36results about How to "Overcome subjectivity" patented technology

The invention discloses a design method capable of giving consideration to both toughening and crack path controllability for a brittle material, and belongs to the technical field of fracture mechanics and artificial intelligence crossing, and the method comprises the steps: employing an input coding method based on space-response double-domain fusion to carry out feature extraction; introducing a channel attention mechanism to perform weighted fusion on the extracted double-domain features, and generating prediction results for different key tasks; evaluating the reliability of a prediction result by adopting a phase field verification and closed-loop adaptive optimization mode, and carrying out iterative training and optimization on the neural network; and taking the optimized neural network as a fitness evaluation function of double-target genetic optimization, and performing multi-target optimization on defect generation parameters. According to the invention, through fusion of the spatial domain and the response domain, geometric morphology and dynamic response characteristics are cooperatively extracted, and the prediction precision of complex fracture behaviors is improved; through non-dominated sorting multi-objective optimization under physical constraints, a high-toughness and path-controllable optimal solution is output, and brittle materials are pushed to cross an active controllable fracture design.

This application relates to a method and system for analyzing microstructural defects in composite materials based on artificial intelligence image recognition. The method includes: acquiring microstructural images of composite material samples; performing multi-scale feature extraction on the microstructural images to obtain a multi-level feature map set; inputting the multi-level feature map set into a preset defect analysis model to obtain a defect segmentation map of the composite material sample; wherein the defect analysis model includes: a topology-aware feature fusion module for analyzing the topological features of each convolutional feature map to obtain topological attribute data of the convolutional feature maps, and fusing the convolutional feature maps based on the topological attribute data to obtain a fused feature map; a morphological constraint decision module connected to the topology-aware feature fusion module for transforming the fused feature map based on preset morphological rule constraint parameters to obtain a defect segmentation map. This method can improve the accuracy and efficiency of defect analysis in composite materials.

This invention relates to the field of bid document management technology, and in particular to an automatic bid document review method and system based on natural language processing. The method includes: identifying target sections; determining historical performance and financial information; judging whether false performance exists; analyzing the complete sentences containing target keywords; calculating material scores; determining price deviation; and determining the review result. This invention achieves multi-dimensional and comprehensive review of bid documents by intelligently analyzing multiple sections such as qualifications, risks, materials, and pricing. This effectively improves the comprehensiveness and depth of the review. Utilizing natural language processing technology, it can accurately identify and extract key information from bid documents, overcoming the subjectivity and limitations of traditional manual review. Quantitative assessment based on risk level significantly improves the accuracy and relevance of risk identification. By comprehensively weighing material scores and price deviation, it significantly improves review efficiency.

The invention belongs to the technical field of power grid disaster prevention and reduction, and particularly relates to a risk assessment method and device for deicing jump of a power transmission line, electronic equipment and a computer readable storage medium. The method comprises the following steps: determining risk influence factors influencing the deicing jump of the power transmission line, and constructing an original parameter data set containing a plurality of deicing working conditions; establishing a conductor deicing model under a plurality of deicing working conditions and carrying out analogue simulation to obtain a risk consequence index; constructing a standardized decision matrix, respectively calculating the comparison intensity and conflict of each risk influence factor, and calculating the comprehensive information amount and objective weight of each risk influence factor to obtain an objective weight vector; and identifying the key dominant factors based on the objective weight vector and carrying out working condition optimization in combination with the risk consequence indexes under each deicing working condition to obtain the optimal deicing working condition with the lowest comprehensive risk. According to the method, the CRITIC method is applied to power transmission line deicing jump risk assessment, and the deicing working condition with small risk can be selected.

The application provides a highway slope meteorological disaster early warning model training method, an early warning method and a system, positive and negative samples are respectively constructed based on historical disaster record and non-disaster record; based on sample slope position and collection time, time series data of multiple meteorological elements before disaster is extracted from meteorological grid point data through spatial interpolation matching to generate meteorological statistical characteristics; initial training sample set is constructed by fusing slope static attributes and meteorological statistical characteristics, adding disaster type and risk level labels; the initial model is pre-trained based on the sample set, the feature importance is evaluated through multi-index, and the feature subset is formed by screening; the features outside the subset are removed from the initial training sample set to form the target training sample set, and multiple candidate models are trained based on this, and the target disaster type recognition model and the target risk level recognition model are determined according to the evaluation index. Through multi-source feature fusion, automatic feature screening and multi-model optimization, the accuracy and reliability of the early warning are improved.

