An inverted river crossing construction deformation seepage automatic monitoring method and system

Abstract

The application discloses a kind of inverted river-crossing construction deformation seepage automatic monitoring method and system, for river pipeline construction safety monitoring field, the method comprises: obtaining the comprehensive key index in inverted river-crossing construction area;Identify the abnormal working condition of pit and pipe body linkage in comprehensive key index, carry out multi-index comprehensive evaluation, obtain comprehensive risk index;Based on comprehensive key index and comprehensive risk index, predict the nonlinear floating trend of pipe body in future preset time period using dynamic trend deduction algorithm, while setting short window emergency prediction strategy;According to the comprehensive risk index and the predicted nonlinear floating trend of pipe body, a collaborative control model is constructed, and each device control instruction is output and executed.The application significantly improves the response accuracy and safety of buoyancy control and leakage plugging in inverted river-crossing construction.

Description

Technical Field

[0001] This application relates to the field of safety monitoring during the construction of pipelines crossing rivers, and more specifically, to an automatic monitoring method and system for deformation and seepage during the construction of inverted siphon pipelines crossing rivers. Background Technology

[0002] Inverted siphon construction refers to an engineering method in which water supply pipelines are laid in a U-shape and pass under the bottom of the riverbed. The pipeline first bends downward, then runs horizontally, and then bends upward, forming an inverted siphon tube shape. It uses the pressure difference between the upstream and downstream sides to drive the water flow, avoiding damage to the river channel for flood control, navigation, and landscape.

[0003] The construction of an inverted siphon across a river involves stages such as cofferdam construction, deep foundation pit excavation, pipeline laying, concrete encapsulation, and backfilling. Because the pipe is U-shaped and located in a high groundwater level area, it is prone to floating as a whole due to the buoyancy of groundwater before it is completely covered with soil. At the same time, the water pressure difference between the inside and outside of the foundation pit may lead to leakage or even backflow of river water.

[0004] Existing technologies have limited monitoring indicators and cannot automatically identify abnormal operating conditions. Risk scoring uses linear weighting, which makes it insensitive to indicators exceeding limits. Trend prediction lacks nonlinear modeling capabilities and has no response mechanism for sudden changes in seepage, resulting in delayed early warning. Equipment control is independent and mostly uses simple start-stop switching control, lacking collaborative closed-loop capabilities based on real-time risk, which can easily lead to control shocks or response delays.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To overcome the above problems, this application aims to propose an automatic monitoring method and system for deformation and seepage during inverted siphon river crossing construction. The purpose is to solve the problems of lack of nonlinear modeling capability in trend prediction and lack of response mechanism for sudden changes in seepage, resulting in delayed early warning; independent equipment control, mostly using simple start-stop switch control, lacking collaborative closed-loop capability based on real-time risk, and easy to generate control shocks or response delays.

[0007] Therefore, the specific technical solution adopted in this application is as follows: In a first aspect, the present invention provides an automatic monitoring method for deformation and seepage during the construction of an inverted siphon crossing a river, the method comprising: Obtain comprehensive key indicators within the inverted siphon river crossing construction area, including real-time pipe body uplift, water level difference inside and outside the foundation pit, dewatering well output, and pipe wall strain. The abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators are identified, and then the floating rate, stress change rate and water discharge rate are comprehensively evaluated to obtain the comprehensive risk index. Based on comprehensive key indicators and comprehensive risk index, the dynamic trend extrapolation algorithm is used to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, a short-window emergency prediction strategy is set to be automatically triggered when sudden seepage changes are detected. Based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, a collaborative control model is constructed for the dynamic matching of the optimal control combination among dewatering, recharge, grouting and ballast equipment, and outputs and executes the control commands of each equipment.

[0008] Optionally, the method for obtaining the comprehensive risk index is as follows: Based on the real-time pipe body uplift, water level difference inside and outside the foundation pit, water output from dewatering wells, and pipe wall strain in the comprehensive key indicators, abnormal working conditions of the linkage between the foundation pit and the pipe body are identified. In abnormal working conditions of the linkage between the foundation pit and the pipe body, the uplift rate is obtained from the real-time uplift of the pipe body, the stress change rate is obtained from the pipe wall strain, the water output rate is obtained from the water output of the dewatering well, and the water level amplitude is obtained from the water level difference inside and outside the foundation pit. A comprehensive risk index is obtained by comprehensively evaluating multiple indicators, including buoyancy rate, stress change rate, outflow rate change rate, and water level fluctuation.

[0009] Optionally, the method for predicting the nonlinear upward trend of the tube within a preset time period using a dynamic trend extrapolation model is as follows: Comprehensive key indicators Preprocessing is performed to obtain a comprehensive dataset; By analyzing the comprehensive dataset using a dynamic trend extrapolation algorithm, the amount of pipe rising at each moment within a preset time period is predicted, and the rising amount change curve is obtained. Obtain the ascent rate at each moment from the ascent rate change curve, obtain the ascent rate curve, and identify the peak ascent rate and its arrival time.

