the present invention includes the following steps:

[0008] Step 1: Extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient;

[0009] Step 2: Obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model;

[0010] Step 3: Based on the digital twin evolution model, extract the effective axial force prediction value and interface debonding displacement increment that change with the duration of fire exposure, and calculate and obtain the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient.

[0011] Step 4: Utilize the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combine the dynamic debonding sensitivity coefficient, synergy factor, and topological redundancy response coefficient to generate the structural topological robustness maintenance rate.

[0012] Step 5: Evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and adjust the concrete aggregate gradation parameter combination.

[0013] Specifically, step one includes:

[0014] A1: Based on the physical properties of the building information model of the beam, extract the surface roughness index, dynamic centroid deviation factor and body surface area change rate of the aggregate in each particle size range in the coupling zone to form the set of structural geometric feature vectors of the prefabricated prestressed continuous box girder.

[0015] A2: Preset the fire source trajectory field function under different fire intensities, map the geometric feature vector of the structure to the discretized finite element topology mesh, and generate a three-dimensional heterogeneous thermal resistance coefficient mapping matrix.

[0016] A3: Construct a digital twin evolution model that includes a prestressed vector field and a thermal conduction gradient, including: introducing the partial tensors of the longitudinal and transverse stress constraint components in the coupling region, obtaining the spatiotemporal evolution law of the temperature field and stress field in the coupling region by iteratively solving the thermo-mechanical coupling differential equations in the computational domain, and constructing a digital twin evolution model.

[0017] Specifically, step two includes:

[0018] B1: Obtain the relative position parameters of the ignition point from the bottom plate under the preset fire scenario, and simulate the bottom plate response curve and web hysteresis displacement curve under different heating phases through the digital twin evolution model to construct the fire response deviation load sequence of the prefabricated prestressed continuous box girder.

[0019] B2: Retrieve the performance degradation data of the same type of prefabricated prestressed continuous box girder under historical fire conditions, and extract the volume ratio of aggregates in each diameter segment, the initial feature vector of the interface transition zone, the aggregate-slurry coating thickness index, the effective contact density decay trajectory of aggregates under the corresponding working conditions, and the shear stiffness retention rate of the beam under the multi-dimensional historical gradation. Construct a training set for the evolution of gradation historical features, and divide the data into the basic training set and the validation set of the gradation generation model in an 8:2 ratio.

[0020] B3: The surface spheroidization factor, contact coordination number distribution gradient and thermal expansion and shrinkage rate difference of aggregates of different particle sizes within a single gradation range are obtained and dimensionlessly processed to form the basic feature mapping space of aggregates of each particle size within the full gradation envelope.

[0021] B4: Construct a gradation parameter generation neural network model, which includes: a topological signal input layer, a non-homogeneous constrained convolutional layer, and a steady-state redundancy mapping layer;

[0022] The non-homogeneous constrained convolutional layer contains a pre-set aggregate displacement interference sensitive weight kernel for the 1.18mm-2.36mm bridging diameter segment;

[0023] B5: Perform dimensionality reduction and standardization on the basic feature mapping space and gradation history feature evolution training set of each aggregate diameter segment, and activate the non-homogeneous constrained hidden layer for tensor correction calculation. The specific formula is as follows:

[0024] ;

[0025] in, For the first The situational output of the j-th mapping unit in the hidden layer. The thermal-mechanical characteristic excitation is for the i-th unit in the previous layer. This refers to the phase offset term of the element under specific prestress constraints. For the first Layer to the first The contribution weight of aggregate dissociation evolution during layer transfer;

[0026] B6: Use the validation set to evaluate the fit of the neural network model for generating the gradation parameters until the sum of the prediction residuals reaches the convergence threshold, thus obtaining the initial gradation generation model that has completed the initial weight training.

[0027] Specifically, step three includes:

[0028] C1: Within the digital twin evolution space defined by the digital twin evolution model constructed in step A3, simultaneously generate attenuation curves for the predicted values ​​of each effective axial force component under the preset fire scenario as a function of fire duration, as well as the temperature gradient at different cross-sections. The trajectory curve that evolves with the duration of fire exposure;

[0029] C2: The system automatically extracts multiple fire evolution processes with different phases where the difference between two or more sets of effective axial force prediction values ​​does not exceed 5%, as a control group of the same magnitude load, and simultaneously extracts the effective contact density of aggregates in the control group. Curves showing the change over time and the phase difference curve of temperature rise ;

[0030] C3: Extract the characteristic fluctuations of the heating gradient feature curve in the process of the evolution of the opposite phase fire, obtain the fluctuation component of the heating rate by removing the trend term, and construct the heating rate interference evaluation envelope set to characterize the temperature field interference.

[0031] Take the temperature rise rate fluctuation within the evaluation envelope set. and the corresponding aggregate-slurry interface debonding displacement increment ;

[0032] Define the set of temperature rise rate fluctuations as: Where the superscript g represents the heat conduction flux correction term at the current moment, and the subscript v represents the v-th discrete calculation period in the out-of-phase fire evolution process; the corresponding set of effective interface debonding displacements is expressed as: ;

[0033] C4: Import the set of heating rate fluctuations and the set of effective interfacial debonding displacements into the interfacial debonding regression function to calculate the dynamic debonding sensitivity coefficient that reflects the robustness of the gradation scheme. The specific calculation formula is as follows:

[0034] ;

[0035] in, This is the arithmetic mean of the temperature rise rate fluctuations during the current assessment period. This represents the arithmetic mean of the interface debonding displacement increments at the corresponding stages. This is used to quantify the negative contribution of thermal conductivity rate fluctuations to aggregate debonding.

