Hydraulic structure safety monitoring data abnormity self-diagnosis method and system

Abstract

The invention relates to a hydraulic structure safety monitoring data abnormity self-diagnosis method and system. The method comprises the following steps: arranging a plurality of variables of monitoring data to form a multivariable time sequence; a dynamic time warping algorithm is adopted to analyze the correlation between the effect variables in the multivariable time sequence and the respective variables; on the basis of correlation analysis and time lag effect analysis results, constructing a multiple regression model between the effect quantity and the independent variable; and performing anomaly identification on the effect quantity and the independent variable measurement value by using a multivariable local anomaly factor, comparing the distribution characteristics and the occurrence time of the multivariable local anomaly factor from different time scales based on a local anomaly factor threshold, judging whether the effect quantity and the independent variable local anomaly factor exceed the threshold at the same time, and if so, executing the next step. And if yes, determining security state abnormity caused by multiple factors, and identifying to obtain a structural state conversion mechanism. According to the method, the accuracy and reliability of abnormality diagnosis are systematically guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring of water conservancy and hydropower engineering, and in particular to a method and system for self-diagnosis of abnormality in safety monitoring data of hydraulic structures. Background Art

[0002] Safety monitoring data is the basis for realizing the digitalization, intelligence and networking of water conservancy and hydropower projects. The structural state response obtained by safety monitoring is affected by multiple factors such as the environment, load and the performance of the structure itself. Safety risk prevention and control mainly relies on identifying anomalies in safety monitoring data, evaluating structural states and taking control measures. Abnormal effect quantity monitoring data may be caused by abnormal structural states or monitoring system failures. Among them, abnormal structural states caused by environmental and load conditions are the key and difficult points of operation management. Taking the uplift pressure of building foundation as an example, as the main load, it affects the anti-sliding stability of the structure. Although curtain grouting and foundation drainage are often taken before construction to reduce the uplift pressure, it is difficult to predict the actual effect and durability due to complex geological conditions. After the project is put into operation, the problem of abnormal local uplift pressure at the dam foundation is gradually exposed. Therefore, timely identification of uplift pressure and analysis of its causes are crucial to dam safety.

[0003] In terms of anomaly identification, for multivariate data such as independent variables and effect quantities, automatic and intelligent identification of abnormal values ​​and analysis of causes are the core requirements for ensuring the reliability of safety monitoring and evaluation conclusions. When the structural state changes, the monitoring effect quantities such as deformation and uplift pressure will have different responses to specific environments and load combinations. For example, the increase in uplift pressure under the same reservoir water level and reservoir water temperature combination may indicate that the performance of the dam foundation anti-seepage drainage system has deteriorated. However, traditional methods for identifying anomalies in the monitoring time series of water conservancy and hydropower engineering effect quantities, such as continuity analysis and consistency analysis, mostly rely on experience and lack in-depth analysis of the evolution mechanism of structural states, which is not conducive to automatic and intelligent identification. In particular, the change of effect quantity is often the result of the coupling between environmental load conditions and self-factors, which requires comprehensive analysis and judgment from the perspective of multivariate data. Only when the anomaly identification method can analyze the interaction between the effect quantity and the corresponding independent variable, it is possible to accurately distinguish and analyze the change trend of the effect quantity and its abnormal situation.

[0004] At present, due to the lack of relevant research on abnormal identification of monitoring effect quantity and abnormal cause analysis considering environmental and load conditions, the reliability, automation and intelligence of abnormal identification of safety monitoring effect quantity of water conservancy and hydropower projects are affected. Therefore, there is an urgent need for a self-diagnosis method for abnormal safety monitoring data of hydraulic structures to assist the construction of digital twin water conservancy projects and intelligent water conservancy projects. Summary of the invention

[0005] To this end, the technical problem to be solved by the present invention is to overcome the technical defects existing in the prior art, and propose a method and system for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures. The method covers the entire process of data processing, from acquiring monitoring data, through multivariate time series construction, smoothing and normalization processing, to using a dynamic time warping algorithm to analyze correlations, and then constructing a regression model based on multivariate correlations and time lag effects, thereby systematically ensuring the accuracy and reliability of abnormality diagnosis.

[0006] In order to solve the above technical problems, the present invention provides a method for self-diagnosis of abnormal safety monitoring data of hydraulic structures, comprising: S10, acquiring monitoring data, arranging multiple variables of the monitoring data in chronological order to form a multivariate time series, and performing smoothing and normalization processing on the multivariate time series; S20, dynamic time warping algorithm was used to analyze the correlation between effect size and respective variables in the processed multivariate time series; S30. Based on the results of correlation analysis and time lag effect analysis, a multivariate regression model between effect size and independent variables was constructed; S40. Use multivariate local abnormal factors to identify abnormalities in effect size and independent variable measurements. Based on the local abnormal factor threshold, compare the distribution characteristics and occurrence time of abnormal values ​​reflected by the local abnormal factors from different time scales to determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, determine that it is a safety state abnormality caused by multiple factors, and identify the transformation mechanism of the structural state.

[0007] In one embodiment of the present invention, in step S20, when analyzing the correlation between the effect size and the respective variables in the multivariate time series, the dynamic time warping algorithm uses the Sakoe-Chuba band constrained warping path.

[0008] In one embodiment of the present invention, the constraint rule of the Sakoe-Chuba band constraint is that the regular path is along the main diagonal i = j At the same time, by the fixed width T 0∈ N To constrain the regular range, based on the Sakoe-Chuba band constraint, the global cost matrix D The calculation formula is: ; In the formula, D represents the total cost matrix, Represent the variables in two time series, respectively, where i ∈ [1, N ], j∈ [1, M ], N , M There are two time series X , Y Length, Represents the given parameter.

