Tunnel deformation prediction method based on multi-source data fusion

Abstract

The application relates to the technical field of tunnel monitoring, in particular to a tunnel deformation prediction method based on multi-source data fusion, which comprises the following steps: dividing a tunnel into a plurality of monitoring sections based on elastic modulus; marking crisis categories of the monitoring sections based on impact representation and water flow change; determining the arrangement density of monitoring sensors used to acquire monitoring data according to the crisis categories; periodically inputting the acquired monitoring data into a deformation prediction model to acquire the relative displacement of each detection point of the tunnel within a preset prediction time length; determining whether the deformation prediction of the tunnel is qualified based on the predicted displacement according to the crisis categories; and increasing the training set of the deformation prediction model when it is determined that the deformation prediction of the tunnel is abnormal. According to the specific geotechnical conditions of the tail water tunnel, the tunnel monitoring parameters are determined in a targeted manner, and the prediction accuracy of the tunnel deformation is improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel monitoring technology, and in particular to a method for predicting tunnel deformation based on multi-source data fusion. Background Technology

[0002] Tailrace tunnels are used to discharge the tailrace water from a hydroelectric power station into the downstream river channel. Tailrace tunnels are an important component of the hydroelectric power station's water flow system, ensuring the smooth discharge of water after power generation and maintaining the normal operation of the power station.

[0003] During or after the tailrace tunnel is constructed, external forces may cause certain damage to the tunnel itself, including deformation. Therefore, it is essential to predict the tunnel deformation before it occurs.

[0004] Chinese Patent Publication No. CN118606731A discloses a method and system for predicting tunnel deformation. The method includes preprocessing tunnel deformation monitoring data; decomposing the processed monitoring data into components; performing search and clustering processing on the decomposed monitoring data; inputting historical training data into a first preset model and a second preset model for training; determining whether the deformation feature data is stable. If the deformation feature data is stable, the deformation feature data is input into the first training prediction model for prediction to output a prediction result; if the deformation feature data is not stable, the deformation feature data is input into the second training prediction model for prediction to output a prediction result. It is evident that the above technical solution has the following problem: it does not consider determining tunnel monitoring parameters specifically based on the specific soil and rock conditions of the tailrace tunnel, affecting the accuracy of tunnel deformation prediction. Summary of the Invention

[0005] To address this issue, the present invention provides a tunnel deformation prediction method based on multi-source data fusion, which overcomes the problem in the prior art that the tunnel monitoring parameters are not specifically determined according to the specific soil and rock conditions of the tailrace tunnel, thus affecting the accuracy of tunnel deformation prediction.

[0006] To achieve the above objectives, this invention provides a tunnel deformation prediction method based on multi-source data fusion, comprising:

[0007] The tunnel is divided into several monitoring sections based on the elastic modulus of several points within the tunnel.

[0008] The crisis category of each monitoring section is marked based on the impact characterization quantity and the change in water flow;

[0009] The deployment density of monitoring sensors used to obtain monitoring data is determined based on the crisis category;

[0010] The acquired monitoring data is periodically input into the deformation prediction model to obtain the predicted relative displacement of each detection point in the tunnel within a preset prediction time.

[0011] The system determines whether the deformation prediction for the tunnel is qualified based on the predicted offset, and increases the sample size in the training set of the deformation prediction model when the deformation prediction for the tunnel is determined to be abnormal, based on the crisis category.

[0012] Furthermore, the impact characterization quantity for a single monitoring segment is determined based on the length of each bending segment in a single monitoring segment and the total length of the single monitoring segment.

[0013] Furthermore, the process of labeling the crisis category of a single monitoring segment based on impact characterization parameters includes:

[0014] The impact characterization quantity is compared with the preset impact characterization quantity to mark the crisis category of a single monitoring segment;

[0015] The preset impact characterization value is adjusted to the corresponding value based on the guiding elastic modulus of a single monitoring segment;

[0016] For a single monitoring segment, the average value of each elastic modulus within the monitoring segment is determined as the guiding elastic modulus of the single monitoring segment.

[0017] Furthermore, when the impact characterization value is less than or equal to the preset impact characterization value, a single monitoring segment is marked as a weak crisis category.

[0018] Furthermore, when the impact characterization value is greater than the preset impact characterization value, the crisis category of a single monitoring segment is marked based on the change in water flow.