[0036] C5: Extract the prestress loss curve obtained during the digital twinning process of the stress path and the predicted displacement phase difference curve at different heating depths d for precast prestressed continuous box girders. ;

[0037] C6: Select an evolution process with a consistent heating gradient to form an anisotropic constraint disturbance set, and extract the prestrain phase adjustment amount caused by prestress fluctuations from this set. and the corresponding increase in the failure frequency of the effective contact network between aggregates. ;

[0038] The set of pre-strain phase adjustment values ​​is set as follows: ;

[0039] Wherein, the superscript pt represents the fitting state between the prestrain and the tension vector, and the subscript u represents the u-th interference calculation sequence;

[0040] The corresponding set of aggregate contact failure increments is represented as follows:

[0041] ;

[0042] C7: The set of pre-strain phase adjustment values ​​and the corresponding set of aggregate contact failure frequencies are imported into the elastic modulus softening regression model to calculate the synergy factor dominated by prestress constraint fluctuations. The specific calculation formula is as follows:

[0043] ;

[0044] in, These are the expected mean values ​​of the corresponding feature sets during the constraint interference period. Used to quantify the probability of instability of aggregate gradation support frames due to non-uniform loss of prestress.

[0045] C8: Extract the structural geometric feature vectors of aggregates from each gradation scheme within the set of structural geometric feature vectors constructed in step A1, and calculate the comprehensive derived feature factor of the load-bearing skeleton of the current gradation sequence. The calculation formula is:

[0046] ;

[0047] in, These are the surface roughness index, dynamic centroid deviation factor, and surface area change rate of the aggregate in this diameter segment, obtained from analysis within the digital twin.

[0048] These are the contribution weights of the feature-derived factors to the thermo-mechanical evolution under the corresponding fire protection phase;

[0049] C9: Extract multiple evolution path sub-items that maintain consistency between the predefined simulated heating gradient and the simulated prestress level, and form a morphological interference set for the evolution of gradation characteristics;

[0050] From the morphological interference set, multiple comprehensive derived feature factors arranged in heterogeneous order and their corresponding beams are simultaneously extracted as the evolution increment of the structural redundancy prediction value at the extreme failure threshold point.

[0051] The set of comprehensive derived feature factors is as follows:

[0052] ;

[0053] The superscript h represents the feature identifier corresponding to the current extraction task, and the subscript e represents the e-th specific mixed gradation topology sequence. The comprehensive derived characteristic factor corresponding to the e-th gradation;

[0054] Among them, the set of continuously evolving effective quantities of the structural redundancy prediction values ​​that match it is:

[0055] ,in The shear load value retained at the end of the fire simulation process corresponding to the e-th specific gradation obtained from the digital space;

[0056] C10: A set of comprehensive derived feature factors and the evolution set of structural redundancy prediction values Import the response co-analysis function for aggregate geometric contribution, and calculate the topological redundancy response coefficient. The specific calculation formula is as follows:

[0057] ;

[0058] in, The average value of the comprehensive derived feature imprint values ​​within the evaluation period of the interference set. This represents the average value of the predicted residual load components within the corresponding structure.

[0059] Specifically, step four includes:

[0060] S101: By... , as well as Weighted summation yields the multi-field fusion adaptive coordination number of three-dimensional asynchronous evolution. ;

[0061] S102: Retrieve the initial gradation generation model trained in step B6, and fuse the adaptive matching numbers from multiple fields. This is mapped to the correction operator of the non-homogeneous constrained convolutional layer in the neural network model;

[0062] This correction operator is used to interfere with the internal structure of the neural network in real time. The situational output of the hidden layer The online correction of the neural network weights is completed to obtain the optimized gradation generation model;

[0063] S103: Input the real-time monitored beam geometry and load sequence into the optimized gradation generation model to predict the effective transfer level of the prestressed load-bearing beam under the current specific gradation parameter combination. ;

[0064] S104: Obtain the structural topological robustness maintenance rate that characterizes the stability of the model scheme. The specific generation strategy is as follows: ;

[0065] in, The effective transmission level of the force path currently generated within the digital twin space under a specific fire phase; The target value for the preset stability redundancy of the coupling zone of the load-bearing beam under the fire phase.

[0066] Specifically, step five includes:

[0067] S201: Online assessment of effective transmission level differences The minimum effective threshold for the stability of the preset structure, when the hierarchical difference When the sampling frequency exceeds the minimum effective stability threshold, the finite element discrete mesh sampling frequency of the digital twin evolution model is increased by 5%, wherein... ;

[0068] S202: Simultaneously calculate the structural robustness maintenance rate under the current aggregate gradation configuration. Compared with the preset target maintenance rate The difference is used to preset the acceptable drift tolerance of the optimized gradation generation model;

[0069] When the difference exceeds the acceptable drift tolerance of the gradation optimization model, it is determined that the gradation scheme has insufficient interlocking force under fire conditions, an alarm signal is issued, and the proportion of medium and large-sized aggregates of 5.0mm-20mm is forcibly increased in the next round of gradation generation.

[0070] In addition, the concrete parameter optimization model generation system of the present invention includes the following modules:

[0071] The module includes a digital twin modeling module, a gradation feature prediction module, a performance parameter calculation module, a robust performance evaluation module, and a gradation early warning and adjustment module.

[0072] The digital twin modeling module is used to extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and to establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient.

[0073] The gradation feature prediction module is used to obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model.

[0074] The performance parameter calculation module extracts the effective axial force prediction value and interface debonding displacement increment that change with the duration of fire based on the digital twin evolution model, and calculates and obtains the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient.

[0075] The robust performance evaluation module uses the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combines the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient to generate the structural topological robustness maintenance rate.

[0076] The gradation early warning and adjustment module is used to evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and to adjust the combination of concrete aggregate gradation parameters.

