UHPC composite beam early warning method, device and equipment based on deep learning and medium

Abstract

The invention discloses a UHPC composite beam early warning method and device based on deep learning, equipment and a medium, and relates to the technical field of bridge structure monitoring, and the method comprises the steps: obtaining the construction information of an ultra-high performance concrete composite beam, inputting the construction information into a preset state prediction model for processing, and obtaining a target feature state and a target performance prediction result; and based on the target feature state and the target performance prediction result, detection is carried out through a multi-level early warning mechanism, a detection result is obtained, and corresponding early warning operation is executed according to the detection result. The deep learning model is used for processing construction information of the UHPC composite beam, key feature states are automatically recognized, performance is predicted, real-time monitoring and risk early warning are achieved through a multi-stage early warning mechanism, the accuracy of feature state recognition and the reliability of performance prediction are improved, early risk detection and active early warning are achieved, and the early warning efficiency is improved. And the structural safety and the design efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge structure monitoring, and in particular to a UHPC composite beam early warning method, device, equipment and medium based on deep learning. Background Art

[0002] With the widespread application of ultra-high performance concrete (UHPC) in bridge engineering, steel-UHPC composite beams have received increasing attention due to their excellent mechanical properties and durability. Domestic and foreign scholars have conducted extensive research on this type of composite beams, and have made significant progress in material properties, structural measures, computational theory, and experimental research. Through systematic research, the constitutive relationship and mechanical properties of UHPC and steel have been established, and a variety of combinations and connection structures have been developed to improve the overall performance of the structure. In addition, a design method system based on cross-sectional analysis has been formed, and a large number of static and fatigue tests have provided reliable data support for theoretical research.

[0003] Existing research methods mainly rely on classical mechanics theory for analysis, which requires a large number of iterative calculations. The process is complex and time-consuming, and it is difficult to meet the needs of rapid design. For characteristic state identification, the current methods mainly rely on preset judgment conditions and engineering experience, lack intelligent means, especially in the state transition and failure mode prediction under complex working conditions. In terms of performance prediction, especially in complex working conditions and long-term performance prediction, the accuracy and reliability of existing methods need to be improved. The early warning mechanism generally has a lag, and early risk identification and active prevention cannot be achieved. More importantly, the large amount of accumulated test data and engineering experience has not been fully utilized, and knowledge inheritance and experience accumulation face bottlenecks. Therefore, an efficient and accurate method is needed to detect and warn UHPC composite beams. Summary of the invention

[0004] The main purpose of this application is to provide a UHPC composite beam early warning method, device, equipment and medium based on deep learning, aiming to solve the technical problem of how to obtain the status of the UHPC composite beam and perform corresponding early warning operations.

[0005] To achieve the above objectives, this application proposes a UHPC composite beam early warning method based on deep learning, the method comprising: Obtain the structural information of ultra-high performance concrete composite beams; Input the construction information into a preset state prediction model for processing to obtain target characteristic state and target performance prediction results, wherein the preset state prediction model is a hybrid neural network model, and the preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer, wherein the feature extraction network includes a first convolution block, a second convolution block and a third convolution block, wherein the first convolution block, the second convolution block and the third convolution block each include two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer, wherein the temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer, wherein the output layer includes a characteristic state classification branch and a performance prediction branch, wherein the characteristic state classification branch outputs the probability distribution of the characteristic state through a three-layer fully connected network, and the performance prediction branch outputs the performance index value through a three-layer fully connected network, wherein the hybrid neural network model performs local feature extraction through the feature extraction network, captures temporal features through the bidirectional long short-term memory network and weights the key features through the attention layer to achieve deep learning of the characteristic state of the ultra-high performance concrete composite beam; Based on the target characteristic state and the target performance prediction result, detection is performed through a multi-level early warning mechanism to obtain a detection result; Execute corresponding warning operations according to the detection results.

[0006] In one embodiment, before the step of inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, the step includes: Obtain sample data of structural information and build an initial state prediction model; The initial state prediction model is trained based on the construction information sample data to obtain a preset state prediction model.

[0007] In one embodiment, the step of training the initial state prediction model based on the construction information sample data to obtain a preset state prediction model includes: Initializing weight and bias parameters of the initial state prediction model; Inputting the structural information sample data into the initial state prediction model for calculation to obtain characteristic state classification results and performance prediction results; Calculate and weight the feature state classification result and the performance prediction result according to the loss function to obtain a total error value; The gradients of the weight and bias parameters are obtained by back propagation algorithm calculation; The weight and bias parameters are iteratively updated through an optimization algorithm according to the gradient until a maximum number of iterations is reached or the total error value converges to a preset threshold, thereby obtaining a preset state prediction model.

[0008] In one embodiment, before the step of inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, the step includes: Inputting the structural information into a preset state prediction model for processing to obtain a characteristic state; Performing an evaluation based on the feature status to obtain a confidence score; When the confidence score exceeds a preset threshold, the feature state is output as a target feature state; When the confidence score does not exceed the preset threshold, an early warning signal is issued, and the step of inputting the construction information into a preset state prediction model for processing to obtain a characteristic state is executed until the confidence score exceeds the preset threshold.

[0009] In one embodiment, the step of performing detection based on the target feature state and the target performance prediction result through a multi-level early warning mechanism to obtain the detection result includes: According to the multi-level warning mechanism, a first threshold, a second threshold and a third threshold are obtained, wherein the first threshold is less than the second threshold, and the second threshold is less than the third threshold; Obtaining a composite beam bending moment according to the target characteristic state and the target performance prediction result; The composite beam bending moment is compared with the first threshold, the second threshold and the third threshold to obtain a detection result.

