Dam construction period displacement monitoring and safety early warning system and method based on physical information deep learning

Abstract

The invention relates to the technical field of water conservancy projects, in particular to a dam construction period displacement monitoring and safety early warning system and method based on physical information deep learning, and the method comprises the following steps: S1, data preprocessing: carrying out the self-adaptive wavelet denoising and auto-encoder reconstruction of original displacement data, and completing the standardization processing; s2, hybrid model displacement prediction: after training a PINN-LSTM / GRU model through historical monitoring data, predicting future displacement, and using NeuralProphet to correct a prediction residual error; and S3, multi-factor dynamic safety evaluation and early warning: adjusting threshold parameters and weight coefficients in combination with a construction stage, dynamically evaluating a safety risk level, and issuing the safety risk level through a graded early warning signal. According to the invention, high-precision displacement prediction and multi-factor dynamic safety early warning are realized through fusion of physical information and deep learning, and the monitoring reliability and early warning response capability of the dam construction period are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy engineering, and in particular to a dam construction period displacement monitoring and safety early warning system and method based on deep learning of physical information. Background Art

[0002] With the rapid advancement of large-scale water conservancy projects, the structural safety of dams during the construction period has received increasing attention. The construction period is the most complex stage in the entire life cycle of a dam. Its displacement evolution process is affected by the coupling of multiple factors and has obvious nonlinear, time-varying and multi-field coupling characteristics. Therefore, conducting research on high-precision and intelligent displacement monitoring and safety early warning technologies has become an important direction to ensure project quality and operational safety.

[0003] Existing monitoring methods still have many shortcomings in dealing with the complex characteristics of the construction period. First, the traditional wavelet denoising method uses a fixed threshold, which is difficult to adapt to the dynamically changing noise environment, resulting in unstable denoising effect and affecting data reliability. Second, the current mainstream prediction models are mostly data-driven and lack physical mechanism support, which is prone to physically unreasonable prediction results and cannot meet the high-precision requirements of safety warnings during the construction period. In addition, the safety assessment method is relatively static, judging risks based only on a single displacement indicator, and failing to fully integrate key influencing factors such as temperature and load, resulting in easy deviations in warning results and affecting the accuracy of risk identification. Summary of the Invention

[0004] The present invention provides a dam construction period displacement monitoring and safety early warning system and method based on deep learning of physical information, so as to improve the monitoring accuracy, reliability and intelligence level and ensure the safety of dam construction.

[0005] The method for displacement monitoring and safety early warning during dam construction period based on physical information deep learning includes the following steps:

[0006] S1, data preprocessing: adaptive wavelet denoising and autoencoder reconstruction are performed on the original displacement data in combination with the construction stage, and standardization is completed;

[0007] S2, hybrid model displacement prediction: Build a PINN-LSTM / GRU model, embed the physical equation constraints of static equilibrium, thermoelastic deformation, and creep into the loss function, train the PINN-LSTM / GRU model with historical monitoring data, and predict future displacements. Use NeuralProphet to correct the prediction residuals.

[0008] S3, multi-factor dynamic safety assessment and early warning: Construct a risk assessment function based on displacement, temperature and load, adjust the threshold parameters and weight coefficients in combination with the construction stage, dynamically assess the safety risk level, and issue it through graded early warning signals.

[0009] Optionally, the data preprocessing in S1 includes:

[0010] S11, dynamic wavelet denoising: Determine the current construction stage based on the construction log, adjust the threshold coefficient of wavelet denoising and select the wavelet basis function based on the construction stage, perform wavelet decomposition on the original displacement data, and perform soft threshold processing on the high-frequency coefficients obtained by the wavelet decomposition;

[0011] S12, autoencoder low-frequency reconstruction: Use convolutional autoencoder to perform nonlinear reconstruction on the low-frequency approximate coefficients obtained after wavelet decomposition. The convolutional autoencoder includes an encoder and a decoder.

[0012] S13, data normalization: performing Z-score normalization on the denoised original displacement data.

