Method for dynamic risk assessment of underground cavern construction period based on deformation prediction

Abstract

The application discloses a method for dynamically evaluating underground cavern construction period risk based on deformation prediction, belongs to the technical field of construction safety, and can solve the problem of insufficient evaluation accuracy. The method comprises the following steps: obtaining deformation time series data and construction time series data of the underground cavern at a plurality of historical time points, and determining a plurality of deformation influence factors of the underground cavern according to the deformation time series data and the construction time series data; obtaining factor time series data of each deformation influence factor at the plurality of historical time points, and constructing an influence factor feature set according to the construction time series data and the factor time series data; predicting the deformation amount of the underground cavern at a first time point by using a convolutional neural network and a long short-term memory network according to the influence factor feature set; determining a construction safety threshold value at the first time point according to the deformation time series data, the construction time series data and construction data at the first time point; and determining the construction risk at the first time point according to the deformation amount at the first time point and the construction safety threshold value. The application is used for evaluating construction risk.

Description

Technical Field

[0001] This invention relates to a dynamic risk assessment method for underground cavern construction based on deformation prediction, belonging to the field of construction safety technology. Background Technology

[0002] During the construction of underground caverns, the deformation of the surrounding rock is a core indicator reflecting the safety status of the cavern construction. Based on the deformation of the surrounding rock, construction risk assessment and early warning are important means to ensure the safety of underground cavern construction and prevent engineering disasters such as cavern collapse and lining cracking.

[0003] Existing construction risk assessment methods typically predict future construction risks based on the current deformation state of the surrounding rock. However, underground cavern construction involves numerous and overlapping procedures, resulting in various construction conditions. These different conditions cause varying degrees of disturbance to the surrounding rock, leading to changes in its deformation rate and amplitude. Current technologies do not consider the impact of different construction conditions on surrounding rock deformation during construction, and therefore cannot perform dynamic risk assessments as construction progresses and conditions change, resulting in insufficient accuracy and specificity in the assessments. Summary of the Invention

[0004] This invention provides a dynamic risk assessment method for underground cavern construction based on deformation prediction, which can solve the problem that existing technologies cannot combine construction progress information for dynamic risk assessment, resulting in insufficient assessment accuracy.

[0005] This invention provides a method for dynamic risk assessment during the construction period of underground caverns based on deformation prediction, the method comprising:

[0006] S1. Obtain deformation time-series data and construction time-series data of the underground cavern at multiple consecutive historical moments, and determine multiple deformation influencing factors of the underground cavern based on the deformation time-series data and the construction time-series data.

[0007] S2. Obtain the time series data of each deformation influencing factor at the multiple historical moments, and construct an influencing factor feature set based on the construction time series data and the factor time series data;

[0008] S3. Based on the influencing factor feature set and the construction data at the current moment, use convolutional neural networks and long short-term memory networks to predict the deformation of the underground cavern at the first moment.

[0009] S4. Determine the construction safety threshold at the first moment based on the deformation time series data, the construction time series data, and the construction data at the first moment.

[0010] S5. Determine the construction risk at the first moment based on the deformation amount and construction safety threshold at the first moment.

[0011] Optionally, in S1, multiple deformation influencing factors of the underground cavern are determined based on the deformation time series data and the construction time series data, specifically as follows:

[0012] Based on the construction timeline data, several candidate influencing factors affecting the deformation of the underground cavern were determined;

[0013] Obtain the candidate factor time series data for each candidate impact factor at the multiple historical times, and determine the multiple deformed impact factors based on the deformed time series data and the candidate factor time series data.

[0014] Optionally, the plurality of deformation influence factors are determined based on the deformation time series data and the candidate factor time series data, specifically as follows:

[0015] Based on the deformation time series data and the candidate factor time series data, determine the time series correlation degree between each candidate influence factor and the deformation amount of the underground cavern;

[0016] Multiple candidate influencing factors with a temporal correlation degree higher than a preset threshold are identified as the multiple deformable influencing factors.