The invention discloses an acoustic radar near-surface wind speed inversion method and system based on entropy weight feature fusion, and the method specifically comprises the steps: carrying out the preprocessing of acoustic radar frequency spectrum data, and obtaining the preprocessed frequency spectrum data; based on the preprocessed spectrum data, constructing a Doppler window dynamic migration model based on turbulence dynamics power law distribution; on the basis of a Doppler window dynamic migration model, candidate spectrum peaks are selected at each height layer, a feature space is constructed, feature weights are dynamically distributed by using an information entropy theory, and an optimal spectrum peak is selected; introducing a symmetry constraint optimization strategy to the optimal spectrum peak, detecting the profile symmetry of the wind spectrum in the symmetric direction, and calibrating an asymmetric result to obtain a target wind speed inversion result; and comparing and evaluating the target wind speed inversion result and the wind speed of the anemometer tower as a reference to complete inversion result verification. According to the method, the wind speed inversion precision of the acoustic radar in a complex environment is effectively improved, and better evaluation and efficient utilization of wind energy resources are promoted.

The application discloses a lightweight feature extraction model and a flotation process running state evaluation method. The model comprises: a first teacher module configured to generate a first roughing image feature and a first scavenging image feature according to the foam image; a second teacher module configured to generate a first data feature according to the foam data; a student module configured to generate a second roughing image feature according to the first roughing image feature, generate a second scavenging image feature according to the first scavenging image feature, and generate a second data feature according to the first data feature; and a fusion module configured to fuse the second roughing image feature, the second scavenging image feature and the second data feature to generate a first fusion feature, and generate a state grade label according to the first fusion feature; and determine the state grade of the current flotation process running state, so as to provide a more lightweight model deployment and a more competitive application in an actual industrial scene.

The invention provides a capability assessment method and device for group discussion participants, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining text data of group discussion; inputting the text data into a pre-trained personal general ability assessment model and a pre-trained team leader assessment model, and outputting a personal general ability assessment result and a team leader assessment result; wherein the personal universal ability evaluation model is a large language model which is trained by performing supervision and fine tuning through a training data set comprising an information node, a decision node and a utility node and combining a low-rank self-adaptive method; the team leader ability evaluation model is used for evaluating the thinking excitation ability, conflict coordination ability and discussion field control ability of the participants based on an event sequence containing participant identifiers, speaking keywords and speaking timestamps. Through the method provided by the invention, the personal general ability can be effectively evaluated, and the team leader ability can be quantitatively evaluated.

The present application relates to the technical field of data processing, and specifically provides a port yard digital management method, system, device and storage medium, comprising: collecting images in a port yard through a plurality of image collection devices arranged above the yard; processing the collected images to identify and extract the cargo stacking area in the images; calculating the proportion of the cargo stacking area in each image based on the extracted cargo stacking area, and calculating the average proportion L of all images; determining whether the yard occupancy meets the standard based on the average proportion L; when it is determined that the yard occupancy does not meet the standard, analyzing the reason for not meeting the standard based on the preset multi-level judgment logic, and generating a processing method corresponding to the reason. Through image recognition and multi-level intelligent judgment logic, the present application realizes the precision, automation and intelligence of port yard management.

The invention discloses a sewage pipe network flow prediction algorithm based on rainfall and liquidometer data, and belongs to the field of flow prediction.The sewage pipe network flow prediction algorithm based on rainfall and liquidometer data comprises the following steps that historical and real-time multi-dimensional data from multiple sources are continuously collected and integrated; performing cleaning, calibration and time alignment processing on the multi-dimensional data to obtain model training data based on historical data and model prediction input data based on real-time data; the method has the advantages that multi-dimensional data are collected and processed systematically, fuzzy manual observation is replaced, historical hydrological laws are converted into a quantifiable sewage pipe network flow prediction model, judgment depending on personal experience is replaced, accurate flow prediction in several hours in the future is carried out based on the model, and the flow prediction efficiency is improved. Passive emergency is converted into active early warning, and the hysteresis and subjective defects caused by dependence on artificial experience in the prior art are overcome fundamentally.

