Durability evaluation method for slope protection layer of flood discharge atomization area
By combining dynamic monitoring and probabilistic modeling with multi-field coupled numerical simulation and intelligent algorithms, the problems of static load assumptions and insufficient simulation accuracy in the durability assessment of the protective layer in the flood discharge atomization zone were solved, realizing dynamic assessment of the protective layer performance and scientific maintenance decision support.
Patent Information
- Application Number
- CN202511346290.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-19
- Publication Date
- 2026-01-23
AI Technical Summary
Existing technologies have problems in assessing the durability of the protective layer in the flood discharge atomization zone, such as damage prediction bias caused by static load assumptions, insufficient simulation accuracy of multi-physics coupling effects, and lack of time-varying reliability assessment.
By monitoring and acquiring dynamic load parameters, performing probabilistic statistical analysis of load characteristics, conducting multi-field coupled numerical simulation analysis, performing intelligent prediction of durability damage, and assessing time-varying reliability, and combining distributed sensors, CFD-DEM models, and LSTM networks, an intelligent prediction model of the dynamic load-material performance degradation relationship is established to quantify the remaining service life of the structure.
It enables accurate assessment of the durability of the protective layer, breaks through the limitations of traditional methods, improves the accuracy of load characteristic description and impact process simulation, and provides a scientific basis for maintenance decisions.
Smart Images

Figure CN121389230A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of water conservancy engineering, and in particular to a durability evaluation method for a slope protection layer of a flood discharge atomization area. BACKGROUND
[0002] In the operation of large-scale water conservancy projects, the high-frequency water particle impact generated by flood discharge atomization is a key environmental load that induces damage to the protection structure, and the durability evaluation thereof is directly related to the safe life cycle management of major infrastructure.
[0003] There are significant deficiencies in the durability evaluation of the protection layer in the pre-flood discharge atomization area: first, the traditional method uses static load assumptions and ignores the amplitude-frequency joint variability in the actual flood discharge process, resulting in damage prediction bias; second, the simulation precision of multi-physical field coupling is insufficient and relies too much on empirical formula simplification; and third, there is a lack of a time-varying reliability evaluation system that integrates long-term monitoring data and intelligent algorithms, making it difficult to provide quantitative basis for differentiated maintenance strategies. SUMMARY
[0004] The main purpose of the present application is to provide a durability evaluation method for a slope protection layer of a flood discharge atomization area to solve the problems raised in the background art.
[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions: A durability evaluation method for a slope protection layer of a flood discharge atomization area, the specific steps being as follows: S1, dynamic load parameter monitoring and collection, by arranging distributed stress sensors, accelerometers and environmental monitoring devices, real-time collection of dynamic load parameters received by the protection layer under flood discharge conditions, and establishment of a long-term monitoring database; S2, load characteristic probability statistical analysis, based on the monitoring data, using the extreme value theory and the Monte Carlo simulation method, establishing a joint probability distribution model of the amplitude-frequency of the dynamic load, and extracting a typical load spectrum through the rain flow counting method; S3, multi-field coupled numerical simulation analysis, constructing a CFD-DEM coupled calculation model, considering fluid dynamics and discrete medium interaction at the same time, simulating the dynamic impact process of atomized water particles on the protection layer, and analyzing the transient stress response characteristics of the protection layer; S4, intelligent prediction of durability damage, using a long short-term memory neural network algorithm, taking the load spectrum characteristic parameters and material performance parameters as input variables, and establishing an intelligent prediction model of the dynamic load-material performance degradation relationship; S5, time-varying reliability evaluation, based on the Gamma process to establish a performance degradation model of the protection layer, combining with a set limit state equation to calculate the reliability index at different service time points, and quantifying the remaining service life of the structure; S6. Maintenance decision optimization output: Based on the reliability assessment results, automatically generate a graded maintenance recommendation plan.
[0006] Preferably, in step S1, the specific steps for monitoring and collecting dynamic load parameters are as follows; S1.1 Layout and calibration of monitoring devices: Distributed stress sensors, triaxial accelerometers and temperature and humidity sensors are deployed at key locations in the flood discharge atomization impact zone. Calibration experiments are conducted to ensure that the sensor range, accuracy and acquisition frequency are adapted to the characteristics of the flood discharge load. S1.2 Multi-source data synchronous acquisition: Dynamic load parameters are collected in real time during flood discharge, and the working period is marked in combination with flood discharge scheduling records; S1.3 Database construction and preprocessing: Time-domain alignment, noise filtering and outlier removal are performed on the raw data, and multi-source data are integrated according to timestamps to construct a structured long-term monitoring database.
[0007] Preferably, in step S2, the specific method for the probabilistic statistical analysis of load characteristics is as follows; S2.1 Monitoring data feature extraction and preprocessing: The dynamic load time history data in the long-term monitoring database is subjected to stationarity test and segmentation processing. Key statistical features such as load amplitude, frequency of action, and duration under each working condition are extracted. The influence of units is eliminated through data standardization, providing standardized input for probabilistic modeling. S2.2 Construction of joint probability distribution model: The generalized extreme value distribution is used to fit the maximum load amplitude distribution to determine the return period of extreme loads. Random samples are generated based on the load characteristic statistical parameters to construct the amplitude-frequency joint probability distribution model. The Kolmogorov-Smirnov test is used to verify the good agreement between the distribution model and the measured data. S2.3 Typical load spectrum generation: Decompose the non-stationary load time history into semi-cyclic events, statistically analyze the amplitude-mean-frequency matrix, merge small amplitude cycles with low damage contribution, generate simplified but equivalent typical load spectra, and output multiple sets of representative load spectra according to the flood discharge conditions.
[0008] Preferably, in step S2.2, the joint probability distribution model is constructed in the following specific way; S2.21. Extreme value statistics and distribution fitting: Extract the maximum load amplitude sequence of each time period from the preprocessed monitoring data, use the generalized extreme value distribution for parameter estimation, determine the shape parameter, scale parameter and location parameter, and calculate the predicted extreme load values corresponding to different return periods. S2.22. Based on the amplitude and frequency edge distribution characteristics of measured data, construct a Copula function to describe its correlation structure, use the Monte Carlo method to generate a random sample set that conforms to the joint distribution characteristics, establish a two-dimensional joint probability density function of amplitude and frequency, and determine the probability of occurrence of different combination conditions. S2.23. Apply the Kolmogorov-Smirnov test to compare the difference between the cumulative distribution function of the model prediction and the measured data, calculate the maximum deviation D statistic, determine the goodness of fit, and iteratively adjust the distribution type or parameters until the test is passed.