[0017] Optionally, based on the deformation time series data and the candidate factor time series data, the time series correlation degree between each candidate influence factor and the deformation amount of the underground cavern is determined, specifically as follows:

[0018] Based on the deformation time series data and the candidate factor time series data, determine the instantaneous correlation between each candidate influence factor and the deformation amount of the underground cavern at each historical moment;

[0019] Calculate the average instantaneous correlation of each candidate influence factor at all historical moments, and use the average value as the temporal correlation between the corresponding candidate influence factor and the deformation of the underground cavern.

[0020] Optionally, in S2, an influencing factor feature set is constructed based on the construction time series data and the factor time series data, specifically as follows:

[0021] Principal component analysis was used to transform the time series data of all deformation-influencing factors into multiple principal component features.

[0022] The multiple principal component features are corrected based on the construction time series data to obtain the influencing factor feature set.

[0023] Optionally, the multiple principal component features are modified based on the construction time series data, specifically as follows:

[0024] The construction progress rate at each historical moment is determined based on the construction time sequence data, and historical moments with a construction progress rate greater than a first preset rate are defined as strong disturbance moments.

[0025] Increase the feature value of the principal component features corresponding to the time of the strong disturbance.

[0026] Optionally, S4 specifically refers to:

[0027] S41. Determine the time series data of deformation rate based on the deformation time series data, and determine the time series data of construction progress rate based on the construction time series data;

[0028] S42. Determine the mapping relationship between deformation rate and construction progress rate based on the time-series data of deformation rate and construction progress rate.

[0029] S43. Determine the construction safety threshold at the first moment based on the mapping relationship and the construction data at the first moment.

[0030] Optionally, S43 specifically refers to:

[0031] S431. Based on the mapping relationship, the construction advancement rate corresponding to the deformation rate that is less than the second preset rate is determined as the conventional advancement rate.

[0032] S432. Determine the construction stability coefficient at the first moment based on the conventional advance rate and the construction data at the first moment;

[0033] S433. Determine the benchmark safety threshold. The product of the benchmark safety threshold and the construction stability coefficient at the first moment is determined as the construction safety threshold at the first moment.

[0034] Optionally, S432 specifically refers to:

[0035] The construction progress rate at the first moment is determined based on the construction data at the first moment.

[0036] Determine whether the construction progress rate at the first moment conforms to the conventional progress rate. If yes, determine that the construction stability coefficient at the first moment is a first value; if no, determine that the construction stability coefficient at the first moment is a second value; the first value is greater than the second value.

[0037] Optionally, prior to S433, the method further includes:

[0038] The construction conditions at the first moment are determined based on the construction data at the first moment.

[0039] Determine whether the construction condition at the first moment is a preset normal condition. If yes, determine that the safety factor of the condition at the first moment is a third value. If no, determine that the safety factor of the condition at the first moment is a fourth value. The third value is greater than the fourth value.

[0040] Accordingly, S433 specifically refers to:

[0041] A baseline safety threshold is determined, and the product of the baseline safety threshold, the construction stability coefficient at the first moment, and the working condition safety coefficient at the first moment is determined as the construction safety threshold at the first moment.

[0042] The beneficial effects that this invention can produce include:

[0043] This invention fully considers the impact of different construction conditions on surrounding rock deformation during construction. It identifies multiple deformation influencing factors by combining historical construction time-series data, then constructs a feature set based on the time-series data of these factors. During feature set construction, the feature values ​​for periods of strong disturbance are increased. Subsequently, based on the feature set and the current construction data, a deep learning model is used to predict the predicted deformation at the first moment in the future. This reflects the impact of construction condition characteristics on the deformation of underground caverns during prediction, improving prediction accuracy. Simultaneously, this invention determines the construction stability coefficient and the working condition safety coefficient based on the construction data at the first moment, and adjusts the construction safety threshold based on these coefficients. This achieves dynamic adaptive adjustment of the risk assessment standard. It allows for dynamic assessment of construction risks based on the construction progress information at the first moment, improving the early warning sensitivity for highly disturbed and high-risk conditions to avoid missed reports, and reducing the early warning sensitivity for stable and low-risk conditions to avoid false alarms. Thus, it enables dynamic risk assessment as construction progresses and construction conditions change, ensuring the accuracy and reliability of the assessment. Attached Figure Description

