Basement exterior wall crack leakage risk intelligent prediction method, storage medium and equipment

Abstract

The invention relates to the technical field of building construction, in particular to a basement exterior wall crack leakage risk intelligent prediction method, a storage medium and equipment, and the method comprises the steps: collecting influence factor data related to basement exterior wall crack leakage; extracting crack image feature data; analyzing the influence factor data and the crack image feature data by using a strong association rule, and identifying potential association factors of crack leakage; constructing a deep learning model, and training the deep learning model by using the influence factor data and the crack image feature data; and inputting basement exterior wall monitoring data needing to be evaluated into the trained deep learning model to output a prediction result. Through the combination of the strong association rule and the deep learning model, a basis is provided for early recognition of the basement exterior wall crack leakage problem, the accuracy and comprehensiveness of a prediction result are improved, the basement crack leakage risk can be timely and effectively prevented and processed, and the safety of a building is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of building construction, and particularly to an intelligent prediction method, a storage medium, and a device for the risk of cracks and leakage in the basement exterior wall. Background Art

[0002] Cracks in the basement exterior wall are one of the common structural defects in buildings. The main reasons include the influence of temperature stress, shrinkage stress, and groundwater, etc., which may lead to leakage problems during the use of the building. If the leakage risk of cracks in the basement exterior wall is not prevented and treated in a timely and effective manner, it will seriously affect the safety of the building and the health and comfort of the occupants, resulting in huge economic losses. Therefore, it is very important to predict the leakage risk of cracks in the basement exterior wall.

[0003] Traditional methods for predicting the leakage risk of cracks in the basement exterior wall mainly rely on the engineering experience and on-site inspections of engineers or experts. A large amount of on-site inspections and manual measurements are required for data collection. Facing a large amount of data, it is impossible to analyze it efficiently and comprehensively. Therefore, these factors limit the comprehensiveness and accuracy of the analysis results. In the prior art, problems can only be discovered when cracks have occurred and obvious leakage phenomena have occurred, and potential risks cannot be predicted in advance. There is an obvious lag, and the safety of buildings and residents cannot be effectively protected. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent prediction method, a storage medium, and a device for the risk of cracks and leakage in the basement exterior wall, which improve the timeliness, accuracy, and comprehensiveness of predicting the leakage risk of cracks in the basement exterior wall by combining strong association rules and deep learning models.

[0005] To solve the above technical problems, an embodiment of the present invention provides a technical solution as follows:

[0006] An intelligent prediction method for the risk of cracks and leakage in the basement exterior wall, comprising the following steps:

[0007] S1: Collect data on influencing factors related to cracks and leakage in the basement exterior wall, where the influencing factor data includes material factors, construction factors, and environmental factors;

[0008] S2: Extract crack image feature data;

[0009] S3: Analyze the influencing factor data and the crack image feature data using strong association rules to generate association rules and identify potential associated factors for crack leakage;

[0010] S4: Build a deep learning model, train the deep learning model using the influencing factor data and the crack image feature data, and optimize the deep learning model in combination with the potential associated factors of crack leakage to improve the model performance;

[0011] S5: Input the monitoring data of the basement exterior wall to be evaluated into the trained deep learning model to output the prediction results.

[0012] Further, the material factors include the elastic modulus of concrete, the linear expansion coefficient, and the slump of concrete; the construction factors include the vertical main reinforcement spacing, the horizontal stirrup spacing, the construction joint setting, the one-time pouring height, the curing time, and the curing quality; the environmental factors include the temperature difference, the shrinkage difference, and the groundwater.

[0013] Further, step S2 includes taking continuous crack images, performing preprocessing on the crack images to reduce image noise and correct distortion, and performing semantic segmentation on the preprocessed images based on the algorithm of the deep neural network to extract the image feature data.