[0019] When the change in water flow is less than or equal to the preset change in water flow, a single monitoring segment will be marked as a weak crisis category;

[0020] When the change in water flow exceeds the preset change in water flow, a single monitoring segment will be marked as a severe crisis category.

[0021] Furthermore, the change in water flow for a single monitoring segment is determined based on the liquid flow rate obtained by each flow meter at several time points within a single monitoring segment.

[0022] Furthermore, the process of determining the deployment density of monitoring sensors for acquiring monitoring data based on the crisis category includes:

[0023] If a single monitoring segment is classified as a weak crisis category, the current deployment density will continue to be used as the deployment parameter for the monitoring sensors.

[0024] If a single monitoring segment is marked as a high-risk category, the deployment density of monitoring sensors will be adjusted to the corresponding value based on the impact characterization quantity.

[0025] The deployment density of monitoring sensors in a single monitoring segment is the ratio of the total number of all monitoring sensors deployed to the total length of the single monitoring segment.

[0026] The monitoring sensors include displacement gauges for measuring the relative displacement of the tunnel structure, flow meters for obtaining the flow rate of liquids inside the tunnel, and strain gauges attached to the surface of the tunnel structure for obtaining the stress on the tunnel surface.

[0027] Furthermore, when adjusting the deployment density of monitoring sensors within a single monitoring segment, the prediction period used to obtain the predicted relative displacement for that single monitoring segment is adjusted to the corresponding value.

[0028] Furthermore, the process of determining whether the deformation prediction for the tunnel is qualified based on the predicted offset, according to the crisis category, includes:

[0029] When a single monitoring segment is classified as a weak crisis category, the current parameters are continuously used to predict the relative displacement of each monitoring point.

[0030] Furthermore, when a single monitoring segment is marked as a high-crisis category, the deformation prediction for the tunnel is determined to be qualified based on the predicted offset;

[0031] The predicted offset for a single monitoring segment is determined based on each predicted relative displacement and each actual relative displacement.

[0032] When the predicted offset is greater than the preset predicted offset, an anomaly in the deformation prediction for the tunnel is identified, and the sample size in the training set of the deformation prediction model is increased based on the predicted offset.

[0033] Compared with existing technologies, the advantages of this invention are as follows: The elastic modulus characterizes the ability of soil and rock to resist elastic deformation. The properties of soil and rock traversed by tunnels are complex and variable; soil and rock with different elastic moduli have different effects on tunnel deformation. By obtaining soil and rock samples to measure the elastic modulus and dividing the tunnel into monitoring sections, the mechanical properties of the soil and rock within each monitoring section are made relatively consistent, facilitating more accurate subsequent monitoring and analysis. The impact characterization quantity characterizes the degree of influence of the bending section within the monitoring section, and the water flow change quantity characterizes the effect of water flow within the tunnel on the tunnel structure. Based on the impact characterization quantity and water flow change quantity, crisis categories are marked, further improving the prediction accuracy of tunnel deformation while comprehensively assessing the specific conditions of the monitoring section.

[0034] Furthermore, the impact characterization value represents the proportion of curved sections in the monitored section. Curved sections alter the direction of water flow, generating additional impact force. The larger the impact characterization value, the greater the impact of the water flow in the tailrace tunnel on the tunnel, and the greater the impact on the stability of the tunnel structure. The preset impact characterization value used to classify impact conditions is adjusted based on the guiding elastic modulus to adapt to monitoring sections with different rock conditions. When the impact characterization value is less than or equal to the preset impact characterization value, the impact on the tunnel structure is relatively weak, and the individual monitoring section is marked as a weak crisis category. When the impact characterization value is greater than the preset impact characterization value, the degree of crisis is further assessed in conjunction with the water flow variation. The water flow variation represents the fluctuation of water flow rate within the tunnel, and fluctuations in water flow rate can cause scouring effects on the tunnel structure. When the change in water flow is less than or equal to the preset change, the change is relatively small, and the water flow within the tunnel is relatively stable, with minimal impact on the tunnel structure from scouring and erosion. Conversely, when the change in water flow exceeds the preset change, the change is significant, potentially causing substantial impact and damage to the tunnel structure. In this case, the individual monitoring segment is marked as a high-risk category. By comprehensively considering both impact characteristics and water flow changes, the hazard level of the monitoring segment is fully assessed. This not only accurately identifies potential hazardous areas but also further improves the accuracy of tunnel deformation prediction.