[0044] Figure 1 A flowchart of a method for dynamic risk assessment during the construction period of underground caverns based on deformation prediction, provided in an embodiment of the present invention. Detailed Implementation

[0045] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] This invention provides a method for dynamic risk assessment during the construction period of underground caverns based on deformation prediction, such as... Figure 1 As shown, the method includes:

[0047] S1. Obtain deformation time-series data and construction time-series data of the underground cavern at multiple consecutive historical moments. Based on the deformation time-series data and construction time-series data, determine multiple deformation influencing factors of the underground cavern, specifically:

[0048] S11. Obtain deformation time-series data and construction time-series data of underground caverns at multiple consecutive historical moments;

[0049] In this embodiment, monitoring equipment is deployed on the arch, left and right sidewalls, or key sections of the surrounding rock convergence in the underground cavern. Deformation data, construction data, and foundation condition data are collected at a fixed sampling frequency. The sampling frequency can be set to 5 minutes / time, 10 minutes / time, 15 minutes / time, etc., and the monitoring process covers the excavation, support, lining, surrounding rock stabilization, and curing processes of the early-stage construction section.

[0050] The deformation data includes the deformation of the surrounding rock of the underground cavern at the current acquisition time relative to the previous acquisition time, and the deformation rate of the surrounding rock at each acquisition time; the construction data includes the process type, process duration, construction technology, construction parameters, and construction advancement rate (i.e., the face advancement rate) corresponding to each acquisition time; the basic condition data includes the ambient temperature, ambient humidity, groundwater seepage pressure, and surrounding rock geological conditions corresponding to each acquisition time.

[0051] Arranging deformation data corresponding to multiple consecutive historical moments in chronological order yields deformation time-series data for multiple consecutive historical moments; arranging construction data corresponding to multiple consecutive historical moments in chronological order yields construction time-series data for multiple consecutive historical moments; and arranging foundation condition data corresponding to multiple consecutive historical moments in chronological order yields foundation condition time-series data for multiple consecutive historical moments.

[0052] S12. Based on deformation time series data and construction time series data, determine multiple deformation influencing factors of the underground cavern, specifically:

[0053] 1) Determine multiple candidate influencing factors affecting the deformation of underground caverns based on construction timeline data;

[0054] Specifically, this embodiment first includes basic condition indicators such as ambient temperature, ambient humidity, groundwater seepage pressure, and surrounding rock geological conditions from the basic condition time series data into candidate influencing factors. Simultaneously, this embodiment fully considers the impact of different construction conditions on surrounding rock deformation, and also includes construction indicators such as process type, process duration, construction technology, construction parameters, and construction progress rate from the construction time series data into candidate influencing factors.

[0055] 2) Obtain the time series data of each candidate impact factor at multiple historical moments, and determine multiple deformed impact factors based on the deformed time series data and the candidate factor time series data;

[0056] First, historical data for each basic condition indicator at multiple historical moments are obtained from the basic condition time series data, and historical data for each construction indicator at multiple historical moments are obtained from the construction time series data. The historical data of each indicator are then arranged in chronological order to obtain the time series data for each candidate influencing factor, i.e., the candidate factor time series data.

[0057] Then, based on the deformation time series data and the candidate factor time series data, the instantaneous correlation between each candidate influencing factor and the deformation of the underground cavern at each historical moment is determined;

[0058] Specifically, this embodiment uses the grey relational analysis method to determine the instantaneous relational degree:

[0059] The first step is to construct the analysis sequence: use the deformed time series data as the reference sequence, and use the time series data of each candidate factor as an independent comparison sequence;

[0060] The second step is to perform dimensionless processing on the reference sequence and all comparison sequences to make the reference sequence comparable to each comparison sequence;

[0061] The third step is to calculate the instantaneous correlation degree using the grey relational analysis method. The calculation formula is as follows:

[0062] (1)

[0063] In equation (1), For the first The candidate impact factor in the th The instantaneous correlation between a historical moment and the deformation of an underground cavern represents the degree of instantaneous correlation between the candidate influencing factor and the deformation of the underground cavern at that historical moment. For the underground cavern in the first The deformation at a given historical moment, i.e., the reference sequence at the _ ... The value at a historical moment; For the first The candidate impact factor in the th Historical data for the _th historical moment, i.e. the _th The comparison sequence corresponding to the candidate impact factor is in the _ ... The value at a historical moment; For the first Deformation at historical moments and the first The absolute difference of historical data for each candidate impact factor; For all historical moments and all candidate impact factors The minimum value in the range, i.e., the global minimum value; For all historical moments and all candidate impact factors The maximum value in the range, i.e., the global maximum value; In this embodiment, the resolution coefficient is used. A fixed value of 0.5 is used to balance the distribution range of instantaneous correlation, compress the range of numerical differences, improve the distinguishability of instantaneous correlation of different candidate influencing factors, and avoid the problem of instantaneous correlation values ​​converging and being difficult to distinguish. The candidate impact factor number; This refers to the sequence number of a historical moment.

[0064] Subsequently, the average instantaneous correlation of each candidate influencing factor across all historical moments is calculated. This average is then used as the temporal correlation between the corresponding candidate influencing factor and the deformation of the underground cavern. The temporal correlation is used to characterize the comprehensive influence of the candidate influencing factor on the deformation over a time period comprised of multiple historical moments.

[0065] Finally, multiple candidate impact factors with a time-series correlation higher than a preset threshold are identified as multiple deformed impact factors. The preset threshold can be flexibly set according to actual conditions, such as 0.6, 0.7, 0.8, etc. When the time-series correlation of a candidate impact factor is lower than or equal to the preset threshold, it indicates that the overall influence of the candidate impact factor and the deformed quantity is low, and it should be eliminated.

[0066] S2. Obtain the time-series data of each deformation influencing factor at multiple historical moments, and construct an influencing factor feature set based on the construction time-series data and the factor time-series data, specifically as follows:

[0067] 1) Obtain historical data of each deformation influencing factor at multiple historical moments from the basic condition time series data and construction time series data, and then arrange them in chronological order to obtain the factor time series data of each deformation influencing factor;

[0068] 2) Use principal component analysis to transform the time series data of all deformed influencing factors into multiple principal component features to obtain the influencing factor feature set.

[0069] Furthermore, to reflect the impact of different construction conditions on deformation during subsequent prediction, this embodiment emphasizes the differences in construction conditions in this step. Specifically, after obtaining multiple principal component features, this step also includes:

[0070] 3) Based on the construction time series data, multiple principal component features are corrected to obtain the influencing factor feature set, including:

[0071] The construction progress rate at each historical moment is determined based on the construction time sequence data, and historical moments with a construction progress rate greater than the first preset rate are defined as strong disturbance moments.

[0072] Increase the eigenvalues ​​corresponding to the strong disturbance times among multiple principal component features to obtain the influence factor feature set.

[0073] Specifically, the first preset rate can be flexibly set according to the actual situation. Generally, the construction progress rate of processes such as excavation and blasting is greater than the first preset rate, resulting in stronger construction disturbance; the construction progress rate of processes such as lining, surrounding rock stabilization, and curing is less than the first preset rate, resulting in relatively smaller construction disturbance.

[0074] S3. Based on the influencing factor feature set and the current construction data, predict the deformation of the underground cavern at the first moment using a convolutional neural network and a long short-term memory network, specifically:

[0075] A deformation prediction model is constructed based on convolutional neural networks and long short-term memory networks.

[0076] The influencing factor feature set and the construction data at the current moment are input into the deformation prediction model, and the deformation prediction model is used to determine the predicted deformation of the underground cavern at the first moment.

