Method for diagnosing service performance of structure based on high dam accident mode and failure path

Abstract

The invention provides a method for diagnosing a structural service behavior based on a high dam accident mode and a failure path, and the method comprises the steps: constructing a high dam accident case database, and extracting accident feature parameters; by applying multi-scale layered adaptive corresponding analysis, through feature weight calculation, risk-out mode layered weighting, physical similarity index establishment and structural difference compensation, the problem of unbalanced feature weight is solved, and a weighted similar project sorting list is generated; constructing a multi-failure mode coupled time sequence adaptive network model, solving the problem that a failure path time sequence evolution mechanism is imperfect, and outputting a time sequence failure path prediction model; and in combination with the similar project sorting list and the time sequence failure path prediction model, performing project analogy analysis, and generating a high dam safety state evaluation report. The technical problems that the operation time of the ultra-high dam is short, the monitoring information is insufficient, the numerical analysis result is difficult to completely reflect the engineering practice and the like are solved, and a new technical path and a new solution are provided for high dam safety assessment and risk management.

Description

Technical Field

[0001] The invention relates to building diagnosis technology, in particular to a method for diagnosing structural service performance based on high dam danger modes and failure paths. Background Art

[0002] As a major water conservancy infrastructure, high dams and large reservoirs play an important role in ensuring flood control, water supply, irrigation and safety. With the increase of service life and the influence of extreme effects such as earthquakes and excessive floods, the probability of high dams suffering from cracks in the dam body and serious leakage increases, causing some losses. Therefore, systematic analysis of high dam failure modes and failure paths and establishment of scientific structural service performance diagnosis methods have important theoretical value and engineering significance for ensuring the safe operation of water conservancy projects, and are hot research directions at home and abroad.

[0003] At present, the safety assessment of high dams at home and abroad mainly relies on in-situ monitoring and on-site testing to obtain disease information, and reveals the physical cause mechanism by establishing monitoring models, conducting simulation tests and structural numerical calculations. For example, the high dam safety assessment method based on neural networks uses historical monitoring data for pattern recognition and early warning; the assessment method based on risk probability analysis calculates the failure probability through Monte Carlo simulation; the assessment method based on finite element analysis analyzes the stress, deformation and seepage characteristics of high dams under different working conditions through numerical simulation. In addition, the multivariate correspondence analysis method is applied to engineering analogy analysis, which provides a reference for safety assessment by comparing the similarities between the high dam to be evaluated and historical cases.

[0004] However, existing technologies still face key technical difficulties in diagnosing high dam accident modes and failure paths. On the one hand, equal weight processing is generally used in traditional multivariate correspondence analysis, ignoring the differences in the degree of influence of different characteristic parameters on accident modes and failure paths, resulting in excessive attention to secondary features and neglect of key factors during engineering analogies, which reduces the accuracy of diagnosis. On the other hand, the existing failure path analysis mainly presents a static topological structure, lacks an accurate description of the temporal evolution relationship between each link in the failure process, and cannot reflect the development rate of the disease and the time dimension characteristics of the failure process, making the high dam operation risk warning and life assessment lack accuracy on the time scale. These problems are particularly prominent in 300m-class ultra-high dam projects. Due to the short construction and operation time and insufficient accumulation of monitoring information, they exceed the existing technical level and specification requirements, and there is an urgent need to develop new diagnostic methods. Summary of the invention

[0005] The purpose of the invention is to provide a method for diagnosing the service performance of a structure based on a high dam's accident mode and failure path, in order to solve at least one technical problem existing in the prior art.

[0006] A technical solution, a method for diagnosing the service performance of a structure based on a high dam failure mode and failure path, comprises the following steps:

[0007] Collect high dam accident case information, build a high dam accident case database, and extract characteristic parameters, accident modes and failure paths to form a high dam accident case data set;

[0008] Based on the high dam accident case data set, multi-scale hierarchical adaptive correspondence analysis is applied to generate a weighted ranking list of similar projects;

[0009] Based on the high dam accident case data set, a time-series adaptive network model with multiple failure mode coupling is constructed, and a time-series failure path prediction model is output;

[0010] Combined with the weighted similar engineering ranking list and the sequential failure path prediction model, an engineering analogy analysis is conducted on the high dam to be evaluated and a service safety diagnosis report is generated.

[0011] Beneficial effects: This application provides a new technical path and solution for high dam safety assessment and risk management. The relevant technical effects will be described in detail with case studies. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a flow chart of the present invention.

[0013] Figure 2 It is a flow chart of the present invention for generating a weighted similar project ranking list.

[0014] Figure 3 It is a flow chart of calculating feature weight vector of the present invention.

[0015] Figure 4 It is a flow chart of constructing a pattern feature weight matrix of the present invention.

[0016] Figure 5 It is a flow chart of the calculation process of correcting the physical similarity index matrix of the present invention. DETAILED DESCRIPTION

[0017] like Figures 1 to 5 As shown, the technical solution of the present application includes the following steps:

[0018] S1. Collect information on high dam accident cases at home and abroad, build a structured database, extract accident characteristic parameters, and form a high dam accident case data set.

[0019] S11. Obtain original case information from engineering documents, accident reports, and monitoring records, including basic parameters of high dams (dam type, dam height, reservoir capacity, construction year, and country of location) and accident information (accident time, accident type, accident location, and treatment measures).

[0020] S12. Perform data cleaning and standardization on the original case information, remove outliers and cases with serious missing data, unify measurement units and classification standards, and generate standardized case data.

[0021] S13. Based on the standardized case data, extract the dam structure characteristics (such as dam type coefficient, dam height ratio, thickness coefficient), hydrogeological characteristics (such as reservoir capacity coefficient, geological conditions), and operating environment characteristics (such as meteorological conditions, operating years) to form a characteristic parameter matrix X.

[0022] S14. Analyze the accident patterns of standardized case data, identify the three main accident patterns of cracks, abnormal deformation, and seepage and their subtypes, and construct the accident pattern classification system Y.

[0023] S15. Analyze the evolution process of accidents in standardized case data, identify the logical relationship between the causes, development process, and failure results of accidents, and construct the initial failure path network Z.

[0024] S16. Integrate the characteristic parameter matrix X, the accident mode classification system Y and the initial failure path network Z into a structured database to form a complete high dam accident case data set.

[0025] S2. Based on the high dam accident case data set, multi-scale hierarchical adaptive correspondence analysis is applied to solve the problem of feature weight imbalance and generate a weighted ranking list of similar projects.

[0026] S21. Extract the characteristic parameter matrix X from the high dam accident case data set and calculate the point bi-column correlation coefficient ρ between each characteristic parameter and the accident mode ij , construct the feature correlation matrix R.

[0027] S22. Based on the feature correlation matrix R, the average correlation coefficient ρj of each feature is calculated and normalized to obtain the initial feature weight vector W0. This solves the limitation of equal weight processing and highlights the importance of key features.

[0028] S23. According to the accident mode classification system Y, the high dam accident case data set is divided into m subsets C1, C2, ..., Cm, and the feature weight is calculated for each subset separately to obtain the pattern feature weight matrix WM. This solves the problem of feature importance differences under different accident modes.

[0029] S24. Construct the risk pattern correlation matrix R, the matrix element r ij It represents the correlation between accident patterns i and j, which is calculated based on co-occurrence frequency and conversion probability. This step considers the evolutionary relationship between accident patterns.

[0030] S25. Construct a physical similarity index system, including n physical indexes such as stress distribution similarity (σ similarity), deformation mode similarity (ε similarity), seepage field similarity (Φ similarity), etc., and establish a mapping relationship F between physical indexes and basic characteristic parameters: characteristic parameter matrix X → physical similarity index matrix P. The introduction of physical mechanism enhances the scientific nature of similarity analysis.

[0031] S26. Apply the mapping relationship F to the characteristic parameter matrix X to obtain the physical similarity index matrix P of all historical cases. For the special structural characteristics of modern ultra-high dams, the structural difference compensation function H is designed to correct the physical similarity index and obtain the corrected physical similarity index matrix P'. This solves the problem of structural representative differences between new high dams and historical cases.

[0032] S27, extract the characteristic parameters of the high dam to be evaluated, construct the characteristic vector X0 of the high dam to be evaluated, and obtain the physical similarity index vector P0 of the high dam to be evaluated through the mapping relationship F and the compensation function H.

[0033] S28. Calculate the weighted Euclidean distance DX between the characteristic vector X0 of the high dam to be evaluated and each case in the characteristic parameter matrix X in the original characteristic space, and determine the weight based on the characteristic correlation matrix R, the pattern characteristic weight matrix WM and the accident pattern correlation matrix R. At the same time, calculate the weighted Euclidean distance DP between the physical similarity index vector P0 of the high dam to be evaluated and the modified physical similarity index matrix P' in the physical similarity space.

[0034] S29. Adopt a multi-scale fusion strategy, integrate DX and DP to calculate the final similarity S = α·f(DX) + (1-α)·g(DP), where α is an adaptive weight factor, and f and g are distance-similarity conversion functions. Sort the historical cases based on the final similarity S and generate a weighted similar project ranking list. Make full use of the advantages of the original feature space and the physical similarity space to improve the reliability of similar project screening.

[0035] S3. Using the failure information in the high dam accident case data set, a time-series adaptive network model with multiple failure modes coupled is constructed to solve the problem of imperfect failure path time-series evolution mechanism and output a time-series failure path prediction model.

[0036] S31. Extract the initial failure path network Z from the high dam accident case data set, define each node in the network as a state Si of the Markov chain, add the normal state S0 and the final failure state Sf, and form a complete state space S.

[0037] S32. Based on the historical cases in the high dam accident case data set, the transition frequencies between states are counted and the state transition probability matrix P is calculated, where P ijRepresents the transition probability from state Si to state Sj. It solves the limitation of only representing the logical sequence without quantifying the probability.

[0038] S33. For each state transition process Si→Sj, extract the time record from the high dam accident case data set and fit its time distribution function T ij (t) and its parameters (mean μ ij , variance σ ij 2 ), construct the state duration distribution matrix D. Introducing the time dimension into failure path analysis solves the problem that static paths cannot reflect timing characteristics.

[0039] S34. Analyze the main external factors (water level, temperature, load, etc.) that affect the high dam failure process, extract relevant monitoring data from the high dam accident case data set, and construct the external factor data matrix F.

[0040] S35. Establishing conditional state transition probability function P ij (F1, F2, ..., Fk), represents the relationship between the state transition probability and external factors, and establishes the mapping relationship between the time distribution parameters and external factors: μ ij =g(F1,F2,...,Fk),σ ij 2 =h(F1,F2,...,Fk). It can adapt to the non-stationary failure process under complex working conditions.

[0041] S36. Model different types of failure modes (cracks, deformation, seepage, etc.) as independent network layers and construct a multi-layer failure mode network M. Define the coupling strength matrix C between failure modes, where C ij It represents the influence of failure mode i on mode j. It can characterize the interaction between failure modes.

[0042] S37. A rule-based path generation engine is used to automatically generate new possible paths in the multi-layer failure mode network M according to the failure mode coupling relationship, and the initial failure path network Z is dynamically updated to form an extended failure path network Z'. This solves the problem that the preset path cannot cover the path mutation.

[0043] S38. Based on the state transition probability matrix P, state duration distribution matrix D, F, multi-layer failure mode network M and extended failure path network Z', a comprehensive time series evolution model is constructed to achieve multi-scale (macro state sequence, meso network propagation, micro physical evolution) simulation of the failure process. The full-scale time series modeling of the high dam failure process is achieved.

[0044] S39. Generate a large number of possible failure timing paths through Monte Carlo simulation method, statistically analyze the time characteristics of typical failure paths, identify key nodes and critical states that may cause path mutations, and form a complete timing failure path prediction model.

[0045] S4. Combined with the weighted similar engineering ranking list and the time series failure path prediction model, an engineering analogy analysis is conducted on the high dam to be evaluated, and a high dam safety status assessment report is generated.

[0046] S41. Select the k projects with the highest similarity from the weighted similar project ranking list as the core analog project set, analyze their failure modes, locations and causes, and predict the potential failure modes of the high dam to be evaluated.

[0047] S42. Input the monitoring data of the high dam to be evaluated as the initial state information into the time series failure path prediction model, simulate and predict its possible failure evolution path and time window, and generate a failure risk time series diagram.