[0035] Furthermore, the deployment density of monitoring sensors used to acquire monitoring data is determined based on the crisis category. For monitoring sections with a high risk of deformation, the deployment density of monitoring sensors is increased to improve the accuracy and reliability of monitoring and to detect potential deformation problems in a timely manner. For monitoring sections with a relatively low risk of deformation, maintaining the existing density can reduce resource waste while ensuring monitoring effectiveness and further improve the prediction accuracy of tunnel deformation.

[0036] Furthermore, based on the crisis category, it is determined whether the deformation prediction for the tunnel is qualified based on the predicted offset. The predicted offset is the average of the absolute values ​​of the differences between each predicted relative displacement and the corresponding actual relative displacement, representing the degree of deviation between the prediction result and the actual situation. Further detailed analysis is conducted only for monitoring sections in the high-crisis category. Even slight deformations in the high-crisis category can have serious consequences, and the detection frequency for high-crisis categories is higher, allowing for timely detection of prediction anomalies. When the predicted offset exceeds the preset predicted offset, monitoring data from other monitoring sections with the same guiding elastic modulus as the individual monitoring section are added as an additional training set. This improves the accuracy of the deformation prediction model and further enhances the prediction precision of tunnel deformation. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating the steps of the tunnel deformation prediction method based on multi-source data fusion according to an embodiment of the present invention.

[0038] Figure 2 This is a logic diagram for determining the crisis category of a single monitoring segment based on impact characterization parameters, as shown in an embodiment of the present invention.

[0039] Figure 3 This is a logic diagram illustrating the process of labeling the crisis category of a single monitoring segment based on changes in water flow, according to an embodiment of the present invention.

[0040] Figure 4 This is a logic diagram for determining the deployment density of monitoring sensors used to obtain monitoring data based on the crisis category in an embodiment of the present invention. Detailed Implementation

[0041] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0042] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0043] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0044] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" 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; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0045] Please see Figure 1 , Figure 2 , Figure 3 as well as Figure 4The diagrams shown are, respectively, a flowchart of the steps of the tunnel deformation prediction method based on multi-source data fusion according to an embodiment of the present invention, a logical decision diagram of marking the crisis category of a single monitoring segment based on impact characterization, a logical decision diagram of the process of marking the crisis category of a single monitoring segment based on water flow change, and a logical decision diagram of determining the deployment density of monitoring sensors used to obtain monitoring data according to the crisis category; the tunnel deformation prediction method based on multi-source data fusion according to an embodiment of the present invention includes:

[0046] S1, the tunnel is divided into several monitoring sections based on the elastic modulus;

[0047] S2, based on the impact characterization quantity and the change in water flow, marks the crisis category of each monitoring section;

[0048] S3, determine the deployment density of monitoring sensors used to obtain monitoring data based on the crisis category;

[0049] S4, periodically input the acquired monitoring data into the deformation prediction model to obtain the predicted relative displacement of each detection point in the tunnel within the preset prediction time.

[0050] S5, determine whether the deformation prediction for the tunnel is qualified based on the predicted offset, according to the crisis category, including,

[0051] When identifying deformation prediction anomalies for tunnels, increase the training set of the deformation prediction model;

[0052] Alternatively, once the deformation prediction for the tunnel is deemed satisfactory, the current parameters can be used to continue predicting the relative displacement of each detection point.

[0053] Specifically, there are no restrictions on the method for obtaining the elastic modulus. Several soil and rock samples from the tunnel can be obtained at equal intervals. In a triaxial pressure chamber, confining pressure and axial pressure are applied to the soil and rock samples to simulate the stress state of the soil and rock in actual engineering. The elastic modulus is determined by measuring the stress-strain relationship under different confining pressures. Alternatively, field tests can be conducted, applying vertical pressure to the soil and rock at equal intervals through a bearing plate, measuring the relationship between settlement and pressure to calculate the elastic modulus. This is existing technology and will not be elaborated further.

