Permeability coefficient dynamic calculation method, system and equipment based on multi-source sensing and medium

Abstract

The invention discloses a permeability coefficient dynamic calculation method, system and equipment based on multi-source sensing and a medium, and belongs to the technical field of seepage monitoring, the permeability coefficient dynamic calculation method based on multi-source sensing comprises the following steps: step 1, according to a layered layout and risk area encryption principle, constructing a construction monitoring area; a distributed optical fiber, a proton nuclear magnetic resonance (NMR) sensor, a micro-seismic sensor, a pore water pressure auxiliary sensor and a solute concentration auxiliary sensor are arranged, and strain, pore water distribution, a micro-seismic event, pore water pressure and solute concentration data are synchronously acquired through an edge computing node and 5G transmission. And a microseismic sensor is matched to capture microcrack initiation and expansion in real time, and macroscopic and microscopic association mapping is constructed.

Description

Technical Field

[0001] This invention relates to the field of seepage monitoring technology, specifically to a method, system, device, and medium for dynamic calculation of permeability coefficient based on multi-source sensing. Background Technology

[0002] The permeability coefficient is a core parameter characterizing the permeability of a medium, and its accurate acquisition directly determines the reliability of engineering seepage safety assessments. Over decades of development, permeability coefficient calculation technology has evolved from "static single-point testing" to "dynamic multi-source inversion," and existing calculation techniques are mainly divided into two categories: static testing techniques and dynamic inversion techniques. With the development of IoT and AI technologies, dynamic calculation methods based on multi-source sensing are gradually becoming mainstream, and current multi-source sensing technology has achieved a breakthrough in dynamic processing.

[0003] Current technologies generally employ a binary combination of macroscopic parameters, which can only characterize the macroscopic influencing factors of permeability coefficient, failing to consider core characteristics that directly determine permeability coefficient, such as microscopic pore water distribution and microstructural damage. For example, when inverting the permeability coefficient of a dam solely using fiber optic strain data, the calculation error jumps from 15% to 22% when the water content of the dam concrete increases from 5% to 8%. In existing shield tunnel construction, ERT and pressure sensing methods are used to monitor the permeability coefficient of sand layers. However, because they fail to capture the development of micro-fractures, abrupt changes in permeability coefficient before water inrush are not identified, leading to construction safety accidents. In soil-rock mixed slope engineering, existing sensing combinations cannot adapt to the heterogeneity of layered media. In weak interlayer areas, the calculation error reaches 25%-35% because it fails to capture changes in pore connectivity and the initiation of micro-fractures, easily triggering shallow landslides. The fundamental reason is that the permeability coefficient is directly related to porosity, pore connectivity, and micro-fracture development. Existing sensing combinations lack microscopic data support, resulting in a break in the physical correlation between data and parameters.

[0004] Therefore, we propose a method, system, device, and medium for dynamic calculation of permeability coefficient based on multi-source sensing, in order to alleviate or solve the above problems.

[0005] The information disclosed above in this background section is only for enhancing the understanding of the background section of this invention, and therefore may include prior art that is not known to those skilled in the art. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method, system, device, and medium for dynamic calculation of permeability coefficient based on multi-source sensing. This solves the problems in existing technologies, such as relying solely on a binary combination of macroscopic parameters, lacking microscopic data support leading to a broken physical correlation between sensor data and permeability coefficient, large calculation errors, ineffective identification of micro-fracture development and abrupt changes in permeability coefficient signals, inability to adapt to the heterogeneity of layered slope media, and lack of correlation with rainfall replenishment coupling effects.

[0007] To achieve the above objectives, this invention provides a method for dynamically calculating the permeability coefficient based on multi-source sensing, comprising the following steps:

[0008] Step 1: In the construction monitoring area, according to the principle of layered deployment and densification in risk areas, deploy distributed optical fibers, proton nuclear magnetic resonance (NMR), microseismic sensors, and auxiliary sensors for pore water pressure and solute concentration. Collect strain, pore water distribution, microseismic events, pore water pressure, and solute concentration data synchronously through edge computing nodes and 5G transmission, with a total transmission delay of ≤10s.

[0009] Step 2: Denoise and align the collected data in layers, including fiber optic data through 8-layer decomposition of db4 wavelet packets and empirical mode decomposition for denoising, NMR data through T2 spectrum double exponential fitting to analyze effective porosity, and microseismic data through energy threshold screening for effective events. Then, the multi-source data are aligned using a dynamic time warping algorithm.

[0010] Step 3: Establish the four-field coupled control equations of seepage, heat, force and solute, introduce the solute adsorption and desorption kinetic model, and perform multi-field coupling through the permeability coefficient correlation function K=K(n,ε,C), where n is the effective porosity, ε is the volumetric strain and C is the solute concentration;

[0011] Step 4: Adaptive Kalman Filter (AKF), Attention-LSTM, and Bayesian Network three-level algorithm are used. AKF initially calculates the penetration coefficient K1, Attention-LSTM optimizes to obtain K2, and Bayesian Network outputs K and 95% confidence interval.