[0010] In one embodiment, the step of performing a corresponding early warning operation according to the detection result includes: When the bending moment of the composite beam is less than the first threshold value, the detection result is obtained as no risk, and normal monitoring and recording of key parameters are performed; When the combined beam bending moment is greater than the first threshold and the combined beam bending moment is less than the second threshold, the detection result is obtained as a first-level warning, and the operation of increasing the frequency of obtaining structural information and preparing an emergency plan is performed; When the combined beam bending moment is greater than the second threshold and the combined beam bending moment is less than the third threshold, the detection result is obtained as a second-level warning, and an operation of sending warning information to relevant personnel through real-time monitoring is performed; When the bending moment of the composite beam is greater than the third threshold, the detection result is a third-level warning, and an operation of triggering an emergency response for maintenance is executed.

[0011] In addition, to achieve the above purpose, the present application also proposes a UHPC composite beam early warning device based on deep learning, and the UHPC composite beam early warning device based on deep learning includes: An acquisition module, used to acquire the structural information of the ultra-high performance concrete composite beam; A processing module, used for inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, wherein the preset state prediction model is a hybrid neural network model, and the preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer, wherein the feature extraction network includes a first convolution block, a second convolution block and a third convolution block, wherein the first convolution block, the second convolution block and the third convolution block each include two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer, wherein the temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer, wherein the output layer includes a characteristic state classification branch and a performance prediction branch, wherein the characteristic state classification branch outputs a probability distribution of the characteristic state through a three-layer fully connected network, and the performance prediction branch outputs a performance index value through a three-layer fully connected network, wherein the hybrid neural network model performs local feature extraction through the feature extraction network, captures temporal features through the bidirectional long short-term memory network and weights key features through the attention layer to achieve deep learning of the characteristic state of the ultra-high performance concrete composite beam; A detection module, used to perform detection through a multi-level early warning mechanism based on the target feature state and the target performance prediction result to obtain a detection result; An execution module is used to execute corresponding warning operations according to the detection results.

[0012] In addition, to achieve the above-mentioned purpose, the present application also proposes a medium, which is a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the UHPC composite beam early warning method based on deep learning as described above are implemented.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the UHPC composite beam early warning method based on deep learning as described above.

[0014] This application obtains the structural information of ultra-high performance concrete composite beams, inputs the structural information into a preset state prediction model for processing, and obtains the target characteristic state and target performance prediction results; based on the target characteristic state and target performance prediction results, a multi-level early warning mechanism is used to detect and obtain the detection results, and the corresponding early warning operation is performed according to the detection results. The deep learning model is used to process the structural information of UHPC composite beams, automatically identify key characteristic states and predict performance, and achieve real-time monitoring and risk warning through a multi-level early warning mechanism, which improves the accuracy of characteristic state identification and the reliability of performance prediction, realizes early risk detection and active warning, and improves structural safety and design efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0016] Figure 1 This is a flow chart of the first embodiment of the UHPC composite beam early warning method based on deep learning in this application; Figure 2 A cross-sectional structural diagram of a UHPC composite beam of the first embodiment of the UHPC composite beam early warning method based on deep learning of the present application; Figure 3 This is a schematic diagram of characteristic state recognition results of the first embodiment of the UHPC composite beam early warning method based on deep learning in this application; Figure 4 This is a schematic diagram of the performance prediction results of the first embodiment of the UHPC composite beam early warning method based on deep learning in this application; Figure 5 This is a flow chart of the second embodiment of the UHPC composite beam early warning method based on deep learning in this application; Figure 6 This is a flow chart of the third embodiment of the UHPC composite beam early warning method based on deep learning in this application; Figure 7 This is a schematic diagram of the module structure of the UHPC composite beam early warning device based on deep learning in an embodiment of the present application; Figure 8 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the UHPC composite beam early warning method based on deep learning in the embodiment of the present application.

[0017] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0019] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0020] With the widespread application of ultra-high performance concrete (UHPC) in bridge engineering, UHPC-steel composite beams have attracted more and more attention due to their excellent mechanical properties and durability. UHPC composite beams not only have high strength and high durability, but also can effectively reduce the deadweight of the structure, improve the service life and economy of the structure. Therefore, UHPC composite beams have been widely used in modern bridge engineering, especially in large bridges and special structures. However, with the widespread application of UHPC composite beams, higher requirements are put forward for real-time monitoring and evaluation of their performance. Traditional analysis and monitoring methods mainly rely on classical mechanics theory and empirical formulas, which have certain limitations in dealing with complex structural behaviors and dynamic changes. In the feature state identification, there is an excessive reliance on manual experience and preset judgment conditions, and there is a lack of automation and intelligent means; secondly, the accuracy of performance prediction is insufficient, especially when dealing with complex working conditions, the reliability of the prediction results needs to be further improved; thirdly, the traditional calculation method is complex and time-consuming, and cannot meet the needs of rapid design; finally, the early warning mechanism has a lag, which makes it difficult to achieve real-time risk identification and active prevention.

[0021] Therefore, the present application proposes a UHPC composite beam early warning method based on deep learning. The main solution of the embodiment of the present application is: by acquiring the structural information of the ultra-high performance concrete composite beam, the structural information is input into a preset state prediction model for processing to obtain the target characteristic state and the target performance prediction results; based on the target characteristic state and the target performance prediction results, a multi-level early warning mechanism is used to perform detection to obtain the detection results, and the corresponding early warning operation is performed according to the detection results.

[0022] Based on this, the embodiment of the present application provides a UHPC composite beam early warning method based on deep learning, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the UHPC composite beam early warning method based on deep learning in this application.

[0023] In this embodiment, the UHPC composite beam early warning method based on deep learning includes steps S10 to S40: Step S10, obtaining structural information of the ultra-high performance concrete composite beam.

[0024] It should be noted that the construction information of ultra-high performance concrete (UHPC) composite beams covers many aspects, including geometric characteristics, material characteristics, and load characteristics. First, in terms of geometric characteristics, the specific parameters of UHPC beams and steel beams are the key. Figure 2From the cross-sectional structural diagram of the UHPC composite beam shown, it can be seen that the flange width of the UHPC beam is 1200mm, the thickness is 120mm, the web width is 120mm, and the height is 480mm; the flange width of the steel beam is 400mm, the thickness is 25mm, the web thickness is 16mm, and the height is 575mm. In addition, detailed reinforcement arrangement parameters are included, such as the top longitudinal reinforcement uses 12 steel bars with a diameter of 12mm and a protective layer thickness of 20mm; the bottom longitudinal reinforcement uses 2 steel bars with a diameter of 16mm and a protective layer thickness of 20mm.

