A method for distinguishing sudden surges of a multi-source heterogeneous feature fusion of a deep foundation pit near a river

Abstract

The application provides a method for discriminating sudden gushing of a river-adjacent deep foundation pit by fusing multi-source heterogeneous characteristics, comprising: acquiring multi-source heterogeneous data of a river-adjacent deep foundation pit project; correcting the confined water head according to the dynamic response relationship between the hydrological data and the confined water head, and constructing multi-source heterogeneous characteristics for reflecting the risk of sudden gushing of the foundation pit according to the hydrological data, stratum data, construction data and dynamic feedback data of on-site patrol; determining the reliability of each data source in the multi-source heterogeneous data, correcting the weight of each data source in the fusion according to the reliability, and fusing the multi-source heterogeneous characteristics into a comprehensive risk index; discriminating the risk level according to the mechanism criterion and the comprehensive risk index, and directly determining the dangerous level when the mechanism criterion reaches the dangerous condition or strong abnormalities appear on site. The application overcomes the problem that the static safety factor method is difficult to reflect the dynamic changes of complex working conditions of the river-adjacent, realizes the comprehensive risk discrimination and graded early warning of the river-adjacent deep foundation pit, and improves the construction safety.

Description

Technical Field

[0001] This invention relates to the fields of geotechnical engineering, underground engineering safety monitoring and foundation pit risk early warning technology, specifically providing a method for identifying sudden surges by fusing multi-source heterogeneous features of deep foundation pits near rivers. Background Technology

[0002] Deep foundation pit projects near rivers are widely used in urban underground space development projects such as rail transit, underground complexes, pumping stations, sluice gates, underground parking garages, and municipal utility tunnels. These foundation pits typically have large excavation depths, long construction periods, and are located near rivers with complex hydrogeological conditions. When the foundation pit is near a river, changes in river water levels can be transmitted to the confined aquifer through seepage channels in the riverbed and bank slopes, causing dynamic fluctuations in the confined water head below the pit bottom over time. If the coupling effect of rising confined water head, thinning of the impermeable layer, and increased excavation unloading is not identified in time during the excavation process, it can easily induce safety accidents such as sudden inrush, piping, sand inrush, and pit bottom heave.

[0003] Existing methods for assessing the resistance to sudden inrush in foundation pits mostly employ the static safety factor method, which compares the self-weight resistance of the impermeable or overburden layer below the pit bottom with the uplift force of the confined water. This method is simple in form and convenient for preliminary engineering verification, but it has significant shortcomings in scenarios involving deep foundation pits near rivers. First, river water levels are not constant; they are affected by tides, upstream scheduling, rainfall runoff, and seasonal fluctuations, causing the confined head to fluctuate continuously over time. Using static water levels for assessment results in significant discrepancies with actual conditions. Second, strata are not homogeneous and continuous; weakly permeable interlayers, sand lenses, and poorly sealed boreholes can create localized inrush channels, and the safety factor method is insufficient to reflect the risks of such vulnerable areas. Third, excavation and unloading alter the effective stress and impermeability of the soil at the pit bottom, and the excavation rate, support timing, and dewatering conditions all influence the risk development process; static assessment methods cannot reflect the dynamic impacts of these construction processes. Fourth, on-site inspections and monitoring alarms contain precursory signs of inrushes such as seepage, sand inrush, and pit bottom heave, which have engineering value for early risk identification, but are not quantified and utilized in traditional safety factor methods. In addition, although some technical solutions have attempted to introduce discrimination parameters such as dynamic confined head, soil tensile strength and crack penetration, they are still limited to the analysis of a single mechanism and have failed to incorporate hydrological changes, geological conditions, construction unloading and on-site inspection information into a unified discrimination framework.

[0004] Therefore, there is a need in this field for a method to identify sudden surges by fusing multi-source heterogeneous features in deep foundation pits near rivers, in order to solve the above problems. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, this invention is proposed, which solves or partially solves the problems in existing methods for identifying anti-surge intrusion in deep foundation pits near rivers, namely, the inability to fuse multi-source heterogeneous feature information and the difficulty of the static safety factor method in reflecting the dynamic changes of complex working conditions near rivers.

[0006] This invention provides a method for identifying sudden surges in deep foundation pits near rivers by fusing multi-source heterogeneous features, comprising the following steps: Acquire multi-source heterogeneous data for deep foundation pit projects along rivers. The multi-source heterogeneous data includes at least hydrological data, stratigraphic data, construction data, and dynamic feedback data from on-site inspections. The confined water head is corrected based on the dynamic response relationship between the hydrological data and the confined water head, and a multi-source heterogeneous feature reflecting the risk of sudden inrush in the foundation pit is constructed based on the hydrological data, the geological data, the construction data and the dynamic feedback data from on-site inspections. The reliability of each data source in the multi-source heterogeneous data is determined, the weight of each data source in the fusion is adjusted according to the reliability, and the multi-source heterogeneous features are fused into a comprehensive risk index according to the adjusted weights. The risk level is determined based on the mechanism criteria and the comprehensive risk index. When the mechanism criteria reach the dangerous condition or a strong anomaly occurs on site, it is directly determined to be a dangerous level.

[0007] Preferably, the steps of correcting the confined hydraulic head based on the dynamic response relationship between the hydrological data and the confined hydraulic head, and constructing a multi-source heterogeneous characteristic reflecting the risk of sudden inrush in the foundation pit based on the hydrological data, the geological data, the construction data, and the dynamic feedback data from on-site inspections, specifically include: Cross-correlation analysis was performed on river water level monitoring data and confined head monitoring data to determine the effect coefficient and response lag time of river water level changes on confined head. The dynamic change of the confined head is determined by combining the current change in river water level, the action coefficient, the response lag time, and the intensity of precipitation pumping. The pressure head at the current judgment moment is corrected based on the dynamic change.

[0008] Preferably, the cross-correlation analysis includes: maximizing the correlation coefficient between river water level monitoring data and confined head monitoring data, using the time offset corresponding to the maximized correlation coefficient as the response lag time, and using the regression coefficient corresponding to the maximized correlation coefficient as the action coefficient.

[0009] Preferably, the step of determining the reliability of each data source in the multi-source heterogeneous data and adjusting the weights of each data source in the fusion based on the reliability specifically includes: The reliability is determined based on data integrity rate, sampling time interval, sensor status, outlier ratio, manual review results, and spatial representativeness. When the reliability of a data source decreases, the fusion weight of that data source decreases.

[0010] Preferably, the step of fusing the multi-source heterogeneous features into a comprehensive risk index according to the corrected weights specifically includes: The comprehensive risk index is obtained by weighting and summing the risk feature values ​​corresponding to each data source with the corrected weights.

[0011] Preferably, the step of determining the risk level based on the mechanism criterion and the comprehensive risk index specifically includes: Determine the risk level corresponding to the aforementioned mechanism criterion; Determine the risk level corresponding to the comprehensive risk index; The highest risk level between the risk level corresponding to the mechanistic criterion and the risk level corresponding to the comprehensive risk index is taken as the final risk level.

[0012] Preferably, the mechanism criteria include anti-surge stability criteria, permeability stability criteria, and fracture penetration criteria.

[0013] Preferably, the strong anomaly at the site includes at least one of water inrush, sand inrush, and continuous turbid seepage at the bottom of the pit; when the strong anomaly at the site occurs, it is directly determined to be a dangerous level without relying on the comprehensive risk index.

