Detection system for stone-laying retaining wall and detection method thereof

Abstract

The invention discloses a detection system for a stone-laying retaining wall and a detection method thereof, and relates to the technical field of stone-laying retaining wall detection. The method comprises the steps that the type of a target stone-laying retaining wall is acquired, and a corresponding environment influence factor set and a special retaining wall data analyzer are extracted; collecting multi-source retaining wall detection data including ultrasonic wave, ground penetrating radar and acoustic emission detection data; retaining wall detection features are extracted through a retaining wall data analyzer, and a first structure safety score is obtained; acquiring historical service environment data according to the environmental impact factor set, predicting long-term environmental impact through time sequence analysis and dynamic time zone division, and acquiring a second structure safety score; and finally, intelligently fusing the two scores to generate a comprehensive structure safety score as a final detection result. According to the method, through dual-path verification and multi-source information fusion, comprehensive and accurate evaluation of the real-time state and long-term performance of the stone-laying retaining wall is realized, and the detection reliability and the engineering guidance value are improved.

Description

Technical Field

[0001] This invention relates to the field of masonry retaining wall testing technology, specifically to a testing system and method for masonry retaining walls. Background Technology

[0002] Masonry retaining walls play an important role in water conservancy, transportation and construction projects, and their structural safety is directly related to the overall stability and safety of the project.

[0003] Currently, the inspection of masonry retaining walls mainly relies on manual inspection and single non-destructive testing (NDT) techniques. Manual inspection is limited by personnel experience and subjective judgment, making it difficult to detect hidden defects inside and deep within the retaining wall. Commonly used NDT methods, such as ultrasonic testing, ground-penetrating radar (GPR), or acoustic emission testing, each have specific advantages but also limitations. Ultrasonic testing is sensitive to internal material cracks, but its detection depth and range are limited; GPR is effective in identifying delamination and voids, but is easily affected by the moisture content of the medium; acoustic emission testing can monitor dynamic damage processes but is difficult to independently complete a static overall assessment. Existing technical solutions typically apply these methods in isolation, lacking effective fusion of multi-source data and failing to incorporate long-term environmental history data of the retaining wall into the analysis framework. This results in biased assessments that cannot comprehensively and accurately reflect the current true safety status of the structure and its evolution trend under time-varying environmental conditions. Summary of the Invention

[0004] This invention addresses the technical problem that existing technologies rely on single detection methods and fail to comprehensively consider long-term environmental impacts, resulting in one-sided and inaccurate safety assessments of masonry retaining walls. It provides a detection system and method for masonry retaining walls.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a detection system for masonry retaining walls, comprising: The retaining wall type extraction module is used to obtain the retaining wall type of the target masonry retaining wall and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; The detection data extraction module is used to extract the retaining wall detection data for the target masonry retaining wall, and the retaining wall detection data includes ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; The first safety scoring module is used to analyze and process ultrasonic detection data, ground-penetrating radar detection data and acoustic emission detection data through the retaining wall data analyzer to obtain retaining wall detection features, and determine the first structural safety score based on the retaining wall detection features. The second safety scoring module is used to obtain historical service environment data of the target masonry retaining wall based on the set of environmental impact factors, perform structural safety prediction of the target masonry retaining wall based on the historical service environment data, and obtain a second structural safety score. The detection result generation module is used to analyze and process the first structural safety score and the second structural safety score to generate the detection result of the target masonry retaining wall.

[0006] Secondly, the present invention provides a method for detecting masonry retaining walls, comprising: Obtain the retaining wall type of the target masonry retaining wall, and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; Extract the retaining wall detection data for the target masonry retaining wall, the retaining wall detection data including ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; The retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features, and determines a first structural safety score based on the retaining wall detection features. Based on the set of environmental impact factors, historical service environment data of the target masonry retaining wall is obtained. Based on the historical service environment data, structural safety prediction of the target masonry retaining wall is performed to obtain a second structural safety score. The first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall.

[0007] The beneficial effects of this invention are: Compared to existing technologies, this invention first achieves deep fusion of multi-source non-destructive testing data, comprehensively utilizing the advantages of ultrasonic, ground-penetrating radar, and acoustic emission technologies to overcome the limitations of single testing methods and more comprehensively reveal the internal structure and current state of masonry retaining walls. Secondly, it innovatively introduces long-term historical service environment data into the evaluation system, quantifying the cumulative evolution of structural performance over time by analyzing the time-varying influence of environmental factors, thus improving the predictability of the assessment. Thirdly, by constructing a dedicated analyzer matched to the retaining wall type, the evaluation model achieves targeting and adaptability, improving the accuracy and reliability of the analysis. Finally, an intelligent fusion mechanism for real-time detection scores and historical prediction scores is designed, dynamically adjusting weights based on data consistency. Ultimately, it outputs more scientific and objective test results that comprehensively reflect the immediate state and long-term evolution trend of the structure, providing strong quantitative support for engineering maintenance decisions. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of a detection system for masonry retaining walls provided by the present invention; Figure 2 This is a flowchart illustrating a detection method for masonry retaining walls provided by the present invention.

[0009] In the attached diagram, the components represented by each number are as follows: The module includes a retaining wall type extraction module 11, a detection data extraction module 12, a first safety scoring module 13, a second safety scoring module 14, and a detection result generation module 15. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0012] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0013] Example 1, as Figure 1 As shown, this embodiment of the invention provides a detection system for masonry retaining walls, comprising: The retaining wall type extraction module 11 is used to obtain the retaining wall type of the target masonry retaining wall and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; First, it is necessary to determine the type of the target masonry retaining wall. The target masonry retaining wall refers to the specific masonry structure undergoing safety testing and evaluation. Its type is defined based on the project's purpose, structural characteristics, and service environment, including categories such as hydraulic retaining walls, irrigation retaining walls, highway retaining walls, mountain highway retaining walls, and municipal slope retaining walls. Since different types of retaining walls differ in material composition, stress mechanisms, common damage patterns, and environmental sensitivity, clearly defining the retaining wall type provides a targeted technical framework and evaluation benchmark for the entire testing and analysis process.

[0014] Simultaneously, it is necessary to extract a set of environmental impact factors associated with the determined retaining wall type. This set of environmental impact factors refers to a series of predefined key parameters that characterize the impact of environmental effects on the long-term performance of the retaining wall. For example, for hydraulic retaining walls, the set of environmental impact factors typically includes water level change cycles, water erosion intensity, and freeze-thaw cycles; for retaining walls along mountain roads, it mainly involves parameters such as rainfall erosion, temperature variation, and seismic activity history. The purpose of extracting this set of environmental impact factors is to clarify the scope of historical environmental data that needs to be collected and processed.

[0015] Secondly, a retaining wall data analyzer corresponding to this type of retaining wall needs to be extracted. This data analyzer is a computational model specifically designed to process inspection data of a particular type of retaining wall and extract structural features and safety status information from it. Because different retaining wall types differ in material composition, construction techniques, and common damage mechanisms, different retaining wall data analyzers are required for each type. For example, an analyzer for irrigation retaining walls might focus on analyzing the acoustic emission signal characteristics induced by seepage, while an analyzer for municipal slope retaining walls might focus on image pattern recognition reflecting overall stability in ground-penetrating radar data.

