A crack grouting self-adaptive regulation method and system based on dynamic water pressure feedback

Abstract

The application provides a crack grouting self-adaptive regulation method and system based on dynamic water pressure feedback, and belongs to the technical field of grouting simulation, and comprises the following steps: collecting grouting area data in real time and obtaining initial characteristics by processing; obtaining crack state prediction values based on the initial characteristics by using a trained dynamic water pressure response model, wherein the dynamic water pressure response model is constructed by fusing fuzzy rules and neural networks; simultaneously obtaining a real-time water pressure change rate from the initial characteristics, and obtaining a water pressure real-time normalization rate by using a constructed water pressure trend normalization model; fusing and calculating a grouting pressure regulation value and a flow dynamic regulation value according to the crack state prediction values and the water pressure real-time normalization rate; and performing regulation according to the grouting pressure regulation value and the flow dynamic regulation value and entering a feedback closed loop. The application can extract key water pressure change characteristics on the basis of monitoring data, judge whether cracks are saturated or blocked, and automatically optimize grouting parameters according to the characteristics, so that closed-loop intelligent control is realized.

Description

Technical Field

[0001] This invention belongs to the field of grouting simulation technology, and particularly relates to an adaptive control method and system for fracture grouting based on dynamic water pressure feedback. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Fissure grouting is an important construction technique commonly used in geotechnical engineering projects such as tunnels, hydraulic structures, and slope reinforcement. It primarily involves injecting grouts, such as cement or water glass, into rock fissures to fill voids, seal leaks, and enhance overall strength. However, in actual engineering environments, the development of fissures often exhibits high uncertainty and heterogeneity. Parameters such as fissure spatial distribution, width variation, connectivity, and permeability are difficult to accurately control. This leads to traditional grouting methods relying on empirical parameters for design, lacking the ability to perceive and respond to real-world conditions. Consequently, problems such as under-grouting, over-grouting, and grout leakage are easily caused, affecting construction effectiveness and cost control.

[0004] To improve the adaptability and intelligence of grouting construction, explorations have gradually emerged in recent years to introduce sensing technology, data acquisition systems, and feedback control into the grouting process. For example, some studies have used monitoring equipment such as water pressure gauges and flow meters deployed in the grouting area to record parameter changes during the grouting process, in order to assist construction personnel in judging the crack filling status. However, such methods are mostly post-event analysis or rely on manual judgment, lacking real-time intelligent data processing, unable to dynamically adjust grouting parameters according to crack response behavior, with coarse control granularity and delayed response.

[0005] Existing control systems are often based on pre-set static grouting schemes, lacking closed-loop feedback to dynamic changes on-site. This is especially problematic in areas with complex geological conditions or drastic fissure changes, where traditional control strategies cannot adapt to the actual situation in a timely manner. Furthermore, while some intelligent grouting systems have attempted to construct grout diffusion models or introduce IoT data acquisition technology, most only achieve monitoring and display functions, lacking efficient information processing mechanisms and intelligent decision-making capabilities, and have not yet formed a systematic framework of "perception-analysis-judgment-control". Summary of the Invention

[0006] To overcome the shortcomings of the existing technologies, this invention proposes an adaptive control method and system for fracture grouting based on dynamic water pressure feedback. Based on real-time water pressure feedback, the method identifies the grouting status of fractures and dynamically adjusts the grouting strategy. The method and system can extract key water pressure change characteristics based on monitoring data, determine whether the fracture is saturated or blocked, and automatically optimize grouting parameters accordingly to achieve closed-loop intelligent control. This improves the accuracy, safety, and automation level of grouting construction, meeting the urgent need for efficient and intelligent grouting construction in current complex engineering environments.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] In a first aspect, the present invention discloses an adaptive control method for fracture grouting based on dynamic water pressure feedback, comprising:

[0009] Real-time acquisition and processing of grouting area data yields initial features;

[0010] Based on the initial features, a trained dynamic water pressure response model is used to obtain the predicted value of the fracture state. The dynamic water pressure response model is constructed by fusing fuzzy rules and neural networks. The fuzzy rules perform fuzzy determination on the state of fracture grouting to obtain a state membership vector. The initial features and the state membership vector are fused and output as the predicted value of the fracture state through a neural network. At the same time, the real-time water pressure change rate is obtained from the initial features, and the real-time water pressure normalization rate is obtained by using the constructed water pressure trend normalization model.

[0011] The grouting pressure adjustment value and flow rate dynamic adjustment value are calculated based on the predicted value of the fracture state and the real-time normalized rate of water pressure.

[0012] The control is executed based on the grouting pressure adjustment value and the flow rate dynamic adjustment value, and then enters the feedback closed loop.