[0025] Specifically, in terms of material characteristics, UHPC has extremely high compressive strength (fcu=145MPa) and tensile strength (ft=10.87MPa), its elastic modulus is 45.7GPa, and it has a specific design compressive strength (fcud=71.05MPa) and design tensile strength (ftd=7.61MPa). The elastic modulus of steel is 206.0GPa, the yield strength reaches 355MPa, the ultimate strength is as high as 512MPa, and its design strength is 286.13MPa. The performance parameters of steel bars are also very important, with an elastic modulus of 200.0GPa, a yield strength of 400MPa, and an ultimate strength of 565MPa. The load characteristics mainly involve positive and negative moment parameters, which are crucial for evaluating the response of the structure under different working conditions. By standardizing these input features, the dimensional effect can be eliminated and the model training process can be optimized. For example, the length dimension is divided by the span reference value, the area parameter is divided by the cross-sectional area reference value, and the strength parameter is divided by the benchmark strength.

[0026] Step S20, inputting the construction information into a preset state prediction model for processing to obtain target characteristic state and target performance prediction results.

[0027] It should be noted that the acquired structural information including geometric features, material features and load characteristics is standardized, including the normalization of length dimensions, area parameters and strength parameters, in order to eliminate the dimensional influence. The time series features are extracted by the sliding window method, and Gaussian noise injection is used to achieve data enhancement to ensure the generalization ability of the model.

[0028] Furthermore, the above-mentioned standardized structural information is input into the preset state deep model. The above-mentioned preset state prediction model is a hybrid neural deep learning model. The preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer. The feature extraction network includes a first convolution block, a second convolution block and a third convolution block. The first convolution block, the second convolution block and the third convolution block all contain two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer. The temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer. The output layer includes a feature state classification branch and a performance prediction branch. The feature state classification branch outputs the probability distribution of the feature state through a three-layer fully connected network, and the performance prediction branch outputs the performance index value through a three-layer fully connected network. According to the hybrid neural network model, the feature extraction network is used to extract local features, the bidirectional long short-term memory network is used to capture temporal features, and the attention layer is used to weight key features to achieve deep learning of the feature state of the ultra-high performance concrete composite beam.

[0029] Furthermore, the hybrid neural network model is adopted because it can make full use of the advantages of different neural network architectures to cope with complex feature extraction and prediction tasks. Specifically, the feature extraction network is good at processing local features and can automatically extract spatial hierarchical structures in data such as images and signals. When processing the geometric features and material features of UHPC composite beams, the feature extraction network can effectively capture local mechanical behavior. The bidirectional long short-term memory network (BiLSTM) can capture long-term dependencies in time series data and is particularly suitable for processing the temporal evolution characteristics of mechanical properties. When analyzing the changes of stress and strain over time, BiLSTM can provide more comprehensive temporal information. The attention mechanism can dynamically highlight the importance of key features, so that the model can adaptively focus on the most helpful parts for prediction. When identifying the key characteristic states of UHPC composite beams (such as cracking, yielding, etc.), the attention mechanism can improve the accuracy and reliability of the model. In this embodiment, the local mechanical features are first extracted by three convolution blocks in series, each of which contains two one-dimensional convolution layers, configured with 64, 128, and 256 channels respectively, with a kernel size of 3, and equipped with a normalization layer and a ReLU activation function. Then, the temporal feature network uses a bidirectional LSTM network to capture the temporal evolution characteristics of mechanical properties. The bidirectional LSTM network consists of two layers of BiLSTM, each with 128 and 64 hidden units, which can effectively identify dependencies on different time scales and is particularly suitable for analyzing the development process of stress and strain. In addition, in order to highlight the influence of key states, the model introduces a multi-head self-attention mechanism, in which 8 attention heads are set, each with a dimension of 64. The feature correlation is calculated through the Query / Key / Value matrix, so that the model can adaptively pay attention to different stress and strain development situations. Finally, in the output layer, the model adopts a multi-task learning framework, which is divided into a feature state classification branch and a performance prediction branch. The feature state classification branch outputs the probability distribution of each feature state through a three-layer fully connected network (256-128-n dimensions respectively), such as the first cracking of UHPC, steel beam yielding and other key states; the performance prediction branch also uses a three-layer fully connected network (linear activation function) to output specific bearing capacity and deformation values. Through this hybrid neural network structure, the advantages of different neural network architectures can be fully utilized to achieve accurate identification of the characteristic state of UHPC composite beams and precise prediction of their performance.

[0030] Further, step S20 also includes: inputting the structural information into the preset state prediction model for processing to obtain the characteristic state; evaluating according to the characteristic state to obtain the confidence score; when the confidence score exceeds the preset threshold, the characteristic state is output as the target characteristic state; when the confidence score does not exceed the preset threshold, an early warning signal is issued, and the step of inputting the structural information into the preset state prediction model for processing to obtain the characteristic state is executed until the confidence score exceeds the preset threshold. Specifically, after obtaining the preliminary characteristic state, these states are evaluated and a confidence score is calculated. This score is based on the accuracy of the model's recognition of various characteristic states and its ability to predict the development of stress and strain under specific working conditions. When the confidence score exceeds the preset threshold (in this embodiment, the preset threshold is set to 0.95), it means that the model has a high degree of credibility in parsing the current structural information, and the output characteristic state is regarded as the target characteristic state, such as the first cracking of UHPC or the yielding of steel beams and other key states. This mechanism ensures that only the state that has been strictly verified will be finally adopted, providing a solid foundation for engineering design. However, if the confidence score does not reach the preset threshold, it means that there is uncertainty or error in the current analysis. In this case, an early warning signal is automatically issued, indicating that further data analysis or model adjustment is needed. Subsequently, the construction information input and processing steps will be re-executed, including adjusting the data preprocessing method (such as improving normalization technology or enhancing the data set), optimizing weights and bias parameters (such as increasing the number of convolutional layers or adjusting the number of LSTM units), and even improving the model architecture itself (such as introducing a more complex attention mechanism). This cycle will continue until the confidence score meets the requirements. This iterative optimization strategy not only improves the robustness and adaptability of the model, but also effectively reduces the risk of misjudgment, ensuring that the target feature state of each output is as close to the actual situation as possible, thereby providing strong protection for structural safety.

