A method for determining the sand content of slurry for caisson shaft excavation in water-rich sand and gravel strata

Abstract

The present invention provides a method for determining the sand content of mud during caisson-type vertical shaft excavation in a water-rich sand and gravel formation, comprising the following steps: using a pipeline-type online laser particle size analyzer for online dynamic scanning to determine the volume concentration of mud sand particles during the caisson-type vertical shaft excavation in a water-rich sand and gravel formation; S 砂‑激光 ; Use dual-outlet sampler for offline verification to calculate the actual sand particle ratio S 砂‑真实 , correcting data deviations from a pipeline-based online laser particle size analyzer; installing an observation window at the bottom of the caisson for visual confirmation, automatically counting sand particles using a machine learning algorithm to trigger an auxiliary verification signal; and determining the sand content in the mud through steps S1, S2, and S3 to perform a three-layer coordinated assessment of formation instability. Through a three-layer monitoring system of "real-time monitoring → precise calibration → visual verification," this invention achieves multi-dimensional, high-precision quantitative analysis of the sand content ≥0.075 mm in the mud, providing a key basis for identifying water and sand inrush risks.

Description

Technical Field

[0001] The invention relates to the technical field of caisson construction safety, and in particular to a method for determining the sand content of slurry for caisson-type vertical shaft excavation in a water-rich sand and gravel stratum. Background Art

[0002] As urban underground space development continues to deepen, the scale and depth of vertical shafts, critical vertical passageways for projects such as underground transportation and integrated pipeline corridors, continue to rise. Caisson-type shaft excavation technology (such as the VSM method) has become a core method for ultra-deep shaft construction due to its advantages such as mechanized excavation, adaptive shaft lining construction, and controlled environmental impact. This method utilizes undrained excavation within the shaft, employing slurry wall protection and pressure balancing techniques to effectively control ground disturbance while reducing retaining structure costs. This provides an innovative solution for shaft construction in complex strata.

[0003] However, the application of this technology in water-rich sandy gravel formations faces significant challenges: the formation has high permeability (permeability coefficient of 10 -2 ~10 -1 cm / s), strong water-richness and low cementation strength. Its porosity can reach 15%-30%, which makes the dynamic balance between the mud pressure at the excavation face and the stratum seepage pressure very easy to break. In this process, the sand content of the mud (the proportion of particles with a particle size ≥ 0.075mm) is the core indicator reflecting the stability of the stratum and the risk of water and sand inrush. Its accurate measurement is directly related to construction safety. The existing monitoring technology system has three key defects: (1) Insufficient real-time monitoring accuracy. The traditional laser particle size analyzer is interfered by the adsorption effect of clay colloids, and the misjudgment rate of sand concentration is as high as 22%; (2) The lack of offline calibration method. The existing technology cannot establish an accurate conversion relationship between the total solid volume of the mud and the sand content; (3) The risk verification method is single and lacks a visual verification mechanism for the sand migration process, making it difficult to capture the early signals of pebble skeleton instability in a timely manner.

[0004] The above technical defects lead to double risks in construction: on the one hand, misjudgment of sand content may lead to invalid warning and cause construction delay; on the other hand, missed judgment risk will directly lead to sudden surge accidents. -2 In sandy and gravel formations with a velocity of 100 cm / s, existing monitoring systems often experience a response lag. Therefore, building a mud sand content determination system that integrates multi-source information perception, dynamic calibration, and visual verification has become an urgent need to overcome the safety bottleneck of caisson construction in water-rich sandy and gravel formations. Summary of the Invention

[0005] To solve the above problems, the present invention aims to propose a method for determining the sand content of mud for caisson-type vertical shaft excavation in water-rich sand and gravel formations. Through a three-layer monitoring system of "real-time monitoring → precise calibration → visual verification", a multi-dimensional, high-precision quantitative analysis of the content of sand particles ≥ 0.075 mm in the mud is achieved, providing a key basis for the risk judgment of water and sand gushing.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] A method for determining the sand content of slurry for caisson-type vertical shaft excavation in a water-rich sand and gravel stratum comprises the following steps:

[0008] S1. Real-time monitoring by laser particle size analyzer: Use pipeline-type online laser particle size analyzer for online dynamic scanning to determine the volume concentration S of mud sand particles during the caisson shaft excavation process in water-rich sand and gravel formations. 砂-激光 ;

[0009] S2. Offline precision calibration: Use a dual-outlet sampler for offline calibration to calculate the actual sand particle ratio and calculate S 砂-真实 , correct the data deviation of pipeline online laser particle size analyzer;

[0010] S3. Visual verification with a bottom observation window: Install an observation window at the bottom of the caisson for visual confirmation. Use a machine learning algorithm to automatically count the number of sand particles and trigger an auxiliary verification signal.

[0011] S4. Risk linkage identification: The three-layer linkage identification of formation instability is performed based on the results of mud sand content determination in steps S1, S2 and S3.