[0014] Further, step S3 includes generating item sets based on the combination of the influencing factor data and the crack image feature data, setting the minimum support and the minimum confidence according to the experience in the field, using the strong association rule mining technology to discover the frequent item sets, generating association rules based on the frequent item sets, and identifying the potential associated factors of crack leakage.

[0015] Further, optimize the potential associated factors of crack leakage based on the expert evaluation suggestions and domain knowledge, optimize and filter the potential associated factors to improve the physical interpretability of the model.

[0016] Further, in step S4, the deep learning model includes at least a convolutional neural network and a recurrent neural network. The convolutional neural network is used to process the crack image feature data, and the recurrent neural network is used to process the influencing factor data. The outputs of the convolutional neural network and the recurrent neural network are spliced and further processed through a fully connected layer to output the crack leakage risk probability.

[0017] Further, the optimization of the deep learning model in combination with the potential associated factors of crack leakage includes deleting the features with little or no impact on the output results, the highly correlated or redundant features, and the features that cannot be accurately measured or explained, adjusting the model parameters according to the domain knowledge or cross-validation results, increasing or decreasing the number of hidden layers, modifying the activation function or the learning rate, and adjusting the regularization parameters.

[0018] Further, in step S5, the output results include the crack leakage risk probability, the estimated crack characteristics, and the maintenance suggestions.

[0019] To solve the technical problems raised by the present invention, the present invention also provides a computer program storage medium, on which a computer program is stored. When the computer program is executed by a processor, it can implement any one of the above-mentioned intelligent prediction methods for the risk of cracks and leakage in the basement exterior wall.

[0020] To solve the technical problems raised by the present invention, the present invention also provides a computer program device, including a computer program. When the computer program is executed by a processor, it can implement any one of the above-mentioned intelligent prediction methods for the risk of cracks and leakage in the basement exterior wall.

[0021] The intelligent prediction method, storage medium and device for the risk of cracks and leakage in the basement exterior wall provided by the present invention, compared with the prior art, through the combination of strong association rules and deep learning models, provides an intelligent prediction method for the risk of cracks and leakage in the basement exterior wall. Through data collection, construction of a deep learning model, and training of the deep learning model, the intelligent prediction of the risk of cracks and leakage in the basement exterior wall can be realized, providing a basis for early identification of the problem of cracks and leakage in the basement exterior wall, improving the timeliness of the risk prediction of cracks and leakage in the basement exterior wall, increasing the accuracy and comprehensiveness of the prediction results, being able to provide decision-making support for building management personnel, being able to prevent and handle the risk of cracks and leakage in the basement in a timely and effective manner, enhancing the safety of the building, and reducing economic losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] One or more embodiments are exemplarily illustrated by the pictures in the corresponding drawings. These exemplary illustrations do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements, unless otherwise stated, and the drawings in the figures do not constitute a proportional limitation.

[0023] Figure 1 is the flowchart of the intelligent prediction method for the risk of cracks and leakage in the basement exterior wall in the embodiment of the present invention;

[0024] Figure 2 is the architecture diagram of the deep learning model in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be described in detail below with reference to the drawings. However, those of ordinary skill in the art can understand that in the embodiments of the present invention, many technical details are provided to help readers better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions required to be protected by the claims of the present application can still be implemented.

[0026] Such as Figure 1As shown, an embodiment of the present invention relates to an intelligent prediction method for the risk of cracks and leakage in the basement exterior wall, comprising the following steps:

[0027] Step S1: Collect data on influencing factors related to cracks and leakage in the basement exterior wall, where the influencing factor data includes material factors, construction factors, and environmental factors.