[0016] Specifically, the retaining wall data analyzer is constructed as follows: The first retaining wall type is determined from multiple preset retaining wall types; Based on the first retaining wall type, retrieve historical retaining wall detection records of multiple masonry retaining wall samples of the same type. The historical retaining wall detection records include historical ultrasonic detection records, historical ground penetrating radar detection records, and historical acoustic emission detection records. Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed. By integrating the ultrasonic feature extraction branch, the ground penetrating radar feature extraction branch, and the acoustic emission feature extraction branch, a retaining wall data analyzer corresponding to the first retaining wall type is obtained. Obtain the retaining wall data analyzer corresponding to the first retaining wall type, and then obtain the retaining wall data analyzers corresponding to the other retaining wall types.

[0017] First, one of several predefined retaining wall types is selected as the first retaining wall type. This first retaining wall type is the initial representative type, such as a hydraulic retaining wall. The multiple predefined retaining wall types refer to a set of masonry retaining wall categories categorized according to common engineering application scenarios, and are determined through comprehensive analysis of industry standards, engineering experience, and historical project databases.

[0018] Secondly, based on the selected first retaining wall type, historical retaining wall detection records for multiple masonry retaining wall samples belonging to the same type are retrieved and obtained. These historical retaining wall detection records form the data foundation for constructing the retaining wall data analyzer, and specifically include historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records.

[0019] Specifically, historical ultrasonic testing records utilize ultrasonic detectors to emit high-frequency sound waves onto the retaining wall surface and record the propagation speed, amplitude attenuation, and waveform characteristics of the sound waves in the masonry medium by receiving the reflected wave signals. Its advantage lies in its high sensitivity to detecting defects such as internal material density, cracks, and delamination of the bonding layer. Historical ground-penetrating radar (GPR) testing records use GPR equipment to emit electromagnetic pulses into the retaining wall and record the two-way travel time, amplitude, and waveform information of the electromagnetic waves by receiving the reflected echoes. Its advantage lies in its ability to non-invasively detect a large area of ​​layered structure, voids, and water-bearing areas within the retaining wall. Historical acoustic emission testing records use high-sensitivity acoustic sensors deployed on the retaining wall surface to continuously monitor and record transient elastic wave signals released when cracks, friction, or deformation occur within the structure under load or environmental factors. Its advantage lies in its ability to dynamically capture the initiation and propagation process of structural damage, reflecting the active damage state of the structure.

[0020] Furthermore, using the acquired historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, three independent feature extraction branches were constructed: an ultrasonic feature extraction branch, a ground-penetrating radar feature extraction branch, and an acoustic emission feature extraction branch. The construction process of each feature extraction branch involves using the corresponding historical detection records and performing machine learning training to enable it to automatically extract deep-level features related to the structural state from new detection data.

[0021] Specifically, based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed, including: Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, sample ultrasonic detection datasets, sample ground-penetrating radar detection datasets, and sample acoustic emission detection datasets are constructed. The retaining wall structure features were annotated on the sample ultrasonic detection dataset, sample ground penetrating radar detection dataset, and sample acoustic emission detection dataset respectively to obtain the sample ultrasonic structure feature set, sample ground penetrating radar structure feature set, and sample acoustic emission structure feature set; Machine learning training is performed based on the sample ultrasonic detection dataset and the sample ultrasonic structural feature set to generate an ultrasonic feature extraction branch. Machine learning training is performed based on the sample ground-penetrating radar detection dataset and the sample ground-penetrating radar structural feature set to generate a ground-penetrating radar feature extraction branch. Machine learning training is performed based on the sample acoustic emission detection dataset and the sample acoustic emission structural feature set to generate an acoustic emission feature extraction branch.

[0022] First, based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, three structured datasets were compiled and constructed: a sample ultrasonic detection dataset, a sample ground-penetrating radar detection dataset, and a sample acoustic emission detection dataset. This step aims to format the raw record data, providing standard input for subsequent annotation and training processes.

[0023] Secondly, the aforementioned ultrasonic testing dataset, ground-penetrating radar dataset, and acoustic emission testing dataset were each professionally annotated with retaining wall structural features. Specifically, retaining wall structural feature annotation is a crucial process that transforms the raw detection signals into structured, quantifiable engineering parameters. This annotation work is completed by engineers with professional knowledge or with the aid of validated auxiliary interpretation tools.

[0024] Preferably, for the ultrasonic testing dataset, the annotation work focuses on identifying and quantifying the internal material defects revealed by the ultrasonic signals. The annotation result of each sample constitutes a multi-dimensional feature vector, whose typical dimensions include defect type, defect depth, defect size, severity, and the planar coordinates of the defect on the retaining wall facade. For example, a fully annotated sample can be described as follows: defect type is mortar void, located at a depth of 15 cm, defect area size is 8 cm × 5 cm, severity is rated as moderate, and its specific planar coordinates are recorded.

[0025] For the ground-penetrating radar (GPR) detection dataset, the annotation work focuses on analyzing the structural anomalies within the retaining wall reflected in the electromagnetic wave images. The annotations for each sample form a feature vector, containing dimensions such as the type of cavity or anomaly, burial depth, estimated volume, distribution density within the detection area, and connectivity between anomalies. For example, a sample might be labeled as detecting a local cavity located 30 cm deep within the wall, with an estimated volume of approximately 200 cubic centimeters, exhibiting a low-density distribution within the area, and existing as an independent, unconnected entity.

[0026] For the acoustic emission detection dataset, the annotation work aims to interpret the structural dynamic damage activity characterized by elastic wave signals. The annotation of each sample generates a feature vector, whose dimensions typically cover the signal activity level, the main damage mode, the damage development trend, the event frequency, and the released energy level. For example, the annotation result of a sample may be: low activity level, shear failure as the dominant damage mode, slow damage development trend, event frequency of 5 times per hour, and low energy level.

[0027] Through the above annotation process, we can obtain the sample ultrasonic structural feature set, sample ground-penetrating radar structural feature set, and sample acoustic emission structural feature set, which correspond one-to-one with the original data.

[0028] Furthermore, based on the obtained sample ultrasound detection dataset and its corresponding sample ultrasound structural feature set, machine learning algorithms are used for training. Machine learning algorithms refer to computer methods capable of automatically learning patterns from data and constructing prediction or feature extraction models, such as convolutional neural networks, support vector machines, or gradient boosting decision trees. This training process aims to enable the model to automatically identify and extract signal patterns or data features highly correlated with the labeled structural features from the original ultrasound detection data. After training, the resulting model becomes the ultrasound feature extraction branch, whose function is to automatically extract features from newly input ultrasound detection data and output structured, numerical feature vectors.

[0029] Similarly, machine learning is trained based on the sample ground-penetrating radar (GPR) detection dataset and its corresponding sample GPR structural feature set to generate a GPR feature extraction branch. This GPR feature extraction branch is used to extract features reflecting the internal structural state from the GPR data.

[0030] Similarly, machine learning training is performed based on the sample acoustic emission detection dataset and its corresponding sample acoustic emission structural feature set to generate an acoustic emission feature extraction branch. This acoustic emission feature extraction branch is used to extract features characterizing the active damage state of the structure from the acoustic emission data.

[0031] For example, since there is a highly nonlinear and complex mapping relationship between the structural defect features contained in the ultrasonic detection signal and its physical parameters, and the neural network model has excellent performance in automatic extraction and abstract representation of signal features, a convolutional neural network model can be selected to construct the ultrasonic feature extraction branch.