[0013] Optionally, the dynamic wavelet denoising in S11 includes:

[0014] S111, construction stage identification: automatically identify the current construction stage based on the construction log;

[0015] S112, wavelet basis function selection: select the wavelet basis function according to the characteristics of the construction stage;

[0016] S113, dynamic threshold calculation: Dynamically calculate the threshold coefficient based on the construction stage and noise level, expressed as:

[0017] θ=b phase ·(1+0.5·η);

[0018] Among them, θ is the threshold coefficient, b phase is the benchmark coefficient of the corresponding stage, η is the noise level;

[0019] S114, high frequency coefficient processing: performing soft threshold processing on the high frequency coefficients obtained by wavelet decomposition, the threshold is expressed as:

[0020]

[0021] Among them, τ is the threshold, α phase is the threshold coefficient related to the construction stage, σ is the estimated value of the noise standard deviation, and N is the signal length;

[0022] S115, wavelet reconstruction: reconstructing the processed wavelet coefficients to obtain displacement data after preliminary denoising.

[0023] Optionally, the hybrid model displacement prediction in S2 includes:

[0024] S21, Build a PINN-LSTM / GRU model: Build a hybrid model that combines a physical information neural network (PINN) with a long short-term memory network (LSTM) or a gated recurrent unit network (GRU). The long short-term memory network (LSTM) or the gated recurrent unit network (GRU) is responsible for extracting temporal features from time series data, and the physical information neural network (PINN) ensures the physical rationality of the prediction results by embedding physical equation constraints.

[0025] S22, embedding physical equation constraints: The static equilibrium equations, thermoelastic deformation equations, and concrete creep equations describing the mechanical behavior of the dam structure are converted into residual terms and embedded into the loss function of the LSTM / GRU model;

[0026] S23, Model Training and Prediction: Use historical monitoring data to train the PINN-LSTM / GRU model. During the training process, use the backpropagation algorithm and optimizer to iteratively update the model parameters to minimize the loss function. After training is complete, use the trained PINN-LSTM / GRU model to predict future displacement values.

[0027] S24, residual correction: For the residuals of the preliminary forecast results, the NeuralProphet model is used to model and correct the trend term and cycle term.

[0028] Optionally, the embedded physical equation constraint in S22 includes:

[0029] Static equilibrium equation constraints:

[0030] Where σ is the stress tensor and F is the body force;

[0031] Thermoelastic deformation equation constraint: ε = ε elastic +αΔT;

[0032] Where ε is the total strain, ε elastic is the elastic strain, α is the thermal expansion coefficient, and ΔT is the temperature change;

[0033] Concrete creep equation constraints:

[0034] Among them, ε creep is the creep strain, σ0 is the initial stress, E0 is the elastic modulus, and φ(t,t0) is the creep coefficient.

[0035] Optionally, the loss function is expressed as:

[0036]

[0037] Among them, y true is the true value, ypred is the model prediction value, is the stress divergence residual calculated from the predicted displacements.

[0038] Optionally, the multi-factor dynamic security assessment and warning in S3 includes:

[0039] S31, construct risk assessment function: construct a risk assessment function that comprehensively considers the multi-factor coupling effects of displacement, temperature, and load;

[0040] S32, establish a dynamic threshold matrix: dynamically adjust the weight coefficients and threshold parameters in the evaluation function according to the changes in the construction stage, and construct a stage-adaptive risk assessment matrix to reflect the sensitivity differences of safety risks in different stages;

[0041] S33, safety risk assessment and graded warning: Combine real-time monitoring data with a dynamic threshold matrix to calculate the current risk assessment value, generate a safety risk level, and issue graded warnings through lights, text messages, and apps to achieve proactive safety response.

[0042] Optionally, the risk assessment function is expressed as:

[0043]

[0044] Among them, u is the current displacement value, u th is the displacement threshold, T is the current temperature value, T0 is the reference temperature, ΔT is the temperature change threshold, F is the current load value, F th is the load threshold, w1, w2, w3 are weight coefficients.

[0045] Optionally, the security risk level classification includes:

[0046] Blue warning: When R<0.5, the security risk level is blue warning;

[0047] Yellow warning: When 0.5≤R<0.7, the safety risk level is yellow warning;

[0048] Orange warning: When 0.7≤R<0.9, the safety risk level is orange warning;

[0049] Red alert: When R ≥ 0.9, the safety risk level is red alert.