[0077] Compared with the prior art, the technical effects of the present invention are as follows:

[0078] 1. This invention is a dynamic local adaptive reconstruction, which effectively addresses the problem of microcrack propagation and aggregate debonding caused by the strong temperature gradient and structural constraint mismatch resulting from the initial high temperature of the bottom plate and subsequent temperature rise of the web plate. It avoids the premature collapse of the skeleton bearing network caused by the failure of traditional models under asynchronous thermal evolution conditions.

[0079] 2. Based on the real structural response, this invention dynamically identifies the aggregate contact degradation trend in key weak areas, which significantly enhances the model's accuracy in capturing and its early warning sensitivity for failure paths caused by constraint-heating-shrinkage incoordination.

[0080] 3. By constructing an online evaluation and incremental proportional correction mechanism, this invention significantly shortens the multi-round trial and error iteration cycle while ensuring computational efficiency. This enables the supplementation strategy of medium and large-sized aggregates (5.0–20mm) to quickly match the actual degree of thermal damage. Thus, without increasing the amount of cement or changing the arrangement of the main reinforcement, it substantially improves the post-fire residual bearing capacity and ductility reserve capacity of key nodes of this type of typical bridge. Attached Figure Description

[0081] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0082] Figure 1 This is a flowchart illustrating a method for generating a concrete parameter optimization model according to the present invention.

[0083] Figure 2 This is a schematic diagram of the structure of a concrete parameter optimization model generation system according to the present invention. Detailed Implementation

[0084] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0085] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0086] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0087] Example 1:

[0088] like Figure 1 As shown in the figure, an embodiment of the present invention provides a method for generating a concrete parameter optimization model, such as... Figure 1 As shown, the specific steps include the following:

[0089] Step 1: Extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient;

[0090] Step one includes:

[0091] A1: Based on the physical properties of the building information model of the beam, extract the surface roughness index, dynamic centroid deviation factor and body surface area change rate of the aggregate in each particle size range in the coupling zone to form the set of structural geometric feature vectors of the prefabricated prestressed continuous box girder.

[0092] A2: Preset the fire source trajectory field function under different fire intensities, map the geometric feature vector of the structure to the discretized finite element topology mesh, and generate a three-dimensional heterogeneous thermal resistance coefficient mapping matrix.

[0093] It should be noted that the three-dimensional heterogeneous thermal resistance coefficient mapping matrix is ​​used to characterize the dynamic evolution trend of thermal resistance of each node inside the beam during the fire process.

[0094] A3: Construct a digital twin evolution model that includes a prestressed vector field and a thermal conduction gradient, including: introducing the partial tensors of the longitudinal and transverse stress constraint components in the coupling region, obtaining the spatiotemporal evolution law of the temperature field and stress field in the coupling region by iteratively solving the thermo-mechanical coupling differential equations in the computational domain, and constructing a digital twin evolution model.

[0095] The thermo-mechanical coupled differential equations include at least the heat conduction equation and the stress balance equation, and take into account the correction effect of the prestress vector field on the heat conduction gradient.

[0096] For example, in this embodiment, step A3 includes:

[0097] First, establish the transient heat conduction differential equation within the coupling zone to determine the temperature field distribution inside the beam under fire conditions. The equation is based on Fourier's law and the principle of conservation of energy, and its specific form is as follows:

[0098] ;

[0099] in, Where C is the density of concrete, c is the specific heat capacity, T is the temperature, t is the firing time, and Q is the internal heat source term, defined when considering hydration reaction or latent heat of phase change, and set to 0 in this embodiment. The thermal conductivity coefficients in the x, y, and z directions are not constant values. Instead, they are interpolated in the computational domain by the three-dimensional heterogeneous thermal resistance coefficient mapping matrix generated in step A2, and are used to characterize the thermal resistance characteristics at the aggregate-slurry interface.

[0100] The initial conditions are: , The initial ambient temperature;

[0101] The boundary conditions are as follows: On the fire-receiving surface, convective and radiative heat transfer boundaries are used, as follows:

[0102] ;

[0103] Where h is the convective heat transfer coefficient. For structural surface temperature and fire smoke temperature, For emissivity and Stefan-Boltzmann constant.

[0104] Then, based on the obtained temperature field T, a thermoelastic (or thermoelastic-plastic) equilibrium differential equation considering the prestress effect is established to calculate the displacement field u and stress field F of the structure under the combined action of temperature and load, as follows: ;

[0105] Meanwhile, considering the effects of thermal strain, elastic strain, and prestress, the constitutive equation is established as follows: ;

[0106] in, Let f be the Cauchy stress tensor, and f be the self-weight. For stiffness tensor, For total strain, These are thermal strain and prestressed equivalent initial strain, respectively.

[0107] Next, the deviatoric tensors of the longitudinal and transverse stress constraint components within the coupling region are introduced. That is, in this embodiment, the total stress tensor is decomposed into the spherical tensor (volume stress) and the deviatoric tensor. (i.e., stress eccentricity): ;in, For average stress, The symbol is Kronecker, which is 1 when i=j and 0 otherwise;

[0108] It should be noted that the change in the deviator tensor directly reflects the failure trend of beam torsion and shear under fire. By sequentially coupling and solving the above heat conduction equation and stress balance equation (i.e., first calculate the temperature field T, and then substitute T as a load into the stress field for solution), and iterating in each time step Δt, the spatiotemporal evolution law of temperature field and stress field in the coupling zone with the duration of fire can be obtained, and finally the digital twin evolution model is constructed.

[0109] It should be noted that the above heat conduction equation and stress balance equation belong to a set of partial differential equations, which are difficult to obtain analytical solutions under complex geometric domains and heterogeneous material conditions. Therefore, in this embodiment, the finite element method is preferably used to discretize and solve the above set of equations.