[0012] Step 5: Obtain the measured value K every 24 hours through in-situ calibration test. If |KK| / K>10%, update the model parameters using the gradient descent method.

[0013] Preferably, in step 1, the distributed optical fiber is laid out along the direction of the principal stress of the project. In dam engineering, it is buried in the joint of the pouring layer; in tunnel engineering, it is pasted on the inner side of the lining; and in slope engineering, it is laid out along the direction of the principal stress of the slope surface. The spacing between the optical fibers in the risk zone is less than 5m. The microseismic sensors are laid out in a triangular array with a burial depth greater than 0.5m and a grouting coupling strength ≥0.8MPa.

[0014] Preferably, in step 3, the four-field coupling governing equations include the seepage field equation and the solute field equation. The seepage field equation is as follows: Solute field equation: In the formula, K is the permeability tensor, h is the piezometric head, ρ is the solid density, and k is the adsorption partition coefficient.

[0015] Preferably, in the third-level algorithm in step 4, AKF updates the Q and R matrices using the recursive least squares method, and the state transition matrix A = 0.98;

[0016] The Attention-LSTM has two hidden layers, and the input features range from K1 to "effective porosity" to "micro-seismic frequency". The Adam optimizer has a learning rate of 0.001, and the Bayesian network is sampled 10,000 times using MCMC.

[0017] Preferably, in the in-situ calibration test in step 5, the soil is subjected to a double-ring water injection test, with an inner ring of φ30cm and an outer ring of φ60cm, and the water head is stable for ≥30min; the rock mass is subjected to a three-stage water pressure test of 0.3, 0.6 and 0.3MPa, with each stage being stable for more than 20min.

[0018] A dynamic permeability coefficient calculation system based on multi-source sensing includes:

[0019] Multi-source sensing module: includes distributed fiber optic unit with BOTDR technology, spatial resolution 1m, strain accuracy ±40με, NMR unit with detection depth 0.5-2m and porosity accuracy ±0.5%, micro-seismic array unit with frequency 1-1000Hz and threshold 10μV, and auxiliary sensing unit for acquiring multi-dimensional data;

[0020] Data preprocessing module: connected to the multi-source sensing module, performs the denoising and alignment operation in step 2 of claim 1;

[0021] Model building module: Includes the four-field coupling model and correlation function from step 3 of claim 1;

[0022] Intelligent inversion module: Deploys the three-level fusion algorithm in step 4 of claim 1;

[0023] Calibration module: Connects to the in-situ calibration device and performs error judgment and parameter update in step 5 of claim 1.

[0024] Preferably, the edge computing nodes of the multi-source sensing module adopt industrial-grade processors and support multi-protocol data access; the correction module includes an error analysis unit and a parameter update unit, and the gradient descent method learning rate is 0.01-0.05 and the iteration is 50-100 times during the update.

[0025] A dynamic permeability coefficient calculation device based on multi-source sensing includes a processor, a memory, and a computer program stored in the memory. When the processor executes the program, it implements the steps of any of the methods described above.

[0026] Preferably, the device further includes a communication interface for connecting the multi-source sensing module and the in-situ calibration device, supporting 5G and Ethernet dual-mode transmission, with a data transmission rate ≥10Mbps.

[0027] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods described above.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] This invention utilizes NMR technology to achieve high-precision analysis of microscopic parameters such as porosity and water content, combined with microseismic sensors to capture the initiation and expansion of microcracks in real time, constructing a correlation mapping between macroscopic and microscopic dimensions. In shield tunnel sand layer monitoring, abrupt changes in permeability coefficient can be identified 10-15 minutes in advance, mitigating the risk of failed water inrush warnings. Furthermore, in field verification of mixed soil and rock slope engineering, the relative errors between calculated and measured values ​​for silty clay at the slope top, sandstone in the middle of the slope, and weak interlayers at the slope toe are all less than 3.4%. It can also detect abrupt changes in permeability coefficient in weak interlayers in advance, avoiding shallow landslide disasters. For chloride ion erosion scenarios in water conservancy projects, the influence of solute concentration on pore structure is quantified through a coupled correlation function K=K(n,ε,C), ensuring that the calculation error of the permeability coefficient of dams operating for many years is controlled within 8%. In deep foundation pit soft soil chemical siltation scenarios, pH value feedback is used to correct solute field parameters, resulting in a long-term permeability coefficient calculation deviation of less than 5%.