[0013] Secondly, this invention discloses an adaptive control system for fracture grouting based on dynamic water pressure feedback, comprising:

[0014] The data acquisition module is configured to: acquire grouting area data in real time and process it to obtain initial features;

[0015] The feature extraction module is configured to: obtain a predicted value of the fracture state based on the initial features using a trained dynamic water pressure response model; the dynamic water pressure response model is constructed by fusing fuzzy rules and a neural network; the fuzzy rules perform fuzzy determination on the state of fracture grouting to obtain a state membership vector; the initial features and the state membership vector are fused and output as a predicted value of the fracture state through a neural network; simultaneously, the real-time water pressure change rate is obtained from the initial features, and the real-time water pressure normalization rate is obtained using the constructed water pressure trend normalization model;

[0016] The parameter calculation module is configured to calculate the grouting pressure adjustment value and the flow dynamic adjustment value based on the predicted fracture state value and the real-time normalization rate of water pressure.

[0017] The control execution module is configured to perform control based on the grouting pressure adjustment value and the flow dynamic adjustment value and enter the feedback closed loop.

[0018] Thirdly, the present invention discloses an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when run by the processor, complete the steps of the above-mentioned adaptive control method for fracture grouting based on dynamic water pressure feedback.

[0019] Fourthly, the present invention discloses a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-mentioned adaptive control method for fracture grouting based on dynamic water pressure feedback.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0021] This invention constructs a dynamic water pressure response model (BP-FRM) and combines multi-dimensional features such as water pressure change rate, fluctuation amplitude, and trend slope to achieve automatic identification of states such as "unfilled, partially filled, near saturation, fully saturated, and abnormal" during the fracture grouting process. This breaks through the traditional method of relying on manual experience for judgment, realizes real-time identification and intelligent judgment of fracture grouting status, and improves identification accuracy and real-time performance.

[0022] This invention introduces a normalized water pressure trend model to improve dynamic adaptability. It uses the normalized water pressure change rate (Rv) to determine the "fast / slow / intense" nature of grouting behavior. Combined with the target rate reference, it enables early warning of abnormal grouting reactions (such as blockage, leakage, and grout bursting), effectively preventing misjudgment and construction risks under extreme working conditions.

[0023] This invention innovatively proposes a linear fusion control model, which integrates state discrimination and trend analysis to calculate quantifiable grouting parameter adjustments. By combining filling state and water pressure trend information, it calculates the dynamic adjustment values ​​(ΔP) of grouting pressure and flow rate. g (ΔQ), upgrading from "qualitative control" to "quantitative control", with higher control precision and a higher level of automation.

[0024] This invention constructs a closed-loop control system of "monitoring—identification—calculation—execution—feedback—optimization," realizing the linkage of the entire process from data acquisition to execution control. The system supports adaptive model updates, can continuously learn and optimize under different geological conditions, and has strong universality and adaptability.

[0025] By introducing physical constraint rules (such as state-water pressure contradiction judgment), a safety error correction mechanism is added on the basis of intelligent recognition results to prevent erroneous termination of grouting or blind pressurization, thereby improving the engineering reliability and control credibility of the system, enhancing the system's safety and engineering interpretability, and facilitating understanding and intervention by on-site construction personnel.

[0026] This invention is applicable to the intelligent grouting control needs under complex geological conditions, and is particularly suitable for complex geological scenarios with uneven fracture development, large permeability variations, and high construction risks, such as deep-buried tunnels, high slopes, and reservoir dam foundations. It significantly improves grouting quality, efficiency, and automation level, and has good prospects for promotion and application.

[0027] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0028] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0029] Figure 1 This is a flowchart of the adaptive control method for fracture grouting based on dynamic water pressure feedback as described in Embodiment 1 of the present invention. Detailed Implementation

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0032] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0033] Example 1

[0034] In one or more embodiments, an adaptive control method for fracture grouting based on dynamic water pressure feedback is disclosed, such as... Figure 1 As shown, it includes the following steps:

[0035] Step S1: Collect and process data from the grouting area in real time to obtain initial features.

[0036] Step S1-1: Real-time data acquisition and processing; Multiple water pressure sensors are deployed in the grouting area, and the following data are collected every Δt seconds: water pressure time series P(t), grouting pressure P g (t), grouting flow rate Q(t), and cumulative grouting time T(t).

[0037] The collected data undergoes preliminary preprocessing, including noise reduction, time synchronization, and sliding window smoothing, for subsequent analysis.

[0038] The water pressure time series P(t) refers to the sequence of water pressure values ​​continuously collected by the system over time during the grouting process, P(t) = {P(t1), P(t2), ..., P(t...} n )}, where P(t) i ) represents the water pressure value at the i-th sampling time, in kPa.

[0039] For example, by collecting pressure data every 10 seconds, a sequence of water pressure data that varies over time will be obtained:

[0040] P(t1), P(t2), P(t3), …, P(t n )

[0041] That is: P = {p1, p2, p3, ..., p n},p i =P(t i )

[0042] Among them, t i It is the i-th sampling time.