[0048] S43. For typical projects in the core analogy project set, combined with the characteristics of the high dam to be evaluated, a finite element numerical model is constructed to conduct a failure mechanism analysis, verify the rationality of the potential accident mode and failure risk time series diagram, and form a verification analysis result.

[0049] S44. Based on the potential failure modes, failure risk time series diagrams and verification analysis results, comprehensively evaluate the safety status of the high dam to be evaluated, determine the risk level, put forward targeted monitoring and intervention strategy recommendations, and generate a high dam safety status assessment report.

[0050] According to one aspect of the present application, step S15 is specifically:

[0051] S151. Extract the accident event sequence of each case from the standardized case data, including the accident phenomenon, occurrence time, development status and impact scope, to form the accident event time series data E.

[0052] S152: Decompose the accident event time series data E, identify the starting event, intermediate development event and final result event in each case, and construct a case-level event chain set L. Through time window segmentation and event correlation analysis, the accurate deconstruction of the accident process is achieved.

[0053] S153, semantic normalization is performed on all events in the event chain set L, and events with different expressions but the same essence are merged and unified to obtain a standardized event type set T. This step adopts an ontology mapping method based on expert knowledge to solve the problem of inconsistent event descriptions in different projects and different documents.

[0054] S154, based on the event type set T, analyze the dependencies between the events in the event chain set L, count the transfer frequencies between the event pairs, and form an event transfer frequency matrix F. The matrix elements f ij Represents the number of historical occurrences from event i to event j.

[0055] S155, normalize the event transfer frequency matrix F and calculate the conditional probability P(j|i)=f ij / ∑fik, construct event transition probability matrix P event . A probabilistic statistical method is introduced to transform the qualitative failure path into a quantifiable probability model.

[0056] S156, based on the event transition probability matrix P event , the minimum spanning tree algorithm is used to extract the main event transfer paths, and the transfer relations with probability lower than the threshold θ are eliminated to obtain the core event transfer network N core .

[0057] S157, combining high dam engineering mechanics theory and expert knowledge, the core event transfer network N core The event transfer relationship in the data is verified for physical rationality, the physically unreasonable transfer edges are deleted, and the transfer edges that must exist physically but are missing in the data are supplemented to form a physically corrected event network N. phy .

[0058] S158. Classify the events in the event type set T in two dimensions according to the risk type (cracks, deformation, seepage, etc.) and development stage (start, development, failure), and construct an event classification matrix C. The matrix provides a structured framework for the multi-dimensional display of failure paths.

[0059] S159, based on physical correction event network N phy The event classification matrix C is used to construct a graph-structured initial failure path network Z, where nodes are event types, edges are event transition relationships, and edge weights are transition probabilities. This network not only retains the topological structure of the failure process, but also integrates the probabilistic characteristics, laying the foundation for subsequent temporal evolution analysis.

[0060] According to one aspect of the present application, step S22 is specifically:

[0061] S221. Read the feature correlation matrix R and extract the correlation coefficient ρ between each feature parameter j and all risk patterns ij , where i represents the accident mode index, j represents the feature parameter index, and constructs the feature-accident correlation coefficient set ρ_set.

[0062] S222. Perform statistical analysis on the feature-risk correlation coefficient set ρ_set, and calculate the average correlation coefficient ρ_avg of each feature jj = (1 / M)∑ρ ij , where M is the total number of accident modes, and the average correlation coefficient vector ρ_avg is obtained. Through averaging processing, the ability of features to influence different accident modes is comprehensively considered.

[0063] S223, perform a significance test on the values ​​in the average correlation coefficient vector ρ_avg, screen out feature parameters with insignificant correlation (p value>0.05), retain a subset of features with significant correlation, and construct a significant feature index set J_sig. This step introduces a statistical test method to improve the scientific nature of feature weight calculation.

[0064] S224, for the features in the significant feature index set J_sig, an expert scoring mechanism is introduced to determine the expert weight coefficient e of each feature based on historical case analysis experience and theoretical analysis results j , forming the expert weight vector E. Combining expert experience with data statistics enhances the rationality of the weight system.

[0065] S225, design the fusion function f, and transform the data-driven average correlation coefficient vector ρ avg Weighted fusion with expert weight vector E: f j = α·ρ avgj + (1-α)·e j , where α is the adaptive fusion coefficient, which is dynamically adjusted according to the data quality. The fusion weight value vector F is obtained. The fusion of data-driven and expert knowledge is realized, overcoming the limitations of a single weight determination method.

[0066] S226. Perform sensitivity analysis on the fusion weight value vector F. By perturbing each feature weight and calculating the degree of influence on the result, the stability influence of each feature is evaluated to obtain the stability coefficient vector S.

[0067] S227. Based on the fusion weight value vector F and the stability coefficient vector S, a stability enhancement function g is designed to calculate the weight value after stability adjustment: g j = F j ·(1+β·S j ), where β is the stability adjustment coefficient. The stable adjustment weight vector G is obtained. The stability of the weight is taken into account, which enhances the robustness of the model.

[0068] S228, normalize the stable adjustment weight vector G: j = g j / ∑g j, ensuring that the sum of all weights is 1, and obtaining the final initial feature weight vector W0. This weight vector breaks through the limitations of the traditional multivariate correspondence analysis medium weight processing and can more accurately reflect the actual impact of different features on the risk mode.

[0069] According to one aspect of the present application, step S23 is specifically:

[0070] S231. Read the accident mode classification system Y and the high dam accident case data set, group the cases according to the main accident modes (cracks, abnormal deformation, seepage, etc.), and form m pattern subsets C1, C2, ..., Cm, each subset contains all cases with the same main accident mode.

[0071] S232, for each pattern subset Ci, extract the case ID list and corresponding feature parameters contained therein, and construct a pattern-specific feature parameter submatrix X i .

[0072] S233, for each feature parameter sub-matrix X i , calculate the point-wise bi-column correlation coefficient between each feature and the sub-pattern, and construct the pattern-specific feature correlation sub-matrix R i . The correlation analysis of accident mode stratification is realized, which can capture the difference in feature importance under specific accident modes.

[0073] S234, for each feature correlation sub-matrix R i , apply the feature weight calculation method of the above-mentioned step S22 (S221-S228) to obtain the feature weight vector W for the specific risk mode i i . Implemented specialized weight calculation for accident patterns.

[0074] S235. Analyze the sample distribution characteristics of each risk pattern subset Ci and calculate the sample size n i , sample coverage c i and sample diversity index d i , construct the pattern sample characteristic matrix D. This matrix is ​​used to evaluate the reliability of the weight calculation of each pattern subset.

[0075] S236. Based on the pattern sample characteristic matrix D, design the reliability evaluation function h and calculate the reliability coefficient r of each pattern weight vector i = h(n i , c i , d i ), construct the model weight reliability vector R mode . A weight reliability evaluation mechanism is introduced to solve the uncertainty problem of weight calculation for small sample patterns.

[0076] S237, for the model weight reliability vector R mode For the mode j whose reliability is lower than the threshold λ, a transfer learning strategy is adopted to transfer weight information from similar modes for reinforcement: W j ' = W j + γ·∑(sim(j,k)·(W_k - W j )), where sim(j,k) is the similarity between patterns j and k, and γ is the migration intensity coefficient. The reinforcement weight vector set W' is obtained. The weight calculation problem of rare accident patterns is solved, and the scope of application of the method is enhanced.

[0077] S238, all the feature weight vectors (W i or the enhanced W j ') into a two-dimensional matrix of pattern-feature, forming a complete pattern-feature weight matrix WM. Each row of the matrix represents an accident pattern, each column represents a feature parameter, and the matrix element WM ij Represents the weight value of feature j under risk mode i.

[0078] S239, based on the pattern feature weight matrix WM, calculate the coefficient of variation CV of each feature weight between different patterns, identify pattern sensitive features (large CV value) and pattern stable features (small CV value), and construct the feature pattern sensitivity vector S feature This vector provides important reference information for the subsequent weight optimization of specific high dams.

[0079] According to one aspect of the present application, step S25 is specifically:

[0080] S251. Read the characteristic parameter matrix X and the accident mode classification system Y, analyze the physical mechanism of high dam failure, determine the key physical quantities affecting the safety of the high dam, including stress field, deformation field, seepage field and temperature field, and form a key physical quantity set K.

[0081] S252. For each physical quantity in the key physical quantity set K, define its similarity metric standard, design a similarity calculation function, and form a physical similarity metric function set Φ. This function set includes the stress distribution similarity function φ σ , deformation pattern similarity function φ ε , seepage field similarity function φ f Etc., each function can convert the characteristics of the corresponding physical field into a similarity value between 0 and 1.

[0082] S253. Based on high dam mechanics theory and historical case analysis, identify the key characteristic parameter combinations that affect the distribution of various physical quantities, and establish a mapping relationship set between characteristic parameters and physical quantities {M σ , M ε , Mf ,...}. For example, the similarity of stress distribution is mainly affected by parameters such as dam type, dam height, and dam thickness ratio; the similarity of seepage field is mainly affected by parameters such as permeability coefficient, anti-seepage facilities, and geological conditions. A mapping mechanism from feature space to physical space is established.

[0083] S254, for stress distribution similarity (σ similarity), extract the feature subset X that affects stress distribution from the feature parameter matrix X σ , including dam type coefficient, dam height, thickness-to-height ratio, bending radius, etc., and applying the nonlinear mapping function M σ Converted into stress field feature vector, and then through the similarity function φ σ Calculate the stress similarity index P σ This index can quantify the similarity of stress distribution patterns of different dam bodies and provide a mechanical perspective for the safety evaluation of high dams.

[0084] S255, for deformation mode similarity (ε similarity), extract the feature subset X that affects the deformation characteristics from the feature parameter matrix X ε , including material elastic modulus, Poisson's ratio, temperature deformation coefficient, etc., and applying the deformation mapping function M ε Converted into deformation field feature vector, and then through the similarity function φ ε Calculate the deformation similarity index P ε This index can compare the similarities of deformation modes of different high dams and is particularly suitable for evaluating abnormal deformation-related accident modes.

[0085] S256. For the seepage field similarity (Φ similarity), extract the feature subset X that affects the seepage characteristics from the feature parameter matrix X f , including water permeability, anti-seepage measures, geological conditions, etc., and applying the seepage mapping function M f Converted into the seepage field feature vector, and then through the similarity function φ f Calculate the seepage similarity index P f This index can evaluate the similarity of seepage characteristics of different high dams and is of great significance for analyzing leakage hazards.

[0086] S257. Combining material science and damage mechanics theory, a damage evolution similarity index (D-similarity) is designed. The characteristic subset X_d that affects the development of material damage is extracted from the characteristic parameter matrix X, including material type, service time, load history, etc. The damage similarity index P_d is calculated using the damage mapping function M_d. The similarity of the service status of high dams can be evaluated from the perspective of material damage.

[0087] S258. For each pair of high dams (i, j), calculate the physical similarity index values ​​to form a physical similarity index vector P ij = [Pσij , P εij , P fij , P_d ij ,…]. The physical similarity index vectors of all high-dam pairs are combined into a three-dimensional tensor, which is converted into a two-dimensional matrix through dimension compression to construct the complete physical similarity index matrix P. Each row in the matrix corresponds to a high-dam case, and each column corresponds to a physical similarity index.

[0088] S259. Analyze the correlation between the indicators in the physical similarity indicator matrix P, identify redundant indicators and complementary indicators, and optimize the physical similarity indicator system. Finally, determine the mapping relationship F between physical indicators and basic characteristic parameters: characteristic parameter matrix X → physical similarity indicator matrix P. The mapping relationship F includes the integration of the mapping functions of various physical quantities, which can transform the engineering characteristics of the high dam into comparable physical similarity indicators, and provide a scientific basis for the subsequent screening of similar projects based on physical mechanisms.

[0089] According to one aspect of the present application, step S26 is specifically:

[0090] S261, read the characteristic parameter matrix X and the mapping relationship F, apply the physical mapping transformation to the characteristic parameters of each historical case, calculate its physical similarity index value, and construct the original physical similarity index matrix P.

[0091] S262. Analyze the differences between modern ultra-high dams and high dams in the historical case library in terms of structural design, material properties, construction technology, etc., identify key difference features, and construct a structural difference feature set D. This set includes the unique features of modern ultra-high dams, such as new dam structures, high-performance concrete materials, special construction technologies, etc. These features may lack direct correspondence in historical cases.

