Steel Structure Safety Damage Prediction Method and System

Abstract

The present invention relates to the technical field of steel structure damage detection, and particularly to a method and system for predicting steel structure safety damage. The method includes: obtaining a historical data set of the steel structure and preprocessing the historical data set, where the historical data set includes historical environmental data, historical load data, historical damage data, and historical maintenance data; constructing a propensity score matching model to extract causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and load conditions; constructing a steel structure equation relationship model to extract the causal effect estimation results of exogenous variables of the steel structure on endogenous variables according to the causal relationship factors; predicting the damage development trend of the steel structure under the current conditions based on the causal effect estimation results; obtaining a real-time data set of the steel structure, extracting the power spectral density of the damaged area and the undamaged area of the steel structure, and optimizing the causal effect estimation model. The purpose is to improve the accuracy of damage prediction for large steel structures such as bridges and elevator shaft installations.

Description

Technical Field

[0001] The present invention relates to the technical field of steel structure damage detection, and specifically to a method and system for predicting the safety damage of steel structures. Background Art

[0002] In the modern engineering field, large steel structures, as the core load-bearing systems of key infrastructure such as elevator shafts, cranes, buildings, bridges, and large stadiums, are widely used in various engineering projects. These steel structures are exposed to complex and variable natural environments and dynamic time-varying load effects for a long time. As the service life increases, the material properties gradually deteriorate, and the structure will inevitably suffer various forms of damage, such as crack initiation and propagation, material corrosion, fatigue damage, etc. Accurately predicting the damage state of steel structures and formulating scientific and reasonable maintenance strategies in advance are of crucial significance for ensuring the structural safety of steel structures, extending their service life, and ensuring the normal operation of infrastructure. Currently, for steel structure damage prediction, a variety of technical means have been developed. For example, the method based on vibration modal analysis identifies potential damage by monitoring changes in the vibration characteristics of the structure; the method based on strain monitoring uses sensors to collect strain data of key parts of the structure in real time, and then evaluates the mechanical properties and damage conditions of the structure.

[0003] However, the existing technologies still expose many limitations in practical applications, making it difficult to meet the actual engineering requirements for the prediction accuracy of large steel structure damage. On the one hand, most existing methods have obvious deficiencies in data utilization, only focusing on single or a few types of data, such as simply relying on load monitoring data or only relying on vibration response analysis, and failing to comprehensively consider the complex influence of historical data on the damage evolution process of steel structures, making it difficult to accurately depict the true service state of steel structures. On the other hand, the existing technologies lack in mining causal relationships, failing to deeply analyze the internal connections between various influencing factors, making it difficult to accurately evaluate the true influence degree of exogenous variables on the endogenous variable of steel structure damage. As a result, when constructing a prediction model, the interaction of various factors cannot be fully considered, and thus a large deviation occurs in the damage trend prediction process. Especially for large steel structures with complex engineering structures and harsh service environments, the existing technologies are difficult to accurately predict the development trend of damage and cannot meet the strict requirements of actual engineering for steel structure safety assurance. Summary of the Invention

[0004] In order to improve the prediction accuracy of large steel structure damage, the present invention provides a method and system for predicting the safety damage of steel structures. The specific technical solutions adopted are as follows:

[0005] The technical solution of the first aspect of the present invention provides a method for predicting the safety damage of steel structures, and the method includes:

[0006] Obtain the historical steel structure dataset and preprocess the historical dataset, where the historical dataset includes historical environmental data, historical load data, historical damage data, and historical maintenance data;

[0007] Construct a propensity score matching model to extract the causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and load conditions;

[0008] Construct a steel structure equation relationship model to extract the causal effect estimation results of exogenous variables on endogenous variables of the steel structure according to the causal relationship factors;

[0009] Predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results;

[0010] Obtain the real-time steel structure dataset, extract the power spectral density of the damaged area and the undamaged area of the steel structure, and optimize the causal effect estimation model.

[0011] Furthermore, construct a propensity score matching model to extract the causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and load conditions, including:

[0012] Based on the preprocessed historical dataset, calculate the propensity scores for maintaining the steel structure under different conditions;

[0013] According to the propensity scores, use the kernel function matching method to extract the non-maintained samples corresponding to the maintained samples under the same environmental and load conditions;

[0014] Use a Bayesian network to extract the causal relationship factors of steel structure maintenance on steel structure damage.