[0030] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0031] Figure 1 This is a flowchart of the dynamic calculation method for the permeability coefficient based on multi-source sensing according to the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. It should be noted that the drawings are schematic and not illustrated to scale. For clarity and convenience, the relative sizes and proportions of the parts shown in the drawings have been exaggerated or reduced in size. Any size is only illustrative and not limiting.

[0033] Example 1: Dynamic calculation of permeability coefficient for mixed soil and rock slopes

[0034] This is a slope engineering project for a mountain highway. The slope is 65m high and 45°. The slope body is divided into three layers from top to bottom: the top layer (0-8m) is silty clay, the layer (8-40m) is moderately weathered sandstone, and the layer (40-65m) is moderately weathered conglomerate. A weak interlayer running downhill is also present in the middle of the slope, with a depth of 15-18m and a thickness of 0.8-1.2m. The area experiences abundant rainfall during the rainy season from May to September and has a history of shallow landslides. Investigation has determined that increased pore water pressure caused by seepage is the main cause of these disasters. To achieve real-time seepage safety monitoring and disaster early warning, it is necessary to obtain dynamic data on the permeability coefficient of different media areas to provide technical support for seepage prevention reinforcement and operational risk management in this project.

[0035] A multi-source sensing system was deployed, with three BOTDR optical fibers arranged along the principal stress direction of the slope in a distributed optical fiber unit; two more fibers were densely arranged in the projection area of ​​the weak interlayer, spaced 5m apart; the fibers were fixed by grouting through φ100mm boreholes, with a fiber optic burial depth of 0.5m in clay areas and 1.0m in sandstone areas. Its spatial resolution is 1m, strain measurement accuracy is ±40με, sampling frequency is 1Hz, and temperature compensation range is -20 to 80℃.

[0036] The proton nuclear magnetic resonance (NMR) unit has three monitoring sections at elevations of 0m at the top of the slope, 25m in the middle of the slope, and 65m at the toe of the slope. Two test points are set up at the projection of the surface layer and the weak interlayer of the slope at each section. The tripod is fixed to ensure that the sensor fits the slope surface with ≥95%. The detection depth is 0.5-2.0m, the porosity measurement accuracy is ±0.5%, the test time is 5min / point, and the T2 spectrum inversion range is 0.1-1000ms.

[0037] The microseismic array unit consists of 12 microseismic sensors arranged in an equilateral triangle around the weak interlayer region, with a sensor spacing of 4m. The sensors are embedded in boreholes and coupled using cement mortar grouting, with a coupling strength ≥0.8MPa. Its frequency response range is 1-1000Hz, trigger threshold is 10μV, sampling frequency is 10kHz, and event location error is ≤0.8m.

[0038] In the auxiliary sensing unit, nine pore water pressure gauges are installed at drilling depths of 5m, 15m, and 30m on the slope, corresponding to clay, sandstone, and weak interlayers, respectively; six solute concentration sensors are installed in the drainage ditch at the toe of the slope to monitor the chloride ion concentration at the seepage outlet. The pore water pressure accuracy is ±0.1kPa, with a sampling frequency of 0.1Hz; the chloride ion concentration accuracy is ±1mg / L, and the measurement range is 0-1000mg / L.

[0039] Two industrial-grade edge servers are deployed in the control room at the bottom of the slope, connected to the slope surface sensor network and the deep sensor network respectively; a UPS power supply is provided to ensure 8 hours of continuous operation during power outages. The processor is an Intel Core i5-12400, with 16GB of memory and a processing speed of 1.1GHz; it supports 5G / Ethernet dual-mode transmission.

[0040] The dynamic calculation method for permeability coefficient includes the following steps:

[0041] Edge computing nodes synchronously receive data from various sensors via industrial Ethernet and 5G network, and use timestamp calibration technology to achieve time synchronization of multi-source data. During the rainy season when the daily rainfall is ≥20mm, the sampling frequency is automatically increased to 2Hz, while it remains at 1Hz during the non-rainy season; the measured data transmission delay for a single cross-section is 7.8s.

[0042] Layered denoising was performed. To address vibration interference caused by vehicle traffic and wind loads on the slope, an 8-layer decomposition using the db4 wavelet basis was adopted to remove high-frequency components with a signal-to-noise ratio (SNR) <10dB. Then, empirical mode decomposition (EMD) was used to decompose the data into 11 intrinsic mode functions (IMFs). The first two high-frequency IMF components were removed to eliminate baseline drift. After reconstruction, the data SNR was improved from 26dB to 43dB.

[0043] The Levenberg-Marquardt algorithm was used to perform a double exponential inversion on the T2 spectrum to separate free water (T2=80-450ms) and bound water (T2=3-10ms). The effective porosity of each medium was calculated, namely n=32% for surface silty clay, n=8.5% for moderately weathered sandstone, and n=38% for weak interlayers.