[0032] Specifically, the ultrasonic feature extraction branch mainly consists of a signal input layer, a convolutional feature extraction layer, and a feature vector output layer. The signal input layer receives a pre-processed ultrasonic time-domain or frequency-domain signal sequence, typically obtained from raw waveform data through filtering, normalization, and segmentation. The convolutional feature extraction layer employs a one-dimensional convolutional neural network structure, capturing local patterns and deep features in the signal through multi-level convolutional operations. A modified linear unit activation function is used after each convolutional layer to introduce non-linear processing capabilities, and pooling layers are added after some network layers to reduce feature dimensionality and enhance feature invariance. The feature vector output layer maps the extracted abstract features into structured feature vectors with the same dimensionality as the sample ultrasonic structural feature set through a fully connected network.

[0033] During training, key hyperparameters included a learning rate of 0.001, 200 training epochs, and a batch size of 32. The learning rate was set to ensure model convergence stability, the number of training epochs ensured sufficient learning of complex defect patterns in ultrasonic signals, and the batch size strikes a balance between training efficiency and gradient update stability. Specifically, a supervised learning approach was adopted, using signal sequences from the sample ultrasonic detection dataset as the input sample set and the labeled feature vectors from the corresponding sample ultrasonic structural feature set as the supervision target. The input sample set and the corresponding labeled feature vectors were divided into training, validation, and test sets according to a preset ratio, such as 7:2:1.

[0034] Next, the signal sequences from the training set are input into the network, and the predicted feature vectors are calculated using forward propagation. The deviation between the predicted feature vectors and the labeled feature vectors is measured using the mean squared error loss function, and the network weight parameters are iteratively updated using the Adam optimizer combined with the backpropagation algorithm. The training process is monitored using a validation set. When the average mean squared error loss function value of the validation set decreases by less than 1‰ over 20 consecutive training epochs, and the error value itself has fallen below a preset threshold, for example, reaching 0.05, the training process will automatically terminate. The model obtained at this point is the converged ultrasonic feature extraction branch. This ultrasonic feature extraction branch, after sufficient training, can automatically and accurately extract structured feature vectors from new ultrasonic detection data, reflecting a series of key information such as the type, size, location, and severity of internal defects.

[0035] Similarly, the construction of the ground-penetrating radar feature extraction branch follows the same principles. However, since its input data is a two-dimensional radar image profile, a two-dimensional convolutional neural network can be used as the core architecture to adapt to the spatial feature extraction requirements of image data. Its training process is also based on the sample ground-penetrating radar detection dataset and its corresponding sample ground-penetrating radar structural feature set, and the model is optimized through supervised learning. In addition, the construction of the acoustic emission feature extraction branch takes into account the characteristics of acoustic emission signals as a time-series event stream. Its input is usually a time series or spectral feature containing event parameters. Therefore, a long short-term memory network or a temporal convolutional network can be used as the basic model to effectively capture the dependencies and patterns of acoustic emission events in the time dimension. Its training process is completed based on the sample acoustic emission detection dataset and its corresponding sample acoustic emission structural feature set.

[0036] Through the above methods, dedicated and efficient feature extraction branches were constructed for the three detection technologies, laying a reliable data foundation for subsequent comprehensive analysis.

[0037] Furthermore, the ultrasonic feature extraction branch, ground-penetrating radar feature extraction branch, and acoustic emission feature extraction branch that have been trained are integrated and fused to form a unified analysis model that can collaboratively process multi-source data. This model is the dedicated retaining wall data analyzer corresponding to the first type of retaining wall.

[0038] Finally, the entire process from selecting the type to model integration is repeated, sequentially constructing a dedicated retaining wall data analyzer for each of the remaining retaining wall types in the preset retaining wall type set, such as irrigation retaining walls and highway retaining walls. This ultimately yields multiple retaining wall data analyzers specifically designed for various masonry retaining walls with different engineering application scenarios and structural characteristics. This allows for the rapid and accurate processing of the appropriate retaining wall data analyzer based on the specific retaining wall type to be detected, improving the targeting of the detection method and the reliability of the evaluation results.

[0039] The detection data extraction module 12 is used to extract the retaining wall detection data for the target masonry retaining wall, and the retaining wall detection data includes ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; Specifically, extracting retaining wall inspection data for a target masonry retaining wall involves comprehensively applying three non-destructive testing technologies based on different physical principles: ultrasonic waves, ground-penetrating radar, and acoustic emission. This enables the collection of comprehensive data on the retaining wall structure, from its surface to its interior, and from static defects to dynamic damage activities.

[0040] First, ultrasonic testing data is extracted. This data is obtained by placing ultrasonic transducers on the surface of the target masonry retaining wall, emitting high-frequency sound waves, and receiving the reflected or transmitted signals after they propagate inside the masonry. The ultrasonic testing data records the velocity, amplitude attenuation, frequency components, and waveform characteristics of the sound waves, and is used to investigate the material homogeneity, crack development, and integrity of the bonding layer inside the retaining wall.

[0041] Secondly, ground-penetrating radar (GPR) detection data is extracted. This data is generated by moving a GPR device along the surface of the target masonry retaining wall, emitting high-frequency electromagnetic pulses into the wall, and receiving reflected echoes from interfaces with different dielectric constants. GPR detection data is typically presented in the form of radar image profiles, containing the two-way travel time, amplitude intensity, and waveform information of the electromagnetic waves. It can effectively reveal the internal layering structure, void distribution, water-bearing areas, and large-scale discontinuities of the retaining wall.

[0042] Next, acoustic emission (AE) data is extracted. This data is obtained by distributing high-sensitivity acoustic sensors on the surface of the target masonry retaining wall to monitor transient elastic wave signals released by the structure due to micro-fractures, friction, or deformation during environmental loads or its own stress adjustment process, either over a long period or in a short period. The AE data is recorded in the form of an event stream, including the arrival time, energy, frequency, rise time, and sensor location information for each AE event, which can dynamically reflect the active damage process and stress redistribution within the retaining wall.

[0043] By simultaneously extracting and aggregating the three types of detection data based on different physical principles, a comprehensive and complementary multi-source information foundation can be provided for subsequent fusion analysis, ensuring that structural condition assessment can take into account multiple dimensions such as surface defects, internal structural anomalies, and dynamic damage activity.

[0044] The first safety scoring module 13 is used to analyze and process ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data through the retaining wall data analyzer to obtain retaining wall detection features, and determine the first structural safety score based on the retaining wall detection features. Specifically, the retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features, and determines a first structural safety score based on these features, including: The retaining wall data analyzer includes an ultrasonic feature extraction branch, a ground-penetrating radar feature extraction branch, and an acoustic emission feature extraction branch; The ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data are processed by ultrasonic feature extraction branch, ground-penetrating radar feature extraction branch, and acoustic emission feature extraction branch, respectively, to obtain ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features. By combining ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features, the retaining wall detection features are obtained. The retaining wall detection features are input into the structural safety evaluator, which outputs the first structural safety score.