[0110] Step 2: Obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model;

[0111] Step two includes:

[0112] B1: Obtain the relative position parameters of the ignition point from the bottom slab under a preset fire scenario, and simulate the bottom slab response curve and web hysteresis displacement curve at different heating phases using the digital twin evolution model to construct the fire response deviation load sequence of the prefabricated prestressed continuous box girder.

[0113] For example, in this embodiment, B1 specifically includes: a preset fire scenario including the location of the ignition point, the power of the fire source, and a standard fire temperature rise curve; extracting the vertical distance from the ignition point to the bottom plate as a relative position parameter and inputting it into the digital twin evolution model constructed in step A3; obtaining the vertical displacement time history curves of the bottom plate under different temperature rise phases and the hysteresis displacement curves of the web plate under the same time series through model calculation; and combining the difference between the bottom plate response curve and the web plate hysteresis displacement curve according to the time series to form a fire response deviation load sequence. It should be noted that this sequence is used to characterize the non-uniform deformation characteristics of the beam coupling zone under fire.

[0114] B2: Retrieve the performance degradation data of the same type of prefabricated prestressed continuous box girder under historical fire conditions, and extract the volume ratio of aggregates in each diameter segment, the initial feature vector of the interface transition zone, the aggregate-slurry coating thickness index, the effective contact density decay trajectory of aggregates under the corresponding working conditions, and the shear stiffness retention rate of the beam under the multi-dimensional historical gradation. Construct a training set for the evolution of gradation historical features, and divide the data into the basic training set and the validation set of the gradation generation model in an 8:2 ratio.

[0115] It should be noted that the "same type of precast prestressed continuous box girder" refers to a beam with the same cross-sectional form, material strength grade, and prestressing process as the current analysis object. The historical fire conditions include different fire durations, different temperature rise curves, and different fire protection conditions. The multidimensional historical gradation refers to different aggregate particle size combinations and is divided into various diameter segments according to a preset sieving interval. The extracted features include the percentage of aggregate in each particle size interval to the total volume, the initial feature vector of the interface transition zone composed of aggregate surface roughness and slurry water-cement ratio, the ratio of the average thickness of the slurry coating on the aggregate surface to the aggregate particle size, the trajectory curve of the decrease in the number of effective contact points between aggregates with fire time, and the rate of change of the ratio of the beam shear stiffness after fire to the initial shear stiffness with fire time. After normalization, all extracted data are randomly divided into a basic training set and a validation set in an 8:2 ratio for subsequent training and validation of the gradation parameter generation neural network model.

[0116] B3: Obtain the surface spheroidization factor, contact coordination number distribution gradient and thermal expansion and shrinkage rate difference of aggregates of different particle size ranges within a single diameter range (e.g., 0.15mm-31.5mm), perform dimensionless processing, and construct the basic feature mapping space of aggregates of each diameter range within the full gradation envelope.

[0117] It should be noted that the surface sphericity factor is used to characterize the sphericity of aggregate particles, the contact coordination number distribution gradient reflects the trend of the number of contact points between aggregates changing with particle size, and the difference in thermal expansion and contraction rate is the difference in the linear expansion coefficient of different aggregates at high temperature.

[0118] B4: Construct a gradation parameter generation neural network model, which includes: a topological signal input layer, a non-homogeneous constrained convolutional layer, and a steady-state redundancy mapping layer;

[0119] The non-homogeneous constrained convolutional layer contains a pre-set aggregate displacement interference sensitive weight kernel for the 1.18mm-2.36mm bridging diameter segment;

[0120] It should be noted that the topological signal input layer is used to receive the geometric features of the beam and the historical attenuation load sequence; the non-homogeneous constrained convolutional layer is used to simulate the non-uniform temperature field and pre-strain phase interference; and the steady-state redundant mapping layer is used to converge the output of the target gradation ratio value.

[0121] For example, in this embodiment, the gradation parameter generation neural network model adopts a three-layer structure: the topological signal input layer receives the structural geometric feature vector (dimension 12) and the fire response deviation load sequence (time step 10), and concatenates them to form a 22-dimensional input tensor; the non-homogeneous constrained convolutional layer adopts one-dimensional convolution with a kernel size of 3, a stride of 1, an output channel number of 16, and an activation function of ReLU. Among them, a sensitive weight kernel for aggregate displacement interference is preset for the 1.18mm-2.36mm bridging diameter segment. The initial value of the weight kernel is set to 1.8 times the weight of the corresponding convolution kernel for this diameter segment (specifically 0.72, and 0.4 for other diameter segments), and an L2 regularization term of the weight kernel is added to the loss function to maintain sensitivity; the steady-state redundant mapping layer consists of two fully connected layers (with 32 and 8 neurons respectively), and the output layer adopts the Softmax activation function to output the volume ratio of aggregate in each particle size segment. The model training uses the Adam optimizer with an initial learning rate of 0.001 and the mean square error between the predicted and actual gradations as the loss function, thereby constructing the neural network structure for generating the initial gradation parameters to be trained.

[0122] B5: Perform dimensionality reduction and standardization on the basic feature mapping space and gradation history feature evolution training set of each aggregate diameter segment, and activate the non-homogeneous constrained hidden layer for tensor correction calculation. The specific formula is as follows:

[0123] ;

[0124] in, For the first The situational output of the j-th mapping unit in the hidden layer. The thermal-mechanical characteristic excitation is for the i-th unit in the previous layer. This refers to the phase offset term of the element under specific prestress constraints. For the first Layer to the first The contribution weight of aggregate dissociation evolution during layer transfer.