[0077] Specifically, the first moment is a future moment. The deformation prediction model takes the feature set of influencing factors as input and predicts the deformation of the underground cavern at the first moment in the future based on the characteristics of the deformation influencing factors and the characteristics of historical construction progress, thus obtaining the predicted deformation.

[0078] Deformation prediction models based on convolutional neural networks and long short-term memory networks are existing technologies.

[0079] Among them, the Long Short-Term Memory (LSTM) network is responsible for mining the temporal evolution pattern of deformation under the linkage of construction progress. Its specific control logic is as follows: the forget gate adaptively filters and forgets invalid and outdated deformation features in the historical data based on the matching degree between the construction data at the current construction time and the influencing factor feature set at the historical time; the input gate accurately accepts the new correlation features of the construction progress, surrounding rock state, and environmental parameters at the current time; the nonlinear transformation and update of temporal features are completed by combining the candidate cell state; the long-term deformation trend is remembered and inherited through the iterative update of the cell state; finally, the output gate filters the effective temporal features and outputs refined temporal evolution features that fit the actual evolution pattern of construction, accurately depicting the long-term trend and short-term abrupt fluctuation characteristics of surrounding rock deformation.

[0080] The Convolutional Neural Network (CNN) is configured with three layers of multi-scale parallel convolutional kernels, with kernel sizes set to 3×3, 5×5, and 7×7, a stride of 1, and a "same" padding method to ensure that the output feature size is consistent with the input. The small-sized convolutional kernel is responsible for extracting the subtle deformation fluctuation features of local monitoring points in the cavern, the medium-sized convolutional kernel is responsible for extracting the collaborative deformation correlation features of adjacent monitoring sections and local areas, and the large-sized convolutional kernel is responsible for extracting the macroscopic deformation trend features of the overall surrounding rock of the cavern. After all convolution operations are completed, the ReLU activation function is used to achieve nonlinear feature mapping, and finally, the multi-scale spatial feature fusion is completed by channel splicing to comprehensively cover the spatial deformation patterns of the cavern at different scales.

[0081] The deformation prediction model can align the one-dimensional temporal evolution features output by LSTM with the two-dimensional multi-scale spatial features output by CNN through dimensional transformation, achieving deep splicing and fusion of temporal and spatial information. It constructs a spatiotemporal joint feature matrix that combines temporal dynamic adaptation capability with multi-scale spatial capture capability. The fused features are then input into a fully connected layer to complete nonlinear regression prediction, accurately adapting to the complex deformation characteristics of the tunnel construction stages, nonlinearity, and strong disturbances. Finally, it outputs continuous, smooth, and high-precision refined deformation temporal prediction results.

[0082] Furthermore, the construction of the deformation prediction model also includes a model training process. In this embodiment, a sample feature set is constructed based on the aforementioned method for constructing the influencing factor feature set. Then, the data in the sample feature set is divided into a training set, a validation set, and a test set according to the chronological order of construction time. The division ratio can be flexibly set according to the actual situation. In this embodiment, the division ratio of the training set, validation set, and test set is 7:1.5:1.5. By dividing the training set, validation set, and test set according to the chronological order of construction time, it can be ensured that the data time sequence of model training, validation, and testing is completely consistent with the actual construction time sequence on site, completely avoiding the problem of time sequence information leakage, thereby avoiding interference of chaotic time sequence information on the model prediction accuracy.

[0083] After training, this embodiment quantitatively evaluates the overall performance of the model from three dimensions: prediction accuracy, operational stability, and robustness against interference. Prediction accuracy is evaluated using two indicators: mean absolute error and root mean square error, which respectively characterize the average deviation level and overall deviation dispersion between the model's predicted values ​​and the actual values. Operational stability is evaluated using the standard deviation of error fluctuations from multiple rounds of cross-validation results; a smaller standard deviation indicates stronger performance consistency across different construction data. Robustness against interference is evaluated using the data perturbation deviation rate; ±5% Gaussian random noise is applied to the input features, and the deviation ratio of the model's prediction results before and after the perturbation is compared; a lower deviation rate indicates stronger anti-interference capability of the model.