[0031] Step S30, performing detection through a multi-level early warning mechanism based on the target feature state and the target performance prediction result to obtain a detection result.

[0032] It should be noted that when the model outputs the target characteristic state (such as the first cracking of UHPC, yielding of steel beams, etc.) and the target performance prediction results (such as the predicted values ​​of bearing capacity and deformation), this information will be input into the preset multi-level early warning system. The above early warning system includes three core modules: state classifier, trend predictor and risk assessor. The state classifier identifies the state of the current structure based on the characteristic points on the M-φ curve; the trend predictor analyzes the changes in the slope of the curve to predict future development trends; and the risk assessor monitors key strain values ​​to assess potential risks. The system sets three levels of early warning thresholds based on the characteristic points on the M-φ curve. Figure 3The characteristic state identification result diagram shown in the figure uses the M-φ curve to display each characteristic state point. The key points such as UHPC cracking state (εtu=0.00293), steel beam yield state (εy=0.001723), steel beam strengthening state (εst=0.0198) and limit state (εu=0.035) are marked on the curve, and the corresponding strain values ​​are given. Figure 4 The performance prediction results shown in the figure show the comparison curves of bearing capacity prediction and deformation prediction. At the beginning of loading, when the lower end of the UHPC web reaches the elastic limit (point B, φ=0.00157 1 / m), the bending moment is 4380.5 kN·m. As loading continues, the lower flange of the steel first reaches the design calculated stress (point I, φ=0.00185 1 / m, M=5161.4 kN·m), and then reaches the yield strain (point E, φ=0.00228 1 / m, M=6331.0 kN·m). When the upper edge of the UHPC reaches the design calculated stress (point H, φ=0.00431 1 / m), the bending moment increases to 8094.6 kN·m. When approaching the limit state, the lower end of the UHPC web reaches the limit strain (point C, φ=0.00837 1 / m, M=9031.3 kN·m), and finally the lower flange of the steel enters the strengthening stage (point F, φ=0.01881 1 / m, M=9131.6 kN·m). In the prediction process, the model comprehensively considers multiple key factors. By analyzing the strain distribution at each characteristic point, such as at the yield point E of the steel beam, the strain of the upper edge is -1.020‰ and the strain of the lower edge is 0.351‰, the stress state of the component can be accurately judged. At the same time, the model also tracks the change of the neutral axis position, and finds that it gradually rises from the initial 448.6mm from the top edge of the section to 147.3mm from the top edge of the section. This change trend reflects the process of cross-sectional stress redistribution. In addition, by analyzing the change in the slope of the M-φ curve, the degradation process of the structural stiffness can be clearly reflected.

[0033] Step S40, executing corresponding warning operations according to the detection results.

[0034] It should be noted that when the bending moment reaches 4380.5 kN·m, the system will automatically increase the data sampling frequency and activate the trend analysis module for more detailed monitoring. At this point, the system begins to prepare emergency plans, including but not limited to notifying relevant engineering personnel to strengthen on-site monitoring and prepare emergency supplies. The main purpose of this stage is to be fully prepared for possible problems so as to respond quickly to any emergencies. As loading continues, when the bending moment increases to 5161.4 kN·m, the system will further upgrade its response measures. At this point, the real-time monitoring mode is activated, and the system will not only continue to track the changes in key strain values, but also call the loading control module to adjust the loading rate or suspend construction activities to avoid further increasing the structural burden. At the same time, the system will send early warning information to all relevant personnel to ensure that everyone can understand the current situation in a timely manner and take necessary protective measures. This early intervention can effectively prevent small problems from turning into major accidents. When the bending moment reaches 8094.6 kN·m, the system will immediately implement emergency protection measures. This may include stopping all construction operations, evacuating on-site personnel, and implementing temporary reinforcement measures.

[0035] This embodiment provides a UHPC composite beam early warning method based on deep learning, which obtains the structural information of the ultra-high performance concrete composite beam, inputs the structural information into a preset state prediction model for processing, and obtains the target characteristic state and target performance prediction results; based on the target characteristic state and target performance prediction results, a multi-level early warning mechanism is used to detect and obtain the detection results, and the corresponding early warning operation is performed according to the detection results. The structural information of the UHPC composite beam is processed using a deep learning model, and the key characteristic states are automatically identified and the performance is predicted. Real-time monitoring and risk early warning are achieved through a multi-level early warning mechanism, which improves the accuracy of characteristic state identification and the reliability of performance prediction, realizes early risk detection and active early warning, and improves structural safety and design efficiency.

[0036] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can refer to the above introduction, and will not be repeated later. Figure 5 The UHPC composite beam early warning method based on deep learning further includes steps S201 to S202 before step S20: Step S201, obtaining structure information sample data and building an initial state prediction model.

[0037] It should be noted that a series of representative sample data are collected, which cover the key geometric features, material properties and load conditions of composite beams. For example, the geometric features include the specific dimensional parameters of UHPC beams and steel beams, such as flange width, thickness, web width and height; the material properties involve the mechanical properties of UHPC, steel and steel bars, such as elastic modulus, compressive strength, yield strength, etc.; the load conditions mainly focus on positive and negative moment parameters. After obtaining these sample data, it is crucial to perform standardization. This step includes normalizing all input features to eliminate dimensional differences, extracting time series features using the sliding window method, and enhancing the data set by injecting Gaussian noise to ensure that the model has good generalization ability. In addition, the stability and accuracy of the model can be further evaluated using the five-fold cross-validation technique.