[0125] B6: Use the validation set to evaluate the fit of the neural network model for generating the gradation parameters until the sum of the prediction residuals reaches the convergence threshold, thus obtaining the initial gradation generation model that has completed the initial weight training.

[0126] It should be noted that the initial gradation generation model is used to predict the path levels that can still effectively transfer prestress within the beam under the current fire conditions.

[0127] Step 3: Based on the digital twin evolution model, extract the effective axial force prediction value and interface debonding displacement increment that vary with the duration of fire exposure, and calculate and obtain the dynamic debonding sensitivity coefficient, synergy factor, and topological redundancy response coefficient corresponding to the micro-reinforcement effect of aggregate to quantify the degree of multi-field coupling interference.

[0128] Step three includes:

[0129] C1: Within the digital twin evolution space defined by the digital twin evolution model constructed in step A3, simultaneously generate attenuation curves of the predicted values ​​of each effective axial force component under this loading condition as a function of fire duration, as well as the temperature gradients of cross-sections at different locations. The trajectory curve that evolves with the duration of fire exposure;

[0130] It should be noted that the effective axial force component refers to the remaining axial force of the prestressed tendons during the fire process, and the temperature gradient is the rate of temperature change along the longitudinal direction of the beam.

[0131] C2: The system automatically extracts multiple fire evolution processes with different phases where the difference between two or more sets of effective axial force prediction values ​​does not exceed 5%, as a control group of the same magnitude load, and simultaneously extracts the effective contact density of aggregates in the control group. Curves showing the change over time and the phase difference curve of temperature rise ;

[0132] It should be noted that the effective contact density of aggregates refers to the number of effective contact points between aggregates per unit volume, and the heating phase difference refers to the phase difference of the heating curves at different locations at the same time.

[0133] C3: Extract the characteristic fluctuations of the heating gradient feature curve in the process of the evolution of the opposite phase fire, obtain the fluctuation component of the heating rate by removing the trend term, and construct the heating rate interference evaluation envelope set to characterize the temperature field interference.

[0134] Take the temperature rise rate fluctuation within the evaluation envelope set. and the corresponding aggregate-slurry interface debonding displacement increment ;

[0135] The set of temperature rise rate fluctuations is represented as: Where the superscript g represents the heat conduction flux correction term at the current moment, and the subscript v represents the v-th discrete calculation period in the out-of-phase fire evolution process; the corresponding set of effective interface debonding displacements is expressed as: ;

[0136] C4: Import the set of heating rate fluctuations and the set of effective interfacial debonding displacements into the interfacial debonding regression function to calculate the dynamic debonding sensitivity coefficient that reflects the robustness of the gradation scheme. The specific calculation formula is as follows:

[0137] ;

[0138] in, This is the arithmetic mean of the temperature rise rate fluctuations during the current assessment period. This represents the arithmetic mean of the interface debonding displacement increments at the corresponding stages. This is used to quantify the negative contribution of thermal conductivity rate fluctuations to aggregate debonding; the larger the value, the more sensitive the debonding is to temperature fluctuations.

[0139] C5: Extract the prestress loss curve obtained during the digital twinning process of the stress path and the predicted displacement phase difference curve at different heating depths d for precast prestressed continuous box girders. ;

[0140] It should be noted that the prestress loss curve refers to the attenuation of the effective prestress of the prestressing tendon with the time of exposure to fire, and the displacement phase difference refers to the phase difference of the displacement response of sections with different heating depths at the same moment.

[0141] C6: Select an evolution process with a consistent heating gradient to form an anisotropic constraint disturbance set, and extract the prestrain phase adjustment amount caused by prestress fluctuations from this set. and the corresponding increase in the failure frequency of the effective contact network between aggregates. ;

[0142] The set of pre-strain phase adjustment values ​​is set as follows: ;

[0143] Wherein, the superscript pt represents the fitting state between the prestrain and the tension vector, and the subscript u represents the u-th interference calculation sequence;

[0144] The corresponding set of aggregate contact failure increments is represented as follows:

[0145] ;

[0146] C7: The set of pre-strain phase adjustment values ​​and the corresponding set of aggregate contact failure frequencies are imported into the elastic modulus softening regression model to calculate the synergy factor dominated by prestress constraint fluctuations. The specific calculation formula is as follows:

[0147] ;

[0148] in, These are the expected mean values ​​of the corresponding feature sets during the constraint interference period. This is used to quantify the probability of instability of the aggregate gradation support frame due to non-uniform prestress loss. The larger the value, the more significant the impact of prestress fluctuation on the stability of the aggregate contact network.

[0149] C8: Extract the structural geometric feature vectors of aggregates from each gradation scheme within the set of structural geometric feature vectors constructed in step A1, and calculate the comprehensive derived feature factor of the load-bearing skeleton of the current gradation sequence. The calculation formula is:

[0150] ;

[0151] in, These are the surface roughness index, dynamic centroid deviation factor, and surface area change rate of the aggregate in this diameter segment, obtained from analysis within the digital twin.

[0152] These are the contribution weights of the feature-derived factors to the thermo-mechanical evolution under the corresponding fire protection phase;

[0153] For example, in this embodiment, the process of obtaining the thermal-mechanical evolution contribution weight in C8 is as follows:

[0154] First, the digital twin evolution model is used to perform virtual evolution simulation under multiple preset fire protection phases. The influence weights of aggregate surface roughness index, dynamic centroid deviation factor and body surface area change rate on the final residual bearing capacity of the structure under the action of heating and stress coupling are calculated by the control variable method.

[0155] Subsequently, the correlation scores between each geometric feature factor and the degree of structural degradation are extracted based on sensitivity analysis algorithms (such as Sobol sensitivity analysis or Pearson correlation coefficient analysis);

[0156] Finally, the correlation scores of each phase are normalized (mapped to sum to 1) to obtain the corresponding evolution contribution weight value for each phase.