[0012] Furthermore, in step S1, the pipeline-type online laser particle size analyzer is connected in series with the straight section of the main return slurry pipeline of the caisson slurry at a distance of 2 to 5 meters from the bottom of the caisson. The pipeline-type online laser particle size analyzer uses a 0.075 mm equivalent aperture screening algorithm to scan the volume concentration of sand particles ≥ 0.075 mm in real time. S 砂-激光 , the monitoring frequency is 1 time / second;

[0013] The pipeline-type online laser particle size analyzer is installed between the two flanges of the main return slurry pipeline of the caisson. The probe is inserted through the flange opening, and a 0.5mm filter is installed at the front end to prevent clogging by large particles.

[0014] The volume concentration of sand particles collected by the pipeline online laser particle size analyzer S 砂-激光 Data generates sand concentration curve and growth rate in real time , when S 砂-激光 >5% and When the level 1 yellow warning is triggered, the offline precision calibration process in step S2 and the visual verification of the bottom hole observation window in step S3 are started simultaneously;

[0015] In step S1, the mud flow is monitored synchronously. When the sand content increases and the flow drops sharply and exceeds 20%, it is determined to be a precursor to pipeline blockage, thereby avoiding misjudgment as sand surge.

[0016] Furthermore, in step S2, an automatic sampling valve is installed on the surface section of the main mud return pipeline, and a dual-outlet sampler is used to collect dual samples, namely, a filtered sample and an original sample.

[0017] Furthermore, the dual-outlet sampler is provided with two outlets, namely outlet A and outlet B;

[0018] Outlet A is a filtration channel with a 0.075mm filter mesh inside. The filtered sample is collected through outlet A, that is, the sand sample is output. The collected filtered sample volume is 100ml. The powder / clay particles <0.075mm are filtered through a 0.075mm stainless steel filter mesh. The sand particles with a particle size of ≥0.075mm in the filtered sample are directly intercepted and then poured into a measuring cylinder and left to stand for 30 minutes to measure the sediment volume. V 1. Wash with clean water twice to remove clay colloid, and then observe the purity of sand particles under a microscope;

[0019] Outlet B is the original mud channel, with no filter inside. The original sample is collected through outlet B, that is, the original sample is output. The volume of the collected original sample is 50 ml. The collected original sample is poured into an evaporating dish and dried in an oven at 105°C to constant weight. The total solid mass is calculated. m 总 ;

[0020] The actual sand particle ratio is calculated according to the test data of the filtered sample and the original sample according to the following formula S 砂-真实 :

[0021] Formula 1

[0022] Where, The natural density in the sand and gravel formation survey report;

[0023] Calibrate twice a day through step S2 to correct the data deviation of the pipeline online laser particle size analyzer in step S1.

[0024] Furthermore, the actual sand particle ratio obtained in step S2 S 砂-真实 The following method was used to calibrate the laser data and correct for clay colloid interference:

[0025] Will S 砂-真实Real-time concentration measured by the simultaneous in-line laser particle size analyzer S 砂-激光 Compare and calculate the coefficient of deviation K = S 砂-真实 / S 砂-激光 ;

[0026] When | K When -1|>5%, add the clay colloid adsorption correction factor in the data analysis software of the pipeline online laser particle size analyzer to reduce the artificially high sand concentration caused by colloid attachment, or use pure sand from the double-outlet sampler as a standard substance, and regularly calibrate the light scattering parameters of the laser equipment to ensure the particle size screening accuracy.

[0027] Furthermore, in step S3, a pressure-bearing transparent observation window with a diameter of 300 mm is set 1.5 m above the cutting foot at the bottom of the caisson, a detachable protective grille is installed on the outside, and a 1080P underwater camera is equipped with near-infrared fill light;

[0028] The pressure-bearing transparent observation window is installed at the upstream position of the groundwater in the caisson shaft;

[0029] Visual observation and confirmation of the mud and sand content at the bottom of the caisson is carried out using underwater cameras, with manual observation and AI image recognition performed separately;

[0030] During manual observation, ensure that visual inspections are carried out twice per shift. If sand particles with a particle size of 0.25 mm or more are found to be flowing in a directional manner with a flow rate greater than 0.5 m / s or if sand is carried by clean water, it is considered a sign of formation instability.

[0031] When performing AI image recognition, the machine learning algorithm automatically counts the number of sand particles ≥0.075mm in the image. When the detection volume is greater than 10 per second, an auxiliary verification signal is triggered.

[0032] Furthermore, in AI image recognition, median filtering or Gaussian filtering is used to remove image noise, combined with Otsu threshold segmentation to separate mud background and sand particles; YOLOv5n is used to recognize sand particle images in mud. The collected mud image is first subjected to denoising and segmentation preprocessing operations, and then the preprocessed image is input into the YOLOv5n model that has been trained on a data set of sand particle images marked with particle size ≥ 0.075mm under different mud concentrations and lighting conditions. Through feature extraction, region proposal generation and bounding box regression steps, the model quickly and accurately identifies the position and size of sand particles in the image, and outputs the detection confidence of each sand particle, thereby counting the number of sand particles and providing a basis for subsequent risk judgment.