[0009] Preferably, in step S2.3, the typical load spectrum is generated in the following specific way; S2.31 Load time history cyclic feature decomposition: Rainflow counting method is used to identify the cyclicity of non-stationary load time history, extract the amplitude, mean and frequency of each cyclic event, and generate an initial amplitude-mean-frequency statistical matrix. S2.32. Calculate the damage contribution of each cycle based on Miner's linear cumulative damage theory, set the damage threshold, merge cycles with small amplitudes and keep the total damage equivalent. S2.33. Associate the flood discharge operation parameters, cluster them according to the operating conditions to generate characteristic load spectra, and store them in a standardized format for subsequent analysis input.
[0010] Preferably, in step S3, the specific method of multi-field coupling numerical simulation analysis is as follows; S3.1 Multiphysics Model Construction: Establish the CFD model of the atomized water phase and the DEM model of the protective layer, and set the data exchange protocol for the coupling interface. S3.2 Coupled calculation and dynamic process simulation: Input typical load spectrum as boundary conditions, solve the fluid-discrete medium interaction equation, and record the time history data of the stress / displacement field of the protective layer during the impact process; S3.3 Response characteristic analysis and verification: Extract key transient response peaks, verify simulation accuracy by comparing with high-speed camera measured data, and generate stress concentration area cloud map and damage evolution curve.
[0011] Preferably, in step S4, the specific method for intelligent prediction of durability damage is as follows; S4.1 Dataset construction and feature engineering: Integrate load spectrum parameters and material performance parameters, perform feature standardization, missing value filling and outlier detection, and divide the training set, validation set and test set according to time series; S4.2 LSTM Model Construction and Training: Design the network structure, set the time-series sliding window, define the loss function and optimizer, and iteratively optimize the weight parameters through backpropagation; S4.3 Model Validation and Degradation Prediction: Use test sets to evaluate prediction accuracy, use sensitivity analysis to determine key influencing parameters, and input new load conditions to output material performance degradation curves.
[0012] Preferably, in step S5, the time-varying reliability assessment is carried out in the following specific way; S5.1 Performance degradation trajectory modeling: Based on damage prediction data, construct a Gamma process model, determine the time-varying characteristics of shape and scale parameters, and establish a degradation stochastic process expression that considers measurement uncertainties; S5.2 Limit state analysis: Define key performance thresholds, establish limit state equations that include degradation, and use the First-Order Reliability Method to calculate time-varying reliability indices. S5.3 Remaining lifetime probability prediction: Set an acceptable minimum reliability threshold, predict the first exceedance probability through Monte Carlo simulation, and output the remaining lifetime range at different confidence levels. Attached Figure Description
[0013] Figure 1 This is a flowchart illustrating the steps of the method described in this application. Detailed Implementation
[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0015] The terms "first," "second," and "third" in this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. All directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationships and movements between components in a specific orientation (as shown in the figures). If the specific orientation changes, the directional indications also change accordingly. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.
[0016] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0017] Example 1: Please refer to Figure 1 A method for evaluating the durability of slope protection layers in flood discharge atomization zones, the specific steps of which are as follows: S1. Dynamic load parameter monitoring and acquisition: By deploying distributed stress sensors, accelerometers and environmental monitoring devices, the dynamic load parameters of the protective layer under flood discharge conditions are collected in real time, and a long-term monitoring database is established. S2. Load characteristic probabilistic statistical analysis: Based on monitoring data, extreme value theory and Monte Carlo simulation method are used to establish a joint probability distribution model of dynamic load amplitude and frequency, and typical load spectrum is extracted by rainflow counting method. S3. Multi-field coupled numerical simulation analysis: Construct a CFD-DEM coupled calculation model, and simultaneously consider the interaction between fluid dynamics and discrete media to simulate the dynamic impact process of atomized water particles on the protective layer and analyze the transient stress response characteristics of the protective layer. S4. Intelligent prediction of durability damage: It adopts a long short-term memory neural network algorithm, takes load spectrum characteristic parameters and material performance parameters as input variables, and establishes an intelligent prediction model of dynamic load-material performance degradation relationship. S5. Time-varying reliability assessment: Based on the Gamma process, a performance degradation model of the protective layer is established. Combined with the set limit state equation, the reliability index at different service time points is calculated to quantify the remaining service life of the structure. S6. Maintenance decision optimization output: Based on the reliability assessment results, automatically generate a graded maintenance recommendation plan.
[0018] In this embodiment: In the prior art, the high-frequency water particle impact generated by the flood discharge atomization of large-scale water conservancy projects poses a serious threat to the durability of protective structures. Traditional assessment methods have limitations due to static load assumptions, failing to accurately reflect the combined changes in load amplitude and frequency during actual flood discharge. Existing simulation technologies often employ simplified physical field models, resulting in insufficient accuracy in simulating multi-field coupling effects. Furthermore, the lack of a time-varying reliability assessment system that integrates long-term monitoring data and intelligent algorithms makes it difficult to provide differentiated decision-making basis for engineering maintenance.
[0019] To address the aforementioned issues, this application proposes a durability assessment method for slope protection layers in flood discharge atomization zones, comprising six core steps: dynamic load parameter monitoring and acquisition, load characteristic probabilistic statistical analysis, multi-field coupled numerical simulation analysis, intelligent prediction of durability damage, time-varying reliability assessment, and optimized output of maintenance decisions.
[0020] Dynamic load parameter monitoring and acquisition refers to the real-time acquisition of the stress state of the protective layer through a distributed sensor network. Specifically, this can be achieved by combining stress sensors, accelerometers, and environmental monitoring devices to realize synchronous acquisition and preprocessing of multi-source data. Load characteristic probabilistic statistical analysis refers to establishing a joint load distribution model based on extreme value theory and Monte Carlo simulation. Specifically, typical load spectra can be extracted using rainflow counting to provide a statistical basis for subsequent analysis. Multi-field coupled numerical simulation analysis refers to constructing a fluid-discrete medium interaction model. Specifically, a CFD-DEM coupled calculation method is used to simulate the impact process of atomized water particles, revealing the dynamic response mechanism of the protective layer. Durability damage intelligent prediction refers to establishing a performance degradation model using a long short-term memory neural network. Specifically, damage evolution prediction is achieved by integrating load characteristics and material parameters. Time-varying reliability assessment refers to establishing a degradation model based on the Gamma process. Specifically, reliability indices for different service periods are calculated using the limit state equation. Maintenance decision optimization output refers to generating graded maintenance schemes based on reliability assessment results. Specifically, different maintenance strategies are triggered by setting thresholds.