[0014] Preferably, the method further includes updating the dynamic response relationship and / or the reliability determination method based on the feedback information of the method's judgment result and the actual monitoring result.

[0015] Preferably, the hydrological data includes river water level time history, rise and fall cycle, rise and fall amplitude, rise and fall rate, flood peak duration, rainfall, rainfall intensity, upstream sluice gate scheduling information, lag time of river water level relative to confined head, initial confined head, real-time confined head, head change rate, correlation between confined head and river water level, and head changes caused by the operation of dewatering wells; The stratigraphic data includes the thickness of each stratigraphic layer, the elevation of the bottom layer, the natural unit weight, the buoyant unit weight, the cohesion, the internal friction angle, the compression modulus, the permeability coefficient, the tensile strength, the continuity of the aquitard, the distribution of weakly permeable interlayers, the elevation of the top plate of the confined aquifer, and the recharge boundary of the confined aquifer. The construction data includes the current excavation depth, layer excavation thickness, excavation time, excavation rate, pit bottom exposure time, support installation time, support axial force, dewatering well start / stop status, pumping volume, recharge volume, construction pause time, and completed excavation conditions; The on-site inspection dynamic feedback data includes the location of seepage at the bottom of the pit, changes in seepage volume, color of the effluent, turbidity of the effluent, whether it contains sand, local bulging, cracks at the bottom of the pit, leakage in the retaining structure, surface cracks, abnormal seepage on the riverbank, turbidity of the effluent from the dewatering well, increase in sand content in the drainage ditch, and monitoring alarm information.

[0016] The beneficial effects of this invention are as follows: During the excavation of deep foundation pits near rivers, this invention constructs and integrates a multi-source heterogeneous feature set to identify the risk of sudden surge. It adjusts the confined head in real time based on the dynamic response relationship between river water level and confined head, allowing the confined head to be dynamically updated according to the river's hydrological conditions and the status of precipitation and drainage. This avoids the problem of delayed judgment results caused by water level fluctuations when using a fixed confined head for identification. Simultaneously, by adjusting the fusion weights based on the reliability of each data source, the multi-source heterogeneous feature set is integrated into a comprehensive risk index. This allows water level monitoring data, geological parameters, construction conditions, and inspection feedback information to be incorporated into a unified judgment framework, overcoming the problem of the inability to integrate multi-source heterogeneous information. This achieves comprehensive risk identification for complex riverside conditions, providing a more accurate basis for identifying actual risk states in foundation pit engineering, thereby improving construction safety.

[0017] Furthermore, this invention introduces mechanistic criteria during the judgment process. Risk levels are jointly determined using anti-surge stability criteria, permeability stability criteria, fracture penetration criteria, and a comprehensive risk index. The highest level is taken as the final risk level. This approach retains the physical determinism of the mechanistic criteria while incorporating dynamic information from multi-source heterogeneous characteristics, achieving collaborative judgment driven by both mechanism and data. When any mechanistic criterion reaches a dangerous condition or a strong anomaly occurs on-site, it is directly determined to be at a dangerous level. This ensures timely identification of surge risks and avoids missed risk assessments due to the fusion averaging effect masking sudden strong danger signals.

[0018] Furthermore, this invention incorporates on-site inspection feedback information as an independent feature source into the fusion discrimination. When strong abnormal phenomena such as water inrush, sand inrush, or continuous turbid seepage occur on-site, the danger level determination is directly triggered, so that the precursor signs of sudden inrush such as water inrush and sand inrush can be reflected in the discrimination results in real time, avoiding the discrimination lag caused by relying solely on instrument monitoring data. At the same time, by updating the dynamic response relationship and reliability determination method based on the feedback of subsequent monitoring data and inspection results, the discrimination method can be continuously optimized as construction progresses, realizing rolling discrimination during the construction period.

[0019] Furthermore, this invention further refines the division of key areas such as those near the river, those with significant geological changes, those with undulating tops of confined aquifers, those with thin aquitards, and those with historical leakage anomalies. This ensures that the spatial distribution of risk identification units matches the actual risk distribution of deep foundation pits near rivers, avoiding the underestimation of local risks due to overly coarse division of identification units, and achieving effective identification and key control of the spatial distribution of risks. Attached Figure Description

[0020] The disclosure of this invention will become more readily understood with reference to the accompanying drawings. It will be readily understood by those skilled in the art that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. Furthermore, similar numbers in the drawings are used to denote similar components, wherein: Figure 1 This is a schematic diagram of the main steps of a method for identifying sudden surges in deep foundation pits near rivers according to an embodiment of the present invention, which involves fusing multi-source heterogeneous features.

[0021] Figure 2 This is a schematic diagram illustrating the identification of sudden surge in a deep foundation pit near a river, according to an embodiment of the present invention. Detailed Implementation

[0022] Some 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.

[0023] like Figure 1 As shown, this invention provides a method for identifying sudden surges in deep foundation pits near rivers by fusing multi-source heterogeneous features, comprising the following steps: S100. Obtain multi-source heterogeneous data for the deep foundation pit project along the river. The multi-source heterogeneous data includes at least hydrological data, stratigraphic data, construction data, and dynamic feedback data from on-site inspections. S200. The confined water head is corrected according to the dynamic response relationship between the hydrological data and the confined water head, and a multi-source heterogeneous feature is constructed to reflect the risk of sudden inrush in the foundation pit based on the hydrological data, the geological data, the construction data and the dynamic feedback data of the on-site inspection. S300. Determine the reliability of each data source in the multi-source heterogeneous data, adjust the weight of each data source in the fusion according to the reliability, and fuse the multi-source heterogeneous features into a comprehensive risk index according to the adjusted weight. S400. Determine the risk level based on the mechanism criteria and the comprehensive risk index, wherein when the mechanism criteria reach the dangerous condition or a strong anomaly occurs on site, it is directly determined to be a dangerous level.

[0024] Preferably, in step S100, the aforementioned multi-source heterogeneous data is collected in the following manner: Hydrological data should include at least periodic river level variation data and confined water dynamic variation data. Periodic river level variation data includes river level time history, rise and fall cycle, rise and fall amplitude, rise and fall rate, flood peak duration, rainfall, rainfall intensity, upstream dam and gate scheduling information, and the lag time of river level relative to the confined head. Confined water dynamic variation data includes the initial confined head, real-time confined head, head change rate, correlation between confined head and river level, and head changes caused by dewatering well operation. River levels can be obtained from automatic level gauges, manual water gauges, or hydrological station data; the sampling interval is determined based on the river level change rate. Confined water monitoring wells should preferably be located on the river side, outside the excavation pit, and within the excavation pit's influence zone. The monitoring frequency is determined based on the construction stage; the monitoring frequency should be increased when excavation reaches sensitive depths, when river levels rise rapidly, or when abnormal signs appear.

[0025] Stratigraphic data should include at least the thickness of each stratum, elevation of the bottom of the stratum, natural unit weight, buoyant unit weight, cohesion, angle of internal friction, compression modulus, permeability coefficient, tensile strength, continuity of aquitards, distribution of weakly permeable interlayers, elevation of the top plate of confined aquifers, and recharge boundary of confined aquifers. Stratigraphic characteristic data are derived from exploration borehole data, in-situ testing, and laboratory test results.