[0157] In a specific application scenario of this embodiment, according to experimental calibration, corresponding to the rapid temperature rise fire prevention stage, The values ​​can be preset to 0.35, 0.45 and 0.20; while in the holding-temperature cooling stage, they are adjusted to 0.40, 0.30 and 0.30 respectively, so as to match the contribution of different thermal evolution states to the aggregate stress frame in real time.

[0158] C9: Extract multiple evolution path sub-items that maintain consistency between the predefined simulated heating gradient and the simulated prestress level, and form a morphological interference set for the evolution of gradation characteristics;

[0159] From the morphological interference set, multiple comprehensive derived feature factors arranged in heterogeneous order and their corresponding beams are simultaneously extracted as the evolution increment of the structural redundancy prediction value at the extreme failure threshold point.

[0160] The set of comprehensive derived feature factors is as follows:

[0161] ;

[0162] The superscript h represents the feature identifier corresponding to the current extraction task, and the subscript e represents the e-th specific mixed gradation topology sequence. The comprehensive derived characteristic factor corresponding to the e-th gradation;

[0163] Among them, the set of continuously evolving effective quantities of the structural redundancy prediction values ​​that match it is:

[0164] ,in The shear load value retained at the end of the fire simulation process corresponding to the e-th specific gradation obtained from the digital space;

[0165] C10: A set of comprehensive derived feature factors and the evolution set of structural redundancy prediction values Import the response co-analysis function for aggregate geometric contribution, and calculate the topological redundancy response coefficient. The specific calculation formula is as follows:

[0166] ;

[0167] in, The average value of the comprehensive derived feature imprint values ​​within the evaluation period of the interference set. This represents the average value of the predicted residual load components within the corresponding structure.

[0168] Step 4: Utilize the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combine the dynamic debonding sensitivity coefficient, synergy factor, and topological redundancy response coefficient to generate the structural topological robustness maintenance rate.

[0169] Step four includes:

[0170] S101: By... , as well as Weighted summation yields the multi-field fusion adaptive coordination number of three-dimensional asynchronous evolution. ;

[0171] S102: Retrieve the initial gradation generation model trained in step B6, and fuse the adaptive matching numbers from multiple fields. This is mapped to the correction operator of the non-homogeneous constrained convolutional layer in the neural network model;

[0172] This correction operator is used to interfere with the internal structure of the neural network in real time. The situational output of the hidden layer The online correction of the neural network weights is completed to obtain the optimized gradation generation model;

[0173] For example, in this embodiment, the online correction of the neural network weights specifically involves:

[0174] ;

[0175] in, This is the status output after online correction.

[0176] S103: Input the real-time monitored beam geometry and load sequence into the optimized gradation generation model to predict the effective transfer level of the prestressed load-bearing beam under the current specific gradation parameter combination. ;

[0177] It should be noted that the predicted effective transfer level of the prestressed load-bearing beam is... Used to characterize the residual topological efficiency of the skeleton's stress path after being exposed to fire;

[0178] S104: Obtain the structural topological robustness maintenance rate that characterizes the stability of the model scheme. The specific generation strategy is as follows: ;

[0179] in, The effective transmission level of the force path currently generated within the digital twin space under a specific fire phase; The target value for the preset stability redundancy of the coupling zone of the load-bearing beam under the fire phase.

[0180] Step 5: Evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and adjust the concrete aggregate gradation parameter combination.

[0181] Step five includes:

[0182] S201: Online assessment of effective transmission level differences The minimum effective threshold for the stability of the preset structure, when the hierarchical difference When the value exceeds the minimum effective stability threshold, the discrete element sampling frequency is increased by 5%. This effectively improves the monitoring accuracy of the probability of gradation stability collapse. ;

[0183] S202: Simultaneously calculate the structural robustness maintenance rate under the current aggregate gradation configuration. Compared with the preset target maintenance rate The difference, the acceptable drift tolerance of the preset gradation optimization model;

[0184] When the difference is greater than the acceptable drift tolerance of the gradation optimization model, it is determined that the gradation scheme has insufficient biting force under fire conditions, an alarm signal is issued, and the proportion of medium and large particle size aggregates of 5.0mm-20mm is forcibly increased in the next round of gradation generation. In this embodiment, the single increase range is 1%-3%.

[0185] Example 2:

[0186] like Figure 2 As shown, an embodiment of the present invention provides a concrete parameter optimization model generation system, such as... Figure 2 As shown, it includes the following modules:

[0187] The module includes a digital twin modeling module, a gradation feature prediction module, a performance parameter calculation module, a robust performance evaluation module, and a gradation early warning and adjustment module.

[0188] The digital twin modeling module is used to extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and to establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient.

[0189] The gradation feature prediction module is used to obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model.

[0190] The performance parameter calculation module extracts the effective axial force prediction value and interface debonding displacement increment that change with the duration of fire based on the digital twin evolution model, and calculates and obtains the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient.

[0191] The robust performance evaluation module uses the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combines the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient to generate the structural topological robustness maintenance rate.

[0192] The gradation early warning and adjustment module is used to evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and to adjust the combination of concrete aggregate gradation parameters.

[0193] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired network and / or wireless network. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).

