Method and system for predicting service performance of cracked lining of complex mountain tunnel

Abstract

The present application relates to the field of tunnel crack identification, and relates to a complex mountainous tunnel crack lining service performance prediction method and system, the method comprising obtaining complex mountainous tunnel crack lining multi-factor model test data; according to the multi-factor model test data, a multi-objective optimization model integrating asymmetric risk and physical monotonicity constraints is constructed to obtain an optimal recommended operation threshold set; according to the optimal recommended operation threshold set, operation period monitoring and layout processing is carried out to obtain real-time updated field monitoring data; according to the real-time updated field monitoring data, a prediction model is constructed to obtain crack lining current service state prediction information; according to the optimal recommended operation threshold set and the crack lining current service state prediction information, service performance evolution deduction processing is carried out to obtain an optimized prediction result, and the present application realizes accurate prediction and scientific operation decision of complex mountainous tunnel crack lining service performance.

Description

Technical Field

[0001] This invention relates to the field of tunnel crack identification, and more specifically, to a method and system for predicting the service performance of cracked linings in tunnels located in complex mountainous areas. Background Technology

[0002] As a crucial component of transportation infrastructure, tunnels in complex mountainous areas are subject to the combined effects of high ground stress, freeze-thaw erosion, and high ground temperature during long-term service. This makes their lining structures prone to cracking and other defects, and the evolution of their service performance is complex. Accurate prediction and safe management are core challenges in ensuring tunnel operation safety. Current technologies typically employ indoor model tests to obtain the correlation between crack characteristics and structural performance, and then assess the service status by comparing crack parameters detected on-site with this correlation. However, these methods suffer from limitations such as a single assessment dimension, insufficient consideration of the combined effects of multiple factors, inadequate characterization of crack parameters, lack of environmental adaptability, unidirectional data flow making it difficult to iteratively update, fixed monitoring points leading to missed critical damage, and insufficient reliability of prediction models due to a lack of physical constraints. These limitations make it difficult to meet the practical needs of accurate prediction and scientific operation and maintenance of cracked linings in complex mountainous environments. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for predicting the service performance of cracked tunnel linings in complex mountainous areas, so as to improve the above-mentioned problems.

[0004] To achieve the above objectives, the embodiments of this application provide the following technical solutions:

[0005] On the one hand, embodiments of this application provide a method for predicting the service performance of cracked linings in tunnels located in complex mountainous areas, the method comprising:

[0006] Obtain multi-factor model test data for cracked tunnel lining in complex mountainous areas;

[0007] Based on the multi-factor model test data, a multi-objective optimization model integrating asymmetric risk and physical monotonicity constraints is constructed to obtain the optimal recommended operation and maintenance threshold set;

[0008] Based on the optimal recommended operation and maintenance threshold set, the monitoring deployment process is carried out during the operation period. The uncertainty of cognition is quantified and the location of the measuring points is optimized through an active learning mechanism to obtain real-time updated on-site monitoring data.

[0009] A prediction model is constructed based on the real-time updated field monitoring data to obtain the predicted information of the current service status of the crack lining;

[0010] Based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, the service performance evolution is extrapolated to obtain the optimized prediction results.

[0011] Secondly, embodiments of this application provide a service performance prediction system for cracked linings in tunnels located in complex mountainous areas, the system comprising:

[0012] The acquisition module is used to acquire multi-factor model test data of cracked lining in complex mountain tunnels;

[0013] The first processing module is used to construct a multi-objective optimization model that integrates asymmetric risk and physical monotonicity constraints based on the multi-factor model test data, and obtain the optimal recommended operation and maintenance threshold set.

[0014] The second processing module is used to perform operational monitoring deployment processing based on the optimal recommended operation and maintenance threshold set, and to quantify cognitive uncertainty and optimize the location of measuring points through an active learning mechanism to obtain real-time updated on-site monitoring data.

[0015] The third processing module is used to construct a prediction model based on the real-time updated field monitoring data to obtain the prediction information of the current service status of the crack lining.

[0016] The fourth processing module is used to perform service performance evolution extrapolation based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, and obtain the optimized prediction results.

[0017] Thirdly, embodiments of this application provide an apparatus, which includes a memory and a processor. The memory stores a computer program; the processor executes the computer program to implement the steps of the above-described method for predicting the service performance of tunnel linings with cracks in complex mountainous areas.

[0018] Fourthly, embodiments of this application provide a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for predicting the service performance of cracked linings in tunnels in complex mountainous areas.

[0019] The beneficial effects of this invention are as follows:

[0020] This invention acquires multi-factor model test data of cracked lining tunnels in complex mountainous areas, constructs a multi-objective optimization model integrating asymmetric risk and physical monotonicity constraints to obtain an optimal recommended operation and maintenance threshold set, conducts operational monitoring deployment based on this threshold set, quantifies cognitive uncertainty and optimizes the location of measuring points using an active learning mechanism to obtain real-time updated field monitoring data, and then uses this field monitoring data to construct a prediction model to obtain the current service status prediction information of the cracked lining. Finally, the optimal recommended operation and maintenance threshold set and the current service status prediction information are combined to extrapolate the service performance evolution and obtain the prediction results of future service performance. This invention effectively solves the problems of single evaluation dimensions, insufficient consideration of multi-factor coupling effects, fixed monitoring points that easily miss key damage, insufficient reliability of prediction models due to lack of physical mechanism constraints, and difficulty in iterative updates due to unidirectional data flow in existing technologies, and achieves accurate prediction of service performance and scientific operation and maintenance decision-making for cracked lining in complex mountainous areas.

[0021] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the service performance prediction method for cracked tunnel lining in complex mountainous areas, as described in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the service performance prediction device for cracked lining of tunnels in complex mountainous areas, as described in an embodiment of the present invention.

[0025] Figure 3 It serves as a multi-factor coupled experimental platform.

[0026] Figure 4 This is a side view of the multi-factor coupling test platform.

[0027] The diagram is labeled as follows: 800, Service performance prediction equipment for tunnels with cracked linings in complex mountainous areas; 801, Processor; 802, Memory; 803, Multimedia component; 804, I / O interface; 805, Communication component. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0029] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] Example 1:

[0031] This embodiment provides a method for predicting the service performance of cracked linings in tunnels located in complex mountainous areas. It is understood that this embodiment can be based on a scenario, such as a highway tunnel traversing an active fault zone in a high-altitude, complex mountainous region showing signs of lining crack development after many years of operation. This area is characterized by high mountains and deep valleys, large diurnal temperature variations, and surrounding rock rich in groundwater. The long-term coupling effect of high ground stress, freeze-thaw cycles, and seepage erosion makes the crack morphology and propagation patterns extremely complex. Existing technologies use traditional static monitoring and experience-based bearing capacity comparison tables for evaluation. However, due to the lack of quantitative analysis of the coupling effect of high ground temperature and freeze-thaw cycles, the model cannot reproduce the abrupt stress concentration at the crack tip, leading to repeated omissions in the accelerated deterioration trend of the cracks.