[0092] S263, for each difference feature d in the structural difference feature set D i , analyze its influence mechanism on each physical similarity index, and construct the difference-impact mapping table M_diff. This table describes the direction and intensity of the influence of each structural difference feature on each physical similarity index. The structural representative differences between modern ultra-high dams and historical cases are quantified.

[0093] S264, based on the difference-impact mapping table M_diff, design a parameterized structural difference compensation function H, which receives the original physical similarity index P and the structural difference feature d i As input, the output is the modified physical similarity index P': P' = H(P, d_1, d_2, ..., d_k). The function H adopts a piecewise continuous nonlinear mapping form, which can provide adaptive compensation for different degrees of structural differences.

[0094] S265. Use a high-precision finite element model to simulate and analyze the impact of typical structural differences and determine the parameter values ​​in the structural difference compensation function H. By comparing the physical field distribution under different structural designs, quantifying the physical similarity deviation caused by structural differences, optimizing the compensation function parameters, and obtaining the calibrated structural difference compensation function H cal .

[0095] S266. Design the dam structure compensation sub-function H for new dam structures in modern ultra-high dams (such as hyperbolic arch dams, RCC gravity dams, etc.). dam , focusing on adjusting the stress distribution and deformation mode similarity indicators related to the dam type. By comparing the mechanical behavior differences between the new dam type and the traditional dam type, the compensation coefficient is determined and the dam type compensation coefficient matrix C is constructed. dam .

[0096] S267. Design material property compensation subfunction H for high-performance material properties (such as low-heat cement, high-strength concrete, etc.) mat This function mainly adjusts the damage evolution and temperature field similarity indicators related to material properties. Based on material science theory and experimental data, the corresponding relationship between material performance parameters and physical behavior is established to form the material compensation coefficient matrix C mat .

[0097] S268. Considering the improvement of modern construction technology and quality control level, design the construction quality compensation sub-function H con , which adjusts the similarity index of seepage field and interface performance related to construction quality. By comparing the quality difference between modern construction and traditional construction, the compensation coefficient is determined and the construction compensation coefficient matrix C is established. con .

[0098] S269, integrating each dedicated compensation sub-function into the main compensation function H cal In the paper, each index value in the physical similarity index matrix P is compensated and corrected to obtain the modified physical similarity index matrix P'. This matrix reflects the physical similarity between historical cases and modern ultra-high dams under the condition of considering structural differences, overcomes the limitation of relying only on the comparison of original characteristic parameters, and provides physical mechanism support for the accurate screening of similar projects.

[0099] According to one aspect of the present application, step S29 is specifically:

[0100] S291, read the weighted Euclidean distance DX of the original feature space and the weighted Euclidean distance DP of the physical similarity space, and construct a dual space distance vector V dist = [DX, DP]. This vector contains the similarity distances of the high dams measured from two different perspectives.

[0101] S292. Design distance-similarity conversion functions f and g to convert distance metric into similarity metric: SX = f(DX), SP = g(DP). The conversion function adopts exponential decay form: f(d) = exp(-d 2 / σ 2 ), where σ is a scale parameter, and the optimal value is determined by cross-validation. The standardization of different spatial distance metrics is achieved, so that they can be effectively compared and integrated.

[0102] S293. Analyze the data distribution characteristics of different types of high dams in the high dam accident case data set, evaluate the data quality and representativeness of the original feature space and physical similarity space, and construct a spatial credibility evaluation index set Q = [QX, QP]. The index set includes evaluation indicators in multiple dimensions such as data coverage, sample consistency, and feature completeness.

[0103] S294. Based on the spatial credibility evaluation index set Q, an adaptive weight factor calculation function h is designed to determine the fusion weight α = h(QX, QP) of the original feature space and the physical similarity space. This function can dynamically adjust the fusion weight according to the data quality of the two spaces and improve the reliability of the similarity calculation. This solves the problem of uneven quality of different data spaces.

[0104] S295. Design a dedicated spatial weight adjustment coefficient β for different accident modes i , adjust the degree of dependence on physical mechanisms according to the danger mode: α i = α·β i For example, for stress-related crack disaster patterns, the weight of the physical similarity space is increased; for environment-related erosion disaster patterns, the weight of the original feature space is increased. The pattern space weight adjustment matrix B is constructed to achieve specialized spatial fusion of disaster patterns.

[0105] S296. For the high dam to be evaluated, the risk probability vector p = [p1, p2, ..., pm] of each risk mode is determined based on the preliminary risk assessment results. The risk probability can be obtained through preliminary monitoring data analysis, expert scoring or historical case statistics. The risk probability vector p is combined with the mode space weight adjustment matrix B to calculate the adaptive weight factor after comprehensive adjustment: α adj = ∑(pi·α i ), which takes into account both data quality and the characteristics of the accident pattern.

[0106] S297, design a multi-scale similarity fusion function, comprehensively consider the original feature space similarity SX and the physical similarity space similarity SP, and calculate the final similarity S = α adj ·SX + (1-α adj)·SP. For special cases, a nonlinear fusion strategy is introduced, and methods such as geometric weighted average or harmonic weighted average are used to enhance the fusion effect, and an enhanced multi-scale fusion similarity S is constructed. enh .

[0107] S298. Aiming at the multi-objective optimization requirements, a hierarchical similarity calculation method is designed to decompose the similarity S into multiple sub-similarity indicators: S struct (structural similarity), S mat (Material Similarity), S env (environmental similarity), etc., to construct a multidimensional similarity vector S multi . It can describe the similarity relationship between different high dams in detail from multiple dimensions, and support more refined screening of similar projects.

[0108] S299, based on the final similarity S or enhanced multi-scale fusion similarity S enh , sort the historical cases, select the k projects with the highest similarity as the core analogy objects, and generate a weighted similar project sorting list. At the same time, based on the multidimensional similarity vector S multi , provide similar project recommendations by category, form a multi-dimensional similar project analysis report, and provide a comprehensive and reliable reference for subsequent engineering analogy analysis. It makes full use of the advantages of the original feature space and physical similarity space to improve the reliability and adaptability of similar project screening.

[0109] According to one aspect of the present application, step S33 is specifically:

[0110] S331. Filter the case subsets containing complete time records from the high dam accident case data set to form a time series complete case set T_cases. It contains cases with clear records of the start and end time of each accident state, providing a data basis for time series analysis.

[0111] S332, for each case in the complete time series case set T_cases, based on the state space S and the state transition probability matrix P, identify the state sequence and state transition process it has experienced, extract the transition time data of each pair of adjacent states Si→Sj, and construct the state transition time data set T data This dataset records the actual occurrence time of each state transition in historical cases and is the basic data for fitting the time distribution function.

[0112] S333, for the state transition time data set T data Perform data cleaning and outlier detection, use the box plot method or Z-score method to identify abnormal time records, replace the outliers with reasonable estimates or mark them as missing, and obtain the cleaned time data set T clean . Improved data quality for time distribution fitting.

[0113] S334, for the cleaned time data set T clean For each pair of state transitions Si→Sj, extract its transition time sample T ij = {t1, t2, ..., tn}, analyze the data distribution characteristics, use the Shapiro-Wilk test and other methods to test its distribution type, determine the appropriate theoretical distribution model, and construct the distribution type mapping table M dist The optional distribution types include exponential distribution, Weibull distribution, gamma distribution, lognormal distribution, etc., which are suitable for describing different types of failure processes.

[0114] S335, based on the distribution type mapping table M dist Determine the distribution type, for each pair of state transition time samples T ij Perform parameter estimation and use maximum likelihood estimation (MLE) or moment method to estimate distribution parameters and obtain the distribution parameter set {θ ij}, construct the distribution parameter table P dist For example, for the exponential distribution, the estimated rate parameter λ is ij ; For the Weibull distribution, estimate the shape parameter k ij and scale parameter λ ij . The conversion from discrete time data to continuous distribution model is realized.

[0115] S336. For state transition processes with insufficient sample size (such as rare accident modes or new high dam-specific states), a parameter transfer learning method is designed to learn parameter information from similar state transitions, use Bayesian inference methods to estimate prior distributions, and combine finite sample data to update the posterior distribution to form a sparse state transition parameter estimation set P_sparse. This solves the problem of temporal distribution modeling of sparse data states.

[0116] S337. Based on the physical failure mechanism and reliability theory, a physical constraint correction function is designed to perform physical rationality verification and correction on the statistically estimated distribution parameters to ensure that the time distribution conforms to the physical laws and obtain the physical correction parameter set P. phy For example, for the crack extension process caused by material fatigue, the distribution parameters are corrected to make it conform to the Paris law; for the penetration erosion process, the parameters are corrected to make it conform to the law of seepage evolution.

[0117] S338. For each state transition process Si→Sj, based on its distribution type and parameters, calculate the key time feature quantities, including the mean μ ij (expected transfer time), variance σ ij 2 (Time volatility), median m ij (typical transfer time) and quantile q ij,α (reliability interval), forming the time feature matrix T feat The matrix comprehensively describes the temporal characteristics of each state transition process and provides a quantitative basis for risk assessment and intervention decisions.

[0118] S339, integrating the time distribution functions and parameters of all state transition processes, and constructing a complete state duration distribution matrix D. Each element D of this matrix ij The time distribution function T containing the state transition Si→Sj ij (t) and its parameters. The static failure path network is transformed into a dynamic model with a time dimension, which realizes the temporal expression of the high dam failure process for the first time and provides a mathematical basis for subsequent temporal evolution analysis and risk warning.

[0119] According to one aspect of the present application, step S35 is specifically:

[0120] S351, read the external factor data matrix F, the state transition probability matrix P and the state duration distribution matrix D, determine the key state transition process that needs to establish a conditional relationship, and form a key transition process set K trans This set includes state transition processes that have significant impact on high dam safety, such as crack initialization → crack expansion, abnormal seepage → piping formation, etc.

[0121] S352, analyze the importance of the factors in the external factor data matrix F, use statistics such as mutual information (MI) and correlation ratio (CR) to evaluate the influence of each factor on the state transition probability and time distribution, select k factors with significant influence, and construct the key external factor set F key = {F1, F2, ..., Fk}. Key factors usually include water level changes, temperature cycles, seismic activity, load history, etc.

[0122] S353、Design data tiering strategy, based on key external factors key The value range of each factor in the factor space is divided into multiple intervals to form a multidimensional grid, and the factor hierarchical framework G is constructed. This framework discretizes the complex multi-factor space into a limited number of factor combination regions, which is convenient for establishing conditional relationships.

[0123] S354. For each key state transition process Si→Sj and each factor combination region g in the factor hierarchical framework G, select state transition cases under corresponding conditions from the high dam accident case data set, count the state transition frequency, calculate the conditional transition probability P(Sj|Si,g), and construct a discrete conditional probability table CP discrete . A preliminary correlation between state transition probability and external factors was achieved.

[0124] S355. Design a conditional probability interpolation function for areas with insufficient discrete sampling points. Based on physical laws and mathematical models, realize continuous estimation of probability in factor space and construct a continuous conditional probability model CP. continuous The interpolation method uses non-parametric models such as radial basis function (RBF) network or Gaussian process regression (GPR), which can effectively process high-dimensional sparse data and solve the problem of incomplete data coverage.

[0125] S356. Analyze the interaction between external factors, design interaction modeling strategies, capture the nonlinear effects of multiple factors, and form an interaction effect model I model For example, the synergistic effects of water level changes and temperature fluctuations may have a greater impact than either alone, and this interactive effect is modeled using interaction terms or kernel methods.

[0126] S357, integrated discrete conditional probability table CP discrete , continuous conditional probability model CP continuous and interaction effect model I model , construct a complete conditional state transition probability function P ij (F1, F2, ..., Fk). This function adopts piecewise continuous mathematical form and can calculate the corresponding state transition probability for any given combination of external factors. It realizes the dynamic modulation of external factors on transition probability.

[0127] S358. Similarly, based on the time parameters (such as the mean μ ij , variance σ ij 2 ) and the factor hierarchical framework G, analyze the relationship between time distribution parameters and external factors, and establish the parameter mapping function μ ij =g(F1,F2,...,Fk) and σ ij 2 =h(F1,F2,...,Fk). The mapping function is implemented using techniques such as multivariate regression, response surface methodology or neural network, and can predict the state transition time characteristics under different external conditions.