[0021] First, sensor arrays are deployed at key locations in the protective layer to collect dynamic load data in real time under flood discharge conditions, constructing a structured monitoring database. Next, probabilistic statistical analysis is performed on the monitoring data to establish an amplitude-frequency joint distribution model, extracting typical load spectra as input conditions. Subsequently, a CFD-DEM coupled model is used to simulate the impact process of atomized water particles, obtaining the transient stress response characteristics of the protective layer. Load spectrum characteristics and material parameters are input into an LSTM network to predict the material performance degradation trajectory. A degradation model is established based on the Gamma process, and time-varying reliability indices are calculated using the limit state equation. Finally, graded maintenance recommendations are generated based on reliability thresholds.
[0022] This method overcomes the limitations of traditional static load assumptions, accurately reflecting the joint variability of loads through dynamic monitoring and probabilistic modeling. Multi-field coupled numerical simulation replaces empirical formulas, significantly improving the simulation accuracy of impact processes. Intelligent algorithms are introduced to construct a degradation prediction model, and combined with stochastic process theory, time-varying reliability quantitative assessment is achieved, providing a scientific basis for maintenance decisions.
[0023] This application effectively solves the core problems of inaccurate load characteristic description, insufficient accuracy of multi-field coupling simulation, and lack of time-varying reliability assessment in traditional methods. The combination of dynamic monitoring and probabilistic modeling improves the accuracy of load characteristic characterization, multi-physics coupling calculation accurately simulates the water particle impact mechanism, and the integration of intelligent prediction model and stochastic process theory enables dynamic assessment of protective layer performance degradation, ultimately providing quantitative decision support for engineering maintenance.
[0024] Example 2: Please refer to Figure 1 In step S1, the specific steps for monitoring and collecting dynamic load parameters are as follows; S1.1 Layout and calibration of monitoring devices: Distributed stress sensors, triaxial accelerometers and temperature and humidity sensors are deployed at key locations in the flood discharge atomization impact zone. Calibration experiments are conducted to ensure that the sensor range, accuracy and acquisition frequency are adapted to the characteristics of the flood discharge load. S1.2 Multi-source data synchronous acquisition: Dynamic load parameters are collected in real time during flood discharge, and the working period is marked in combination with flood discharge scheduling records; S1.3 Database construction and preprocessing: Time-domain alignment, noise filtering and outlier removal are performed on the raw data, and multi-source data are integrated according to timestamps to construct a structured long-term monitoring database.
[0025] In this embodiment, this application further proposes a specific implementation method for monitoring and acquiring dynamic load parameters in step S1, including three stages: deployment and calibration of monitoring devices, synchronous acquisition of multi-source data, and database construction and preprocessing. Distributed stress sensors, triaxial accelerometers, and temperature and humidity sensors are deployed at key locations in the flood discharge atomization impact zone. Calibration experiments are conducted to ensure that the sensor range, accuracy, and acquisition frequency are adapted to the characteristics of the flood discharge load. Dynamic load parameters are acquired synchronously in real time during flood discharge, and the working period is marked in conjunction with the flood discharge scheduling records. The raw data is time-domain aligned, noise filtered, and outlier removed. Multi-source data is integrated according to timestamps to construct a structured long-term monitoring database.
[0026] Distributed stress sensors are sensing devices capable of measuring surface stress of protective layers at multiple points. They can be implemented using fiber optic grating sensors or piezoelectric sensors, and are used to capture the spatial distribution characteristics of impact loads. Triaxial accelerometers are devices with three-dimensional vibration monitoring capabilities, and can be implemented using MEMS accelerometers, used to acquire the dynamic response of the protective layer under flood discharge impact. Temperature and humidity sensors are devices capable of simultaneously monitoring ambient temperature and humidity. They can be implemented using a combination module of capacitive humidity sensors and thermistors, and are used to correlate the influence of environmental factors on load characteristics. Calibration experiments refer to the process of determining the sensor's range and accuracy through laboratory loading tests. This can be achieved by applying a standard load spectrum using a hydraulic servo loading system, ensuring that the sensor parameters match the characteristics of the flood discharge load. Time-domain alignment refers to a processing method to eliminate timestamp discrepancies in multi-source data. This can be achieved using GPS clock synchronization technology, ensuring data time consistency. Noise filtering refers to the data processing process of eliminating high-frequency interference signals. This can be achieved using wavelet threshold denoising algorithms, improving the data signal-to-noise ratio.
[0027] When deploying sensor networks at key locations on the slope of the flood discharge atomization zone, stress concentration areas need to be determined based on fluid dynamics simulation results. For example, sensors should be densely deployed on the water-facing side of the protective layer, at structural joints, and at slope turning points. In calibration experiments, the dynamic response characteristics of the sensors are verified by simulating the impact load spectrum under flood discharge conditions, such as using vibration loads with a frequency range of 5-200Hz and an amplitude range of 0.1-5MPa. During multi-source data synchronous acquisition, a GPS module is used to provide a unified clock reference for all sensor nodes; for example, a PPS second pulse signal is used to achieve microsecond-level synchronization accuracy. During database construction, the raw vibration signals collected by the accelerometers are subjected to Butterworth low-pass filtering, for example, setting the cutoff frequency to 500Hz to eliminate high-frequency noise. Simultaneously, box plot methods are used to identify and remove abnormal stress peak data.
[0028] Traditional methods often employ single-point sensors and lack system calibration, resulting in an inability to accurately capture the spatial distribution characteristics of impact loads. Existing technologies frequently suffer from time synchronization issues in data acquisition; for example, the sampling time deviation between accelerometers and stress sensors can reach the second level, affecting load characteristic correlation analysis. This solution, through the coordinated deployment and precise calibration of multiple types of sensors, can comprehensively acquire the spatiotemporal distribution information of the load. GPS clock synchronization technology controls the time deviation of multi-source data to within milliseconds. Combined filtering algorithms effectively eliminate environmental noise interference with useful signals.
[0029] This application addresses the issue of insufficient data integrity in traditional monitoring methods, enabling accurate acquisition of the amplitude, frequency, and spatiotemporal distribution characteristics of dynamic loads. Through multi-sensor collaborative calibration and synchronous acquisition, it ensures the temporal consistency of monitoring data for different physical quantities, providing reliable input for subsequent probabilistic statistical analysis. The construction of a structured database achieves effective integration of multi-dimensional data, supporting the mining and analysis of long-term performance degradation patterns.
