A method for assessing the cascading failure of underground water supply pipeline structures

Abstract

This invention relates to the field of data processing and pipeline structure safety assessment fusion technology, and discloses a method for assessing the cascading failure of underground water supply pipeline structures. The method includes: acquiring multi-source data of the pipeline to be assessed; constructing a structural failure knowledge graph and rule base, generating rule consistency constraints and boundary condition update rules; loading a knowledge-data driven proxy model to obtain the predicted results of the pipeline's structural response indicators; constructing a unified limit state model for multiple failure modes, obtaining the failure probability calculation results corresponding to each failure mode, and determining the dominant failure mode; when the failure probability reaches a preset threshold, performing cascading propagation deduction to obtain the cascading propagation deduction results; and generating the final assessment result of the pipeline structure cascading failure based on the predicted results of the structural response indicators, the calculated failure probability results, and the cascading propagation deduction results. This invention can achieve computable and interpretable assessments even with incomplete data and dynamically changing operating conditions.

Description

Technical Field

[0001] This invention relates to the field of data processing and pipeline structure safety assessment fusion technology, specifically to a method for assessing the cascading failures of underground water supply pipeline structures. Background Technology

[0002] Underground water supply pipelines are subjected to a complex environment involving internal pressure fluctuations, traffic loads, external pressure from overburden and groundwater, uneven ground settlement, third-party disturbances, and corrosion degradation. This makes them prone to defects such as corrosion thinning, crack propagation, ellipticization, localized voids, and loosened interface constraints, leading to stress concentration and reduced load-bearing capacity. In engineering practice, the common risk is not the instantaneous rupture of a single pipe segment, but rather the abrupt change in surrounding supports and interface constraints after a segment enters a state of failure or severe damage. This causes a redistribution of load paths and boundary conditions in adjacent pipe segments, resulting in stress / deformation propagating to neighboring areas and accelerating damage in adjacent segments, ultimately leading to a progressively expanding cascading failure of structural performance along the pipeline.

[0003] Several solutions exist in the existing technology:

[0004] (1) Establish a pipe-soil finite element model, generate training samples, train a neural network to quickly predict stress, and further calculate the structural reliability. Its advantage is accelerated calculation, but it often still takes "single-segment response" to "single-segment reliability" as the core output, and it is difficult to explicitly characterize the cascading expansion driven by boundary mutation.

[0005] (2) Based on crack images and vibration data, a crack parameter and stress mapping is established through graph convolutional network, and the ductility of cracks in adjacent pipelines is scored, which is more inclined to "crack identification-stress estimation-trend warning" and image association with adjacent cracks.

[0006] Furthermore, methods such as health assessment at the water supply network level and critical valve identification are mostly based on statistical indicators and water supply functions, which are not at the same level as the mechanical boundary update mechanism of cascade failure of structural performance.

[0007] Therefore, a technical solution is needed to propose a closed loop that can form the "failure triggering-boundary mutation-response redistribution-cascade expansion" process under conditions of incomplete multi-source data and dynamic changes in operating conditions, and output calculable and interpretable cascade failure assessment results for underground water supply pipeline structures. Summary of the Invention

[0008] To address the technical problems of existing technologies, such as high data dependence and difficulty in reflecting the mutual influence between pipe segments and the evolution of structural performance, this invention provides a method for assessing the cascading failure of underground water supply pipeline structures, comprising:

[0009] Multi-source data of the pipeline to be evaluated is acquired, spatiotemporally aligned, and standardized multi-source data and multi-source data fitting error are obtained. The multi-source data includes the geometric dimensions and material parameters of the pipeline asset, operating internal pressure and external load conditions, defect detection information, and structural monitoring and historical operation and maintenance data.

[0010] Based on the standardized multi-source data and combined with the physical and mechanical mechanisms of pipeline structures, a structural failure knowledge graph and rule base are constructed to generate rule consistency constraints and boundary condition update rules. The structure of the structural failure knowledge graph is: pipeline component—defect—operating condition—failure mode—boundary condition change—evidence; the structure of the rule base is: triggering criterion—reasoning result.

[0011] Load a knowledge-data driven proxy model, which is used to obtain the prediction results of the structural response index of the pipeline; input the standardized multi-source data of the pipeline to be evaluated and the structural boundary parameters of the pipeline into the knowledge-data driven proxy model, and output the prediction results of the structural response index of the pipeline to be evaluated; the structural response index includes physical and mechanical mechanism parameters, which include stress, deformation and boundary parameters.