[0009] In one embodiment of the present invention, in step S30, the method of constructing a multivariate regression model between effect size and independent variables includes: S31. Construct a data matrix based on independent variables and effect sizes, and establish a multiple linear regression model framework based on the variable relationships in the data matrix; S32, solving the unknown coefficients of the multiple linear regression model framework by the least square method; S33. Use the stepwise regression analysis method to screen the independent variables in the multiple linear regression model according to the degree of influence of the independent variables on the effect size, determine the independent variables with high influence and eliminate the independent variables with low influence, and obtain a simplified multiple linear regression model; S34. Based on the simplified multiple linear regression model, the relative contribution of each independent variable to the explained effect size variance is calculated through relative weights to determine the relative importance ranking of the independent variables; S35. Incorporate the relative importance ranking information of the independent variables and construct a multiple linear regression model between the effect size and the independent variables.

[0010] In one embodiment of the present invention, in step S40, the method of using multivariate local abnormal factors to identify abnormalities in effect size and independent variable measurements includes: S41. Integrate the independent variables and effect size time series to form a data set, construct a multivariate time series, and perform smoothing and balancing on the multivariate time series; S42, calculating the multivariate local anomaly factor for the processed multivariate time series, and determining a threshold value based on the calculation result of the multivariate local anomaly factor; S43, calculating the multivariate local abnormal factor values ​​of the multivariate time series again, and obtaining the effect size and the distribution of the local abnormal factor values ​​of the independent variables with strong correlation by comparison; S44, comparing the local anomaly factor value of each data point with a threshold value, if the local anomaly factor value of the data point is greater than the threshold value, the data point is an outlier; S45. Compare the distribution characteristics and occurrence time of the abnormal values ​​reflected by the local abnormal factors from different time scales, and determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state abnormality.

[0011] In one embodiment of the present invention, in step S42, a threshold value of the local anomaly factor value is determined by a kernel density estimation method.

[0012] In one embodiment of the present invention, after obtaining the abnormal information of the structural state, the cause of the abnormal structural effect size and the mechanism of the structural state transformation are analyzed based on the abnormal information obtained by multivariate local abnormality identification.

[0013] In addition, the present invention also provides a hydraulic structure safety monitoring data abnormality self-diagnosis system, comprising: A data processing module, which is used to obtain monitoring data, arrange multiple variables of the monitoring data in chronological order to form a multivariate time series, and perform smoothing and normalization processing on the multivariate time series; A correlation analysis module, which is used to analyze the correlation between the effect size and the respective variables in the processed multivariate time series using a dynamic time warping algorithm; Model building module, which builds a multiple regression model between effect size and independent variables based on the results of correlation analysis and time lag effect analysis; The structural anomaly identification module uses multivariate local anomaly factors to identify anomalies in effect sizes and independent variable measurements. Based on the local anomaly factor threshold, it compares the distribution characteristics and occurrence time of the anomaly values ​​reflected by the local anomaly factors from different time scales, and determines whether the effect size and the local anomaly factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state anomaly caused by multiple factors, and the structural state transformation mechanism is identified.

[0014] Furthermore, the present invention also provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method when executing the program.

[0015] Furthermore, the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above-mentioned method are implemented.

[0016] The above technical solution of the present invention has the following advantages compared with the prior art: 1. The present invention provides a method and system for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures. The method covers the entire process of data processing, from acquiring monitoring data, constructing multivariate time series, smoothing and normalizing, to analyzing correlation using dynamic time warping algorithm, and then constructing a regression model based on multivariate correlation and time lag effect, which systematically guarantees the accuracy and reliability of abnormal diagnosis. And it proposes a recommended value of time lag effect for key effect quantities, providing clear guidance for actual engineering applications, and is highly practical; 2. The present invention provides a method and system for self-diagnosis of abnormal data of hydraulic structure safety monitoring, which supports multi-scale method improvement, meets the abnormal identification requirements of different time scales by decomposing multi-variable time series, and shows good scalability. At the same time, it uses multi-variable local abnormal factors to identify abnormalities from different time scales, and comprehensively and accurately captures abnormal data; the overall system is complete and each link is closely connected, which facilitates the realization of automated and intelligent abnormal diagnosis, and significantly improves monitoring efficiency and accuracy; 3. The present invention provides a method and system for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures, which fully consider the characteristics of different hydraulic structures. The constructed regression model fits the actual operation of hydraulic structures, accurately describes the internal connection between variables, provides a scientific reference for abnormality identification, and effectively ensures the safe and stable operation of hydraulic structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments of the present invention in conjunction with the accompanying drawings.

[0018] Figure 1 The present invention is a flowchart of a method for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures.

[0019] Figure 2 It is a flowchart diagram of step S30 proposed by the present invention.

[0020] Figure 3 It is a flowchart diagram of step S40 proposed by the present invention.

[0021] Figure 4 It is a schematic diagram of the distribution of measuring points of the dam foundation uplift pressure of the concrete gravity dam in the case proposed by the present invention.

[0022] Figure 5 This is the process line of the uplift pressure time series of the dam foundation uplift pressure measuring point UP4 proposed by the present invention.

[0023] Figure 6 This is the process line of the uplift pressure time series of the dam foundation uplift pressure measuring point UP5 proposed by the present invention.

[0024] Figure 7 This is the process line of the uplift pressure time series of the dam foundation uplift pressure measuring point UP6 proposed by the present invention.

[0025] Figure 8 This is the process line of the uplift pressure time series of the dam foundation uplift pressure measuring point UP7 proposed by the present invention.

[0026] Fig. 9 It is the process line of the time series of daily average reservoir water level and reservoir water temperature near the dam foundation proposed by the present invention.

[0027] Fig.10 This is the mapping relationship between the uplift pressure at the uplift pressure measuring point UP4 and the reservoir water level time series based on the SC bandwidth constraint proposed by the present invention.

[0028] Fig.11 This is the mapping relationship between the uplift pressure at the uplift pressure measuring point UP5 and the reservoir water level time series based on the SC bandwidth constraint proposed by the present invention.