[0045] First, the retaining wall data analyzer is invoked to analyze and process the previously obtained ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data. Specifically, the retaining wall data analyzer contains three feature extraction branches: ultrasonic feature extraction, ground-penetrating radar feature extraction, and acoustic emission feature extraction. Each feature extraction branch is a pre-trained dedicated model designed to process the corresponding type of data and extract depth features.

[0046] Secondly, the ultrasonic feature extraction branch, the ground-penetrating radar (GPR) feature extraction branch, and the acoustic emission feature extraction branch are launched in parallel. Specifically, the ultrasonic feature extraction branch receives the raw ultrasonic detection data as input, processes it through its internal network, and outputs a set of quantitative features characterizing the internal defect status of the material, i.e., ultrasonic detection features. The GPR feature extraction branch receives the raw GPR detection data as input, processes it, and outputs a set of quantitative features reflecting the structural anomalies and spatial distribution within the wall, i.e., GPR detection features. The acoustic emission feature extraction branch receives the raw acoustic emission detection data as input, analyzes it, and outputs a set of quantitative features describing the dynamic damage activity and stress state of the structure, i.e., acoustic emission detection features.

[0047] Furthermore, a feature combination operation is performed on the ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features acquired separately. Specifically, this combination operation concatenates the three feature vectors along their feature dimensions, fusing them into a single, higher-dimensional comprehensive feature vector. This comprehensive feature vector integrates information from multiple physical fields, including acoustics, electromagnetics, and elastic wave dynamics, and covers multi-scale state representations ranging from micro-defects in materials to macro-structural anomalies, thus constituting a retaining wall detection feature capable of comprehensively and deeply describing the current overall structural state of the target masonry retaining wall.

[0048] Finally, the generated retaining wall detection features are input into a pre-trained structural safety evaluator. This structural safety evaluator is a machine learning model whose core function is to establish a complex nonlinear mapping relationship from multi-source fused features to the overall safety status of the retaining wall. After receiving the input retaining wall detection features, the structural safety evaluator performs calculations and inferences through its internal network layers or functions, and finally outputs a continuously distributed quantitative value, namely the first structural safety score.

[0049] Specifically, the first structural safety score comprehensively reflects the immediate structural integrity, damage level, and stability of the target retaining wall as assessed based on current multi-source non-destructive testing data. A higher first structural safety score indicates a better current structural condition of the retaining wall, fewer internal defects, and stronger overall stability; a lower first structural safety score indicates that the retaining wall currently has significant structural damage or potential safety hazards, and its integrity or load-bearing capacity may have significantly decreased.

[0050] Specifically, the structural safety evaluator is constructed as follows: Based on the retaining wall type, historical retaining wall detection features and corresponding structural safety score annotation data of multiple masonry retaining walls of the same type are collected to construct a sample retaining wall detection feature set and a sample structural safety score set. A structural safety evaluator is built using machine learning; Using the sample retaining wall detection feature set as input features and the sample structural safety score set as supervision labels, the structural safety evaluator is trained until convergence, thus completing the construction.

[0051] First, based on the retaining wall type, historical retaining wall inspection features and corresponding structural safety score annotation data for multiple masonry retaining walls of the same type were collected. Specifically, the structural safety score annotation data was obtained through expert evaluation, inversion of long-term instrument monitoring data, or comprehensive analysis combined with historical maintenance records, based on quantitative safety indicators determined by professional engineering evaluation standards, long-term structural health monitoring conclusions, or authoritative inspection and appraisal reports. Using the historical retaining wall inspection features and corresponding structural safety score annotation data, two corresponding datasets can be constructed: a sample retaining wall inspection feature set and a sample structural safety score set. Each sample in the sample retaining wall inspection feature set is a feature vector that integrates multi-source information, while each sample in the sample structural safety score set is a scalar score representing the overall safety status.

[0052] Secondly, a suitable machine learning algorithm framework is selected to establish the initial model of the structural safety evaluator. Machine learning algorithms are computational methods that can automatically learn from data and establish a mapping relationship between inputs and outputs, such as deep neural networks, gradient boosting regression trees, or support vector regression. This structural safety evaluator needs to be able to process high-dimensional feature inputs and output continuous numerical scores to achieve accurate quantification and evaluation of the complex structural state information contained in the comprehensive retaining wall detection features.

[0053] Finally, the sample retaining wall detection feature set is used as input, and the sample structural safety score set is used as the target label for supervised learning to train the structural safety evaluator. This training process continues until convergence, at which point the construction of the structural safety evaluator is complete. The convergence condition is set based on a combination of the loss function value during training and the model's evaluation metrics on the validation set. For example, it can be set to ensure that the mean absolute error on the validation set decreases by less than 1‰ for 10 consecutive training epochs. The trained structural safety evaluator can effectively infer from the retaining wall detection features and output the corresponding first structural safety score.

[0054] Finally, the aforementioned retaining wall detection features are input into the trained structural safety evaluator, which calculates and outputs the first structural safety score.

[0055] The second safety scoring module 14 is used to obtain historical service environment data of the target masonry retaining wall based on the set of environmental impact factors, perform structural safety prediction of the target masonry retaining wall based on the historical service environment data, and obtain a second structural safety score. Specifically, historical service environment data of the target masonry retaining wall are obtained based on the set of environmental impact factors; structural safety prediction of the target masonry retaining wall is performed based on the historical service environment data; and a second structural safety score is obtained, including: A first environmental impact factor is obtained from the set of environmental impact factors, and first impact factor data is extracted from historical service environment data based on the first environmental impact factor. Time series analysis is performed on the first impact factor data, and the time segmentation points corresponding to the first environmental impact factor are determined based on the numerical change characteristics to generate a first time interval sequence. Following the method of obtaining the first time interval sequence corresponding to the first environmental impact factor, the time interval sequences corresponding to the remaining environmental impact factors are obtained, resulting in multiple time interval sequences. The multiple time interval sequences are merged to generate an environmental change time zone sequence, which includes multiple environmental change time zones. Environmental characteristic parameters of each environmental change time zone are obtained based on the historical service environment data. Obtain the initial structural safety score of the target masonry retaining wall; The initial structural safety score and environmental characteristic parameters of each environmental change time zone are input into the retaining wall structure analyzer to obtain the second structural safety score.

[0056] During long-term service, the material properties and structural integrity of masonry retaining walls are affected by a combination of environmental factors, including but not limited to periodic fluctuations in water levels, cyclical temperature changes, fluctuations in external loads such as traffic or surcharges, and potential chemical corrosion. Different environmental factors often follow their own independent patterns of change. For example, water level changes experienced by hydraulic retaining walls may exhibit seasonal flood and drought cycles, while temperature changes are mainly affected by diurnal and seasonal alternations, and load changes may be related to traffic flow or agricultural activity cycles.

[0057] Therefore, to overcome the above limitations, it is necessary to conduct independent and refined time-series analysis on each environmental factor in order to accurately capture and quantify the actual change rhythm and cumulative effect of each environmental force, thereby providing an input basis for the subsequent establishment of an accurate long-term prediction model for structural performance.

[0058] First, the first factor to be analyzed is extracted from the aforementioned set of environmental impact factors, namely the first environmental impact factor. Based on this first environmental impact factor, the corresponding chronologically recorded sequence of observation values ​​is selected from the complete historical service environment data, namely the first impact factor data. This first impact factor data is a one-dimensional time series reflecting the evolution of this specific environmental condition over time, including specific values ​​collected or recorded at different historical time points and their corresponding timestamps. It can be used to reveal the change cycle, fluctuation amplitude, trend characteristics of this environmental impact factor, and identify key time nodes when its state changes.