[0030] Example 3: Please refer to Figure 1 In step S2, the specific method of load characteristic probabilistic statistical analysis is as follows; S2.1 Monitoring data feature extraction and preprocessing: The dynamic load time history data in the long-term monitoring database is subjected to stationarity test and segmentation processing. Key statistical features such as load amplitude, frequency of action, and duration under each working condition are extracted. The influence of units is eliminated through data standardization, providing standardized input for probabilistic modeling. S2.2 Construction of joint probability distribution model: The generalized extreme value distribution is used to fit the maximum load amplitude distribution to determine the return period of extreme loads. Random samples are generated based on the load characteristic statistical parameters to construct the amplitude-frequency joint probability distribution model. The Kolmogorov-Smirnov test is used to verify the good agreement between the distribution model and the measured data. S2.3 Typical load spectrum generation: Decompose the non-stationary load time history into semi-cyclic events, statistically analyze the amplitude-mean-frequency matrix, merge small amplitude cycles with low damage contribution, generate simplified but equivalent typical load spectra, and output multiple sets of representative load spectra according to the flood discharge conditions.
[0031] In this embodiment: This application further proposes a durability assessment method for the slope protection layer of the flood discharge atomization zone. The specific method of load characteristic probabilistic statistical analysis in step S2 is as follows: The stationarity test and segmentation processing of the dynamic load time history data in the long-term monitoring database are performed, and the key statistical features of load amplitude, frequency of action, and duration under each working condition are extracted. The influence of dimensions is eliminated by data standardization. The maximum load amplitude distribution is fitted by the generalized extreme value distribution to determine the return period of extreme loads. Random samples are generated based on the load characteristic statistical parameters, and a joint probability distribution model of amplitude and frequency is constructed. The Kolmogorov-Smirnov test is used to verify the consistency between the distribution model and the measured data. The non-stationary load time history is decomposed into semi-cyclic events, the amplitude-mean-frequency matrix is statistically analyzed, and small amplitude cycles with low damage contribution are merged to generate a simplified but equivalent typical load spectrum. Multiple sets of representative load spectra are output according to the flood discharge working conditions.
[0032] Generalized extreme value distribution refers to a probabilistic distribution model used to describe the statistical characteristics of extreme events. Specifically, maximum likelihood estimation can be used for parameter fitting to accurately characterize the distribution patterns of extreme load amplitudes during flood discharge. Copula functions are probabilistic tools used to construct multivariate joint distributions. Gaussian Copula or t-Copula functions can be used to jointly model load amplitude and frequency by separating marginal distributions from related structures. Monte Carlo methods are probabilistic simulation techniques based on random number generation. Specifically, Latin hypercube sampling strategies can be used to generate load sample sets that conform to joint distribution characteristics. Rainflow counting is a cyclic load identification algorithm. Specifically, the four-point method can be used to decompose complex load time histories into quantifiable damage semi-cyclic events. Miner's linear cumulative damage theory is a fatigue damage superposition criterion based on stress cycles. Specifically, a damage contribution threshold can be set, and the load spectrum can be simplified by equivalently merging small-amplitude cycles.
[0033] After the dynamic load time history data underwent stationarity testing, it was segmented using the sliding window method to extract statistical characteristic parameters for each time period. The data standardization process employed Z-score normalization to eliminate the influence of parameters with different dimensions on probabilistic modeling. When constructing the joint probability distribution model, the marginal distribution of the maximum load amplitude was first fitted using a generalized extreme value distribution. Then, a correlation model between amplitude and frequency was established using the Copula function, ultimately forming a two-dimensional joint probability density function. In the generation of typical load spectra, based on the cyclic event matrix extracted using the rainflow counting method, low-damage cycles were merged using the damage equivalence principle, forming a simplified spectrum that reflects actual load characteristics and is convenient for engineering applications.
[0034] Traditional methods typically employ single-variable probability models or empirical correlation assumptions, which struggle to accurately reflect the joint variability of load amplitude and frequency. This application constructs a joint probability distribution model based on measured data, comprehensively characterizing the multidimensional statistical features of dynamic loads and eliminating evaluation biases caused by neglecting parameter correlations in traditional methods. Regarding load spectrum generation, existing technologies often use fixed amplitude thresholds for cyclic merging, while this application's dynamic threshold setting method based on damage contribution achieves a balance between load spectrum simplification and damage equivalence.
[0035] This application effectively solves the damage prediction bias problem caused by traditional static load assumptions and improves the engineering applicability of load statistical models. The established joint probability distribution model can accurately reflect the amplitude-frequency joint characteristics of flood discharge loads, providing accurate input conditions for subsequent numerical simulations. The load spectrum simplification method based on the damage equivalence principle significantly reduces computational complexity while ensuring analytical accuracy, providing an efficient and reliable data foundation for durability assessment.
[0036] Example 4: Please refer to Figure 1In step S2.2, the joint probability distribution model is constructed in the following specific way; S2.21. Extreme value statistics and distribution fitting: Extract the maximum load amplitude sequence of each time period from the preprocessed monitoring data, use the generalized extreme value distribution for parameter estimation, determine the shape parameter, scale parameter and location parameter, and calculate the predicted extreme load values corresponding to different return periods. S2.22. Based on the amplitude and frequency edge distribution characteristics of measured data, construct a Copula function to describe its correlation structure, use the Monte Carlo method to generate a random sample set that conforms to the joint distribution characteristics, establish a two-dimensional joint probability density function of amplitude and frequency, and determine the probability of occurrence of different combination conditions. S2.23. Apply the Kolmogorov-Smirnov test to compare the difference between the cumulative distribution function of the model prediction and the measured data, calculate the maximum deviation D statistic, determine the goodness of fit, and iteratively adjust the distribution type or parameters until the test is passed.
[0037] In this embodiment: This application further proposes a specific method for constructing the joint probability distribution model in step S2.2, including: extracting the maximum load amplitude sequence for each time period from the preprocessed monitoring data, using the generalized extreme value distribution for parameter estimation, determining the shape parameter, scale parameter, and location parameter, and calculating the predicted extreme load values corresponding to different return periods; based on the amplitude and frequency marginal distribution characteristics of the measured data, constructing a Copula function to describe its correlation structure, using the Monte Carlo method to generate a random sample set that conforms to the joint distribution characteristics, establishing a two-dimensional joint probability density function of amplitude-frequency, and determining the probability of occurrence of different combined working conditions; applying the Kolmogorov-Smirnov test to compare the difference between the model prediction and the cumulative distribution function of the measured data, calculating the maximum deviation D statistic, determining the goodness of fit, and iteratively adjusting the distribution type or parameters until the test is passed.