[0038] Furthermore, an initial state prediction model is constructed based on a deep learning method. The model contains several key stages: first, a feature extraction network, which consists of three convolution blocks, each of which has two one-dimensional convolution layers (with 64, 128, and 256 channels, respectively), normalization layers, ReLU activation functions, and maximum pooling layers, which are used to gradually extract local features in the stress-strain field. Then, a bidirectional LSTM network is used to capture the temporal evolution characteristics of mechanical properties. This design is particularly suitable for dealing with situations where structural responses change over time. In order to highlight the impact of key states, a multi-head self-attention mechanism is introduced, which enables the model to adaptively focus on different stress-strain developments. Finally, a multi-task learning framework is used in the output layer, and feature state classification branches and performance indicator prediction branches are set up respectively. The classification branch outputs the probability distribution of each feature state through a three-layer fully connected network, while the regression branch predicts the specific bearing capacity and deformation values.

[0039] Step S202: training the initial state prediction model based on the construction information sample data to obtain a preset state prediction model.

[0040] It should be noted that when training the initial state prediction model, first initialize the weight and bias parameters of the initial state prediction model; input the constructed information sample data into the initial state prediction model for calculation to obtain the feature state classification result and the performance prediction result; calculate and weight-process the feature state classification result and the performance prediction result according to the loss function to obtain the total error value; calculate through the backpropagation algorithm to obtain the gradients of the weight and bias parameters; update the weight and bias parameters iteratively according to the gradients through the optimization algorithm until the maximum number of iterations is reached or the total error value converges to a preset threshold to obtain the preset state prediction model. Specifically, when initializing the weight and bias parameters of the model, a random initialization method or the parameters of a pre-trained model are usually used as the starting point. Then, input the prepared constructed information sample data into the initial state prediction model, and through forward propagation calculation, obtain the feature state classification result (such as the first cracking of UHPC, the yield of the steel beam, etc.) and the performance prediction result (such as the bearing capacity and deformation value). This process utilizes the convolutional blocks in the model to extract local mechanical features, the bidirectional LSTM network to capture temporal evolution features, and the multi-head self-attention mechanism to highlight the influence of key features. Subsequently, based on the loss function, evaluate and weight-process the feature state classification result and the performance prediction result to obtain the total error value. The loss function used here is usually a combination of cross-entropy loss (for classification tasks) and mean squared error loss (for regression tasks), corresponding to the results of the feature state classification branch and the performance index prediction branch respectively. The weighting process ensures that the relative importance of the two types of tasks is appropriately considered in the calculation of the total error value. Next, use the backpropagation algorithm to calculate the gradients of the weights and bias parameters of each layer of the model according to the total error value. This process starts from the output layer, propagates the error layer by layer forward, and calculates the contribution degree of the parameters of each layer to the total error. Then, based on the calculated gradients, use an optimization algorithm (such as Adam or SGD) to iteratively update the model parameters. The selection and adjustment of the optimization algorithm are crucial for accelerating convergence and improving the model performance. The entire training process will continue to iterate until the predetermined maximum number of iterations is reached or the total error value converges to a preset threshold. During this period, the model gradually learns how to more accurately identify the feature state and predict the structural performance while reducing the error. To prevent overfitting, regularization techniques such as Dropout or L2 regularization are usually introduced, and the performance of the model is monitored using the validation set to ensure its generalization ability.

[0041] Finally, the fully trained model becomes the preset state prediction model, which can efficiently and accurately complete the feature state recognition and performance prediction tasks in practical applications, providing strong technical support for the safety monitoring of UHPC composite beams.

[0042] After obtaining the preset state prediction model, it also includes: evaluating the preset state prediction model to obtain an evaluation result; if the evaluation result does not meet the detection requirements, retraining the preset state prediction model until the detection requirements are met.

[0043] This embodiment collects UHPC composite beam structural information, constructs and trains an initial state prediction model, and obtains a preset state prediction model by iterating the optimization algorithm until the error converges. This improves the accuracy of feature state recognition and the reliability of performance prediction, achieves early risk warning, and enhances structural safety and design efficiency.

[0044] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above description, and will not be described in detail later. Figure 6 The UHPC composite beam early warning method based on deep learning, step S30, also includes steps S301 to S303: Step S301, obtaining a first threshold, a second threshold and a third threshold according to a multi-level warning mechanism.

[0045] It should be noted that the first threshold is smaller than the second threshold, and the second threshold is smaller than the third threshold. In the present embodiment, the first threshold corresponds to a bending moment of 4380.5 kN·m, the second threshold corresponds to a bending moment of 5161.4 kN·m, and the third threshold corresponds to a bending moment of 8094.6 kN·m.

[0046] Step S302, obtaining the composite beam bending moment according to the target characteristic state and the target performance prediction result.

[0047] It should be noted that after the deep learning model processes the structural information samples, it will output a series of target characteristic states (such as the first cracking of UHPC, yielding of steel beams, etc.) and target performance prediction results (such as bearing capacity and deformation). These prediction results not only provide the response of the structure under different stress and strain conditions, but also help identify the key stress stages of the structure.

[0048] Specifically, by analyzing the characteristic points on the M-φ curve, the moment value of the composite beam under specific working conditions can be accurately obtained. For example, when the lower end of the UHPC web reaches the elastic limit, the corresponding moment value is 4380.5 kN·m, which indicates that the structure begins to enter the nonlinear working range; as the loading continues, when the lower flange of the steel reaches the design calculated stress, the moment increases to 5161.4 kN·m, indicating that the structure is subjected to a greater load; further, when the lower flange of the steel reaches the yield strain, the moment value rises to 6331.0 kN·m, which is an important sign that the structure is close to its ultimate bearing capacity; finally, when the upper edge of the UHPC reaches the design calculated stress, the moment value can be as high as 8094.6 kN·m, and the maximum moment value when reaching the ultimate strain is 9031.3 kN·m.