[0054] Specifically, the specific methods for dividing a tunnel into several monitoring segments based on the elastic modulus can be as follows: tunnel segments with the same and adjacent elastic moduli can be identified as the same monitoring segment; alternatively, cluster analysis can be used to cluster data points of several elastic moduli into several classes to divide the tunnel into several monitoring segments. This will not be elaborated further.

[0055] Specifically, for a single monitoring segment, the average value of each elastic modulus within the monitoring segment is determined as the guiding elastic modulus of the single monitoring segment.

[0056] Specifically, the monitoring sensors include displacement gauges for measuring the relative displacement of the tunnel structure, flow meters for obtaining the flow rate of liquid inside the tunnel, and strain gauges attached to the surface of the tunnel structure for obtaining the stress on the tunnel surface.

[0057] Specifically, the input data of the deformation prediction model includes the monitoring data acquired by the monitoring sensors, and the output data of the deformation prediction model includes the predicted relative displacement of each detection point in the tunnel within a preset prediction time.

[0058] Specifically, each detection point in the tunnel can be a location where displacement gauges are installed.

[0059] Specifically, there are no restrictions on the deformation prediction model. An LSTM model can be constructed to predict the deformation of the tailrace tunnel. For example, by training an LSTM model, patterns and rules in the input data can be learned to predict the tunnel deformation; alternatively, an SVM model can be trained to obtain the predicted tunnel deformation; or LSTM and SVM can be fused to construct a fused prediction model. First, the LSTM model is used to make a preliminary prediction of the tunnel deformation, and then the predicted results and the original data are input into the SVM model for a secondary prediction, fusing the prediction results of the two models. This is an existing technology and will not be elaborated further.

[0060] Specifically, the process of labeling the crisis category of a single monitoring segment based on impact characterization metrics includes:

[0061] The impact characterization quantity is compared with the preset impact characterization quantity to mark the crisis category of a single monitoring segment;

[0062] The ratio of the total length of each bending segment in a single monitoring segment to the total length of the monitoring segment is determined as the impact characterization quantity for a single monitoring segment.

[0063] The pre-set impact characterization value is adjusted to the corresponding value based on the guiding elastic modulus of a single monitoring segment.

[0064] Specifically, there is no limitation on the method for determining the total length of each curved segment within a single monitoring section. One approach is to import the tailrace tunnel design drawings into CAD software, calibrate using known dimensions, and then directly obtain the length of each curved segment using the software's measurement function. Alternatively, a 3D laser scanner can be used to scan each monitoring section, acquiring 3D point cloud data of the tunnel's interior. This point cloud data can then be imported into 3D processing software, and the centerline data of the curved segments can be extracted by fitting the tunnel's centerline. The software's measurement function can then be used to calculate the length of the centerline, thus obtaining the length of each curved segment. This will not be elaborated further.

[0065] Specifically, based on the guiding elastic modulus of a single monitoring segment, the preset impact characterization quantity is adjusted to the corresponding value, wherein,

[0066] The increase in the preset impact characterization is proportional to the guiding elastic modulus.

[0067] In this embodiment, optionally,

[0068] The guiding elastic modulus is compared with the first preset guiding modulus and the second preset guiding modulus;

[0069] If the guiding elastic modulus is less than or equal to the first preset guiding modulus, the preset impact characterization value is adjusted to 1.1 times the initial preset impact characterization value;

[0070] If the guiding elastic modulus is less than or equal to the second preset guiding modulus and greater than the first preset guiding modulus, then the preset impact characterization value is adjusted to 1.18 times the initial preset impact characterization value.

[0071] If the guiding elastic modulus is greater than the second preset guiding modulus, the preset impact characterization value is adjusted to 1.25 times the initial preset impact characterization value;

[0072] The first preset guiding modulus is 0.7Y0, and the second preset guiding modulus is 0.8Y0, where Y0 is the average value of the obtained elastic moduli.

[0073] Specifically, the elastic modulus characterizes the ability of soil and rock to resist elastic deformation. The soil and rock masses traversed by tunnels have complex and variable properties; soil and rock masses with different elastic moduli have varying effects on tunnel deformation. By obtaining soil and rock samples to measure the elastic modulus and dividing the tunnel into monitoring sections, the mechanical properties of the soil and rock masses within each monitoring section are made relatively consistent, facilitating more accurate subsequent monitoring and analysis. Impact parameters characterize the degree of influence of bending within the monitoring section, while water flow variation characterizes the effect of water flow on the tunnel structure. Based on the impact parameters and water flow variation, crisis categories are marked, further improving the accuracy of tunnel deformation prediction while comprehensively assessing the specific conditions of the monitoring sections.