[0059] Secondly, a specialized time series analysis is performed on the data of the first impact factor. Specifically, the core purpose of this time series analysis is to identify the changing patterns and trend characteristics of the impact factor's numerical sequence. Preferably, a change point detection algorithm can be used to automatically locate the key time nodes where the impact factor experiences state transitions or trend reversals, defining them as time segment points. Based on all identified time segment points, the entire historical observation period of the first impact factor is divided into several consecutive time segments. Within each time segment, the value of the first impact factor remains relatively stable or exhibits a consistent changing trend, thereby generating the first time interval sequence corresponding to the first environmental impact factor.

[0060] Furthermore, following the same principles and methods, each remaining environmental impact factor in the set of environmental impact factors is analyzed independently. That is, each environmental impact factor is treated as the current analysis object, its specific time segmentation points are identified from its historical data, and its own independent time interval series are generated. Ultimately, multiple time interval series generated by all environmental impact factors are obtained.

[0061] Furthermore, the aforementioned multiple time interval sequences are merged. Specifically, the identified time segmentation points in all time interval sequences are extracted, merged into a single set, sorted chronologically, and duplicate points are removed, thus forming a unified sequence of key time points spanning the entire historical period. Based on this unified sequence of key time points, the historical service period is re-divided into a series of continuous time periods, each time period being called an environmental change time zone. All time zones arranged chronologically constitute the final environmental change time zone sequence.

[0062] This merging process ensures that all environmental impact factors considered are in a relatively stable state within each final defined environmental change time zone. Based on this, for each environmental change time zone, representative statistical characteristics of each environmental factor within that period are extracted from the original historical service environment data, such as mean, peak, or cumulative values. The extracted characteristic values ​​are combined to form environmental characteristic parameters that characterize the comprehensive environmental conditions of that time zone.

[0063] Secondly, obtain the quantitative value of the structural safety status of the target masonry retaining wall at the start of service or at a known assessment benchmark point. This value serves as the starting point for simulation prediction and is called the initial structural safety score.

[0064] Finally, the obtained initial structural safety score, along with the environmental characteristic parameters corresponding to each environmental change time zone arranged chronologically, are sequentially input into a pre-trained retaining wall structure analyzer. This retaining wall structure analyzer can simulate the evolution and degradation of the safety status of a masonry retaining wall structure over time under specific environmental conditions. Through iterative calculations from the first environmental change time zone to the last, the retaining wall structure analyzer can ultimately output a predicted structural safety score reflecting the cumulative impact of long-term environmental history, namely the second structural safety score.

[0065] Specifically, the retaining wall structure analyzer is constructed as follows: Based on the retaining wall type, multiple historical retaining wall service records of the same type are retrieved. Each historical retaining wall service record includes the starting structural safety score, environmental characteristic parameters, time zone duration, and ending structural safety score of multiple historical environmental change time zones. Based on the multiple historical retaining wall service records, a sample initial score set, a sample environmental feature parameter set, a sample time zone duration set, and a sample termination score set are constructed. The retaining wall structure analyzer is generated by using the sample initial score set, sample environmental feature parameter set, and sample time zone duration set as inputs, and the sample final score set as the output target, through machine learning training.

[0066] First, based on the retaining wall type, historical service records of multiple masonry retaining walls of the same type were retrieved and obtained. Each complete historical retaining wall service record contains comprehensive data for each historical environmental change time zone experienced by the retaining wall during its service life, specifically including: the initial structural safety score at the start of the historical environmental change time zone, statistically extracted environmental characteristic parameters within the historical environmental change time zone, the duration of the historical environmental change time zone, and the final structural safety score at the end of the historical environmental change time zone. In summary, the historical retaining wall service record describes the evolution of the retaining wall structure's safety status from start to finish under the continuous action of specific environmental conditions.

[0067] Secondly, the collected historical retaining wall service records were structured and organized. Specifically, data units from different retaining walls and different time zones were decomposed and classified to construct four one-to-one corresponding sample datasets: a sample starting score set consisting of starting scores for all time zones; a sample environmental feature parameter set consisting of environmental feature parameter vectors corresponding to each time zone; a sample time zone duration set consisting of duration for each time zone; and a sample termination score set consisting of termination scores for all time zones.

[0068] Finally, a machine learning approach is used to train and generate a retaining wall structure analyzer. During training, the initial score set, the environmental feature parameter set, and the time zone duration set are used as input features to describe the initial state at the start of a time zone and the environmental conditions within that time zone. Simultaneously, the final score set is used as the output target, i.e., the supervision label. Machine learning algorithms, such as gradient boosting regression trees, deep neural networks, or support vector regression, are used to train the model. The training process allows the model to learn and internalize the complex patterns of structural safety score changes under the combined effects of environmental conditions and time duration until convergence. The convergence condition is set based on the model's performance on the independent validation set; for example, it can be set to a minimum decrease of 0.5% in the root mean square error of the validation set over 20 consecutive training cycles. The trained retaining wall structure analyzer can predict the final structural safety score of the retaining wall after experiencing a given initial score, environmental feature parameters, and time zone duration.

[0069] Furthermore, the initial structural safety score and environmental characteristic parameters for each environmental change time zone are input into the retaining wall structure analyzer to obtain a second structural safety score, including: The initial structural safety score, the environmental characteristic parameters of the first environmental change time zone, and the time zone duration of the first environmental change time zone are input into the retaining wall structure analyzer to obtain the first-stage structural safety score. The first-stage structural safety score, the environmental characteristic parameters of the second environmental change time zone, and the time zone duration of the second environmental change time zone are input into the retaining wall structure analyzer to obtain the second-stage structural safety score. The process is iterated until the structural safety score of the N-1th stage, the environmental characteristic parameters of the Nth environmental change time zone, and the time zone duration of the Nth environmental change time zone are input into the retaining wall structure analyzer to obtain the structural safety score of the Nth stage. The structural safety score of the Nth stage is used as the second structural safety score.

[0070] Specifically, the process for obtaining the second structural safety score starts with the initial structural safety score of the retaining wall. First, the initial structural safety score, the environmental characteristic parameters corresponding to the first environmental change time zone, and the duration of the first environmental change time zone are input into the trained retaining wall structure analyzer. The retaining wall structure analyzer can simulate and calculate the safety state of the retaining wall after experiencing the environmental effects of the first environmental change time zone, and its output is the first-stage structural safety score.

[0071] Secondly, the first-stage structural safety score calculated in the previous step is used as the new starting state. This score, along with the environmental characteristic parameters and duration of the second environmental change time zone, are input again into the retaining wall structure analyzer. The retaining wall structure analyzer calculates the safety status of the retaining wall after experiencing the environmental effects of the first two consecutive environmental change time zones and outputs the second-stage structural safety score.

[0072] The above steps are performed iteratively. In each iteration, the structural safety score output from the previous iteration is used as the starting score for the current iteration. This score is combined with the characteristic parameters and duration of the current environmental change time zone and input into the retaining wall structure analyzer to calculate the structural safety score for the next stage. This iterative process continues until the last historical service period division, i.e., the Nth environmental change time zone, has been processed. Specifically, the structural safety score obtained from the (N-1)th iteration, the environmental characteristic parameters of the Nth time zone, and the duration of the Nth time zone are input into the retaining wall structure analyzer to calculate the structural safety score for the Nth stage.