[0084] This embodiment presets thresholds for various indicators: mean absolute error no greater than 0.3 mm, root mean square error no greater than 0.4 mm, standard deviation of error fluctuation no greater than 0.05, and data perturbation deviation rate no greater than 3%, while the model's time-series trend fit is no less than 0.95. Only when all the above indicators simultaneously meet the preset threshold requirements is the model deemed qualified, and the model can be used to predict the deformation at the first moment; otherwise, the model undergoes further training. This ensures the model's predictive accuracy and reliability.

[0085] S4. Determine the construction safety threshold for the first moment based on the deformation time series data, construction time series data, and construction data at the first moment, specifically as follows:

[0086] S41. Determine the time series data of deformation rate based on deformation time series data, and determine the time series data of construction progress rate based on construction time series data;

[0087] S42. Determine the mapping relationship between deformation rate and construction progress rate based on the time-series data of deformation rate and construction progress rate; specifically, this embodiment determines the mapping relationship between deformation rate and construction progress rate at the same moment.

[0088] S43. Determine the construction safety threshold for the first moment based on the mapping relationship and the construction data at the first moment, including:

[0089] S431. Based on the mapping relationship, the construction advancement rate corresponding to the deformation rate that is less than the second preset rate is determined as the conventional advancement rate.

[0090] S432. Determine the construction stability coefficient at the first moment based on the conventional advance rate and the construction data at the first moment;

[0091] Specifically, firstly, the construction progress rate at the first moment is determined based on the construction data at the first moment; then, it is determined whether the construction progress rate at the first moment conforms to the conventional progress rate. If so, the construction stability coefficient at the first moment is determined to be a first value; if not, the construction stability coefficient at the first moment is determined to be a second value; the first value is greater than the second value.

[0092] S433. Determine the baseline safety threshold. The product of the baseline safety threshold and the construction stability coefficient at the first moment shall be determined as the construction safety threshold at the first moment.

[0093] Specifically, the baseline safety threshold is a baseline deformation amount set based on actual geological conditions and construction safety requirements. The second value can be greater than or equal to 1, and can be flexibly selected within the range of (0,1).

[0094] Furthermore, prior to S433, the method also includes:

[0095] The construction conditions at the first moment are determined based on the construction data at the first moment.

[0096] Determine whether the construction condition at the first moment is the preset normal condition. If so, determine that the safety factor of the condition at the first moment is the third value. If not, determine that the safety factor of the condition at the first moment is the fourth value. The third value is greater than the fourth value.

[0097] Accordingly, S433 specifically refers to:

[0098] Determine the baseline safety threshold, and the product of the baseline safety threshold, the construction stability coefficient at the first moment, and the working condition safety coefficient at the first moment shall be determined as the construction safety threshold at the first moment;

[0099] Specifically, the preset conventional working conditions can be flexibly set according to actual geological conditions and construction safety requirements. For example, the working conditions that involve lining, surrounding rock stabilization, and curing are defined as preset conventional working conditions. In contrast, the working conditions corresponding to highly disturbing processes such as excavation and blasting are non-conventional working conditions. The third value can be greater than or equal to 1, and the fourth value can be flexibly selected within the range of (0,1).

[0100] By setting a construction stability coefficient and a working condition safety coefficient, the construction safety threshold at the first moment can be dynamically adjusted based on the construction progress rate and working conditions at that moment. Specifically, when the construction progress rate is higher than the conventional progress rate, the construction stability coefficient is smaller, which will lower the construction safety threshold; when the working conditions do not conform to the preset conventional working conditions, the working condition safety coefficient is smaller, which will also lower the construction safety threshold. This results in the construction safety threshold for unconventional working conditions corresponding to highly disturbing processes such as excavation and blasting being lower than the benchmark safety threshold, while the construction safety threshold for conventional working conditions corresponding to processes such as lining, surrounding rock stabilization, and curing is higher than the benchmark safety threshold. In subsequent risk assessments, the construction safety threshold for highly disturbed and high-risk working conditions can be automatically tightened to improve the sensitivity of early warning for minor anomalies and avoid missed reports; while the construction safety threshold for stable and low-risk working conditions can be automatically loosened to avoid over-warning and invalid false alarms, thereby achieving dynamic adaptive adjustment of the construction safety threshold as a risk assessment standard.