[0012] Based on the prediction results of the structural response index, a unified limit state model for multiple failure modes is constructed; the unified limit state model obtains the failure probability calculation results corresponding to each failure mode according to the failure probability threshold and failure probability index, and determines the master failure mode; the failure probability index includes a single failure mode and a unified failure probability.

[0013] When the failure probability reaches a threshold, a cascading propagation simulation is performed based on the structural failure knowledge graph and rule base. The cascading propagation simulation results are obtained according to the master failure mode. The cascading propagation simulation results include the cascading propagation path of structural failure and key intervention windows.

[0014] Based on the predicted results of the structural response indicators, the calculated failure probability results, and the cascading propagation simulation results, the final assessment results of the pipeline structure cascading failure are generated; the assessment results include the pipe segment structural risk indicators, the cascading propagation path and its probability of occurrence, the key triggering pipe segments, and the intervention priority to block the cascading failure.

[0015] Furthermore, before loading the knowledge-data-driven agent model, the knowledge-data-driven agent model is constructed, including the following:

[0016] The existing pipeline's multi-source data is acquired, spatiotemporally aligned, and standardized multi-source data and multi-source data fitting error of the existing pipeline are obtained.

[0017] A pipe-soil coupled structural model and a semi-analytical empirical model are loaded, and standardized multi-source data of the existing pipeline are input to generate multiple multi-fidelity samples to form a multi-fidelity training set; the multi-fidelity includes high fidelity and low fidelity.

[0018] Based on the structural failure knowledge graph and rule base, obtain the rule consistency constraints and boundary condition update rules of existing pipeline standardized multi-source data;

[0019] Define a knowledge-data driven agent model, which is a multi-input multi-output feedforward neural network structure. The input includes standardized multi-source data of the pipeline and boundary condition update rules, and the output is the structural response index of the pipeline.

[0020] Define the physical constraints and cascading consistency constraints of the knowledge-data driven agent model;

[0021] Based on the multi-source data fitting error, rule consistency constraints, physical constraints, and cascade consistency constraints, a joint training objective is formed. The knowledge-data driven agent model is then trained using the boundary condition update rules and the multi-fidelity training set, thus completing the construction of the knowledge-data driven agent model.

[0022] Furthermore, the process of forming the multi-fidelity training set includes:

[0023] Load the pipe-soil coupled structure model and the semi-analytical empirical model, input the standardized multi-source data of the existing pipeline, the pipe-soil coupled structure model is used to generate multiple high-fidelity samples; the semi-analytical empirical model is used to generate multiple low-fidelity samples; the high-fidelity samples and low-fidelity samples are fused to generate multiple multi-fidelity samples, forming a multi-fidelity training set.

[0024] Furthermore, alternative models to the pipe-soil coupled structure model include the elastic foundation model and the pipe-soil contact finite element model.

[0025] Furthermore, the definitions of the physical constraints and cascading consistency constraints of the knowledge-data driven agent model include:

[0026] The physical constraints are physical constraints used to describe the structural response of underground water supply pipelines, including mechanical equilibrium constraints and boundary condition constraints.

[0027] The cascaded consistency constraint is used to describe the consistency relationship that should be maintained between the structural response increments of adjacent pipe segments when the structural boundary conditions change, reflecting the structural response transmission effect caused by boundary changes.

[0028] Furthermore, the process of obtaining the failure probability calculation results corresponding to each failure mode includes the following:

[0029] Based on the prediction results of the structural response index, a unified limit state model with multiple failure modes is constructed, wherein the multiple failure modes include yielding, rupture, fracture, fatigue, and external pressure buckling.

[0030] Determine the criteria for multiple failure modes, map the criteria for multiple failure modes to a unified failure probability index, set the failure probability threshold for each failure probability index, and calculate the failure probability corresponding to each failure mode.

[0031] The failure mode of the master controller is determined based on the failure probability.

[0032] Furthermore, the process of obtaining the cascading propagation deduction results includes:

[0033] When the failure probability reaches the failure probability threshold, rule reasoning is performed based on the structural failure knowledge graph and rule base to generate boundary condition update rules.

[0034] Based on the boundary condition update rule, the structural response index prediction and failure probability calculation are performed iteratively. A maximum iteration threshold is set. When the number of iterations reaches the iteration threshold, the currently determined master failure mode is output.