[0073] Finally, the structural safety score of stage N obtained from the last iteration is used as the predicted safety state of the retaining wall under the cumulative effects of all historical environmental sequences; this score is the second structural safety score. In summary, this iterative mechanism effectively simulates the dynamic evolution of the structural safety state under the influence of time-series environmental factors.

[0074] The detection result generation module 15 is used to analyze and process the first structural safety score and the second structural safety score to generate the detection result of the target masonry retaining wall.

[0075] Specifically, the first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall, including: Calculate the score difference between the first structural safety score and the second structural safety score; When the score difference is less than or equal to a preset difference threshold, the average of the first structural safety score and the second structural safety score is calculated to obtain a comprehensive structural safety score. When the score difference is greater than a preset difference threshold, the historical inspection record corresponding to the retaining wall type is obtained. The historical inspection record includes multiple historical first structural safety scores and multiple corresponding historical actual structural safety scores. The detection accuracy rate of the first structural safety score is calculated based on the historical detection records, and the detection accuracy rate is used as the first weighting coefficient. The second weighting coefficient is determined based on the first weighting coefficient. The first structural safety score and the second structural safety score are weighted and calculated based on the first weighting coefficient and the second weighting coefficient to obtain the comprehensive structural safety score. The comprehensive structural safety score is used as the test result for the target masonry retaining wall.

[0076] First, calculate the absolute difference between the first structural safety score and the second structural safety score to obtain the score difference.

[0077] When the score difference is less than or equal to a preset difference threshold, it indicates that the assessment results based on real-time detection and the assessment results based on historical environmental predictions have a high degree of consistency, and both are considered to provide reliable references. In this case, an arithmetic mean calculation method is used to directly calculate the average of the first structural safety score and the second structural safety score, and this average is taken as the comprehensive structural safety score.

[0078] The preset difference threshold is a pre-defined numerical limit used to determine whether two scores are consistent. It represents that within the acceptable error range for engineering, the results of the two evaluation methods can be considered mutually corroborative. This preset difference threshold is set comprehensively based on the safety level requirements of the retaining wall type, the typical error range of the evaluation model, and engineering practice experience. For example, it can be set to 5% of the total score range or specifically 5 points.

[0079] When the score difference exceeds a preset threshold, it indicates a significant discrepancy between the conclusions reached by the two independent evaluation paths. To scientifically integrate the first and second structural safety scores, a dynamic weight allocation strategy based on historical performance is initiated. Specifically, historical inspection records of retaining walls of the same type as the target retaining wall are acquired. These historical inspection records include historical first structural safety scores obtained through real-time inspection paths from multiple past inspections, as well as corresponding historical actual structural safety scores confirmed through subsequent more in-depth professional assessments, long-term monitoring, or actual damage assessments.

[0080] Furthermore, based on historical detection records, the historical reliability of the real-time detection path is evaluated. Specifically, each historical first structural safety score is compared with its corresponding historical actual structural safety score. If the difference between the two is within the allowable error range, i.e., the difference is less than or equal to a preset difference threshold, then the detection is considered accurate. The number of accurate detections in all historical detections is counted, and the proportion of accurate detections to the total number of detections is calculated. This proportion is the detection accuracy rate of the first structural safety score. The calculated detection accuracy rate is directly set as the first weighting coefficient used for weighted fusion.

[0081] Secondly, the second weighting coefficient is determined based on the first weighting coefficient. The second weighting coefficient is obtained by subtracting the first weighting coefficient from 1, ensuring that the sum of the first and second weighting coefficients is always 1, which meets the mathematical requirements of weighted calculation.

[0082] Finally, based on the determined first and second weighting coefficients, the first and second structural safety scores are weighted and calculated respectively. Specifically, the comprehensive structural safety score is calculated using the following formula: Comprehensive Structural Safety Score = First Structural Safety Score × First Weighting Coefficient + Second Structural Safety Score × Second Weighting Coefficient. The final comprehensive structural safety score is a fused and quantified structural safety status index, characterizing the overall safety level and risk grade of the target masonry retaining wall under the comprehensive consideration of the current immediate monitoring status and the long-term historical environmental cumulative effects.

[0083] In summary, the core advantage of this fusion mechanism lies in its adaptability. For retaining wall types where real-time detection technology has a stable historical performance and high accuracy, it automatically assigns higher weights to the real-time detection results. Conversely, when the historical reliability of real-time detection is low, the weights based on long-term environmental prediction results are increased accordingly. As detection data accumulates, the weight coefficients can be dynamically updated, thereby continuously optimizing the accuracy of the fusion strategy.

[0084] Ultimately, the comprehensive structural safety score is used as the inspection result for the target masonry retaining wall, quantitatively and comprehensively reflecting its structural safety status. This inspection result integrates real-time assessment based on multi-source real-time detection data with predictive analysis based on long-term service environment history. Through a decision-making mechanism, it provides users with clear, reliable, and easily understandable quantitative indicators of safety status. These indicators can be directly used to guide engineering maintenance decisions, risk classification management, and the development of preventative maintenance strategies, achieving a scientific assessment of the entire chain of masonry retaining wall structural safety, from data acquisition to decision support.

[0085] In summary, the embodiments of this application have at least the following technical effects: Compared to existing technologies, this application implements a dual-path verification and intelligent fusion detection mechanism. First, it integrates multiple non-destructive testing technologies, including ultrasonic, ground-penetrating radar, and acoustic emission, to achieve a multi-dimensional and three-dimensional assessment of the current structural state of the masonry retaining wall. This overcomes the limitations of single detection methods and improves the comprehensiveness and accuracy of real-time state detection. Second, it innovatively constructs a long-term performance prediction path based on historical service environment data. By dynamically dividing time zones, it accurately quantifies the cumulative environmental effects, making the assessment results predictable and compensating for the shortcomings of traditional methods that ignore long-term time-varying influences. Third, it designs an adaptive weight fusion strategy based on the reliability of historical detection, which intelligently processes the dual-path assessment results, ensuring that the final comprehensive score combines data objectivity and engineering rationality, thus enhancing the reliability of the assessment results.

[0086] Finally, this application provides a systematic and automated detection process from type identification to result output, which can automatically adapt to special analysis models for different retaining wall types. This not only improves detection efficiency and reduces reliance on subjective experience, but also provides universal and accurate quantitative decision support for the preventive maintenance and life cycle management of masonry retaining walls.

[0087] Example 2, as Figure 2 As shown, this embodiment of the invention also provides a method for detecting masonry retaining walls, comprising: Obtain the retaining wall type of the target masonry retaining wall, and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; Extract the retaining wall detection data for the target masonry retaining wall, the retaining wall detection data including ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; The retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features, and determines a first structural safety score based on the retaining wall detection features. Based on the set of environmental impact factors, historical service environment data of the target masonry retaining wall is obtained. Based on the historical service environment data, structural safety prediction of the target masonry retaining wall is performed to obtain a second structural safety score. The first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall.