[0044] The energy threshold method is used to screen valid events, and the signal energy E=∫|x(t)| is calculated. 2 dt, where t=1s sampling window, excluding samples with E<10 -3 Environmental disturbance events accounted for 91% of the total number of events; the remaining valid events, after location analysis, showed that 92% were concentrated around the weak interlayer.

[0045] Using the 1Hz timestamp of fiber optic data as a reference, the Dynamic Time Warping (DTW) algorithm is used to align NMR and microseismic data, and the NMR time nodes are completed by linear interpolation; finally, a 3×100-dimensional feature vector is generated.

[0046] Basic parameters were determined through indoor experiments and field surveys, and four-field coupled control equations and correlation functions were established. First, the basic parameters were determined:

[0047] Initial permeability coefficient K0: determined by indoor variable head permeability test, silty clay = 1.5 × 10⁻⁶ -5 cm / s, moderately weathered sandstone = 3.2 × 10 -6 cm / s, weak interlayer = 8.6 × 10 -5 cm / s;

[0048] Mechanical parameters: Lamé constant λ = 1.2 × 10⁻⁶, determined by indoor triaxial compression tests. 5MPa, shear modulus μ=0.7×10 5 MPa, effective stress coefficient α=0.92;

[0049] Solute parameters: Determined through adsorption experiments, the solute adsorption coefficient k_d0 = 0.78, and the standard specific surface area S0 = 3.0 m². 2 / g.

[0050] In the four-field coupled control equations, the seepage field equation introduces the rainfall source-sink term Q (Q = rainfall intensity × slope projected area), considering the influence of the stress field on seepage, and the expression is:

[0051]

[0052] In the formula, K is the permeability tensor; h is the water head; n is the porosity; S is the degree of saturation; ρ is the water density; g is the gravitational acceleration; ε is the volumetric strain; and D is the solute diffusion coefficient, taken as 1.1 × 10⁻⁶. -5 cm 2 / s, where C is the solute concentration and Q is the source / sink term.

[0053] The solute field equation, combined with NMR measurements, indicates that the specific surface area of ​​the weak interlayer pores is S = 3.4 m². 2 / g, the corrected adsorption coefficient k_d = 0.78 × (3.4 / 3.0) = 0.88, the expression is:

[0054] In the formula: ρ is the density of the medium, taken as 2600 kg / m³. 3 v is the seepage velocity.

[0055] By fitting the coupling parameters of each medium using field data (m=4.0 for silty clay, m=2.8 for sandstone, m=4.5 for weak interlayers, a=6.0, b=0.004), a correlation function between permeability coefficient and porosity, strain, and solute concentration was established.

[0056]

[0057] In the formula, n is the initial porosity, m is the porosity influence coefficient, a is the strain influence coefficient, and b is the solute concentration influence coefficient.

[0058] A three-level intelligent inversion architecture is adopted to improve the accuracy of permeability coefficient calculation and realize uncertainty quantification. Adaptive Kalman Filter (AKF) preliminary calculation:

[0059] The equation of state is: K(t) = 0.98 × K(t-1) + w(t), where the coefficient 0.98 is determined based on the daily fluctuation range of the slope permeability coefficient being ≤2%.

[0060] The observation equation is: Z(t) = 1 × K(t) + v(t), where the observed value Z(t) is the permeability coefficient derived from Darcy's law, v = K × i, and i is the hydraulic gradient;

[0061] Noise covariance update: The process noise covariance Q = 1.1 × 10⁻⁶ is updated in real time using the recursive least squares method. -13 The observation noise covariance R = 8.2 × 10⁻⁶ -14 ;

[0062] Based on the pore water pressure data at the slope toe monitoring point, P1 = 0.15 MPa, P2 = 0.145 MPa, and the seepage path length L = 15 m, the hydraulic gradient i = (0.15 - 0.145) × 10⁶ / (1000 × 9.8 × 15) = 0.0034 is calculated. Substituting this into Darcy's law, we get K1: silty clay 1.6 × 10⁶ MPa. -5 cm / s, weak interlayer 9.2×10 -5 cm / s.

[0063] Attention-LSTM optimization calculations were performed, with the input features being a 3×100-dimensional feature vector [K1, effective porosity n, microseismic frequency f]; the training set consisted of 6000 sets of measured data from the first 45 days, including 150 sets of weak interlayer anomaly data; the test set consisted of 2000 sets of data from the last 15 days.

[0064] The weighting of microseismic frequency characteristics was strengthened, with the frequency of microseismic events in the weak interlayer f=4 times / 100s and the weight α_f=0.65; after 450 iterations, the mean square error (MSE) converged to 1.8×10 -13 After output optimization, K2: 1.5 × 10⁻⁶ silty clay -5 cm / s, weak interlayer 8.8×10 -5 cm / s.