[0050] The dam construction period displacement monitoring and safety early warning system based on deep learning of physical information is used to implement the above-mentioned dam construction period displacement monitoring and safety early warning method based on deep learning of physical information, and includes the following modules:

[0051] Data preprocessing module: Adaptive wavelet denoising and autoencoder reconstruction are performed on the original displacement data in combination with the construction stage, and standardization is completed;

[0052] Hybrid model displacement prediction module: This module builds a PINN-LSTM / GRU hybrid model, embeds static equilibrium, thermoelastic deformation, and creep equation constraints into the loss function, completes model training and predicts future displacements using historical monitoring data, and uses the NeuralProphet model to correct prediction residuals.

[0053] Multi-factor dynamic safety assessment and early warning module: Constructs a risk assessment function that integrates displacement, temperature, and load factors, dynamically adjusts threshold parameters and weight coefficients according to the construction stage, and outputs corresponding safety risk levels and graded early warning signals.

[0054] Beneficial effects of the present invention:

[0055] The present invention combines wavelet analysis with convolutional autoencoders to achieve adaptive multi-scale noise suppression during the construction phase, effectively improving the quality and stability of the original displacement monitoring data. By introducing the PINN-LSTM / GRU hybrid model, physical constraints such as static equilibrium, thermoelastic deformation, and concrete creep are embedded in the deep learning framework, significantly enhancing the physical consistency and engineering interpretability of the displacement prediction results. At the same time, the NeuralProphet model is combined with periodic correction of the residuals to further improve the accuracy and timeliness of the prediction, meeting the dynamic prediction needs under complex construction stages.

[0056] The present invention realizes adaptive safety level identification and multi-level early warning issuance mechanism by constructing a risk assessment function that integrates multiple factors such as displacement, temperature, and load, and dynamically adjusting thresholds and weight parameters according to the construction stage. This method not only breaks through the limitations of traditional methods that rely on static thresholds and single indicators, significantly improves the accuracy and responsiveness of safety assessments, but also provides real-time, visual intelligent early warning means for construction sites, and has good engineering practical value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0058] Figure 1 This is a flowchart of an early warning method according to an embodiment of the present invention;

[0059] Figure 2Schematic diagram of system function modules according to an embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0061] It should be noted that references in the specification to "one embodiment," "an embodiment," "an exemplary embodiment," "some embodiments," etc. indicate that the described embodiments may include specific features, structures, or characteristics, but not every embodiment necessarily includes such specific features, structures, or characteristics. In addition, when specific features, structures, or characteristics are described in conjunction with an embodiment, it is within the knowledge of persons skilled in the relevant art to implement such features, structures, or characteristics in conjunction with other embodiments (whether or not explicitly described).

[0062] In general, terms can be understood, at least in part, from their use in context. For example, depending at least in part on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending at least in part on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0063] like Figure 1 As shown in FIG, the displacement monitoring and safety early warning method for dam construction period based on deep learning of physical information includes the following steps:

[0064] S1, data preprocessing:

[0065] S1.1, dynamic wavelet denoising: Determine the current construction stage based on the construction log, and adaptively adjust the wavelet denoising threshold coefficient and select the wavelet basis function based on the construction stage. Perform wavelet decomposition on the original displacement data and perform soft threshold processing on the high-frequency coefficients obtained from the decomposition to reduce the impact of high-frequency noise. The implementation steps are as follows:

[0066] (1) Construction stage identification: The current construction stage is automatically identified based on the construction log, which serves as the basis for dynamically adjusting the wavelet denoising parameters.

[0067] (2) Wavelet basis function selection: Select appropriate wavelet basis functions according to the characteristics of the construction stage.

[0068] (3) Dynamic threshold calculation: The threshold coefficient is dynamically calculated based on the construction stage and noise level. The specific implementation is as follows:

[0069] θ=b phase ·(1+0.5·η);

[0070] Among them, θ is the threshold coefficient, b phase is the benchmark coefficient of the corresponding stage, η is the noise level;

[0071] (4) High-frequency coefficient processing: Soft threshold processing is performed on the high-frequency coefficients obtained by wavelet decomposition. The threshold τ calculation formula is:

[0072]

[0073] Among them, α phase is the threshold coefficient related to the construction stage, σ is the estimated value of the noise standard deviation, and N is the signal length.

[0074] (5) Wavelet reconstruction: reconstruct the processed wavelet coefficients to obtain the displacement data after preliminary denoising.