[0026] Construction data should include at least the current excavation depth, layer excavation thickness, excavation time, excavation rate, pit bottom exposure time, support installation time, support axial force, dewatering well start / stop status, pumping volume, recharge volume, construction pause time, and completed excavation conditions. This type of data reflects the impact of the gradual removal of the overburden at the pit bottom, the reduction in the self-weight resistance of the overburden layer, and stress release on the risk of sudden inrush. It can be obtained from construction logs, machinery operation records, monitoring platforms, or manual data entry.

[0027] On-site inspection data includes dynamic feedback such as pit bottom seepage, location of water outlets, water color, turbidity, presence of fine sand or silt, localized uplift, pit bottom cracks, leakage in the retaining structure, surface cracks, abnormal riverbank seepage, turbidity of dewatering well water, increased sand content in drainage ditches, and monitoring alarm information. Inspection data should be entered using standardized forms and should ideally include photos, videos, inspection time, and personnel information.

[0028] The aforementioned data sources have different temporal frequencies and spatial scales. River water levels and confined head are generally continuous data at the minute or hour level, formation parameters are usually static spatial data, construction conditions are phased process data, and field inspections are event-type or graded data. This invention transforms these data into feature samples under the same discrimination time and discrimination unit through subsequent preprocessing.

[0029] Furthermore, the river level function is first determined based on historical river water levels, real-time water levels during the construction period, and meteorological and rainfall data. If the river water level has obvious periodicity, it can be expressed by superimposing one or more periodic terms on the base water level; if the river water level is mainly affected by rainfall or upstream scheduling, the measured water level curve and the rate of rise and fall can be directly used in the calculation.

[0030] River water level cycle characteristics include the increment of the current water level relative to the design normal water level, the degree of proximity of the current water level to the warning water level, the increase per unit time, the duration of continuous rise, the amplitude of cycle fluctuations, the predicted peak value, and the remaining time to reach the peak value. These indicators can reflect the strength and sustainability of the recharge of the confined aquifer by the river's lateral boundary.

[0031] The dynamic characteristics of confined water are obtained through confined water monitoring wells. For each discrimination unit, the current confined water head is calculated. Increase in confined water head Rate of change of water head The continuous rise time, the head difference between the confined head and the bottom elevation of the pit, and the response coefficient α and lag time τ between the river water level and the confined head.

[0032] For projects with multiple confined aquifer monitoring wells, the flow direction and hydraulic gradient of confined aquifers can be calculated using the head difference between the monitoring wells in the affected areas on the river side, outside the pit, and inside the pit. If the confined aquifer head near the pit is higher than that far from the pit, and the rate of head change near the bottom of the pit increases, the dynamic risk characteristic value of the confined aquifer for that discrimination unit should be increased.

[0033] Preferably, before step S100, the method further includes: engineering object filing and discrimination unit division.

[0034] Specifically: First, establish an archive of the deep foundation pit project near the river. The archive includes the foundation pit's planar boundary, excavation depth, support structure type, internal support or anchor cable arrangement, dewatering well arrangement, river location, distance between the riverbank and the foundation pit, riverbed elevation, design water level, historical highest water level, historical lowest water level, depth of confined aquifer, location of confined water monitoring wells, and surrounding environmental protection objects.

[0035] To facilitate dynamic assessment, the excavation pit is divided into several assessment units along the horizontal plane. The division principles are as follows: units closer to the river are appropriately densified; areas with significant geological changes are divided separately; areas covered by the influence range of the confined water monitoring wells are considered as the same or adjacent assessment units; and the excavation zones are kept as consistent as possible with the assessment units. Each assessment unit establishes independent data records and risk indices, ultimately yielding the distribution of sudden inrush risk across the entire excavation pit.

[0036] The spatial information of the discrimination unit can be expressed using two-dimensional or three-dimensional coordinates. For engineering practice, the corner points of the foundation pit, the axis of the retaining piles, the riverbank, borehole points, monitoring wells, and inspection points can be unified into the construction coordinate system. If a BIM model, GIS model, or foundation pit monitoring platform exists on site, the discrimination unit can be bound to model components or spatial grids, allowing the risk level to be displayed in the model.

[0037] Furthermore, a unified data base is established for each discrimination unit. This data base includes at least spatial coordinates, the design elevation of the pit bottom, the current excavation elevation, the elevation of the top plate of the confined aquifer, the thickness of each soil layer between the pit bottom and the confined aquifer, the distance to the riverbank, information on the support structure, the influence range of the dewatering wells, and the location of the inspection points. Through this data base, river level monitoring points, confined aquifer monitoring wells, exploration boreholes, excavation condition records, and on-site inspection records are linked to the corresponding discrimination unit.

[0038] Preferably, after step S100, a step of preprocessing the collected multi-source heterogeneous data is further included.

[0039] Specifically, the river water level data undergoes missing value processing, outlier removal, smoothing filtering, and periodic feature extraction. For short-term missing data, interpolation between adjacent time periods can be used to fill in the gaps; for data with abrupt sensor fluctuations or that clearly does not conform to the hydrological process, corrections are made by combining manual water gauge records or data from adjacent monitoring points. After processing, the amplitude and rate of rise and fall of river water levels, the dominant frequency of the period, the duration of the flood peak, and the degree to which the current water level exceeds the warning level are extracted. For suspected abnormal data, priority is given to verification through adjacent measuring points, manual records, and on-site inspection records; if the verification confirms anomalies, the data is not included in parameter fitting and risk calculation, or interpolated values ​​are used as substitutes.

[0040] Synchronize data at different frequencies. Set a discrimination time step Δt, such as 10 min, 30 min, 1 h, or determine according to project needs. Map river water level, confined head, rainfall, construction conditions, and inspection events to the discrimination time. For high-frequency monitoring data, the mean, maximum value, rate of change, and peak value within the time window can be used; for low-frequency inspection data, its status can be maintained for the effective duration until the next inspection or review update.

[0041] Spatial data is mapped. Boreholes, monitoring wells, inspection points, and construction zones are mapped to discrimination units. Formation parameters can be obtained through interpolation from nearby boreholes, inference from layered profiles, or assignment by engineering geological zones. Confined water head can be spatially weighted based on monitoring well distance, aquifer connectivity, and seepage direction. Inspection events are bound to corresponding discrimination units according to their location, and influence attenuation is applied to adjacent units.

[0042] Perform head conversion, datum unification, and lag response analysis on the dynamic data of confined water. The recorded values ​​of pressure level gauges should be uniformly converted to the confined water head under the same elevation datum, and correlation analysis should be conducted with the river water level sequence to determine the river water-confined water response coefficient and lag time. When the confined water head is significantly affected by the operation of dewatering wells, the pumping rate, well start / stop status, and dewatering influence coefficient should be included in the correction.

[0043] Spatial interpolation and weak layer identification were performed on the stratigraphic data. Based on borehole data, in-situ tests, and laboratory test results, stratigraphic columnar models were established for each discrimination unit to determine the remaining overburden thickness, equivalent thickness of the aquitard, location of weakly permeable interlayers, permeability coefficient zones, and soil strength parameters between the pit bottom and the top of the confined aquifer. For areas with large borehole spacing or significant stratigraphic undulations, the aquitard thickness was adjusted according to the least favorable principle.

[0044] The excavation and unloading data are processed into a construction time sequence. Each excavation, support installation, dewatering well start / stop, and pit bottom exposure time is converted into a time axis event and aligned with the judgment time t to obtain the current excavation depth, remaining impermeable layer thickness, excavation rate, support lag time, and dewatering operation status.