[0010] Optionally, the process of obtaining the comprehensive dataset includes: The comprehensive key indicators are filtered, denoised, and normalized, and the indicators are time-series aligned according to the same sampling period to obtain multidimensional monitoring data. After performing second-order difference processing on the pipe body uplift amount in the multidimensional monitoring data, the uplift acceleration is extracted, and the water level fluctuation intensity is obtained by using sliding window statistics on the water level difference inside and outside the foundation pit. The extracted buoyancy acceleration, water level fluctuation intensity, and dewatering well output and pipe wall strain are integrated to obtain an extended monitoring index set. This extended monitoring index set is then integrated with the comprehensive risk index to construct a comprehensive dataset.

[0011] Optionally, the method for obtaining the buoyancy change curve is as follows: According to the preset historical window length, the continuous monitoring values ​​of each time step are extracted sequentially from the comprehensive dataset to obtain the continuous monitoring value sequence; The continuously monitored numerical sequence is sequentially fed into the Long Short-Term Memory network, and the internal state is updated step by step to obtain the historical internal state sequence. The importance weight of each time step is calculated through a temporal importance assessment mechanism. The historical internal state sequence is then weighted and fused to obtain a comprehensive state feature. This feature is then integrated with the latest internal state and used as the initial state of the decoding and prediction unit. The expression for the calculation process is as follows: In the formula, Indicates the first t Attention weights for each time step; Represents the natural exponential function; Indicates the first t Attention score at each time step; Represents the hyperbolic tangent activation function; Indicates the first One time step; Indicates the first Attention score at each time step; Represents the parameter vector determined through training. Transpose of; This represents the weight matrix determined through training; This represents the internal state of the Long Short-Term Memory network at the U-th time step; This represents the bias feature determined through training; This indicates the total number of time steps in the history window; Indicates the first t One time step; ) represents the vector concatenation operation; This indicates the initial state of the decoding prediction unit; c Indicates the overall state characteristics; The decoding and prediction unit starts from the initial state, outputs the predicted value of the upward float step by step, and uses the current predicted value as the input for the next step. The cycle continues until the predicted values ​​for all moments within the preset time period are generated, and the future upward float change curve is obtained by arranging them in chronological order.

[0012] Optionally, the process of looping until the predicted values ​​for all moments within a preset time period are generated, and arranging them in chronological order to obtain the future upward change curve, includes: Let the prediction window length be P, and the decoding prediction unit be... As the initial state, for the th n Step, and n =1, 2, ..., P, the predicted buoyancy values ​​are generated sequentially according to time steps, and the expression for the calculation process of obtaining the future buoyancy change curve by arranging them in time order is as follows: In the formula, P represents the total number of time steps in the prediction window; This represents the output function, used to map the current state to the predicted upward float value; This represents the state update function, which is used to calculate the next state by combining the current state with the current prediction value. Indicates the first P The predicted amount of pipe body buoyancy; Indicates the first Update status after step completion; This represents the curve showing the change in upward fluctuation, which is composed of all predicted values ​​arranged in chronological order.

[0013] Optionally, the setting of a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected includes: Real-time monitoring of the water level difference inside and outside the foundation pit and the rate of change of the water output from the dewatering well. When any rate of change exceeds the preset threshold, it is determined to be a sudden change in seepage characteristics. Trigger the short window emergency prediction strategy, switch the prediction window from the normal duration to the preset short window duration, and correspondingly increase the data sampling frequency to the preset high frequency value; Using the comprehensive key indicators within the most recent sampling period as input, the state parameters established in the previous prediction process are reused, and the decoding prediction unit is used to generate the predicted value of the upward amount within a preset short window. The upward buoyancy value and arrival time are compared with the conventional prediction results, and the maximum upward buoyancy value and the earliest arrival time are taken as the basis for early warning.

[0014] Optionally, the construction of the collaborative control model for dynamically matching the optimal control combination among precipitation, reinjection, grouting, and ballast equipment includes: Based on the comprehensive risk index, the current working condition type is identified, and the dewatering, recharge, grouting and ballast equipment are divided into main control equipment set and auxiliary equipment set according to response speed and effect. Based on the predicted nonlinear upward trend of the pipe body, the initial ratio of the required control quantities of the main control equipment and auxiliary equipment is calculated; The comprehensive risk index is mapped to a stepless adjustment coefficient and the initial ratio is corrected to obtain the target control amount for each device. Using the final target control amount as a constraint, a dynamic collaborative mapping relationship is established among precipitation, reinjection, grouting, and ballast equipment, and a collaborative control model is constructed. The expression for its calculation process is as follows: In the formula, Indicates a collection of main control devices; Represents a collection of auxiliary equipment; Represents the initial control quantity allocation vector; This indicates the initial ratio of the required control quantities between the main control equipment and the auxiliary equipment; This represents the peak value of the predicted nonlinear upward trend of the tube. Indicates the predicted peak arrival time; Indicates the stepless adjustment coefficient; This represents the overall risk index; This represents the final target control vector after modification. This represents a specific allocation vector for high-risk operating conditions; Indicates time The device status feedback information; Indicates time Output control commands.