[0015] Furthermore, the expression for calculating the propensity scores for maintaining the steel structure under different conditions is:

[0016]

[0017] In the formula, represents the propensity score for maintaining the steel structure under the given condition ; represents the mapping function; represents the total number of steel structure characteristics considered for calculating the propensity scores; represents the th preset weight of the feature; represents the th feature function under the given condition ; represents the discount factor at time ; represents the environmental load state at time Take maintenance decisions The reward value; Represents a time-varying adjustment coefficient; Represents the time step.

[0018] Furthermore, construct a steel structure equation relationship model, and extract the causal effect estimation results of the exogenous variables of the steel structure on the endogenous variables according to the causal relationship factors, including:

[0019] Determine the exogenous variables and endogenous variables of the steel structure equation relationship model according to the causal relationship factors;

[0020] Use the maximum likelihood estimation method to estimate the causal effect strength between the exogenous variables and the endogenous variables, and establish a steel structure equation relationship model between the exogenous variables and the endogenous variables;

[0021] Based on the causal effect strength, extract the causal effect estimation results of the exogenous variables of the steel structure on the endogenous variables.

[0022] Furthermore, use the maximum likelihood estimation method to estimate the causal effect strength between the exogenous variables and the endogenous variables, and establish a steel structure equation relationship model between the exogenous variables and the endogenous variables, including:

[0023] Establish a linear relationship between the endogenous variable vector and the exogenous variable vector;

[0024] Calculate the likelihood function and the log-likelihood function according to the linear relationship between the endogenous variable vector and the exogenous variable vector;

[0025] Solve the log-likelihood function to obtain the estimated values of the intercept term, the exogenous variable coefficient vector, and the error term variance.

[0026] Furthermore, predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results, including:

[0027] Obtain the current environmental data, load data, and maintenance data of the steel structure, and input them into the steel structure equation relationship model;

[0028] According to the causal effect estimation values of the variables in the steel structure equation relationship model, predict the changes in the damage degree and damage development speed of the steel structure over time under the current conditions.

[0029] Furthermore, obtain the real-time data set of the steel structure, extract the power spectral density of the damaged area and the undamaged area of the steel structure, and optimize the causal effect estimation model, including:

[0030] Obtain the state information of the damaged area and the undamaged area of the steel structure;

[0031] Calculate the power spectral density of the signals in the damaged area and the undamaged area;

[0032] Based on a neural network, optimize the causal effect estimation model according to the power spectral density;

[0033] Use the optimized causal effect estimation model to extract the final damage development trend.

[0034] The technical solution of the second aspect of the present invention provides a steel structure safety damage prediction system, which adopts the steel structure safety damage prediction method described in the technical solution of the first aspect of the present invention. The system includes:

[0035] A data acquisition module configured to acquire a historical dataset of the steel structure and preprocess the historical dataset. The historical dataset includes historical environmental data, historical load data, historical damage data, and historical maintenance data;

[0036] A causal relationship analysis module configured to construct a propensity score matching model and extract the causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and load conditions;

[0037] A causal effect estimation module configured to construct a steel structure equation relationship model and extract the causal effect estimation results of exogenous variables on endogenous variables of the steel structure according to the causal relationship factors;

[0038] A damage trend prediction module configured to predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results;

[0039] An optimization module configured to acquire a real-time dataset of the steel structure, extract the power spectral density of the damaged area and the undamaged area of the steel structure, and optimize the causal effect estimation model.

[0040] The technical solution of the third aspect of the present invention provides an electronic device, which includes: a processor and a memory communicatively connected to the processor; wherein, the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can execute the steps of the steel structure safety damage prediction method described in the technical solution of the first aspect of the present invention.

[0041] The technical solution of the fourth aspect of the present invention provides a computer-readable storage medium, on which a program for implementing the steel structure safety damage prediction method is stored, and the program for implementing the steel structure safety damage prediction method is executed by a processor to implement the steps of the steel structure safety damage prediction method described in the technical solution of the first aspect of the present invention.

[0042] The present invention has the following beneficial effects:

[0043] The steel structure safety damage prediction method provided by the present invention constructs a propensity score matching model to accurately extract the causal relationship factors of steel structure maintenance on damage under the same environment and load conditions, realizing the extraction of the internal connection between steel structure maintenance and damage. By constructing a structural equation relationship model, the causal effect estimation result of exogenous variables on endogenous variables is obtained based on the causal relationship factors, thereby providing a more scientific quantitative basis for steel structure damage prediction. Finally, the damage development trend under the current conditions is predicted, and the causal effect estimation model is optimized by extracting the power spectral density of damaged and undamaged areas, enabling the model to continuously adjust according to the real-time state of the steel structure, continuously correct the prediction deviation, always maintain a high prediction accuracy, effectively improve the damage prediction accuracy of large steel structures, and ensure the safe and stable operation of steel structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0045] Figure 1 It is the method flow chart of the steel structure safety damage prediction method provided by an embodiment of the present invention;

[0046] Figure 2 It is the structural schematic diagram of the steel structure safety damage prediction system provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of a steel structure safety damage prediction method and system according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0049] The following specifically describes the specific solutions of a steel structure safety damage prediction method and system provided by the present invention in combination with the drawings.