[0128] S359, verify the performance of the conditional transition probability and time parameter mapping model, use the cross-validation method to evaluate the model prediction accuracy, determine the response characteristics of the model to changes in external factors through sensitivity analysis, and form the final non-stationary Markov model parameter set Θ. This parameter set fully describes the law of change of state transition characteristics with external factors, provides a mathematical tool for simulating the high dam failure process under complex working conditions, and is the core of realizing the prediction of non-stationary failure evolution.

[0129] According to one aspect of the present application, step S36 is specifically:

[0130] S361. Extract the main failure mode types from the high dam accident case data set and the accident mode classification system Y, including crack mode (CM), deformation mode (DM), seepage mode (FM), etc., to form the failure mode type set MODE = {CM, DM, FM, ...}. Each failure mode represents a typical physical manifestation of high dam accident.

[0131] S362, for each failure mode m in the failure mode type set MODE, extract the state nodes and transfer relationships related to the mode from the initial failure path network Z, construct a mode-specific subnetwork Zm, and form a mode subnetwork set Z sub = {ZCM, ZDM, ZFM, ...}. Each sub-network focuses on describing the evolution of a specific failure mode.

[0132] S363. For each mode sub-network Zm, based on the high dam engineering mechanics theory and historical case analysis, the state nodes and transfer relationships that may exist but are not fully reflected in the data are supplemented, the network topology is optimized, and an enhanced mode enhanced sub-network Z'm is formed. This makes up for the insufficiency of historical data and enables the sub-network to more completely describe the failure evolution process.

[0133] S364. Define network parameters for each pattern enhancement subnetwork Z'm, including node attributes (such as state duration, degree of harm) and edge attributes (such as transition probability, transition time), and construct a pattern network parameter set Pm. These parameters transform the pattern subnetwork from a qualitative description to a quantitative model, which can support numerical simulation and quantitative analysis.

[0134] S365. All mode enhancement sub-networks Z'm are taken as independent layers to construct a multi-layer network structure to form an initial multi-layer failure mode network M0. This network separates different types of failure modes into relatively independent layers, which is convenient for analyzing the internal evolution law of each mode and provides a basic framework for subsequent coupling analysis between modes.

[0135] S366. Based on high dam engineering practice and mechanical principles, the physical interaction mechanisms between different failure modes are analyzed, key inter-mode influence paths are identified, and the mode coupling path set L is constructed. For example, crack development may affect the seepage path (CM→FM), and abnormal seepage may accelerate material degradation and lead to new cracks (FM→CM). These coupling relationships reflect the mutual promotion or inhibition effects between failure modes.

[0136] S367. For each coupling path l(i,j) (from mode i to mode j) in the mode coupling path set L, a coupling strength evaluation method is designed to comprehensively consider physical mechanism analysis, historical case statistics and expert knowledge to calculate the inter-mode influence strength C. ij, construct the failure mode coupling strength matrix C. The coupling strength uses a standardized value in the interval [-1,1], where a positive value indicates a promoting effect, a negative value indicates an inhibiting effect, and the absolute value indicates the degree of influence.

[0137] S368. Design a coupled dynamics equation to describe the dynamic adjustment mechanism of the state transition probability under mode coupling: P' ij =P ij + ∑(Ckm·f(Sk,Sm)), where f(Sk,Sm) is the state interaction function. This equation can calculate the corrected transfer probability after considering mode coupling and construct the coupling adjustment probability model CP. This achieves a quantitative description of the state evolution due to the interaction between failure modes.

[0138] S369, integrate the failure mode coupling intensity matrix C and the coupling adjustment probability model CP into the multi-layer failure mode network M0, establish inter-layer connection and dynamic adjustment mechanism, and form a complete multi-layer failure mode network M. This network is a multi-layer, weighted, dynamic complex network structure, which not only describes the internal evolution law of each failure mode, but also depicts the interaction relationship between modes. It breaks through the limitations of traditional single failure path analysis, can capture the synergy and competition relationship between failure modes, and provides a new perspective for understanding and predicting complex failure processes.

[0139] According to one aspect of the present application, step S37 is specifically:

[0140] S371, read the multi-layer failure mode network M, the failure mode coupling strength matrix C and the initial failure path network Z, analyze the network structure characteristics, including node connectivity, path length distribution and key node distribution, and construct a network feature statistical set E. This statistical set describes the topological characteristics of the failure path network and provides a reference for subsequent path generation.

[0141] S372. Design a path generation rule base, including physical rationality rules (constraints based on mechanical principles), historical similarity rules (references based on case patterns), and structural integrity rules (based on network topology requirements), to form a path generation rule set R. The rule set is described in a formal language, such as "if state A appears and environmental conditions meet X, it may change to state B", and each rule has corresponding trigger conditions and execution actions.

[0142] S373. Based on the path generation rule set R, a rule execution engine is designed. Using a forward reasoning mechanism, starting from the initial state, possible state transition sequences are iteratively generated according to the rule set, and a rule-driven path set P_rule is constructed. The engine can automatically explore the state space allowed by the rules and discover failure paths that may exist but have not appeared in historical data.

[0143] S374. Combined with the failure mode coupling mechanism, a cross-layer path generation algorithm is designed. The algorithm can jump between mode layers and generate a composite path containing multiple failure mode interactions based on the failure mode coupling strength matrix C to form a cross-mode path set P_cross. For example, from a certain state of the fracture layer to the relevant state of the seepage layer through the coupling relationship, and then jump back to the fracture layer, a composite failure path of "fracture-seepage-fracture" is formed. It can discover collaborative failure paths that are difficult to identify in single mode analysis.

[0144] S375. Aiming at the special structural characteristics of modern ultra-high dams, a specialized path generation module is designed. Combining finite element analysis and expert knowledge, a special failure path for ultra-high dam structural characteristics is generated, and a special path set P_high for ultra-high dams is constructed. This module takes into account special factors such as temperature control, segmented casting, and stress distribution of ultra-high dams, and can generate failure paths that are more in line with the actual ultra-high dam project.

[0145] S376. Perform physical rationality verification on the newly generated path sets (rule-driven path set P_rule, cross-mode path set P_cross and special path set P_high for ultra-high dams), and use a combination of mechanical models and expert review to evaluate the physical feasibility of each path, filter out unreasonable paths, and retain reasonable paths to form a verification path set P_valid. Ensure that the automatically generated paths conform to physical laws and avoid invalid paths that violate mechanical principles.

[0146] S377. Design a path scoring function, comprehensively consider physical rationality, historical similarity and risk impact, calculate the importance score for each path in the validation path set P_valid, and construct a path importance ranking table S_path. The scoring function adopts a multi-index weighted summation form, which can identify the key failure paths that pose the greatest threat to high dam safety.

[0147] S378. Based on the path importance ranking table S_path, select the paths with scores exceeding the threshold τ as the core extension paths, merge them with the initial invalid path network Z, delete redundant paths and low-importance paths, optimize the network structure, and form a simplified extension path network Z_opt. This network retains the most valuable extension paths to avoid excessive network complexity that affects computational efficiency.

[0148] S379. Parameter estimation is performed on the simplified and extended path network Z_opt. Based on historical data, physical models and expert knowledge, reasonable transfer probabilities and time distribution parameters are assigned to the newly added paths, the network is quantified, and finally a complete extended failure path network Z' is formed. This network not only includes the failure paths observed in historical cases, but also incorporates potential paths automatically generated based on rules, which greatly expands the coverage of failure path analysis and can cope with complex and changeable failure scenarios, especially rare path mutations. It breaks through the limitations of the preset paths and provides more comprehensive path coverage for high dam safety assessment.

[0149] According to one aspect of the present application, step S38 is specifically:

[0150] S381, read the state transition probability matrix P, state duration distribution matrix D, external factor data matrix F, multi-layer failure mode network M and extended failure path network Z', design a data integration framework, convert data from different sources and formats into a unified model input format, and construct a comprehensive model input data set I.

[0151] S382. Design a macro-scale state sequence simulation module. Based on the state transition probability matrix P and the state duration distribution matrix D, use the Monte Carlo method to simulate the state transition sequence and time process of high dam failure, realize the overall evolution simulation from normal state to final failure state, and generate a macro evolution model M_macro. This model simulates the overall trend of the failure process through random sampling, and can predict the approximate path and time frame of failure.

[0152] S383. Develop a mesoscale network spreading model. Based on the multi-layer failure mode network M and percolation theory, simulate the propagation and spread of failures in the network, consider the interaction and cascade effect between nodes, and construct a mesoscale spreading model M_meso. This model regards the failure process as information or energy propagation in the network, and can capture the spatial distribution characteristics and spreading dynamics of failures.

[0153] S384. Construct a micro-scale physical evolution model, combining the principles of material mechanics, fracture mechanics and fluid mechanics, and based on the environmental conditions in the external factor data matrix F, simulate the physical field evolution process in the local area of ​​the high dam, including stress distribution, crack expansion and seepage changes, to form a micro-physical model M_micro. This model describes the spatiotemporal evolution of the physical field through a set of partial differential equations, and can provide a detailed physical mechanism explanation of the failure process.

[0154] S385. Design a multi-scale model coupling framework, build an information transmission and feedback mechanism between macro, meso and micro scale models, realize bidirectional coupling between upper and lower scales, and form a multi-scale coupling architecture C. For example, the crack extension results of the microscopic physical model can be passed to the mesoscopic network model as a node state update; the failure propagation of the mesoscopic network can be passed to the macroscopic state model as a basis for adjusting the transfer probability. It breaks through the limitations of a single scale model and can provide both overall trends and local details.

[0155] S386. Based on the extended failure path network Z', a dynamic adjustment algorithm for path priority is designed. The activation probability of each failure path is evaluated in real time according to the system state and environmental conditions during the simulation process, the path weight is adjusted dynamically, and the path priority model P_prior is constructed. The algorithm can adaptively adjust the importance of the failure path according to the changes in external conditions and the evolution of the system state, and realize the dynamic optimization of the failure path.

[0156] S387, integrate the macro-evolution model M_macro, the meso-spread model M_meso, the micro-physical model M_micro, the multi-scale coupling architecture C and the path priority model P_prior, build a complete multi-scale temporal evolution framework, and form a comprehensive temporal evolution model M_evol. This model can simulate the time evolution, spatial distribution and physical mechanism of the high dam failure process according to the given initial state and external conditions, and provide a comprehensive dynamic perspective for safety assessment.

[0157] S388, design model solving engine, use hybrid numerical algorithm, including Monte Carlo sampling, Markov chain solving, network dynamics calculation and finite element analysis, realize efficient calculation of comprehensive time evolution model M_evol, generate simulation result data set simulation result set R_sim. The engine selects appropriate solution strategy according to the characteristics of different scale models, optimizes calculation efficiency and result accuracy.

[0158] S389, develop result analysis and visualization modules, process simulation result set R_sim, extract key time series features, identify critical states and early warning indicators, generate multi-dimensional visualization of failure process, and form time series evolution analysis report A_evol. The report includes failure time prediction, path probability distribution, key node identification and risk level assessment, providing intuitive and comprehensive information for decision support. It realizes full-scale, dynamic and quantitative simulation of high dam failure process, providing strong technical support for high dam safety assessment.

[0159] According to one aspect of the present application, step S42 is specifically:

[0160] S421. Collect real-time monitoring data of the high dam to be evaluated, including deformation monitoring, seepage monitoring, stress monitoring and environmental parameter monitoring data, perform data cleaning, outlier processing and standardization, and construct a high dam monitoring data set M. data This dataset reflects the current working status and environmental conditions of the high dam and is the basic input for failure path prediction.

[0161] S422, based on high dam monitoring data set M data , combined with the high dam design parameters and operation history, evaluate the current state of the high dam, identify potential abnormal characteristics and initial signs of disease, locate the specific state node in the state space S, and form the initial state information S i This information contains the current status assessment results of the high dam and the probability distribution of various possible states, reflecting the uncertainty of the initial state.

[0162] S423. Predict the changing trend of the external environment of the high dam to be evaluated in the future, including water level changes, temperature fluctuations, rainfall conditions and possible extreme events (such as floods, earthquakes, etc.), and construct a prediction set of future environmental conditions E f This prediction set takes into account a variety of possible environmental scenarios and provides external condition input for subsequent path simulation.

[0163] S424, read the timing failure path prediction model (i.e., the comprehensive timing evolution model M_evol generated in step S38), and convert the initial state information S i nit and future environmental condition prediction set E f uture as input, perform multi-scenario failure path simulation, generate a large number of possible failure evolution trajectory samples, and construct a failure trajectory sample set T_sample. This sample set contains thousands of simulation trajectories, covering the possible failure development paths of the high dam under different initial states and environmental conditions.