The invention discloses a pavement engineering construction data processing and construction quality control method and system, and belongs to the technical field of pavement engineering construction. The method comprises the following steps: collecting multi-source construction data in real time through sensors arranged on materials, equipment and a site; predicting a future trend value of each quality influence factor by using a gray model; determining the comprehensive weight of each factor by adopting an analytic hierarchy process; the correlation degree between the factors is calculated through a grey correlation analysis method; constructing a multi-level grey comprehensive evaluation system based on the predicted value, the weight and the correlation degree; according to the prediction result, generating a plurality of construction parameter combination schemes in a preset step length to form a candidate scheme set; sorting the candidate schemes by adopting the comprehensive evaluation system, and selecting an optimal construction scheme; and finally, the parameters of the optimal scheme are issued to corresponding construction equipment through a control interface, the equipment operation parameters are dynamically adjusted, and real-time closed-loop control over the construction quality is achieved. According to the invention, accurate, dynamic and intelligent control of the construction process is realized.

The present application relates to the technical field of intelligent surveying and mapping, in particular to a real estate surveying and mapping optimization method and system based on AI. In the present application, boundary error data is generated by quantifying the deviation between global navigation satellite system coordinates and cadastral vector data, the spatial features of the real estate subject boundary are deeply extracted using a convolutional neural network, and targeted boundary fitting correction is performed on the super-limit point position in combination with the error data, overcoming the passive acceptance and subjective judgment of errors in traditional data processing, realizing intelligent perception and adaptive optimization of the boundary geometric form, and subsequently using a Bayesian inference model, taking the corrected boundary as a prior distribution and fusing new observation data for probability updating, dynamically obtaining the boundary point position coordinates with minimized uncertainty, solving the problem that static data coverage cannot effectively fuse multi-source information, and then performing consistency screening on multi-temporal cadastral data based on the updated boundary.

The embodiment of the invention relates to a method and device for training a model based on reinforcement learning, and the method comprises the steps: inputting a sample question into a to-be-trained target model, and obtaining a first target reply; inputting the sample questions into a plurality of trained contrast models to obtain a plurality of contrast answers; the sample question comprises text content; inputting the sample question, the first target reply and any first contrast reply in the plurality of contrast replies into an evaluation model to obtain D relative preference scores of the first target reply on D preset evaluation dimensions relative to the first contrast reply; performing aggregation operation on the respective D relative preference scores of the first target reply relative to each contrast reply, and determining a first reward score corresponding to the first target reply; and updating a parameter value of the target model at least according to the first reward score.

This application discloses a method and apparatus for predicting and controlling dissolved gas precipitation at the end of brine discharge in a salt cavern gas storage facility, comprising the following steps: acquiring basic data required for prediction and control at the end of brine discharge; establishing or calling a table of equilibrium solubility data of the stored gas in the brine based on the composition, pressure, temperature, and salinity of the stored gas; determining the initial dissolved gas state of the brine at the interface, using the boundary condition that the brine and stored gas reach dissolution equilibrium at the gas-liquid interface of the salt cavern; establishing a coupled calculation model of dissolved gas precipitation and wellbore gas-liquid two-phase flow during the brine return process to predict the gas phase volume fraction of the fluid returning from the wellhead; generating a safe operating envelope diagram of back pressure-discharge rate at different gas-liquid interface positions, using the wellhead back pressure as the main control variable and the brine discharge flow rate as the auxiliary correction variable, with the wellhead back pressure as the main control variable and the brine discharge flow rate as the auxiliary correction variable; and outputting a suggested point for terminating brine discharge when any wellhead safety constraint cannot be met.

This invention proposes a dynamic self-optimization and adaptive learning method for large-scale water conservancy system models, comprising steps such as data acquisition and preprocessing, model construction and initialization, real-time monitoring and model evaluation, difference analysis and adaptive adjustment strategy generation, model updating and optimization, and continuous learning and knowledge accumulation. This method, through dynamic data-driven approaches, intelligent difference analysis, and adaptive optimization, achieves a transformation of water conservancy system models from "static and passive" to "dynamic and proactive." Compared to existing technologies, it offers greater accuracy, real-time performance, economy, and sustainability, providing a feasible technical path for the intelligent upgrading of water conservancy systems and possessing broad application potential in fields such as flood control and disaster reduction, agricultural irrigation, and clean energy production.