[0033] Furthermore, a DN50 sampling tube is arranged 30 cm below the pressure-bearing transparent observation window.

[0034] Furthermore, in step S4, risk linkage identification specifically refers to:

[0035] when S 砂-真实 When the flow rate is >5% and the observation window confirms abnormal sand flow, it is comprehensively judged as a risk of water and sand gushing, and emergency grouting is automatically started, while the mud density is adjusted to 1.2~1.25g / cm³;

[0036] Determine the risk level, S 砂-真实 >8% is high risk, 5%< S 砂-真实 <8% is medium risk;

[0037] Sensor data is integrated through the Internet of Things platform, and BIM technology is used to build a three-dimensional model of the construction area, mapping monitoring parameters in real time. Machine learning algorithms are used to identify abnormal data patterns, automatically trigger graded warnings, and simultaneously push them to the construction management system.

[0038] Furthermore, the S 砂-真实 The sand content threshold is determined in actual engineering through trial excavation tests, which specifically include the following steps: 1) Determine the sand content of normal excavation mud, and collect data on the normal cutting stage of the caisson-type shaft boring machine in the water-rich sand and gravel formation through a pipeline online laser particle size analyzer and a dual-outlet sampler. After several rounds of excavation cycles, calculate the average value of the sand concentration during the normal cutting process. and , determine the mud sand content in the normal cutting process by weighted average method ,α is the weight of the two test methods, which can be allocated according to the accuracy and reliability of the two measurement methods;

[0039] 2) Determine the threshold value of sand content in unstable sand influx slurry, according to S 砂-正常 Determine the medium risk mud sand content threshold S 砂-中 = (2~5) S 砂-正常 and high-risk mud sand content threshold S 砂-高 = (5~10) S 砂-正常 .

[0040] Beneficial effects: 1. Three-dimensional monitoring of mud sand content is achieved, with real-time warning (response time <1 second) through the pipeline-type online laser particle size analyzer, dual-export precise calibration (error <3%), and observation window visual verification (identification of particle size ≥0.075mm). The missed judgment rate can be significantly reduced compared with the traditional single method; 2. Compliance with the monitoring and evaluation method standards: The sampling process complies with GB / T 50123-2019 "Standard for Geotechnical Test Methods", and laser monitoring complies with SL42-2010 "River Sediment Particle Analysis Procedure", forming a standardized monitoring process. 3. Intelligent linkage control of risk assessment: Three layers of data are synchronized in real time, and the risk level is automatically determined. The actual sand particle ratio is calculated S 砂-真实 Determine the method for distinguishing medium and high risks, significantly reduce emergency response time, and improve construction safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings, which constitute part of the present invention, are provided to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are provided to explain the present invention and do not constitute an undue limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 This is a main flow chart of a method for determining the sand content of slurry for caisson-type vertical shaft excavation in a water-rich sand and gravel formation according to an embodiment of the present invention;

[0043] Figure 2 This is a working principle diagram of mud processing and separation (sand content) during the excavation process of a caisson-type shaft boring machine in the method for determining the sand content of slurry for caisson-type shaft boring in water-rich sand and gravel formations according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0045] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0046] The excavation face of a caisson-type shaft boring machine (TBM) is the interface between the strata directly cut by the cutterhead (milling head, or cutting drum). This area is directly subject to the dynamic balance between mud pressure and ground water and soil pressure. If the mud pressure is insufficient (lower than the ground water and soil pressure), groundwater will seep through the gaps between the pebbles in the excavation face, carrying fine sand with it and forming sand influx (the critical particle size is generally ≥0.075mm). Furthermore, when the cutterhead cuts large pebbles or causes vibration to loosen the ground, the excavation face can become locally unstable, creating seepage channels, allowing sand to flow into the shaft with the water. Furthermore, during the caisson's sinking process, irregular gaps between the shaft wall and the surrounding strata are created by friction, overexcavation, or pebble collisions (especially when pebble edges scrape against the shaft wall, causing localized gaps to widen). If the drag-reducing mud fails to fill these gaps in time (e.g., due to insufficient mud viscosity or excessive sinking speed), groundwater may seep into the shaft through these gaps, carrying sand particles (especially fine sand with a particle size of less than 0.075mm) within them. Therefore, the mud sand content is the core indicator for judging the stability of the excavation face.

[0047] Example 1

[0048] Based on the above research background, see Figure 1-2 This embodiment provides a method for determining the sand content of slurry for caisson-type vertical shaft excavation in a water-rich sand and gravel formation, comprising the following steps:

[0049] S1. Real-time monitoring by laser particle size analyzer: Use pipeline-type online laser particle size analyzer for online dynamic scanning to determine the volume concentration of mud sand particles during the caisson shaft excavation process in water-rich sand and gravel formations. S 砂-激光 ;

[0050] S2. Offline precision calibration: Use dual-outlet sampler for offline calibration to calculate the actual sand particle ratio S 砂-真实 , correct the data deviation of pipeline online laser particle size analyzer;

[0051] S3. Visual verification with a bottom observation window: Install an observation window at the bottom of the caisson for visual confirmation. Use a machine learning algorithm to automatically count the number of sand particles and trigger an auxiliary verification signal.