[0075] S1.2, Autoencoder Low-Frequency Reconstruction: Use convolutional autoencoder to perform nonlinear reconstruction on the low-frequency approximate coefficients obtained after wavelet decomposition to further eliminate low-frequency interference signals and improve data quality. The specific implementation is as follows:

[0076] (1) Autoencoder architecture: It adopts a convolutional autoencoder structure, which consists of two parts: encoder and decoder.

[0077] (2) Low-frequency signal reconstruction: The low-frequency approximate coefficients obtained by wavelet decomposition are input into the trained autoencoder. Through the nonlinear mapping capability of the autoencoder, the low-frequency interference signal is further eliminated to obtain the reconstructed low-frequency signal.

[0078] S1.3, Data Standardization: Perform Z-score standardization on the denoised displacement data to eliminate dimension and value range differences, accelerate model training and improve model performance. The calculation formula is:

[0079]

[0080] Among them, μ is the data mean and σ is the data standard deviation.

[0081] S2, hybrid model displacement prediction:

[0082] S2.1, Build a PINN-LSTM / GRU Model: Build a hybrid model that combines a physical information neural network (PINN) with a long short-term memory network (LSTM) or a gated recurrent unit network (GRU). The LSTM / GRU network extracts temporal features from time series data, while the PINN model ensures the physical plausibility of the prediction results by embedding physical equation constraints.

[0083] S2.2, Embedding Physical Equation Constraints: The physical equations describing the mechanical behavior of the dam structure, such as the static equilibrium equation, the thermoelastic deformation equation, and the concrete creep equation, are converted into residual terms and embedded into the loss function of the LSTM / GRU model. The design of the loss function must consider both data-driven prediction accuracy and the degree of satisfaction of physical constraints, including:

[0084] (1) Static equilibrium equation constraints:

[0085]

[0086] Where σ is the stress tensor and F is the body force.

[0087] (2) Thermoelastic deformation equation constraints:

[0088] ε=ε elastic +αΔT (4)

[0089] Where ε is the total strain, ε elastic is the elastic strain, α is the thermal expansion coefficient, and ΔT is the temperature change.

[0090] (3) Concrete creep equation constraints:

[0091]

[0092] Among them, ε creep is the creep strain, σ0 is the initial stress, E0 is the elastic modulus, and φ(t,t0) is the creep coefficient.

[0093] The loss function designed in this invention comprehensively considers the prediction accuracy of data-driven and the satisfaction degree of physical constraints, and is specifically implemented as follows:

[0094]

[0095] Among them, y true is the true value, y pred is the model prediction value, is the stress divergence residual calculated from the predicted displacement, and 0.7 and 0.3 are weight coefficients that control the balance between data fitting and physical constraints.

[0096] S2.3, Model Training and Prediction: A PINN-LSTM / GRU hybrid model is trained using historical monitoring data. During training, the model parameters are iteratively updated using a backpropagation algorithm and optimizer to minimize the loss function, enabling the model to simultaneously fit the monitoring data and satisfy the physical equation constraints. After training is complete, the trained model is used to predict displacement values ​​for a period of time in the future.

[0097] S2.4, NeuralProphet residual correction: In order to further improve the prediction accuracy and separate the periodic and trend components in the prediction results, the NeuralProphet model is used to model and correct the prediction residual of the PINN-LSTM / GRU model to obtain the final displacement prediction result.

[0098] S3, multi-factor dynamic security assessment and early warning:

[0099] S3.1, Constructing a risk assessment function: Construct a risk assessment function that comprehensively considers the coupling effects of multiple factors such as displacement, temperature, and load. The design of the risk assessment function must be able to reflect the different degrees of impact of different factors on dam safety, expressed as:

[0100]

[0101] in:

[0102] u is the current displacement value, u th is the displacement threshold.

[0103] T is the current temperature value, T0 is the reference temperature, and ΔT is the temperature change threshold.

[0104] F is the current load value, F th is the load threshold.

[0105] w1, w2, and w3 are weight coefficients, which are dynamically adjusted according to the construction stage.

[0106] S3.2, Establish a dynamic threshold matrix: Dynamically adjust the weight coefficients and threshold parameters in the risk assessment function according to the construction stage to establish an adaptive risk assessment matrix for the construction stage. Different thresholds and weights correspond to different construction stages to adapt to the safety risk characteristics of different stages, as shown in the following table:

[0107] Table 1

[0108]

[0109] Different construction stages correspond to different thresholds and weights to adapt to the safety risk characteristics of different stages.