[0015] A second aspect of the present invention provides an automatic monitoring system for deformation and seepage during the construction of an inverted siphon crossing a river, the system comprising: The integrated key indicator perception module is used to acquire comprehensive key indicators in the inverted siphon construction area, including the real-time uplift of the pipe body, the water level difference inside and outside the foundation pit, the water output of the dewatering well, and the pipe wall strain. The risk identification and assessment module is used to identify abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators. Then, it comprehensively evaluates multiple indicators such as the buoyancy rate, stress change rate and water flow rate to obtain the comprehensive risk index. The trend prediction module, based on comprehensive key indicators and comprehensive risk index, uses a dynamic trend extrapolation algorithm to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, it sets a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected. The collaborative control and execution module is used to construct a collaborative control model for dynamically matching the optimal control combination among dewatering, recharge, grouting and ballast equipment based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, and output and execute the control commands of each equipment.

[0016] Compared with the prior art, this application has the following beneficial effects: 1. This application uses multi-source deformation-seepage sensors to automatically collect data on pipe uplift, water level difference, water output and strain, identify abnormal working conditions of the pit and pipe linkage, and adopts an exponential weighted fusion method to nonlinearly penalize the out-of-limit indicators, thereby achieving accurate quantification of the comprehensive risk index and effectively distinguishing different working conditions such as construction disturbance, normal settlement, seepage and uplift and buoyancy change.

[0017] 2. This application introduces an attention mechanism and LSTM autoregressive decoding into time series prediction, and uses historical multi-index sequences to accurately model the nonlinear evolution trend of pipe floating, so as to realize multi-step prediction within a preset time period in the future; at the same time, it designs a short window emergency prediction strategy for sudden seepage changes, which significantly improves the response speed and reliability of sudden risks.

[0018] 3. This application maps the comprehensive risk index to a stepless adjustment coefficient, and smoothly transitions between conventional mix proportions and high-risk special mix proportions through linear interpolation. It constructs a collaborative control model for dewatering, reinjection, grouting, and ballast equipment, realizing continuous stepless adjustment from risk quantification to control execution. This avoids the impact and lag of traditional switch control, forms a closed-loop intelligent management and control system, and significantly improves the response accuracy and safety of buoyancy control and leakage sealing in inverted siphon river crossing construction. Attached Figure Description

[0019] The above-mentioned features, characteristics, and advantages of this application, as well as their implementation methods, will become clearer and more understandable in conjunction with the following description of the embodiments, which are illustrated in detail with reference to the accompanying drawings. Schematic diagrams are shown here: Figure 1 This is a flowchart of the automatic monitoring method for deformation and seepage during inverted siphon river crossing construction in this application; Figure 2 This is a diagram of the short-window emergency prediction strategy in the automatic monitoring method for deformation and seepage during inverted siphon river crossing construction in this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0021] This embodiment provides an automatic monitoring method for deformation and seepage during inverted siphon construction. It automatically collects data on pipe uplift, water level difference, water flow, and strain using multi-source deformation-seepage sensors. This identifies abnormal conditions linking the foundation pit and the pipe, and employs an exponential weighted fusion method to apply nonlinear penalties to exceed limits, achieving precise quantification of the comprehensive risk index. This effectively distinguishes between different conditions such as construction disturbance, normal settlement, seepage and uplift, and sudden buoyancy changes. Figure 1 As shown, the method includes: Obtain comprehensive key indicators within the inverted siphon river crossing construction area, including real-time pipe uplift, water level difference inside and outside the foundation pit, dewatering well output, and pipe wall strain.

[0022] It should be explained that the real-time buoyancy of the pipe is obtained by laying a hydrostatic level along the axis of the inverted siphon and directly measuring the vertical displacement of the U-shaped pipe under the action of buoyancy. The strain of the pipe wall is obtained by welding a vibrating wire strain gauge to the outer wall of the pipe (especially in stress concentration areas such as the lower bend and upper bend), which reflects the degree of tensile and compressive deformation of the pipe wall material, and at the same time helps to distinguish different working conditions such as buoyancy and uneven settlement. The difference in water level inside and outside the foundation pit is obtained by measuring the water level inside the foundation pit (dewatering area) and outside the foundation pit (riverbed and groundwater area) with a piezometer and calculating the water level difference. This difference is a direct measure of the driving force of groundwater seepage. The larger the water level difference, the stronger the buoyancy and seepage pressure that the pipe body bears. The output of the precipitation well is obtained by monitoring the pumping flow rate of the precipitation well using an electromagnetic flow meter. This indicator reflects the intensity of active precipitation, and combined with the water level difference, it can be used to determine whether the precipitation effect is abnormal. The project is implemented as follows: a static level is arranged every 6 to 10 meters along the pipe axis; strain gauges are installed on the outer surface of the lower bend, upper bend and middle flat section of the pipe; piezometers are respectively installed inside and outside the foundation pit and in the boreholes on both banks, with a depth flush with the bottom of the pipe; flow meters are installed on the main pipe of the dewatering well. All sensors automatically collect data via low-power wireless modules at a frequency of 0.1Hz to 1Hz. The data is then aggregated to the on-site acquisition unit and uploaded to the cloud. The acquisition frequency can be dynamically adjusted according to the construction stage, such as increasing the frequency during excavation and decreasing it after backfilling. All indicators are aligned with a unified timestamp to form a multi-dimensional time-series dataset.