[0074] Specifically, if the impact characterization value is less than or equal to the preset impact characterization value, then a single monitoring segment will be marked as a weak crisis category.

[0075] If the impact characterization value is greater than the preset impact characterization value, the crisis category of a single monitoring segment is marked based on the change in water flow.

[0076] Specifically, the preset impact characterization quantity Z0 is selected within the interval [0.2, 0.3].

[0077] Specifically, the process of labeling the crisis category of a single monitoring segment based on changes in water flow includes:

[0078] Solve for the variance of liquid flow rate obtained by a single flow meter at several time points within a single monitoring segment;

[0079] The mean of the variance of each flow rate within a single monitoring segment is calculated to obtain the change in water flow for that single monitoring segment.

[0080] If the change in water flow is less than or equal to the preset change in water flow, then the individual monitoring segment will be marked as a weak crisis category.

[0081] If the change in water flow exceeds the preset change in water flow, the individual monitoring segment will be marked as a severe crisis category.

[0082] Specifically, the preset water flow change is selected within the range [0.05, 0.15], and its unit is (m^3 / s)².

[0083] Specifically, the impact characterization value represents the proportion of curved sections in the monitored section. Curved sections alter the direction of water flow, generating additional impact force. The larger the impact characterization value, the greater the impact of the water flow in the tailrace tunnel on the tunnel, and the greater the impact on the stability of the tunnel structure. The preset impact characterization value used to classify impact conditions is adjusted based on the guiding elastic modulus to adapt to monitoring sections with different rock conditions. When the impact characterization value is less than or equal to the preset impact characterization value, the impact on the tunnel structure is relatively weak, and the individual monitoring section is marked as a weak crisis category. When the impact characterization value is greater than the preset impact characterization value, the degree of crisis is further assessed in conjunction with the water flow variation. The water flow variation value represents the fluctuation of water flow within the tunnel, and fluctuations in water flow can cause scouring effects on the tunnel structure. When the water flow variation value is less than or equal to the preset water flow variation value, the water flow variation is small, and the water flow within the tunnel is relatively stable, with minimal impact on the tunnel structure from scouring and erosion. When the water flow variation value is greater than the preset water flow variation value, the water flow variation is large, and this will cause significant impact and damage to the tunnel structure; in this case, the individual monitoring section is marked as a strong crisis category. By comprehensively considering both impact characteristics and water flow changes, the degree of danger of the monitoring section is fully assessed, which not only accurately identifies potential dangerous areas but also further improves the prediction accuracy of tunnel deformation.

[0084] Specifically, the process of determining the deployment density of monitoring sensors for acquiring monitoring data based on the crisis category includes:

[0085] If a single monitoring segment is classified as a weak crisis category, the current deployment density will continue to be used as the deployment parameter for the monitoring sensors.

[0086] If a single monitoring segment is marked as a high-risk category, the deployment density of monitoring sensors will be adjusted to the corresponding value based on the impact characterization quantity.

[0087] Specifically, the deployment density of monitoring sensors in a single monitoring segment is the ratio of the total number of all monitoring sensors deployed to the total length of the single monitoring segment.

[0088] Specifically, the deployment density of monitoring sensors used to obtain monitoring data is determined based on the crisis category. In the case of a high crisis category, the risk of deformation in the monitoring section is high, so the deployment density of monitoring sensors is increased to improve the accuracy and reliability of monitoring and to detect potential deformation problems in a timely manner. In the case of a low crisis category, the risk in the monitoring section is relatively low, and maintaining the existing density can reduce resource waste while ensuring the monitoring effect, and further improve the prediction accuracy of tunnel deformation.

[0089] Specifically, the deployment density of monitoring sensors is adjusted to a corresponding value based on the impact characterization parameters, wherein,

[0090] The increase in deployment density is directly proportional to the impact characterization quantity.