The invention discloses a model preferential method, device and equipment based on feature weight and a storage medium. The method comprises the following steps: acquiring historical predicted power data and historical model input data corresponding to at least two candidate power prediction models, and historical real power data corresponding to the historical predicted power data; on the basis of the historical predicted power data, the historical model input data and the historical real power data, generating a performance feature vector corresponding to each candidate power prediction model; and determining a target power prediction model from at least two candidate power prediction models based on the performance feature vector and the to-be-optimized performance feature weight corresponding to each performance feature. According to the technical scheme, the prediction precision of power prediction is improved.

The application provides a silver ore oxidation degree analysis method, belongs to the field of mineral processing and process mineralogy, and selects a to-be-measured silver ore sample, performs solidification treatment after cutting, and obtains a primary sample; performs grinding and polishing and carbon spraying treatment on the primary sample to obtain an automatic mineralogy analysis sample; performs automatic mineralogy analysis on the automatic mineralogy analysis sample, measures silver mineral content and silver content of each silver mineral, calculates an ore theoretical silver grade; measures the length of an exposed surface of a single silver mineral particle, determines the centroid position of each silver mineral particle, and calculates a centroid distance; obtains an actual silver grade of the to-be-measured silver ore sample through gravity separation, and calculates a correction coefficient; and based on mineral spatial geometric relations and density parameters, the overall oxidation rate of the silver ore and the oxidation rate of a specific silver mineral phase are calculated. The application provides a quantitative analysis method based on automatic mineralogy technology and spatial geometric relations, and realizes accurate, objective and efficient determination of the silver ore oxidation degree.

The utility model relates to connector detection technical field discloses a kind of connector detection system, by being respectively arranged first detection foot and second detection foot on male seat and female seat, and using the detection port of mainboard to carry out whether consistency monitoring to the level state of these two detection feet, whether the connection state of connector can be judged quickly, accurately. This detection mode overcomes the subjectivity and inefficiency problem of traditional manual visual inspection, also avoids the limitation of infrared detection equipment on identification accuracy, can automatically, objectively complete detection task, effectively improves the accuracy and stability of detection result, reduces the electronic equipment failure caused by connector connection bad, so as to significantly improve the overall performance and service life of electronic equipment, provides a kind of efficient, reliable detection means for the production, assembly and quality control of electronic equipment.

The invention discloses a business process evaluation method, which is used for solving the problems that an existing health degree evaluation method cannot effectively capture a dynamic dependency relationship and a time sequence change among multiple systems, so that an evaluation result is inaccurate and incomplete, and the real-time sensing capability on the health state of the whole process is lacked. The method comprises the steps that according to acquired operation data of a to-be-evaluated business process, a heterogeneous atlas corresponding to the to-be-evaluated business process is constructed, nodes in the heterogeneous atlas are used for representing business systems corresponding to the to-be-evaluated business process, and edges in the heterogeneous atlas are used for representing the calling relation between the business systems; inputting the heterogeneous atlas into a pre-trained time-space diagram neural network model to obtain a business process health degree score output by the time-space diagram neural network model; and determining the health state of the to-be-evaluated business process according to the business process health degree score, and determining a corresponding operation and maintenance strategy according to the health state.

The invention discloses a multi-virtual-person simulated teaching method and a related device, and relates to the field of virtual teaching, multi-modal interaction data of a target object in a virtual classroom is acquired, a target dimension influenced by the multi-modal interaction data is determined, a state value of a virtual student under the target dimension is updated according to the multi-modal interaction data, and the state value of the virtual student under the target dimension is updated according to the updated state value. And the virtual students are driven to make related behavior feedback based on the updated state values, the multi-modal interaction data are obtained again until the teaching is finished, and teaching evaluation data are generated according to the multi-modal interaction data and the behavior feedback obtained in the teaching process. According to the method, state value updating is only carried out on the influenced target dimension, the lesson listening state jumping efficiency is improved, the training scene is more consistent with the real teaching situation, and teaching in the virtual training scene is safer and more controllable; and evaluation is carried out based on objective data of the teaching process, so that objectivity and credibility of teaching evaluation data are improved.