[0043] Step S1-2: Water pressure feature extraction and normalization; Within the time window [t-kΔt,t], extract the following features, including the rate of water pressure change, the amplitude of water pressure fluctuation, and the slope of the water pressure trend:

[0044] ① The rate of change of water pressure is expressed as follows:

[0045] (1)

[0046] Where v(t) is the rate of change of water pressure at the current moment, usually expressed in kPa / s; P(t) i P(t) represents the water pressure value at the current moment; i -Δt) is the water pressure value at the previous sampling time; Δt is the time interval between two adjacent samplings, in seconds (s).

[0047] ② The standard deviation of water pressure fluctuation is expressed as follows:

[0048] (2)

[0049] Where σ(t) is the standard deviation of water pressure within the current window, which measures the intensity of water pressure fluctuation; k is the number of sampling points included in the sliding time window (e.g., if there are 10 samples in the past 30 seconds, then k=10); P(i) is the water pressure value of the i-th sampling point. The average water pressure inside the window is calculated as follows:

[0050] (3)

[0051] ③ The slope of the water pressure trend was extracted by using linear regression to fit the water pressure sequence within the window as a straight line:

[0052] P(t)=a(t)•t+b(t) (4)

[0053] Where a(t) is the slope of the water pressure change trend within the current time window, i.e. the fitted increase rate per unit time; b(t) is the intercept of the fitted line corresponding to the current moment (in this embodiment, this parameter is not used for judgment, but is only a mathematical term).

[0054] In this embodiment, the least squares method is used to fit the points (t). i ,P i );

[0055] When a(t)>0: water pressure is increasing (cracks are being filled); a(t)≈0: water pressure is stable (close to saturation); a(t)<0: water pressure is decreasing (there may be pressure relief or seepage escape).

[0056] Steps S1-3: Construct initial features X(t)=[v(t),σ(t),a(t),P] based on the above data and features. g (t),Q(t),T(t)].

[0057] In this embodiment, the rate of change of water pressure reflects the diffusion speed of the grout, and the amplitude of water pressure fluctuation reflects the disturbance of the flow field. Large fluctuations may indicate instability of the fracture structure or problems such as grout impact or flow around the fracture. This embodiment considers parameters such as the rate of change of water pressure, the amplitude of water pressure fluctuation, and the trend slope to jointly construct a feature vector. Specifically, the rate of change of water pressure v(t) reflects the rate of change of water pressure per unit time, indirectly representing the diffusion efficiency of the grout in the fracture system, and is an important indicator for judging the effectiveness of grouting. When v(t) is too small, it may indicate that the fracture has not been effectively filled or that there is excessive leakage; when v(t) increases significantly, it can enter the steady injection or saturation judgment stage. The amplitude of water pressure fluctuation σ(t) represents the degree of fluctuation of water pressure within the sliding time window and is an indicator of the stability of the flow field. When the fluctuation amplitude is large, it may indicate drastic changes in the fracture channel, grout obstruction, impact flow, or backflow, indicating that the system needs to slow down the grouting speed or switch to the pressure holding mode to prevent abnormal seepage or grout bursting. The trend slope a(t) is the slope value obtained by fitting the water pressure time series trend line, used to characterize the overall trend of water pressure change. If a(t) > 0, it indicates that the water pressure continues to rise and the cracks are still being filled; if a(t) ≈ 0, the water pressure is stabilizing and the grouting may be close to saturation; if a(t) < 0, there may be leakage, escape channels, or system depressurization. These three features together constitute the system's dynamic perception mechanism of the grouting response state, playing a core role in model recognition and adaptive control. If any one of these features is ignored, the system's recognition ability will be limited. For example, without considering water pressure fluctuations, it will be impossible to effectively identify the risk of crack splitting or unstable seepage; without considering the trend slope, the termination time may be misjudged, resulting in insufficient grouting or material waste. Therefore, the synergistic extraction and joint input of the three features are the fundamental guarantee for achieving high-reliability control in this embodiment.

[0058] Step S2: Based on the initial features, a trained dynamic water pressure response model is used to obtain the predicted value of the fracture state. The dynamic water pressure response model is constructed by fusing fuzzy rules and neural networks. The fuzzy rules perform fuzzy judgment on the state of fracture grouting to obtain the state membership vector. The initial features and the state membership vector are fused and output as the predicted value of the fracture state through a neural network. At the same time, the real-time water pressure change rate is obtained from the initial features. Based on the real-time water pressure change rate, the constructed water pressure trend normalization model is used to obtain the real-time water pressure normalization rate.

[0059] Step S2-1: Construct a mapping model of "water pressure characteristics → fracture state" by fusing fuzzy rules and neural networks: dynamic water pressure response model, namely neural network-fuzzy rule model (BP-FRM model).

[0060] The dynamic water pressure response model consists of a fuzzy rule layer, a neural network prediction layer, and a physical constraint correction layer connected in sequence.