[0050] Please refer to Figure 1, which shows the method flow chart of the steel structure safety damage prediction method provided by an embodiment of the present invention. The method includes:

[0051] Step S100: Obtain the historical data set of the steel structure and preprocess the historical data set. The historical data set includes historical environmental data, historical load data, historical damage data, and historical maintenance data. Specifically, in this embodiment, a large steel structure bridge built for 10 years in a certain city has suffered different degrees of damage and been maintained due to heavy traffic flow and humid climate in the past. In this embodiment, a historical data set of the environment, load, maintenance, and damage for 12 consecutive months is collected. Specifically, the number, type, and driving speed of passing vehicles can be counted through traffic flow monitoring equipment to obtain historical load data. Non-destructive testing techniques, such as ultrasonic flaw detectors, are used to regularly detect key nodes and components of the bridge, record the crack length, width, and position, and measure the corrosion degree of steel through corrosion detection equipment to collect historical damage data. At the same time, historical maintenance data such as maintenance time, maintenance method, and materials used for maintenance are extracted from the bridge maintenance files. In this embodiment, at least cleaning is performed on the historical data set to remove duplicates, errors, and missing values. The preprocessed data is sorted according to time sequence and data type to construct a historical data matrix of the steel structure. The rows of the historical data matrix represent different time points, and the columns represent various environmental, load, damage, and maintenance data.

[0052] Through comprehensive and systematic data collection and preprocessing in this embodiment, these data can truly reflect the damage evolution process of the steel structure under complex environments and loads and the effect of maintenance intervention.

[0053] Step S200: Construct a propensity score matching model to extract the causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and load conditions.

[0054] Step S200 specifically includes:

[0055] Step S210: Based on the preprocessed historical data set, calculate the propensity score for maintaining the steel structure under different conditions. Specifically, after completing the preprocessing of the historical data set in step S100 to obtain a high-quality historical data matrix, obtain the environmental load state, maintenance decision, and damage state change in the target period.

[0056] Step S211: In this embodiment, considering the influence of maintenance cost, damage type, and environmental factors, a reward function is constructed, which can be expressed as:

[0057]

[0058] In the formula, represents the environmental load state when taking the maintenance decision After that, according to the change of the damage state The obtained reward value; Indicates the environmental impact coefficient dynamically determined by fuzzy logic based on environmental data; And Respectively represent the weights of crack and corrosion damage, reflecting the importance of different damage types to the structural safety; Indicates the cost adjustment coefficient, which is used to balance the maintenance cost and the damage improvement effect; Indicates the welding repair cost, Indicates the coating protection cost; Indicates the change in crack length, that is, the difference in crack length between this detection and the previous detection; Indicates the change in corrosion area, that is, the difference in corrosion area between this detection and the previous detection;

[0059] Step S212: Based on reinforcement learning, update the action value function, which can be expressed as:

[0060]

[0061] In the formula, Indicates the learning rate, and its value range is between 0 and 1, which is used to control the update degree of new experience to the old estimate; Indicates the discount factor, and its value range is between 0 and 1, which is used to balance the importance of recent rewards and long-term rewards; Indicates The maximum action value among all possible maintenance decisions at time

[0062] Step S213: Based on the action value function, the expression for calculating the propensity score for maintaining the steel structure under different conditions is:

[0063]

[0064] In the formula, Indicates the given condition The propensity score for maintaining the steel structure under, including at least the environmental load state and the damage state; Indicates the mapping function, which is used to map the score to the interval from 0 to 1; Indicates the total number of steel structure characteristics considered for calculating the propensity score; Indicates the th preset weight of the feature; Indicates the given condition Under the th feature function; Indicates The discount factor at time, which is used to reflect the importance difference of reward values at different times; Represents The environmental load state at a certain moment Maintenance decisions are made The reward value; Represents the adjustment coefficient that changes over time; Represents the number of time steps;

[0065] The expression of the adjustment coefficient is:

[0066]

[0067] In the formula, Represents the expected reward value, which can be set according to historical experience or engineering goals;