[0035] Based on the currently determined master failure mode, obtain the cascading propagation deduction results.

[0036] Furthermore, the failure probability is used as a triggering criterion for cascading propagation deduction.

[0037] Furthermore, the boundary condition update rule is used to update the boundary conditions of adjacent pipe segments, and the boundary conditions include support stiffness and interface constraint conditions.

[0038] Furthermore, the spatiotemporal alignment includes spatial alignment of multi-source data, time axis alignment, and error consistency labeling, specifically including the following:

[0039] The multi-source data is spatially aligned using coordinate mapping;

[0040] The data from different sampling frequencies in the multi-source data are unified to the same time axis by time interpolation or resampling.

[0041] Uncertainty labeling is applied to the multi-source data errors to obtain the multi-source data fitting error.

[0042] The beneficial effects of this invention are as follows: By integrating structural failure knowledge with multi-source monitoring, detection and simulation data, this invention constructs a knowledge-data driven model that incorporates physical and mechanical constraints, enabling efficient assessment of the structural response and failure risk of underground water supply pipelines. While ensuring assessment accuracy, it reduces dependence on large-scale sample data, and can achieve calculable and interpretable assessments even with incomplete data and dynamic changes in operating conditions, making it highly applicable to engineering projects. Attached Figure Description

[0043] Figure 1 This is a flowchart of the cascade failure assessment method for underground water supply pipeline structures provided by the present invention. Detailed Implementation

[0044] The technical solution of the present invention is further described below, but the scope of protection is not limited to what is described.

[0045] This invention provides a method for assessing the cascading failure of underground water supply pipeline structures, such as... Figure 1 As shown, it includes:

[0046] Step S100: Obtain multi-source data of the pipeline to be evaluated, perform spatiotemporal alignment, and obtain standardized multi-source data and multi-source data fitting error; the multi-source data includes the geometric dimensions and material parameters of the pipeline asset, operating internal pressure and external load conditions, defect detection information, and structural monitoring and historical operation and maintenance data;

[0047] The spatiotemporal alignment includes spatial alignment of multi-source data, time axis alignment, and error consistency labeling, specifically including the following:

[0048] The multi-source data is spatially aligned using coordinate mapping;

[0049] The data from different sampling frequencies in the multi-source data are unified to the same time axis by time interpolation or resampling.

[0050] Uncertainty labeling is applied to the multi-source data errors to obtain the multi-source data fitting error.

[0051] Step S200: Based on the standardized multi-source data and combined with the physical and mechanical mechanisms of the pipeline structure, construct a structural failure knowledge graph and a rule base, and generate rule consistency constraints and boundary condition update rules; wherein, the structure of the structural failure knowledge graph is: pipeline component—defect—operating condition—failure mode—boundary condition change—evidence; the structure of the rule base is: triggering criterion—reasoning result;

[0052] The boundary condition update rule is used to update the boundary conditions of adjacent pipelines, and the boundary conditions include support stiffness and interface constraint conditions.

[0053] When specific evidence or operating conditions are met, the rule base outputs one or a combination of the following calculation results:

[0054] (1) Correction of the prior weights of failure modes (i.e., cascade weights);

[0055] (2) Suggestions for updating structural boundary conditions (such as support stiffness or interface constraints);

[0056] (3) Correction of the distribution of uncertain parameters.

[0057] The rule base is used, on the one hand, to constrain the prediction results during the training of the surrogate model to not violate the structural failure mechanism, and on the other hand, to perform reasoning on the boundary condition update rules during the cascade propagation deduction process.

[0058] Step S300: Load the knowledge-data driven proxy model, which is used to obtain the prediction results of the structural response index of the pipeline; input the standardized multi-source data of the pipeline to be evaluated and the structural boundary parameters of the pipeline into the knowledge-data driven proxy model, and output the prediction results of the structural response index of the pipeline to be evaluated; the structural response index includes physical and mechanical mechanism parameters, which include stress, deformation and boundary parameters.

[0059] Before loading the knowledge-data-driven agent model, the knowledge-data-driven agent model is trained. The training process includes the following:

[0060] Step S310: Obtain the multi-source data of the existing pipeline, perform spatiotemporal alignment, and obtain the standardized multi-source data and multi-source data fitting error of the existing pipeline;

[0061] Step S320: Load the pipe-soil coupled structure model and the semi-analytical empirical model, input the standardized multi-source data of the existing pipeline, generate multiple multi-fidelity samples, and form a multi-fidelity training set; the multi-fidelity includes high fidelity and low fidelity.