[0032] See Figure 1 The figure shows that the method includes steps S1-S5.

[0033] Step S1: Obtain multi-factor model test data for cracked lining of tunnels in complex mountainous areas;

[0034] Step S1 further includes steps S11-S14, which specifically include:

[0035] Step S11: Based on the characteristics of complex mountainous environments, conduct multi-factor coupled experimental design to obtain a set of experimental conditions;

[0036] To address the multi-field coupled environment characteristics of complex mountain tunnels, including high ground stress, freeze-thaw erosion, high ground temperature, and seepage, the initial crack conditions (including crack location, length, width, and morphology) are the main factor. Combined with other influencing factors such as surrounding rock type, ambient temperature, ground stress level, and number of freeze-thaw cycles, orthogonal experiments or uniform design methods are used to arrange test conditions. Orthogonal experiments cover the interaction of multiple factors with fewer experiments by reasonably allocating the level combinations of each factor. Uniform design emphasizes the uniform distribution of factor levels to further reduce the number of experiments. The final set of test conditions needs to cover the main service scenarios of complex mountain tunnels, while reasonably controlling the number of test groups to balance test costs and data representativeness.

[0037] Step S12: Perform physical model test execution processing according to the set of test conditions to obtain structural response data under multiple conditions;

[0038] In this step, a multi-factor coupled test platform is first built, including an extreme temperature environment and seepage simulation system, a ground stress loading system, a lining structure model and a crack prefabrication module. Specifically, a tunnel lining model was designed based on geometric similarity, and poured using gypsum or similar concrete materials. Cracks of different locations, lengths, widths, and shapes were pre-fabricated in the model by embedding removable thin steel or plastic sheets. Surrounding rock similar materials were prepared and filled using cement, gypsum, and sand in proportions according to actual rock physical and mechanical properties. Subsequently, the lining model and surrounding rock similar materials were loaded into the three-dimensional loading frame of the geostress loading system. Multi-physics field monitoring sensors were deployed at key sections of the lining and inside the surrounding rock. The geostress loading system was activated, applying vertical and horizontal loads to simulate high geostress using servo-controlled jacks. Simultaneously, an extreme temperature environment and seepage simulation system was activated, adjusting the temperature of the environmental chamber to simulate high ground temperature and freeze-thaw cycles, and applying controllable water pressure through the bottom seepage pipe to simulate groundwater seepage. After the temperature field, stress field, and seepage field reached a relatively stable state, the mechanical response characteristics of the lining under various working conditions were recorded, obtaining structural response data under multiple working conditions. A multi-factor coupled test platform was used, such as... Figure 3 and Figure 4 As shown.

[0039] Step S13: Perform multi-physics monitoring processing on the structural response data under the multiple working conditions to obtain the original monitoring dataset;

[0040] During the experiment, multi-dimensional data were collected synchronously through multi-physics field monitoring. Specifically, resistance strain gauges were attached to key locations on the inner surface of the lining, such as the arch crown, arch waist, arch foot, and crack tips, to monitor the stress-strain distribution of the lining; earth pressure cells were buried at the contact surface between the surrounding rock and the lining to monitor the surrounding rock pressure and the frost heave pressure generated by freeze-thaw cycles; vibrating wire crack gauges were deployed in existing and potential crack areas, and distributed optical fibers were laid along the circumferential and longitudinal directions of the key cross-section of the lining to obtain crack width, length, and propagation rate; thermometer strings were arranged along the axial and radial directions inside the surrounding rock to monitor the evolution of the temperature field; and piezometers were deployed on the lining surface to monitor seepage pressure. All sensors synchronously collected and stored data through a data acquisition system, forming a raw monitoring dataset containing crack parameters, environmental variables, and structural response parameters under various working conditions.

[0041] Step S14: Preprocess the original monitoring dataset to obtain the multi-factor model test data.

[0042] In this step, preprocessing includes noise reduction, normalization, and standardization.

[0043] Step S2: Based on the multi-factor model test data, construct a multi-objective optimization model that integrates asymmetric risk and physical monotonicity constraints to obtain the optimal recommended operation and maintenance threshold set;

[0044] Step S2 further includes steps S21-S24, which specifically include:

[0045] Step S21: Perform physical feature screening processing on the multi-factor model test data to obtain the screened feature set;

[0046] Step S21 further includes steps S211-S214, which specifically include:

[0047] Step S211: Perform multi-field energy gradient calculation processing based on the multi-factor model experimental data to obtain the multi-field energy gradient coefficients of each feature;

[0048] In this step, the multi-field energy gradient coefficients are specifically as follows:

[0049] ;

[0050] In the formula: express Multiple energy gradients; This represents the k-th feature in the feature set; This represents the mechanical strain energy density of the lining structure. This represents the thermal strain energy density of the lining structure. This represents the energy density of the seepage field around the lining.

[0051] Step S212: Perform physical enhancement feature difference function reconstruction based on the multi-field energy gradient coefficients of each feature to obtain the feature difference metric;

[0052] In this step, the physical enhancement feature difference function is specifically as follows:

[0053] ;

[0054] In the formula, This represents the physical enhancement feature difference function, used to quantify any two samples. and In the Differences in individual characteristics; express Multiple energy gradients; , They represent the first and second digits of the entire sample set, respectively. The maximum and minimum values ​​of each feature; , Representing samples respectively and In the The values ​​that can be taken on each feature.

[0055] Step S213: Perform feature weight update processing based on the feature difference metric to obtain the influence weight value of each feature;

[0056] Randomly select target samples from the sample set. Find those similar to it. nearest neighbor samples ( ) and different types The nearest neighbor samples (denoted as ) ).feature weight The update formula is refactored as follows:

[0057] ;

[0058] In the formula: Indicates the updated number Each feature weight; Indicates the number before the update Each feature weight; This represents the difference function of physical enhancement features; This represents a randomly selected target sample. Indicates the target sample similar The number of nearest neighbor samples; This represents the total number of sampling iterations of the algorithm; Indicates the target sample The first of its kind The nearest neighbor samples; Indicates the target sample Different types of the first The nearest neighbor samples; Indicates sample The category or tag it belongs to; This represents the prior probability of category C; Indicates the target sample Different types The number of nearest neighbor samples.

[0059] Step S214: Perform feature filtering based on the influence weight values ​​of each feature to obtain the filtered feature set.

[0060] In this step, the features are sorted in descending order according to their weights, a physical contribution threshold is set, and secondary features and environmental noise variables with weights less than the physical contribution threshold are removed. Finally, a subset of high-weight master features is output, resulting in the filtered feature set.

[0061] Step S22: Construct a LightGBM regression model based on the selected feature set. By introducing an asymmetric risk loss function, apply a high-weight penalty to the error of overestimating structural safety to obtain the service performance mapping model.