[0049] Step S303, comparing the composite beam bending moment with the first threshold, the second threshold and the third threshold to obtain a detection result.

[0050] It should be noted that the bending moment value predicted by the deep learning model is compared and analyzed with the threshold value corresponding to the set key feature point to obtain the detection result.

[0051] Furthermore, step S302 also includes: when the bending moment of the composite beam is less than the first threshold, the detection result is obtained as risk-free, and normal monitoring and recording of key parameters are performed. Specifically, when the bending moment value of the UHPC composite beam is less than 4380.5 kN·m, the detection result shows that the structure is currently in a risk-free state. This means that the structure is still within the elastic working range and no signs of potential failure or damage have appeared. In this case, the system will continue to perform normal monitoring and record key parameters to ensure that the health of the structure can be continuously tracked.

[0052] When the composite beam bending moment is greater than the first threshold and the composite beam bending moment is less than the second threshold, the detection result is a first-level warning, and the operation of increasing the frequency of obtaining structural information and preparing an emergency plan is performed. Specifically, when the bending moment value of the UHPC composite beam is greater than 4380.5 kN·m and less than 5161.4 kN·m, the system detection result is a first-level warning. This means that the structure has begun to enter the nonlinear working range. Although it has not yet reached a dangerous state, potential risk signals have appeared, requiring closer attention and preparation of countermeasures. In this case, the system will perform a series of operations to increase the frequency of obtaining structural information and start the preparation of emergency plans. First, the data sampling frequency is significantly increased to more accurately capture any subtle changes in the structural response. This includes not only the monitoring of key parameters such as bending moment, strain and curvature, but may also be extended to other related factors such as temperature changes, humidity effects, etc., to ensure a comprehensive understanding of the current status of the structure. At the same time, the system will activate the trend analysis module to predict the behavior change trend of the structure in the short term in the future through a deep learning model. This module uses historical data and real-time collected data for comparative analysis to identify possible development patterns and potential risk points. Based on these analysis results, engineers can develop more detailed emergency plans, covering everything from temporary reinforcement measures to emergency evacuation plans, to ensure a rapid response if the situation deteriorates. In addition, at the first-level warning stage, the system will notify all relevant parties, including project managers, on-site engineers, and safety supervisors, to ensure that everyone is aware of the current situation and is prepared accordingly.

[0053] When the bending moment of the composite beam is greater than the second threshold and the bending moment of the composite beam is less than the third threshold, the detection result is a second-level warning, and the operation of sending warning information to relevant personnel through real-time monitoring is executed. Specifically, when the bending moment value of the UHPC composite beam is greater than 5161.4 kN·m and less than 8094.6 kN·m, the system detection result is a second-level warning. This indicates that the structure has been subjected to a large load, close to its design limit, and there are significant risk signals, requiring more stringent monitoring and response measures. In the second-level warning state, the system will start the real-time monitoring mode, collect key parameters such as bending moment, strain, and curvature data at a higher frequency, and perform real-time analysis. This high-density data collection helps to capture any subtle changes and ensure that potential problems can be discovered in a timely manner. At the same time, the system will send warning information to all relevant personnel, including project managers, on-site engineers, and safety supervisors, to ensure that everyone involved can understand the current situation in a timely manner and be prepared to respond.

[0054] When the bending moment of the composite beam is greater than the third threshold, the detection result is a level 3 warning, and the operation of triggering an emergency response for maintenance is performed. Specifically, when the bending moment value of the UHPC composite beam exceeds 8094.6 kN·m, the system detection result is a level 3 warning, which indicates that the structure has approached or reached its ultimate bearing capacity and there is a serious risk. In this case, the emergency response mechanism is immediately triggered to ensure the safety of personnel and prevent further deterioration of structural damage. The system automatically performs a series of emergency maintenance operations. All ongoing construction activities will be immediately suspended and on-site personnel will be quickly evacuated to a safe area. At the same time, the pre-established emergency plan is activated, including but not limited to the implementation of temporary reinforcement measures, the setting of support structures, and the necessary local repair work to stabilize the structural state and avoid catastrophic damage.

[0055] This embodiment sets thresholds based on a multi-level warning mechanism, predicts the bending moment of the composite beam through a deep learning model and compares it with the threshold, thereby achieving graded warning, improving the accuracy and response speed of structural safety monitoring, and realizing early risk identification and intervention.

[0056] This application also provides a UHPC composite beam early warning device based on deep learning, please refer to Figure 7 , the device comprises: The acquisition module 10 is used to acquire the structural information of the ultra-high performance concrete composite beam.

[0057] The processing module 20 is used to input the construction information into the preset state prediction model for processing to obtain the target characteristic state and the target performance prediction results. The preset state prediction model is a hybrid neural network model. The preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer. The feature extraction network includes a first convolution block, a second convolution block and a third convolution block. The first convolution block, the second convolution block and the third convolution block all include two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer. The temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer. The output layer includes a characteristic state classification branch and a performance prediction branch. The characteristic state classification branch outputs the probability distribution of the characteristic state through a three-layer fully connected network. The performance prediction branch outputs the performance index value through a three-layer fully connected network. According to the hybrid neural network model, the feature extraction network is used to extract local features, the bidirectional long short-term memory network is used to capture temporal features, and the attention layer is used to weight key features to achieve deep learning of the characteristic state of the ultra-high performance concrete composite beam.

[0058] The detection module 30 is used to perform detection through a multi-level early warning mechanism based on the target feature state and the target performance prediction result to obtain the detection result.

[0059] The execution module 40 is used to execute corresponding warning operations according to the detection results.