The invention relates to the technical field of session data analysis, and discloses a time axis fusion analysis-based multi-modal session data analysis method and system, and the method comprises the steps: obtaining multi-modal session data, and enabling each modal session data in the multi-modal session data to have a corresponding modal session time axis; according to a preset anti-interference fitting algorithm, all modal session time axes are calibrated, and a unified wall clock time axis of the multi-modal session data is generated; identifying key information nodes in the multi-modal session data based on a unified wall clock time axis, and calculating an emotion vector corresponding to each moment in the multi-modal session data to obtain an emotion intensity curve of the multi-modal session data; and according to all the key information nodes and the emotional intensity curve, generating an interactive emotional map of the multi-modal session data. Therefore, by implementing the multi-modal session data analysis method, the traceability of the analysis result can be improved while the analysis accuracy and objectivity of the multi-modal session data are improved.

The invention discloses a component quality detection system for an assembly type steel structure, and relates to the technical field of engineering quality detection, and the system comprises a digital twinborn body construction module which is used for generating a first detection report through simulation analysis based on a digital twinborn body; the data acquisition module is used for guiding the detection equipment to perform entity detection based on the first detection report and acquiring actually measured data of the component; the analysis and diagnosis module is used for comparing the measured data with a theoretical prediction value to generate deviation data and diagnosing the root of a quality problem; the prognosis prediction module is used for predicting the performance degradation path and the remaining service life of the component under multiple damage coupling by using an artificial intelligence model; and the feedback optimization module is used for outputting an instruction for optimizing a subsequent component design, a manufacturing process or a detection strategy based on the diagnosis result, the performance degradation path and the remaining service life. On the basis, the technical problem that an existing quality detection method is insufficient in accuracy and perspective is solved.

The application relates to the technical field of computer vision and environment monitoring, and discloses a microplastic component rapid identification method based on multi-modal data driving, which comprises the following steps: firstly, acquiring morphological sequence images and chemical component maps of microplastics, and constructing a standardized three-dimensional voxel dataset through metadata alignment; then, constructing a non-coupling training set and performing three-dimensional data enhancement based on spatial transformation; training a component identification network by using the enhanced data, and establishing a mapping relationship between the three-dimensional morphological characteristics of the microplastics and the chemical components; finally, collecting morphological images of samples to be measured and inputting the images into the network, and outputting three-dimensional segmentation masks with chemical component codes. The application realizes rapid and accurate identification of the chemical components and three-dimensional distribution of microplastics only according to low-cost morphological images by mining the deep connection between the morphology and the components, solves the problems of high cost and slow speed of traditional chemical imaging, and is suitable for efficient monitoring of environmental microplastic pollution.

This invention discloses a method and system for constructing a multi-category product sustainability index (SPI) that integrates subjective and objective information with fuzzy processing, belonging to the field of product eco-design and environmental management technology. The method first constructs a unified evaluation framework encompassing four attributes—product function, resource and energy, environmental performance, and health and safety—along with specific measurement indicators. Then, it uses the analytic hierarchy process (AHP) to calculate subjective weights, the entropy weight method to calculate objective weights, and a game theory model to optimally integrate the subjective and objective weights to obtain a comprehensive weight. For qualitative indicators, fuzzy matter-element analysis is used to transform their linguistic evaluations into calculable quantitative values. Finally, the standardized data and quantitative values ​​of all indicators are combined with the comprehensive weight to calculate the product score and normalize it into the SPI. This invention overcomes the shortcomings of existing methods, such as strong subjectivity, inability to comprehensively quantify, and difficulty in horizontal comparison. It provides a scientific, universal, and comparable comprehensive evaluation scheme for product sustainability, offering an intuitive quantitative decision-making tool for green design, procurement, and regulation.

The application discloses a digital laser level and an event-based measurement recording method based on a specification threshold. The ruler body is hollow inside and is provided with a light path channel. A light source is arranged at the bottom of the ruler body for projecting light to the gap between the bottom of the ruler and the surface to be measured, and a light and shade image with light and shade contrast is formed at the fitting surface with flatness deviation. The light path assembly in the light path channel conducts the light and shade image to the camera module. The processing and storage module extracts the flatness feature of the collected light and shade image, converts it into quantitative data and stores it. The application converts the invisible interface state after the ruler body is attached into quantifiable and traceable quality evidence, fundamentally overcomes the limitation that the existing level can only read the angle or the levelness but cannot quantify the interface state of the attachment, improves the flatness detection precision, and provides direct basis for the after-check of the single measurement result by cooperating with the event-based measurement recording method based on the specification threshold.