[0062] In this step, the general-purpose LightGBM uses Mean Squared Error (MSE) to penalize both positive and negative deviations equally. However, in complex mountain tunnels, overestimating the remaining bearing capacity can lead to catastrophic collapse, while underestimating it only increases maintenance costs. Therefore, this invention reconstructs the loss function of the LightGBM, introduces an asymmetric risk penalty term for tunnel safety, and constructs a service performance mapping model.

[0063] In the t-th iteration, the model update formula is:

[0064] ;

[0065] In the formula: The prediction function is the ensemble function for the first t-1 rounds; The learning rate of the model; Let L be a single regression tree model trained in round t. To fit the asymmetry of security risk, the objective loss function L is defined as:

[0066] ;

[0067] In the formula: This represents the total number of training samples; Indicates the first The actual service performance index values ​​of each sample; Indicates the first The model prediction value for each sample; This represents a penalty factor for overestimating security risks (i.e., the predicted value is higher than the actual value); This represents the penalty factor for underestimating safety risks (i.e., the actual value is higher than the predicted value), and satisfies... ; For regularization terms, the specific regularization terms are:

[0068] ;

[0069] In the formula: Let t be the leaf node tree of the t-th tree; For the first The weight of each leaf node; and This is the regularization parameter.

[0070] The final prediction output is a weighted sum of all tree predictions:

[0071] ;

[0072] In the formula: This represents the final predicted value of the model, namely the predicted service performance index of the cracked lining of the tunnel in the complex mountainous area. Indicates the initial predicted value; This indicates the total number of regression trees; This refers to a single regression tree model trained in round t. This represents the learning rate of the model.

[0073] Step S23: Based on the service performance mapping model, perform multi-field coupling objective function construction processing to obtain the comprehensive objective function;

[0074] To accurately reflect the nonlinear degradation of lining performance under the coupled effects of temperature, seepage, and stress in complex mountain tunnels, this step constructs a comprehensive objective function based on a multi-field coupled degradation mechanism, specifically including:

[0075] ;

[0076] In the formula: Represent the overall objective function; , , , These represent the coupled structure's safety bearing capacity index, coupled crack control index, structural safety reserve index, and durability index, respectively. The coupled structure's safety bearing capacity index considers the nonlinear weakening effect of freeze-thaw damage and crack activity on the bearing capacity, and is reconstructed by introducing a cross-index penalty term. Its specific form is as follows:

[0077] ;

[0078] In the formula: This represents the remaining bearing capacity. This is the initial bearing capacity; The degree of freeze-thaw damage; This refers to the crack propagation rate; To allow for expansion rate; This is the sensitivity coefficient for load-bearing capacity coupling degradation. When freeze-thaw damage... With crack propagation At the same time, when the level is high, the function can force a precipitous drop in the bearing capacity of the lining, guiding the algorithm to avoid the risk zone of coupled disaster.

[0079] The coupled crack control index considers the frost heave and splitting effect caused by water entering the lining through microcracks and freezing and expanding. The crack propagation rate is nonlinearly amplified using freeze-thaw damage, specifically in the following form:

[0080] ;

[0081] In the formula: This refers to the crack propagation rate; To allow for expansion rate; This is the frost heave expansion factor; The degree of freeze-thaw damage; For nonlinear coupling exponent (typically, This indicator reflects the catalytic effect of environmental degradation on the development of structural defects.

[0082] The structural safety reserve index takes into account the spatial weakening of the overall structural stability caused by the geometric size of macroscopic cracks, and geometrically reduces the theoretical safety factor. Specifically, it takes the following form:

[0083] ;

[0084] In the formula: This represents the current theoretical safety factor. The target safety factor; These represent the current crack width and length, respectively. This corresponds to the maximum allowable value; This is the spatial damage reduction factor.

[0085] The durability index takes into account the mechanism that the opening of micro-fractures under high ground stress accelerates freeze-thaw erosion, and introduces the ground stress ratio for weighted penalty, specifically in the form of:

[0086] ;

[0087] In the formula: The degree of freeze-thaw damage; To allow for the maximum freeze-thaw damage; The maximum compressive stress at the critical section of the lining; It refers to the compressive strength of concrete; This is the stress-freeze-thaw coupling penalty coefficient.

[0088] Step S24: Process the comprehensive objective function using the non-dominated sorting whale optimization algorithm to obtain the optimal recommended operation and maintenance threshold set.

[0089] By integrating the objective function, the NSDWOA algorithm strictly follows the multi-field coupling disaster mechanism of geotechnical underground engineering when calculating fitness in multi-dimensional space, thus eliminating pseudo-optimal solutions that meet individual indicators but face systemic failure as a whole.

[0090] By taking the main control parameters affecting structural service life and maintenance decisions as decision variables to be optimized, a threshold solution vector is constructed:

[0091] ;

[0092] In the formula: Represent decision variables; Indicates the length of the crack; Indicates the crack width; Indicates the crack propagation rate; Indicates the degree of freeze-thaw damage.

[0093] To ensure that the multi-objective optimization algorithm searches within the real physical space of complex mountain tunnel engineering, the following hard multi-field boundary constraints are set based on current tunnel design specifications and material mechanics constitutive models.

[0094] Crack length: ,in , These represent the minimum perceptible crack length and the maximum limit control length allowed by operational specifications; crack width: ,in Maximum allowable crack width; Remaining bearing capacity: , This is the minimum remaining bearing capacity. The initial theoretical bearing capacity under undamaged conditions; safety factor: , For the minimum permissible safety factor, Design target safety factor; crack propagation rate: To ensure that crack evolution within the optimization interval is in a convergent or controllable state; freeze-thaw damage degree: Maximum compressive stress at key sections: ,in For concrete compressive strength, To account for the material strength reduction factor under long-term high ground stress, and to ensure that the structure does not suffer sudden crushing failure.

[0095] NSDWOA incorporates a non-dominated sorting mechanism into the whale optimization algorithm. During initialization, the algorithm generates a population of size P, with each whale representing a set of decision variables. Because the material damage and crack propagation in the lining of tunnels with cracks in complex mountainous areas exhibit irreversible deterioration characteristics under long-term high ground stress and freeze-thaw environments, the free-jump search of traditional optimization algorithms is prone to producing pseudo-safety material understandings that violate the laws of rock mass mechanics evolution. Therefore, in each iteration, this invention, in addition to the three conventional update mechanisms, introduces an additional geological evolution monotonicity constraint vector M.

[0096] ;

[0097] In the formula: This is the vector constraining the monotonicity of geological evolution. Representing vectors The Each element.