[0052] S4. Risk linkage identification: The three-layer linkage identification of formation instability is performed based on the results of mud sand content determination in steps S1, S2 and S3.

[0053] This embodiment has the following advantages:

[0054] Multi-dimensional monitoring system: Combining real-time scanning, offline calibration, and visual verification to form a three-layer data verification system, it avoids the limitations of a single monitoring method and significantly improves the accuracy of mud sand content detection.

[0055] ‌Risk linkage mechanism: Identify the risk of stratum instability through multi-source data fusion and enhance the dynamic response capability of construction in complex strata.

[0056] This embodiment solves the "real-time" problem through laser, solves the "accuracy" problem through offline precise calibration, and solves the "intuitive" problem through the observation window, filling the gap in the multi-dimensional, high-precision quantitative monitoring system of mud sand content in the mechanical caisson construction in water-rich sand and gravel formations. It is of great significance to promote the advancement of underground engineering construction safety monitoring technology, enhance the risk prevention and control capabilities of mechanical caisson construction in complex formations, and promote the efficient and safe development of related engineering construction.

[0057] It should be noted that the water-rich sand and gravel formation of this embodiment is a formation with abundant groundwater, high sand and gravel porosity, low sand and gravel cementation strength, and large permeability coefficient. The caisson-type vertical shaft excavation of the water-rich sand and gravel formation adopts a caisson-type vertical shaft boring machine. The water-rich sand and gravel formation is synchronously suspended and sunk by the caisson-type vertical shaft boring machine during the excavation process.

[0058] The lower end of the caisson shaft is a steel blade angle ring, and the vertical distance between the milling working surface of the caisson shaft boring machine and the lower edge of the steel blade angle ring is H =1~1.5m;

[0059] The milling radius of the caisson shaft boring machine R Greater than the outer radius of the caisson shaft r The horizontal over-excavation of the caisson shaft boring machine is R - r The water-rich sand and gravel stratum and the outer wall of the caisson shaft are filled with drag-reducing mud;

[0060] The caisson shaft is filled with mud, and the mud is circulated continuously and stably through a mud circulation system. The mud circulation system includes a mud tank, a main slurry discharge pump, a mud pipeline system, and a mud-water separation station. The mud pipeline system includes a mud supply pipeline, a main mud return pipeline, a mud circulation branch pipeline, and a flushing slurry feed pipeline. The main mud return pipeline is the main return pipeline from the outlet of the main slurry discharge pump to the mud-water separation station.

[0061] In a specific example, in step S1, the pipeline-type online laser particle size analyzer is connected in series to the straight section of the main return slurry pipeline of the caisson slurry at a distance of 2 to 5 meters from the bottom of the caisson. The pipeline-type online laser particle size analyzer uses a 0.075 mm equivalent aperture screening algorithm to scan the volume concentration of sand particles ≥ 0.075 mm in real time. S 砂-激光 , the monitoring frequency is 1 time / second;

[0062] The pipeline-type online laser particle size analyzer is installed between the two flanges of the main return slurry pipeline of the caisson. The probe is inserted through the flange opening, and a 0.5mm filter is installed at the front end to prevent clogging by large particles.

[0063] The volume concentration of sand particles collected by the pipeline online laser particle size analyzer S 砂-激光 Data generates sand concentration curve and growth rate in real time ,when S 砂-激光 >5% and When the level 1 yellow warning is triggered, the offline precision calibration process in step S2 and the visual verification of the bottom hole observation window in step S3 are started simultaneously;

[0064] In step S1, the mud flow is monitored synchronously. When the sand content increases and the flow drops sharply and exceeds 20%, it is determined to be a precursor to pipeline blockage, thereby avoiding misjudgment as sand surge.

[0065] Dynamic warning logic: Based on the volume concentration of sand particles ( S 砂-激光 Dual indicators of flow rate drop (>5%) and flow rate drop (>20%) distinguish pipeline blockage from sand inrush risks and reduce false alarms.

[0066] It should be noted that testing and sampling in the pipeline avoids the problem of "the mud in the caisson is turbid and difficult to sample at a single point", ensuring that the sample directly reflects the characteristics of the sand particles in the entire caisson;

[0067] The pipeline-type online laser particle size analyzer is connected in series with the straight section of the main return slurry pipeline of the caisson to avoid flow interference from elbows or diameter-changing sections. If conditions permit, multi-point sampling can be set to form a sand content gradient monitoring to eliminate local disturbance misjudgment.

[0068] The pipeline-type online laser particle size analyzer automatically filters out silt / clay particles with a size less than 0.075mm using a 0.075mm equivalent aperture screening algorithm. The scientific basis for the critical particle size of 0.075mm is that this particle size is the "quicksand starting particle size" in soil mechanics. Particles smaller than this value are constrained by mud viscosity and are difficult to migrate. Particles larger than this value are prone to forming piping channels under the action of seepage forces (verified by Terzaghi's seepage theory). The automatic use of the pipeline-type online laser particle size analyzer avoids the randomness of manual sampling and ensures that the samples are synchronized with the excavation conditions.