[0164] S425. Perform cluster analysis on the failure trajectory sample set T_sample, identify typical failure evolution patterns, extract representative trajectories and key features of each type of failure mode, and construct a typical failure mode set T_typical. This set compresses a large number of trajectory samples into a small number of representative patterns, which is convenient for subsequent analysis and decision-making.

[0165] S426. For each typical failure mode, calculate the key time characteristics of the failure process, including the expected duration of each state, the time window of state transfer, the time point of the key node and the overall failure time distribution, and form a time characteristic analysis table T. feat This table provides a quantitative description of the failure process in the time dimension and is the core basis for the failure risk timing assessment.

[0166] S427, based on the typical failure mode set T_typical and time characteristic analysis table T feat , evaluate the occurrence probability and risk level of each failure mode, and combine the severity assessment of the failure consequences to construct the risk assessment matrix R. This matrix combines the failure probability and consequence severity to give a comprehensive risk level assessment result.

[0167] S428. Design a timing risk visualization solution to integrate the failure path, time window and risk level information into the same view, use multi-dimensional visualization technology (such as timeline diagram, heat map, Sankey diagram, etc.) to express the timing characteristics and risk distribution of the failure process, and generate a preliminary failure risk timing diagram G_risk.

[0168] S429. The failure risk time series diagram G_risk is optimized and enhanced, with interactive functions, multi-level detailed information and warning threshold marks added to improve readability and practicality, forming the final failure risk time series diagram. This diagram intuitively shows the possible failure evolution path of the high dam to be evaluated, the time window of each stage, the key turning points and the risk level distribution, providing a decision-making basis in the time dimension for safety monitoring and risk management. It breaks through the limitations of traditional static risk assessment, can reveal the dynamic characteristics of the failure process from the time dimension, and provides precise time window guidance for risk warning and intervention decisions.

[0169] According to one aspect of the present application, step S44 is specifically:

[0170] S441. Read potential accident modes, failure risk time series diagrams and verification analysis results, design a multi-source information fusion framework, convert and normalize the evaluation results from different sources according to unified standards, and construct a comprehensive evaluation input set I_comp. This input set integrates the results of similar engineering analogy analysis and time series evolution prediction, providing a rich information basis for comprehensive evaluation.

[0171] S442. Based on the comprehensive evaluation input set I_comp, the analytic hierarchy process (AHP) is used to construct a safety evaluation index system, set index weights, and form an evaluation index system H. The index system includes three categories of indicators: structural safety, operational reliability, and risk controllability. Each category of indicators has multiple secondary and tertiary indicators to form a complete evaluation framework.

[0172] S443. For each indicator in the evaluation index system H, a quantitative score is given based on the relevant information in the comprehensive evaluation input set I_comp. The uncertainty of the score is handled by the fuzzy comprehensive evaluation method, and the fuzzy score vector of each indicator is obtained to construct the indicator score matrix S. This matrix reflects the performance of the high dam in each evaluation indicator and is the basic data for the comprehensive evaluation.

[0173] S444. Combine the weights of the evaluation index system H and the index scoring matrix S to calculate the comprehensive score of the high dam's safety status and the scores of each subsystem, determine the overall safety level and sub-item safety level based on the scoring results, and form the safety level assessment result L. The safety level is usually divided into four levels: normal, caution, alert and dangerous, reflecting the current overall safety status of the high dam.

[0174] S445. Based on the failure risk time series diagram and verification analysis results, identify the most likely failure modes and weak links of the high dam, analyze their physical causes and development trends, and construct a key risk point analysis table K. This table describes in detail the location, nature, severity and development rate of each potential risk point, providing a clear target for targeted intervention.

[0175] S446. For each risk point identified in the key risk point analysis table K, a monitoring enhancement plan is designed, including the layout of monitoring points, adjustment of monitoring frequency, new monitoring items and setting of early warning thresholds, to form a targeted monitoring plan M. This strengthens the monitoring coverage of key risk points and improves the ability to detect abnormalities early.

[0176] S447. Based on the safety level assessment results L and the key risk point analysis table K, combined with the actual conditions of the project and technical and economic feasibility, a hierarchical and phased intervention strategy is designed, including immediate intervention measures, mid-term reinforcement plans and long-term maintenance plans, to construct an intervention strategy matrix I. This matrix provides differentiated intervention plans for different risk levels and problem types, taking into account both the rapid response to emergency risks and the systematic solution to long-term safety.

[0177] S448. Design a dynamic risk monitoring framework, establish a real-time data-driven risk assessment update mechanism based on the time-series failure path prediction model and targeted monitoring plan M, determine the regular assessment cycle and trigger assessment conditions, and form a dynamic risk monitoring plan D. The plan establishes a closed-loop feedback loop between monitoring data and risk assessment, and realizes continuous tracking and timely updating of the safety status of the high dam.

[0178] S449. Integrate the safety level assessment results L, key risk point analysis table K, targeted monitoring plan M, intervention strategy matrix I and dynamic risk monitoring plan D to prepare a structured high dam safety status assessment report.

[0179] Taking an arch dam as the research object, the accident mode and failure path diagnosis are carried out. The arch dam is a 300m-class extra-high arch dam with a relatively short operation time and insufficient monitoring information accumulation. It is necessary to use engineering analogy and failure path analysis for safety assessment. The existing medium-weighted multivariate correspondence analysis and static failure path analysis methods cannot meet its assessment needs.

[0180] Step S1: Construction of high dam accident case database

[0181] Step S11: collect basic information and accident records of 235 high dam accident cases at home and abroad from engineering documents, accident reports, and monitoring records, including dam type, dam height, reservoir capacity, construction year, country, accident time, accident type, accident location, treatment measures, etc., to form original case information. The collected cases include 73 arch dams, 89 gravity dams, and 73 earth-rock dams, with a construction period ranging from 1920 to 2015.

[0182] Step S12: Data cleaning of the original case information is performed to process missing values ​​and outliers. For some missing parameters, such as material parameters, the average values ​​of high dams of the same type are used to supplement them; outliers are identified and processed using the 3σ principle. Unify parameter units, such as unifying dam height to meters and reservoir capacity to 100 million cubic meters. After processing, standardized case data is formed, including 235 complete records.

[0183] Step S13, based on the standardized case data, extract 27 characteristic parameters of each case: structural characteristics, including dam type coefficient (1.01.5), dam height (30m300m), dam crest length, dam bottom width, dam thickness coefficient (0.10.5), reservoir capacity (0.1500 billion m³); material characteristics, including concrete strength grade, elastic modulus, Poisson's ratio, permeability coefficient; environmental characteristics, including geological conditions (divided into 5 levels), climate characteristics (divided into 4 categories), earthquake areas (divided into 4 areas); service characteristics, including service life, historical water level fluctuations, and historical temperature changes; all extracted characteristic parameters form a characteristic parameter matrix X, the matrix dimension is 235×27, each row represents a high dam case, and each column represents a characteristic parameter.

[0184] Step S14, analyze the accidents in the standardized case data and identify three major types of accident modes: crack mode (112 cases): including 5 subtypes such as dam heel cracks, dam surface cracks, and dam top cracks; abnormal deformation mode (78 cases): including 4 subtypes such as uneven settlement, excessive deformation, and top displacement; seepage mode (45 cases): including 6 subtypes such as dam body leakage, dam foundation leakage, and joint leakage; construct an accident mode classification system Y, using a two-level classification structure, the first level is the main accident mode, and the second level is the specific accident type.

[0185] Step S15: Construction of failure path network

[0186] Step S151: extract the accident event sequence of each case from the standardized case data, including the accident phenomenon, occurrence time, development status and impact range, to form the accident event time series data E. For example, the Kolnbrein arch dam records the cracks that appeared at the dam heel when it was first filled with water in 1977, as well as the complete process of subsequent crack expansion and leakage intensification.

[0187] Step S152: Decompose the time series data E of the accident event to identify the starting event, intermediate development event and final result event in each case. Taking the Kolnbrein arch dam as an example, the event chain is identified: initial water storage → stress concentration at the dam heel → cracking at the dam heel → crack expansion → leakage intensification → reduction of structural safety factor. Perform similar analysis on all cases to construct a case-level event chain set L.

[0188] Step S153: semantically normalize all events in the event chain set L, and merge and unify events with different expressions but the same essence. For example, "dam heel crack", "bottom crack", "dam heel tension crack" and so on are unified into "dam heel crack" event. After normalization, a standardized event type set T is obtained, which includes 45 standard event types.

[0189] Step S154: Based on the event type set T, analyze the dependencies between the events in the event chain set L, count the transfer frequencies between the event pairs, and form an event transfer frequency matrix F. For example, the transfer frequency from "dam heel cracking" to "crack extension" is 56 times, and the transfer frequency from "crack extension" to "leakage intensification" is 42 times.

[0190] Step S155: normalize the event transfer frequency matrix F and calculate the conditional probability P(j|i) = f ij / ∑fik; where: P(j|i) is the conditional probability of event j occurring after event i occurs; f ij is the number of historical occurrences from event i to event j; ∑fik represents the total number of occurrences from event i to all possible events k; i, j, k are event indexes. For example, the conditional probability of "dam heel cracking" to "crack extension" P("crack extension" | "dam heel cracking") = 56 / 75 = 0.747. Calculate the conditional probabilities of all event pairs and construct the event transition probability matrix P event .

[0191] Steps S156-S159: Based on the event transition probability matrix P event , the minimum spanning tree algorithm is used to extract the main event transfer paths, and the transfer relations with probability lower than the threshold θ=0.15 are eliminated to obtain the core event transfer network N core Combining high dam engineering mechanics theory and expert knowledge, the network is physically verified for rationality, and the transfer edges that must exist physically but are missing in the data are added to form a physical correction event network N. phy The events are classified in two dimensions according to the risk type and development stage, and the event classification matrix C is constructed. Based on this matrix and the correction network, the final initial failure path network Z is constructed. The network is a directed weighted graph structure, with nodes as event types, edges as event transfer relationships, and edge weights as transfer probabilities.

[0192] Step S16: Integrate the characteristic parameter matrix X, the accident mode classification system Y and the initial failure path network Z into a structured database to form a complete high dam accident case data set. The data set adopts a relational database structure, including a case basic information table, a characteristic parameter table, an accident mode table and a failure path table, and establishes an association relationship through the case ID.

[0193] Implementation step S2: Multi-scale hierarchical adaptive correspondence analysis based on physical mechanism

[0194] Step S21: extract the characteristic parameter matrix X from the high dam accident case data set, and calculate the point bi-column correlation coefficient ρ between each characteristic parameter and the accident mode ij , construct the feature correlation matrix R. Point bi-column correlation coefficient calculation formula ρ ij = (φ ij -φi+φ+j) / √(φi+(1-φi+)φ+j(1-φ+j)); where: ρ ij is the point biserial correlation coefficient between feature j and accident mode i; φ ij is the proportion of cases that simultaneously meet feature j and failure mode i; φi+ is the marginal proportion of failure mode i; φ+j is the marginal proportion of feature j; i is the failure mode index; j is the characteristic parameter index. For example, the correlation coefficient between the dam type coefficient and the failure mode of dam heel cracks is calculated, and ρ11 = 0.78 is obtained, indicating that the two have a strong positive correlation.

[0195] Step S22: Calculation of feature weights

[0196] Steps S221-S222: read the feature correlation matrix R, extract the correlation coefficient between each feature parameter j and all accident patterns, and form a feature-accident correlation coefficient set ρ_set. Calculate the average correlation coefficient ρ_avg of each feature j j = (1 / M)∑ρ ij ; Among them: ρ_avg j is the average correlation coefficient of feature j; ρ ij is the point-wise bi-column correlation coefficient between feature j and accident mode i; M is the total number of accident modes; ∑ represents the summation operation from i=1 to M. For the dam type coefficient (j=1), the average correlation coefficient ρ_avg_1 = 0.76 is calculated; for the dam height ratio (j=2), the average correlation coefficient ρ_avg_2 = 0.82 is calculated; for the elastic modulus ratio (j=3), the average correlation coefficient ρ_avg_3 = 0.65 is calculated. The average correlation coefficients of all features are summarized to form the average correlation coefficient vector ρ_avg.