[0029] Fig.12 This is the mapping relationship between the uplift pressure at the uplift pressure measuring point UP6 and the reservoir water level time series based on the SC bandwidth constraint proposed by the present invention.

[0030] Fig.13 This is the mapping relationship between the uplift pressure at the uplift pressure measuring point UP7 and the reservoir water level time series based on the SC bandwidth constraint proposed by the present invention.

[0031] Fig.14 It is a diagnostic diagram of the RM model of the uplift pressure at the uplift pressure measuring point UP4 constructed by the present invention.

[0032] Fig.15 It is a diagnostic diagram of the RM model of the uplift pressure at the uplift pressure measuring point UP7 constructed by the present invention.

[0033] Fig.16 It is the fitting effect of the RM model of the uplift pressure at the uplift pressure measuring point UP4 constructed by the present invention.

[0034] Fig.17 It is the fitting effect of the RM model of the uplift pressure at the uplift pressure measuring point UP5 constructed by the present invention.

[0035] Fig.18 It is the fitting effect of the RM model of the uplift pressure at the uplift pressure measuring point UP6 constructed by the present invention.

[0036] Fig.19 It is the fitting effect of the RM model of the uplift pressure at the uplift pressure measuring point UP7 constructed by the present invention.

[0037] Fig. 20 It is the LOF value distribution diagram of the reservoir water level time series proposed by the present invention.

[0038] Fig.21 It is the LOF value distribution diagram of the reservoir water temperature time series proposed by the present invention.

[0039] Fig. 22 It is a distribution diagram of the LOF value of the uplift pressure time series of the uplift pressure measuring point Up4 proposed by the present invention.

[0040] Fig.23 It is a distribution diagram of the LOF value of the uplift pressure time series of the uplift pressure measuring point Up7 proposed by the present invention. DETAILED DESCRIPTION

[0041] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0042] Reference Figure 1 As shown, an embodiment of the present invention provides a method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures, comprising the following steps: S10, acquiring monitoring data, arranging multiple variables of the monitoring data in chronological order to form a multivariate time series, and performing smoothing and normalization processing on the multivariate time series; S20, dynamic time warping algorithm was used to analyze the correlation between effect size and respective variables in the processed multivariate time series; S30. Based on the results of correlation analysis and time lag effect analysis, a multivariate regression model between effect size and independent variables was constructed; S40. Use multivariate local abnormal factors to identify abnormalities in effect size and independent variable measurements. Based on the local abnormal factor threshold, compare the distribution characteristics and occurrence time of abnormal values ​​reflected by the local abnormal factors from different time scales to determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, determine that it is a safety state abnormality caused by multiple factors, and identify the transformation mechanism of the structural state.

[0043] The present invention provides a method for self-diagnosis of abnormal data of hydraulic structure safety monitoring, which covers the whole process of data processing, from obtaining monitoring data, constructing multivariate time series, smoothing and normalization, to analyzing correlation using dynamic time warping algorithm, and then constructing regression model based on multivariate correlation and time lag effect, which systematically guarantees the accuracy and reliability of abnormal diagnosis. And the time lag effect recommended value is proposed for key effect quantity, which provides clear guidance for actual engineering application and has strong practicality.

[0044] Among them, in step S10, assuming that the monitoring data does not have gross errors and missing values ​​caused by the monitoring system, the data preprocessing here is mainly the processing process required to convert the monitoring data into appropriate data suitable for analysis. Preferably, the Moving Average Method (MA) can be used to smooth short-term fluctuations and highlight long-term change trends or periodicity, wherein the MA parameters are set according to the time lag effect of the monitoring data. Due to the fluctuations of environmental and load variables such as reservoir water level and reservoir water temperature, as well as the evolution of the physical and mechanical properties of the anti-seepage drainage system and structural filling materials during long-term service, the deformation and seepage properties of hydraulic structures will also fluctuate. Therefore, considering the inconsistent change amplitudes between independent variables such as environment and load and effect size, this embodiment normalizes multiple variables to the interval [0,1] to eliminate the impact caused by the difference in change amplitudes.

[0045] Among them, in step S20, the dynamic time warping algorithm (DTW) allows flexible transformation of time series, which can minimize the impact of time series offset and distortion, and provide strong robustness for correlation measurement. Given a length of N and M Two time series, a test series and a reference series, are represented as and ; In the formula, x i and y j Respectively represent the time series X and Y In i and j The goal of the DTW algorithm is to O ( NM ) time to find the optimal solution, and the optimal alignment results of different time series are transformed into arranging all time series points by minimizing the cost function (also known as "distance"). The DTW algorithm first constructs a distance matrix C ∈R N×M ,express X and Y All pairwise distances between the two sequences are called X and Y The local cost matrix of the comparison is defined as follows: ; In the formula, C Represents local cost N × M Matrix, the matrix's first ( i , j ) elements are measuring points xi and y j The distance between c ij (usually Euclidean distance), i and j They are X and Y As can be seen from the location of the above measurement points, when the time series correlation is good, the distance value is small; when the time series correlation is weak, the distance value is large. Once the local cost matrix is ​​established, the DTW algorithm can find the regular path with the minimum distance under the constraint of the alignment criterion. The regular path defines the element x i ∈ X arrive y j ∈ Y Follow X and Y The first and last elements of the boundary conditions match each other.