[0065] Bayesian network quantization was performed, with fiber strain error ±40με, NMR porosity error ±0.5%, and model parameter error ±8%, which were converted into normal distribution priors. The Metropolis-Hastings algorithm was used to sample 10,000 times, and the first 2,000 combustion period samples were discarded, resulting in an effective sample acceptance rate of 42%.

[0066] Output the mean permeability coefficient and 95% confidence interval: Silty clay K = 1.5 × 10 -5 cm / s ([1.4×10 -5 1.6×10 -5 [cm / s], weak interlayer K=8.8×10 -5 cm / s ([8.5×10 -5 9.1×10 -5 [cm / s]

[0067] To ensure long-term monitoring accuracy, a dynamic correction mechanism was established. In-situ calibration tests were conducted, with three sets of calibration devices deployed at the slope top, middle, and toe. For soil, a double-ring water injection test was used (inner ring φ30cm, outer ring φ60cm). For sandstone, a three-stage water pressure test was used (pressures 0.3MPa, 0.6MPa, and 0.3MPa), with each stage stabilizing for ≥20 minutes. Measured calibration values ​​(K_cal): Silty clay 1.45×10⁻⁶ -5 cm / s, sandstone 3.1×10 -6 cm / s, weak interlayer 8.6×10 -5 cm / s.

[0068] The relative error was calculated as |K - K_cal| / K_cal×100%, with results of 3.4% for silty clay, 3.2% for sandstone, and 2.3% for weak interlayers. All these values ​​are greater than the design threshold of 10%, thus requiring no model parameter updates. Continuous monitoring during the 30-day rainy season showed that the relative errors in each region remained stable between 2.3% and 4.8%, all exceeding the design requirement of ±8%, validating the long-term reliability of this method.

[0069] Three typical media areas of the slope were selected, and the calculated values ​​of this embodiment, the calculated values ​​of the traditional method, and the measured values ​​were compared. The experimental results are shown in Table 1.

[0070] Table 1:

[0071] Monitoring area Conventional method calculated value (10 -5 cm / s The example calculates a value (10 -5 cm / s)]]> Found (10 -5 cm / s) Relative error Slope top silty clay 1.8±0.5 1.5±0.1 1.45 3.4% Slope middle sandstone 4.2±0.8 3.0±0.2 3.1 3.2% Slope toe weak interlayer 11.5±1.6 8.8±0.3 8.6 2.3%

[0072] Note: The traditional method is the "single pore water pressure derivation method", which does not take into account the characteristics of micropores and micro-damage.

[0073] In summary, during periods of heavy rainfall in the rainy season, the system successfully captured the permeability coefficient of the weak interlayer from 8.8 × 10⁻⁶. -5 cm / s changed to 12.5×10 -5 The anomaly of cm / s was detected with a response time of 9.2 seconds, and a leakage warning was issued 12 minutes in advance, guiding the on-site initiation of emergency drainage measures and preventing shallow landslide disasters. Based on the calculated permeability coefficients of each area, the anchor bolt support density was optimized. The anchor bolt spacing in the weak interlayer area was increased from 2.0m to 1.5m, while it remained at 2.0m in other areas, reducing reinforcement costs by 18%. Monitoring data during the one-year operation period showed that the slope seepage flow remained stable below 0.02L / (s·m), and the permeability coefficient showed no significant increasing trend, verifying the effectiveness of the seepage prevention and reinforcement measures.

[0074] Example 2: Dynamic calculation of the permeability coefficient of shield tunnels with lining surrounding rock.

[0075] The 6.2m diameter subway shield tunnel traverses a composite stratum of silty clay and sand, and is lined with C50 precast concrete segments. During construction, it is necessary to monitor the permeability coefficient of the surrounding rock ahead of the tunnel face and the joints of the segments to prevent the risk of water inrush.

[0076] A multi-source sensing system is deployed, with two FBG optical fibers bonded circumferentially along the tunnel segment in a distributed optical fiber unit, with a spacing of 1.2m between each ring; a BOTDR optical fiber is buried in a pre-drilled hole at the working face. The FBG wavelength accuracy is ±1pm, the BOTDR spatial resolution is 1m, and the sampling frequency is 2Hz.

[0077] In the NMR unit, a portable NMR is mounted behind the tunnel boring machine cutterhead, and the tunnel face is tested once every 10m of tunneling. Its detection depth is 0.5-1.5m, and the test takes 3 minutes per test.

[0078] In the microseismic array unit, one set of microseismic sensors is arranged every 5 rings on the inner side of the tube segment, forming a square array. Its frequency response is 5-1500Hz, the trigger threshold is 8μV, and the sampling frequency is 15kHz.

[0079] In the auxiliary sensing unit, 12 pore water pressure gauges are installed at the joints of the tunnel segments; 4 moisture content sensors are deployed at the working face. The pressure accuracy is ±0.1 kPa, the moisture content accuracy is ±0.5%, and the sampling frequency is 0.5 Hz.