The invention relates to the crossing field of public health and geographic information technology, and discloses a plague risk assessment method and system combined with a MaxEnt model, and the method comprises the steps: obtaining a multi-source environment variable, and carrying out the standardization; reducing a colinear variable through variance threshold preliminary screening, a Pearson's correlation coefficient and plague point significance test; performing variable importance sorting by using recursive feature elimination and a support vector machine, and dynamically determining an optimal variable subset; and inputting a MaxEnt model to train ecological niche probability distribution, and dividing risk levels based on historical occurrence point probability quantiles. The system comprises corresponding function modules. According to the system, through a three-stage automatic variable screening mechanism, the generalization ability, interpretability and prediction precision of the model are remarkably improved, and reliable support is provided for accurate prevention and control of plague.

This invention relates to the fields of neuroengineering and medical device technology, specifically to a method and system for optimizing the electrode placement of a brain-computer interface based on deep learning. The method includes: acquiring historical patient MRI images, surgical tasks, and intraoperative multi-channel neural electrical signals, and recording the three-dimensional spatial coordinates of each electrode; performing independent component analysis on the multi-channel neural electrical signals after preprocessing to evaluate the quality of each electrode and generate an electrode subset with spatial location information; training a deep learning model using MRI images and surgical tasks as training data and the electrode subset as labels; inputting the target patient's MRI images and surgical tasks into the model, and outputting the optimal electrode placement. This invention integrates multimodal data, accurately evaluates electrodes through independent component analysis, and automatically generates personalized optimal electrode placement using deep learning. This overcomes the shortcomings of traditional methods that rely on subjective experience or fixed combinations, significantly enhancing the adaptability and accuracy of intraoperative monitoring.

This invention proposes a quality evaluation method for aluminothermic welded joints based on welding process data, belonging to the field of welding technology. The method includes: collecting on-site parameters during the aluminothermic welding process; calculating the source quality score of the aluminothermic welded joint based on the on-site parameters; and formulating a handling strategy for the aluminothermic welded joint based on the source quality score. This method significantly improves the efficiency and consistency of quality evaluation, overcomes the subjectivity and lag of manual review, accurately identifies the risk levels of different types of defects, and provides data support for differentiated flaw detection and maintenance strategies, thereby comprehensively improving the quality control capabilities of railway welded joints and the level of traffic safety assurance.

The invention discloses a park low-carbon transformation financial evaluation and path optimization method based on a dynamic benchmark, and relates to the technical field of financial decision making, and the method comprises the steps: obtaining multi-source heterogeneous data of a park enterprise platform, and carrying out the credibility evaluation and timeliness analysis of the multi-source heterogeneous data; based on financial evaluation data and real-time monitoring data of Internet of Things equipment, a dynamic reference model is established by analyzing industry carbon emission trend and regional characteristics; based on the abnormal early warning signal and the elastic risk control strategy instruction set, automatically allocating a response event to a corresponding responsibility subject; generating a graded response instruction through a three-stage fusing mechanism; and synthesizing the hierarchical response instruction and the historical execution effect data, carrying out feature extraction and mode mining on the historical cases, and converting and outputting a transformation path scheme. According to the method, through multi-source data fusion and intelligent analysis, the risk identification and decision-making efficiency in the park low-carbon transformation process is remarkably improved, and it is ensured that the financial evaluation result has both accuracy and real-time performance.

The application provides a salinization permafrost railway embankment adaptive design method and system, and the method comprises the following steps: obtaining permafrost type information and slope rock-soil body properties of a permafrost section; calculating the comprehensive condition grade of the permafrost section, determining embankment design information according to the comprehensive condition grade, the permafrost type information and the slope rock-soil body properties of the permafrost section; determining embankment drainage design information and slope protection design information according to the permafrost type information and the slope rock-soil body properties of the permafrost section. The application realizes the precision and scientific nature of embankment design by establishing a permafrost parameter quantitative scoring and comprehensive grading system. Based on the matching difference of the comprehensive grade and the permafrost type, differential engineering measures are adopted to strengthen the fortification of high-risk sections, optimize the design of low-risk sections, and build a thermal-mechanical-hydraulic collaborative protection system, which can significantly improve the engineering economy while ensuring the long-term safety of the embankment.