[0098] First, individual whales calculate an initial updated position set through prey defense mechanisms, bubble-web attack mechanisms, and random search mechanisms. After obtaining the initial updated position set, a geological evolution monotonicity constraint vector is applied to obtain the final updated position, specifically:

[0099] ;

[0100] In the formula: Indicates the final update position; This indicates an initial update to the location set; This represents the Hadamard product. For decision variables involving time accumulation or crack evolution... If the calculated new position That is, a reverse shrinkage phenomenon that violates the laws of physical degradation occurs on the time axis, then the corresponding constraint element Forcefully reject the mathematical update for that dimension; otherwise... This approach strictly constrains the purely mathematical optimization path within the tunnel geotechnical deterioration criterion. This indicates the initial update of the position set. Each element.

[0101] After each iteration, the algorithm calculates the fitness of individuals in the population based on the aforementioned multi-field coupled objective function group, uses a fast non-dominated sorting mechanism to divide them into multiple Pareto front layers, and calculates the crowding distance to maintain the diversity of the solution set in a highly nonlinear, strongly coupled space. Through multiple generations of iterative iterations, a converged Pareto optimal operational threshold solution set is finally output. Each optimal solution in this set is not only strictly controlled by the monotonic physical boundary of geological evolution in its optimization path, but also avoids pseudo-safe physical states where individual indicators appear to be within the safe threshold, but there is a risk of disaster under the superposition of multiple coupled factors due to the nonlinear cross-penalty mechanism of the objective function. The various combinations of crack parameters and environmental factors in the solution set represent the optimal engineering equilibrium strategy with different weight preferences under the multi-field strongly coupled mechanism of "bearing capacity weakening - frost heave expansion - spatial vulnerability - stress concentration", which can provide a scientific decision-making basis for predicting the service performance of tunnels in complex mountainous areas throughout their entire life cycle.

[0102] Furthermore, in conjunction with current tunnel design and maintenance specifications, the boundary conditions in the Pareto solution set are rigorously examined: all out-of-bounds solutions are eliminated if the remaining bearing capacity is below the minimum limit, the crack propagation rate exceeds the allowable value, the safety factor is below the minimum allowable value, the freeze-thaw damage exceeds the allowable extreme value, or the maximum principal stress of local key sections exceeds the limit. Subsequently, considering the economic efficiency of reinforcement and the feasibility of emergency construction operations throughout the entire life cycle of tunnels in complex mountainous areas, customized recommended operation and maintenance thresholds, such as appropriate crack width warning values, crack length control values, and freeze-thaw cycle limits, are extracted from the optimal solution set selected through rigorous multi-field constraint screening. These thresholds will be directly used to guide on-site operation and maintenance decisions during the operational period, and relevant model experimental data and optimization results will be uniformly compiled into a benchmark database.

[0103] This embodiment uses physical enhancement feature screening based on multi-field energy gradients to accurately eliminate environmental noise and identify cracks and environmental factors that play a dominant role in structural safety, thus avoiding model distortion caused by data redundancy from the source. On this basis, the LightGBM model with an asymmetric risk loss function is introduced to impose a huge penalty on fatal errors that overestimate structural safety, establishing an assessment baseline that conforms to the conservative principle of civil engineering disaster prevention. Finally, by incorporating the monotonicity constraint of geological evolution, the improved NSDWOA multi-objective optimization rigidly blocks the mathematical optimization path that violates the common sense of geotechnical mechanics, ensuring that under extreme conditions of strong multi-field coupling, such as freeze-thaw-soil stress co-deterioration, the output Pareto maintenance threshold solution set has both mathematical optimality and solid physical and mechanical confidence.

[0104] Step S3: Perform operational monitoring deployment processing based on the optimal recommended operation and maintenance threshold set, quantify cognitive uncertainty and optimize the location of measurement points through an active learning mechanism, and obtain real-time updated on-site monitoring data;

[0105] Step S3 further includes steps S31-S34, which specifically include:

[0106] Step S31: Perform initial monitoring system deployment processing based on the optimal recommended operation and maintenance threshold set to obtain the initial field monitoring database;

[0107] The following monitoring instruments were installed at key sections of the tunnel lining and near existing cracks:

[0108] (1) Temperature field monitoring: A string of temperature sensors is deployed inside the surrounding rock to monitor the temperature distribution characteristics of the surrounding rock and lining under high geothermal conditions, and to record the temperature evolution data during the freeze-thaw cycle. The sensors are deployed in a multi-point manner in both the horizontal and vertical directions, with the density increasing from the outer surrounding rock towards the lining, following the deployment method in the model test. The monitoring data correspond to the temperature field evolution law of the surrounding rock and lining in the multi-factor coupled test platform.

[0109] (2) Stress and strain monitoring: Vibrating wire strain gauges are installed on key locations such as the arch crown, arch waist, sidewalls, and the tips and vicinity of existing cracks on the inner and outer surfaces of the lining structure to monitor the stress and strain distribution and variation patterns of the lining during crack propagation and freeze-thaw cycles. The strain gauges must possess freeze-thaw resistance and long-term stability to operate stably in complex mountainous environments. The monitoring data correspond to the structural stress and strain development patterns in the multi-factor coupled test platform.

[0110] (3) Crack Development Monitoring: Vibrating wire crack gauges are used as the primary monitoring method at existing cracks and key locations where new cracks may form. The sensor is fixed at both ends to the sides of the crack to directly measure changes in crack opening and closing. Simultaneously, distributed optical fibers are deployed along the circumferential and longitudinal directions of the lining's key sections to achieve continuous, distributed monitoring of crack location and development range. Furthermore, resistance strain gauges can be selectively attached to stress concentration areas such as crack tips to monitor local strain conditions. These monitoring methods complement each other, and combined with periodic inspections and image recognition technology, key parameters such as crack width, length, and propagation rate are obtained. The monitoring data correspond to crack propagation characteristics in a multi-factor coupled test platform.

[0111] (4) Seepage monitoring: Piezometers and flow meters are installed at key locations on the lining surface to monitor changes in seepage pressure and flow rate. Piezometers are densely installed in sections where cracks develop and seepage may intensify. The piezometers are of vibrating wire type to meet the needs of long-term monitoring.

[0112] (5) Monitoring of surrounding rock pressure and frost heave pressure: Earth pressure cells are installed at the contact surface between the lining and the surrounding rock, near existing cracks, and in freeze-thaw sensitive areas to monitor the surrounding rock pressure and the frost heave pressure generated by soil freezing during freeze-thaw cycles. The monitoring items of the above monitoring instruments are matched one-to-one with those of the multi-factor coupled test platform. The initial monitoring frequency and the distribution of measuring points are set according to specifications and engineering experience. Preliminary data are collected and an initial field monitoring database is formed based on this data.