[0028] In one embodiment, the influencing factors affecting cracks and leakage in the basement exterior wall mainly include material factors, construction factors, and environmental factors. Among them, material factors include the elastic modulus of concrete, coefficient of linear expansion, slump of concrete, etc. The elastic modulus is an important mechanical property index of concrete, and its magnitude directly affects the deformation and stress distribution of the structure. Concrete with a lower elastic modulus is more likely to deform when subjected to external forces, which may lead to the appearance of cracks. When collecting data, the specific value of its elastic modulus can be obtained through laboratory tests before the production and pouring of concrete. The coefficient of linear expansion reflects the degree of expansion and contraction of concrete when the temperature changes. When the ambient temperature in the basement changes, concrete with a larger coefficient of linear expansion will produce a larger expansion and contraction deformation. If this deformation is restricted, it is easy to generate temperature stress in the wall, which may further trigger cracks. Data collection can be carried out during the concrete material inspection stage and tested according to relevant standards. The slump is an important index to measure the workability of concrete. Too large or too small slump will affect the construction performance and final quality of concrete. Too large slump may lead to segregation and bleeding of concrete, making the internal structure of the wall uneven and the strength reduced, thus increasing the risk of cracks. Too small slump will make it difficult to pour the concrete densely, which also affects the wall quality. The slump test can be carried out after the concrete is mixed and transported to the construction site and before pouring, and the relevant data can be recorded.

[0029] Construction factors include the spacing of vertical main reinforcement bars, the spacing of horizontal stirrups, the setting of construction joints, the one-time pouring height, the curing time, and the curing quality, etc. The size of the spacing of vertical main reinforcement bars affects the restraint effect of the reinforcement on the concrete. If the spacing is too large, the restraint force of the reinforcement on the concrete weakens, and when subjected to external loads or temperature changes, etc., the concrete is more likely to crack. If the spacing is too small, it will increase the construction difficulty and cost. The actual spacing of vertical main reinforcement bars can be obtained through on-site measurement during the steel bar binding and installation process. Similarly, the spacing of horizontal stirrups also affects the restraint effect of the reinforcement on the concrete. A reasonable stirrup spacing can effectively limit the lateral deformation of the concrete and reduce the generation of cracks. The data collection method is similar to that of the spacing of vertical main reinforcement bars, and it is also measured and recorded during the steel bar construction stage. A construction joint is a temporary discontinuity left during the construction process due to technical or organizational reasons. The setting position, treatment method, etc. of the construction joint will all affect the integrity of the basement exterior wall. If the construction joint is not properly treated, such as not being cleaned thoroughly or the joint being not compact, it is likely to become a weak link for leakage. The data collection requires detailed recording of relevant information such as the position, width, and treatment method of the construction joint during the setting and treatment of the construction joint. If the one-time pouring height is too high, it will increase the lateral pressure of the concrete on the formwork, which may cause the formwork to deform or leak, affecting the wall quality. At the same time, the temperature gradient inside the concrete will also increase, and temperature cracks are likely to occur. The data collection can be recorded according to the one-time pouring height determined by the construction plan before the concrete pouring. If the curing time is insufficient, the strength development of the concrete is not sufficient, and its early crack resistance is poor, and it is easy to crack when affected by external adverse factors. The data collection can determine the specific start and end times of curing according to the construction records after the concrete pouring is completed. Good curing quality can ensure that the concrete maintains appropriate temperature and humidity conditions during the hardening process, which is beneficial to improving the strength and durability of the concrete and reducing the generation of cracks. The data collection can evaluate and record the curing quality by checking the implementation of curing measures, such as the sprinkling frequency and the integrity of the covering material.