[0101] S5. Determine the construction risk at the first moment based on the deformation amount and construction safety threshold at the first moment;

[0102] Specifically, this embodiment divides the construction safety threshold into four levels of refined risk levels based on dynamic adaptive adjustment: safe level, early warning level, and high-risk level, forming a multi-gradient risk level:

[0103] When the predicted deformation at the first moment is less than 80% of its construction safety threshold, the construction risk at the first moment is determined to be at the safe level.

[0104] When the predicted deformation at the first moment is greater than or equal to 80% but less than 100% of the construction safety threshold, the construction risk at the first moment is determined to be at the warning level, triggering a low-level warning. Intensive monitoring measures can be implemented, such as increasing the original sampling frequency from 10 minutes / time to 2 minutes / time, continuously tracking the deformation evolution trend, and intervening in prevention and control in advance.

[0105] When the predicted deformation at the first moment is greater than or equal to 100% of its construction safety threshold, the construction risk at the first moment is determined to be high-risk, triggering a high-level early warning. Pre-set risk intervention plans can be implemented, such as support reinforcement, construction speed reduction, hidden danger investigation, and stress release.

[0106] This embodiment fully considers the impact of different construction conditions on surrounding rock deformation during construction. It identifies multiple deformation influencing factors based on historical construction time-series data, then constructs a feature set of these factors. During feature set construction, the feature values ​​for periods of strong disturbance are increased. Subsequently, based on the feature set and current construction data, a deep learning model is used to predict the deformation at the first moment in the future. This reflects the impact of construction condition characteristics on the deformation of the underground cavern during prediction, improving prediction accuracy. Simultaneously, this embodiment determines the construction stability coefficient and working condition safety coefficient based on the construction data at the first moment, and adjusts the construction safety threshold based on these coefficients. This achieves dynamic adaptive adjustment of the risk assessment standard, allowing for dynamic assessment of construction risks based on the construction progress information at the first moment. This improves the early warning sensitivity for highly disturbed, high-risk conditions, avoiding missed reports, and reduces the early warning sensitivity for stable, low-risk conditions, avoiding false alarms. Thus, dynamic risk assessment can be performed as construction progresses and construction conditions change, ensuring the accuracy and reliability of the assessment.

[0107] The above description is merely a few embodiments of this application and is not intended to limit this application in any way. Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any changes or modifications made by those skilled in the art without departing from the scope of the technical solution of this application using the disclosed technical content are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

[0108] In the embodiments of this application, unless otherwise explicitly specified and limited, the terms "first," "second," etc., are used to distinguish similar objects and not to describe a specific order or sequence. The terms "connected" and "linked" should be interpreted broadly; for example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application based on the specific circumstances. "Above" or "below" a second feature can mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediate medium. Furthermore, "above," "on top of," and "over" a second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" a second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature. The terms “center,” “longitudinal,” “lateral,” “up,” “down,” “front,” “back,” “left,” “right,” “vertical,” “horizontal,” “top,” “bottom,” “inner,” and “outer,” etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and do not indicate or imply that the position or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations of this invention.

Claims

1. A dynamic risk assessment method for underground cavern construction based on deformation prediction, characterized in that, The method includes: S1. Obtain deformation time-series data and construction time-series data of the underground cavern at multiple consecutive historical moments, and determine multiple deformation influencing factors of the underground cavern based on the deformation time-series data and the construction time-series data. S2. Obtain the time series data of each deformation influencing factor at the multiple historical moments, and construct an influencing factor feature set based on the construction time series data and the factor time series data; S3. Based on the influencing factor feature set and the construction data at the current moment, use convolutional neural networks and long short-term memory networks to predict the deformation of the underground cavern at the first moment. S4. Determine the construction safety threshold at the first moment based on the deformation time series data, the construction time series data, and the construction data at the first moment. S5. Determine the construction risk at the first moment based on the deformation amount and construction safety threshold at the first moment.