[0194] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0195] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0196] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0197] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0199] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for generating a concrete parameter optimization model, characterized in that, The method includes: Step 1: Extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient; Step one includes: A1: Based on the physical properties of the building information model of the beam, extract the surface roughness index, dynamic centroid deviation factor and body surface area change rate of the aggregate in each particle size range in the coupling zone to form the set of structural geometric feature vectors of the prefabricated prestressed continuous box girder. A2: Preset the fire source trajectory field function under different fire intensities, map the geometric feature vector of the structure to the discretized finite element topology mesh, and generate a three-dimensional heterogeneous thermal resistance coefficient mapping matrix. A3: Construct a digital twin evolution model that includes a prestressed vector field and a thermal conduction gradient, including: introducing the partial tensors of the longitudinal and transverse stress constraint components in the coupling region, obtaining the spatiotemporal evolution law of the temperature field and stress field in the coupling region by iteratively solving the thermo-mechanical coupling differential equations in the computational domain, and constructing a digital twin evolution model. Step 2: Obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model; Step 3: Based on the digital twin evolution model, extract the effective axial force prediction value and interface debonding displacement increment that change with the duration of fire exposure, and calculate and obtain the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient. Step three includes: C1: Within the digital twin evolution space defined by the digital twin evolution model constructed in step A3, simultaneously generate attenuation curves for the predicted values ​​of each effective axial force component under the preset fire scenario as a function of fire duration, as well as the temperature gradient at different cross-sections. The trajectory curve that evolves with the duration of fire exposure; C2: The system automatically extracts multiple fire evolution processes with different phases where the difference between two or more sets of effective axial force prediction values ​​does not exceed 5%, as a control group of the same magnitude load, and simultaneously extracts the effective contact density of aggregates in the control group. Curves showing the change over time and the phase difference curve of temperature rise ; C3: Extract the characteristic fluctuations of the heating gradient feature curve in the process of the evolution of the opposite phase fire, obtain the fluctuation component of the heating rate by removing the trend term, and construct the heating rate interference evaluation envelope set to characterize the temperature field interference. Take the temperature rise rate fluctuation within the evaluation envelope set. and the corresponding aggregate-slurry interface debonding displacement increment ; Define the set of temperature rise rate fluctuations as follows: Where the superscript g represents the heat conduction flux correction term at the current moment, and the subscript v represents the v-th discrete calculation period in the out-of-phase fire evolution process; the corresponding set of effective interface debonding displacements is expressed as: ; C4: Import the set of heating rate fluctuations and the set of effective interfacial debonding displacements into the interfacial debonding regression function to calculate the dynamic debonding sensitivity coefficient that reflects the robustness of the gradation scheme. The specific calculation formula is as follows: ； in, This is the arithmetic mean of the temperature rise rate fluctuations during the current assessment period. This represents the arithmetic mean of the interface debonding displacement increments at the corresponding stages. Used to quantify the negative contribution of thermal conductivity rate fluctuations to aggregate debonding; Step three also includes: C5: Extract the prestress loss curve obtained during the digital twinning process of the stress path and the predicted displacement phase difference curve at different heating depths d for precast prestressed continuous box girders. ; C6: Select an evolution process with a consistent heating gradient to form an anisotropic constraint disturbance set, and extract the prestrain phase adjustment amount caused by prestress fluctuations from this set. and the corresponding increase in the failure frequency of the effective contact network between aggregates. ; The set of pre-strain phase adjustment values ​​is set as follows: ; Wherein, the superscript pt represents the fitting state between the prestrain and the tension vector, and the subscript u represents the u-th interference calculation sequence; The corresponding set of aggregate contact failure increments is represented as follows: ； C7: The set of prestress phase adjustment values ​​and the corresponding set of aggregate contact failure frequencies are imported into the elastic modulus softening regression model to calculate the synergy factor dominated by prestress constraint fluctuations. The specific calculation formula is as follows: ； in, These are the expected mean values ​​of the corresponding feature sets during the constraint interference period. Used to quantify the probability of instability of aggregate gradation support frames due to non-uniform loss of prestress; Step three also includes: C8: Extract the structural geometric feature vectors of aggregates from each gradation scheme within the set of structural geometric feature vectors constructed in step A1, and calculate the comprehensive derived feature factor of the load-bearing skeleton of the current gradation sequence. The calculation formula is: ； in, These are the surface roughness index, dynamic centroid deviation factor, and surface area change rate of the aggregate in this diameter segment, obtained from analysis within the digital twin. These are the contribution weights of the feature-derived factors to the thermo-mechanical evolution under the corresponding fire protection phase; C9: Extract multiple evolution path sub-items that maintain consistency between the predefined simulated heating gradient and the simulated prestress level, and form a morphological interference set for the evolution of gradation characteristics; From the morphological interference set, multiple comprehensive derived feature factors arranged in heterogeneous order and their corresponding beams are simultaneously extracted as the evolution increment of the structural redundancy prediction value at the extreme failure threshold point. The set of comprehensive derived feature factors is as follows: ； The superscript h represents the feature identifier corresponding to the current extraction task, and the subscript e represents the e-th specific mixed gradation topology sequence. The comprehensive derived characteristic factor corresponding to the e-th gradation; Among them, the set of continuously evolving effective quantities of the structural redundancy prediction values ​​that match it is: ,in The shear load value retained at the end of the fire simulation process corresponding to the e-th specific gradation obtained from the digital space; C10: A set of comprehensive derived feature factors and the evolution set of structural redundancy prediction values Import the response co-analysis function for aggregate geometric contribution, and calculate the topological redundancy response coefficient. The specific calculation formula is as follows: ； in, The average value of the comprehensive derived feature imprint values ​​within the evaluation period of the interference set. This represents the average value of the predicted residual load components within the corresponding structure. Step 4: Utilize the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combine the dynamic debonding sensitivity coefficient, synergy factor, and topological redundancy response coefficient to generate the structural topological robustness maintenance rate. Step 5: Evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and adjust the concrete aggregate gradation parameter combination.