[0088] Furthermore, the retaining wall data analyzer is constructed as follows: The first retaining wall type is determined from multiple preset retaining wall types; Based on the first retaining wall type, retrieve historical retaining wall detection records of multiple masonry retaining wall samples of the same type. The historical retaining wall detection records include historical ultrasonic detection records, historical ground penetrating radar detection records, and historical acoustic emission detection records. Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed. By integrating the ultrasonic feature extraction branch, the ground penetrating radar feature extraction branch, and the acoustic emission feature extraction branch, a retaining wall data analyzer corresponding to the first retaining wall type is obtained. Obtain the retaining wall data analyzer corresponding to the first retaining wall type, and then obtain the retaining wall data analyzers corresponding to the other retaining wall types.

[0089] Specifically, based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed, including: Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, sample ultrasonic detection datasets, sample ground-penetrating radar detection datasets, and sample acoustic emission detection datasets are constructed. The retaining wall structure features were annotated on the sample ultrasonic detection dataset, sample ground penetrating radar detection dataset, and sample acoustic emission detection dataset respectively to obtain the sample ultrasonic structure feature set, sample ground penetrating radar structure feature set, and sample acoustic emission structure feature set; Machine learning training is performed based on the sample ultrasonic detection dataset and the sample ultrasonic structural feature set to generate an ultrasonic feature extraction branch. Machine learning training is performed based on the sample ground-penetrating radar detection dataset and the sample ground-penetrating radar structural feature set to generate a ground-penetrating radar feature extraction branch. Machine learning training is performed based on the sample acoustic emission detection dataset and the sample acoustic emission structural feature set to generate an acoustic emission feature extraction branch.

[0090] Specifically, the retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features, and determines a first structural safety score based on these features, including: The retaining wall data analyzer includes an ultrasonic feature extraction branch, a ground-penetrating radar feature extraction branch, and an acoustic emission feature extraction branch; The ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data are processed by ultrasonic feature extraction branch, ground-penetrating radar feature extraction branch, and acoustic emission feature extraction branch, respectively, to obtain ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features. By combining ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features, the retaining wall detection features are obtained. The retaining wall detection features are input into the structural safety evaluator, which outputs the first structural safety score.

[0091] Specifically, the structural safety evaluator is constructed as follows: Based on the retaining wall type, historical retaining wall detection features and corresponding structural safety score annotation data of multiple masonry retaining walls of the same type are collected to construct a sample retaining wall detection feature set and a sample structural safety score set. A structural safety evaluator is built using machine learning; Using the sample retaining wall detection feature set as input features and the sample structural safety score set as supervision labels, the structural safety evaluator is trained until convergence, thus completing the construction.

[0092] Further, based on the set of environmental impact factors, historical service environment data of the target masonry retaining wall is obtained. Based on this historical service environment data, structural safety prediction is performed on the target masonry retaining wall to obtain a second structural safety score, including: A first environmental impact factor is obtained from the set of environmental impact factors, and first impact factor data is extracted from historical service environment data based on the first environmental impact factor. Time series analysis is performed on the first impact factor data, and the time segmentation points corresponding to the first environmental impact factor are determined based on the numerical change characteristics to generate a first time interval sequence. Following the method of obtaining the first time interval sequence corresponding to the first environmental impact factor, the time interval sequences corresponding to the remaining environmental impact factors are obtained, resulting in multiple time interval sequences. The multiple time interval sequences are merged to generate an environmental change time zone sequence, which includes multiple environmental change time zones. Environmental characteristic parameters of each environmental change time zone are obtained based on the historical service environment data. Obtain the initial structural safety score of the target masonry retaining wall; The initial structural safety score and environmental characteristic parameters of each environmental change time zone are input into the retaining wall structure analyzer to obtain the second structural safety score.

[0093] Specifically, the retaining wall structure analyzer is constructed as follows: Based on the retaining wall type, multiple historical retaining wall service records of the same type are retrieved. Each historical retaining wall service record includes the starting structural safety score, environmental characteristic parameters, time zone duration, and ending structural safety score of multiple historical environmental change time zones. Based on the multiple historical retaining wall service records, a sample initial score set, a sample environmental feature parameter set, a sample time zone duration set, and a sample termination score set are constructed. The retaining wall structure analyzer is generated by using the sample initial score set, sample environmental feature parameter set, and sample time zone duration set as inputs, and the sample final score set as the output target, through machine learning training.

[0094] Specifically, the initial structural safety score and environmental characteristic parameters of each environmental change time zone are input into the retaining wall structure analyzer to obtain a second structural safety score, including: The initial structural safety score, the environmental characteristic parameters of the first environmental change time zone, and the time zone duration of the first environmental change time zone are input into the retaining wall structure analyzer to obtain the first-stage structural safety score. The first-stage structural safety score, the environmental characteristic parameters of the second environmental change time zone, and the time zone duration of the second environmental change time zone are input into the retaining wall structure analyzer to obtain the second-stage structural safety score. The process is iterated until the structural safety score of the N-1th stage, the environmental characteristic parameters of the Nth environmental change time zone, and the time zone duration of the Nth environmental change time zone are input into the retaining wall structure analyzer to obtain the structural safety score of the Nth stage. The structural safety score of the Nth stage is used as the second structural safety score.

[0095] Further, the first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall, including: Calculate the score difference between the first structural safety score and the second structural safety score; When the score difference is less than or equal to a preset difference threshold, the average of the first structural safety score and the second structural safety score is calculated to obtain a comprehensive structural safety score. When the score difference is greater than a preset difference threshold, the historical inspection record corresponding to the retaining wall type is obtained. The historical inspection record includes multiple historical first structural safety scores and multiple corresponding historical actual structural safety scores. The detection accuracy rate of the first structural safety score is calculated based on the historical detection records, and the detection accuracy rate is used as the first weighting coefficient. The second weighting coefficient is determined based on the first weighting coefficient. The first structural safety score and the second structural safety score are weighted and calculated based on the first weighting coefficient and the second weighting coefficient to obtain the comprehensive structural safety score. The comprehensive structural safety score is used as the test result for the target masonry retaining wall.

[0096] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0097] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0098] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A detection system for masonry retaining walls, characterized in that, include: The retaining wall type extraction module is used to obtain the retaining wall type of the target masonry retaining wall and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; The detection data extraction module is used to extract the retaining wall detection data for the target masonry retaining wall, and the retaining wall detection data includes ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; The first safety scoring module is used to analyze and process ultrasonic detection data, ground-penetrating radar detection data and acoustic emission detection data through the retaining wall data analyzer to obtain retaining wall detection features, and determine the first structural safety score based on the retaining wall detection features. The second safety scoring module is used to obtain historical service environment data of the target masonry retaining wall based on the set of environmental impact factors, perform structural safety prediction of the target masonry retaining wall based on the historical service environment data, and obtain a second structural safety score. The detection result generation module is used to analyze and process the first structural safety score and the second structural safety score to generate the detection result of the target masonry retaining wall.

2. The detection system for masonry retaining walls according to claim 1, characterized in that, The retaining wall data analyzer is constructed as follows: The first retaining wall type is determined from multiple preset retaining wall types; Based on the first retaining wall type, retrieve historical retaining wall detection records of multiple masonry retaining wall samples of the same type. The historical retaining wall detection records include historical ultrasonic detection records, historical ground penetrating radar detection records, and historical acoustic emission detection records. Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed. By integrating the ultrasonic feature extraction branch, the ground penetrating radar feature extraction branch, and the acoustic emission feature extraction branch, a retaining wall data analyzer corresponding to the first retaining wall type is obtained. Obtain the retaining wall data analyzer corresponding to the first retaining wall type, and then obtain the retaining wall data analyzers corresponding to the other retaining wall types.