[0197] Steps S223-S224: Perform a significance test on the values ​​in the average correlation coefficient vector ρ_avg, calculate the p value, screen out features with p>0.05, retain 21 significantly correlated features, and form a significant feature index set J_sig. For the features in the significant feature index set J_sig, introduce 10 experts to perform importance scoring (0-1 points), and calculate the average score as the expert weight coefficient e j , construct the expert weight vector E. For example, the expert weight of dam type coefficient e_1 = 0.85, and the expert weight of dam height ratio e_2 = 0.92.

[0198] Steps S225-S228: Design a fusion function to perform weighted fusion of the data-driven average correlation coefficient vector ρ_avg and the expert weight vector E: Fusion weight calculation formula f j = α·ρ_avg j + (1-α)·e j ; Among them: f j is the fusion weight value of feature j; ρ_avg j is the average correlation coefficient of feature j; e j is the expert weight coefficient of feature j; α is the adaptive fusion coefficient, and its value range is [0,1].

[0199] According to the data quality assessment, α = 0.7 is determined. For the dam type coefficient (j=1), the fusion weight f_1 = 0.7×0.76 + 0.3×0.85 = 0.787 is calculated. The fusion weights of all features are calculated to form a fusion weight value vector F.

[0200] Sensitivity analysis and stability adjustment are performed, and finally the stability adjustment weight vector G is normalized to obtain the initial characteristic weight vector W0, where the dam type coefficient weight w_1 = 0.092, the dam height ratio weight w_2 = 0.105, and the elastic modulus ratio weight w_3 = 0.078.

[0201] Step S23: Layered weighting of risk patterns

[0202] According to the accident mode classification system Y, the 235 high dam cases are divided into three categories: crack mode (C1, 112 cases), abnormal deformation mode (C2, 78 cases) and seepage mode (C3, 45 cases), and the characteristic parameter submatrices X specific to each mode are constructed. i For each feature parameter submatrix X i , calculate the correlation coefficient between features and sub-patterns, and construct the pattern-specific feature correlation sub-matrix R i For each feature correlation submatrix R i, apply the feature weight calculation method of step S22 to obtain the feature weight vector W for the specific risk mode i i .

[0203] For example, the characteristic weight vector W_1 of the fracture mode (i=1) focuses on stress-related features, such as dam type coefficient (w_11 = 0.126), dam height ratio (w_12 = 0.135), and bending radius (w_13 = 0.114). The characteristic weight vector W_3 of the seepage mode (i=3) focuses on permeability-related features, such as material permeability coefficient (w_31 = 0.152), anti-seepage measures (w_32 = 0.138), and geological conditions (w_33 = 0.127).

[0204] Analyze the sample characteristics of each pattern subset and calculate the sample size n i , sample coverage c i and sample diversity index d i , construct the pattern sample characteristic matrix D. Based on the pattern sample characteristic matrix D, evaluate the reliability coefficient r of each pattern weight vector i , construct the mode weight reliability vector R_mode. For mode j whose reliability is lower than the threshold λ=0.6, such as the percolation mode (r_3= 0.58), a transfer learning strategy is used to reinforce the weight W j ' = W j + γ·∑(sim(j,k)·(W_k - W j )); where: W j ' is the weight vector after reinforcement; W j is the original weight vector; sim(j,k) is the similarity between modes j and k; γ is the migration strength coefficient; ∑ represents the sum of all similar modes k. For the seepage mode, γ = 0.4 is used to migrate the weight information from the crack mode and deformation mode to obtain the reinforcement weight vector W_3'. The characteristic weight vectors of all modes are combined into an m×n dimensional matrix to form a complete mode characteristic weight matrix WM. Each row of the matrix represents a risk mode, each column represents a characteristic parameter, and the matrix element WM ij Represents the weight value of feature j under accident mode i. Calculate the coefficient of variation CV of each feature weight between different modes, identify mode sensitive features (large CV value) and mode stable features (small CV value), and construct the feature mode sensitivity vector S feature .

[0205] Step S24: Constructing the risk pattern correlation matrix

[0206] The relationship between different accident modes is analyzed, the conversion probability and co-occurrence frequency between modes are calculated, and the accident mode correlation matrix R is constructed. For example, the correlation coefficient between the fracture mode and the seepage mode is r13 = 0.65, indicating that the two modes have a strong correlation.

[0207] Step S25: Construction of physical similarity index system

[0208] The physical mechanism of high dam failure is analyzed, and four key physical quantities are selected: stress field (σ), deformation field (ε), seepage field (Φ) and damage evolution (D), and a key physical quantity set K is constructed. For each physical quantity, a similarity measurement function is designed to form a physical similarity measurement function set Φ. For example, the stress distribution similarity function φ σ , deformation pattern similarity function φ ε wait.

[0209] Identify the key characteristic parameter combinations that affect the distribution of each physical quantity, and establish a mapping relationship set between characteristic parameters and physical quantities {M σ , M ε , M f , M_d}. For example:

[0210] Stress Mapping σ :Mainly consider parameters such as dam type, dam height, thickness-to-height ratio, and bending radius;

[0211] Percolation Mapping f : Mainly consider parameters such as permeability coefficient, anti-seepage facilities, and geological conditions;

[0212] According to the similarity of stress distribution (σ similarity), the calculation function P is designed σ (i,j) = exp(-||Γσ(X σi ) - Γσ(X σj )|| 2 / 2θσ 2 );where: P σ (i, j) is the stress distribution similarity between high dams i and j; X σi is the stress-related feature subset of high dam i; Γσ is the stress field feature mapping function; ||·|| represents the Euclidean distance; θσ is the scaling parameter.

[0213] For an arch dam and Kolnbrein arch dam, stress-related features are extracted: σ _1 = [1.2, 294,0.32, 456]; Kolnbrein Arch Dam X σ_2 = [1.15, 200, 0.28, 425]; respectively represent [dam type coefficient, dam height (m), thickness-to-height ratio, bending radius (m)]. Apply the stress field characteristic mapping function Γσ transformation to obtain the stress field characteristic vector and calculate the stress similarity P σ (1,2) = 0.875. Similarly, the deformation similarity P is calculated. ε (1,2) = 0.823, seepage similarity P f (1,2) = 0.762, damage similarity P_d(1,2) = 0.798. Combining the physical similarity indices, a physical similarity index vector P_12 = [0.875, 0.823, 0.762, 0.798] is formed. By calculating the physical similarity indices for all high dam pairs, a complete physical similarity index matrix P is constructed.

[0214] Analyze the correlation between the indicators in the physical similarity indicator matrix P, identify redundant indicators and complementary indicators, and optimize the physical similarity indicator system. Finally, determine the mapping relationship F between physical indicators and basic characteristic parameters: characteristic parameter matrix X → physical similarity indicator matrix P.

[0215] Steps S26 to S28, take the characteristic parameter matrix X and the mapping relationship F, and calculate the original physical similarity index matrix P. Analyze the structural differences between a certain arch dam as a modern ultra-high arch dam and historical cases, identify three key difference features: ultra-high dam body type (d_1), low-heat cement material (d_2) and intelligent temperature control construction (d_3), and construct a structural difference feature set D. Evaluate the impact of each difference feature on each physical similarity index, and construct a difference-impact mapping table M_diff. For example, the influence coefficient of the ultra-high dam body type (d_1) on stress similarity is c_11 = 1.15, and the influence coefficient on deformation similarity is c_12 = 1.23.

[0216] Design structural difference compensation function P'_k(i,j) = P_k(i,j)·∏(1+δ_km·d_m); where: P'_k(i,j) is the corrected physical similarity index k; P_k(i,j) is the original physical similarity index k; δ_km is the influence coefficient of difference feature m on physical index k; d_m is the intensity value of difference feature m; ∏ represents the multiplication operation.

[0217] Compensation for the stress similarity between a certain arch dam and the Kolnbrein arch dam: the original stress similarity P σ (1,2) = 0.875; the difference in the shape of the super-high dam d_1 = 0.45; the influence coefficient δ σ 1 = 0.15; Modified stress similarity P' σ(1,2) =0.875×(1+0.15×0.45) = 0.934; Similarly, the correction values ​​of other physical indicators are calculated to obtain the modified physical similarity index vector P'_12 = [0.934, 0.893, 0.762, 0.846]. By performing similar compensation on all high dam pairs, the modified physical similarity index matrix P' is obtained.

[0218] 27 characteristic parameters of an arch dam are extracted to construct the characteristic vector X0 of the high dam to be evaluated. The physical similarity index vector P0 of the high dam to be evaluated is obtained by applying the physical mapping relationship F and the compensation function H. In the original characteristic space, the weighted Euclidean distance between an arch dam and the historical case is calculated, DX(0,j) = √(∑(wi·(X0i-Xji) 2 )); where: DX(0,j) is the distance between the high dam 0 to be evaluated and the historical case j in the original feature space; wi is the weight coefficient of feature i; X0i is the feature i value of the high dam to be evaluated; Xji is the feature i value of the historical case j; ∑ represents the sum of all features i. For example, the feature space distance DX(0,2) = 0.263 is calculated for a certain arch dam and the Kolnbrein arch dam.

[0219] In the physical similarity space, the weighted Euclidean distance between the high dam to be evaluated and the historical case is calculated, that is, the physical space distance calculation formula DP(0,j) = √(∑(vk·(P0k-P'jk) 2 )); where: DP(0,j) is the distance between the high dam 0 to be evaluated and the historical case j in the physical similarity space; vk is the weight coefficient of the physical similarity index k; P0k is the physical similarity index k value of the high dam to be evaluated; P'jk is the modified physical similarity index k value of the historical case j; ∑ represents the sum of all physical indicators k. Calculate the physical space distance DP(0,2) = 0.118 between a certain arch dam and the Kolnbrein arch dam.

[0220] Step S29: Multi-scale fusion strategy

[0221] Read the original feature space distance DX and the physical similarity space distance DP, and construct the dual space distance vector V dist =[DX, DP].

[0222] Design a distance-similarity conversion function to convert distance into similarity SX = exp(-DX 2 / 2σX 2 ), SP = exp(-DP 2 / 2σP 2), where: SX is the similarity in the original feature space; SP is the similarity in the physical similarity space; DX is the distance in the original feature space; DP is the distance in the physical similarity space; σX and σP are scaling parameters.

[0223] For a certain arch dam and Kolnbrein arch dam, σX = 0.5, σP = 0.3, the feature space similarity SX = exp(-0.263 2 / 0.5) = 0.78; physical space similarity SP = exp(-0.118 2 / 0.3) = 0.89; evaluate the data quality of the two spaces, determine the original feature space credibility QX = 0.65, the physical space credibility QP = 0.82, and calculate the adaptive weight factor α = QX / (QX+QP); where: α is the fusion weight factor; QX is the credibility evaluation index of the original feature space; QP is the credibility evaluation index of the physical similarity space. Calculated α = 0.65 / (0.65+0.82) = 0.44.

[0224] According to the risk probability vector p = [0.65, 0.25, 0.10] (representing the risk probability of cracks, deformation, and seepage, respectively) of a certain arch dam and the mode space weight adjustment matrix B, the weight factor α after comprehensive adjustment is calculated. adj = 0.42.

[0225] Calculate the multi-scale fusion similarity S(i,j) = α adj ·SX(i,j) + (1-α adj )·SP(i,j); where: S(i,j) is the final similarity between dams i and j; SX(i,j) is the similarity in the original feature space; SP(i,j) is the similarity in the physical similarity space; α adj is the adaptive weight factor after comprehensive adjustment.

[0226] For a certain arch dam and the Kolnbrein arch dam, the final similarity S(1,2) = 0.42×0.78 + 0.58×0.89 = 0.843 is calculated.

[0227] To meet the needs of multi-objective optimization, calculate the multi-dimensional similarity vector S multi = [S struct , S mat , S env ], describing the similarity relationship in detail from three dimensions: structural similarity, material similarity, and environmental similarity.

[0228] Based on the final similarity, the historical cases were sorted and the top five most similar projects were obtained: Kolnbrein arch dam (0.843), Mica arch dam (0.812), Ertan arch dam (0.795), Hoover arch dam (0.766), and Mauvoisin arch dam (0.754), forming a weighted ranking list of similar projects.