[0068] Step S220: According to the propensity score, use the kernel function matching method to extract the unmaintained samples corresponding to the maintained samples under the same environment and load conditions; specifically, for each maintained sample, its propensity score is , in this embodiment, it is necessary to find matching samples among the unmaintained samples to maximize the kernel function value, and the kernel function can be expressed as:

[0069]

[0070] In the formula, Represents the kernel function, which is used to measure the similarity of the propensity scores of two samples; Is the bandwidth parameter, which can be determined by cross-validation; in this way, corresponding unmaintained samples can be found for each maintained sample under the same environment and load conditions, providing a suitable control sample for subsequent causal relationship analysis;

[0071] Step S230: Use a Bayesian network to extract the causal relationship factors of steel structure maintenance on steel structure damage; specifically, in this embodiment, a Bayesian network is constructed, where the nodes include environmental factors, load factors, maintenance factors, and damage factors; using the preprocessed historical data and the matched sample data, a method combining maximum likelihood estimation and Bayesian estimation is used to estimate the network parameters; for nodes X and Y, their conditional probability distribution Is updated through data statistics and Bayes' formula:

[0072]

[0073] In the formula, Represents the conditional probability that Y occurs under the condition that X occurs; Represents the conditional probability that X occurs under the condition that Y occurs; Represents the prior probability that Y occurs; Represents the prior probability that X occurs;

[0074] Taking a large steel structure bridge as an example, the Bayesian network constructed in this embodiment includes the following nodes:

[0075] Environmental factors (X1): including temperature, humidity, and wind force;

[0076] Load factors (X2): such as traffic flow and vehicle weight;

[0077] Maintenance factors (X3): maintenance methods (welding repair, coating protection), and maintenance frequency;

[0078] Damage factors (Y): crack length, corrosion area;

[0079] In this embodiment, by continuously updating the conditional probability distribution between nodes, the causal relationship between various factors can be deeply analyzed. For example, through calculation, it is found that , that is, the probability of damage occurring under the condition of taking a certain maintenance measure is significantly different from , which indicates that the maintenance factor has an important causal impact on the damage factor. If is significantly less than , it indicates that this maintenance measure has a positive effect on reducing damage; otherwise, it means that the maintenance measure is not effective. In this way, we can accurately extract the causal relationship factors of steel structure maintenance on steel structure damage, providing a more scientific basis for subsequent damage prediction and maintenance decision-making;

[0080] In summary, in this embodiment, by constructing a propensity score matching model, considering maintenance costs, damage types, and environmental factors comprehensively, the propensity scores for maintaining the steel structure under different conditions are accurately calculated. The kernel function matching method is used to extract control samples under the same environmental and load conditions, providing a reliable data basis for subsequent causal relationship analysis. The Bayesian network is used to mine the causal relationship factors between steel structure maintenance and damage, enabling in-depth understanding of the influence mechanism of various factors on steel structure damage. It effectively improves the analysis accuracy of the causal relationship of steel structure damage, laying a foundation for subsequent construction of an accurate structural equation relationship model and prediction of damage development trends.

[0081] Step S300: Construct a steel structure equation relationship model, and extract the causal effect estimation results of exogenous variables of the steel structure on endogenous variables according to the causal relationship factors;

[0082] Step S300 specifically includes:

[0083] Step S310: Determine the exogenous variables and endogenous variables of the steel structure equation relationship model according to the causal relationship factors; specifically, according to the causal relationship obtained from the Bayesian network analysis, determine the exogenous variables and endogenous variables. For example:

[0084] Exogenous variables include:

[0085] Environmental factors: temperature, humidity, concentration of atmospheric corrosion medium; for example, in a high-temperature and humid environment, steel structures are more likely to corrode;

[0086] Load factors: daily traffic flow, passing frequency and weight of heavy vehicles; for example, frequent passing of heavy-load vehicles will increase the stress of steel structures and accelerate the development of damage;

[0087] Maintenance factors: maintenance methods (such as welding repair, coating protection, etc.), maintenance frequency; for example, regular maintenance can effectively delay the damage of steel structures;

[0088] Endogenous variables:

[0089] Degree of damage: measured by crack length and corrosion area indicators; for example, the longer the crack length, the more serious the degree of damage of the steel structure;

[0090] Damage development rate: calculated by the ratio of the damage index to the time interval in two adjacent detections;

[0091] Step S320: Use the maximum likelihood estimation method to estimate the causal effect strength between exogenous variables and endogenous variables, and establish a steel structure equation relationship model between exogenous variables and endogenous variables; specifically, first establish a linear relationship between endogenous variables and exogenous variables:

[0092]

[0093] In the formula, represents the endogenous variable vector, which is a two-dimensional vector, represents the degree of damage, represents the damage development rate; represents the exogenous variable vector; represents the intercept term, which is a constant; is the coefficient vector of exogenous variables, represents the number of exogenous variables, is the error term, which follows a normal distribution with a mean of 0 and a variance of ;

[0094] Calculate the likelihood function and log-likelihood function according to the linear relationship between the endogenous variable vector and the exogenous variable vector. The likelihood function can be expressed as:

[0095]

[0096] In the formula, represents the sample size; represents the th sample value of the endogenous variable; represents the th sample value of the exogenous variable;

[0097] The logarithmic likelihood function can be expressed as:

[0098]

[0099] Solve the logarithmic likelihood function to obtain the estimated values of the intercept term, the coefficient vector of exogenous variables, and the variance of the error term; specifically, by taking the partial derivatives of the logarithmic likelihood function and setting them to 0, the solutions are obtained for , and These three parameters can be specifically solved using the Newton-Raphson method;

[0100] Step S330: Based on the causal effect strength, extract the causal effect estimation results of the exogenous variables of the steel structure on the endogenous variables; specifically, the obtained is the estimated value of the causal effect strength of the exogenous variable on the endogenous variable. For example, if is the coefficient corresponding to the exogenous variable of temperature, when , it indicates that an increase in temperature will lead to an increase in the degree of damage or the rate of damage development; when , it indicates that an increase in temperature will cause a decrease in the degree of damage or the rate of damage development; The larger the value, the greater the influence of temperature on damage. Based on these estimated values, the direction and degree of the influence of exogenous variables on endogenous variables can be comprehensively analyzed;

[0101] In summary, in this embodiment, by constructing a steel structure equation relationship model, the influences of exogenous variables such as environmental factors, load factors, and maintenance factors on endogenous variables such as the degree of damage and the rate of damage development are comprehensively considered. The causal effect strength between exogenous variables and endogenous variables is accurately estimated using the maximum likelihood estimation method, which can clearly reveal the direction and degree of the influence of each exogenous variable on the damage of the steel structure. This helps to deeply understand the formation mechanism of large steel structure damage, and thus can more accurately consider the comprehensive influence of various factors when predicting damage, avoiding the prediction deviation caused by insufficient mining of causal relationships in the prior art, and greatly improving the prediction accuracy. The constructed structural equation relationship model can predict damage based on the dynamic changes of exogenous variables and adapt to different service stages and environments. For example, when the environmental temperature rises or the traffic flow increases, the model can accurately predict the corresponding changes in the degree of damage and the rate of damage development of the steel structure according to the estimated causal effect strength. Compared with traditional static prediction methods, this dynamic prediction model can better adapt to the damage changes of large steel structures in different service stages and different environmental conditions, thereby improving the prediction accuracy and reliability. Based on the causal effect estimation results, it can also scientifically guide maintenance decisions, take targeted measures to delay the development of damage, and ensure prediction accuracy and structural safety through multi-faceted coordination.

[0102] Step S400: Predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results;

[0103] Step S400 specifically includes:

[0104] Step S410: Obtain the current environmental data, load data, and maintenance data of the steel structure and input them into the steel structure equation relationship model; specifically, organize these obtained current environmental data, load data, and maintenance data into vector form and input them as exogenous variable data into the steel structure equation relationship model established in Step S300;

[0105] Step S420: Predict the changes in the damage degree and damage development speed of the steel structure over time under the current conditions according to the causal effect estimation values of the variables in the steel structure equation relationship model;

[0106] Specifically, the steel structure equation relationship model can be expressed as:

[0107]

[0108] Among them, is The predicted value of the damage degree or damage development speed at time ; for the damage degree, indicators such as crack length and corrosion area can be used to measure it; for the damage development speed, it can be the crack propagation length per unit time or the increase in corrosion area; is The exogenous variable at time ; Represents the intercept term estimation value obtained by maximum likelihood estimation; Represents the coefficient vector estimation value of the exogenous variable; for example: obtained through Step S300 ; , corresponding to the three exogenous variables of temperature, traffic flow, and maintenance frequency respectively; the current , corresponding to a temperature of 35°C, a traffic flow of 100 vehicles per hour, and the most recent maintenance being 1 month ago, then can be calculated , this value represents the currently predicted damage degree;

[0109] Secondly, considering the dynamics of damage development in this embodiment, an autoregressive integrated moving average model (ARIMA) is introduced to correct the prediction results, and its expression is:

[0110]

[0111] Among them, Represents the difference operator, is the order of differencing, used to make the time series data stationary; Represents the autoregressive order, indicating the number of past values of itself used in the model; denotes the order of the moving average, representing the number of past error terms used in the model; and respectively denote the estimated parameters of the first model and the second model, denotes the error term, which follows a normal distribution with a mean of 0; denotes the constant term; in this embodiment, in order to determine the parameters of the ARIMA model 、 、 , the grid search method combined with the AIC (Akaike Information Criterion) criterion can be used for model selection. Based on the historical damage data, different parameter combinations are tried, and the combination that makes the AIC value the smallest is selected as the optimal parameter, and then the corrected predicted value is obtained;

[0112] In this embodiment, by combining the steel structure equation relationship model and time series analysis, the accuracy and reliability of large steel structure damage prediction are significantly improved. First, the current environmental, load, and maintenance data are input into the equation relationship model for preliminary prediction, fully considering the real-time impact of exogenous variables on damage, making the prediction results more in line with the actual situation. Then, the ARIMA model is introduced to correct the prediction results, effectively capturing the dynamics and time series characteristics of damage development, and making up for the deficiency of the equation relationship model in dealing with dynamic changes. This comprehensive prediction method can more accurately predict the degree of steel structure damage and the change of damage development speed over time, provide a more scientific and timely decision-making basis for the maintenance and management of steel structures, help take measures in advance to prevent damage from further deteriorating, ensure the safe and stable operation of large steel structures, and reduce maintenance costs and safety risks.

[0113] Step S500: Obtain the real-time steel structure data set, extract the power spectral density of the damaged area and the undamaged area of the steel structure, and optimize the causal effect estimation model;

[0114] Step S500 specifically includes:

[0115] Step S510: Obtain the status information of the damaged area and the undamaged area of the steel structure; specifically, use ultrasonic testing equipment to detect the key connection parts, welds and other areas that may be damaged of the steel structure. When ultrasonic waves encounter damage (such as cracks) during propagation, reflected waves will be generated. By analyzing the characteristics of the reflected waves, the location and approximate size of the damage can be determined. At the same time, use magnetic particle testing technology to detect the surface of the steel structure. When there are defects on the surface or near the surface of the steel structure, applying magnetic powder will form obvious magnetic marks at the defects, thus discovering the damage; then reasonably arrange strain sensors, displacement sensors, etc. in the damaged area and the undamaged area of the steel structure. The strain sensor can monitor the stress and strain conditions of the steel structure in real time, and the displacement sensor is used to measure the displacement change of the structure. These sensors will collect status information such as stress, strain, and displacement according to a certain sampling frequency to form a real-time data set.

[0116] Step S520: Calculate the power spectral density of the signals in the damaged area and the undamaged area; use wavelet packet transform to calculate the power spectral density. For the signals collected from the sensors , decompose the signal through wavelet packet decomposition to obtain the th wavelet packet coefficient . Wavelet packet decomposition can refine the analysis of the signal at different frequencies and time scales, and can better capture the local characteristics of the signal;

[0117] The th wavelet packet coefficient has the following expression for the power spectral density:

[0118]

[0119] where represents the signal length, represents the frequency. By calculating the power spectral density at different frequencies, the frequency characteristic distribution of the signals in the damaged area and the undamaged area can be obtained;

[0120] Step S530: Optimize the causal effect estimation model based on the neural network according to the power spectral density. Specifically, construct an MLP neural network with an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is equal to the number of power spectral density features of the damaged area and the undamaged area. For example, if the power spectral density is calculated at 10 different frequencies, then there are 20 neurons in the input layer. The number of neurons in the output layer is equal to the number of parameters of the causal effect estimation model. Normalize the calculated power spectral density features so that their value range is between [0, 1] to improve the training effect of the neural network. Then, use the real-time dataset to train the MLP neural network. During the training process, the backpropagation algorithm is used to update the network weights. By continuously adjusting the weights, make the output of the neural network as close as possible to the true adjustment amount of the parameters of the causal effect estimation model.

[0121] Step S540: Use the optimized causal effect estimation model to extract the final damage development trend. Specifically, in step S400, we constructed a preliminary prediction model. Now, recalculate using the optimized parameters. For example, input the current temperature, traffic flow, maintenance frequency, etc. data into the optimized model, and predict the final steel structure damage development trend according to the output of the model, including the changes in the damage degree and the damage development speed over time. We can plot the curves of the damage degree and the damage development speed changing with time to visually display the steel structure damage development trend.