[0062] The pipe-soil coupled structural model and the semi-analytical empirical model are used to analyze the stress and deformation relationship between the underground water supply pipeline and the surrounding soil, and generate structural response samples; the structural response samples are multi-fidelity samples.

[0063] Alternative models for the pipe-soil coupled structure model include the elastic foundation model, the pipe-soil contact finite element model, and other equivalent techniques.

[0064] The process of forming the multi-fidelity training set includes:

[0065] A pipe-soil coupled structure model and a semi-analytical empirical model are loaded, and standardized multi-source data of the existing pipeline are input. The pipe-soil coupled structure model is used to generate multiple high-fidelity samples; the semi-analytical empirical model is used to generate multiple low-fidelity samples; the high-fidelity samples and low-fidelity samples are fused to generate multiple multi-fidelity samples, forming a multi-fidelity training set; in order to reduce the dependence on the calculation of a large number of high-fidelity samples.

[0066] Step S330: Based on the structural failure knowledge graph and rule base, obtain the rule consistency constraints and boundary condition update rules of the existing pipeline standardized multi-source data;

[0067] Step S340: Define a knowledge-data driven agent model. The knowledge-data driven agent model is a multi-input multi-output feedforward neural network structure. The input includes standardized multi-source data of the pipeline and boundary condition update rules. The output is the structural response index of the pipeline.

[0068] Step S350: Define the physical constraints and cascading consistency constraints of the knowledge-data driven agent model;

[0069] The physical constraints are physical constraints used to describe the structural response of underground water supply pipelines, including mechanical equilibrium constraints and boundary condition constraints.

[0070] The cascaded consistency constraint is used to describe the consistency relationship that should be maintained between the structural response increments of adjacent pipe segments when the structural boundary conditions change, reflecting the structural response transmission effect caused by boundary changes.

[0071] Step S360: Based on the multi-source data fitting error, rule consistency constraints, physical constraints, and cascade consistency constraints, a joint training objective is formed. The knowledge-data driven agent model is trained by combining the boundary condition update rules and the multi-fidelity training set, thus completing the construction of the knowledge-data driven agent model.

[0072] Step S400: Based on the prediction results of the structural response index, a unified limit state model for multiple failure modes is constructed; the unified limit state model obtains the failure probability calculation results corresponding to each failure mode according to the failure probability threshold and the failure probability index, and determines the master failure mode; the failure probability index includes a single failure mode and a unified failure probability.

[0073] The failure probability is used as a triggering criterion for cascading propagation.

[0074] The process of obtaining the failure probability calculation results corresponding to each failure mode includes the following:

[0075] Based on the prediction results of the structural response index, a unified limit state model with multiple failure modes is constructed, wherein the multiple failure modes include yielding, rupture, fracture, fatigue, and external pressure buckling.

[0076] Determine the criteria for multiple failure modes, map the criteria for multiple failure modes to a unified failure probability index, set the failure probability threshold for each failure probability index, and calculate the failure probability corresponding to each failure mode.

[0077] The failure mode of the master controller is determined based on the failure probability.

[0078] Step S500: When the failure probability reaches the threshold, cascading propagation is performed based on the structural failure knowledge graph and rule base. The cascading propagation results are obtained according to the master failure mode. The cascading propagation results include the cascading propagation path of structural failure and key intervention windows.

[0079] The process of obtaining the cascading propagation results includes:

[0080] When the failure probability reaches the failure probability threshold, rule reasoning is performed based on the structural failure knowledge graph and rule base to generate boundary condition update rules.

[0081] Based on the boundary condition update rule, the structural response index prediction and failure probability calculation are performed iteratively. A maximum iteration threshold is set. When the number of iterations reaches the iteration threshold, the currently determined master failure mode is output.

[0082] Based on the currently determined master failure mode, obtain the cascading propagation deduction results.

[0083] Step S600: Based on the predicted results of the structural response index, the calculated failure probability, and the cascading propagation simulation results, the final assessment result of the pipeline structure cascading failure is generated; the assessment result includes the pipe segment structural risk index, the cascading propagation path and its probability of occurrence, the key triggering pipe segments, and the intervention priority for blocking the cascading failure.

[0084] An example is provided below, using three adjacent sections of a city water supply main as the objects to be evaluated:

[0085] Pipe section number is Connected sequentially according to mileage, the interface set is as follows .