[0229] Implementing step S3, time-adaptive network evolution with multiple failure mode coupling

[0230] Step S31: extract the initial failure path network Z from the high dam accident case data set, and define the state space S ={S0, S1, S2, ..., Sf}, where S0 represents the normal state, S1 represents the microcracks at the dam heel, S2 represents the crack extension, S3 represents the abnormal seepage, and Sf represents the failure state. A total of 12 key state nodes are identified to form the complete state space S.

[0231] Step S32: Based on the historical cases in the high dam accident case data set, the transition frequencies between the states are counted, and the state transition probability matrix P is calculated. For example, P(S1|S0) = 0.08 (the probability of the normal state turning into microcracks at the dam heel), P(S2|S1) = 0.32 (the probability of the microcracks at the dam heel developing into crack extension).

[0232] Step S33: Time distribution function fitting

[0233] We screened cases with complete time records from the high dam accident case dataset to form a time series complete case set T_cases, which contains 87 cases. For each case, we identified the state sequence and transition process it experienced, extracted the transition time data of each pair of adjacent states Si→Sj, and constructed the state transition time dataset T data .

[0234] For the state transition time dataset T data Perform data cleaning and process outliers. Use the box plot method to identify abnormal time records, replace the values ​​that exceed the upper and lower limits with the critical values, and obtain the cleaned time data set T clean .

[0235] For each pair of state transitions, extract its time sample and analyze the data distribution characteristics. Use the Shapiro-Wilk test to determine the distribution type and construct the distribution type mapping table M dist For example, the time sample of S1→S2 (the development of microcracks at the dam heel into crack expansion) is found to be consistent with the log-normal distribution; the time sample of S2→S3 (the development of crack expansion into seepage anomaly) is consistent with the Weibull distribution. Parameters of each pair of state transition time samples are estimated, and the distribution parameters are fitted using the maximum likelihood estimation method to form a distribution parameter table P distFor example, the lognormal distribution parameters for S1→S2 are: μ12 = 1.8, σ12 2 = 0.5; Weibull distribution parameters of S2→S3: shape parameter k23 = 2.1, scale parameter λ23 = 2.8. For state transitions with insufficient sample size (such as S3→Sf), the parameter transfer learning method is used to learn parameter information from similar state transitions and estimate the parameters: μ3f = 2.3, σ3f 2 = 0.7. Based on the physical failure mechanism, the physical rationality check and correction of the statistically estimated distribution parameters are carried out to ensure that the time distribution conforms to the physical laws, and the physical correction parameter set P is obtained. phy .

[0236] For each state transition process, based on its distribution type and parameters, calculate the key time characteristic quantity: mean μ ij (expected transfer time); variance σ ij 2 (Time volatility); median m ij (typical transfer time); quantile q ij ,α (reliability interval); forming the time feature matrix T feat , comprehensively describing the time characteristics of each state transition process. Integrating the time distribution functions and parameters of all state transition processes, constructing a complete state duration distribution matrix D. Each element D of this matrix ij The time distribution function T containing the state transition Si→Sj ij (t) and its parameters.

[0237] Step S34: External factor data matrix construction

[0238] The main external factors affecting the high dam failure process are analyzed, historical data of water level changes, temperature fluctuations, load history and other factors are collected, and the external factor data matrix F is constructed. This matrix contains the time series data of each external factor and provides a basis for conditional state transition probability modeling.

[0239] Step S35: Establishment of conditional state transition probability function

[0240] Read the external factor data matrix F, the state transition probability matrix P and the state duration distribution matrix D, determine the key state transition process that needs to establish a conditional relationship, and form the key transition process set K trans The importance of external factors is analyzed to evaluate the influence of each factor on state transfer. Water level change (F1), temperature change (F2) and load history (F3) are selected as key factors to construct the key external factor set F key .

[0241] Design a data stratification strategy, divide the factor space into multiple intervals, form a multidimensional grid, and construct a factor stratification framework G. Count the state transition frequency under each factor combination area, calculate the conditional transition probability, and construct a discrete conditional probability table CP discrete For areas with insufficient sampling points, a conditional probability interpolation function is designed, and a radial basis function network is used to achieve continuous estimation, and a continuous conditional probability model CP is constructed. continuous . Analyze the interaction between external factors, design interaction modeling strategies, capture the nonlinear effects of multiple factors, and form an interaction effect model I model .

[0242] Integrate the above models to construct a complete conditional state transition probability function P ij (F1,F2,F3) = P ij ·(1+β1·(F1-F1_ref) / F1_max + β2·(F2-F2_ref) / F2_max + β3·(F3-F3_ref) / F3_max + β12·F1·F2 / F1_max / F2_max + β13·F1·F3 / F1_max / F3_max + β23·F2·F3 / F2_max / F3_max); where: P ij (F1, F2, F3) is the conditional transition probability from state i to state j under the influence of external factors F1, F2, F3; P ij is the benchmark transfer probability; β1, β2, β3 are the first-order influence coefficients; β12, β13, β23 are the interaction influence coefficients; F1_ref, F2_ref, F3_ref are the reference values; F1_max, F2_max, F3_max are the normalization coefficients.

[0243] For the S1→S2 transfer process of a certain arch dam (microcracks in the dam heel develop into crack expansion), the relevant parameters are: baseline probability P12 = 0.32; water level influence coefficient β1 = 0.65; temperature influence coefficient β2 = 0.43; load influence coefficient β3 = 0.28; water-temperature interaction coefficient β12 = 0.15; when the water level ratio (F1-F1_ref) / F1_max = 0.8 and the temperature deviation (F2-F2_ref) / F2_max = 0.5, the conditional transfer probability is calculated: P12(0.8,0.5,0) = 0.32×(1+0.65×0.8+0.43×0.5+0.15×0.8×0.5) = 0.32×1.698 = 0.543.

[0244] Similarly, the mapping relationship between time distribution parameters and external factors is established, and the formula is: time parameter mapping function μ ij(F1,F2,F3) = μ ij ·(1-γ1·(F1-F1_ref) / F1_max - γ2·(F2-F2_ref) / F2_max - γ3·(F3-F3_ref) / F3_max - γ12·F1·F2 / F1_max / F2_max); where: μ ij (F1, F2, F3) is the expected transition time from state i to state j under the influence of external factors F1, F2, F3; μ ij is the benchmark expected time; γ1, γ2, γ3 are the first-order influence coefficients; γ12 is the interaction influence coefficient; other parameters are the same as above.

[0245] For the S1→S2 transfer process, the parameters are: benchmark expected time μ12 = 1.8 years; water level influence coefficient γ1 = 0.45; temperature influence coefficient γ2 = 0.32; water-temperature interaction coefficient γ12 = 0.12; when the water level ratio is 0.8 and the temperature deviation is 0.5, the conditional expected time is calculated: μ12(0.8,0.5) = 1.8×(1-0.45×0.8-0.32×0.5-0.12×0.8×0.5) = 1.8×0.588 = 1.06 years.

[0246] Verify the model performance and form the final non-stationary Markov model parameter set Θ, which fully describes the change of state transfer characteristics with external factors.

[0247] Step S36: Multi-layer failure mode network construction

[0248] The main failure modes, namely crack mode (CM), deformation mode (DM) and seepage mode (FM), are extracted from the high dam accident case data set and the accident mode classification system Y, and the failure mode type set MODE is constructed.

[0249] For each failure mode, relevant state nodes and transfer relationships are extracted from the initial failure path network Z to construct a mode subnetwork to form a mode subnetwork set Z sub .

[0250] Optimize the structure of each sub-network, supplement possible state nodes and transition relationships, and form a pattern-enhanced sub-network Z'm.

[0251] Define network parameters for each sub-network, including node attributes and edge attributes, and construct a model network parameter set Pm.

[0252] All enhanced sub-networks are taken as independent layers to construct the initial multi-layer failure mode network M0 to achieve separate representation of different types of failure modes.

[0253] Analyze the physical interaction mechanism between failure modes, identify the key inter-mode influence paths, and construct the mode coupling path set L. For example, the impact of crack development on seepage (CM→FM), and the feedback of seepage anomaly on crack development (FM→CM).

[0254] For each coupling path, the influence strength between modes is evaluated and the failure mode coupling strength matrix C is constructed. ij = η·φ ij ·(1+δ ij ·S ij );Among them: C ij is the coupling strength of failure mode i to mode j; η is the global coupling coefficient; φ ij is the physical mechanism influence coefficient; δ ij is the historical data adjustment coefficient; S ij Assign weights to the expert scores.

[0255] For the influence of fracture mode on seepage mode, the parameters are: physical mechanism influence coefficient φCF = 0.75; historical data adjustment coefficient δCF = 0.6; expert scoring weight SCF = 0.8; global coupling coefficient η = 0.5; calculated coupling strength CCF = 0.5×0.75×(1+0.6×0.8) = 0.45. For the influence of seepage mode on fracture mode, the calculated coupling strength CFC = 0.38.

[0256] Design coupled dynamic equations to describe the modified transfer probability after considering mode coupling, where the coupled dynamic equation P' ij =P ij + ∑(Ckm·f(Sk,Sm)); where: P' ij is the modified transition probability after considering mode coupling; P ij is the original transfer probability; Ckm is the coupling strength of failure mode k to mode m; f(Sk,Sm) is the state interaction function; ∑ represents the sum of all relevant mode pairs (k,m). For the S1→S2 transfer (microcracks at the dam heel develop into crack extension) in a crack network of an arch dam, the original transfer probability P12 = 0.32, considering the coupling effect of seepage mode on crack mode CFC·f(SF,SC) = 0.38×0.25 = 0.095, the corrected transfer probability P'12 = 0.32 + 0.095 = 0.415 is calculated.

[0257] The coupling relationship is integrated into the multi-layer network to form a complete multi-layer failure mode network M, which describes the internal evolution law of each failure mode and the interaction relationship between the modes.

[0258] Step S37: Adaptive path generation

[0259] Read the multi-layer failure mode network M, the failure mode coupling strength matrix C and the initial failure path network Z, analyze the network characteristics, and construct the network feature statistics set E. Design a path generation rule base, which contains 30 specific rules to form a path generation rule set R. For example, the rule "If a dam heel crack appears and the water level change rate exceeds 0.5, it may develop into a crown crack". Based on the rule set, perform forward reasoning, iteratively generate possible state transition sequences from the initial state, and construct a rule-driven path set P_rule, which contains 45 newly generated possible paths. Design a cross-layer path generation algorithm, jump between mode layers, generate a composite path containing multiple failure mode interactions, and form a cross-mode path set P_cross, which contains 28 cross-mode paths. For example, the path of "dam heel crack → seepage anomaly → crown crack" describes the complex failure process of cracks inducing seepage and then inducing new cracks. According to the characteristics of a certain ultra-high arch dam, generate a special failure path, consider special factors such as temperature control and the joint action of arch and beam, and form a special path set P_high for ultra-high dams, which contains 12 special paths. The physical rationality of the newly generated path set is verified by using a method combining finite element model and expert review to evaluate the physical feasibility of each path. Reasonable paths are retained to form the verification path set P_valid, which contains 63 valid paths in total.

[0260] The path scoring function is designed, and the physical rationality, historical similarity and risk impact are comprehensively considered. The importance score of each path in the verification path set P_valid is calculated, and the path importance ranking table S_path is constructed. The paths with scores exceeding the threshold τ=0.65 are selected as core extension paths, which are merged with the initial failure path network Z to form a simplified extension path network Z_opt, which contains a total of 75 original and newly added core paths. Parameters of the simplified extension path network Z_opt are estimated, and the transfer probability and time distribution parameters are assigned to the newly added paths. The network is quantified to form a complete extended failure path network Z'.

[0261] Step S38: Construction of multi-scale temporal evolution model

[0262] Integrate various types of data to construct a comprehensive model input data set I as a unified input for the multi-scale model. Design a macro-scale state sequence simulation module, use the Monte Carlo method to simulate the state transition sequence and time process of high dam failure, and generate a macro-evolution model M_macro. Develop a meso-scale network propagation model, based on the multi-layer failure mode network M and permeation theory, simulate the propagation and spread of failure in the network, and construct a meso-propagation model M_meso. Construct a micro-scale physical evolution model, combine the principles of material mechanics, fracture mechanics and fluid mechanics, simulate the physical field evolution process of the local area of ​​the high dam, and form a micro-physical model M_micro.