[0038] Generalized extreme value distribution refers to a distribution model used to describe the statistical characteristics of extreme events. Specifically, maximum likelihood estimation can be used for parameter estimation, capturing the tail characteristics of load amplitude. Copula function is a probability connection function used to construct a multivariate joint distribution. Specifically, Gaussian Copula or t-Copula can be used, separating marginal distributions from correlation structures. Monte Carlo method is a random sampling technique based on a probability model. Specifically, Latin hypercube sampling can be used to generate sample sets, efficiently constructing multidimensional joint distribution characteristics. Kolmogorov-Smirnov test is a nonparametric statistical test method. Specifically, by calculating the maximum difference between the empirical distribution and the theoretical distribution, it can quantitatively assess the fitting accuracy of the probability model.
[0039] In the joint probabilistic modeling of dynamic loads, the extreme characteristics of the maximum load amplitude are first fitted using a generalized extreme value distribution. For example, shape parameters reflect the tail shape of the load distribution, and scale parameters control the degree of data dispersion. Based on the marginal distribution characteristics of amplitude and frequency, a Copula function is used to establish the nonlinear dependency between them. For example, the frequency may exhibit a nonlinear decay trend as the load amplitude increases. Random samples satisfying the joint distribution characteristics are generated using the Monte Carlo method. For example, each sample contains a matching combination of amplitude and frequency, forming a two-dimensional probability density surface. Finally, the Kolmogorov-Smirnov test is applied to verify the model accuracy. For example, the model is considered to have passed the validation if the D statistic is less than the critical value; otherwise, the Copula type or marginal distribution parameters need to be adjusted.
[0040] Traditional methods often employ single-variable distribution assumptions or ignore parameter correlations; for example, using only a normal distribution to describe load amplitude leads to an underestimation of extreme events. This scheme combines extreme value theory with Copula functions to simultaneously characterize the extreme characteristics of load amplitude and the amplitude-frequency coupling relationship, accurately reflecting the load combination probability under a 100-year flood discharge condition. The combined application of Monte Carlo simulation and statistical testing ensures the model possesses both physical rationality and mathematical rigor; for instance, the generated joint distribution both conforms to the characteristics of measured data and passes statistical significance tests.
[0041] This application can accurately quantify the amplitude-frequency joint distribution characteristics of dynamic loads, solving the damage prediction bias problem caused by traditional static load assumptions. The established two-dimensional probabilistic model can provide accurate load input parameters for the reliability analysis of protective layers. For example, the probability of extreme load combinations under different return periods can be directly used for limit state equation calculation. The model verification mechanism ensures the matching degree between the probability distribution and actual working conditions. For example, through iterative adjustments, the maximum deviation between the theoretical distribution and the measured data is controlled within the allowable range.
[0042] Example 5: Please refer to Figure 1 In step S2.3, the specific method for generating the typical load spectrum is as follows; S2.31 Load time history cyclic feature decomposition: Rainflow counting method is used to identify the cyclicity of non-stationary load time history, extract the amplitude, mean and frequency of each cyclic event, and generate an initial amplitude-mean-frequency statistical matrix. S2.32. Calculate the damage contribution of each cycle based on Miner's linear cumulative damage theory, set the damage threshold, merge cycles with small amplitudes and keep the total damage equivalent. S2.33. Associate the flood discharge operation parameters, cluster them according to the operating conditions to generate characteristic load spectra, and store them in a standardized format for subsequent analysis input.
[0043] In this embodiment: This application further proposes the following specific method for generating the typical load spectrum in step S2.3: load time history cyclic feature decomposition, using rainflow counting method to identify non-stationary load time histories in cycles, extracting the amplitude, mean, and frequency of each cycle event, and generating an initial amplitude-mean-frequency statistical matrix; calculating the damage contribution of each cycle based on Miner's linear cumulative damage theory, setting damage thresholds, merging small amplitude cycles and maintaining total damage equivalence; associating flood discharge operation parameters, clustering according to operating conditions to generate characteristic load spectra, and standardizing and storing them as input formats for subsequent analysis.
[0044] Rainflow counting is a counting method that decomposes the time history of non-stationary loads into semi-cyclic events. Specifically, it can be implemented using the cyclic identification algorithm recommended by the ASTM E1049 standard to accurately extract the amplitude-mean characteristics in the load time history.
[0045] Miner's linear cumulative damage theory refers to a fatigue cumulative damage calculation method that uses linear superposition of damage amounts from each cycle. Specifically, it can use the damage equivalence principle to merge small-amplitude cycles with low damage contributions, thereby simplifying the load spectrum while maintaining damage equivalence.
[0046] The damage threshold is the minimum damage contribution threshold for judging cycle merging. Specifically, a percentage threshold can be set according to the material fatigue characteristics to screen out key cycle events that retain high damage contributions.
[0047] Flood discharge operation parameters refer to operating parameters including flood discharge flow, gate opening, and operating period. Specifically, these parameters can be analyzed by correlating scheduling records with load spectra to generate characteristic load spectra that reflect different flood discharge conditions.
[0048] Non-stationary load time history data is first decomposed into multiple semi-cyclic events using the rainflow counting method, and the amplitude, mean, and frequency of each cycle are extracted to form an initial statistical matrix. Then, based on Miner's theory, the damage amount corresponding to each cycle is calculated. After setting a damage threshold, cycles with small amplitudes below the threshold are merged to ensure that the total damage amount is equivalent to the original data. Finally, the processed load spectrum is correlated with flood discharge operation parameters, and cluster analysis is performed according to different operating conditions to generate a characteristic load spectrum in a standardized storage format.
[0049] Existing methods typically employ raw load time histories or simple amplitude statistics, which cannot effectively handle the cyclic characteristics and damage equivalence issues of non-stationary loads. This proposed solution, however, combines rainflow counting with damage equivalence analysis, significantly reducing data size while preserving key damage characteristics. Furthermore, it generates multiple characteristic load spectra by incorporating load parameters, better meeting the multi-load condition analysis needs of practical engineering projects.
[0050] This application can accurately capture the cyclic characteristics and damage contribution distribution of dynamic loads, effectively reducing computational complexity while ensuring analytical accuracy. The generated standardized characteristic load spectrum can directly support subsequent multi-field coupled simulation and lifetime prediction, providing reliable load input conditions for the durability assessment of the protective layer.