[0045] The on-site inspection data is categorized and spatially matched. Text descriptions, photographic records, and manual judgments are converted into standardized inspection levels. Different phenomena, such as "small amount of clear water seepage," "continuous seepage of turbid water," "water discharge with sand," "pit bottom bulging," and "sudden water inrush," are assigned values ​​and matched to the corresponding discrimination unit based on their location. If the same discrimination unit shows anomalies multiple times within a short period, its on-site feedback risk value is increased.

[0046] In a preferred embodiment, step S200, "correcting the confined head based on the dynamic response relationship between the hydrological data and the confined head," specifically includes: S210. Synchronize the river water level monitoring data with the confined head monitoring data in time to determine the lag time between the river water level change and the confined head response. S220. The dynamic change of the confined head is determined by combining the change in river water level, lag time, and precipitation pumping intensity. S230. Correct the pressure head at the current judgment time based on the dynamic changes.

[0047] In step S210, the river water level monitoring data and the confined aquifer head monitoring data are aligned according to the discrimination time, and the river water-confined aquifer response coefficient and lag time are determined through correlation analysis. Since it takes time for changes in river water level to propagate to the confined aquifer, determining the lag time is crucial for accurately reflecting the true changes in the confined aquifer head. If the river water level has risen but the confined aquifer head has not yet increased significantly, the lag time can indicate subsequent risks; if the river water level has receded but the confined aquifer head remains high, it indicates that the confined aquifer discharge is delayed, and the risk of sudden inrush may still persist.

[0048] In step S220, the dynamic change of the confined head is determined in the following way: taking the initial confined head as a reference, the change in river water level at the current moment is obtained; the driving term of the river water level change on the confined head is determined according to the response coefficient between the change in river water level and the confined head; the driving term is delayed according to the lag time; and the weakening term of precipitation pumping on the confined head is determined according to the precipitation pumping intensity; and the dynamic change of the confined head at the current judgment moment is obtained by combining the above items.

[0049] In step S230, the initial confined head is superimposed with the dynamic change to obtain the corrected confined head. Through this method, the confined head can be updated in real time according to changes in river water level and precipitation pumping status, avoiding the problem of delayed judgment results caused by water level fluctuations when using a fixed confined head for judgment.

[0050] In a preferred embodiment, step S200, "constructing multi-source heterogeneous characteristics to reflect the risk of foundation pit inrush based on the hydrological data, the geological data, the construction data, and the dynamic feedback data from on-site inspections," specifically includes: Stratigraphic characteristics are the basis for assessing the risk of sudden inrush. For each assessment unit, the soil layer combination between the pit bottom and the top of the confined aquifer is determined, including the thickness, buoyant unit weight, permeability coefficient, cohesion, internal friction angle, tensile strength, and compression modulus of each layer. If a continuous thick layer of cohesive soil exists below the pit bottom, the water-retaining capacity is relatively strong; if silt, fine sand, sand lenses, weak interlayers, or poorly sealed borehole sections exist, the water-retaining capacity is reduced.

[0051] This invention defines the thickness of the remaining waterproof layer. This refers to the thickness of the soil layer between the bottom of the excavated pit and the top of the confined aquifer, which can effectively resist sudden heave. As the excavation depth increases... Gradually decrease. If a local weak area exists in a certain discrimination unit, then The minimum effective thickness for control within the unit should be taken, or it should be reduced proportionally to the area of ​​the weak zone.

[0052] The water-impermeability characteristics of a formation can be composed of multiple sub-indicators, including thickness indicators. Severe indicators Penetration indicators Intensity indicators and continuity indicators The smaller the thickness, the larger the permeability coefficient, the lower the strength, and the worse the continuity, the higher the risk of sudden surge. The above sub-indicators are normalized to obtain the formation risk characteristics. .

[0053] For confined water surges, in addition to overall anti-buoyancy or surge stability, local hydraulic fracturing and fracture penetration should also be considered. If the tensile stress generated by the confined water pressure at weak points exceeds the tensile strength of the soil, or if the tensile stress generated by the bending deformation at the bottom of the pit, combined with the water pressure, reaches the conditions for fracture development, a hydraulic channel may form. This invention allows for the tensile strength... The fracture penetration criterion is used as a hard criterion in the discrimination.

[0054] Stratigraphic characteristic data generally originates from the exploration stage, but during the construction stage, it can still be corrected through supplementary drilling, pit sampling, particle analysis of dewatering well water, and inversion of monitoring data. If on-site inspections reveal continuous turbid water or sand inrush at a certain location, and the original exploration data indicates that the aquitard layer at that location is relatively thick, it should be considered that there is an unrevealed weak channel in the area, and the stratigraphic continuity index should be reduced accordingly.

[0055] Excavation depth ratio It can be defined as the current excavation depth D(t) and the designed final excavation depth. The ratio of unloading stress. It can be defined as the unloading stress corresponding to the excavated soil. Total stress of the overlying soil at the bottom of the pit before excavation The ratio of the two values. The excavation rate influence coefficient. The risk characteristic value can be determined based on the incremental excavation depth per unit time. The faster the excavation, the less sufficient the dissipation of soil pore pressure and the adjustment of the support system, and the greater the risk characteristic value.

[0056] Support lag effect coefficient This is used to reflect situations where the supports fail to provide effective restraint in a timely manner after the excavation face exceeds the designed support installation elevation. If the design requires immediate support construction after excavation to a certain elevation, but the actual support installation is delayed, the deformation of the retaining structure and the redistribution of stress at the bottom of the pit may be exacerbated, and the unloading risk characteristic should be improved.

[0057] Rainfall Operation Influence Coefficient This indicator reflects the reliability of depressurization wells, pressure relief wells, or drainage systems in controlling confined water heads. Insufficient pumping, pump shutdown, rising head in monitoring wells, or increased sand content in the discharged water indicate a decline in the depressurization system's ability to control confined water, and the risk characteristic value should be increased.

[0058] Excavation and unloading characteristics can be to The weighted average is obtained. For deep foundation pits near rivers, when the river level rises rapidly and the excavation unloading increases rapidly at the same time, a coupling amplification term can be set to make the comprehensive risk index more sensitive to this unfavorable combination.

[0059] On-site inspections are a crucial step in identifying precursory water surges. Some water surge risks may appear as slow changes in monitoring data, but on-site, they may first manifest as localized seepage at the bottom of the pit, turbid effluent, fine sand precipitation, increased sand content in drainage ditches, bulging or cracks at the bottom of the pit, etc. Failure to incorporate these information into the assessment can easily lead to delayed early warnings.

[0060] This invention quantifies on-site inspection information into The inspection checklist may include the following items: The state of seepage at the bottom of the pit is assigned values ​​based on no seepage, wetness, linear seepage, and concentrated water outflow. The turbidity of the effluent is assigned as follows: clear water, slightly turbid, obviously turbid water, and sand-laden water. For cases of sand inrush or particle carryover, assign values ​​according to whether they are absent, small amount, continuous, or rapid. For pit bottom bulges or cracks, assign values ​​as none, slight, obvious, or rapidly developing; For abnormalities in the support structure or precipitation system, assign values ​​as no abnormality, slight abnormality, obvious abnormality, or severe abnormality.

[0061] Each inspection item can be assigned a weight. For sudden inrush detection, sand inrush, sand-laden muddy water, concentrated water inrush, and rapid heaving of the pit bottom usually have higher weights; general wetness or seepage of a small amount of clear water has lower weights. If the inspection record is accompanied by photos or videos or reviewed by two or more people, the reliability increases; if there is only a verbal description or the location is unclear, the reliability decreases.