[0122] In step S500, by obtaining the real-time state information of the steel structure, calculate the power spectral density of the damaged area and the undamaged area. The power spectral density can reflect the energy distribution of the signal at different frequencies and contains rich information about the damage state of the steel structure. The neural network has a powerful non-linear mapping ability and can learn the complex relationship between the power spectral density and the parameters of the causal effect estimation model. Compared with traditional linear models, it can capture more features and rules hidden in the data, thus more accurately adjusting the parameters of the causal effect estimation model, improving the accuracy of damage prediction, and ultimately significantly enhancing the accuracy and reliability of steel structure damage prediction.

[0123] As another implementation, this method is also applicable to the safety damage prediction of elevator steel structures. For example, the elevator operation management system records environmental data such as the temperature and humidity in the machine room, obtains the load as load data, the maintenance records provide maintenance data, and the regular inspection records can be used as damage data. When constructing the propensity score matching model, consider the elevator maintenance cost, different damage types (such as guide rail deformation, car bracket cracks), and the impact of machine room environmental factors on the maintenance effect and damage development. Then, it is reasonable and feasible to calculate the propensity score. Using the kernel function to match samples and the Bayesian network to extract causal relationship factors is also applicable to analyzing the relationship between elevator maintenance and damage. Construct a steel structure equation relationship model, set the environment, load, maintenance, etc. as exogenous variables, and set the damage degree and development speed of the elevator steel structure as endogenous variables. Use maximum likelihood estimation to establish the model and extract the causal effect estimation results. In the model optimization stage, obtain the state information of the damaged and undamaged areas of the elevator steel structure through detection equipment, calculate the power spectral density, and then optimize the model with a neural network, which can effectively improve the prediction accuracy and provide strong support for the safe operation and maintenance of the elevator.

[0124] In summary, the introduction of real-time data in this embodiment enables the model to timely reflect the latest state of the steel structure. The calculation of the power spectral density mines the frequency characteristics of the signal and provides richer information for model optimization. The application of the neural network can adaptively adjust the parameters of the causal effect estimation model to better adapt to the actual damage situation of the steel structure. The finally obtained optimized model can more accurately predict the damage development trend of the steel structure, provide a more scientific and timely basis for the maintenance decision of the steel structure, help ensure the safe operation of the steel structure, and reduce the maintenance cost and safety risks.

[0125] Please refer to Figure 2 , which shows a schematic structural diagram of a steel structure safety damage prediction system provided by the technical solution of the second aspect of the present invention. The system includes:

[0126] A data acquisition module configured to acquire a historical data set of the steel structure and preprocess the historical data set. The historical data set includes historical environmental data, historical load data, historical damage data, and historical maintenance data;

[0127] A causal relationship analysis module configured to construct a propensity score matching model and extract causal relationship factors of the steel structure maintenance on the steel structure damage under the same environmental and load conditions;

[0128] A causal effect estimation module configured to construct a steel structure equation relationship model and extract causal effect estimation results of exogenous variables of the steel structure on endogenous variables according to the causal relationship factors;

[0129] A damage trend prediction module configured to predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results;

[0130] An optimization module, configured to obtain a real-time steel structure data set, extract the power spectral density of the damaged area and the undamaged area of the steel structure, and optimize the causal effect estimation model.

[0131] The technical solution of the third aspect of the present invention provides an electronic device, which includes: a processor and a memory communicatively connected to the processor; wherein, the memory stores instructions executable by the processor, and when the instructions are executed by the processor, the processor can execute the steps of the steel structure safety damage prediction method according to the technical solution of the first aspect of the present invention.

[0132] The technical solution of the fourth aspect of the present invention provides a computer-readable storage medium, on which a program for implementing the steel structure safety damage prediction method is stored, and when the program for implementing the steel structure safety damage prediction method is executed by a processor, the steps of the steel structure safety damage prediction method according to the technical solution of the first aspect of the present invention are implemented.