[0080] In the edge computing node, a vehicle-mounted edge server is deployed in the tunnel boring machine's control room, connecting all sensors. Its processor is an AMD Ryzen 5 with a processing speed of 1.0GHz and a 5G transmission latency of ≤9s.

[0081] The dynamic calculation method for permeability coefficient based on multi-source sensing includes the following steps:

[0082] To address vibration interference during tunnel boring machine (TBM) excavation, first exclude tunnels with vibrations less than 5 × 10⁻⁶. -3 For the J event, tunneling machinery noise was removed by 5-500Hz bandpass filtering; the NMR data were adjusted for the sand layer by adjusting the double exponential fitting parameters, the free water T2 threshold was reduced to 80ms, and the accuracy of effective porosity calculation was improved to ±0.4%.

[0083] Due to the unloading effect during tunnel excavation, the force field equation is modified as follows: σ = λεδ + 2με - αpδ - σ, where σ is the initial ground stress, measured to be 0.8 MPa. The correlation function is adjusted for sand layers with m = 3.8 and a = 6.2, resulting in:

[0084] Through borehole water injection tests and numerical inversion calibration water pressure tests, namely, drilling ahead of the working face to carry out water injection tests, recording flow rate and head curves, and combining three-dimensional seepage numerical models to invert K; the calibration cycle is shortened to 12 hours, the learning rate of the gradient descent method is increased to 0.05, and the parameters are updated quickly by iterating 80 times.

[0085] The calculated permeability coefficient of the sand layer ahead of the tunnel face is K = 3.2 × 10⁻⁶. -4 cm / s, measured value 3.1×10 -4 cm / s, error 3.2%; segment joint K=4.5×10 -6 cm / s, measured value 4.3×10 -6 cm / s, error 4.7%; response time 9.0s, successfully provided early warning of a water inrush risk at a working face, and guided advanced grouting reinforcement.

[0086] Example 3: Dynamic Calculation of Permeability Coefficient for Deep Foundation Pit in Soft Soil Foundation

[0087] Ultra-deep foundation pit, with an excavation depth of 28m and a pit area of ​​1200m². 2 The support structure consists of a diaphragm wall and internal bracing, and the foundation is silty clay with a water content of 45%. It is necessary to monitor the seepage prevention performance of the diaphragm wall and the permeability coefficient of the soil at the bottom of the pit to prevent the risk of piping.

[0088] A multi-source sensing system was deployed. The microseismic sensors were pre-embedded and buried simultaneously during the pouring of the continuous wall, with a burial depth of 2-5m. The NMR was conducted using ground-based testing, with the test point 1.5m away from the continuous wall and a detection depth of 1.0-2.0m.

[0089] Because soft soil is prone to chemical clogging, a pH sensor is added to the seepage outlet to correct the diffusion coefficient in the solute field equation: D = D × (1 + 0.1 × (7 - pH)) (D = 1.0 × 10⁻⁶). -5 cm 2 / s);

[0090] To address the large fluctuations in soft soil parameters, the accuracy error of the pore water pressure gauge was added as prior information, the number of MCMC samplings was increased to 15,000, and the confidence interval width was reduced by 30%.

[0091] To verify Example 3, because soft soil foundations have poor resistance to disturbance, traditional borehole water pressure tests are prone to damaging the soil structure and causing data distortion. Therefore, a double-ring water injection test was selected as the method for obtaining measured values, covering three types of core seepage prevention areas. The steps are as follows:

[0092] One φ50mm test hole was drilled at 5m, 15m, and 25m from the top of the diaphragm wall, penetrating the wall thickness. Rubber sealing devices were used to seal the hole walls to prevent leakage. Three sets of test points were laid out in a 20m×20m grid at the bottom of the pit. After the pit bottom was excavated and leveled, a 5mm thick geotextile was laid to isolate disturbance, and the test device was placed in the center of the geotextile. Test points were set at the midpoints of the joints of three adjacent pipe sections at the diaphragm wall joints. After cleaning the joint surface, it was sealed with cement-based sealant to prevent surface water seepage. Each test used a double-ring device with an inner ring of φ30cm and an outer ring of φ60cm. The inner ring was used to measure the actual seepage flow in the soil, and the outer ring was used to isolate surrounding water interference. The device was inserted into the soil to a depth of 10cm.

[0093] Water was injected into the inner and outer rings simultaneously to the design head, and the head was kept stable. Initially, the flow rate was recorded once every 2 minutes. After the flow rate fluctuation was less than 5%, the record was changed to once every 5 minutes. The total stabilization time was greater than 30 minutes.

[0094] To address the slow seepage characteristics of soft soil, the test cycle was extended to 60 minutes to ensure complete flow stability. Each test was repeated three times, outliers were removed, and the average value was taken as the basic measured data for the area.