[0046] Formally speaking, the regular path constructed by the DTW algorithm p is a set of continuous matrix elements that defines X and Y For the mapping between l ∈[1: K ] (max( N , M )≤ K ≤ N + M -1), p No. l The elements are defined as p l =( p i , p j )∈[1: N ]×[1: M ], K represents the length of the regular path. Regular paths are usually subject to three conditions: (1) boundary conditions, p 1=( p 1, p 1) and p K =( p N , p M ), which requires that the regular path starts and ends in the diagonal cells of the matrix; (2) Monotonicity condition, given p l = ( pa , p b ),but p l-1 = ( p a’ , p b’ ),in a - a '≥0, b - b '≥0, which makes p The points in are monotonically spaced in time; (3) Continuity condition: given p l = ( p a , p b ),but p l-1 = ( p a’ , p b’ ),in a - a '≤1, b - b '≤1, which restricts the allowed steps in the regular path to adjacent cells (including diagonal adjacent cells). The total cost function associated with the regular path is calculated relative to the local cost matrix representing all pairwise distances: ; In the formula, c ( x li , y lj ) indicates the regularized path l two data point indices in the elements (one from X , one from Y ). It can be found that there is an exponential regularization path that satisfies the above conditions. However, the two time series X and Y The similarity measure between D ij is the best regularized path that minimizes the total cost function value p ∗ .

[0047] In order to overcome the challenge brought by the exponential growth of the optimal path, this embodiment uses dynamic programming (DP) to calculate the optimal matching path between two given time series with constraints. The DP part of the DTW algorithm uses the DTW distance function: ; In the formula, p ∗ ={ p 1 ∗ , p 2 ∗ , …, p l ∗ ,…, p K ∗}, P N×M Represents the set of all possible regular paths and constructs the total cost matrix D , which is defined as follows: First line: ; First column: ; Other elements: , , ; In the formula, D Represents the total cost matrix ， Represent the variables in two time series, respectively, where i ∈ [1, N ], j ∈ [1, M ], N , M There are two time series X , Y Length, c express The distance between two data points.

[0048] Once the total cost matrix determines the regular path, we can start from point p end = ( p N , p M )arrive p start = ( p 1; p 1) to find it by simple backtracking. In some cases, this approach may have some undesirable effects. For example, because the DTW algorithm tries to find the minimum distance, this may cause the time series to X A large number of points are mapped to another time series YTo overcome this shortcoming, the present embodiment uses constraints to regularize the path. By using global constraints to limit the allowed regularized paths, preferably, the constraints used in the present embodiment are Sakoe-Chuba (SC) band constraints. The SC band constraints require that the paths along the main diagonal be i = j To regularize, by fixed width T 0∈ N To constrain the scope of regulation.

[0049] The above SC band constraint restricts the regular path to the threshold T 0 represents a specific depth for regularization and element correspondence, so that the global cost matrix D It can be calculated as follows: ; In the formula, D represents the total cost matrix, Represent the variables in two time series, respectively, where i ∈ [1, N ], j ∈ [1, M ], N , M There are two time series X , Y Length, Represents the given parameter.

[0050] It should be noted that when the lengths of the two time series are N and M When there is no order of magnitude difference, the SC band constraint shows good performance, but it is not suitable when there is a significant difference in the length of the two time series. Therefore, when executing global constraints, attention should be paid to the compatibility of the time series length. For the safety monitoring of water conservancy and hydropower projects, the technical specifications for safety monitoring require the simultaneous monitoring of the environment, load independent variables and effect size. When there are missing values, the missing values ​​can be interpolated before preprocessing, or a well-balanced time series data set can be obtained through data processing.

[0051] For the time series of hydraulic structures, the width of the SC band can be fixed at 10% of the length of the time series. In order to determine the appropriate SC bandwidth, all possible SC bandwidths of global constraints should be tested as much as possible, and the bandwidth with the maximum test accuracy should be selected as the model input bandwidth. For the time series of hydraulic structures, each effect quantity can be associated with the characteristics of the independent variables such as special environment and load. For example, for gravity dams, due to the low permeability of the rock itself and the anti-seepage performance of the grouting curtain of the dam foundation, the change of the foundation uplift pressure may significantly lag behind the change of the reservoir water level. At the same time, the changes in the foundation permeability and the anti-seepage performance of the grouting curtain lead to different time lag intervals between the change of the reservoir water level and the uplift pressure at different parts. For the same effect quantity with different spatial distribution of measuring points, it is also necessary to test in order to obtain a more reasonable SC bandwidth. The process of determining the optimal SC bandwidth is shown in Table 1. Since the change of uplift pressure usually lags behind the reservoir water level by about 10 days, the SC bandwidth can be set to 15 days first. In the test, each step 1 gradually reduces the bandwidth to find the optimal SC band, thereby obtaining the best regular path.

[0052] Table 1 Process for determining the optimal SC bandwidth

[0053] Among them, in step S30, if Figure 2 As shown in Figure 2, the method for constructing a multiple regression model between effect size and independent variables includes the following steps: S31. Construct a data matrix based on independent variables and effect sizes, and establish a multiple linear regression model framework based on the variable relationships in the data matrix; S32, solving the unknown coefficients of the multiple linear regression model framework by the least square method; S33. Use the stepwise regression analysis method to screen the independent variables in the multiple linear regression model according to the degree of influence of the independent variables on the effect size, determine the independent variables with high influence and eliminate the independent variables with low influence, and obtain a simplified multiple linear regression model; S34. Based on the simplified multiple linear regression model, the relative contribution of each independent variable to the explained effect size variance is calculated through relative weights to determine the relative importance ranking of the independent variables; S35. Incorporate the relative importance ranking information of the independent variables and construct a multiple linear regression model between the effect size and the independent variables.

[0054] Specifically, the multiple linear regression model is designed to simulate the relationship between the effect quantity and the independent variable. This method can clarify the contribution of independent variables such as environment and load to the structural state response quantity (such as effect quantities such as seepage and deformation), and subdivide the structural state response quantity into different components, each of which is associated with specific environmental and load conditions. For example, in engineering practice, environmental and load independent variables such as reservoir water level and reservoir water temperature will affect the structural response quantities such as deformation and seepage of hydraulic structures. The study also found that the structural response quantity is also affected by time effects such as the evolution of the performance of the dam body and dam foundation materials. For example, the foundation seepage state will be affected by time effects such as the degradation of the anti-seepage drainage system.