[0030] Environmental factors include temperature difference, shrinkage difference, groundwater, etc. The temperature difference between the inside and outside of the basement and the temperature change inside the basement will both affect the exterior wall of the basement. A large temperature difference will cause temperature stress in the wall. When the temperature stress exceeds the crack resistance of the wall, cracks will occur. Data collection can be carried out by real-time monitoring of the environmental temperature inside and outside the basement and recording the temperature change within a certain time range, such as the daily maximum temperature, minimum temperature, and temperature change rate, etc. Shrinkage difference refers to the shrinkage deformation that occurs in concrete during the hardening process due to reasons such as water evaporation. The shrinkage degree of concrete in different parts may vary, and this shrinkage difference will cause tensile stress inside the wall, thus triggering cracks. Data collection can be carried out by embedding sensors such as strain gauges in the concrete to monitor the shrinkage deformation of the concrete in real time and calculate the shrinkage difference. The existence of groundwater will generate water pressure and seepage on the exterior wall of the basement. The water pressure will cause the wall to be subjected to an outward thrust. If the wall strength is insufficient or there are defects, cracks are likely to occur. The seepage effect may dissolve harmful substances in the wall, resulting in a decrease in wall strength. At the same time, it will also corrode the steel bars, further affecting the crack resistance of the wall. Data collection can be carried out by monitoring the groundwater level to understand the change of groundwater. At the same time, pressure sensors can also be embedded in the wall to monitor the pressure of groundwater on the wall.

[0031] The time range for data collection of influencing factors is set as follows: For newly built basement projects, data collection should start from the geological exploration and design stages in the early stage of the project, run through the entire construction process, and continue for a period of time after the project completion acceptance, so as to monitor and evaluate the performance of the basement in the initial use stage. For the prediction of crack leakage risks of existing basements, the time range for data collection can start from the present and trace back a certain period of time to obtain sufficient historical data for analysis and modeling. The collection of historical data can be carried out by means of consulting drawings, on-site investigation, experimental testing, data mining, research interviews, and literature review.

[0032] The geographical range for data collection of influencing factors should be determined according to specific engineering projects. For a single basement project, the geographical range is mainly concentrated in the project location and a certain area around it to obtain environmental, geological, and other data related to the project. If research or prediction on the crack leakage risks of regional basement exterior walls is carried out, the geographical range can be expanded to the entire city or a specific area to collect relevant data within this area, including the groundwater level, geological conditions, climate characteristics, etc. in different sections, so as to more comprehensively analyze and predict the regional differences in crack leakage risks of basement exterior walls.

[0033] Step S2: Extract crack image feature data.

[0034] In one embodiment, extracting the crack image feature data specifically includes the following steps: taking consecutive crack images, where there may be overlap or non - overlap between consecutive crack images, and the overlapping area of the overlapping part accounts for 30% - 50% of the entire image; pre - processing the taken crack images, including reducing image noise and correcting distortion; performing semantic segmentation on the pre - processed images based on the deep neural network algorithm to identify and distinguish each crack area, forming a two - dimensional image of the crack, and recording the pixel values, the number of pixel points, and the position information of each crack area; extracting the image feature data, including the crack length, width, distribution position, and the positional relationship between the crack and other structures; preferably, collecting the crack feature data through sensors arranged around the exterior wall, and combining the sensor data to accurately describe the morphological features of the crack, including the depth, angle of the crack, and the three - dimensional geometric shape of the crack and the structural interface, etc., to form the crack image feature data.

[0035] S3: Use strong association rules to analyze the influencing factor data and the crack image feature data, generate association rules, and identify the potential associated factors of crack leakage. By using the strong association rule mining technology to find the potential associated factors affecting crack leakage, it helps to identify and eliminate potential risk factors, achieve intelligent, predictive, and precise decision - making support, helps to improve the service life and safety of the basement exterior wall, and realizes intelligent maintenance.

[0036] In one embodiment, generate item sets based on the combination of the influencing factor data and the crack image feature data, set the minimum support and minimum confidence according to the experience in the field, use the strong association rule mining technology to find frequent item sets, generate association rules based on the frequent item sets, and identify the potential associated factors of crack leakage. Preferably, based on expert evaluation suggestions and domain knowledge, optimize and filter the potential associated factors to improve the physical interpretability of the model.