2. The method for generating a concrete parameter optimization model according to claim 1, characterized in that, Step two includes: B1: Obtain the relative position parameters of the ignition point from the bottom plate under the preset fire scenario, and simulate the bottom plate response curve and web hysteresis displacement curve under different heating phases through the digital twin evolution model to construct the fire response deviation load sequence of the prefabricated prestressed continuous box girder. B2: Retrieve performance degradation data of the same type of prefabricated prestressed continuous box girder under historical fire conditions, and extract the volume ratio of aggregates in each diameter segment, the initial feature vector of the interface transition zone, the aggregate-slurry coating thickness index, the effective contact density decay trajectory of aggregates under the corresponding conditions, and the shear stiffness retention rate of the beam from the multi-dimensional historical gradation. This constitutes a training set for the evolution of gradation historical features. The data is divided into a basic training set and a validation set for the gradation generation model in an 8:2 ratio.

3. The method for generating a concrete parameter optimization model according to claim 2, characterized in that, Step two also includes: B3: The surface spheroidization factor, contact coordination number distribution gradient and thermal expansion and shrinkage rate difference of aggregates of different particle sizes within a single gradation range are obtained and dimensionlessly processed to form the basic feature mapping space of aggregates of each particle size within the full gradation envelope. B4: Construct a gradation parameter generation neural network model, which includes: a topological signal input layer, a non-homogeneous constrained convolutional layer, and a steady-state redundancy mapping layer; The non-homogeneous constrained convolutional layer contains a pre-set aggregate displacement interference sensitive weight kernel for the 1.18mm-2.36mm bridging diameter segment; B5: Perform dimensionality reduction and standardization on the basic feature mapping space and gradation history feature evolution training set of each aggregate diameter segment, and activate the non-homogeneous constrained hidden layer for tensor correction calculation. The specific formula is as follows: ； in, For the first The situational output of the j-th mapping unit in the hidden layer. The thermal-mechanical characteristic excitation is for the i-th unit in the previous layer. This refers to the phase offset term of the element under specific prestress constraints. For the first Layer to the first The contribution weight of aggregate dissociation evolution during layer transfer; B6: Use the validation set to evaluate the fit of the neural network model for generating the gradation parameters until the sum of the prediction residuals reaches the convergence threshold, thus obtaining the initial gradation generation model that has completed the initial weight training.

4. The method for generating a concrete parameter optimization model according to claim 3, characterized in that, Step four includes: S101: By... , as well as Weighted summation yields the multi-field fusion adaptive coordination number of three-dimensional asynchronous evolution. ; S102: Retrieve the initial gradation generation model trained in step B6, and fuse the adaptive matching numbers from multiple fields. This is mapped to the correction operator of the non-homogeneous constrained convolutional layer in the neural network model; This correction operator is used to interfere with the internal structure of the neural network in real time. The situational output of the hidden layer The online correction of the neural network weights is completed to obtain the optimized gradation generation model; S103: Input the real-time monitored beam geometry and load sequence into the optimized gradation generation model to predict the effective transfer level of the prestressed load-bearing beam under the current specific gradation parameter combination. ; S104: Obtain the structural topological robustness maintenance rate that characterizes the stability of the model scheme. The specific generation strategy is as follows: ; in, The effective transmission level of the force path currently generated within the digital twin space under a specific fire phase; The target value for the preset stability redundancy of the coupling zone of the load-bearing beam under the fire phase.

5. The method for generating a concrete parameter optimization model according to claim 4, characterized in that, Step five includes: S201: Online assessment of effective transmission level differences The minimum effective threshold for the stability of the preset structure, when the hierarchical difference When the sampling frequency exceeds the minimum effective stability threshold, the finite element discrete mesh sampling frequency of the digital twin evolution model is increased by 5%, wherein... ; S202: Simultaneously calculate the structural robustness maintenance rate under the current aggregate gradation configuration. Compared with the preset target maintenance rate The difference is used to preset the acceptable drift tolerance of the optimized gradation generation model; When the difference exceeds the acceptable drift tolerance of the gradation optimization model, it is determined that the gradation scheme has insufficient interlocking force under fire conditions, an alarm signal is issued, and the proportion of medium and large-sized aggregates of 5.0mm-20mm is forcibly increased in the next round of gradation generation.

6. A concrete parameter optimization model generation system, used to implement the concrete parameter optimization model generation method as described in any one of claims 1-5, characterized in that, The system includes the following modules: The module includes a digital twin modeling module, a gradation feature prediction module, a performance parameter calculation module, a robust performance evaluation module, and a gradation early warning and adjustment module. The digital twin modeling module is used to extract the set of structural geometric feature vectors of the web-bottom coupling zone of the prefabricated prestressed continuous box girder, and to establish a digital twin evolution model that integrates the prestressed vector field and the heat conduction gradient. The gradation feature prediction module is used to obtain the fire response deviation load sequence under the preset fire scenario, simultaneously retrieve the gradation history feature evolution training set for aggregates of different diameters, and map the load sequence and feature evolution training set to the preset gradation parameter generation neural network model. The performance parameter calculation module extracts the effective axial force prediction value and interface debonding displacement increment that change with the duration of fire based on the digital twin evolution model, and calculates and obtains the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient. The robust performance evaluation module uses the gradation parameters to generate non-homogeneous constrained convolutional layers in the neural network model to perform tensor correction calculations, and combines the dynamic debonding sensitivity coefficient, synergy factor and topological redundancy response coefficient to generate the structural topological robustness maintenance rate. The gradation early warning and adjustment module is used to evaluate the deviation tolerance between the maintenance rate generated by the current model and the target maintenance rate online, and to adjust the combination of concrete aggregate gradation parameters.

Images