[0263] Design a multi-scale model coupling framework, build an information transmission and feedback mechanism between the macro, meso and micro scale models, realize bidirectional coupling between upper and lower scales, and form a multi-scale coupling architecture C. Based on the extended failure path network Z', design a dynamic adjustment algorithm for path priority, evaluate the activation probability of each failure path in real time according to the system state and environmental conditions in the simulation process, and build a path priority model P_prior. Integrate each sub-model and framework to build a complete multi-scale temporal evolution framework to form a comprehensive temporal evolution model M_evol. This model can simulate the time evolution, spatial distribution and physical mechanism of the high dam failure process according to the given initial state and external conditions. Design a model solving engine, adopt a hybrid numerical algorithm to achieve efficient calculation of the model, and generate a simulation result data set simulation result set R_sim. Develop a result analysis and visualization module to extract key temporal features, identify critical states and early warning indicators, and generate a temporal evolution analysis report A_evol.

[0264] Implement step S4: Engineering analogy analysis and safety assessment

[0265] Step S41: Select the five projects with the highest similarity from the weighted similar project ranking list as the core analog project set, among which the Kolnbrein arch dam (similarity 0.843) is the most core analog object. Analyze the historical accidents of the Kolnbrein arch dam: the dam heel cracked when it was first filled with water in 1977, mainly located at the contact surface of the arch dam foundation, caused by temperature stress and water pressure stress concentration. Based on the analogy analysis, the potential accident mode of a certain arch dam is predicted to be "crack risk caused by stress concentration in the dam heel area".

[0266] Step S42: collect real-time monitoring data of an arch dam, including deformation monitoring, seepage monitoring, stress monitoring, etc., and construct a high dam monitoring data set M data . Evaluate the current state of the high dam, identify potential abnormal features, locate the specific state node, and determine the initial state information S i nit. At present, an arch dam is in the normal state S0, but the stress level in the dam heel area has reached 85% of the design value, and there is a potential risk of developing into S1 (microcracks in the dam heel). Predict the trend of future environmental conditions, including water level changes, temperature fluctuations, etc., and construct a future environmental condition prediction set E f uture, including normal working conditions, extreme working conditions and other scenarios.

[0267] The initial state information S i nit and future environmental condition prediction set E f uture inputs the timing failure path prediction model, performs simulation calculations, generates 1000 possible failure evolution trajectories, and constructs the failure trajectory sample set T_sample.

[0268] Cluster analysis was performed on the trajectory samples to identify three typical failure modes: dam heel crack-dominated type (probability 65%); dam body deformation-dominated type (probability 25%); and seepage anomaly-dominated type (probability 10%).

[0269] Extract the key time characteristics of each mode in the typical failure mode set T_typical and construct the time characteristic analysis table T feat For the most likely failure mode dominated by cracks at the dam heel, the key time characteristics are: from the initial state to microcracks at the dam heel (S0→S1): expected time 4.2 years, 90% confidence interval [2.8 years, 6.1 years]; from microcracks to obvious cracks (S1→S2): expected time 1.6 years, 90% confidence interval [0.9 years, 2.5 years]; from obvious cracks to abnormal seepage (S2→S3): expected time 2.3 years, 90% confidence interval [1.5 years, 3.4 years];

[0270] Assess the risk level of each failure mode, construct the risk assessment matrix R, combine the failure probability and consequence severity, and give the risk level assessment result. Design a time series risk visualization solution, integrate the failure path, time window and risk level information, and generate a preliminary failure risk time series diagram G_risk. Optimize and enhance the time series diagram, add interactive functions, multi-level detailed information and warning threshold marks, and form the final failure risk time series diagram. The diagram intuitively shows the possible failure evolution path of a certain arch dam, the time window of each stage and the risk level distribution, and marks three key warning points: dam heel stress over-limit warning point (time T1); crack initialization warning point (time T2); abnormal seepage warning point (time T3);

[0271] Step S43: For typical projects in the core analogy project set, combined with the characteristics of a certain arch dam, a finite element numerical model is constructed to conduct a failure mechanism analysis, verify the rationality of the potential failure mode and failure risk time series diagram, and form a verification analysis result. The verification analysis shows that under extreme water level fluctuation conditions, stress concentration may indeed occur in the heel area of ​​a certain arch dam, which is basically consistent with the failure mechanism of the analogy project Kolnbrein arch dam, confirming the reliability of the engineering analogy results.

[0272] Step S44, read the potential accident mode, failure risk sequence diagram and verification analysis results, design a multi-source information fusion framework, and construct a comprehensive evaluation input set I_comp. Use the hierarchical analysis method to construct a safety assessment index system, set the index weights, and form an evaluation index system H. The index system includes: structural safety (weight 0.4): stress state, deformation control, seismic resistance, etc.; operational reliability (weight 0.35): monitoring system, operation management, response speed, etc.; risk controllability (weight 0.25): early warning mechanism, emergency measures, backup plans, etc.; Quantitatively score each indicator, use the fuzzy comprehensive evaluation method to deal with the uncertainty of the score, and construct the index scoring matrix S.

[0273] The comprehensive score and the score of each subsystem were calculated: structural safety score: 85 points (safety level A); operational reliability score: 78 points (safety level B); risk controllability score: 82 points (safety level A); comprehensive safety level: A (good); the safety level assessment result L was determined to be A (good), indicating that the overall safety status of a certain arch dam is currently good, but potential problems in operational reliability still need to be paid attention to.

[0274] Based on the failure risk time series diagram and verification analysis results, key risk points are identified and key risk point analysis table K is constructed. The main risk points include: stress concentration points in the dam heel area (risk coefficient 0.78); dam body temperature control area (risk coefficient 0.65); grouting curtain leakage area (risk coefficient 0.52);

[0275] A monitoring enhancement plan is designed for key risk points, including: adding 5 stress monitoring points in the dam heel area; adjusting the existing strain monitoring frequency from quarterly to monthly; adding an acoustic wave detection system in potential crack areas; and forming a targeted monitoring plan M to strengthen monitoring coverage of high-risk areas.

[0276] Design hierarchical and phased intervention strategies and construct intervention strategy matrix I: Short-term strategy: regular ultrasonic testing once a quarter; Medium-term strategy: optimize water level control scheme to avoid drastic water level changes; Long-term strategy: refer to the reinforcement experience of Kolnbrein arch dam and reserve the design scheme of the support structure behind the dam;

[0277] Design a dynamic risk monitoring framework, establish a data-driven risk assessment update mechanism, determine a regular assessment cycle every six months, and form a dynamic risk monitoring plan D. Integrate all assessment results and suggestions, and compile a structured high dam safety status assessment report to provide a scientific basis for the safe operation and management of a certain arch dam.

[0278] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A method for diagnosing structural service performance based on high dam failure modes and failure paths, characterized in that: The following steps are involved: Collect high dam accident case information, build a high dam accident case database, and extract characteristic parameters, accident modes and failure paths to form a high dam accident case data set; Based on the high dam accident case data set, multi-scale hierarchical adaptive correspondence analysis is applied to generate a weighted ranking list of similar projects; Based on the high dam accident case data set, a time-series adaptive network model with multiple failure mode coupling is constructed, and a time-series failure path prediction model is output; Combined with the weighted similar engineering ranking list and the sequential failure path prediction model, an engineering analogy analysis is conducted on the high dam to be evaluated and a service safety diagnosis report is generated.

2. The method according to claim 1, characterized in that The steps to generate a ranked list of weighted similar projects include: Extract the feature parameter matrix from the high dam accident case data set, calculate the feature correlation matrix, and calculate the feature weight vector based on it; According to the classification system of accident patterns, the high dam accident case data set is divided into at least two pattern subsets, and their feature weights are calculated one by one to construct the pattern feature weight matrix; Construct a physical similarity index system, establish the mapping relationship between characteristic parameters and physical similarity indexes, calculate the physical similarity index matrix, and obtain the modified physical similarity index matrix through structural difference compensation; Combining the outputs of the above three processes, the similarities between high dams are calculated and a weighted ranking list of similar projects is generated.

3. The method according to claim 2, characterized in that The steps to calculate the feature weight vector include: Calculate the average correlation coefficient between each feature and the accident pattern to form an average correlation coefficient vector; Perform a significance test on the average correlation coefficient vector, screen the significant correlation features to form a significant feature index set, and obtain the expert weight vector through expert scoring; Perform weighted fusion of the average correlation coefficient vector and the expert weight vector to form a fusion weight value vector; The fusion weight value vector is stabilized and normalized to obtain a feature weight vector.

4. The method according to claim 2, characterized in that The steps to construct the pattern feature weight matrix include: According to the accident pattern classification system, the high dam accident case data set is divided into multiple pattern subsets, and the characteristic parameter sub-matrix is ​​extracted for each pattern subset; Calculate the correlation coefficient between the feature and the pattern in each feature parameter submatrix, construct the feature correlation submatrix, and calculate the pattern feature weight vector; Analyze the sample distribution characteristics of each pattern subset, calculate the sample size, coverage and diversity index, and construct the pattern sample characteristic matrix; The feature weight vectors of all patterns are combined into a pattern feature weight matrix.

5. The method according to claim 2, characterized in that The calculation process of the modified physical similarity index matrix includes: Combined with the physical mechanism of high dam failure, key physical quantities are determined and similarity measurement functions are constructed to establish the mapping relationship between characteristic parameters and physical quantities; According to the mapping relationship, the physical similarity index value is calculated from the characteristic parameter matrix and the physical similarity index matrix is ​​constructed; Identify the structural difference characteristics between modern ultra-high dams and historical cases, construct a set of structural difference characteristics, evaluate the impact of each difference characteristic on the physical similarity index, and form a structural difference compensation function; According to the structural difference compensation function, the physical similarity index matrix is ​​compensated and corrected to obtain the corrected physical similarity index matrix.

6. The method according to claim 2, characterized in that The steps to generate a ranked list of weighted similar projects include: Calculate the distance of the high dam in the original feature space and the physical similarity space respectively, and construct the dual-space distance vector; Convert the distance into similarity to obtain the original feature space similarity and physical similarity space similarity; evaluate the data quality and credibility of each space one by one, and combine the risk probability of the accident mode to obtain the comprehensively adjusted fusion weight factor; Based on the fusion weight factor, the original feature space similarity and the physical similarity space similarity are weightedly fused to calculate the multi-scale fusion similarity; Sort historical cases according to multi-scale fusion similarity and generate a weighted ranking list of similar projects.

7. The method according to claim 1, characterized in that The steps to form a timing failure path prediction model include: Extract failure path information from high dam accident case data set, and construct state space and state transition probability matrix; Fit the time distribution function to the state transition process and construct the state duration distribution matrix; Analyze the impact of external factors on state transition and construct conditional state transition probability function; Construct different failure modes into multi-layer networks, extract the coupling relationship between failure modes, and generate an extended failure path network; Combining the outputs of the above steps, a multi-scale timing evolution model is constructed, and a timing failure path prediction model is output through simulation calculation.

8. The method according to claim 7, characterized in that The steps to construct the state duration distribution matrix include: Select cases with complete time records from the high dam accident case data set and extract state transition time data; Clean the time data, analyze its distribution characteristics, determine the theoretical distribution type, and fit the distribution parameters; Based on the physical failure mechanism, the distribution parameters are rationally corrected and the expected transfer time, variance and reliability interval are calculated; Integrate the time distribution functions and parameters of all state transition processes to construct the state duration distribution matrix.

9. The method according to claim 7, characterized in that The steps to construct the conditional state transition probability function include: Determine the key state transition process that requires the establishment of conditional relationships, evaluate the impact of each external factor, and select key external factors; Design a factor space hierarchical framework based on key external factors, count the state transition frequencies under each factor combination area, and construct a discrete conditional probability table; Design interpolation functions for areas with insufficient sampling points, analyze the interactions between factors, and form a continuous conditional probability model; Integrate discrete conditional probability tables and continuous conditional probability models to construct conditional state transition probability functions.

10. The method according to claim 7, characterized in that The steps to build coupling relationships between failure modes include: Extract the main failure mode types from the high dam accident case data set and construct a failure mode type set; For each failure mode, extract relevant state nodes and transfer relationships, build and optimize the mode sub-network; Define node attributes and edge attributes for each sub-network, and build a multi-layer network with all sub-networks as independent layers; The physical interaction mechanism between failure modes is analyzed, the influence paths between modes are identified, a set of mode coupling paths is constructed and the influence intensity between modes is evaluated, a failure mode coupling intensity matrix is ​​constructed, the coupling relationship is obtained, and a multi-layer failure mode network is formed.

Images