[0110] S3.3, Safety Risk Assessment and Graded Warning: Based on real-time monitoring data and a dynamic threshold matrix, the risk assessment function is calculated to assess the current safety risk level of the dam. Warning levels are set based on the risk level, and corresponding graded warning signals are output, such as blue warning, yellow warning, orange warning, and red warning. Warning signals can be issued in various forms, including lighting, text messages, and app push notifications. Specifically,

[0111] (1) Risk level classification:

[0112] R<0.5: safe state (blue warning)

[0113] 0.5≤R<0.7: Attention status (yellow warning)

[0114] 0.7≤R<0.9: Alert state (orange warning)

[0115] R≥0.9: Dangerous state (red alert)

[0116] (2) Early warning signal issuance:

[0117] Early warning signals can be issued through various forms such as lights, text messages, and App push, and different levels of warnings correspond to different handling measures and response processes.

[0118] like Figure 2 As shown, the dam construction period displacement monitoring and safety early warning system based on physical information deep learning is used to implement the above-mentioned dam construction period displacement monitoring and safety early warning method based on physical information deep learning, and includes the following modules:

[0119] Data preprocessing module: Adaptive wavelet denoising and autoencoder reconstruction are performed on the original displacement data in combination with the construction stage, and standardization is completed;

[0120] Hybrid model displacement prediction module: This module builds a PINN-LSTM / GRU hybrid model, embeds static equilibrium, thermoelastic deformation, and creep equation constraints into the loss function, completes model training and predicts future displacements using historical monitoring data, and uses the NeuralProphet model to correct prediction residuals.

[0121] Multi-factor dynamic safety assessment and early warning module: Constructs a risk assessment function that integrates displacement, temperature, and load factors, dynamically adjusts threshold parameters and weight coefficients according to the construction stage, and outputs corresponding safety risk levels and graded early warning signals.

[0122] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0123] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A method for displacement monitoring and safety early warning during dam construction based on deep learning of physical information, characterized by: The following steps are involved: S1, data preprocessing: adaptive wavelet denoising and autoencoder reconstruction are performed on the original displacement data in combination with the construction stage, and standardization is completed; S2, hybrid model displacement prediction: Build a PINN-LSTM / GRU model, embed the physical equation constraints of static equilibrium, thermoelastic deformation, and creep into the loss function, train the PINN-LSTM / GRU model with historical monitoring data, and predict future displacements. Use NeuralProphet to correct the prediction residuals. S3, multi-factor dynamic safety assessment and early warning: Construct a risk assessment function based on displacement, temperature and load, adjust the threshold parameters and weight coefficients in combination with the construction stage, dynamically assess the safety risk level, and issue it through graded early warning signals.

2. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 1 is characterized in that: The data preprocessing in S1 includes: S11, dynamic wavelet denoising: Determine the current construction stage based on the construction log, adjust the threshold coefficient of wavelet denoising and select the wavelet basis function based on the construction stage, perform wavelet decomposition on the original displacement data, and perform soft threshold processing on the high-frequency coefficients obtained by the wavelet decomposition; S12, autoencoder low-frequency reconstruction: Use convolutional autoencoder to perform nonlinear reconstruction on the low-frequency approximate coefficients obtained after wavelet decomposition. The convolutional autoencoder includes an encoder and a decoder. S13, data normalization: performing Z-score normalization on the denoised original displacement data.

3. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 2 is characterized in that: The dynamic wavelet denoising in S11 includes: S111, construction stage identification: automatically identify the current construction stage based on the construction log; S112, wavelet basis function selection: select the wavelet basis function according to the characteristics of the construction stage; S113, dynamic threshold calculation: Dynamically calculate the threshold coefficient based on the construction stage and noise level, expressed as: θ=b phase ·(1+0.5·h); Among them, θ is the threshold coefficient, b phase is the benchmark coefficient of the corresponding stage, η is the noise level; S114, high frequency coefficient processing: performing soft threshold processing on the high frequency coefficients obtained by wavelet decomposition, the threshold is expressed as: Among them, τ is the threshold, α phase is the threshold coefficient related to the construction stage, σ is the estimated value of the noise standard deviation, and N is the signal length; S115, wavelet reconstruction: reconstructing the processed wavelet coefficients to obtain displacement data after preliminary denoising.

4. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 3 is characterized in that: The hybrid model displacement prediction in S2 includes: S21, Build a PINN-LSTM / GRU model: Build a hybrid model that combines a physical information neural network with a long short-term memory network or a gated recurrent unit network. The long short-term memory network or the gated recurrent unit network is responsible for extracting temporal features from time series data, and the physical information neural network ensures the physical rationality of the prediction results by embedding physical equation constraints. S22, embedding physical equation constraints: The static equilibrium equations, thermoelastic deformation equations, and concrete creep equations describing the mechanical behavior of the dam structure are converted into residual terms and embedded into the loss function of the LSTM / GRU model; S23, Model Training and Prediction: Use historical monitoring data to train the PINN-LSTM / GRU model. During the training process, use the backpropagation algorithm and optimizer to iteratively update the model parameters to minimize the loss function. After training is complete, use the trained PINN-LSTM / GRU model to predict future displacement values. S24, residual correction: For the residuals of the preliminary forecast results, the NeuralProphet model is used to model and correct the trend term and cycle term.

5. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 4 is characterized in that: The embedded physical equation constraints in S22 include: Static equilibrium equation constraints: Where σ is the stress tensor and F is the body force; Thermoelastic deformation equation constraint: ε = ε elastic +αΔT; Where ε is the total strain, ε elastic is the elastic strain, α is the thermal expansion coefficient, and ΔT is the temperature change; Concrete creep equation constraints: Among them, ε creep is the creep strain, σ0 is the initial stress, E0 is the elastic modulus, and φ(t,t0) is the creep coefficient.

6. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 5 is characterized in that: The loss function is expressed as: Among them, y true is the true value, y pred is the model prediction value, is the stress divergence residual calculated from the predicted displacements.

7. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 6 is characterized in that: The multi-factor dynamic security assessment and early warning in S3 include: S31, construct risk assessment function: construct a risk assessment function that comprehensively considers the multi-factor coupling effects of displacement, temperature, and load; S32, establish a dynamic threshold matrix: dynamically adjust the weight coefficients and threshold parameters in the evaluation function according to the changes in the construction stage, and construct a stage-adaptive risk assessment matrix to reflect the sensitivity differences of safety risks in different stages; S33, safety risk assessment and graded warning: Combine real-time monitoring data with a dynamic threshold matrix to calculate the current risk assessment value, generate a safety risk level, and issue graded warnings through lights, text messages, and apps to achieve proactive safety response.

8. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 7 is characterized in that: The risk assessment function is expressed as: Among them, u is the current displacement value, u th is the displacement threshold, T is the current temperature value, T0 is the reference temperature, ΔT is the temperature change threshold, F is the current load value, F th is the load threshold, w1, w2, w3 are weight coefficients.

9. The method for displacement monitoring and safety early warning during dam construction period based on deep learning of physical information according to claim 8 is characterized in that: The security risk level classification includes: Blue warning: When R<0.5, the security risk level is blue warning; Yellow warning: When 0.5≤R<0.7, the safety risk level is yellow warning; Orange warning: When 0.7≤R<0.9, the safety risk level is orange warning; Red alert: When R ≥ 0.9, the safety risk level is red alert.

10. A dam construction period displacement monitoring and safety early warning system based on deep learning of physical information, used to implement the dam construction period displacement monitoring and safety early warning method based on deep learning of physical information as described in any one of claims 1 to 9, characterized in that: Includes the following modules: Data preprocessing module: Adaptive wavelet denoising and autoencoder reconstruction are performed on the original displacement data in combination with the construction stage, and standardization is completed; Hybrid model displacement prediction module: This module builds a PINN-LSTM / GRU hybrid model, embeds static equilibrium, thermoelastic deformation, and creep equation constraints into the loss function, completes model training and predicts future displacements using historical monitoring data, and uses the NeuralProphet model to correct prediction residuals. Multi-factor dynamic safety assessment and early warning module: Constructs a risk assessment function that integrates displacement, temperature, and load factors, dynamically adjusts threshold parameters and weight coefficients according to the construction stage, and outputs corresponding safety risk levels and graded early warning signals.

Images