[0051] Example 6: Please refer to Figure 1 In step S3, the specific method of multi-field coupling numerical simulation analysis is as follows; S3.1 Multiphysics Model Construction: Establish the CFD model of the atomized water phase and the DEM model of the protective layer, and set the data exchange protocol for the coupling interface. S3.2 Coupled calculation and dynamic process simulation: Input typical load spectrum as boundary conditions, solve the fluid-discrete medium interaction equation, and record the time history data of the stress / displacement field of the protective layer during the impact process; S3.3 Response characteristic analysis and verification: Extract key transient response peaks, verify simulation accuracy by comparing with high-speed camera measured data, and generate stress concentration area cloud map and damage evolution curve.
[0052] In this embodiment, the following specific method for multi-field coupled numerical simulation analysis is proposed: Multiphysics model construction: establishing a CFD model of the atomized water phase and a DEM model of the protective layer, and setting a data exchange protocol for the coupling interface; Coupled calculation and dynamic process simulation: inputting typical load spectra as boundary conditions, solving the fluid-discrete medium interaction equation, and recording the time history data of the stress / displacement field of the protective layer during the impact process; Response characteristic analysis and verification: extracting key transient response peaks, verifying the simulation accuracy by comparing with high-speed camera measured data, and generating stress concentration area cloud maps and damage evolution curves.
[0053] CFD models refer to computational fluid dynamics models. Specifically, they can use the finite volume method to discretize the governing equations and combine them with turbulence models to simulate the flow characteristics of atomized water particles, thereby capturing the velocity and pressure field distributions of the fluid phase.
[0054] The DEM model refers to the discrete element model, which can specifically use an aggregate of spherical particles to characterize the protective layer material and calculate the interaction forces between particles through a contact mechanics model to simulate the dynamic response behavior of discrete media.
[0055] The coupling interface data exchange protocol refers to the data transmission rules for the interaction between fluid and discrete medium. Specifically, a two-way coupling algorithm can be used to realize the momentum exchange between CFD and DEM models, ensuring that the drag force of the fluid on the particles and the reaction force of the particles on the fluid are updated synchronously.
[0056] Using typical load spectrum as boundary conditions means inputting load characteristic parameters obtained from probability statistics into the simulation model. Specifically, time-varying pressure distribution functions can be used to define the inlet boundary conditions of the fluid domain, so that the impact process in numerical simulation is dynamically consistent with the actual flood discharge conditions.
[0057] High-speed camera measurement data comparison and verification refers to obtaining the dynamic deformation of the protective layer surface through optical measurement methods. Specifically, digital image correlation technology can be used to process high-speed camera images, extract displacement field measurement data and simulation results for quantitative error analysis.
[0058] First, a CFD model of the atomized water phase and a DEM model of the protective layer were established, and a data exchange protocol for the coupling interface was set up to achieve accurate modeling of the interaction between fluid dynamics and discrete media. Then, a typical load spectrum was used as the boundary condition input model to solve the fluid-discrete media interaction equations, fully recording the time-history changes of the stress and displacement fields of the protective layer during dynamic impact. Finally, by extracting key transient response peak values and comparing them with high-speed camera measurement data, contour maps and damage evolution curves reflecting stress concentration areas were generated, providing high-precision input data for subsequent durability assessment.
[0059] Traditional methods often employ single-physics models or empirical formulas to simplify multi-field coupling effects, resulting in an inability to accurately capture the transient mechanical response during water particle impact. This proposed approach, however, constructs a CFD-DEM coupled model to fully characterize the bidirectional interaction between the fluid and the discrete medium. Furthermore, it combines measured data to dynamically verify the simulation results, significantly improving the accuracy of transient stress response calculations.
[0060] This application effectively solves the technical defect of traditional simulation methods that ignore the coupling effect of multiple physics fields. It can accurately simulate the dynamic impact process of atomized water particles, accurately identify the stress concentration area and damage evolution path of the protective layer, and provide a reliable transient response data basis for durability damage prediction.
[0061] Example 7: Please refer to Figure 1 In step S4, the specific method for intelligent prediction of durability damage is as follows; S4.1 Dataset construction and feature engineering: Integrate load spectrum parameters and material performance parameters, perform feature standardization, missing value filling and outlier detection, and divide the training set, validation set and test set according to time series; S4.2 LSTM Model Construction and Training: Design the network structure, set the time-series sliding window, define the loss function and optimizer, and iteratively optimize the weight parameters through backpropagation; S4.3 Model Validation and Degradation Prediction: Use test sets to evaluate prediction accuracy, use sensitivity analysis to determine key influencing parameters, and input new load conditions to output material performance degradation curves.
[0062] In this embodiment, this application further proposes a specific method for intelligent prediction of durability damage, including three steps: dataset construction and feature engineering, LSTM model construction and training, and model validation and degradation prediction. In dataset construction and feature engineering, load spectrum parameters and material performance parameters are integrated, and feature standardization, missing value imputation, and outlier detection are performed. The training set, validation set, and test set are divided according to time series. In LSTM model construction and training, the network structure is designed, a time-series sliding window is set, the loss function and optimizer are defined, and the weight parameters are iteratively optimized through backpropagation. In model validation and degradation prediction, the prediction accuracy is evaluated using the test set, sensitivity analysis is performed to determine key influencing parameters, and new load conditions are input to output material performance degradation curves.
[0063] Load spectrum parameters refer to the set of dynamic load characteristic parameters obtained through monitoring and numerical simulation. Specifically, they can be implemented using combinations of parameters such as amplitude, frequency, and duration of action, and are used to characterize the load action patterns under different flood discharge conditions. Material performance parameters refer to the set of indicators characterizing the mechanical properties of the protective layer material. Specifically, they can be implemented using parameters such as elastic modulus, tensile strength, and fatigue limit, and are used to quantify the performance degradation law of materials under long-term loads. Feature standardization refers to a data processing method that converts input parameters with different dimensions into a unified scale. Specifically, it can be implemented using Z-score standardization or maximum-minimum normalization methods, and is used to eliminate the influence of parameter dimension differences on model training. Temporal sliding window refers to a method of dividing continuous time series data into fixed-length subsequences. Specifically, it can be implemented using a fixed-step sliding truncation method, and is used to extract the temporal correlation features between load and material performance. Sensitivity analysis is an analytical method that evaluates the degree of influence of input parameters on output results. Specifically, it can be implemented using parameter perturbation methods or variance decomposition methods, and is used to identify key parameters that play a dominant role in predicting material degradation.