[0113] Step S32: Based on the initial field monitoring database, perform monitoring blind zone uncertainty quantification processing based on active learning to obtain the uncertainty distribution of each unknown measuring point in the whole space;

[0114] In this step, as the operating environment evolves nonlinearly, fixedly deployed sensors struggle to capture randomly emerging new defects. Using a trained and converged improved LightGBM regression model as a surrogate evaluator, an active learning mechanism is employed to quantify the cognitive uncertainty of each unknown measurement point within the tunnel's entire spatial domain. Specifically:

[0115] ;

[0116] In the formula: This represents the uncertainty of various unknown measurement points within the entire spatial domain. The larger the value, the less certain the existing model is about the security status of the area, and the higher the potential physical information gain contained at that point. Represents the three-dimensional spatial coordinate vector of the tunnel lining; Indicates the total number of random samples; express The subsampling model for spatial coordinates Predicted output values ​​of service performance indicators; express The arithmetic mean of the predicted outputs.

[0117] Step S33: Perform measurement point optimization and recommendation processing based on the uncertainty distribution to obtain the optimal coordinates of the newly added monitoring points;

[0118] Since conventional active learning involves a global search in three-dimensional space, the recommended highest uncertainty point is prone to falling into the deep rock mass of the tunnel, making it difficult to drill and install sensors during actual operation. Therefore, this step constructs a special spatial dimensionality reduction constraint mechanism to forcibly map and restrict the candidate space to the topological manifold formed by the accessible inner surface of the tunnel lining and the existing two-dimensional crack propagation surface. The optimization objective function is expressed as:

[0119] ;

[0120] In the formula: This represents the coordinates of the optimal new monitoring point recommended by the active learning algorithm. This refers to a two-dimensional internal surface topological manifold domain that is accessible to maintenance personnel without damaging the existing structure during the operation period. Indicates at this point in space The engineering construction cost or safety risk function for deploying sensors at a location; This represents the cost penalty weighting coefficient. This mechanism allows for the precise selection of monitoring points that provide the maximum security information gain and are fully feasible for implementation. Maintenance personnel dynamically add monitoring items based on the output coordinates of the optimal new monitoring points, and the collected high-value data is imported into the initial field monitoring database in real time.

[0121] Step S34: Obtain real-time updated field monitoring data based on the coordinates of the optimal newly added monitoring points.

[0122] Step S4: Construct a prediction model based on the real-time updated field monitoring data to obtain the prediction information of the current service status of the crack lining;

[0123] Step S4 further includes steps S41-S44, which specifically include:

[0124] Step S41: Based on the real-time updated field monitoring data, perform multi-field coupling control equation definition processing to obtain composite physical constraints;

[0125] In this step, the composite physical constraints specifically include:

[0126] (1) Residual of the heat conduction equation containing latent heat of phase change:

[0127] ;

[0128] In the formula: The residuals of the heat conduction equation containing the latent heat of phase change; Density of the surrounding rock; Specific heat capacity; For temperature, For time; For heat conduction, Thermal conductivity; For latent heat of phase transition, This refers to the unfrozen water content; For temperature gradient.

[0129] (2) Residuals of the elastic equilibrium equations for thermo-hydraulic-mechanical coupling:

[0130] ;

[0131] In the formula: The residuals of the elastic equilibrium equations representing the thermo-hydraulic-mechanical coupling; It is a divergence operator; The fourth-order elastic stiffness tensor represents the lining or rock mass medium. Representing a fourth-order tensor Condensation multiplication operation between the second-order elastic strain tensor; To represent the second-order total strain tensor of the medium; The coefficient of thermal expansion is The change in temperature This represents a second-order unit tensor, used to convert scalar thermal strain into a tensor form with isotropic volumetric expansion characteristics. Density of the surrounding rock; This is the acceleration due to gravity.

[0132] Step S42: Under the composite physical constraints, reconstruct and pre-train the physical information extreme learning machine network with embedded crack tip singularity residuals to obtain the initialized network weight parameters;

[0133] To address the discontinuities in cracked linings, this step, based on linear elastic fracture mechanics, embeds the singularity residuals at the crack tip into the total loss function of the network, thus reconstructing the loss function. The specific process is as follows:

[0134] ;

[0135] In the formula: Represents the loss function after reconstruction; The mean square error of the fit to the sensor's measured data; , , Penalty weighting coefficients for balancing gradients of multiple physical dimensions; The squared L2 norm of the collocation points of the continuum equation; This represents the singularity residual at the crack tip.

[0136] The singularity residual at the crack tip is specifically:

[0137] ;

[0138] In the formula: This represents the total number of spatially arranged points within a tiny neighborhood of the known crack tip. This represents the local polar coordinate system position of the j-th configuration point with the crack tip as the origin; This represents the stress tensor at that point in the network's predicted output; , These represent the pure Type I and pure Type II stress intensity factors, respectively; , This represents a triangular distribution function specific to fracture mechanics.

[0139] During the initialization of the Physical Information Extreme Learning Machine (PISM) network, the input layer weights and hidden layer biases are randomly generated and fixed. First, the PISM network is pre-trained using multi-factor model experimental data. The initial hidden layer output matrix and initial output weight matrix are then obtained by minimizing the reconstructed loss function.

[0140] Step S43: Perform domain difference quantization processing based on the initialized network weight parameters to obtain inter-domain difference index;

[0141] Existing technologies typically use Karl Fischer divergence (KL divergence) to quantify the difference between the source and target domains. However, in this scenario, the probability distributions of multi-factor model experimental data and complex mountainous field monitoring data often exhibit non-overlapping support sets, which can easily lead to KL divergence failure or gradient vanishing. Therefore, this step uses Wasserstein distance, based on physical energy dissipation cost, as the domain difference quantification and migration triggering indicator. The specific process includes:

[0142] ;

[0143] In the formula: This represents the optimal physical energy transmission distance obtained. This represents the distribution of experimental data from a multi-factor coupling test platform. Indicates the distribution of on-site monitoring data; and Let P and Q represent the service performance characteristic sample vectors, respectively. Let represent the set of all joint probability distributions satisfying marginal distributions P and Q (i.e., the set of all possible transmission schemes); This indicates that under a specific transmission scheme, from the sample Transfer to sample Quality weights; This represents all possible transmission schemes. Find the optimal transmission path with the lowest cost; This represents the physical transmission cost function.

[0144] Furthermore, considering the physical degradation laws of tunnels, this step reconstructs the physical transmission cost function into the absolute difference of thermal-mechanical-seepage multi-field strain energy dissipation that takes into account spatial vulnerability:

[0145] ;

[0146] In the formula: , , Representing samples respectively and The difference between them in mechanical strain energy density, thermal strain energy density and seepage field energy density; This is a spatial vulnerability weight matrix preset based on the mechanical properties of the lining. High weights, such as 0.8~0.9, are assigned to the arch crown, arch foot, and the front edge of existing crack propagation, while low weights, such as 0.1~0.2, are assigned to conventional sidewalls.