[0086] S1 acquires multi-source data for the pipeline to be evaluated:

[0087] 1) Pipe type: Welded steel pipe (This method is also applicable to ductile iron pipe, PE pipe and reinforced concrete pipe. When implementing it, only the corresponding material parameters and corresponding failure mode library need to be changed).

[0088] 2) Geometric parameters: outer diameter Nominal wall thickness Single segment length .

[0089] 3) Material parameters: elastic modulus Poisson's ratio Yield strength fracture toughness .

[0090] 4) Typical burial depth range: In this example, Soil weight .

[0091] 5) Operating internal pressure: Taking the quasi-steady-state fluctuation, the evaluation time in this example is taken as... .

[0092] 6) Equivalent additional vertical stress from traffic dynamic loads: (Equivalent value, used to form distributed load; can be obtained from monitoring / load model).

[0093] 7) Test data:

[0094] Pipe section 1 has an axial corrosion defect, with a defect length of [missing information]. The corrosion depth distribution along the axial direction is set to a parabolic shape (obtained by fitting discrete detection points):

[0095] (1)

[0096] in, The corrosion depth is a positional function along the axial direction; Maximum corrosion depth; Length of axial corrosion defects.

[0097] This example takes The remaining wall thickness at the critical section

[0098] (2)

[0099] in, This represents the remaining wall thickness of the cross-section. This refers to the nominal wall thickness of the pipeline.

[0100] Pipe section 2 showed no significant corrosion, but at the interface... Evidence of constraint slack was detected at the site (such as misalignment / increased opening), accompanied by evidence of local voiding and settlement (work order + ground settlement).

[0101] 8) Monitoring data:

[0102] Strain monitoring and settlement monitoring are used to calibrate the initial value range of boundary parameters (the initial values ​​are given directly in this example to illustrate the method flow).

[0103] S2 constructs a structural failure knowledge graph and rule base, generating rule consistency constraints and boundary condition update rules:

[0104] This example calls the rule library. Output two types of key quantities: failure mode prior weights and boundary condition update rules.

[0105] 1) Boundary condition update rules, represented by boundary condition parameter vectors (per segment):

[0106] (3)

[0107] in, Let be the boundary condition parameter vector for the k-th segment of the pipeline; This refers to the equivalent support stiffness between the pipe section and the soil. , These are the axial and rotational constraint stiffness of the interface, respectively.

[0108] Initial settings for this example:

[0109] (4)

[0110] (5)

[0111] These correspond to the equivalent support stiffness and rotational constraint stiffness of pipe segments 1, 2, and 3 with the soil, respectively.

[0112] 2) Prior weights (i.e., cascaded weights): Because and Shared Interface , and Shared Interface Take the neighborhood The prior weights between adjacent pipe segments are obtained by combining "shared interface + distance + sensitivity" normalization.

[0113] (6)

[0114] in, This represents the prior weight (i.e., cascade weight) between pipe segments 1 and 2; This represents the prior weight (i.e., cascade weight) between pipe segments 2 and 3.

[0115] 3) Reasoning results: Pipe segment 2 has a combination of evidence: "voiding + interface loosening + significant dynamic load". The rule base increases the prior weights of buckling / bending control and outputs updated boundary conditions. In this example, the updated boundary conditions are updated using the correction coefficients.

[0116] (7)

[0117] in, The evidence for the removal of voids mainly weakens the support constraint correction factor. The interface relaxation mainly weakens the rotation constraint correction coefficient.

[0118] S3 obtains the predicted results of structural response indices:

[0119] This example uses a combination of "low-fidelity approximation (beam-elastic foundation model) + physical constraint proxy model" to replace the pipe-soil coupled structure model and the semi-analytical empirical model, where bending stress... The results are provided by a computable beam-elastic foundation model (or can be directly output from a trained knowledge-data driven surrogate model); the calculated predictions include:

[0120] 1) Equivalent distributed load: The vertical stress is converted into a load per unit length, taking...

[0121] (8)

[0122] in, Represents the equivalent vertical stress; Indicates soil weight; Indicates the burial depth of the pipeline; This represents the equivalent additional vertical stress caused by traffic dynamic loads.

[0123] (9)

[0124] Where q=p eq , represents the equivalent vertical stress; Indicates the outer diameter of the pipe.