[0061] Step S2-1-1: Construct a fuzzy rules model (FRM) layer. Construct membership degrees based on empirical rules. In the system, the state of crack grouting is divided into multiple levels, resulting in a state membership vector μ=[μ0,μ1,μ2,μ3,μ4], where μ0 is unfilled, μ1 is partially filled, μ2 is close to saturation, μ3 is fully saturated, and μ4 is abnormal or blocked.

[0062] It should be understood that μ represents the value of the "membership function," which is a numerical value used in fuzzy logic systems to describe the degree to which an input value belongs to a certain state. Its numerical range is μ∈[0,1]. At a certain time t, after the system performs a fuzzy determination on the state of the fracture grouting, it outputs a state membership vector: μ(t)=[μ0(t),μ1(t),μ2(t),μ3(t),μ4(t)]. In fuzzy logic, each state (e.g., "unfilled," "partially filled," etc.) corresponds to a membership function μ(x), which maps a certain input variable x (e.g., the rate of water pressure increase) to a membership value between [0,1].

[0063] The value of the membership function It is obtained through the following process:

[0064] Water pressure change rate As input variables, fuzzy mappings are constructed using triangular and trapezoidal membership functions, which are applicable to actual grouting conditions.

[0065] Unfilled state The calculation is performed using a trapezoidal function at a relatively low rate of water pressure rise (before the slurry affects the sensing point), as shown in the following formula:

[0066] (5)

[0067] Partially filled state Using a moderate to low slurry flow rate, the slurry begins to enter the cracks but does not diffuse uniformly. The calculation is performed using trigonometric functions, as shown in the following formula:

[0068] (6)

[0069] Approaching saturation The calculation shows that as water pressure steadily increases, flow resistance increases. Using trigonometric functions, the following formula is used:

[0070] (7)

[0071] Fully saturated state Calculation: With almost no increase in water pressure, the fluid cannot expand further. The trapezoidal function is used for calculation, as shown in the following formula:

[0072] (8)

[0073] Abnormal state Calculation: An abnormally high rate of water pressure change may indicate blockage, backflow, or sluice gate burst. Use an S-curve or incremental trapezoidal function.

[0074] (9)

[0075] In this embodiment, the membership function value μ provides the state probability distribution. Unlike hard classification, which only determines which state a system belongs to, fuzzy membership can express "fuzzy transitions". This further improves the robustness of the system. Multiple μ values ​​can be used to make weighted judgments to avoid misjudgments due to errors. As input to the subsequent neural network, μ can be combined with the original features to form a fusion vector, improving prediction accuracy.

[0076] It should be understood that the membership function in this embodiment adopts an overlapping design. The purpose is to achieve a smooth transition between different states, improve the model's fault tolerance to state boundaries, and avoid state jumps or misjudgments caused by feature perturbations. Especially during the crack grouting process, the field measurement data inevitably contains fluctuations and delays. If there is no overlapping area, the model will exhibit discontinuous responses in the boundary region, reducing control stability.

[0077] Step S2-1-2: Construct the neural network prediction layer (BP);

[0078] First, construct the input combination vector for the prediction layer of the neural network:

[0079] Z(t)=[X(t),μ(t)] (10)

[0080] The input combination vector Z(t) integrates the physical perception layer X(t) and the cognitive logic layer μ(t), unifying hard measurements (such as v, P) and soft judgments (such as "may be a partial fill"), and is the core input of the BP-FRM model.

[0081] Furthermore, formula (10) can be written as:

[0082] Z(t)=[X(t),μ(t)]=[v(t),σ(t),a(t),P g (t),Q(t),T(t),μ0,μ1,μ2,μ3,μ4] (11)

[0083] The above formula is the input feature vector used for the state recognition model, with a dimension of 11.

[0084] Secondly, modeling can be done using feedforward neural networks (such as backpropagation neural networks):

[0085] S(t)=f(Z(t);θ) (12)

[0086] Where f is the trained neural network function (activation + weighting + bias); θ is the model parameters, including the weight matrix and bias term; and S(t) is the predicted value of the fracture state.

[0087] Specifically, the neural network structure includes an input layer with 11 nodes (Z vector dimension), a hidden layer with 1 to 2 layers (e.g., 8 or 6 neurons), and an output layer with 1 node (state value).

[0088] For the forward propagation of a neural network, assuming the hidden layer output is h, then:

[0089] The formula for hidden layers is:

[0090] h=ReLU(W1Z(t)+b1) (13)

[0091] The formula for the output layer is:

[0092] S(t)=σ(W2h+b2) (14)

[0093] Where W1 and W2 are weight matrices; b1 and b2 are bias vectors; ReLU(•) is the activation function; and σ(•) is the output activation function.