[0023] S2. Identify abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators, and then conduct a comprehensive evaluation of multiple indicators such as the floating rate, stress change rate and water output change rate to obtain the comprehensive risk index.

[0024] Preferably, the method for obtaining the comprehensive risk index is as follows: Based on the real-time pipe body uplift, water level difference inside and outside the foundation pit, water output from dewatering wells, and pipe wall strain in the comprehensive key indicators, abnormal working conditions of the linkage between the foundation pit and the pipe body are identified. In abnormal working conditions of the linkage between the foundation pit and the pipe body, the uplift rate is obtained from the real-time uplift of the pipe body, the stress change rate is obtained from the pipe wall strain, the water output rate is obtained from the water output of the dewatering well, and the water level amplitude is obtained from the water level difference inside and outside the foundation pit. By comprehensively evaluating multiple indicators such as buoyancy rate, stress change rate, outflow rate variability, and water level fluctuation, a comprehensive risk index is obtained. The expression for its calculation process is as follows: In the formula, This represents the overall risk index; Indicators; Indicates the rate at which the tube rises; This represents the rate of change of pipe wall stress; Indicates the rate of change in water output from precipitation wells; Indicates the fluctuation range of water levels inside and outside the foundation pit; Indicates the first i Weighting coefficients for each risk indicator; Indicates the sensitivity coefficient; Indicates the first i The relative risk ratio of each risk indicator.

[0025] It should be explained that identifying abnormal conditions in the linkage between the foundation pit and the pipe body refers to simultaneously monitoring the changing trends of the pipe body's buoyancy rate and the water level difference between the inside and outside of the foundation pit: when the buoyancy rate continues to rise and the water level difference increases synchronously, it is determined to be an abnormal buoyancy condition; when the water level difference is stable but the outflow suddenly increases, it is determined to be an abnormal leakage condition; when only the buoyancy rate fluctuates and the water level difference changes slowly, it is classified as construction vibration disturbance or normal consolidation settlement. The buoyancy rate, stress change rate, water output change rate, and water level change amplitude were obtained by performing first-order difference analysis on the time series data of the real-time buoyancy of the pipe body, pipe wall strain, water output from the dewatering well, and water level difference inside and outside the foundation pit, respectively. That is, the difference between adjacent time points is divided by the sampling time interval.

[0026] S3. Based on comprehensive key indicators and comprehensive risk index, use dynamic trend extrapolation algorithm to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, set a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected.

[0027] Preferably, the method for predicting the nonlinear upward trend of the tube within a preset time period using a dynamic trend extrapolation model is as follows: Comprehensive key indicators Preprocessing is performed to obtain a comprehensive dataset; By analyzing the comprehensive dataset using a dynamic trend extrapolation algorithm, the amount of pipe rising at each moment within a preset time period is predicted, and the rising amount change curve is obtained. Obtain the ascent rate at each moment from the ascent rate change curve, obtain the ascent rate curve, and identify the peak ascent rate and its arrival time.

[0028] Preferably, the process of obtaining the comprehensive dataset includes: The comprehensive key indicators are filtered, denoised, and normalized, and the indicators are time-series aligned according to the same sampling period to obtain multidimensional monitoring data. After performing second-order difference processing on the pipe body uplift amount in the multidimensional monitoring data, the uplift acceleration is extracted, and the water level fluctuation intensity is obtained by using sliding window statistics on the water level difference inside and outside the foundation pit. The extracted buoyancy acceleration, water level fluctuation intensity, and dewatering well output and pipe wall strain are integrated to obtain an extended monitoring index set. This extended monitoring index set is then integrated with the comprehensive risk index to construct a comprehensive dataset.

[0029] Preferably, the method for obtaining the buoyancy change curve is as follows: According to the preset historical window length, the continuous monitoring values ​​of each time step are extracted sequentially from the comprehensive dataset to obtain the continuous monitoring value sequence; The continuously monitored numerical sequence is sequentially fed into the Long Short-Term Memory network, and the internal state is updated step by step to obtain the historical internal state sequence. The importance weight of each time step is calculated through a temporal importance assessment mechanism. The historical internal state sequence is then weighted and fused to obtain a comprehensive state feature. This feature is then integrated with the latest internal state and used as the initial state of the decoding and prediction unit. The expression for the calculation process is as follows: In the formula, Indicates the first t Attention weights for each time step; Represents the natural exponential function; Indicates the first t Attention score at each time step; Represents the hyperbolic tangent activation function; Indicates the first One time step; Indicates the first Attention score at each time step; Represents the parameter vector determined through training. Transpose of; This represents the weight matrix determined through training; This represents the internal state of the Long Short-Term Memory network at the U-th time step; This represents the bias feature determined through training; This indicates the total number of time steps in the history window; Indicates the first t One time step; ) represents the vector concatenation operation; This indicates the initial state of the decoding prediction unit; c Indicates the overall state characteristics; The decoding and prediction unit starts from the initial state, outputs the predicted value of the upward float step by step, and uses the current predicted value as the input for the next step. The cycle continues until the predicted values ​​for all moments within the preset time period are generated, and the future upward float change curve is obtained by arranging them in chronological order.