[0069] In this embodiment S 砂-激光 >5% corresponds to the critical concentration of fine sand in soil mechanics for initiating migration (Terzaghi seepage theory: when the volume proportion of sand particles is greater than 5%, the seepage force can overcome the friction between particles and form a piping channel).

[0070] In a specific example, in step S2, an automatic sampling valve is installed on the surface section of the main mud return pipeline, and a dual-outlet sampler is used to collect dual samples, namely, a filtered sample and an original sample. The dual-outlet sampler is provided with two outlets, namely, outlet A and outlet B.

[0071] Outlet A is a filtration channel with a 0.075mm filter mesh inside. The filtered sample is collected through outlet A, that is, the sand sample is output. The collected filtered sample volume is 100ml. The powder / clay particles <0.075mm are filtered through a 0.075mm stainless steel filter mesh. The sand particles with a particle size of ≥0.075mm in the filtered sample are directly intercepted and then poured into a measuring cylinder and left to stand for 30 minutes to measure the sediment volume. V 1. Rinse twice with clean water to remove clay colloids, and then observe the purity of the sand particles under a microscope (if 90% of the sand is found to be fine sand with a particle size of >0.075-0.3mm after washing);

[0072] Outlet B is the original mud channel, without a filter inside. The original sample is collected through outlet B, that is, the original sample is output. The volume of the collected original sample is 50 ml. The collected original sample is poured into an evaporating dish and dried in an oven at 105°C to constant weight (4 hours). The total solid mass is calculated. m 总 ;

[0073] The actual sand particle ratio is calculated according to the test data of the filtered sample and the original sample according to the following formula S 砂-真实 :

[0074] Formula 1

[0075] Where, The natural density in the sand and gravel formation survey report;

[0076] Calibrate twice a day through step S2 to correct the data deviation of the pipeline online laser particle size analyzer in step S1.

[0077] Dual-sample comparison calibration: Through a combined physical-chemical analysis of filtered samples (separating sand particles ≥ 0.075 mm) and original samples (oven-dried total solids), the true sand fraction is accurately calculated (Formula 1), eliminating falsely inflated values ​​caused by clay colloid adsorption.

[0078] Standardized operating procedures: Twice-daily calibration frequency ensures the long-term reliability of laser particle size analyzer data and adapts to the dynamic changes in mud composition in water-rich formations.

[0079] The dual-outlet sampler of this embodiment has two outlets, and can quickly estimate the content of sand particles with a particle size >0.075mm through outlet A, which is used for early risk control, avoiding the risk information delay caused by the 4-hour drying requirement of outlet B, and solving the problem of long-term full-volume mud analysis (traditional drying method requires 4 hours), and is used for microscopic observation and rapid risk judgment.

[0080] Rapidly separate particles of critical size, focusing on sand particles (≥0.075mm) that have the greatest impact on seepage stability, and combine with the total solid volume to form a quantitative indicator of "risk particle concentration".

[0081] In a specific example, the actual sand grain ratio obtained in step S2 is S 砂-真实 The following method was used to calibrate the laser data and correct for clay colloid interference:

[0082] Will S 砂-真实 Real-time concentration measured by the simultaneous in-line laser particle size analyzer S 砂-激光 Compare and calculate the coefficient of deviation K = S 砂-真实 / S 砂-激光 ;

[0083] when |K When -1|>5%, add the clay colloid adsorption correction factor in the data analysis software of the pipeline online laser particle size analyzer to reduce the artificially high sand concentration caused by colloid attachment, or use pure sand particles (confirmed by microscope that there is no clay attachment) from the double-outlet sampler as standard materials, and regularly calibrate the light scattering parameters of the laser equipment to ensure the particle size screening accuracy.

[0084] Dynamic correction algorithm‌: Through the deviation coefficient ( K ) Real-time adjustment of the light scattering parameters of the laser particle size analyzer or addition of clay colloid correction factors can improve the particle size screening accuracy.

[0085] Equipment self-calibration capability: Using samples from the dual-outlet sampler as standard materials, the equipment is regularly calibrated to reduce drift errors caused by aging of optical components or environmental fluctuations.

[0086] In a specific example, in step S3, a 300mm diameter pressure-bearing transparent observation window (e.g., tempered glass, with a pressure resistance of 1MPa) is installed 1.5m above the cutting edge of the caisson bottom. A removable protective grille (to prevent pebble impact) is installed on the outside. A 1080P underwater camera (measuring that 0.075mm particles are ≥5px in the image, reducing the impact of mud transmittance, ensuring clear particles in the field of view, and waterproof level IP68) is equipped with near-infrared fill light.