[0094] The neural network automatically learns complex nonlinear mapping relationships by fitting a function f(Z) to training data. This embodiment employs a neural network model that can be trained and updated for different strata types, fracture characteristics, and grouting parameters. It exhibits strong adaptability and dynamic adjustability, does not rely on fixed empirical rules, and is suitable for complex and ever-changing underground engineering environments. Through multi-layer nonlinear transformations, it possesses strong expressive power, enabling the network to capture the deep coupling relationships between multi-dimensional features such as water pressure rate, fluctuations, and trends. This achieves high-resolution identification of ambiguous boundary states and improves the accuracy of filling state prediction.

[0095] Furthermore, a properly trained and validated neural network has strong generalization ability, can maintain stable output on unlabeled new data, has good generalization ability, and reduces the risk of misjudgment; the model can continuously update parameters through actual feedback data on site, and achieve closed-loop learning and performance evolution through sustainable optimization, providing sustainable intelligent support for long-term engineering applications.

[0096] The neural network prediction layer outputs predicted crack states:

[0097] S(t)∈[0,4] (15)

[0098] This yields a five-level filling state code for the crack state.

[0099] Step S2-1-3: Construct a physical constraint correction layer to perform physical logic verification and backoff correction on the filled state value S(t) output by the neural network prediction layer, thereby enhancing the engineering controllability and safety of the model under extreme working conditions or prediction errors. This layer introduces a "consistency judgment mechanism between predicted values ​​and physical measured trends" by embedding water pressure response rules based on engineering experience.

[0100] Utilize: if 2.8<=S(t)<=3.2 and dP / dt>ε:

[0101] S(t) := 2.5# or directly reset to 2

[0102] Here, ε is a threshold value for the rate of change of water pressure, measured in kPa / s, representing the “acceptable range” of water pressure increase, and is usually set based on historical data or construction experience.

[0103] The above judgment logic is as follows: when S(t)∈[2.8,3.2], the model considers it to be in a fuzzy state of "close to complete saturation"; and dP / dt is greater than the preset threshold, indicating that the water pressure is still rising rapidly, which does not conform to the physical law that saturation should be stable; then the forced rollback S(t):=2.5 is made to adjust the judgment, making the system more conservative and avoiding premature stop betting.

[0104] Specifically, the different measures taken for different values ​​of S(t) are as follows:

[0105] 0≤S(t)<1.0: No filling (slow water pressure rise, low flow resistance), increase grouting pressure and flow rate to quickly advance filling;

[0106] 1.0≤S(t)<2.0: Partial filling (grout has entered part of the cracks), maintain pressure and slightly increase flow rate, observe water pressure response;

[0107] 2.0≤S(t)<2.8: Approaching saturation (water pressure rises slowly, trend stabilizes), maintain current parameters or appropriately reduce them to enter the stable injection stage;

[0108] 2.8≤S(t)≤3.2: Complete saturation (preliminary judgment). If dP / dt<ε, prepare to terminate grouting. If the water pressure continues to rise, the judgment is reversed to 2.5.

[0109] S(t)>3.2 and dP / dt>ε: Saturation is misjudged, water pressure rises sharply, physical rollback is executed: the state is forcibly corrected to 2.5 or 2.0, and the parameters are reduced to avoid bursting;

[0110] S(t)≈4.0 and σ(t) is large: Abnormal state (blockage, reversal), immediately stop grouting and alarm to prompt manual inspection.

[0111] When the neural network predicts a state of "complete saturation" or "near saturation" (i.e., S(t) > 2.8), but the rate of increase in water pressure, dP / dt, is still significantly greater than the safety threshold ε (e.g., 0.05 kPa / s), it is determined that there is a physical conflict in the current fracture state assessment. Understandably, theoretically, if saturation is achieved, the water pressure should tend to stabilize. However, if the measured pressure continues to rise sharply, it indicates that the model risks prematurely terminating the assessment. In this case, the system automatically reverts S(t) to a lower state value (e.g., 2.0 or 2.5) and adjusts the control parameters to avoid erroneous termination of grouting or causing a grout burst. Similarly, if the predicted state is too low (e.g., S(t) < 1), but the water pressure remains consistently high and stable, the system can trigger an upward adjustment or freeze assessment mechanism to avoid over-injection.

[0112] The physical constraint correction layer enhances the physical consistency of the model: it integrates the objective laws of the grouting process with the data-driven model to ensure that the output results meet the basic hydraulic logic; it provides a fallback judgment in abnormal states (such as blockage, depressurization, and sudden expansion) to prevent misadjustment or loss of control and improve engineering safety; it flexibly adjusts the neural network prediction results to avoid small errors causing drastic parameter fluctuations and improves fault tolerance and control stability; this layer is in the form of explicit rules, which improves interpretability and debuggability, facilitates human understanding, adjustment and intervention, and facilitates system integration and practical promotion.

[0113] Step S2-2: Construct a water pressure trend normalization model to determine the rate of change of the current water pressure, and quantify it as a normalization index R. v (t).