[0030] Preferably, the process of looping until the predicted values ​​for all moments within a preset time period are generated, and arranging them in chronological order to obtain the future upward fluctuation curve, includes: Let the prediction window length be P, and the decoding prediction unit be... As the initial state, for the th n Step, and n=1, 2, ..., P, the predicted buoyancy values ​​are generated sequentially according to time steps, and the expression for the calculation process of obtaining the future buoyancy change curve by arranging them in time order is as follows: In the formula, P represents the total number of time steps in the prediction window; This represents the output function, used to map the current state to the predicted upward float value; This represents the state update function, which is used to calculate the next state by combining the current state with the current prediction value. Indicates the first P The predicted amount of pipe body buoyancy; Indicates the first Update status after step completion; This represents the curve showing the change in upward fluctuation, which is composed of all predicted values ​​arranged in chronological order.

[0031] Preferably, the setting of a short-window emergency prediction strategy that is automatically triggered when a sudden change in seepage characteristics is detected includes: Real-time monitoring of the water level difference inside and outside the foundation pit and the rate of change of the water output from the dewatering well. When any rate of change exceeds the preset threshold, it is determined to be a sudden change in seepage characteristics. Trigger the short window emergency prediction strategy, switch the prediction window from the normal duration to the preset short window duration, and correspondingly increase the data sampling frequency to the preset high frequency value; Using the comprehensive key indicators within the most recent sampling period as input, the state parameters established in the previous prediction process are reused, and the decoding prediction unit is used to generate the predicted value of the upward amount within a preset short window. The upward buoyancy value and arrival time are compared with the conventional prediction results, and the maximum upward buoyancy value and the earliest arrival time are taken as the basis for early warning.

[0032] It should be noted that the inverted siphon pipeline project across a river in a certain city involves a 3.0m diameter reinforced concrete pipe with an axial length of 120m. During the construction phase, which is in the dewatering and excavation stage of the foundation pit, the pipe bottom elevation is 8m below the riverbed, and the target water level for dewatering within the foundation pit is 6m below the riverbed. The riverbed water level fluctuates drastically due to the flood season. To control the risk of pipe floating and seepage damage, the sensors are deployed as follows: Static level: One unit is arranged every 10m along the pipe axis, for a total of 12 units, to monitor the real-time uplift of the pipe body. The normal sampling frequency is 0.5Hz (sampling interval 2s), and the frequency is increased to 1Hz during the excavation stage. Vibrating wire strain gauge: 16 measuring points are arranged on the outer wall of the lower bend, upper bend and middle flat section of the tube to monitor the strain of the tube wall; Piezometers: Eight measuring points are installed inside the foundation pit, on the riverbed outside the foundation pit, and in boreholes on both banks (the depth is flush with the bottom of the pipe) to monitor the water level difference inside and outside the foundation pit; Electromagnetic flow meter: installed on the main pipe of the dewatering well to monitor the water output of the dewatering well.

[0033] Measured data were collected during a specific monitoring period in the flood season (10:00-10:10 on June 8, 20XX, a total of 10 minutes, with 300 sampling points) to conduct monitoring and risk assessment. The basic experimental data are shown in Table 1 below: Table 1. Measured Basic Data Based on the table above, first-order differencing (difference between adjacent time points / sampling interval) is performed on the time-series data to obtain various derived indices, as follows: tube buoyancy d≈ 0.0033 mm / s ; Pipe wall stress change rate s= 2.5 kPa / s ; Variation rate of water output from precipitation wells g= 180 m 3 / h 2 ; Water level fluctuation inside and outside the foundation pit h =5.0−4.0=1m; The buoyancy rate continued to rise (from 0 to 0.0033 mm / s), and the water level difference increased simultaneously (from 4.0 m to 5.0 m), which met the criteria for determining abnormal buoyancy conditions. Therefore, a multi-indicator risk assessment was initiated. Based on engineering experience, normal thresholds and benchmark values ​​for each indicator were set, and the relative risk ratios were calculated, as shown in Table 2 below: Table 2 Relative Risk Ratio Table As shown in Table 2, weighting coefficients are set according to the degree of influence of the indicators on risk: Float rate weight w d =0.3, stress change rate weight w s =0.25, weight of water flow rate variation w g =0.3, water level fluctuation weight w h =0.3, weight sum is 1; Sensitivity coefficient k=2 amplifies the impact of high-risk indicators; Based on the degree of influence of the indicators on risk, weighting coefficients are set, and the comprehensive risk index is calculated by substituting them into the formula. S ; Wd =1.5645; W s =0.8725; W g =1.21; W h =1.3235; Total numerators: 1.5645 + 0.8725 + 1.210 + 1.3235 = 4.970; Sum of denominators: 0.3 + 0.25 + 0.2 + 0.25 = 1.0; The overall risk index S = 4.9705 / 1.0 ≈ 4.97, which is in the high-risk range (S > 4 is set as high risk). The raw monitoring data is filtered to remove noise, normalized, and aligned according to a 2-second time interval to obtain multidimensional monitoring data. The second-order difference of the buoyancy is used to extract the buoyancy acceleration, approximately 1.1 * 10⁻⁶. -5 mm / s 2 The water level fluctuation intensity was obtained by performing a sliding window statistical analysis on the water level difference, with a variance of 0.15m. 2 ; By integrating extended indicators (buoyancy acceleration, water level fluctuation intensity) with the comprehensive risk index S≈4.97, a comprehensive dataset is constructed. Set the historical window length U=30, take the past 60s of data, the prediction window length P=10, predict the next 20s, 10 time steps; Assuming the calculations show that the average attention score for the first 27 time steps is 0.5, and the scores for the last 3 time steps (closest to the current time) are 0.8, 1.0, and 1.2 respectively; ; Comprehensive state characteristics ; initial state ; The decoding unit generates predicted values ​​according to a cyclic formula; Suppose that the uplift prediction sequence generated by the output function Outpu() (current uplift is 3.0mm) is as follows: mm; The prediction curve shows that the ascent rate reaches a peak of 0.007 mm / s at the 8th time step (16s), which requires close monitoring; The monitoring showed that the water level difference between the inside and outside of the foundation pit suddenly increased from 5.0m to 6.0m within 2s, with a change rate of 0.5m / s, which exceeded the preset threshold (0.1m / s) and was judged to be a sudden change in seepage. Prediction window switched to short window (In the next 10 seconds, the sampling frequency will be increased to 1Hz, with an interval of 1 second). Using the monitoring data from the most recent 10 seconds as input, and reusing the trained state parameters, a short-window prediction sequence is generated. ; Compared with the results of the first 5 steps of conventional prediction In comparison, the maximum upward float of 3.20 mm and the earliest arrival time (5 seconds) were taken as the basis for early warning, and emergency measures such as opening more dewatering wells and replenishing water in the foundation pit were immediately initiated.