[0087] The pressure-bearing transparent observation window is installed at the upstream position of the groundwater in the caisson shaft;

[0088] Visual observation and confirmation of the mud and sand content at the bottom of the caisson is carried out using underwater cameras, with manual observation and AI image recognition performed separately;

[0089] During manual observation, ensure that visual inspections are carried out twice per shift. If sand particles with a particle size of 0.25 mm or more are found to be flowing in a directional manner with a flow rate greater than 0.5 m / s or if sand is carried by clean water, it is considered a sign of formation instability.

[0090] When performing AI image recognition, the machine learning algorithm automatically counts the number of sand particles ≥0.075mm in the image. When the detection volume is greater than 10 per second, an auxiliary verification signal is triggered.

[0091] In AI image recognition, median filtering or Gaussian filtering is used to remove image noise, combined with Otsu threshold segmentation to separate mud background and sand particles. YOLOv5n is used to identify sand particles in mud. The collected mud images are first preprocessed by denoising and segmenting. The preprocessed images are then input into a YOLOv5n model trained on a dataset of sand images labeled with particle sizes ≥ 0.075mm under different mud concentrations and lighting conditions. Through feature extraction, region proposal generation, and bounding box regression, the model quickly and accurately identifies the location and size of sand particles in the image. It also outputs the detection confidence level for each sand particle, which is used to count the sand particles and provide a basis for subsequent risk assessment.

[0092] Visual direct observation: The bottomhole observation window (300mm pressure-bearing transparent design) combines near-infrared fill light and protective grille to achieve intuitive monitoring of sand flow in complex underwater environments, compensating for the limitations of instrument data.

[0093] AI-assisted decision-making: High-speed sand particle recognition (≥0.075mm, >10 trigger signals / second) based on the YOLOv5n model, combined with manual visual inspection (0.25mm sand particle velocity threshold), forms a "human-machine dual-core" verification mechanism to improve the timeliness of risk assessment.

[0094] A DN50 sampling tube is arranged 30 cm below the pressure-bearing transparent observation window.

[0095] When the water flow in the observation window is found to be turbid, in this embodiment, the original mud at the bottom of the well can be manually extracted through the DN50 sampling tube to avoid interference from the circulating mud.

[0096] In a specific example, in step S4, risk linkage identification specifically refers to:

[0097] when S砂-真实 When the flow rate is >5% and the observation window confirms abnormal sand flow, it is comprehensively judged as a risk of water and sand gushing, and emergency grouting is automatically started, while the mud density is adjusted to 1.2~1.25g / cm³;

[0098] Determine the risk level, S 砂-真实 >8% is high risk, 5%< S 砂-真实 <8% is medium risk;

[0099] Sensor data is integrated through the IoT platform, and BIM technology is used to build a three-dimensional model of the construction area, mapping monitoring parameters in real time. Machine learning algorithms are used to identify abnormal data patterns, automatically triggering graded warnings (yellow / orange / red), and simultaneously pushing them to the construction management system.

[0100] The relevant S 砂-真实 The threshold value can be dynamically adjusted according to the specific geological characteristics of the site.

[0101] In a specific example, the S 砂-真实 The sand content threshold is determined in actual engineering through trial excavation tests, which specifically include the following steps:

[0102] 1) Determine the sand content of the normal excavation slurry. Use a pipeline-type online laser particle size analyzer and a dual-outlet sampler to collect data from the caisson-type shaft boring machine during the normal cutting (milling) stage in the water-rich sand and gravel formation. After several rounds of excavation cycles, calculate the average value of the sand concentration during the normal cutting process. and , determine the mud sand content during normal cutting (milling) by weighted average method ,α is the weight of the two test methods, which can be allocated according to the accuracy and reliability of the two measurement methods;

[0103] 2) Determine the threshold value of sand content in unstable sand influx slurry, according to S 砂-正常 Determine the medium risk mud sand content threshold S 砂-中 =(2~5) S 砂-正常 and high-risk mud sand content threshold S 砂-高 =(5~10) S 砂-正常 .

[0104] Dynamic calibration of threshold value: Determine the sand content threshold value through trial excavation test ( S砂-正常 、 S 砂-中 、 S 砂-高 ), avoid the disconnection between the fixed threshold and engineering practice, and enhance the formation adaptability of the method.

[0105] Smart Emergency Response: Based on the integration of BIM 3D models and IoT data, this system automatically links risk grading (high / medium risk) with grouting parameter adjustment (mud density 1.2-1.25g / cm³), shortening emergency response time.

[0106] In a specific application, the shaft of a certain caisson shaft excavation project is designed to be 50m deep, with a shaft diameter of 12m. The milling working surface is 1.2m away from the lower edge of the steel blade angle ring, with a horizontal over-excavation of 20cm. The gap between the shaft wall and the formation is filled with bentonite slurry with a density of 1.1g / cm³. The shaft is buried at a depth of 29m. The porosity of deep sand and gravel is 25%, and the natural density is 20cm. ρ 土 =2.68g / cm³, permeability coefficient k =5×10 - ²cm / s, indicating a high-risk, water-rich formation. A pipeline-type online laser particle size analyzer (installed 3m from the bottom of the well on the main return slurry pipeline), a dual-outlet sampler (calibrated once daily at 8:00 and 16:00), and a bottomhole observation window were installed on the mud circulation pipeline.