[0133] It should be noted that: the above-mentioned sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0134] Each embodiment in this specification is described in a progressive manner. The same parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

Claims

1. A method for predicting safety damage of steel structures, characterized in that: The method comprises: Acquire and preprocess the historical data set of steel structures, where the historical data set includes historical environmental data, historical load data, historical damage data, and historical maintenance data; A propensity score matching model was constructed to extract the causal factors of steel structure maintenance on steel structure damage under the same environmental and load conditions, including: Based on the preprocessed historical data set, the propensity score for maintaining the steel structure under different conditions is calculated, and the expression is: In the formula, Indicates a given condition The propensity score for maintaining the steel structure is: represents a mapping function; represents the total number of steel structure characteristics considered for calculating the propensity score; Indicates The preset weights of the features; Indicates a given condition Next characteristic functions; express Discount factor for the moment; express Environmental load status at all times Maintenance decision making The reward value of represents the adjustment coefficient that changes over time; represents the number of time steps; According to the propensity score, the kernel function matching method is used to extract the unmaintained samples corresponding to the maintained samples under the same environmental and load conditions; Bayesian network is used to extract the causal factors of steel structure maintenance on steel structure damage; Construct a steel structure equation relationship model, and extract the causal effect estimation results of the steel structure exogenous variables on the endogenous variables based on the causal relationship factors; Predict the damage development trend of steel structures under current conditions based on the causal effect estimation results; Acquire real-time data sets of steel structures, extract the power spectrum density of damaged and undamaged areas of steel structures, and optimize the causal effect estimation model.

2. The method for predicting safety damage of steel structure according to claim 1, characterized in that: Construct a steel structure equation relationship model, and extract the causal effect estimation results of the steel structure exogenous variables on the endogenous variables based on the causal relationship factors, including: According to the causal factors, the exogenous and endogenous variables of the steel structure equation relationship model are determined; The maximum likelihood estimation method is used to estimate the causal effect strength between exogenous variables and endogenous variables, and a steel structure equation relationship model between exogenous variables and endogenous variables is established; Based on the strength of causal effect, the causal effect estimation results of exogenous variables of steel structure on endogenous variables are extracted.

3. The method for predicting safety damage of steel structure according to claim 2, characterized in that: The maximum likelihood estimation method is used to estimate the causal effect strength between exogenous variables and endogenous variables, and a steel structure equation relationship model between exogenous variables and endogenous variables is established, including: Establish a linear relationship between the endogenous variable vector and the exogenous variable vector; Calculate the likelihood function and log-likelihood function based on the linear relationship between the endogenous variable vector and the exogenous variable vector; Solving the log-likelihood function yields estimates of the intercept term, the coefficient vector of the exogenous variables, and the variance of the error term.

4. The method for predicting safety damage of a steel structure according to any one of claims 1 to 3, characterized in that: Based on the causal effect estimation results, the damage development trend of the steel structure under the current conditions is predicted, including: Obtain the current environmental data, load data and maintenance data of the steel structure, and input them into the steel structure equation relationship model; According to the estimated values ​​of the causal effects of each variable in the steel structure equation relationship model, the changes of the damage degree and damage development speed of the steel structure over time under the current conditions are predicted.

5. The method for predicting safety damage of steel structure according to claim 4, characterized in that: Acquire real-time data sets of steel structures, extract power spectrum density of damaged and undamaged areas of steel structures, and optimize causal effect estimation models, including: Obtain status information of damaged and undamaged areas of steel structures; Calculate the power spectral density of the signals in the damaged area and the undamaged area; Based on neural network, the causal effect estimation model is optimized according to the power spectrum density; The final damage development trend is extracted using the optimized causal effect estimation model.

6. Steel structure safety damage prediction system, characterized by: The method for predicting safety damage of a steel structure according to any one of claims 1 to 5 is adopted, wherein the system comprises: A data acquisition module is configured to acquire a historical data set of the steel structure and pre-process the historical data set, the historical data set including historical environmental data, historical load data, historical damage data, and historical maintenance data; A causal relationship analysis module, configured to build a propensity score matching model to extract the causal relationship factors of steel structure maintenance on steel structure damage under the same environmental and loading conditions; A causal effect estimation module is configured to construct a steel structure equation relationship model and extract the causal effect estimation results of the steel structure exogenous variables on the endogenous variables based on the causal relationship factors; A damage trend prediction module, configured to predict the damage development trend of the steel structure under the current conditions based on the causal effect estimation results; The optimization module is configured to obtain a real-time data set of the steel structure, extract the power spectral density of the damaged area and the undamaged area of ​​the steel structure, and optimize the causal effect estimation model.

7. An electronic device, characterized in that: The electronic device includes: a processor and a memory communicatively connected to the processor; wherein the memory stores instructions executable by the processor, and the instructions are executed by the processor so that the processor can execute the steps of the steel structure safety damage prediction method described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program for implementing the method for predicting safety damage to a steel structure, and the program for implementing the method for predicting safety damage to a steel structure is executed by a processor to implement the steps of the method for predicting safety damage to a steel structure as described in any one of claims 1 to 5.

Images