[0062] On-site inspection feedback is used not only to calculate the comprehensive risk index but also to correct other characteristics. When turbid water or sand-laden water is observed on-site but the pressurized head does not increase significantly, it may indicate that the monitoring well location does not cover a local channel or that there are unrevealed weak interlayers in the formation; in this case, the formation continuity risk can be increased and more frequent monitoring is recommended. When the sand content in the dewatering well water increases on-site, the dewatering operation impact coefficient can be increased, and it is recommended to check the filter pipes, well wall sealing, and pumping intensity.

[0063] Specifically, for each discrimination unit u and discrimination time t, a multi-source heterogeneous feature vector is constructed: In the formula, This refers to the periodic changes in river water levels. The dynamic characteristics of confined water Characteristics of the formation's water-retaining capacity. For excavation and unloading characteristics, These are dynamic feedback features from on-site inspections. All of the above features have been normalized to ensure that data of different dimensions can participate in unified fusion calculations.

[0064] Characteristics of periodic changes in river water level The river level is determined by its relative height, fluctuation amplitude, rate of rise, dominant frequency of the period, duration of the flood peak, and intensity of rainfall disturbance. A higher rate of water level rise, a longer flood peak duration, and a current water level closer to or exceeding the warning level all indicate a greater likelihood of flooding. The larger the value, the greater the value. For tidal channels, the duration of high tide and the amplitude of tidal range are also extracted; for channels affected by dam and gate operations, the amplitude of water level fluctuations caused by dam and gate releases is also extracted.

[0065] Dynamic characteristics of confined water It is composed of confined water head height, head increment, head rise rate, river water-confined water response coefficient, lag time, and the degree of influence of dewatering wells. Higher confined water head, faster rise rate, stronger correlation between river water and confined water, and poorer control effect of dewatering wells all contribute to this effect. The larger the value, the better.

[0066] In the formula, This is the corrected pressure head; This is the initial confined head; To take lag time into account The subsequent river water level; The baseline river water level; The response coefficient between river water and confined water; This refers to rainfall or upstream water disturbance. This is a correction factor for rainfall replenishment. The equivalent pumping intensity of a dewatering well or a pressure-reducing well; This represents the impact coefficient of the drainage process. This treatment allows the confined water discrimination value to dynamically change with river water level and construction precipitation conditions.

[0067] Characteristics of formation water-impermeability The remaining overburden thickness, equivalent buoyant weight, equivalent cohesion, equivalent tensile strength, permeability coefficient, degree of development of weakly permeable interlayers, and degree of undulation of the confined aquifer's top plate are all factors that contribute to its strength. The thinner the remaining overburden, the higher the permeability coefficient, the more developed the weakly permeable interlayers, and the closer the confined aquifer's top plate is to the pit bottom, the better. The larger the value, the greater the effective resistance of the pit bottom cover layer. For multi-layer cover layers, the equivalent resistance of the pit bottom cover layer is calculated based on the effective thickness and mechanical parameters of each layer.

[0068] In the formula, The equivalent resistance of the pit bottom cover layer; Let be the buoyant unit weight of the i-th soil layer; To determine the effective thickness of the i-th layer at time i; Equivalent cohesion of the capping layer; This represents the equivalent tensile strength of the coating layer. and These are the reduction factors for cohesion and tensile strength, respectively.

[0069] In the formula, The upward thrust of the pressurized water; The specific gravity of water; This refers to the water head corresponding to the pit bottom elevation or the benchmark surface for determining sudden inrush. When there are potential channels such as localized weak interlayers, poorly sealed exploration boreholes, or old wells in the strata, [this is used for...]. Reduce the amount.

[0070] Among them, the excavation unloading influence coefficient The following formula is used to determine it: In the formula, This is the current excavation depth; To design the final excavation depth; This refers to the amount of vertical stress released due to excavation unloading. The total vertical stress at the bottom of the pit before excavation; This refers to the excavation rate; To support the impact of delayed or exposed pit bottom items; , , and For the corresponding weights.

[0071] On-site inspection dynamic feedback characteristics The severity and duration of the anomaly are determined by its level and persistence. A small amount of clear water seepage corresponds to a low-level feedback; continuous seepage of turbid water, increased water flow, and increased sand content in drainage ditches correspond to medium-to-high-level feedback; water discharge with sand, pit bottom heaving, crack expansion, sudden water inrush, or localized quicksand correspond to strong anomaly feedback. Strong anomaly feedback not only participates in the comprehensive index calculation but also serves as a hard trigger condition for risk level assessment.

[0072] Of course, the specific composition of the above five risk characteristics is only exemplary. Those skilled in the art can add or adjust the characteristic sub-items according to the actual situation of the project, as long as they can fully reflect the risk of sudden inrush in the foundation pit.

[0073] Different data sources have varying degrees of integrity and reliability; therefore, the reliability of each type of data source is calculated before merging. Reliability is determined by a combination of factors, including data integrity, sampling time interval, sensor status, outlier rate, manual verification results, and spatial representativeness. Reliability decreases when sensors fail, data is missing for extended periods, the measurement point is too far from the discrimination unit, or the inspection record lacks photographic evidence; reliability increases with manual verification or cross-verification from multiple measurement points.

[0074] In a preferred embodiment, reliability The reliability of a data source is expressed as a score, ranging from 0 to 1. Data sources with online sensors, complete measurement points, minimal time delay, data fluctuations conforming to physical laws, and manually verified are considered highly reliable. Data sources with faulty sensors, numerous missing measurements, incomplete manual descriptions, or contradictions with other data are considered less reliable. As the reliability of a data source decreases, its fusion weight decreases. When serious anomalies are discovered during on-site inspections and confirmed through verification, both the reliability and weight of the on-site feedback features can be increased. Therefore, the comprehensive surge risk index I(t) can adapt to changes in data quality.

[0075] Reliability-corrected dynamic weights The following formula is used to determine it: In the formula, The basic weights of the j-th type of features; Let be the reliability of the j-th type of data source; The dynamic weights are adjusted for reliability, and m is the total number of features involved in the fusion calculation.

[0076] Basic weights The weighting can be adjusted based on the importance of the project. For example, in projects where river water levels have a significant impact and the confined aquifer is highly connected to the river channel, the weighting of river water level and confined water dynamic characteristics should be increased; in projects with drastic geological changes or sand lenses, the weighting of geological characteristics should be increased; and in projects with rapid excavation speeds and complex support transitions, the weighting of excavation unloading characteristics should be increased. The sum of the basic weights is 1.

[0077] In a preferred embodiment, step S300, "integrating multi-source heterogeneous data into a comprehensive risk index based on the modified weights," specifically includes: weighting and summing the risk characteristic values ​​corresponding to each data source using the modified weights to obtain the comprehensive risk index. The comprehensive risk index... The following formula is used to determine it: In the formula, The risk characteristics after normalization; ξ is the hard trigger term, which takes a larger value when the mechanism criterion is dangerous or strong abnormal feedback occurs on site; ξ is the amplification factor of the hard trigger term.

[0078] In this way, when a certain type of data is missing or unreliable, its impact is automatically reduced, but the surge detection will not be interrupted; when a strong anomaly occurs on site, the risk index can be quickly increased.