2. The method according to claim 1, characterized in that, In S1, multiple deformation influencing factors of the underground cavern are determined based on the deformation time series data and the construction time series data, specifically: Based on the construction timeline data, several candidate influencing factors affecting the deformation of the underground cavern were determined; Obtain the candidate factor time series data for each candidate impact factor at the multiple historical times, and determine the multiple deformed impact factors based on the deformed time series data and the candidate factor time series data.

3. The method according to claim 2, characterized in that, The plurality of deformation influence factors are determined based on the deformation time series data and the candidate factor time series data, specifically as follows: Based on the deformation time series data and the candidate factor time series data, determine the time series correlation degree between each candidate influence factor and the deformation amount of the underground cavern; Multiple candidate influencing factors with a temporal correlation degree higher than a preset threshold are identified as the multiple deformable influencing factors.

4. The method according to claim 3, characterized in that, Based on the deformation time series data and the candidate factor time series data, the time series correlation degree between each candidate influence factor and the deformation amount of the underground cavern is determined, specifically as follows: Based on the deformation time series data and the candidate factor time series data, determine the instantaneous correlation between each candidate influence factor and the deformation amount of the underground cavern at each historical moment; Calculate the average instantaneous correlation of each candidate influence factor at all historical moments, and use the average value as the temporal correlation between the corresponding candidate influence factor and the deformation of the underground cavern.

5. The method according to claim 1, characterized in that, In S2, an influencing factor feature set is constructed based on the construction time series data and the factor time series data, specifically as follows: Principal component analysis was used to transform the time series data of all deformation-influencing factors into multiple principal component features. The multiple principal component features are corrected based on the construction time series data to obtain the influencing factor feature set.

6. The method according to claim 5, characterized in that, The multiple principal component features are corrected based on the construction time series data, specifically as follows: The construction progress rate at each historical moment is determined based on the construction time sequence data, and historical moments with a construction progress rate greater than a first preset rate are defined as strong disturbance moments. Increase the feature value of the principal component features corresponding to the time of the strong disturbance.

7. The method according to claim 1, characterized in that, S4 specifically refers to: S41. Determine the time series data of deformation rate based on the deformation time series data, and determine the time series data of construction progress rate based on the construction time series data; S42. Determine the mapping relationship between deformation rate and construction progress rate based on the time-series data of deformation rate and construction progress rate. S43. Determine the construction safety threshold at the first moment based on the mapping relationship and the construction data at the first moment.

8. The method according to claim 7, characterized in that, S43 specifically refers to: S431. Based on the mapping relationship, the construction advancement rate corresponding to the deformation rate that is less than the second preset rate is determined as the conventional advancement rate. S432. Determine the construction stability coefficient at the first moment based on the conventional advance rate and the construction data at the first moment; S433. Determine the benchmark safety threshold. The product of the benchmark safety threshold and the construction stability coefficient at the first moment is determined as the construction safety threshold at the first moment.

9. The method according to claim 8, characterized in that, S432 specifically refers to: The construction progress rate at the first moment is determined based on the construction data at the first moment. Determine whether the construction progress rate at the first moment conforms to the conventional progress rate. If yes, determine that the construction stability coefficient at the first moment is a first value; if no, determine that the construction stability coefficient at the first moment is a second value; the first value is greater than the second value.

10. The method according to claim 8, characterized in that, Prior to S433, the method further includes: The construction conditions at the first moment are determined based on the construction data at the first moment. Determine whether the construction condition at the first moment is a preset normal condition. If yes, determine that the safety factor of the condition at the first moment is a third value. If no, determine that the safety factor of the condition at the first moment is a fourth value. The third value is greater than the fourth value. Accordingly, S433 specifically refers to: A baseline safety threshold is determined, and the product of the baseline safety threshold, the construction stability coefficient at the first moment, and the working condition safety coefficient at the first moment is determined as the construction safety threshold at the first moment.

Images