[0107] When the excavation reaches a certain depth, the sand concentration is obtained by counting the 2h data through the laser particle size analyzer. S 砂-激光 The average value is 4.2%, with an increase of 0.1% / min. The mud flow rate is stable at 80m³ / h, without sudden drops (excluding pipeline blockage). The analysis data is stable, and it is determined to be a normal cutting state. When the laser particle size analyzer is used for statistics, a dual-outlet sampler is used for offline synchronous calibration. Outlet A collects filtered samples. 100mL of mud is filtered through a 0.075mm filter. After standing for 30 minutes, the volume of sand particles settled V 1=4.5mL; Microscope observation: Sand purity 92% (fine sand with a particle size of 0.075-0.3mm accounts for 85%), which is consistent with normal cutting characteristics. The original sample was collected at outlet B, and the total solid mass of 50mL mud after drying was m 总 =134g, the actual sand particle ratio is obtained by calculation S 砂-真实 =4.5%. The calculated deviation coefficient K=1.07, the clay colloid correction factor (correction coefficient 1.07) was added to the laser particle size analyzer software. The mud at the bottom of the well was uniformly turbid and no directional flow of sand particles was visible. Based on this, S 砂-真实=4.5%<5%, no abnormality in the observation window, it is determined to be a safe state, and the current tunneling parameters (speed 15mm / min, mud density 1.15g / cm³) are maintained.

[0108] In summary, this embodiment is suitable for risk control of caisson-type vertical shaft excavation in high-permeability, highly disturbed, water-rich strata and has the following comprehensive advantages:

[0109] High precision and anti-interference: Multi-source data fusion (laser particle size analyzer + dual sample calibration + AI image recognition) reduces single errors and improves the reliability of sand particle detection.

[0110] Real-time and automation: Second-level monitoring frequency and machine learning algorithms enable instant risk warnings, reducing manual reliance.

[0111] Engineering Applicability: The threshold calibration and pressure observation window design based on trial excavation are suitable for the complex working conditions of water-rich sandy and gravel formations.

[0112] Closed-loop risk prevention and control: A complete closed-loop system from data collection, calibration, verification to emergency linkage effectively prevents water and sand gushing accidents and ensures the safety of caisson construction.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel formations, characterized in that: The following steps are involved: S1. Real-time monitoring by laser particle size analyzer: Use pipeline-type online laser particle size analyzer for online dynamic scanning to determine the volume concentration of mud sand particles during the caisson shaft excavation process in water-rich sand and gravel formations. S 砂-激光 In step S1, the pipeline-type online laser particle size analyzer is connected in series with the straight section of the main return slurry pipeline of the caisson slurry at a distance of 2 to 5 m from the bottom of the caisson. The pipeline-type online laser particle size analyzer uses a 0.075 mm equivalent aperture screening algorithm to perform real-time Scan the volume concentration of sand particles ≥ 0.075 mm S 砂-激光 , the monitoring frequency is 1 time / second; The pipeline-type online laser particle size analyzer is installed between the two flanges of the main return slurry pipeline of the caisson. The probe is inserted through the flange opening, and a 0.5mm filter is installed at the front end to prevent clogging by large particles. The volume concentration of sand particles collected by the pipeline online laser particle size analyzer S 砂-激光 Data generates sand concentration curve and growth rate in real time , when S 砂-激光 >5% and When the level 1 yellow warning is triggered, the offline precision calibration process in step S2 and the visual verification of the bottom hole observation window in step S3 are started simultaneously; In step S1, the mud flow is monitored simultaneously. When the sand content increases and the flow drops sharply by more than 20%, it is determined to be a precursor to pipeline blockage, thereby avoiding misjudgment as sand inrush. S2. Offline precision calibration: Use dual-outlet sampler for offline calibration to calculate the actual sand particle ratio S 砂-真实 , correct the data deviation of pipeline online laser particle size analyzer; 砂-真实 The real-time concentration S measured by the pipeline online laser particle size analyzer at the same time 砂-激光 Compare and calculate the deviation coefficient K=S 砂-真实 / S 砂-激光 ; When |K-1|>5%, add the clay colloid adsorption correction factor in the data analysis software of the pipeline online laser particle size analyzer to reduce the falsely high sand concentration caused by colloid attachment; S3. Visual verification with a bottom observation window: Install an observation window at the bottom of the caisson for visual confirmation. Use a machine learning algorithm to automatically count the number of sand particles and trigger an auxiliary verification signal. S4. Risk linkage identification: S 砂-真实 When the flow rate is >5% and the observation window confirms abnormal sand flow, it is comprehensively judged as a risk of water and sand gushing, and emergency grouting is automatically started. At the same time, the mud density is adjusted to 1.2~1.25g / cm 3 ; Determine the risk level, S 砂-真实 >8% is high risk, 5%< S 砂-真实 <8% is medium risk; Sensor data is integrated through the Internet of Things platform, and BIM technology is used to build a three-dimensional model of the construction area, mapping monitoring parameters in real time. Machine learning algorithms are used to identify abnormal data patterns, automatically trigger graded warnings, and simultaneously push them to the construction management system.

2. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 1 is characterized in that: In step S2, an automatic sampling valve is installed on the surface section of the main mud return pipeline, and a dual-outlet sampler is used to collect dual samples, namely, a filtered sample and an original sample.

3. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 2 is characterized in that: The dual-outlet sampler is provided with two outlets, namely outlet A and outlet B; Outlet A is a filtration channel with a 0.075mm filter mesh inside. The filtered sample is collected through outlet A, that is, the sand sample is output. The collected filtered sample volume is 100ml. The powder / clay particles <0.075mm are filtered through a 0.075mm stainless steel filter mesh. The sand particles with a particle size of ≥0.075mm in the filtered sample are directly intercepted and then poured into a measuring cylinder and left to stand for 30 minutes to measure the sediment volume. V 1. Wash with clean water twice to remove clay colloid, and then observe the purity of sand particles under a microscope; Outlet B is the original mud channel, with no filter inside. The original sample is collected through outlet B, that is, the original sample is output. The volume of the collected original sample is 50 ml. The collected original sample is poured into an evaporating dish and dried in an oven at 105°C to constant weight. The total solid mass is calculated. m 总 ; The actual sand particle ratio is calculated according to the test data of the filtered sample and the original sample according to the following formula S 砂-真实 : Formula 1 Where, The natural density in the sand and gravel formation survey report; Calibrate twice a day through step S2 to correct the data deviation of the pipeline online laser particle size analyzer in step S1.

4. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 3 is characterized in that: The actual sand grain ratio obtained by step S2 S 砂-真实 The following method was used to calibrate the laser data and correct for clay colloid interference: Will S 砂-真实 Real-time concentration measured by the simultaneous in-line laser particle size analyzer S 砂-激光 Compare and calculate the coefficient of deviation K = S 砂-真实 / S 砂-激光 ; Pure sand particles from the double-outlet sampler are used as standard materials to regularly calibrate the light scattering parameters of the laser equipment to ensure the accuracy of particle size screening.

5. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 1 is characterized in that: In step S3, a pressure-bearing transparent observation window with a diameter of 300 mm is set 1.5 m above the cutting foot at the bottom of the caisson, a detachable protective grille is installed on the outside, and a 1080P underwater camera is equipped with near-infrared fill light; The pressure-bearing transparent observation window is installed at the upstream position of the groundwater in the caisson shaft; Visual observation and confirmation of the mud and sand content at the bottom of the caisson is carried out using underwater cameras, with manual observation and AI image recognition performed separately; During manual observation, ensure that visual inspections are carried out twice per shift. If sand particles with a particle size of 0.25 mm or more are found to be flowing in a directional manner with a flow rate greater than 0.5 m / s or if sand is carried by clean water, it is considered a sign of formation instability. When performing AI image recognition, the machine learning algorithm automatically counts the number of sand particles ≥0.075mm in the image. When the detection volume is greater than 10 per second, an auxiliary verification signal is triggered.

6. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel formations according to claim 5, characterized in that: In AI image recognition, median filtering or Gaussian filtering is used to remove image noise, combined with Otsu threshold segmentation to separate mud background and sand particles; YOLOv5n is used for image recognition of sand particles in mud. The collected mud images are first preprocessed by denoising and segmentation. Then, the preprocessed images are input into the YOLOv5n model that has been trained on a dataset of sand images with particle sizes ≥ 0.075 mm under different mud concentrations and lighting conditions. Through feature extraction, region proposal generation, and bounding box regression steps, the model quickly and accurately identifies the position and size of sand particles in the image, and outputs the detection confidence of each sand particle. This is used to count the number of sand particles and provide a basis for subsequent risk assessment.

7. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 5, characterized in that: A DN50 sampling tube is arranged 30 cm below the pressure-bearing transparent observation window.

8. The method for determining the sand content of slurry for caisson-type vertical shaft excavation in water-rich sand and gravel strata according to claim 1, characterized in that: described S 砂-真实 The sand content threshold is determined through trial excavation tests in actual projects, which specifically include the following steps: 1) Determine the sand content of the mud during normal excavation. Use a pipeline-type online laser particle size analyzer and a dual-outlet sampler to collect data from the caisson-type shaft boring machine during the normal cutting stage in the water-rich sand and gravel formation. After several rounds of excavation cycles, calculate the average sand concentration during the normal cutting process. and , determine the mud sand content in the normal cutting process by weighted average method , α The weights of the two test methods can be assigned based on the accuracy and reliability of the two measurement methods; 2) Determine the threshold value of sand content in unstable sand influx slurry, according to S 砂-正常 Determine the medium risk mud sand content threshold S 砂-中 = (2~5) S 砂-正常 and high-risk mud sand content threshold S 砂-高 =(5~10) S 砂-正常 .

Images