3. The detection system for masonry retaining walls according to claim 2, characterized in that, Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, ultrasonic feature extraction branches, ground-penetrating radar feature extraction branches, and acoustic emission feature extraction branches are constructed, including: Based on historical ultrasonic detection records, historical ground-penetrating radar detection records, and historical acoustic emission detection records, sample ultrasonic detection datasets, sample ground-penetrating radar detection datasets, and sample acoustic emission detection datasets are constructed. The retaining wall structure features were annotated on the sample ultrasonic detection dataset, sample ground penetrating radar detection dataset, and sample acoustic emission detection dataset respectively to obtain the sample ultrasonic structure feature set, sample ground penetrating radar structure feature set, and sample acoustic emission structure feature set; Machine learning training is performed based on the sample ultrasonic detection dataset and the sample ultrasonic structural feature set to generate an ultrasonic feature extraction branch. Machine learning training is performed based on the sample ground-penetrating radar detection dataset and the sample ground-penetrating radar structural feature set to generate a ground-penetrating radar feature extraction branch. Machine learning training is performed based on the sample acoustic emission detection dataset and the sample acoustic emission structural feature set to generate an acoustic emission feature extraction branch.

4. The detection system for masonry retaining walls according to claim 1, characterized in that, The retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features. Based on these features, a first structural safety score is determined, including: The retaining wall data analyzer includes an ultrasonic feature extraction branch, a ground-penetrating radar feature extraction branch, and an acoustic emission feature extraction branch; The ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data are processed by ultrasonic feature extraction branch, ground-penetrating radar feature extraction branch, and acoustic emission feature extraction branch, respectively, to obtain ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features. By combining ultrasonic detection features, ground-penetrating radar detection features, and acoustic emission detection features, the retaining wall detection features are obtained. The retaining wall detection features are input into the structural safety evaluator, which outputs the first structural safety score.

5. The detection system for masonry retaining walls according to claim 4, characterized in that, The structural safety assessor is constructed as follows: Based on the retaining wall type, historical retaining wall detection features and corresponding structural safety score annotation data of multiple masonry retaining walls of the same type are collected to construct a sample retaining wall detection feature set and a sample structural safety score set. A structural safety evaluator is built using machine learning; Using the sample retaining wall detection feature set as input features and the sample structural safety score set as supervision labels, the structural safety evaluator is trained until convergence, thus completing the construction.

6. The detection system for masonry retaining walls according to claim 1, characterized in that, Based on the environmental impact factor set, historical service environment data of the target masonry retaining wall is obtained. Based on this historical service environment data, structural safety prediction of the target masonry retaining wall is performed, and a second structural safety score is obtained, including: A first environmental impact factor is obtained from the set of environmental impact factors, and first impact factor data is extracted from historical service environment data based on the first environmental impact factor. Time series analysis is performed on the first impact factor data, and the time segmentation points corresponding to the first environmental impact factor are determined based on the numerical change characteristics to generate a first time interval sequence. Following the method of obtaining the first time interval sequence corresponding to the first environmental impact factor, the time interval sequences corresponding to the remaining environmental impact factors are obtained, resulting in multiple time interval sequences. The multiple time interval sequences are merged to generate an environmental change time zone sequence, which includes multiple environmental change time zones. Environmental characteristic parameters of each environmental change time zone are obtained based on the historical service environment data. Obtain the initial structural safety score of the target masonry retaining wall; The initial structural safety score and environmental characteristic parameters of each environmental change time zone are input into the retaining wall structure analyzer to obtain the second structural safety score.

7. The detection system for masonry retaining walls according to claim 6, characterized in that, The retaining wall structure analyzer is constructed as follows: Based on the retaining wall type, multiple historical retaining wall service records of the same type are retrieved. Each historical retaining wall service record includes the starting structural safety score, environmental characteristic parameters, time zone duration, and ending structural safety score of multiple historical environmental change time zones. Based on the multiple historical retaining wall service records, a sample initial score set, a sample environmental feature parameter set, a sample time zone duration set, and a sample termination score set are constructed. The retaining wall structure analyzer is generated by using the sample initial score set, sample environmental feature parameter set, and sample time zone duration set as inputs, and the sample final score set as the output target, through machine learning training.

8. The detection system for masonry retaining walls according to claim 7, characterized in that, The initial structural safety score and environmental characteristic parameters for each environmental change time zone are input into the retaining wall structure analyzer to obtain the second structural safety score, including: The initial structural safety score, the environmental characteristic parameters of the first environmental change time zone, and the time zone duration of the first environmental change time zone are input into the retaining wall structure analyzer to obtain the first-stage structural safety score. The first-stage structural safety score, the environmental characteristic parameters of the second environmental change time zone, and the time zone duration of the second environmental change time zone are input into the retaining wall structure analyzer to obtain the second-stage structural safety score. The process is iterated until the structural safety score of the N-1th stage, the environmental characteristic parameters of the Nth environmental change time zone, and the time zone duration of the Nth environmental change time zone are input into the retaining wall structure analyzer to obtain the structural safety score of the Nth stage. The structural safety score of the Nth stage is used as the second structural safety score.

9. The detection system for masonry retaining walls according to claim 1, characterized in that, The first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall, including: Calculate the score difference between the first structural safety score and the second structural safety score; When the score difference is less than or equal to a preset difference threshold, the average of the first structural safety score and the second structural safety score is calculated to obtain a comprehensive structural safety score. When the score difference is greater than a preset difference threshold, the historical inspection record corresponding to the retaining wall type is obtained. The historical inspection record includes multiple historical first structural safety scores and multiple corresponding historical actual structural safety scores. The detection accuracy rate of the first structural safety score is calculated based on the historical detection records, and the detection accuracy rate is used as the first weighting coefficient. The second weighting coefficient is determined based on the first weighting coefficient. The first structural safety score and the second structural safety score are weighted and calculated based on the first weighting coefficient and the second weighting coefficient to obtain the comprehensive structural safety score. The comprehensive structural safety score is used as the test result for the target masonry retaining wall.

10. A method for detecting masonry retaining walls, characterized in that, The method, applied to the detection system for masonry retaining walls according to any one of claims 1-9, comprises: Obtain the retaining wall type of the target masonry retaining wall, and extract the environmental impact factor set and retaining wall data analyzer corresponding to the retaining wall type; Extract the retaining wall detection data for the target masonry retaining wall, the retaining wall detection data including ultrasonic detection data, ground penetrating radar detection data and acoustic emission detection data; The retaining wall data analyzer analyzes and processes ultrasonic detection data, ground-penetrating radar detection data, and acoustic emission detection data to obtain retaining wall detection features, and determines a first structural safety score based on the retaining wall detection features. Based on the set of environmental impact factors, historical service environment data of the target masonry retaining wall is obtained. Based on the historical service environment data, structural safety prediction of the target masonry retaining wall is performed to obtain a second structural safety score. The first structural safety score and the second structural safety score are analyzed and processed to generate the detection results of the target masonry retaining wall.

Images