[0060] The UHPC composite beam early warning device based on deep learning provided by the present application adopts the UHPC composite beam early warning method based on deep learning in the above-mentioned embodiment, which can solve the technical problem of how to obtain the state of the UHPC composite beam and perform the corresponding early warning operation. Compared with the prior art, the beneficial effects of the UHPC composite beam early warning device based on deep learning provided by the present application are the same as the beneficial effects of the UHPC composite beam early warning method based on deep learning provided by the above-mentioned embodiment, and the other technical features of the UHPC composite beam early warning device based on deep learning are the same as the features disclosed in the above-mentioned embodiment method, which will not be repeated here.

[0061] In one embodiment, the processing module 20 is further used to obtain construction information sample data and construct an initial state prediction model; and train the initial state prediction model based on the construction information sample data to obtain a preset state prediction model.

[0062] In one embodiment, the processing module 20 is also used to initialize the weights and bias parameters of the initial state prediction model; input the construction information sample data into the initial state prediction model for calculation to obtain the feature state classification results and the performance prediction results; calculate and weight the feature state classification results and the performance prediction results according to the loss function to obtain the total error value; calculate through the back propagation algorithm to obtain the gradient of the weight and bias parameters; iteratively update the weight and bias parameters through the optimization algorithm according to the gradient until the maximum number of iterations is reached or the total error value converges to the preset threshold, thereby obtaining the preset state prediction model.

[0063] In one embodiment, the processing module 20 is also used to input the construction information into a preset state prediction model for processing to obtain a characteristic state; perform an evaluation based on the characteristic state to obtain a confidence score; when the confidence score exceeds a preset threshold, output the characteristic state as a target characteristic state; when the confidence score does not exceed the preset threshold, issue a warning signal, and execute the step of inputting the construction information into the preset state prediction model for processing to obtain the characteristic state until the confidence score exceeds the preset threshold.

[0064] In one embodiment, the detection module 30 is also used to obtain a first threshold, a second threshold and a third threshold according to the multi-level warning mechanism, wherein the first threshold is less than the second threshold, and the second threshold is less than the third threshold; obtain the bending moment of the composite beam according to the target characteristic state and the target performance prediction results; compare the bending moment of the composite beam with the first threshold, the second threshold and the third threshold to obtain the detection result.

[0065] In one embodiment, the execution module 40 is also used to, when the combined beam bending moment is less than the first threshold, obtain a detection result of no risk, and perform normal monitoring and record key parameters; when the combined beam bending moment is greater than the first threshold and the combined beam bending moment is less than the second threshold, obtain a detection result of a first-level warning, and perform an operation of increasing the frequency of obtaining structural information and preparing an emergency plan; when the combined beam bending moment is greater than the second threshold and the combined beam bending moment is less than the third threshold, obtain a detection result of a second-level warning, and perform an operation of sending warning information to relevant personnel through real-time monitoring; when the combined beam bending moment is greater than the third threshold, obtain a detection result of a third-level warning, and perform an operation of triggering an emergency response for maintenance.

[0066] The present application provides a UHPC composite beam early warning device based on deep learning, and the UHPC composite beam early warning device based on deep learning includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the UHPC composite beam early warning method based on deep learning in the above-mentioned embodiment 1.

[0067] Reference below Figure 8 , which shows a schematic diagram of the structure of a UHPC composite beam warning device based on deep learning suitable for implementing the embodiment of the present application. The UHPC composite beam warning device based on deep learning in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The deep learning-based UHPC composite beam early warning device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0068] like Figure 8The UHPC composite beam early warning device based on deep learning may include a processing device 1001 (such as a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a ROM (Read Only Memory) 1002 or a program loaded from a storage device 1003 to a RAM (Random Access Memory) 1004. Various programs and data required for the operation of the UHPC composite beam early warning device based on deep learning are also stored in the RAM 1004. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 1003 including, for example, a tape, a hard disk, etc.; and communication devices 1009. The communication device 1009 can allow the UHPC composite beam early warning device based on deep learning to communicate with other devices wirelessly or by wire to exchange data. Although the UHPC composite beam early warning device based on deep learning with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems may be implemented or have alternatively.

[0069] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method described in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0070] The UHPC composite beam early warning device based on deep learning provided by the present application adopts the UHPC composite beam early warning method based on deep learning in the above-mentioned embodiment, which can solve the technical problem of how to obtain the state of the UHPC composite beam and perform the corresponding early warning operation. Compared with the prior art, the beneficial effects of the UHPC composite beam early warning device based on deep learning provided by the present application are the same as the beneficial effects of the UHPC composite beam early warning method based on deep learning provided by the above-mentioned embodiment, and the other technical features of the UHPC composite beam early warning device based on deep learning are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0071] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0072] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0073] The present application provides a computer-readable medium having computer-readable program instructions (i.e., computer programs) stored thereon for calculation, and the obtained computer-readable program instructions are used to execute the UHPC composite beam early warning method based on deep learning in the above-mentioned embodiment.

[0074] The computer-readable medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the calculation is performed to obtain a machine-readable medium that may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0075] The above-mentioned computer-readable medium may be included in the UHPC composite beam early warning device based on deep learning; or it may exist independently without being assembled into the UHPC composite beam early warning device based on deep learning.

[0076] The computer-readable medium may carry one or more programs. When the one or more programs are executed by the UHPC composite beam early warning device based on deep learning, the UHPC composite beam early warning device based on deep learning can be written in one or more programming languages ​​or a combination thereof to perform computer program codes for the operation of the present application. The programming languages ​​include object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0077] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0078] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0079] The readable medium provided in the present application is a computer-readable medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned UHPC composite beam early warning method based on deep learning, and can solve the technical problem of how to obtain the state of the UHPC composite beam and perform corresponding early warning operations. Compared with the prior art, the beneficial effects of the computer-readable medium provided in the present application are the same as the beneficial effects of the UHPC composite beam early warning method based on deep learning provided in the above-mentioned embodiment, and will not be repeated here.

[0080] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the UHPC composite beam early warning method based on deep learning as described above.

[0081] The computer program product provided in this application can solve the technical problem of how to obtain the state of the UHPC composite beam and perform the corresponding early warning operation. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the UHPC composite beam early warning method based on deep learning provided in the above embodiment, which will not be repeated here.