[0114] Step S2-2-1: Calculate the reference rate of rise, as shown in the following formula:

[0115] (16)

[0116] Among them, v ref The reference rise rate is the average rate of water pressure change required for the fracture to rise from its initial state to its target state under ideal conditions; P 目标 The target water pressure is the water pressure value that the system is judged to be "near saturation" or "fully saturated" based on design or experience; P 起始 The initial water pressure is the average value of the water pressure at the beginning of the grouting section (before grouting begins or in the first minute); T 估算 The theoretical filling time estimate, i.e. the estimated time required to reach the target water pressure after grouting, can be estimated from historical data, grouting volume, etc.

[0117] Step S2-2-2: Calculate the real-time normalized rate, as shown in the following formula:

[0118] (17)

[0119] Among them, R v(t) represents the real-time normalized rate of water pressure; v(t) represents the real-time rate of change of water pressure; V ref For reference rate.

[0120] This embodiment calculates R. v (t), if R v If (t) < 0.5, it indicates that the actual rate of water pressure change is too slow, possibly due to weak grouting seepage or large, unfilled cracks; if 0.5 ≤ R v (t)≤1.5, in normal filling state; if R v If (t)>2.0, it indicates a sharp increase in water pressure, which may be a risky state such as closure, abnormality, or blockage.

[0121] Step S3: Calculate the grouting parameter control value based on the predicted value of the fracture state and the real-time normalized rate of water pressure.

[0122] The fusion calculation of grouting parameter control values ​​is based on the state judgment result S(t) and the trend normalization rate R. v (t), calculate the grouting parameter control values, including the grouting pressure control value and the dynamic control value of the flow rate.

[0123] The grouting pressure adjustment value is calculated using the following formula:

[0124] (18)

[0125] Where, Δ P g (t) Δ represents the grouting pressure adjustment value, i.e., the recommended increase or decrease in grouting pressure by the system at the current moment. P g (t) >0 indicates that the pressure should be increased, Δ P g (t) <0 indicates that the pressure should be reduced; α and β are adjustment weight coefficients, which can be set according to engineering experience; α is the state weight coefficient, which controls the influence of "filling state" on pressure regulation, and is generally taken as 0.3~0.7 according to engineering experience; β is the rate weight coefficient, which controls the influence of "water pressure rising trend" on regulation, and is generally taken as 0.3~0.7 according to engineering experience.

[0126] The formula for calculating the grouting flow rate adjustment value is as follows:

[0127] (19)

[0128] Wherein, ΔQ(t) is the grouting flow rate adjustment value, that is, the adjustment that should be made to the grouting flow rate at present. ΔQ(t)>0 indicates that the flow rate should be increased, and ΔQ(t)<0 indicates that the flow rate should be decreased; γ and δ are adjustment weight coefficients, which can be set according to engineering experience; γ is the state weight coefficient, which is the degree of influence of the filling state on the grouting rate. It is generally taken as 0.3~0.7 according to engineering experience; δ is the rate weight coefficient, which is the influence of the water pressure rise rate on the grouting rate. It is generally taken as 0.3~0.7 according to engineering experience.

[0129] This embodiment implements a linear fusion control strategy through formulas (18) and (19). If the current state is not yet saturated (S is small) and the water pressure rises slowly (R... v If the pressure is less than 1, the system should actively increase the pressure or flow rate; if the state is close to saturation (S is large), or the water pressure rises too quickly (R... v If the pressure is >1), the system should reduce or slow down the grouting. The linear fusion control strategy considers both the current "grouting filling status" and "water pressure dynamic response" simultaneously, and standardizes them into [0,1] indicators, combining them with weights for decision-making.

[0130] Step S4: Execute control according to the grouting parameter control value and enter the feedback closed loop.

[0131] Step S4-1: Execute control and enter the feedback loop. The system controller will adjust ΔP. g ΔP(t) and ΔQ(t) are converted into control commands, which are applied to the execution equipment (i.e., the grouting equipment) to adjust the actual parameters. g The system issues an execution instruction based on the sign of ΔQ(t) and ΔQ(t).

[0132] If ΔP g (t)>0: Increase grouting pressure;

[0133] If ΔP g (t)<0: Reduce pressure or pause;

[0134] If ΔQ(t) > 0: Increase the grouting rate;

[0135] If ΔQ(t) < 0: slow down grouting or switch to pressure holding mode.

[0136] Control commands can be applied to actuators such as frequency converter pumps, electric valves, and grouting pump controllers.

[0137] Step S4-2, Feedback Loop and Model Update. The status after execution will be fed back to the system through the next round of data collection and compared with the expected results:

[0138] If the deviation exceeds the set threshold, the model will be automatically fine-tuned or retrained; if an anomaly occurs (such as a sudden increase in water pressure), a safety protection mode will be entered. The system runs continuously in increments of seconds, achieving closed-loop control throughout the entire process. This embodiment constructs a closed-loop feedback control system for the grouting process, which has self-updating capabilities.