[0095] According to the formula: Where Q: steady-state seepage flow rate (unit: cm³) 3 / s), take the average of 3 stable readings; R: outer ring radius (30cm), r: inner ring radius (15cm); H: test head (30cm), t: test duration.

[0096] The final measured values ​​for the three types of areas were: 3.1 × 10⁻⁶ for continuous wall. -7 cm / s, soil mass at the bottom of the pit: 8.2×10 -7 cm / s, continuous wall joint 12.0×10 -7 cm / s.

[0097] Based on the calculated values ​​derived from the entire technology chain, the optimization steps for soft soil characteristics are as follows:

[0098] Microseismic sensors are pre-embedded in the reinforcing cage during continuous wall construction, secured by binding to prevent displacement. Grouting coupling strength is ≥0.8MPa to minimize soft soil disturbance. NMR sensors are ground-based, with a detection depth of 1.0-2.0m, avoiding disturbed areas at the bottom of the pit. Ground debris is cleared before testing to ensure a fit greater than 95%. Among auxiliary sensors, a pore water pressure gauge is embedded at half the borehole depth at the bottom of the pit, and a pH sensor is added at the seepage outlet to monitor the chemical siltation effect of soft soil. Data transmission utilizes edge computing nodes and a 5G mode, with a single data transmission latency of 9.2s.

[0099] Data preprocessing was performed. The pore water in soft soil was mainly bound water. The T2 spectrum double exponential fitting threshold was adjusted, with free water T2≥60ms ​​and bound water T2≤10ms. The effective porosity n was obtained as follows: 2.8% for continuous wall and 32% for pit bottom soil. The microseismic signal of soft soil foundation was weak. A dual mechanism of energy threshold and low frequency filtering was adopted to eliminate interference from construction machinery, and the effective event recognition rate was improved to 82%. Based on the pore water pressure gauge sampling frequency of 0.5Hz, the dynamic time warping (DTW) algorithm was used to align NMR (10min / point) and microseismic (10kHz) data. Data synchronization and alignment were achieved by interpolating the time nodes three times.

[0100] Four-field coupled modeling and inversion revealed that soft soil is prone to chemical clogging. The solute field diffusion coefficient D was corrected as follows: D = D0 × (1 + 0.1 × (7 - pH)), where D0 = 1.0 × 10⁻⁶. -5 cm 2 / s, measured pH value 7.8, calculated D = 1.08 × 10 -5 cm 2 / s;

[0101] By fitting the parameters of the coupling correlation function through indoor soft soil seepage tests, the influence coefficients of soft soil porosity (m) and concentration (a) were found to be 4.2, 6.8, and 0.005, respectively. The correlation function is as follows:

[0102]

[0103] The Bayesian network adds the pore water pressure gauge accuracy error as prior information, and the number of MCMC samplings is increased to 15,000. The first 3,000 combustion period samples are removed to ensure the convergence of the probability distribution.

[0104] The calibration cycle was shortened from 24 hours to 12 hours to adapt to the excavation progress; the learning rate of the gradient descent method was increased to 0.04, and the parameters were updated quickly with 80 iterations, finally outputting the mean and standard deviation of the calculated values;

[0105] Continuous wall structure: 3.2±0.3×10 -7 cm / s; Soil mass at the bottom of the pit: 8.5±0.6×10 -7 cm / s; Continuous wall joint: 12.3±0.9×10 -7 cm / s.

[0106] Quantitative verification is performed, and the relative error is calculated using the following formula:

[0107]

[0108] The error of the traditional method is obtained by simultaneously using the single pore water pressure derivation method to calculate the permeability coefficient, with the formula K=v / i, where v is the seepage velocity and i is the hydraulic gradient. The error of the traditional method is then obtained by comparing it with the measured values.

[0109] For continuous wall structures, the traditional calculated value is 3.7 × 10⁻⁶. -7 cm / s, error (3.7-3.1) / 3.1×100%≈18.7%; for the soil at the bottom of the pit, the traditional calculated value is 10.0×10 cm / s. -7 cm / s, error (10.0-8.2) / 8.2×100%≈22.1%; for continuous wall joints, the traditional calculated value is 15.1×10 -7 The speed was cm / s, and the error was (15.1-12.0) / 12.0×100%≈25.4%. The experimental results are shown in Table 2.