[0055] When conducting multiple linear regression model analysis, the first step is to determine potential independent variables such as environment and load. Taking the foundation uplift pressure of hydraulic structures as an example, its potential independent variables include reservoir water temperature, reservoir water level and time effect. On this basis, a data matrix of independent variables and effect size is constructed, and the unknown coefficients are fitted by the least squares method.

[0056] The following takes the uplift pressure as an example to explain the construction process of the multiple linear regression model. The uplift pressure can be expressed as: ; ; ; ; In the formula, a 0, a i , b i , c 1 and c 2 is the unknown coefficient, H h represents the component related to the hydrostatic pressure (reservoir water level), H T It represents the component caused by the change of ambient temperature such as reservoir water temperature. H θ represents the time effect size component, h 0 represents the dam foundation elevation, H h and H T They are and The linear combination of H θ By linear terms θ and the nonlinear term ln θ The combination of represents the irreversible component; among them, and represent the average values ​​of reservoir water level and ambient temperature (reservoir water temperature) in the previous few days determined by the monitoring day and hysteresis effect, respectively. θ equal t / 100, t Indicates the number of days from the start monitoring date to the current monitoring date.

[0057] The above unknown coefficients ( a 0, a i , b i , c 1 and c 2) The least square method can be used to minimize the difference between the actual monitoring value and the effect size, and the model can be further simplified with the help of stepwise regression analysis to extract the main factors affecting the effect size, such as environment, load and time. It should be noted that the possible lag time should be considered, that is, a series of possible lag times should be assumed, and appropriate and scientific lag times should be selected from them to achieve the best correlation between the environment, load independent variables and the effect size.

[0058] It is important to analyze the relative importance of the independent variables to the effect size, and a simple way to do this is to compare the standardized regression coefficients. Here, relative weights are used to measure the relative importance of the independent variables. Relative weights are calculated by calculating the squared zero-order correlation between the independent and dependent variables (i.e., the standardized beta weight) and dividing that value by a multiple of ². Relative weights have the significant advantage of being able to rank the contribution of individual independent variables.

[0059] First, for a given independent variable X , the structural coefficient is ; In the formula, Indicates the independent variable X and the dependent variable Y The bivariate correlation between R Represents the multivariate correlation of the regression including all independent variables. Based on the fitting coefficients, a multiple linear regression model of the effect size can be constructed. At the same time, the relative contribution of each independent variable to explaining the variance of the effect size can be obtained through relative weight analysis.

[0060] Among them, in step S40, if Figure 3 As shown, the method for identifying abnormalities of effect size and independent variable measurement values ​​using multivariate local abnormal factors includes the following steps: S41. Integrate the independent variables and effect size time series to form a data set, construct a multivariate time series, and perform smoothing and balancing on the multivariate time series; S42, calculating the multivariate local anomaly factor for the processed multivariate time series, and determining a threshold value based on the calculation result of the multivariate local anomaly factor; S43, calculating the multivariate local abnormal factor values ​​of the multivariate time series again, and obtaining the effect size and the distribution of the local abnormal factor values ​​of the independent variables with strong correlation by comparison; S44, comparing the local anomaly factor value of each data point with a threshold value, if the local anomaly factor value of the data point is greater than the threshold value, the data point is an outlier; S45. Compare the distribution characteristics and occurrence time of the abnormal values ​​reflected by the local abnormal factors from different time scales, and determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state abnormality.

[0061] Specifically, firstly, the relevant independent variables and effect size time series are formed into a data set to construct a multivariate time series; then, the multivariate local outlier factor (MVLOF) is calculated for the multivariate time series after smoothing and balancing; based on the MVLOF calculation value, a small probability value is selected from the statistical value as the threshold; finally, the MVLOF value of the multivariate time series is calculated again, and the distribution of the LOF values ​​of the calculated effect size and the independent variables with strong correlation is compared to identify anomalies on a small time scale, especially the top-ranked outliers. On this basis, it is analyzed whether the outliers conform to the actual physical meaning of the response of hydraulic structures.

[0062] Unlike traditional anomaly methods that treat anomalies as binary attributes, LOF is density-based and has no distribution assumptions. The brief calculation algorithm of LOF is as follows: For the training dataset X =[ x 1, x 2, …, x N ]∈R D×N , find out according to the following formula x a ( a = 1, 2, …, N )of k neighbors, and then we can x a The neighborhood set of is identified as KNN ( x a ): ; In the formula, x a 、x b Representation dataset X The elements in ( a , b,n = 1, 2, …, N ,and a ≠b ); d ( x a , x b )express x a 、x b the distance between the two; D , N is a constant.

[0063] x a of k The nearest neighbor is represented by k - distance ( x a ), defined as k Nearest Neighbors x a The Euclidean distance of d ( x a , x b ) ( a ≠ b , x b ∈KNN( x a )) k Minimum value, x a Relative to x b The reachable distance is defined as: ; In the formula, express x a Relative to x b The reachable distance; k - distance ( x a )express x a of k Neighbors, k is a constant.

[0064] x a The local reachability density of is given by: ; In the formula, express x a The local accessibility density of .

[0065] x a LOF by x a The local reachability density and x b of k The reachability density of the nearest neighbors is calculated as: ; In the formula, LOF( x a )express x a The local anomaly coefficient value.

[0066] The anomaly coefficient factor reflects the degree of anomaly. LOF is KNN( x a ) and the average density x a The density ratio of LOF( x a ) value, the more likely it is an outlier, because x a and KNN( x a ) are similar in density, if x a Not abnormal, LOF( x a ) should be close to 1, but if x a is an outlier, LOF( x a ) should be significantly greater than 1, because x a With KNN( x a ) has a very small relative density. Specifically, if LOF( x a ) is less than or equal to 1, then x a However, a LOF value of "1" implies an ideal condition where the local accessibility density of the sample is equal to its k The average density of the nearest neighbors is the same.