[0147] After calculating the optimal physical energy transmission distance, a distance threshold is set. A fine-tuning update of the model is triggered only when the physical energy state in the high-risk weighted region experiences a true drift, causing the optimal physical energy transmission distance to exceed the distance threshold. This avoids the mathematical divergence defects of the general domain adaptation algorithm when handling non-overlapping extreme operating conditions, while ensuring the physical rigor of the migration triggering mechanism.

[0148] Step S44: Perform migration fine-tuning triggering processing based on the inter-domain difference index to obtain the current service status prediction information of the crack lining.

[0149] In this step, after triggering transfer fine-tuning, the field monitoring database is imported into the pre-trained physical information extreme learning machine network. During this process, the network structure remains unchanged. The newly added monitoring data matrix and corresponding hidden state matrix from the field monitoring database are used to quickly update the network's output layer weights via the generalized inverse matrix. Specifically:

[0150] ;

[0151] In the formula, This indicates that the output layer weights of the network are updated quickly using the generalized inverse matrix; Represents the hidden state matrix; This represents the newly added monitoring data matrix. Throughout the fine-tuning process, the crack tip singularity residual and multiphysics residual in the reconstructed loss function are always used as penalty regularization terms to forcibly constrain the update direction of the output weights.

[0152] After fine-tuning, the Physical Information Extreme Learning Machine (PIM) network acquires a dedicated evaluation capability for the current tunnel evolution state. When the on-site monitoring system collects new multi-dimensional feature vectors of environmental and structural service, the network instantly calculates and outputs the final corrected prediction value—the predicted information of the current service state of the cracked lining—using the forward propagation formula. This prediction information accurately fits the variability of the on-site data and strictly follows the extreme evolution laws of geotechnical mechanics and fracture mechanics, providing a highly reliable computational benchmark for subsequent performance comparisons and closed-loop early warning.

[0153] This implementation resolves the fundamental contradiction between indoor model testing and extreme on-site conditions in complex mountainous areas: by embedding heat conduction equations containing latent heat of phase change, heat-water-mechanical coupling equilibrium equations, and stress singularity residuals at crack tips, the prediction model is forced to strictly follow the physical laws of multi-field coupling and the extreme characteristics of fracture mechanics. This fundamentally avoids the underestimation of failure risk caused by mathematical smoothing in high-risk areas such as crack tips by purely data-driven models. At the same time, it innovatively adopts PE-Wasserstein distance quantization based on physical energy dissipation costs to differentiate the domains between model testing and on-site monitoring. This breaks through the limitation of traditional statistical divergence in non-overlapping distribution scenarios, accurately capturing physical drift caused by high ground stress redistribution or freeze-thaw damage, and triggering an efficient migration fine-tuning mechanism so that the model can dynamically adapt to the real evolution state on-site without retraining. This significantly improves the physical confidence and engineering reliability of service performance prediction in complex environments.

[0154] Step S5: Based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, perform service performance evolution extrapolation to obtain the optimized prediction results.

[0155] As the tunnel's operational period progresses, the optimized monitoring sensor network deployed on a two-dimensional accessible topological manifold continuously feeds measured data into the on-site monitoring database. It calculates the optimal physical energy transmission distance for the current physical environment in real time. When the optimal physical energy transmission distance is less than or equal to a distance threshold, it indicates that no drastic physical drift has occurred in the current on-site conditions. The output weights of the current physical information limit learning machine network are then used to quickly output the structural service status indicators through the forward propagation formula. When the optimal physical energy transmission distance exceeds the distance threshold, a migration fine-tuning mechanism is triggered to update the data. This ensures that the model's predicted trends accurately reflect the actual service status evolution process at the engineering site.

[0156] Meanwhile, since the tunnel service environment is a long-term dynamically evolving system, the initially constructed LightGBM model, constrained by the static boundaries of multi-factor model test data, will experience decision drift in its Pareto solution set as the tunnel ages. Therefore, a cross-model co-evolution mechanism was established in this step: after the Physical Information Extreme Learning Machine (PISM) network completes cross-domain fine-tuning in the field, under the constraints of the current field environment, the PISM network performs rapid forward propagation on a large number of randomly generated candidate crack parameter combinations, outputting labels for residual bearing capacity and deterioration rate with real fracture mechanics residual information, generating a high-fidelity dataset. This high-fidelity dataset is then mixed with the field monitoring database, and the mixed dataset is used to incrementally learn and reconstruct the hyperparameters of the LightGBM model, forcibly correcting the model's underlying decision surface to conform to the current real multi-field coupling decay law in the field. Based on the reconstructed and improved LightGBM model, the NSDWOA algorithm, which incorporates geological evolution monotonicity constraints, is restarted for global multi-objective optimization. Since the underlying evaluation benchmark has been corrected, the algorithm will output a new Pareto front solution set and updated recommended operation and maintenance thresholds. Through the aforementioned closed-loop structure, this invention not only achieves accurate assessment of the current state of the cracked lining, but also endows the entire decision-making system with the autonomous learning capability of "co-evolution" throughout the entire life cycle of the tunnel.

[0157] Furthermore, this invention further classifies the security status levels and corresponding operation and maintenance measures, specifically as follows:

[0158] (1) Safety. When both the measured and predicted values ​​are within the stable range of the Pareto front, the inter-domain difference index has not triggered the fine-tuning threshold, and the maximum principal stress of the key section is lower than the reduced compressive strength and the crack propagation rate tends to stabilize, the system determines that the structure is in a safe state. At this time, routine inspections should be maintained, and the monitoring frequency should be executed according to the original cycle. The system mainly relies on the physical information extreme learning machine network for routine inference, and no special physical intervention is required.

[0159] (2) Potential insecurity. If a single measured index does not exceed the allowable threshold, but multiple indices show a synchronous deterioration trend under coupled evaluation, or if the inter-domain difference index approaches the trigger threshold, indicating a potential risk of increased damage, the system determines that the structure has entered a potentially unsafe state. In this case, it is necessary to increase the monitoring frequency of key measurement points on site, dynamically reduce the cost weight coefficient of the active learning objective function, force the algorithm to output recommended coordinates of newly added sensors in high-risk and vulnerable areas of accessible topological manifolds (such as the crack propagation front) to implement dynamic densification of measurement points, and formulate preventive measures based on the development trend.

[0160] (3) Unsafe. When the measured value or high-confidence predicted value substantially exceeds the physical boundary constraints set by the LightGBM model, the system determines that the structure is in an unsafe state. At this time, a structural warning must be issued immediately, and traffic must be restricted or closed. The system will simultaneously retrieve the solutions in the Pareto set that satisfy the current weight preference and directly output reinforcement suggestions with engineering feasibility to assist the management department in activating the emergency response plan. When multiple indicators correspond to different levels, the highest level is taken as the comprehensive safety level. The thresholds corresponding to each level can be appropriately adjusted according to the importance level of the project, the design safety factor, and the requirements of the specifications.