[0091] In this embodiment, optionally,

[0092] The impact characterization quantity is compared with the first preset impact comparison threshold and the second preset impact comparison threshold;

[0093] If the impact characterization value is less than or equal to the first preset impact comparison threshold, the deployment density of the monitoring sensors in a single monitoring segment will be adjusted to 1.1 times the initial deployment density.

[0094] If the impact characterization value is less than or equal to the second preset impact comparison threshold and greater than the first preset impact comparison threshold, the deployment density of the monitoring sensors in a single monitoring segment will be adjusted to 1.2 times the initial deployment density.

[0095] If the impact characterization value is greater than the second preset impact comparison threshold, the deployment density of the monitoring sensors in a single monitoring segment will be adjusted to 1.3 times the initial deployment density.

[0096] The first preset impact comparison threshold is set to 1.25Z0, and the second preset impact comparison threshold is set to 1.44Z0.

[0097] Specifically, when adjusting the deployment density of monitoring sensors within a single monitoring segment, the prediction period for that single monitoring segment will be adjusted to the corresponding value.

[0098] In this embodiment, optionally, the prediction period for a single monitoring segment is adjusted to 0.8 times the initial prediction frequency.

[0099] Specifically, the prediction period is the period used to obtain the relative displacement of each detection point in the tunnel within a preset prediction time.

[0100] Specifically, depending on the crisis category, it is determined whether the deformation prediction for the tunnel is qualified based on the predicted offset, including:

[0101] If a single monitoring segment is classified as a weak crisis category, the current parameters will continue to be used to predict the relative displacement of each monitoring point.

[0102] If a single monitoring segment is marked as a high-crisis category, the qualification of the deformation prediction for the tunnel is determined based on the predicted offset.

[0103] Specifically, the process of determining whether the deformation prediction for the tunnel is qualified based on the predicted offset includes:

[0104] The absolute value of the difference between each predicted relative displacement and the corresponding actual relative displacement is calculated, and the average value of each calculated absolute value is determined as the predicted offset for a single monitoring segment.

[0105] If the predicted offset is less than or equal to the preset predicted offset, the deformation prediction for the tunnel is deemed qualified, and the current parameters are used to continue to predict the relative displacement of each detection point.

[0106] If the predicted offset is greater than the preset predicted offset, it is determined that the deformation prediction for the tunnel is abnormal, and the training set of the deformation prediction model is increased based on the predicted offset.

[0107] Specifically, the preset prediction offset P0 is selected within the interval [0.02, 0.05], and its unit is mm.

[0108] Specifically, the process of increasing the training set of the deformation prediction model based on the predicted offset includes:

[0109] The increase in the amount of training data is proportional to the prediction offset.

[0110] Specifically, the guiding elastic modulus of a single monitoring segment is obtained, and the monitoring data and actual relative displacement of each monitoring segment with the same guiding elastic modulus are selected as an additional training set.

[0111] In this embodiment, optionally,

[0112] The predicted offset is compared with the first preset offset comparison threshold and the second preset offset comparison threshold;

[0113] If the predicted offset is less than or equal to the first preset offset comparison threshold, the amount of training data will be increased to 1.13 times the initial amount of data.

[0114] If the predicted offset is less than or equal to the second preset offset comparison threshold and greater than the first preset offset comparison threshold, the amount of training set data will be increased to 1.23 times the initial amount of data.

[0115] If the predicted offset is greater than the second preset offset comparison threshold, the amount of training set data will be increased to 1.33 times the initial amount of data.

[0116] The first preset offset comparison threshold is 2P0, and the second preset offset comparison threshold is 5P0.

[0117] Specifically, the selection of each data point is based on statistical analysis of a large amount of monitoring data from actual tunnel engineering projects. It is understood that those skilled in the art can select the data points themselves based on the actual application.

[0118] Specifically, based on the crisis category, it is determined whether the deformation prediction for the tunnel is qualified based on the predicted offset. The predicted offset is the average of the absolute values ​​of the differences between each predicted relative displacement and the corresponding actual relative displacement, representing the degree of deviation between the prediction result and the actual situation. Further detailed analysis is conducted only for monitoring sections in the high-crisis category. Even slight deformation in the high-crisis category can have serious consequences, and the detection frequency for the high-crisis category is higher, allowing for timely detection of prediction anomalies. When the predicted offset exceeds the preset predicted offset, monitoring data from other monitoring sections with the same guiding elastic modulus as the individual monitoring section are added as an additional training set. This improves the accuracy of the deformation prediction model and further enhances the prediction precision of tunnel deformation.