[0067] Since the acceptable range of LOF values ​​varies from application to application, a threshold is needed to determine whether the value of the monitoring point is abnormal. Kernel density estimation (KDE) is a non-parametric method for estimating probability density functions from discrete samples. It can be introduced to evaluate the threshold of LOF values. By summing the individual kernels of each sample, it estimates the probability distribution function over the entire sample. If LOF( x a ) is higher than the 99.3% threshold determined by KDE using LOF, then x a Therefore, this implementation method uses KDE to select control limits and uses MVLOF to compare local anomalies of multiple variables such as environment, load, and effect size to determine the properties of the anomaly.

[0068] Furthermore, after obtaining the abnormal information of the structural state, the abnormal information obtained by multivariate local anomaly identification is used to analyze the causes of the abnormal structural effect quantity and the mechanism of the structural state transformation. Specifically, the local anomaly is judged from the abnormalities of the independent variables related to the effect quantity, and the trend change is judged in combination with the time effect. In this way, it is analyzed whether the anomaly is a short-term anomaly caused by an unfavorable load combination, or a trend change in an unfavorable direction caused by the coupling of unfavorable loads and the evolution of the structure's own performance, so that engineering managers can make corresponding risk control measures.

[0069] The present invention provides a method for self-diagnosis of abnormal data of hydraulic structure safety monitoring, which supports multi-scale method improvement, meets the abnormal identification requirements of different time scales by decomposing multi-variable time series, and shows good scalability. At the same time, it uses multi-variable local abnormal factors to identify abnormalities from different time scales, and captures abnormal data comprehensively and accurately; the overall system is complete and each link is closely connected, which is convenient for realizing automated and intelligent abnormal diagnosis, and significantly improves monitoring efficiency and accuracy.

[0070] The present invention provides a method for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures, which fully considers the characteristics of different hydraulic structures. The constructed regression model fits the actual operation of hydraulic structures, accurately describes the internal connection between variables, provides a scientific reference for abnormality identification, and effectively ensures the safe and stable operation of hydraulic structures.

[0071] The following is a detailed description of a method for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures proposed by the present invention through a specific case.

[0072] Take Reservoir A as an example. Reservoir A is a concrete gravity dam with multi-year regulation capacity. The dam crest and normal water level elevations are 153.00m and 147.00m respectively, with a maximum height of 71.0m. The two spillway dam sections are located in the middle of the riverbed, and 11 retaining dam blocks are located on both sides of the spillway dam section. Figure 4 As shown. The project began to store water in July 2005. The dam is located on bedrock with complex geological conditions. The river valley is V-shaped. Steep joints and faults are developed in the dam foundation rock mass. The depth of the relatively impermeable layer with a permeability of less than 3.0Lu on the left and right banks is 4.4~20.2m and 2.3~16m respectively. In order to reduce the uplift pressure of the dam foundation, curtain grouting was carried out on the upstream side of the dam body. The grouting curtain hole spacing was 2.0m, and the depth was determined by the depth of the relatively impermeable layer. The grouting curtain extended to 14.0m and 18.0m to the left and right dam shoulders respectively. The dam foundation drainage consists of a row of drainage holes with a diameter of 110mm, a hole spacing of 2.0m, and extending to about 12.0m in the bedrock. The dam foundation drainage is introduced into the dam foundation observation gallery for discharge.

[0073] The dam safety monitoring system was installed in June 2007. The dam foundation pressure gauge was installed 1.0m below the foundation surface. The monitored environmental quantities include reservoir water level and reservoir water temperature. The dam foundation pressure measurement points and reservoir water temperature measurement points are arranged as follows: Figure 4 As shown in Figure 2. The time series of dam foundation uplift pressure is from July 2006 to July 2014. However, due to the failure of the automation system in 2010, the time series from January 2011 to June 2014 was interrupted. Considering this factor, the time series from January 2011 to June 2014 was selected. The uplift pressure values ​​of the dam foundation uplift pressure measuring points Up4 and Up5 experienced a sudden increase in March 2012. The change patterns of Up6 and Up7 were consistent with those of other measuring points, and the temporal consistency was maintained. The time series process lines of the uplift pressure measuring points Up4, Up5, Up6 and Up7 are shown in Figure 2. Figure 5 , Figure 6 , Figure 7 and Figure 8 As shown in the figure, the time series process lines of the daily average reservoir water level and the reservoir water temperature near the dam foundation (TW1) are as follows: Fig. 9 As shown, the reservoir water temperature shows an annual periodicity, and the reservoir water level presents the basic characteristics of multi-year regulation.

[0074] 1. Data preprocessing The moving average (MA) method is used to smooth the time series. The processed data is as follows: Figure 5 , Figure 6 , Figure 7 and Figure 8 According to the time lag effect of the dam foundation uplift pressure, the smoothing time window is selected as 10. Figure 5 , Figure 6 , Figure 7 and Figure 8The process line becomes smooth after medium MA processing, but still maintains the trend of seasonal changes and overall changes, and the smoothed data is normalized to eliminate the impact of the amplitude.

[0075] 2. Analysis of multivariate correlation By using the SC bandwidth-constrained DTW algorithm to perform similarity analysis on the reservoir water level and uplift pressure measurement points, considering the time lag effect, the time window is set to 15 to 0, and it is found that the best regularized path can be obtained when the width is set to 7 days. The SC bandwidth-constrained DTW algorithm can maintain the similarity of time series at different scales, which is crucial for safety monitoring. The relationship between mapping the time series of uplift pressure measurement points Up4, Up5, Up6 and Up7 to the time series of reservoir water level is as follows: Fig.10 , Fig.11 , Fig.12 and Fig.13 As shown, from Fig.12 and Fig.13 It can be seen that the uplift pressure at the uplift pressure measuring points Up6 and Up7 is consistent with the change pattern of reservoir water level.