[0082] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A UHPC composite beam early warning method based on deep learning, characterized in that: The method comprises: Obtain the structural information of ultra-high performance concrete composite beams; Input the construction information into a preset state prediction model for processing to obtain target characteristic state and target performance prediction results, wherein the preset state prediction model is a hybrid neural network model, and the preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer, wherein the feature extraction network includes a first convolution block, a second convolution block and a third convolution block, wherein the first convolution block, the second convolution block and the third convolution block each include two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer, wherein the temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer, wherein the output layer includes a characteristic state classification branch and a performance prediction branch, wherein the characteristic state classification branch outputs the probability distribution of the characteristic state through a three-layer fully connected network, and the performance prediction branch outputs the performance index value through a three-layer fully connected network, wherein the hybrid neural network model performs local feature extraction through the feature extraction network, captures temporal features through the bidirectional long short-term memory network and weights the key features through the attention layer to achieve deep learning of the characteristic state of the ultra-high performance concrete composite beam; Based on the target characteristic state and the target performance prediction result, detection is performed through a multi-level early warning mechanism to obtain a detection result; Execute corresponding warning operations according to the detection results.

2. The method according to claim 1, characterized in that Before the step of inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, the method includes: Obtain sample data of structural information and build an initial state prediction model; The initial state prediction model is trained based on the construction information sample data to obtain a preset state prediction model.

3. The method according to claim 2, characterized in that The step of training the initial state prediction model based on the construction information sample data to obtain a preset state prediction model includes: Initializing weight and bias parameters of the initial state prediction model; Inputting the structural information sample data into the initial state prediction model for calculation to obtain characteristic state classification results and performance prediction results; Calculate and weight the feature state classification result and the performance prediction result according to the loss function to obtain a total error value; The gradients of the weight and bias parameters are obtained by back propagation algorithm calculation; The weight and bias parameters are iteratively updated through an optimization algorithm according to the gradient until a maximum number of iterations is reached or the total error value converges to a preset threshold, thereby obtaining a preset state prediction model.

4. The method according to claim 1, characterized in that Before the step of inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, the method includes: Inputting the structural information into a preset state prediction model for processing to obtain a characteristic state; Performing an evaluation based on the feature status to obtain a confidence score; When the confidence score exceeds a preset threshold, the feature state is output as a target feature state; When the confidence score does not exceed the preset threshold, an early warning signal is issued, and the step of inputting the construction information into a preset state prediction model for processing to obtain a characteristic state is executed until the confidence score exceeds the preset threshold.

5. The method according to claim 1, characterized in that The step of performing detection through a multi-level early warning mechanism based on the target feature state and the target performance prediction result to obtain the detection result comprises: According to the multi-level warning mechanism, a first threshold, a second threshold and a third threshold are obtained, wherein the first threshold is less than the second threshold, and the second threshold is less than the third threshold; Obtaining a composite beam bending moment according to the target characteristic state and the target performance prediction result; The composite beam bending moment is compared with the first threshold, the second threshold and the third threshold to obtain a detection result.

6. The method according to claim 5, characterized in that After the step of comparing the composite beam bending moment with the first threshold, the second threshold and the third threshold to obtain a detection result, the method further comprises: When the bending moment of the composite beam is less than the first threshold value, the detection result is obtained as no risk, and normal monitoring and recording of key parameters are performed; When the combined beam bending moment is greater than the first threshold and the combined beam bending moment is less than the second threshold, the detection result is obtained as a first-level warning, and the operation of increasing the frequency of obtaining structural information and preparing an emergency plan is performed; When the combined beam bending moment is greater than the second threshold and the combined beam bending moment is less than the third threshold, the detection result is obtained as a second-level warning, and an operation of sending warning information to relevant personnel through real-time monitoring is performed; When the bending moment of the composite beam is greater than the third threshold, the detection result is a third-level warning, and an operation of triggering an emergency response for maintenance is executed.

7. A UHPC composite beam early warning device based on deep learning, characterized in that: The device comprises: An acquisition module, used to acquire the structural information of the ultra-high performance concrete composite beam; A processing module, used for inputting the construction information into a preset state prediction model for processing to obtain a target characteristic state and a target performance prediction result, wherein the preset state prediction model is a hybrid neural network model, and the preset state prediction model includes a feature extraction network, a temporal feature network, an attention layer and an output layer, wherein the feature extraction network includes a first convolution block, a second convolution block and a third convolution block, wherein the first convolution block, the second convolution block and the third convolution block each include two one-dimensional convolution layers, a normalization layer, an activation function layer and a maximum pooling layer, wherein the temporal feature network includes a bidirectional long short-term memory network layer and a normalization layer, wherein the output layer includes a characteristic state classification branch and a performance prediction branch, wherein the characteristic state classification branch outputs a probability distribution of the characteristic state through a three-layer fully connected network, and the performance prediction branch outputs a performance index value through a three-layer fully connected network, wherein the hybrid neural network model performs local feature extraction through the feature extraction network, captures temporal features through the bidirectional long short-term memory network and weights key features through the attention layer to achieve deep learning of the characteristic state of the ultra-high performance concrete composite beam; A detection module, used to perform detection through a multi-level early warning mechanism based on the target feature state and the target performance prediction result to obtain a detection result; An execution module is used to execute corresponding warning operations according to the detection results.

8. A UHPC composite beam early warning device based on deep learning, characterized in that: The device comprises: a memory, a processor, and a deep learning-based UHPC composite beam early warning program stored in the memory and running on the processor, wherein the deep learning-based UHPC composite beam early warning program is configured to implement the steps of the deep learning-based UHPC composite beam early warning method as described in any one of claims 1 to 6.

9. A medium, characterized in that The medium stores a UHPC composite beam early warning program based on deep learning, and when the UHPC composite beam early warning program based on deep learning is executed by the processor, the steps of the UHPC composite beam early warning method based on deep learning as described in any one of claims 1 to 6 are implemented.

Images