[0110] Table 2:

[0111] Monitoring position The example calculates a value (10 -7 cm / s)]]> Double loop water injection measured value (10 -7 cm / s) Relative error Error of traditional method Continuous wall wall 3.2±0.3 3.1 3.2% 18.7% Pit bottom soil body 8.5±0.6 8.2 3.7% 22.1% Continuous wall joint 12.3±0.9 12.0 2.5% 25.4%

[0112] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0113] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dynamic calculation method of permeability coefficient based on multi-source sensing, characterized in that, The method comprises the following steps: Step 1: In the construction monitoring area, according to the principle of hierarchical distribution and risk area encryption, distributed optical fiber, proton nuclear magnetic resonance (NMR), microseismic sensor and auxiliary sensor of pore water pressure and solute concentration are arranged, and strain, pore water distribution, microseismic event, pore water pressure and solute concentration data are synchronously collected through edge computing nodes and 5G transmission; Step 2: The collected data is denoised and time-space aligned, including 8-layer decomposition of optical fiber data by db4 wavelet packet, empirical mode decomposition denoising, NMR data is fitted by T2 spectrum double exponential to analyze effective porosity, microseismic data is screened by energy threshold to obtain effective events, and multi-source data is aligned by dynamic time warping algorithm; Step 3: The four-field coupling control equations of seepage, heat, force and solute are established, the solute adsorption and desorption kinetics model is introduced, and the multi-field coupling is carried out through the permeability coefficient correlation function K=K(n,ε,C), wherein n is the effective porosity, ε is the volume strain, and C is the solute concentration; Step 4: The three-level algorithm of adaptive Kalman filter (AKF), attention mechanism LSTM and Bayesian network is adopted, AKF is used to calculate the permeability coefficient K1, Attention-LSTM is used to optimize K2, and Bayesian network is used to output K and 95% confidence interval; Step 5: The measured value K is obtained by in-situ calibration test every 24 hours, and if |K-K| / K>10%, the model parameters are updated by gradient descent method.

2. The multi-source sensing based dynamic calculation method of the permeability coefficient according to claim 1, characterized in that: In step 1, the distributed optical fiber is arranged along the main stress direction of the project, the dam project is buried in the pouring layer joint, the tunnel project is pasted on the inner side of the lining, and the slope project is arranged along the main stress direction of the slope surface, and the interval of the risk area is less than 5m.

3. The multi-source sensing based dynamic calculation method of the permeability coefficient according to claim 1, characterized in that: In step 3, the four-field coupling control equations include the seepage field equation and the solute field equation, the seepage field equation: ; and the solute field equation: ; wherein K is the permeability tensor, h is the piezometric head, p is the solid phase density, and k is the adsorption partition coefficient; Wherein, the source and sink term Q is suitable for the rainfall recharge scene of slope in rainy season, and the value is rainfall intensity x slope projection area.

4. The multi-source sensing based dynamic calculation method of permeability coefficient according to claim 1, characterized in that, In the three-level algorithm of step 4, AKF updates Q and R matrix by recursive least squares method, and state transition matrix A=0.98; Attention-LSTM contains 2 hidden layers, the input features are K1 to "effective porosity" to "microseismic frequency", and the learning rate of Adam optimizer is 0.001; Bayesian network is sampled by MCMC for 10000 times.

5. The multi-source sensing based dynamic calculation method of permeability coefficient according to claim 1, characterized in that, In the in-situ calibration test of step 5, the double-ring water injection test is used for soil body, the inner ring is φ30cm, the outer ring is φ60cm, and the water head is stable for ≥30min; the rock mass is tested by 0.3, 0.6 and 0.3MPa three-stage pressure water test, and each stage is stable for more than 20min.

6. A dynamic calculation system of permeability coefficient based on multi-source sensing, characterized in that, It comprises: Multi-source sensing module: containing distributed optical fiber unit, NMR unit, microseismic array unit and auxiliary sensing unit, used for collecting multi-dimensional data; Data preprocessing module: connected with the multi-source sensing module, and performing the denoising and alignment operation of step 2 in claim 1; Model building module: built-in four-field coupling model and correlation function of step 3 in claim 1; Intelligent inversion module: deploying the three-level fusion algorithm of step 4 in claim 1; Correction module: connected with the in-situ calibration device, and performing the error judgment and parameter updating of step 5 in claim 1.

7. The multi-source sensing based dynamic hydraulic conductivity calculation system of claim 6, wherein, The edge computing node of the multi-source sensing module adopts an industrial-grade processor and supports multi-protocol data access; the correction module comprises an error analysis unit and a parameter updating unit, and the learning rate of the gradient descent method is 0.01-0.05 and the iteration is 50-100 times during updating.

8. A multi-source sensing based dynamic permeability coefficient calculation device, characterized in that, The computer program is stored in the memory and comprises computer program code which, when executed by the processor, causes the processor to carry out the steps of the method according to any one of claims 1-5.

9. The apparatus of claim 8, wherein, The device further comprises a communication interface for connecting the multi-source sensing module and the in-situ calibration device, supporting 5G and Ethernet dual-mode transmission, and the data transmission rate is greater than or equal to 10 Mbps.

10. A computer-readable storage medium, characterized in that, The medium stores a computer program, and the program is executed by the processor to realize the steps of the method according to any one of claims 1-5.

Images