[0125] 2) Moment of inertia of section 1 (using the remaining wall thickness) ):

[0126] (10)

[0127] (11)

[0128] in, This refers to the inner diameter of the pipe. Let be the moment of inertia of the cross section 1.

[0129] Take the distance between the outer fibers of the cross section (12)

[0130] 3) Equivalent span length after detachment: Initial moment The length of the unsupported span of pipe section 1 is given by the inspection. Maximum bending moment under simplified beam conditions:

[0131] (13)

[0132] in, This indicates the maximum bending moment of pipe segment 1; This indicates the equivalent span length of pipe segment 1 after it is emptied.

[0133] Substituting, we get:

[0134] (14)

[0135] Bending stress (low-fidelity approximation output):

[0136] (15)

[0137] in, This indicates the bending stress in pipe section 1; This indicates the distance from the outer fiber of the cross section to the neutral axis.

[0138] 4) Circumferential and axial membrane stresses:

[0139] (16)

[0140] in, This indicates the circumferential membrane stress in pipe section 1; Indicates the internal pressure during operation; This represents the remaining wall thickness of the cross section.

[0141] (17)

[0142] in, This indicates the axial stress in pipe section 1.

[0143] 5) Von Mises equivalent stress (taking shear stress) (Conservative simplification)

[0144] (18)

[0145] in, This indicates the equivalent stress of pipe segment 1 (von Mises). Indicates circumferential membrane stress; This indicates axial stress.

[0146] S4 constructs a unified limit state model with multiple failure modes, obtains the failure probability calculation results corresponding to each failure mode, and determines the master failure mode.

[0147] 1) Yield limit state:

[0148] (19)

[0149] in, This indicates the yield limit state of pipe section 1; Indicates yield strength; This indicates the equivalent stress of pipe segment 1 von Mises.

[0150] 2) Corrosion cracking (residual strength) limit state:

[0151] This example calculates the defect area ratio using a general method:

[0152] (20)

[0153] in, Indicates the effective area of ​​corrosion defects; The corrosion depth is a positional function along the axial direction; Maximum corrosion depth; Length of axial corrosion defects.

[0154] (twenty one)

[0155] in, Indicates the nominal area that has not been corroded; Length of axial corrosion defects; Indicates the nominal wall thickness of the pipe; This indicates the defect area ratio.

[0156] The geometric magnification factor is:

[0157] (twenty two)

[0158] in, Indicates the geometric magnification factor of the defect; Geometric magnification parameters; Length of axial corrosion defects.

[0159] Substitution have to Equivalent failure strength parameter values ​​under corrosion cracking limit state Then the rupture pressure is:

[0160] (twenty three)

[0161] in, This indicates the internal pressure of pipe 1 after corrosion and rupture. Indicates flow stress; This represents the geometric magnification factor of the defect.

[0162] achievable Much larger Therefore, in this case, rupture was not the primary controlling factor at the initial moment. This conclusion is consistent with the yield / bending control commonly seen under the conditions of "low internal pressure / strong external load + defect coupling" in water supply mains.

[0163] 3) Unifying extreme states and master failure modes:

[0164] (twenty four)

[0165] in, To represent the unified limit state function value of pipe segment 1 at the initial time t=0; These are respectively represented as the yield failure limit state function, internal pressure rupture failure limit state function, buckling failure limit state function, and fracture failure limit state function of pipe segment 1 at t=0.

[0166] This example is in The master failure mode at any given time is "small yield margin but still safe".

[0167] 4) Failure probability (instantiated calculation): , We can approximate it as a normally distributed random variable. Let... Considering the uncertainty of internal pressure and corrosion, ( (This standard deviation can be obtained from the S1 uncertainty label and the S3 sample statistics), then

[0168] (25)

[0169] in, The function representing the yield failure limit state; The mathematical expectation of the yield limit state function; It represents the standard deviation.

[0170] Reliable indicators:

[0171] (26)

[0172] in, Indicates structural reliability indicators; The mathematical expectation of the yield limit state function; Indicates standard deviation; This indicates the probability that pipe segment 1 will yield and fail under the current condition; It is the standard normal distribution function.

[0173] Note: Under the coupling of corrosion and external load, although pipe segment 1 has not failed, it is in a sensitive state with "a low probability of failure but close to the threshold", which provides a criterion for subsequent cascading triggering.

[0174] S5 performs cascading propagation simulation to obtain cascading propagation simulation results, which include the cascading propagation path of structural failure and key intervention windows.