[0161] Example 2:

[0162] This embodiment provides a service performance prediction system for cracked lining tunnels in complex mountainous areas. The system includes an acquisition module, a first processing module, a second processing module, a third processing module, and a fourth processing module, specifically comprising:

[0163] The acquisition module is used to acquire multi-factor model test data of cracked lining in complex mountain tunnels;

[0164] The first processing module is used to construct a multi-objective optimization model that integrates asymmetric risk and physical monotonicity constraints based on the multi-factor model test data, and obtain the optimal recommended operation and maintenance threshold set.

[0165] The second processing module is used to perform operational monitoring deployment processing based on the optimal recommended operation and maintenance threshold set, and to quantify cognitive uncertainty and optimize the location of measuring points through an active learning mechanism to obtain real-time updated on-site monitoring data.

[0166] The third processing module is used to construct a prediction model based on the real-time updated field monitoring data to obtain the prediction information of the current service status of the crack lining.

[0167] The fourth processing module is used to perform service performance evolution extrapolation based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, and obtain the optimized prediction results.

[0168] In one specific embodiment of this disclosure, the acquisition module includes:

[0169] The first processing unit is used to perform multi-factor coupled experimental design processing based on the characteristics of complex mountainous environments to obtain a set of experimental conditions.

[0170] The second processing unit is used to perform physical model test execution processing according to the set of test conditions to obtain structural response data under multiple conditions.

[0171] The third processing unit is used to perform multi-physics field monitoring processing on the structural response data under the multiple working conditions to obtain the original monitoring dataset.

[0172] The fourth processing unit is used to preprocess the original monitoring dataset to obtain the multi-factor model test data.

[0173] In one specific embodiment of this disclosure, the first processing module includes:

[0174] The fifth processing unit is used to perform physical feature screening processing based on the multi-factor model test data to obtain the screened feature set;

[0175] The sixth processing unit is used to construct a LightGBM regression model based on the filtered feature set, and to apply a high-weight penalty to the error of overestimating structural safety by introducing an asymmetric risk loss function to obtain a service performance mapping model.

[0176] The seventh processing unit is used to perform multi-field coupling objective function construction processing based on the service performance mapping model to obtain a comprehensive objective function;

[0177] The eighth processing unit is used to process the comprehensive objective function using the non-dominated sorting whale optimization algorithm to obtain the optimal recommended operation and maintenance threshold set.

[0178] In one specific embodiment of this disclosure, the fifth processing unit includes:

[0179] The ninth processing unit is used to perform multi-field energy gradient calculation processing based on multi-factor model experimental data to obtain the multi-field energy gradient coefficients of each feature;

[0180] The tenth processing unit is used to perform physical enhancement feature difference function reconstruction processing based on the multi-field energy gradient coefficients of each feature to obtain the feature difference metric.

[0181] The eleventh processing unit is used to perform feature weight update processing based on the feature difference metric to obtain the influence weight value of each feature.

[0182] The twelfth processing unit is used to perform feature filtering based on the influence weight values ​​of each feature to obtain the filtered feature set.

[0183] In one specific embodiment of this disclosure, the second processing module includes:

[0184] The thirteenth processing unit is used to perform initial monitoring system deployment processing based on the optimal recommended operation and maintenance threshold set to obtain an initial on-site monitoring database.

[0185] The fourteenth processing unit is used to perform monitoring blind zone uncertainty quantification based on active learning according to the initial field monitoring database, and obtain the uncertainty distribution of each unknown measuring point in the whole domain space;

[0186] The fifteenth processing unit is used to perform measurement point optimization and recommendation processing based on the uncertainty distribution to obtain the optimal coordinates of the newly added monitoring points.

[0187] The sixteenth processing unit is used to obtain real-time updated field monitoring data based on the coordinates of the optimal newly added monitoring points.

[0188] In one specific embodiment of this disclosure, the third processing module includes:

[0189] The seventeenth processing unit is used to perform multi-field coupling control equation definition processing based on the real-time updated field monitoring data to obtain composite physical constraints;

[0190] The eighteenth processing unit is used to reconstruct and pre-train a physical information extreme learning machine network with embedded crack tip singularity residuals under the composite physical constraints, and obtain initialized network weight parameters.

[0191] The nineteenth processing unit is used to perform domain difference quantization processing based on the initialized network weight parameters to obtain inter-domain difference indicators.

[0192] The twentieth processing unit is used to perform migration fine-tuning triggering processing based on the inter-domain difference index to obtain the current service status prediction information of the crack lining.

[0193] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0194] Example 3:

[0195] Corresponding to the above method embodiments, this embodiment also provides a service performance prediction device for cracked lining of tunnels in complex mountainous areas. The service performance prediction device for cracked lining of tunnels in complex mountainous areas described below can be referred to in correspondence with the service performance prediction method for cracked lining of tunnels in complex mountainous areas described above.

[0196] Figure 2 This is a block diagram illustrating a service performance prediction device 800 for cracked lining of tunnels in complex mountainous areas, according to an exemplary embodiment. Figure 2 As shown, the service performance prediction device 800 for tunnels with cracked linings in complex mountainous areas may include: a processor 801 and a memory 802. The device 800 may also include one or more of the following: a multimedia component 803, an I / O interface 804, and a communication component 805.

[0197] The processor 801 controls the overall operation of the service performance prediction device 800 for tunnels with cracked linings in complex mountainous areas, to complete all or part of the steps in the aforementioned service performance prediction method for tunnels with cracked linings in complex mountainous areas. The memory 802 stores various types of data to support the operation of the service performance prediction device 800 for tunnels with cracked linings in complex mountainous areas. This data may include, for example, instructions for any application or method operating on the service performance prediction device 800 for tunnels with cracked linings in complex mountainous areas, as well as application-related data, such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the service performance prediction device 800 for tunnels with cracked linings in complex mountainous areas and other devices. Wireless communication methods include Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, and an NFC module.

[0198] In an exemplary embodiment, the service performance prediction device 800 for cracked lining of tunnels in complex mountainous areas can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned service performance prediction method for cracked lining of tunnels in complex mountainous areas.

[0199] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the above-described method for predicting the service performance of tunnel linings with cracks in complex mountainous areas. For example, the computer-readable storage medium may be the memory 802 including the program instructions, which may be executed by the processor 801 of the device 800 for predicting the service performance of tunnel linings with cracks in complex mountainous areas to complete the above-described method for predicting the service performance of tunnel linings with cracks in complex mountainous areas.

[0200] Example 4:

[0201] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the service performance prediction method for cracked lining of tunnels in complex mountainous areas described above.

[0202] A readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method described above for predicting the service performance of cracked tunnel linings in complex mountainous areas.

[0203] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0204] 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.