[0139] Understandably, the "safety protection mode" is an emergency control state that the system automatically switches to when it detects abnormal behavior or sudden changes in state, ensuring the stability and safety of grouting operations. This mode consists of the following triggering logic, response actions, and recovery mechanisms:

[0140] I. Triggering conditions:

[0141] If any of the following abnormal indicators are met during continuous system operation, the security protection mode will be triggered:

[0142] 1. Sudden increase in water pressure: The rate of change of water pressure per unit time, dP / dt, exceeds the upper limit threshold εmax, for example, greater than 0.5 kPa / s;

[0143] 2. High-amplitude fluctuations: The water pressure fluctuation amplitude σ(t) exceeds the safety threshold;

[0144] 3. The model predicts an abnormal state: such as S(t) > 3.8 and continues to exceed the set period;

[0145] 4. Accumulated systematic error: The difference between the model prediction and the actual behavior continues to accumulate and exceed the threshold (such as the root mean square error continuously exceeding the standard, as shown in equation (20)).

[0146] (20)

[0147] Where RMSE is the root mean square error, y i It is the actual value (such as actual water pressure, actual state). It is the model prediction value (such as the predicted state S(t)), and n is the number of sample points;

[0148] 5. Abnormal feedback from the actuator: such as alarms triggered by signals like grouting pump backflow or valve overload.

[0149] II. Response Actions:

[0150] Once the system enters secure protection mode, the following operations will be performed immediately:

[0151] 1. Automatic grouting pause: Sends a pump stop / valve shut-off command to the execution unit to prevent further grout injection;

[0152] 2. Pressure maintenance and stabilization: Maintain the current water pressure to prevent negative pressure suction;

[0153] 3. Record abnormal data: Automatically label the current time window as a basis for subsequent model review and learning or on-site investigation;

[0154] 4. Alarm notification: Sends graphic / voice / text alarms to the monitoring terminal to prompt construction personnel to intervene;

[0155] 5. System lock and wait for manual confirmation: The system will not automatically resume operation before the safety flag is removed to prevent accidental startup.

[0156] III. Exit Mechanism:

[0157] The system can automatically or manually exit protection mode under the following conditions:

[0158] 1. Monitoring data continuously returns to normal and stabilizes (e.g., dP / dt recovers to less than 0.05 kPa / s and fluctuations decrease);

[0159] 2. The operator confirms on-site that the "fault is controllable" and manually cancels the warning;

[0160] 3. The system restarts and reloads the parameters.

[0161] Example 2

[0162] In one or more embodiments, a fracture grouting adaptive control system based on dynamic water pressure feedback is disclosed, specifically including:

[0163] The data acquisition module is configured to: acquire grouting area data in real time and process it to obtain initial features;

[0164] The feature extraction module is configured to: obtain a predicted value of the fracture state based on the initial features using a trained dynamic water pressure response model; the dynamic water pressure response model is constructed by fusing fuzzy rules and a neural network; the fuzzy rules perform fuzzy determination on the state of fracture grouting to obtain a state membership vector; the initial features and the state membership vector are fused and output as a predicted value of the fracture state through a neural network; simultaneously, the real-time water pressure change rate is obtained from the initial features, and the real-time water pressure normalization rate is obtained using the constructed water pressure trend normalization model;

[0165] The parameter calculation module is configured to calculate the grouting pressure adjustment value and the flow dynamic adjustment value based on the predicted fracture state value and the real-time normalization rate of water pressure.

[0166] The control execution module is configured to perform control based on the grouting pressure adjustment value and the flow dynamic adjustment value and enter the feedback closed loop.

[0167] Example 3

[0168] This embodiment provides an electronic device, including a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When the computer instructions are executed by the processor, they complete the steps of the above-mentioned adaptive control method for fracture grouting based on dynamic water pressure feedback.

[0169] Example 4

[0170] This embodiment provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the steps of the above-described adaptive control method for fracture grouting based on dynamic water pressure feedback.

[0171] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0172] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0173] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment, whereby a series of operational steps are performed to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0174] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0175] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for adaptive control of fracture grouting based on dynamic water pressure feedback, characterized in that, include: Real-time acquisition and processing of grouting area data yields initial features; Based on the initial features, a trained dynamic water pressure response model is used to obtain the predicted value of the fracture state. The dynamic water pressure response model is constructed by fusing fuzzy rules and neural networks. The fuzzy rules perform fuzzy determination on the state of fracture grouting to obtain a state membership vector. The initial features and the state membership vector are fused and output as the predicted value of the fracture state through a neural network. At the same time, the real-time water pressure change rate is obtained from the initial features, and the real-time water pressure normalization rate is obtained by using the constructed water pressure trend normalization model. The grouting pressure adjustment value and flow rate dynamic adjustment value are calculated based on the predicted value of the fracture state and the real-time normalized rate of water pressure. The control is executed based on the grouting pressure adjustment value and the flow rate dynamic adjustment value, and then enters the feedback closed loop; The dynamic water pressure response model includes a fuzzy rule layer, a neural network prediction layer, and a physical constraint correction layer connected in sequence. The fuzzy rule layer divides the state of crack grouting into multiple levels to obtain a state membership vector. Specifically, it calculates the water pressure change rate based on the water pressure time series P(t) and calculates the membership values ​​of unfilled state, partially filled state, near-saturated state, fully saturated state, and abnormal state in segments based on the water pressure change rate. The initial features and state membership vector are input into the neural network prediction layer to obtain the crack state prediction value; The predicted fracture state is adjusted using a physical constraint correction layer. The water pressure trend normalization model determines the rate of change of current water pressure and quantifies it into a normalization index: Calculate the reference rate of rise: Among them, v ref For reference rate of rise, P 目标 For the target water pressure, P 起始 T is the initial water pressure. 估算 This is the estimated time required for grouting to reach the target water pressure; Calculate the real-time normalized rate based on the reference rise rate: Among them, R v v(t) is the real-time normalized rate of water pressure; v(t) is the real-time rate of change of water pressure.