[0076] 3. Determination of the main independent variables affecting the effect size RM was used to fit the interaction mechanism between effect size and independent variables, and regression diagnostics were used to assess the reliability of the constructed model. The results are shown in Fig.14 and Fig.15 As shown, from the fitted value ~ residual plot, it can be seen that the residual and the estimated value are basically unrelated. From the Quantile-Quantile plot (QQ plot), it can be seen that the points in the plot are in a straight line, and the normality of the data is relatively good. From the diagnostic results, the model fitting effect is good.

[0077] The relevant results of the regression model of each measuring point are listed in Table 2. It can be seen that the complex correlation coefficient of the model of each measuring point R 2 All are greater than 0.70, the model fitting accuracy is high, and the confidence interval of the mean value is calculated. p The value is small enough, and there is a strong correlation between the effect size and the independent variable. In comparison, the fitting accuracy of the Up6 and Up7 measuring points is higher. The fitting effects of the regression model of the uplift pressure measuring points Up4, Up5, Up6 and Up7 are as follows: Fig.16 , Fig.17 , Fig.18 and Fig.19 shown.

[0078] Table 2 Fitting results of regression models at each measuring point

[0079] As shown in Table 3, for the uplift pressure measuring point Up4, the reservoir water level isR 2 The relative importance of reservoir water level to the uplift pressure measuring point Up6 is 36.49%, the time effect is 43.19%, and the reservoir water temperature is 20.33%. On the contrary, for the uplift pressure measuring point Up6, the reservoir water level is 61.74%, and the time effect is 35.75%. It can be seen that the reservoir water level has the greatest relative importance for the uplift pressure measuring point Up6, followed by the time effect, but for the uplift pressure measuring point Up4, the time effect has the greatest relative importance, followed by the reservoir water level and the reservoir water temperature.

[0080] Table 3 Relative importance of the influence of environment, load and other independent variables on the dam foundation uplift pressure

[0081] It can be found that the uplift pressure is significantly affected by the reservoir water level; in addition, the uplift pressure is also affected by the reservoir water temperature and time effect, and the uplift pressure measuring points Up4 and Up5 are particularly significant.

[0082] 4. Identification of multivariate local anomalies The calculation results of LOF values ​​of uplift pressure and related independent variables time series are listed in Table 4. The calculation results show that the ideal reachable distance of reservoir water temperature is 60, but that of other variables is 8. The LOF values ​​of multiple variables are calculated, and then the threshold of KDE is obtained. At the same time, the LOF values ​​of reservoir water level and reservoir water temperature time series are calculated respectively, and the MVLOF values ​​are compared to identify the local anomalies of multiple variables.

[0083] Table 4 Calculation results of LOF values ​​of uplift pressure and related independent variables time series

[0084] When the pressure performance is abnormal, the independent variables and the LOF values ​​of the effect size show the binary relationship between the variables. Table 4 lists the five largest abnormal points. Figure 20~Figure 23 The effect size and LOF values ​​of the independent variables were compared. As can be seen from Table 4, the local anomaly of uplift pressure is consistent with the local anomaly of reservoir water level and reservoir water temperature, and the uplift pressure is affected by the coupling of reservoir water level and reservoir water temperature. Figure 20~Figure 23 This is also confirmed by the LOF value distribution diagram in . When comparing the LOF values ​​of the independent variables and the effect size, when an independent variable shows an abnormal value, the uplift pressure will also be abnormal. For example, in March and August 2011, the reservoir water level dropped and rose respectively, and the uplift pressure showed large LOF values ​​in these two time periods. In March 2012, due to the coupling effect of low reservoir water temperature and high reservoir water level, the LOF value reached the maximum value.

[0085] 5. Analysis of abnormal causes The uplift pressure changes seasonally with the reservoir water level and temperature. In addition to the outflow water level, the change of reservoir water temperature will affect the opening of discontinuities such as faults and joints in the dam foundation, and thus affect the size of the uplift pressure on the dam foundation. When encountering continuous low temperature and high water level, the uplift pressure will deviate from the expected value. Compared with the low temperature and high water level conditions encountered in history, the foundation anti-seepage curtain where the uplift pressure measuring points Up4 and Up5 are located may be weakened. It is recommended to carry out local reinforcement grouting.

[0086] Corresponding to the above method embodiment, the present invention also provides a hydraulic structure safety monitoring data abnormality self-diagnosis system, comprising: A data processing module, which is used to obtain monitoring data, arrange multiple variables of the monitoring data in chronological order to form a multivariate time series, and perform smoothing and normalization processing on the multivariate time series; A correlation analysis module is used to analyze the correlation between the effect size and the respective variables in the processed multivariate time series using a dynamic time warping algorithm; Model building module, which builds a multiple regression model between effect size and independent variables based on the results of correlation analysis and time lag effect analysis; The structural anomaly identification module uses multivariate local anomaly factors to identify anomalies in effect sizes and independent variable measurements. Based on the local anomaly factor threshold, it compares the distribution characteristics and occurrence time of the anomaly values ​​reflected by the local anomaly factors from different time scales, and determines whether the effect size and the local anomaly factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state anomaly caused by multiple factors, and the structural state transformation mechanism is identified.

[0087] The present invention provides a hydraulic structure safety monitoring data abnormality self-diagnosis system, which systematically guarantees the accuracy and reliability of abnormality diagnosis from acquiring monitoring data, constructing multivariate time series, smoothing and normalizing, to analyzing correlation using dynamic time warping algorithm, and then constructing a regression model based on multivariate correlation and time lag effect. And for the key effect quantity, the time lag effect recommended value is proposed, providing clear guidance for actual engineering application, and it is highly practical.

[0088] The hydraulic structure safety monitoring data anomaly self-diagnosis system of this embodiment is used to implement the aforementioned hydraulic structure safety monitoring data anomaly self-diagnosis method, so its specific implementation method can refer to the description of the corresponding various partial embodiments and will not be introduced in detail here.