[0064] First, a dataset containing multidimensional features is constructed by integrating load spectrum parameters from a long-term monitoring database and material performance parameters obtained from laboratory tests. Missing value imputation and outlier detection are performed on the dataset to ensure data quality meets modeling requirements. Standardization methods are used to transform feature parameters of different dimensions to a uniform numerical range, avoiding model training bias caused by differences in magnitude. Subsequently, the processed data is divided into training, validation, and test sets in chronological order, preserving the continuity of the time series. In the model building phase, an LSTM network structure containing input, hidden, and output layers is designed, and the sliding window length is set to match the load cycle characteristics. By defining a mean squared error loss function and an Adam optimizer, the backpropagation algorithm is used to iteratively update the network weight parameters, and the validation set is used to monitor the model training process to prevent overfitting. After model training is complete, the prediction accuracy is evaluated using the test set, and the error index between predicted and measured values is calculated. Sensitivity analysis is used to identify key parameters that significantly affect material performance degradation, forming a parameter importance ranking. Finally, the trained model is applied to new load conditions, outputting material performance degradation curves, providing input data for subsequent reliability assessment.
[0065] Traditional methods typically employ static regression models or empirical formulas for damage prediction, which struggle to capture the temporal correlation characteristics of dynamic loads and fail to consider the time-varying properties of material performance parameters. This proposed solution, by integrating a temporal sliding window and an LSTM neural network, effectively extracts the dynamic correlation between load application and material degradation. Existing technologies often employ simple filtering in data preprocessing, failing to systematically address missing data and outlier issues. This solution, however, significantly improves data quality through feature standardization combined with missing value imputation techniques. Furthermore, traditional sensitivity analyses are often based on theoretical assumptions rather than measured data. This solution combines parameter perturbation with measured data, making the identification of key parameters more valuable for engineering guidance.
[0066] This application addresses the problem that traditional damage prediction methods cannot effectively handle the temporal characteristics of dynamic loads, overcomes prediction biases caused by data quality defects, and establishes a material performance degradation prediction model based on actual monitoring data. By integrating intelligent algorithms and engineering parameters, it achieves quantitative prediction of protective layer durability damage, providing data support for formulating precise maintenance strategies.
[0067] Example 8: Please refer to Figure 1 In step S5, the specific method for time-varying reliability assessment is as follows; S5.1 Performance degradation trajectory modeling: Based on damage prediction data, construct a Gamma process model, determine the time-varying characteristics of shape and scale parameters, and establish a degradation stochastic process expression that considers measurement uncertainties; S5.2 Limit state analysis: Define key performance thresholds, establish limit state equations that include degradation, and use the First-Order Reliability Method to calculate time-varying reliability indices. S5.3 Remaining lifetime probability prediction: Set an acceptable minimum reliability threshold, predict the first exceedance probability through Monte Carlo simulation, and output the remaining lifetime range at different confidence levels.
[0068] In this embodiment: This application further proposes a specific method for time-varying reliability assessment in step S5, including performance degradation trajectory modeling, limit state analysis, and remaining lifetime probability prediction. Performance degradation trajectory modeling constructs a Gamma process model based on damage prediction data, determines the time-varying characteristics of shape and scale parameters, and establishes a degradation stochastic process expression considering measurement uncertainties; Limit state analysis defines key performance thresholds and establishes limit state equations that include degradation. The First-Order Reliability Method is used to calculate time-varying reliability indices. Remaining lifetime probability prediction sets an acceptable minimum reliability threshold and predicts the first exceedance probability through Monte Carlo simulation, outputting the remaining lifetime range at different confidence levels.
[0069] The Gamma process model is a stochastic process model applicable to monotonically increasing degradation processes. It uses the maximum likelihood estimation method to determine the time functions of shape and scale parameters, thereby effectively characterizing the nonlinear degradation law of the protective layer material performance over time.
[0070] The First-Order Reliability Method is a reliability calculation method based on Taylor expansion. Specifically, it can be implemented by combining design checkpoint search with gradient calculation to solve reliability indices under complex limit state equations. Monte Carlo simulation is a numerical calculation method based on probability sampling. Specifically, it can use Latin hypercube sampling to generate a random sample set and predict remaining lifetime by statistically analyzing the time distribution of the first exceedance event.
[0071] In the performance degradation trajectory modeling stage, the material performance degradation data output from the intelligent prediction model is first input into the Gamma process. The functional relationships between shape and scale parameters over time are determined using parameter estimation methods; for example, the shape parameter can be described using an exponential function to describe the accelerated degradation stage. When constructing the stochastic degradation process, a measurement error term is further introduced to reflect the actual uncertainty of the sensor-acquired data. In the limit state analysis stage, critical failure thresholds are defined based on the protective layer's functional requirements, such as using the maximum crack width or minimum compressive strength as criteria, and a limit state function incorporating time-varying degradation is constructed. When calculating reliability indices using the first-order reliability method, an iterative algorithm is used to search for design verification points and calculate sensitivity coefficients, outputting reliability change curves at different time points. In the remaining lifetime prediction stage, a minimum acceptable reliability threshold is set, for example, 0.85 is taken as the safety threshold. Large-scale random sampling is used to simulate the time distribution of the performance degradation path first exceeding the threshold, ultimately outputting a statistically significant remaining lifetime confidence interval.
[0072] Traditional methods often employ deterministic degradation models that neglect the time-varying characteristics of parameters, such as assuming linear degradation of material properties, leading to overly optimistic remaining life predictions. This proposed solution, however, captures the randomness and time-varying nature of performance degradation through the Gamma process and combines it with Monte Carlo simulation to quantify the impact of uncertainties, making the life prediction results closer to engineering realities. Existing reliability analyses are mostly limited to static condition assessments; this solution introduces a time-varying reliability index to dynamically reflect the evolution of the structure's safety status, providing a basis for decision-making in full life-cycle management.
[0073] This application can accurately quantify the stochastic characteristics of protective layer performance degradation, dynamically assess the evolution of structural reliability over service time, and predict the confidence interval of remaining service life based on probabilistic methods. This solves the problem that traditional deterministic models cannot reflect the uncertainty of the actual degradation process, provides a scientific basis for formulating differentiated maintenance strategies, and effectively avoids resource waste caused by premature maintenance or safety hazards caused by delayed maintenance.
[0074] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
[0075] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.