2. The adaptive control method for fracture grouting based on dynamic water pressure feedback as described in claim 1, characterized in that, The initial features are a feature vector including the rate of change of water pressure, the amplitude of water pressure fluctuation, the slope of water pressure trend, grouting pressure, grouting flow rate, and cumulative grouting time.

3. The adaptive control method for fracture grouting based on dynamic water pressure feedback as described in claim 1, characterized in that, The grouting pressure adjustment value and flow rate dynamic adjustment value are calculated based on the predicted value of the fracture state and the real-time normalized rate of water pressure. The grouting pressure adjustment value is: Where, Δ P g (t) Here, α is the grouting pressure adjustment value, β is the state weighting coefficient, and R is the rate weighting coefficient. v (t) represents the real-time normalization rate of water pressure, and S(t) represents the predicted value of fracture state. The dynamic flow adjustment value is: Where ΔQ(t) is the grouting flow rate adjustment value, γ is the state weighting coefficient, and δ is the rate weighting coefficient.

4. The adaptive control method for fracture grouting based on dynamic water pressure feedback as described in claim 1, characterized in that, The adjustment is performed according to the grouting parameter adjustment value, specifically as follows: According to the grouting pressure adjustment value ΔP g The positive and negative values ​​of the grouting flow rate adjustment value ΔQ(t) are used to implement control. If ΔP g If (t)>0, increase the grouting pressure; If ΔP g If (t) < 0, reduce the pressure or pause; If ΔQ(t) > 0, then increase the grouting rate; If ΔQ(t) < 0, then slow down the grouting or switch to the pressure holding mode.

5. An adaptive control system for fracture grouting based on dynamic water pressure feedback, characterized in that, include: The data acquisition module is configured to: acquire grouting area data in real time and process it to obtain initial features; The feature extraction module is configured to: obtain a predicted value of the fracture state based on the initial features using a trained dynamic water pressure response model; the dynamic water pressure response model is constructed by fusing fuzzy rules and a neural network; the fuzzy rules perform fuzzy determination on the state of fracture grouting to obtain a state membership vector; the initial features and the state membership vector are fused and output as a predicted value of the fracture state through a neural network; simultaneously, the real-time water pressure change rate is obtained from the initial features, and the real-time water pressure normalization rate is obtained using the constructed water pressure trend normalization model; The parameter calculation module is configured to calculate the grouting pressure adjustment value and the flow dynamic adjustment value based on the predicted fracture state value and the real-time normalization rate of water pressure. The control execution module is configured to: perform control based on the grouting pressure adjustment value and the flow dynamic adjustment value and enter the feedback closed loop; The dynamic water pressure response model includes a fuzzy rule layer, a neural network prediction layer, and a physical constraint correction layer connected in sequence. The fuzzy rule layer divides the state of crack grouting into multiple levels to obtain a state membership vector. Specifically, it calculates the water pressure change rate based on the water pressure time series P(t) and calculates the membership values ​​of unfilled state, partially filled state, near-saturated state, fully saturated state, and abnormal state in segments based on the water pressure change rate. The initial features and state membership vector are input into the neural network prediction layer to obtain the crack state prediction value; The predicted fracture state is adjusted using a physical constraint correction layer. The water pressure trend normalization model determines the rate of change of current water pressure and quantifies it into a normalization index: Calculate the reference rate of rise: Among them, v ref For reference rate of rise, P 目标 For the target water pressure, P 起始 T is the initial water pressure. 估算 This is the estimated time required for grouting to reach the target water pressure; Calculate the real-time normalized rate based on the reference rise rate: Among them, R v v(t) is the real-time normalized rate of water pressure; v(t) is the real-time rate of change of water pressure.

6. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the adaptive control method for fracture grouting based on dynamic water pressure feedback as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the adaptive control method for fracture grouting based on dynamic water pressure feedback as described in any one of claims 1-4.

Images