[0034] S4. Based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, construct a collaborative control model for dynamic matching of the optimal control combination among dewatering, recharge, grouting and ballast equipment, and output and execute the control commands of each equipment.

[0035] Preferably, the construction of the collaborative control model for dynamically matching the optimal control combination among precipitation, reinjection, grouting, and ballast equipment includes: Based on the comprehensive risk index, the current working condition type is identified, and the dewatering, recharge, grouting and ballast equipment are divided into main control equipment set and auxiliary equipment set according to response speed and effect. Based on the predicted nonlinear upward trend of the pipe body, the initial ratio of the required control quantities of the main control equipment and auxiliary equipment is calculated; The comprehensive risk index is mapped to a stepless adjustment coefficient and the initial ratio is corrected to obtain the target control amount for each device. Using the final target control amount as a constraint, a dynamic collaborative mapping relationship is established among precipitation, reinjection, grouting, and ballast equipment, and a collaborative control model is constructed. The expression for its calculation process is as follows: In the formula, Indicates a collection of main control devices; Represents a collection of auxiliary equipment; Represents the initial control quantity allocation vector; This indicates the initial ratio of the required control quantities between the main control equipment and the auxiliary equipment; This represents the peak value of the predicted nonlinear upward trend of the tube. Indicates the predicted peak arrival time; Indicates the stepless adjustment coefficient; This represents the overall risk index; This represents the final target control vector after modification. This represents a specific allocation vector for high-risk operating conditions; Indicates time The device status feedback information; Indicates time Output control commands.

[0036] It should be explained that the main control equipment set M consists of dewatering wells and recharge wells (with fast response speed and water volume adjustment in seconds). Auxiliary equipment set B consists of grouting equipment and ballast equipment (slow response speed, requires advance planning). Based on the predicted peak rise =3.2mm, peak arrival time =5s, through the initial proportion function Calculate the initial control amount; set up: For engineering experience mapping functions, when , At that time, the initial matching vector is: ; Meaning: First, quickly balance the water level difference through precipitation and reinjection; grouting and ballast injection will not be initiated for the time being. Given a comprehensive risk index S = 4.9, normalize it and map it to an adjustment coefficient. =0.994, set the maximum risk index to 5, and linearly map it to the 0~1 range; High-risk working conditions special proportioning vector (Emergency plan for sudden changes in seepage): ; Based on the above parameters, calculate = ; Control command generation ; fb ( t The current equipment status feedback is as follows: The current water output from the dewatering well is 80m³. 3 / h、Recharge well 0m 3 / h, grouting machine 0L / min, ballast equipment 0t / h; For the equipment control mapping function, a tiered control command is generated based on the difference between the target quantity and the current state. Rainwater wells: with a depth of 20m 3 The step size was increased by / h, from 80m in 3 steps. 3 / h increased to 150m 3 / h; Recharge well: at 20m 3 The step size increases by / h, from 0m in 2 steps. 3 / h increased to 60m 3 / h; Grouting equipment: Direct start, grouting at a rate of 20L / min; Counterweight equipment: Start the loader and load sandbags at a rate of 5t / h; The final output control commands are the step-by-step control sequences of the above-mentioned equipment, which are automatically executed by the field PLC system.