[0175] This example sets a failure probability threshold. Due to the combination of evidence of "vacuuming + interface slack," segment 2 is classified as high-risk by the rule base. Then the failure severity factor is:

[0176] (27)

[0177] in, Indicators representing the severity of failure; This indicates the probability that pipe segment 2 will yield and fail under the current condition;

[0178] 1) Boundary condition update (for adjacent segment 1):

[0179] (28)

[0180] in, Indicates the equivalent foundation support stiffness; Indicates the cascading weights between adjacent pipe segments; Indicates severity factor; This indicates the rule trigger decay coefficient.

[0181] (29)

[0182] in, Indicates the equivalent rotational stiffness of the interface; Indicates the cascading weights between adjacent pipe segments; Indicates severity factor; This indicates the rule trigger decay coefficient.

[0183] The engineering implications are: Neighborhood failure / high risk leads to a significant weakening of the support and rotation constraints of pipe section 1.

[0184] 2) Span length update and bending stress growth: Weakening of the support leads to an expansion of the effective span length of the void (which can be updated by the rule base or foundation model). In this example, we take: ,but:

[0185] (30)

[0186] (31)

[0187] 3) Updated equivalent stress and yield criterion:

[0188] (32)

[0189] (33)

[0190] (34)

[0191] Therefore, after the first round of cascaded updates, pipe section 1 entered the failure control state, and the main failure mode changed from "near yield" to "yield failure (bending-dominated)".

[0192] 4) Failure probability update: Similarly, assume... ,but:

[0193] (35)

[0194] in, This represents the mean; Indicates standard deviation; Indicates a reliable indicator; This indicates the probability of failure.

[0195] That is, the failure probability of pipe segment 1 increases dramatically, realizing the structural cascading trigger of "high-risk segment 2 → boundary mutation → segment 1 yielding".

[0196] 5) Output of cascade propagation simulation results: In High-risk collection at all times ,exist time Therefore, the first two steps of the cascaded expansion path are: Continuing to perform the same update on segment 3, we can obtain... Or the path that terminates after intervention, used to provide a key intervention window.

[0197] S6 generates the final assessment results of the cascading failure of the pipeline structure:

[0198] In this example, the final evaluation result of the cascading failure of the pipeline structure is: if in Before implementing "support restoration (backfilling and grouting / foundation reinforcement) or interface constraint restoration (reinforcing sleeves / replacing connectors)" on pipe section 1, No expansion or If it is not lower than the threshold, then it can be Controlled within approximately Level, thus maintaining To achieve path The blocking. Output "critical intervention window" (before the first round of cascading updates) and "priority" (segment 2 is the triggering segment, segment 1 is the highly sensitive affected segment).

[0199] The above-disclosed embodiments are merely specific examples of the present invention. However, the present invention is not limited thereto, and any variations that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A method for assessing the cascading failure of underground water supply pipeline structures, characterized in that, include: Multi-source data of the pipeline to be evaluated is acquired, spatiotemporally aligned, and standardized multi-source data and multi-source data fitting error are obtained. The multi-source data includes the geometric dimensions and material parameters of the pipeline asset, operating internal pressure and external load conditions, defect detection information, and structural monitoring and historical operation and maintenance data. Based on the standardized multi-source data and combined with the physical and mechanical mechanisms of pipeline structures, a structural failure knowledge graph and rule base are constructed to generate rule consistency constraints and boundary condition update rules. The structure of the structural failure knowledge graph is: pipeline component—defect—operating condition—failure mode—boundary condition change—evidence; the structure of the rule base is: triggering criterion—reasoning result. Load a knowledge-data driven proxy model, which is used to obtain the prediction results of the structural response index of the pipeline; input the standardized multi-source data of the pipeline to be evaluated and the structural boundary parameters of the pipeline into the knowledge-data driven proxy model, and output the prediction results of the structural response index of the pipeline to be evaluated; the structural response index includes physical and mechanical mechanism parameters, which include stress, deformation and boundary parameters. Based on the prediction results of the structural response index, a unified limit state model for multiple failure modes is constructed; the unified limit state model obtains the failure probability calculation results corresponding to each failure mode according to the failure probability threshold and failure probability index, and determines the master failure mode; the failure probability index includes a single failure mode and a unified failure probability. When the failure probability reaches a threshold, a cascading propagation simulation is performed based on the structural failure knowledge graph and rule base. The cascading propagation simulation results are obtained according to the master failure mode. The cascading propagation simulation results include the cascading propagation path of structural failure and key intervention windows. Based on the predicted results of the structural response indicators, the calculated failure probability results, and the cascading propagation simulation results, the final assessment results of the pipeline structure cascading failure are generated; the assessment results include the pipe segment structural risk indicators, the cascading propagation path and its probability of occurrence, the key triggering pipe segments, and the intervention priority to block the cascading failure. The process of obtaining the failure probability calculation results corresponding to each failure mode includes the following: Based on the prediction results of the structural response index, a unified limit state model with multiple failure modes is constructed, wherein the multiple failure modes include yielding, rupture, fracture, fatigue, and external pressure buckling. Determine the criteria for multiple failure modes, map the criteria for multiple failure modes to a unified failure probability index, set the failure probability threshold for each failure probability index, and calculate the failure probability corresponding to each failure mode. The failure mode of the master controller is determined based on the failure probability.

2. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 1, characterized in that, Before loading the knowledge-data-driven agent model, construct the knowledge-data-driven agent model, including the following: The existing pipeline's multi-source data is acquired, spatiotemporally aligned, and standardized multi-source data and multi-source data fitting error of the existing pipeline are obtained. The pipe-soil coupled structural model and the semi-analytical empirical model are loaded, and the standardized multi-source data of the existing pipeline are input to generate multiple multi-fidelity samples to form a multi-fidelity training set; the multi-fidelity includes high fidelity and low fidelity. Based on the structural failure knowledge graph and rule base, obtain the rule consistency constraints and boundary condition update rules of existing pipeline standardized multi-source data; Define a knowledge-data driven agent model, which is a multi-input multi-output feedforward neural network structure. The input includes standardized multi-source data of the pipeline and boundary condition update rules, and the output is the structural response index of the pipeline. Define the physical constraints and cascading consistency constraints of the knowledge-data driven agent model; Based on the multi-source data fitting error, rule consistency constraints, physical constraints, and cascade consistency constraints, a joint training objective is formed. The knowledge-data driven agent model is then trained using the boundary condition update rules and the multi-fidelity training set, thus completing the construction of the knowledge-data driven agent model.

3. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 2, characterized in that, The process of forming the multi-fidelity training set includes: Load the pipe-soil coupled structure model and the semi-analytical empirical model, input the standardized multi-source data of the existing pipeline, the pipe-soil coupled structure model is used to generate multiple high-fidelity samples; the semi-analytical empirical model is used to generate multiple low-fidelity samples; the high-fidelity samples and low-fidelity samples are fused to generate multiple multi-fidelity samples, forming a multi-fidelity training set.

4. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 2, characterized in that, Alternative models for the pipe-soil coupled structure model include the elastic foundation model and the pipe-soil contact finite element model.

5. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 2, characterized in that, The definitions of the physical constraints and cascading consistency constraints of the knowledge-data driven agent model include: The physical constraints are physical constraints used to describe the structural response of underground water supply pipelines, including mechanical equilibrium constraints and boundary condition constraints. The cascaded consistency constraint is used to describe the consistency relationship that should be maintained between the structural response increments of adjacent pipe segments when the structural boundary conditions change, reflecting the structural response transmission effect caused by boundary changes.

6. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 1, characterized in that, The process of obtaining the cascading propagation results includes: When the failure probability reaches the failure probability threshold, rule reasoning is performed based on the structural failure knowledge graph and rule base to generate boundary condition update rules. Based on the boundary condition update rule, the structural response index prediction and failure probability calculation are performed iteratively. A maximum iteration threshold is set. When the number of iterations reaches the iteration threshold, the currently determined master failure mode is output. Based on the currently determined master failure mode, obtain the cascading propagation deduction results.

7. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 6, characterized in that, The failure probability is used as a triggering criterion for cascading propagation.

8. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 1, characterized in that, The boundary condition update rule is used to update the boundary conditions of adjacent pipelines, and the boundary conditions include support stiffness and interface constraint conditions.

9. The method for assessing the cascading failure of underground water supply pipeline structures as described in claim 1, characterized in that, The spatiotemporal alignment includes spatial alignment of multi-source data, time axis alignment, and error consistency labeling, specifically including the following: The multi-source data is spatially aligned using coordinate mapping; The data from different sampling frequencies in the multi-source data are unified to the same time axis by time interpolation or resampling. Uncertainty labeling is applied to the errors of the multi-source data to obtain the multi-source data fitting error.

Images