Claims
1. A method for evaluating the durability of slope protection layers in flood discharge atomization zones, characterized in that: The specific steps are as follows: S1. Dynamic load parameter monitoring and acquisition: By deploying distributed stress sensors, accelerometers and environmental monitoring devices, the dynamic load parameters of the protective layer under flood discharge conditions are collected in real time, and a long-term monitoring database is established. S2. Load characteristic probabilistic statistical analysis: Based on monitoring data, extreme value theory and Monte Carlo simulation method are used to establish a joint probability distribution model of dynamic load amplitude and frequency, and typical load spectrum is extracted by rainflow counting method. S3. Multi-field coupled numerical simulation analysis: Construct a CFD-DEM coupled calculation model, and simultaneously consider the interaction between fluid dynamics and discrete media to simulate the dynamic impact process of atomized water particles on the protective layer and analyze the transient stress response characteristics of the protective layer. S4. Intelligent prediction of durability damage: It adopts a long short-term memory neural network algorithm, takes load spectrum characteristic parameters and material performance parameters as input variables, and establishes an intelligent prediction model of dynamic load-material performance degradation relationship. S5. Time-varying reliability assessment: Based on the Gamma process, a performance degradation model of the protective layer is established. Combined with the set limit state equation, the reliability index at different service time points is calculated to quantify the remaining service life of the structure. S6. Maintenance decision optimization output: Based on the reliability assessment results, automatically generate a graded maintenance recommendation plan.
2. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 1, characterized in that, In step S1, the specific steps for monitoring and collecting dynamic load parameters are as follows; S1.1 Layout and calibration of monitoring devices: Distributed stress sensors, triaxial accelerometers and temperature and humidity sensors are deployed at key locations in the flood discharge atomization impact zone. Calibration experiments are conducted to ensure that the sensor range, accuracy and acquisition frequency are adapted to the characteristics of the flood discharge load. S1.2 Multi-source data synchronous acquisition: Dynamic load parameters are collected in real time during flood discharge, and the working period is marked in combination with flood discharge scheduling records; S1.3 Database construction and preprocessing: Time-domain alignment, noise filtering and outlier removal are performed on the raw data, and multi-source data are integrated according to timestamps to construct a structured long-term monitoring database.
3. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 2, characterized in that, In step S2, the specific method for probabilistic statistical analysis of load characteristics is as follows; S2.1 Monitoring data feature extraction and preprocessing: The dynamic load time history data in the long-term monitoring database is subjected to stationarity test and segmentation processing. Key statistical features such as load amplitude, frequency of action, and duration under each working condition are extracted. The influence of units is eliminated through data standardization, providing standardized input for probabilistic modeling. S2.2 Construction of joint probability distribution model: The generalized extreme value distribution is used to fit the maximum load amplitude distribution to determine the return period of extreme loads. Random samples are generated based on the load characteristic statistical parameters to construct the amplitude-frequency joint probability distribution model. The Kolmogorov-Smirnov test is used to verify the good agreement between the distribution model and the measured data. S2.3 Typical load spectrum generation: Decompose the non-stationary load time history into semi-cyclic events, statistically analyze the amplitude-mean-frequency matrix, merge small amplitude cycles with low damage contribution, generate simplified but equivalent typical load spectra, and output multiple sets of representative load spectra according to the flood discharge conditions.
4. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 3, characterized in that, In step S2.2, the joint probability distribution model is constructed in the following specific way; S2.
21. Extreme value statistics and distribution fitting: Extract the maximum load amplitude sequence of each time period from the preprocessed monitoring data, use the generalized extreme value distribution for parameter estimation, determine the shape parameter, scale parameter and location parameter, and calculate the predicted extreme load values corresponding to different return periods. S2.
22. Based on the amplitude and frequency edge distribution characteristics of measured data, construct a Copula function to describe its correlation structure, use the Monte Carlo method to generate a random sample set that conforms to the joint distribution characteristics, establish a two-dimensional joint probability density function of amplitude and frequency, and determine the probability of occurrence of different combination conditions. S2.
23. Apply the Kolmogorov-Smirnov test to compare the difference between the cumulative distribution function of the model prediction and the measured data, calculate the maximum deviation D statistic, determine the goodness of fit, and iteratively adjust the distribution type or parameters until the test is passed.
5. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 4, characterized in that, In step S2.3, the specific method for generating the typical load spectrum is as follows; S2.31 Load time history cyclic feature decomposition: Rainflow counting method is used to identify the cyclicity of non-stationary load time history, extract the amplitude, mean and frequency of each cyclic event, and generate an initial amplitude-mean-frequency statistical matrix. S2.
32. Calculate the damage contribution of each cycle based on Miner's linear cumulative damage theory, set the damage threshold, merge cycles with small amplitudes and keep the total damage equivalent. S2.
33. Associate the flood discharge operation parameters, cluster them according to the operating conditions to generate characteristic load spectra, and store them in a standardized format for subsequent analysis input.
6. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 5, characterized in that, In step S3, the specific method of multi-field coupled numerical simulation analysis is as follows; S3.1 Multiphysics Model Construction: Establish the CFD model of the atomized water phase and the DEM model of the protective layer, and set the data exchange protocol for the coupling interface. S3.2 Coupled calculation and dynamic process simulation: Input typical load spectrum as boundary conditions, solve the fluid-discrete medium interaction equation, and record the time history data of the stress / displacement field of the protective layer during the impact process; S3.3 Response characteristic analysis and verification: Extract key transient response peaks, verify simulation accuracy by comparing with high-speed camera measured data, and generate stress concentration area cloud map and damage evolution curve.
7. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 6, characterized in that, In step S4, the specific method for intelligent prediction of durability damage is as follows; S4.1 Dataset construction and feature engineering: Integrate load spectrum parameters and material performance parameters, perform feature standardization, missing value filling and outlier detection, and divide the training set, validation set and test set according to time series; S4.2 LSTM Model Construction and Training: Design the network structure, set the time-series sliding window, define the loss function and optimizer, and iteratively optimize the weight parameters through backpropagation; S4.3 Model Validation and Degradation Prediction: Use test sets to evaluate prediction accuracy, use sensitivity analysis to determine key influencing parameters, and input new load conditions to output material performance degradation curves.
8. The method for evaluating the durability of the slope protection layer in the flood discharge atomization zone according to claim 7, characterized in that, In step S5, the specific method for time-varying reliability assessment is as follows; S5.1 Performance degradation trajectory modeling: Based on damage prediction data, construct a Gamma process model, determine the time-varying characteristics of shape and scale parameters, and establish a degradation stochastic process expression that considers measurement uncertainties; S5.2 Limit state analysis: Define key performance thresholds, establish limit state equations that include degradation, and use the First-Order Reliability Method to calculate time-varying reliability indices. S5.3 Remaining lifetime probability prediction: Set an acceptable minimum reliability threshold, predict the first exceedance probability through Monte Carlo simulation, and output the remaining lifetime range at different confidence levels.
Citation Information
Cited By
Green building life cycle evaluation method and system based on pile foundation data
CN121881482A