[0037] In addition, it should be noted that, such as Figure 2 As shown, the short-window emergency prediction strategy is a rapid prediction mechanism to deal with sudden changes in seepage: when the rate of change of the water level difference inside and outside the foundation pit or the water output of the dewatering well exceeds the preset threshold, the regular prediction window (e.g., 2-6 hours) is automatically switched to a preset short window (e.g., 15-30 minutes), while the data sampling frequency is increased; then, the decoder state parameters established in the previous prediction process are reused, and the comprehensive key indicators in the most recent sampling period are used as input to quickly generate the predicted value of pipe body uplift within the short window. The prediction result and the regular prediction result are output according to the strict principle (taking the larger peak value and the earlier time) as the basis for early warning, thereby realizing an early response to sudden hydraulic risks.

[0038] This embodiment also provides an automatic monitoring system for deformation and seepage during inverted siphon construction, the system comprising: The integrated key indicator perception module is used to acquire comprehensive key indicators in the inverted siphon construction area, including the real-time uplift of the pipe body, the water level difference inside and outside the foundation pit, the water output of the dewatering well, and the pipe wall strain. The risk identification and assessment module is used to identify abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators. Then, it comprehensively evaluates multiple indicators such as the buoyancy rate, stress change rate and water flow rate to obtain the comprehensive risk index. The trend prediction module, based on comprehensive key indicators and comprehensive risk index, uses a dynamic trend extrapolation algorithm to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, it sets a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected. The collaborative control and execution module is used to construct a collaborative control model for dynamically matching the optimal control combination among dewatering, recharge, grouting and ballast equipment based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, and output and execute the control commands of each equipment.

[0039] It should be noted that the calculation formulas and all parameters involved in the calculations in this application have been dimensionless beforehand. The process of dimensionless processing is well known in the industry and will not be described here.

[0040] Although the present application has disclosed the preferred embodiments above, the embodiments are merely examples for the purpose of illustration and are not intended to limit the present application. Those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the present application. The scope of protection claimed by the present application should be determined by the claims.

Claims

1. An automatic monitoring method for deformation and seepage during inverted siphon river crossing construction, characterized in that, The method includes: Obtain comprehensive key indicators within the inverted siphon construction area, including real-time pipe uplift, water level difference inside and outside the foundation pit, dewatering well output, and pipe wall strain. The abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators are identified. Then, the floating rate, stress change rate and water discharge rate are comprehensively evaluated to obtain the comprehensive risk index. Based on comprehensive key indicators and comprehensive risk index, the dynamic trend extrapolation algorithm is used to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, a short-window emergency prediction strategy is set to be automatically triggered when sudden seepage changes are detected. Based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, a collaborative control model is constructed for the dynamic matching of the optimal control combination among dewatering, recharge, grouting and ballast equipment, and outputs and executes the control commands of each equipment.

2. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 1, characterized in that, The method for obtaining the comprehensive risk index is as follows: Based on the real-time pipe body uplift, water level difference inside and outside the foundation pit, water output from dewatering wells, and pipe wall strain in the comprehensive key indicators, abnormal working conditions of the linkage between the foundation pit and the pipe body are identified. In abnormal working conditions of the linkage between the foundation pit and the pipe body, the uplift rate is obtained from the real-time uplift of the pipe body, the stress change rate is obtained from the pipe wall strain, the water output rate is obtained from the water output of the dewatering well, and the water level amplitude is obtained from the water level difference inside and outside the foundation pit. A comprehensive risk index is obtained by comprehensively evaluating multiple indicators, including buoyancy rate, stress change rate, outflow rate change rate, and water level fluctuation.

3. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 1, characterized in that, The method for predicting the nonlinear upward trend of the tube within a preset time period using a dynamic trend extrapolation model is as follows: Comprehensive key indicators Preprocessing is performed to obtain a comprehensive dataset; By analyzing the comprehensive dataset using a dynamic trend extrapolation algorithm, the amount of pipe rising at each moment within a preset time period is predicted, and the rising amount change curve is obtained. Obtain the ascent rate at each moment from the ascent rate change curve, obtain the ascent rate curve, and identify the peak ascent rate and its arrival time.

4. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 3, characterized in that, The process of obtaining the comprehensive dataset includes: The comprehensive key indicators are filtered, denoised, and normalized, and the indicators are time-series aligned according to the same sampling period to obtain multidimensional monitoring data. After performing second-order difference processing on the pipe body uplift amount in the multidimensional monitoring data, the uplift acceleration is extracted, and the water level fluctuation intensity is obtained by using sliding window statistics on the water level difference inside and outside the foundation pit. The extracted buoyancy acceleration, water level fluctuation intensity, and dewatering well output and pipe wall strain are integrated to obtain an extended monitoring index set. This extended monitoring index set is then integrated with the comprehensive risk index to construct a comprehensive dataset.

5. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 4, characterized in that, The method for obtaining the upward buoyancy change curve is as follows: According to the preset historical window length, the continuous monitoring values ​​of each time step are extracted sequentially from the comprehensive dataset to obtain the continuous monitoring value sequence; The continuously monitored numerical sequence is sequentially fed into the Long Short-Term Memory network, and the internal state is updated step by step to obtain the historical internal state sequence. The importance weight of each time step is calculated through a temporal importance assessment mechanism. The historical internal state sequence is then weighted and fused to obtain a comprehensive state feature. This feature is then integrated with the latest internal state and used as the initial state of the decoding and prediction unit. The expression for the calculation process is as follows: In the formula, Indicates the first t Attention weights for each time step; Represents the natural exponential function; Indicates the first t Attention score at each time step; Represents the hyperbolic tangent activation function; Indicates the first One time step; Indicates the first Attention score at each time step; Represents the parameter vector determined through training. transpose; This represents the weight matrix determined through training; This represents the internal state of the Long Short-Term Memory network at the U-th time step; This represents the bias feature determined through training; This indicates the total number of time steps in the history window; Indicates the first t One time step; ) represents the vector concatenation operation; This indicates the initial state of the decoding prediction unit; c Indicates the overall state characteristics; The decoding and prediction unit starts from the initial state, outputs the predicted value of the upward float step by step, and uses the current predicted value as the input for the next step. The cycle continues until the predicted values ​​for all moments within the preset time period are generated, and the future upward float change curve is obtained by arranging them in chronological order.

6. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 5, characterized in that, The process of looping until the predicted values ​​for all moments within the preset time period are generated, and arranging them in chronological order to obtain the future upward fluctuation curve, includes: Let the prediction window length be P, and the decoding prediction unit be... As the initial state, for the th n Step, and n =1, 2, ..., P, the predicted buoyancy values ​​are generated sequentially according to time steps, and the expression for the calculation process of obtaining the future buoyancy change curve by arranging them in time order is as follows: In the formula, P represents the total number of time steps in the prediction window; This represents the output function, used to map the current state to the predicted upward float value; This represents the state update function, which is used to calculate the next state by combining the current state with the current prediction value. Indicates the first P The predicted amount of pipe body buoyancy; Indicates the first Update status after step completion; This represents the curve showing the change in upward fluctuation, which is composed of all predicted values ​​arranged in chronological order.

7. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 6, characterized in that, The aforementioned setting includes a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected, including: Real-time monitoring of the water level difference inside and outside the foundation pit and the rate of change of the water output from the dewatering well. When any rate of change exceeds the preset threshold, it is determined to be a sudden change in seepage characteristics. Trigger the short window emergency prediction strategy, switch the prediction window from the normal duration to the preset short window duration, and correspondingly increase the data sampling frequency to the preset high frequency value; Using the comprehensive key indicators within the most recent sampling period as input, the state parameters established in the previous prediction process are reused, and the decoding prediction unit is used to generate the predicted value of the upward amount within a preset short window period. The upward buoyancy value and arrival time are compared with the conventional prediction results, and the maximum upward buoyancy value and the earliest arrival time are taken as the basis for early warning.

8. The automatic monitoring method for deformation and seepage during inverted siphon river crossing construction according to claim 1, characterized in that, The collaborative control model for dynamically matching the optimal control combination among precipitation, reinjection, grouting, and ballast equipment includes: Based on the comprehensive risk index, the current working condition type is identified, and the dewatering, recharge, grouting and ballast equipment are divided into main control equipment set and auxiliary equipment set according to response speed and effect. Based on the predicted nonlinear upward trend of the pipe body, the initial ratio of the required control quantities of the main control equipment and auxiliary equipment is calculated; The comprehensive risk index is mapped to a stepless adjustment coefficient and the initial ratio is corrected to obtain the target control amount for each device. Using the final target control amount as a constraint, a dynamic collaborative mapping relationship is established among precipitation, reinjection, grouting, and ballast equipment, and a collaborative control model is constructed. The expression for its calculation process is as follows: In the formula, Indicates a collection of main control devices; Represents a collection of auxiliary equipment; Represents the initial control quantity allocation vector; This indicates the initial ratio of the required control quantities between the main control equipment and the auxiliary equipment; This represents the peak value of the predicted nonlinear upward trend of the tube. Indicates the predicted peak arrival time; Indicates the stepless adjustment coefficient; This represents the overall risk index; This represents the final target control vector after modification. This represents a specific allocation vector for high-risk operating conditions; Indicates time The device status feedback information; Indicates time Output control commands.

9. An automatic monitoring system for deformation and seepage during inverted siphon construction, used to implement the automatic monitoring method for deformation and seepage during inverted siphon construction as described in any one of claims 1-8, characterized in that, The system includes: The integrated key indicator perception module is used to acquire comprehensive key indicators in the inverted siphon construction area, including the real-time uplift of the pipe body, the water level difference inside and outside the foundation pit, the water output of the dewatering well, and the pipe wall strain. The risk identification and assessment module is used to identify abnormal working conditions of the pit and pipe body linkage in the comprehensive key indicators. Then, it comprehensively evaluates multiple indicators such as the buoyancy rate, stress change rate and water flow rate to obtain the comprehensive risk index. The trend prediction module, based on comprehensive key indicators and comprehensive risk index, uses a dynamic trend extrapolation algorithm to predict the nonlinear upward trend of the pipe body within a preset time period in the future. At the same time, it sets a short-window emergency prediction strategy that is automatically triggered when sudden changes in seepage characteristics are detected. The collaborative control and execution module is used to construct a collaborative control model for dynamically matching the optimal control combination among dewatering, recharge, grouting and ballast equipment based on the comprehensive risk index and the predicted nonlinear upward trend of the pipe body, and output and execute the control commands of each equipment.

Images