[0119] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0120] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A tunnel deformation prediction method based on multi-source data fusion, characterized in that, The method comprises the following steps: dividing the tunnel into a plurality of monitoring sections based on the elastic modulus of a plurality of points in the tunnel; determining an impact characteristic of a single monitoring section based on the length of each curved section in the single monitoring section and the total length of the single monitoring section, and marking the crisis category of each monitoring section based on the impact characteristic and the water flow change; determining the layout density of monitoring sensors for obtaining monitoring data according to the crisis category, wherein the monitoring sensors comprise a displacement meter for measuring the relative displacement of the tunnel structure, a flow meter for obtaining the flow rate of the liquid in the tunnel, and a strain gauge attached to the surface of the tunnel structure for obtaining the stress of the tunnel surface; periodically inputting the obtained monitoring data into a deformation prediction model to obtain the predicted relative displacement of each detection point of the tunnel within a preset prediction time; determining whether the deformation prediction of the tunnel is qualified based on the predicted displacement according to the crisis category, and increasing the sample size in the training set of the deformation prediction model when it is determined that the deformation prediction of the tunnel is abnormal.

2. The tunnel deformation prediction method based on multi-source data fusion according to claim 1, characterized in that, The process of marking the crisis category of a single monitoring section based on the impact characteristic comprises: comparing the impact characteristic with a preset impact characteristic to mark the crisis category of the single monitoring section; adjusting the preset impact characteristic to a corresponding value based on the guide elastic modulus of the single monitoring section; determining the average value of each elastic modulus in the single monitoring section as the guide elastic modulus of the single monitoring section.

3. The tunnel deformation prediction method based on multi-source data fusion according to claim 2, characterized in that, When the impact characteristic is less than or equal to the preset impact characteristic, the single monitoring section is marked as a weak crisis category.

4. The tunnel deformation prediction method based on multi-source data fusion according to claim 3, characterized in that, When the impact characteristic is greater than the preset impact characteristic, the crisis category of the single monitoring section is marked based on the water flow change; When the water flow change is less than or equal to a preset water flow change, the single monitoring section is marked as a weak crisis category; When the water flow change is greater than the preset water flow change, the single monitoring section is marked as a strong crisis category.

5. The tunnel deformation prediction method based on multi-source data fusion according to claim 4, characterized in that, The water flow change of a single monitoring section is determined based on the liquid flow obtained by each flow meter in the single monitoring section at a plurality of time nodes.

6. The tunnel deformation prediction method based on multi-source data fusion according to claim 5, characterized in that, The process of determining the layout density of monitoring sensors for obtaining monitoring data according to the crisis category comprises: if the single monitoring section is a weak crisis category, the current layout density is continuously used as the layout parameter of the monitoring sensors; if the single monitoring section is marked as a strong crisis category, the layout density of the monitoring sensors is adjusted to a corresponding value based on the impact characteristic; the layout density of the monitoring sensors of a single monitoring section is the ratio of the total number of the laid-out monitoring sensors to the total length of the single monitoring section.

7. The tunnel deformation prediction method based on multi-source data fusion according to claim 6, characterized in that, When the adjustment of the layout density of the monitoring sensors in a single monitoring section is completed, the prediction period for obtaining the predicted relative displacement of the single monitoring section is adjusted to a corresponding value.

8. The tunnel deformation prediction method based on multi-source data fusion according to claim 7, characterized in that, The process of determining whether the deformation prediction of the tunnel is qualified based on the predicted displacement according to the crisis category comprises: when the single monitoring section is a weak crisis category, the current parameters are continuously used to complete the prediction of the relative displacement of each detection point.

9. The tunnel deformation prediction method based on multi-source data fusion according to claim 8, characterized in that, when the single monitoring section is marked as a strong crisis category, whether the deformation prediction of the tunnel is qualified is determined based on the predicted displacement; the predicted displacement of a single monitoring section is determined based on each predicted relative displacement and each actual relative displacement. When the predicted offset is greater than the preset predicted offset, a deformation prediction anomaly for the tunnel is determined, and a sample quantity in a training set of the deformation prediction model is increased based on the predicted offset.

Images