[0089] In addition, since the hydraulic structure safety monitoring data anomaly self-diagnosis system of this embodiment is used to implement the aforementioned hydraulic structure safety monitoring data anomaly self-diagnosis method, its function corresponds to that of the aforementioned method and will not be repeated here.

[0090] Corresponding to the above method embodiment, an embodiment of the present invention further provides a computer device, including: A memory for storing computer programs; A processor is used to implement the steps of the above-mentioned method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures when executing a computer program.

[0091] In the embodiment of the present invention, the processor may be a central processing unit (CPU), an application specific integrated circuit, a digital signal processor, a field programmable gate array or other programmable logic devices.

[0092] The processor may call a program stored in the memory. Specifically, the processor may execute the operations in the above-mentioned embodiment of the method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures.

[0093] The memory is used to store one or more programs, which may include program codes, and the program codes include computer operating instructions.

[0094] In addition, the memory may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device or other volatile solid-state storage device.

[0095] Corresponding to the above method embodiment, the embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned method for self-diagnosis of abnormalities in safety monitoring data of hydraulic structures are implemented.

[0096] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0097] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures, characterized by: include: S10, acquiring monitoring data, arranging multiple variables of the monitoring data in chronological order to form a multivariate time series, and performing smoothing and normalization processing on the multivariate time series; S20, dynamic time warping algorithm was used to analyze the correlation between effect size and respective variables in the processed multivariate time series; S30. Based on the results of correlation analysis and time lag effect analysis, a multivariate regression model between effect size and independent variables was constructed; S40. Use multivariate local abnormal factors to identify abnormalities in effect size and independent variable measurements. Based on the local abnormal factor threshold, compare the distribution characteristics and occurrence time of abnormal values ​​reflected by the local abnormal factors from different time scales to determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, determine that it is a safety state abnormality caused by multiple factors, and identify the transformation mechanism of the structural state.

2. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 1, characterized in that: In step S20, the dynamic time warping algorithm uses the Sakoe-Chuba band constrained warping path when analyzing the correlation between the effect size and the respective variables in the multivariate time series.

3. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 2, characterized in that: The constraint rule of the Sakoe-Chuba belt constraint is that the regular path is along the main diagonal i = j At the same time, by the fixed width T 0∈ N To constrain the regular range, based on the Sakoe-Chuba band constraint, the global cost matrix D The calculation formula is: ； In the formula, D represents the total cost matrix, Represent the variables in two time series, respectively, where i ∈ [1, N ], j ∈ [1, M ], N , M There are two time series X , Y Length, Represents the given parameter.

4. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 1 or 2, characterized in that: In step S30, the method of constructing a multivariate regression model between effect size and independent variables includes: S31. Construct a data matrix based on independent variables and effect sizes, and establish a multiple linear regression model framework based on the variable relationships in the data matrix; S32, solving the unknown coefficients of the multiple linear regression model framework by the least square method; S33. Use the stepwise regression analysis method to screen the independent variables in the multiple linear regression model according to the degree of influence of the independent variables on the effect size, determine the independent variables with high influence and eliminate the independent variables with low influence, and obtain a simplified multiple linear regression model; S34. Based on the simplified multiple linear regression model, the relative contribution of each independent variable to the explained effect size variance is calculated through relative weights to determine the relative importance ranking of the independent variables; S35. Incorporate the relative importance ranking information of the independent variables and construct a multiple linear regression model between the effect size and the independent variables.

5. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 4, characterized in that: In step S40, the method of using multivariate local abnormal factors to identify abnormalities in effect size and independent variable measurements includes: S41. Integrate the independent variables and effect size time series to form a data set, construct a multivariate time series, and perform smoothing and balancing on the multivariate time series; S42, calculating the multivariate local anomaly factor for the processed multivariate time series, and determining a threshold value based on the calculation result of the multivariate local anomaly factor; S43, calculating the multivariate local abnormal factor values ​​of the multivariate time series again, and obtaining the effect size and the distribution of the local abnormal factor values ​​of the independent variables with strong correlation by comparison; S44, comparing the local anomaly factor value of each data point with a threshold value, if the local anomaly factor value of the data point is greater than the threshold value, the data point is an outlier; S45. Compare the distribution characteristics and occurrence time of the abnormal values ​​reflected by the local abnormal factors from different time scales, and determine whether the effect size and the local abnormal factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state abnormality.

6. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 5, characterized in that: In step S42, a threshold value of the local anomaly factor value is determined by a kernel density estimation method.

7. A method for self-diagnosis of abnormality in safety monitoring data of hydraulic structures according to claim 4, characterized in that: After obtaining the abnormal information of the structural state, the causes of the abnormal structural effect size and the mechanism of the structural state transformation are analyzed based on the abnormal information obtained by multivariate local anomaly identification.

8. A hydraulic structure safety monitoring data abnormality self-diagnosis system, characterized by: include: A data processing module, which is used to obtain monitoring data, arrange multiple variables of the monitoring data in chronological order to form a multivariate time series, and perform smoothing and normalization processing on the multivariate time series; A correlation analysis module, which is used to analyze the correlation between the effect size and the respective variables in the processed multivariate time series using a dynamic time warping algorithm; Model building module, which builds a multiple regression model between effect size and independent variables based on the results of correlation analysis and time lag effect analysis; The structural anomaly identification module uses multivariate local anomaly factors to identify anomalies in effect sizes and independent variable measurements. Based on the local anomaly factor threshold, it compares the distribution characteristics and occurrence time of the anomaly values ​​reflected by the local anomaly factors from different time scales, and determines whether the effect size and the local anomaly factors of the independent variables exceed the threshold at the same time. If so, it is determined to be a safety state anomaly caused by multiple factors, and the structural state transformation mechanism is identified.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images