[0079] Multi-source fusion is not a simple averaging process, but rather follows the rules of "prioritizing mechanistic criteria, adjusting dynamic weights, and triggering on-site anomalies." River water level and confined head are used to determine the external hydraulic drive, geological features are used to determine the pit bottom resistance capacity, excavation unloading is used to determine the intensity of construction disturbance, and on-site inspections are used to verify the consistency between the model's judgment and the actual on-site conditions. These five types of characteristics jointly determine the comprehensive surge risk index.

[0080] In a preferred embodiment, step S400, "determining the risk level based on the mechanism criterion and the comprehensive risk index," specifically includes: S410. Determine the risk level corresponding to each mechanism criterion and the risk level corresponding to the comprehensive risk index respectively; S420, take the highest level among them as the final risk level.

[0081] In step S410, the mechanism criteria include the anti-surge stability criterion, the permeability stability criterion, and the fracture penetration criterion.

[0082] The surge resistance stability criterion is based on the surge resistance safety factor. The expression is based on the equivalent resistance of the remaining overburden at the bottom of the pit. Upward jacking force of pressurized water The ratio is determined as follows: when When the value is less than a preset safety factor threshold, the anti-surge stability criterion is deemed to have reached a dangerous condition. The preset safety factor threshold can be determined based on the engineering grade, geological conditions, and relevant specifications.

[0083] It should be noted that when there are potential channels such as localized weak interlayers, poorly sealed exploration boreholes, or old wells in the strata, the equivalent resistance of the remaining overburden at the bottom of the pit should be considered. The reduction is performed. Specifically, if a local weak area exists in a certain discrimination unit, the remaining waterproof layer thickness is... The minimum effective thickness for control within the unit should be taken, or reduced proportionally to the area of ​​the weak zone.

[0084] The seepage stability criterion is based on the actual hydraulic gradient between the bottom of the pit and the confined aquifer. The expression is determined by dividing the difference between the confined water head and the pit bottom elevation by the remaining overburden thickness: in, The remaining cover thickness is between the bottom of the pit and the top of the confined aquifer.

[0085] like Approaching or exceeding the critical hydraulic gradient This indicates that fine particles may be carried away by seepage, posing a risk of piping or quicksand. The critical hydraulic gradient can be determined based on the buoyant unit weight of the soil, the unit weight of water, and the fine particle content, or it can be determined by combining specifications or test results.

[0086] Criteria for fracture penetration This is used to determine whether confined water pressure may induce tensile cracking in the overburden layer at the bottom of the pit and the upward propagation of cracks. For soil layers that may undergo hydraulic fracturing under the action of confined water, the relationship between confined water pressure, lateral earth pressure, bending tensile stress at the bottom of the pit, and tensile strength of the soil is determined. When the combined effect of water pressure and tensile stress causes cracks to propagate from the confined water side towards the bottom of the pit and penetrate to the excavation face at the bottom of the pit, it is determined to be a crack penetration hazard.

[0087] The above-mentioned mechanistic criteria are hard criteria. As long as any hard criterion meets the dangerous conditions, even if the comprehensive risk index has not yet reached the highest level, the judgment unit should be upgraded to the warning or danger level; if strong abnormal phenomena such as sandy water discharge, continuous turbid water, and pit bottom heave are found during on-site inspection, it is directly judged that there is a risk of sudden surge or imminent surge.

[0088] Preferably, the hard criteria have priority. When Less than the lower threshold, or Greater than ,or When the conditions for connection are met, or when strong anomalies such as continuous sand inrush, sudden water inrush, or muddy water carrying sand are observed during on-site inspections, the danger level will be directly output even if the comprehensive risk index has not yet reached the danger threshold. This setting can avoid the fusion averaging effect from masking sudden strong anomalies.

[0089] In one embodiment, the final risk level is determined jointly based on hard criteria and a comprehensive risk index. The risk of sudden inrush can be divided into four levels: safe, concerning, warning, and dangerous. The safe level indicates that the pressure from the confined water is significantly less than the resistance of the overburden layer at the bottom of the pit, and there are no abnormalities on site. The concerning level indicates that the river water level or confined water head is trending upwards, or local geological parameters are unfavorable, but the mechanistic criteria are still met. The warning level indicates that the safety factor against sudden inrush is close to the threshold, the hydraulic gradient is too large, the impact of excavation unloading is significant, or there are abnormalities such as continuous turbid water or increased seepage on site. The dangerous level indicates that the hard criteria are not met, the possibility of fracture penetration is high, or phenomena such as sand-bearing water discharge, pit bottom heave, sand inrush, or sudden water inrush have occurred on site.

[0090] Specifically, the level corresponding to the comprehensive index is determined based on I(t). For ease of calculation, the values ​​for Safety, Attention, Warning, and Danger are assigned to 0, 1, 2, and 3 respectively. The corresponding levels of the comprehensive index are then determined according to the table below.

[0091] Where I(t) < 0.45 indicates a lower overall risk, corresponding to =0; 0.45≤I(t)<0.65 indicates that risk factors have begun to accumulate, corresponding to =1; 0.65≤I(t)<0.80 indicates a significant risk of sudden surge, corresponding to =2; I(t)≥0.80 indicates that the comprehensive risk index has reached a dangerous level, corresponding to =3. The above thresholds can be calibrated based on engineering grade, geological conditions, monitoring samples, and historical treatment experience.

[0092] At the same time, the corresponding levels of the anti-surge safety factor are determined based on the mechanism criteria and on-site inspection results. Hydraulic gradient corresponding level Criteria for determining fracture penetration and corresponding levels Corresponding level to on-site inspection feedback Among them, when the anti-surge safety factor is lower than the preset safety threshold, The danger level is determined by the following: when the actual hydraulic gradient exceeds the critical hydraulic gradient. Determine the hazard level; when the crack penetration criterion meets the hazard conditions. The danger level is determined by the following: when severe abnormal phenomena occur on site, such as sandy water discharge, continuous turbid water, pit bottom heave, sand inrush, or sudden water inrush, the danger level is determined by the following: Determine the danger level.

[0093] Final risk level Determined jointly by the hard criterion level and the comprehensive index level: In the formula, The final risk level; The corresponding level of the anti-surge safety factor; This refers to the corresponding level of hydraulic gradient; The corresponding level for determining the continuity of the fracture; Provide feedback on the corresponding level during on-site inspections; This is to determine the corresponding levels of the comprehensive surge risk index. Adopting the highest level principle avoids the comprehensive average masking of a single strong danger signal.

[0094] when When =0, it is determined to be a safe level, and normal excavation and routine monitoring should be maintained; when When =1, it is judged as a level of concern, the frequency of monitoring of confined water head and inspection of the pit bottom is increased, and the operating status of the dewatering wells is reviewed; when When the value is 2, it is determined to be a warning level. The excavation speed should be controlled, the judgment time step shortened, and if necessary, a pressure relief well should be activated, monitoring intensified, or partial plugging implemented. When the value is 3, it is determined to be a dangerous level. Excavation in the relevant area should be suspended, and measures such as reducing the pressure of the water, applying back pressure to the bottom of the pit, grouting and sealing, backfilling and weighting, or emergency drainage should be taken.

[0095] when , , or If any of the following conditions is met, the final risk level should be adjusted even if I(t) has not yet reached 0.80. It is classified as a hazardous level. This rule reflects the judgment principle of "prioritizing mechanistic criteria and prioritizing strong on-site anomalies".