[0205] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for predicting the service performance of cracked tunnel linings in complex mountainous areas, characterized in that, include: Obtain multi-factor model test data for cracked tunnel lining in complex mountainous areas; Based on the multi-factor model test data, a multi-objective optimization model integrating asymmetric risk and physical monotonicity constraints is constructed to obtain the optimal recommended operation and maintenance threshold set; Based on the optimal recommended operation and maintenance threshold set, the monitoring deployment process is carried out during the operation period. The uncertainty of cognition is quantified and the location of the measuring points is optimized through an active learning mechanism to obtain real-time updated on-site monitoring data. A prediction model is constructed based on the real-time updated field monitoring data to obtain the predicted information of the current service status of the crack lining; Based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, the service performance evolution is extrapolated to obtain the optimized prediction results.

2. The service performance prediction method for cracked lining of tunnels in complex mountainous areas according to claim 1, characterized in that, Obtain multi-factor model test data for cracked lining in complex mountain tunnels, including: Based on the characteristics of complex mountainous environments, a multi-factor coupled experimental design was carried out to obtain a set of experimental conditions. Based on the set of test conditions, physical model tests are performed to obtain structural response data under multiple conditions. Based on the structural response data under the aforementioned multiple operating conditions, multi-physics field monitoring processing is performed to obtain the original monitoring dataset; The original monitoring dataset is preprocessed to obtain the multi-factor model test data.

3. The service performance prediction method for cracked lining of tunnels in complex mountainous areas according to claim 1, characterized in that, Based on the multi-factor model experimental data, a multi-objective optimization model integrating asymmetric risk and physical monotonicity constraints is constructed, including: Based on the multi-factor model test data, physical feature screening is performed to obtain the screened feature set; A LightGBM regression model is constructed based on the selected feature set. By introducing an asymmetric risk loss function, a high-weight penalty is applied to the error of overestimating structural safety, resulting in a service performance mapping model. Based on the service performance mapping model, a multi-field coupled objective function is constructed to obtain a comprehensive objective function; The comprehensive objective function is processed using the non-dominated sorting whale optimization algorithm to obtain the optimal recommended operation and maintenance threshold set.

4. The service performance prediction method for cracked lining of tunnels in complex mountainous areas according to claim 3, characterized in that, Physical feature screening processing is performed based on the multi-factor model test data, including: Multi-field energy gradient calculations were performed on the multi-factor model experimental data to obtain the multi-field energy gradient coefficients of each feature. Based on the multi-field energy gradient coefficients of each feature, a physical enhancement feature difference function reconstruction process is performed to obtain the feature difference metric. The feature weights are updated based on the feature difference metric to obtain the influence weight values ​​of each feature. Feature filtering is performed based on the influence weight values ​​of each feature to obtain the filtered feature set.

5. The service performance prediction method for cracked lining of tunnels in complex mountainous areas according to claim 1, characterized in that, Operational monitoring deployment is performed based on the optimal recommended maintenance threshold set, including: The initial monitoring system deployment is carried out based on the optimal recommended operation and maintenance threshold set to obtain the initial field monitoring database. Based on the initial field monitoring database, the uncertainty of the monitoring blind zone is quantified using active learning to obtain the uncertainty distribution of each unknown measuring point in the entire space. Based on the uncertainty distribution, the measurement point optimization and recommendation process is performed to obtain the optimal coordinates of the newly added monitoring points; Real-time updated field monitoring data is obtained based on the coordinates of the optimal newly added monitoring points.

6. A service performance prediction system for cracked lining tunnels in complex mountainous areas, characterized in that, include: The acquisition module is used to acquire multi-factor model test data of cracked lining in complex mountain tunnels; The first processing module is used to construct a multi-objective optimization model that integrates asymmetric risk and physical monotonicity constraints based on the multi-factor model test data, and obtain the optimal recommended operation and maintenance threshold set. The second processing module is used to perform operational monitoring deployment processing based on the optimal recommended operation and maintenance threshold set, and to quantify cognitive uncertainty and optimize the location of measuring points through an active learning mechanism to obtain real-time updated on-site monitoring data. The third processing module is used to construct a prediction model based on the real-time updated field monitoring data to obtain the prediction information of the current service status of the crack lining. The fourth processing module is used to perform service performance evolution extrapolation based on the optimal recommended operation and maintenance threshold set and the current service status prediction information of the crack lining, and obtain the optimized prediction results.

7. The service performance prediction system for cracked tunnel linings in complex mountainous areas according to claim 6, characterized in that, The acquisition module includes: The first processing unit is used to perform multi-factor coupled experimental design processing based on the characteristics of complex mountainous environments to obtain a set of experimental conditions. The second processing unit is used to perform physical model test execution processing according to the set of test conditions to obtain structural response data under multiple conditions. The third processing unit is used to perform multi-physics field monitoring processing on the structural response data under the multiple working conditions to obtain the original monitoring dataset. The fourth processing unit is used to preprocess the original monitoring dataset to obtain the multi-factor model test data.

8. The service performance prediction system for cracked tunnel linings in complex mountainous areas according to claim 6, characterized in that, The first processing module includes: The fifth processing unit is used to perform physical feature screening processing based on the multi-factor model test data to obtain the screened feature set; The sixth processing unit is used to construct a LightGBM regression model based on the filtered feature set, and to apply a high-weight penalty to the error of overestimating structural safety by introducing an asymmetric risk loss function to obtain a service performance mapping model. The seventh processing unit is used to perform multi-field coupling objective function construction processing based on the service performance mapping model to obtain a comprehensive objective function; The eighth processing unit is used to process the comprehensive objective function using the non-dominated sorting whale optimization algorithm to obtain the optimal recommended operation and maintenance threshold set.

9. The service performance prediction system for cracked lining of tunnels in complex mountainous areas according to claim 8, characterized in that, The fifth processing unit includes: The ninth processing unit is used to perform multi-field energy gradient calculation processing based on multi-factor model experimental data to obtain the multi-field energy gradient coefficients of each feature; The tenth processing unit is used to perform physical enhancement feature difference function reconstruction processing based on the multi-field energy gradient coefficients of each feature to obtain the feature difference metric. The eleventh processing unit is used to perform feature weight update processing based on the feature difference metric to obtain the influence weight value of each feature. The twelfth processing unit is used to perform feature filtering based on the influence weight values ​​of each feature to obtain the filtered feature set.

10. The service performance prediction system for cracked lining of tunnels in complex mountainous areas according to claim 6, characterized in that, The second processing module includes: The thirteenth processing unit is used to perform initial monitoring system deployment processing based on the optimal recommended operation and maintenance threshold set to obtain an initial on-site monitoring database. The fourteenth processing unit is used to perform monitoring blind zone uncertainty quantification based on active learning according to the initial field monitoring database, and obtain the uncertainty distribution of each unknown measuring point in the whole domain space; The fifteenth processing unit is used to perform measurement point optimization and recommendation processing based on the uncertainty distribution to obtain the optimal coordinates of the newly added monitoring points. The sixteenth processing unit is used to obtain real-time updated field monitoring data based on the coordinates of the optimal newly added monitoring points.

Images