[0096] In step S400, the logic for comprehensive judgment is as follows: S401. Determine whether the anti-surge safety factor is less than the preset safety factor threshold; if yes, directly determine it as a dangerous level; if no, proceed to S402. S402. Determine whether the actual hydraulic gradient is greater than the critical hydraulic gradient; if so, directly determine it as a dangerous level; if not, proceed with S403. S403. Determine whether the criteria for the penetration of the fracture at the bottom of the pit meet the dangerous conditions; if yes, directly determine it to be dangerous; if no, proceed with S404. S404. Determine if a strong anomaly has occurred at the scene; if so, directly classify it as a danger level; if not, proceed to S405. S405, Based on the comprehensive risk index The numerical value determines the corresponding risk level.

[0097] Among these, strong anomalies on-site include at least one of the following: water inrush at the bottom of the pit, sand inrush, and continuous turbid seepage. When any of these strong anomalies occur on-site, regardless of the value of the comprehensive risk index, they are directly classified as hazardous. This setting reflects the judgment principle of "prioritizing mechanistic criteria and prioritizing strong on-site anomalies," and can avoid the masking of sudden strong hazard signals by the fusion averaging effect.

[0098] It should be noted that the order of steps S401 to S405 above is merely illustrative and not restrictive. In practical applications, the judgment steps can be executed in parallel or in different orders, as long as the risk level is ultimately output according to the highest level principle.

[0099] Furthermore, when the final risk level is at the safe level, normal excavation and routine monitoring will be maintained; when the final risk level is at the concern level, the frequency of confined water head monitoring and pit bottom inspection will be increased, and the operation status of dewatering wells will be reviewed; when the final risk level is at the warning level, the excavation speed will be controlled, the judgment time step will be shortened, and if necessary, pressure relief wells will be activated, monitoring will be intensified, or local sealing will be carried out; when the final risk level is at the danger level, excavation in the relevant area will be suspended, and measures such as reducing confined water, backpressure at the pit bottom, grouting and sealing, backfilling and weighting, or emergency drainage will be taken.

[0100] In a preferred embodiment, the method further includes: after obtaining the final risk level, calculating the contribution value of each feature to the comprehensive risk index, and identifying the controlling factors. One or more features with the largest contribution values ​​are identified as the controlling factors. For example, if... and A high contribution value indicates that the risk is mainly caused by rising river levels and confined head response. Priority measures should be given to lowering the confined head, increasing water level monitoring, and paying close attention to river channel management information. If... A high contribution value indicates that the risk is mainly caused by a thin impermeable layer, weak interlayer development, or localized weak channels. Priority should be given to measures such as bottom reinforcement, grouting sealing, or adjusting excavation zones. A high contribution value indicates that the risk is mainly caused by excessively rapid excavation and unloading, delayed support, or prolonged exposure of the pit bottom. In such cases, the construction pace should be adjusted first, and the pit bottom should be sealed promptly. A high contribution value indicates that the on-site anomaly is obvious, and manual verification and emergency response should be prioritized.

[0101] The recommendations are not fixed rules, but are automatically generated based on a combination of key controlling factors. For example, if the main control factor is rising river water levels, recommendations may include monitoring river channel management, increasing the frequency of confined water monitoring, pre-lowering confined water levels, or delaying excavation. If the main control factor is rising confined water head, recommendations may include activating pressure relief wells, checking pumps, and verifying pumping volume and head reduction. If the main control factor is weak geological formations, recommendations may include reinforcing the pit bottom, sealing weak passages, or conducting supplementary exploration. If the main control factor is excavation unloading, recommendations may include reducing the thickness of layered excavation, lowering the excavation rate, and timely installation of supports. If the main control factor is abnormal inspection findings, recommendations may include immediately sealing outflow points, laying a filter layer, backfilling with ballast, and organizing expert review.

[0102] When "river water level rises + confined water head increases + pit bottom seepage water becomes turbid" occurs simultaneously, it is recommended to increase the operating intensity of depressurization wells, suspend excavation on the riverside side, and strengthen pit bottom inspections; when "impermeable layer thins + excavation approaches the foundation + support lags behind" occurs simultaneously, it is recommended to reduce the thickness of single-layer excavation, shorten the pit bottom exposure time, and pre-reinforce weak areas; when "dewatering well water contains sand + confined water head control fails" occurs simultaneously, it is recommended to check well pipe damage, filter material failure, and sealing quality.

[0103] In a preferred embodiment, the method further includes: during construction, continuously writing back subsequent monitoring data, inspection results, and treatment effects to the data base, and updating the river water-confined water response coefficient, lag time, precipitation impact coefficient, unloading impact coefficient, inspection level assignment, and risk classification threshold. If precipitation measures are taken after a warning, resulting in a significant decrease in the confined water head and the disappearance of on-site anomalies, the confined water driving characteristics in subsequent judgments are correspondingly reduced; if turbid water or sand-laden phenomena repeatedly occur in a certain judgment unit, the on-site inspection feedback weight of that unit is increased. Through dynamic feedback, this invention can achieve continuous application from static review in the design stage to real-time rolling judgment in the construction stage.

[0104] In a preferred embodiment, the method further includes: the surge risk assessment is not calculated all at once, but is updated continuously throughout the construction process. This invention writes each assessment result, actual monitoring changes, on-site inspection verification, and handling effectiveness into a feature database.

[0105] If precipitation measures are taken after an early warning to reduce the confined head, the precipitation operation impact coefficient is updated; if the peak lag time of the confined head after the river level rises is different from the original estimate, the lag time τ is updated; if seepage or turbid water occurs repeatedly in a certain area, the stratigraphic continuity reduction coefficient of that area is updated.

[0106] Dynamic feedback can be implemented through manual or automatic verification. Manual verification involves technicians confirming whether model parameters need adjustment based on site conditions; automatic verification corrects the river water-confined water response coefficient, drainage influence coefficient, inspection weight, and risk threshold based on the discrepancy between monitoring and predicted data. Through dynamic feedback, the judgment method can adapt to the constantly changing engineering conditions during the construction period.

[0107] The project management platform can display the assessment results using color, level, or text prompts. A safety level indicates a normal status; a monitoring level indicates enhanced monitoring; a warning level indicates on-site handling and technical review; and a hazard level indicates suspension of excavation and activation of the emergency response plan. The platform also saves the assessment criteria for future reference.

[0108] The method of the present invention will be described below with reference to specific engineering examples.

[0109] Example 1 A deep foundation pit for an underground structure located near a river was excavated to a depth of 18 meters, with one side approximately 45 meters from the riverbank. Survey data indicated the presence of a silty clay impermeable layer and a fine sand confined aquifer below the pit bottom. During construction, the river water level was significantly affected by upstream water management and rainfall, and the confined aquifer monitoring wells showed a delayed rise in water head. During implementation, river water level monitoring points were set up on the riverbank side, and multiple confined aquifer monitoring wells were deployed along the river direction outside the foundation pit. The density of detection units was increased in the pit section closest to the river. The construction unit recorded the excavation depth, support installation status, and dewatering well operation status daily. Inspection personnel used mobile terminals to report pit bottom seepage, effluent turbidity, and sand content in the drainage ditch.

[0110] When the excavation reached below the second support, the river level rose continuously due to rainfall, and the confined head began to rise after a lag of several hours. The system corrected the confined head according to step S200, and the corrected confined head was significantly higher than the initial value. The anti-surge safety factor had not yet fallen below the hard criterion threshold, and the comprehensive risk index entered the attention level. The system prompted for increased monitoring of the confined head and required on-site inspections to be increased to once per shift.

[0111] Subsequently, excavation of the foundation pit continued, reducing the thickness of the remaining impermeable layer at the bottom of the pit and increasing the unloading stress ratio. A site inspection by a certain judgment unit revealed localized linear seepage at the bottom of the pit, with slightly turbid effluent. The system adjusted the on-site feedback characteristics from low risk to medium risk and improved the reliability of the data source; the comprehensive risk index rose to the warning level. The main controlling factors were identified as the dynamic increase in confined water head and the superposition of excavation unloading. Based on this, the excavation rate was reduced, a backup pressure relief well was activated, and reverse filtration was implemented at the seepage point. After the treatment, the confined water head gradually decreased, the seepage point did not continue to expand, and the comprehensive risk index returned to the concern level. The system recorded the effect of this treatment and updated the drainage impact coefficient and the inspection feedback weight. This embodiment demonstrates that the present invention can identify the risk accumulation process before a sudden surge occurs and uses monitoring data, construction status, and inspection feedback together for risk judgment.

[0112] Example 2 In another construction phase, a certain discrimination unit was excavated to near the design pit bottom, and the confined water head rapidly increased due to rising river water levels and a malfunction in the dewatering well. In step S401, the anti-surge safety factor was lower than the preset safety factor threshold, and in step S404, on-site inspection revealed continuous fine sand particles in the drainage ditch. This invention directly outputs the hazard level based on steps S401 and S404, without waiting for the smoothing result of the comprehensive risk index. After the hazard level is output, the system simultaneously provides the main control factors: dewatering failure, rapid rise in confined water head, insufficient remaining cover thickness at the pit bottom, and signs of sand inrush on-site. Corresponding handling suggestions include suspending excavation in the area, resuming or increasing decompression well pumping, setting up reverse filter pressure, sealing concentrated water outlets, increasing the density of confined water head monitoring, and verifying stratigraphic data. This embodiment illustrates that the fusion discrimination of this invention is not a simple averaging of various features, but rather retains the priority of hard mechanistic criteria and strong on-site anomalies based on the fusion of multi-source heterogeneous features, providing sufficient safety redundancy for sudden surge risks with strong consequences.

[0113] It should be noted that the above embodiments are merely preferred embodiments of the present invention, used only to illustrate the principle of the method of the present invention, and are not intended to limit the scope of protection of the present invention. In practical applications, those skilled in the art can allocate the above functions to different steps as needed, that is, further decompose or combine the steps in the embodiments of the present invention. For example, the steps in the above embodiments can be combined into one step, or further divided into multiple sub-steps to complete all or part of the functions described above. The names of the steps involved in the embodiments of the present invention are merely for distinguishing the various steps and are not considered as limitations on the present invention.

[0114] 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 original technical features, and the technical solutions resulting from these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for identifying sudden surges in deep foundation pits near rivers by fusing multi-source heterogeneous features, characterized in that... Includes the following steps: Acquire multi-source heterogeneous data for deep foundation pit projects along rivers. The multi-source heterogeneous data includes at least hydrological data, stratigraphic data, construction data, and dynamic feedback data from on-site inspections. The confined water head is corrected based on the dynamic response relationship between the hydrological data and the confined water head, and a multi-source heterogeneous feature reflecting the risk of sudden inrush in the foundation pit is constructed based on the hydrological data, the geological data, the construction data and the dynamic feedback data from on-site inspections. The reliability of each data source in the multi-source heterogeneous data is determined, the weight of each data source in the fusion is adjusted according to the reliability, and the multi-source heterogeneous features are fused into a comprehensive risk index according to the adjusted weights. The risk level is determined based on the mechanism criteria and the comprehensive risk index. When the mechanism criteria reach the dangerous condition or a strong anomaly occurs on site, it is directly determined to be a dangerous level.

2. The method according to claim 1, characterized in that, The steps for correcting the confined head based on the dynamic response relationship between the hydrological data and the confined head specifically include: Cross-correlation analysis was performed on river water level monitoring data and confined head monitoring data to determine the effect coefficient and response lag time of river water level changes on confined head. The dynamic change of the confined head is determined by combining the current change in river water level, the action coefficient, the response lag time, and the intensity of precipitation pumping. The pressure head at the current judgment moment is corrected based on the dynamic change.

3. The method according to claim 2, characterized in that, The cross-correlation analysis includes: maximizing the correlation coefficient between river water level monitoring data and confined head monitoring data, using the time offset corresponding to the maximized correlation coefficient as the response lag time, and using the regression coefficient corresponding to the maximized correlation coefficient as the action coefficient.

4. The method according to claim 1, characterized in that, The steps of determining the reliability of each data source in the multi-source heterogeneous data and adjusting the weights of each data source in the fusion process based on the reliability specifically include: The reliability is determined based on data integrity rate, sampling time interval, sensor status, outlier ratio, manual review results, and spatial representativeness. When the reliability of a data source decreases, the fusion weight of that data source decreases.

5. The method according to claim 1, characterized in that, The specific steps of fusing the multi-source heterogeneous features into a comprehensive risk index based on the corrected weights include: The comprehensive risk index is obtained by weighting and summing the risk feature values ​​corresponding to each data source with the corrected weights.

6. The method according to claim 1, characterized in that, The steps for determining the risk level based on the mechanistic criteria and the comprehensive risk index specifically include: Determine the risk level corresponding to the aforementioned mechanism criterion; Determine the risk level corresponding to the comprehensive risk index; The highest risk level between the risk level corresponding to the mechanistic criterion and the risk level corresponding to the comprehensive risk index is taken as the final risk level.

7. The method according to claim 6, characterized in that, The mechanistic criteria include the anti-surge stability criterion, the permeability stability criterion, and the fracture penetration criterion.

8. The method according to claim 1, characterized in that, The strong anomalies at the site include at least one of water inrush, sand inrush, and continuous turbid seepage at the bottom of the pit; when the strong anomalies at the site occur, they are directly determined to be at a dangerous level without relying on the comprehensive risk index.

9. The method according to claim 1, characterized in that, Also includes: Based on the feedback information from the method's judgment result and the actual monitoring result, update the dynamic response relationship and / or the reliability determination method.

10. The method according to claim 1, characterized in that, The hydrological data includes river water level time history, rise and fall cycle, rise and fall amplitude, rise and fall rate, flood peak duration, rainfall, rainfall intensity, upstream sluice gate and dam scheduling information, lag time of river water level relative to confined head, initial confined head, real-time confined head, head change rate, correlation between confined head and river water level, and head changes caused by the operation of dewatering wells. The stratigraphic data includes the thickness of each stratigraphic layer, the elevation of the bottom layer, the natural unit weight, the buoyant unit weight, the cohesion, the internal friction angle, the compression modulus, the permeability coefficient, the tensile strength, the continuity of the aquitard, the distribution of weakly permeable interlayers, the elevation of the top plate of the confined aquifer, and the recharge boundary of the confined aquifer. The construction data includes the current excavation depth, layer excavation thickness, excavation time, excavation rate, pit bottom exposure time, support installation time, support axial force, dewatering well start / stop status, pumping volume, recharge volume, construction pause time, and completed excavation conditions; The on-site inspection dynamic feedback data includes the location of seepage at the bottom of the pit, changes in seepage volume, color of the effluent, turbidity of the effluent, whether it contains sand, local bulging, cracks at the bottom of the pit, leakage in the retaining structure, surface cracks, abnormal seepage on the riverbank, turbidity of the effluent from the dewatering well, increase in sand content in the drainage ditch, and monitoring alarm information.

Images