[0037] The use of the strong association rule algorithm aims to mine the potential correlation between the crack influencing factor data and the crack characteristics, so as to identify the key factors leading to crack leakage. According to industry experience, there is a correlation between material factors and cracks: material properties such as the elastic modulus and linear expansion coefficient of concrete directly affect the formation and development of cracks. For example, concrete with a lower elastic modulus is more likely to deform when subjected to external forces, which may lead to the appearance of cracks; concrete with a larger linear expansion coefficient will produce larger expansion and contraction deformations when the temperature changes. If this deformation is restricted, it is easy to generate temperature stress in the wall, which in turn triggers cracks.

[0038] Construction factors are related to cracks: Parameters during the construction process, such as the spacing of vertical main reinforcement, the spacing of horizontal stirrups, the slump of concrete, the setting of construction joints, the height of one-time pouring, the curing time, and the curing quality, etc., have an important impact on the generation of cracks. Reasonable spacing of vertical main reinforcement and horizontal stirrups can effectively restrain the deformation of concrete and reduce the occurrence of cracks; an appropriate slump of concrete is conducive to ensuring the pouring quality of concrete and avoiding cracks caused by problems such as segregation and bleeding; the reasonable setting and treatment of construction joints, appropriate pouring height, and good curing conditions all contribute to improving the integrity and crack resistance of the wall.

[0039] Environmental factors are related to cracks: Environmental factors such as temperature difference, shrinkage difference, and the influence of groundwater are also important inducements for crack formation. A large temperature difference will cause temperature stress in the wall, and when it exceeds the crack resistance of the wall, cracks will appear; the shrinkage difference will cause tensile stress inside the wall, thereby triggering cracks; the pressure and seepage action of groundwater may cause the wall to deform or erode the wall material, reducing the strength and crack resistance of the wall, thus leading to the generation and leakage of cracks.

[0040] By applying strong association rules, frequent itemsets are discovered, as shown in the following examples: Crack width and coefficient of linear expansion. When the temperature changes, concrete with a larger coefficient of linear expansion is more likely to produce wider cracks because its expansion and contraction deformation is larger, and the temperature stress generated after being restricted is also larger, thus leading to an increase in the width of the cracks; therefore, there may be an association between a larger crack width (such as greater than 0.3 mm) and a larger coefficient of linear expansion of concrete (such as greater than 1×10⁻ 5 / ℃).

[0041] Crack depth and the spacing of vertical main reinforcement. When the spacing of vertical main reinforcement is larger, the restraint effect of the reinforcement on the concrete weakens, and the concrete is more likely to produce deeper cracks when subjected to external forces or its own shrinkage, etc. There may be an association between a deeper crack depth (such as exceeding 1 / 3 of the wall thickness) and a larger spacing of vertical main reinforcement (such as greater than 20 cm).

[0042] Crack distribution density and the slump of concrete. An excessive slump may cause segregation and bleeding of concrete, making the internal structure of the wall uneven and reducing the strength, thus increasing the crack distribution density; too small a slump will make it difficult to pour the concrete densely, which also affects the wall quality and leads to an increase in the crack distribution density. There may be an association between a higher crack distribution density (such as the number of cracks on the wall surface per square meter exceeding 10) and a larger (such as greater than 20 cm) or smaller (such as less than 5 cm) slump of concrete.

[0043] Crack leakage probability and groundwater level height. When the groundwater level is relatively high, the water pressure on the basement wall increases. If there are cracks in the wall, water is more likely to seep into the room through the cracks, thus increasing the crack leakage probability. Therefore, there may be a correlation between a relatively high crack leakage probability (such as exceeding 50%) and a relatively high groundwater level height (such as more than 1 m above the basement floor).

[0044] Crack length and one-time pouring height: If the one-time pouring height is too high, it will increase the lateral pressure of the concrete on the formwork, which may cause the formwork to deform or leak slurry, affecting the wall quality; at the same time, the temperature gradient inside the concrete will also increase, making it easy to generate temperature cracks, thus increasing the crack length. Therefore, there may be a correlation between a relatively long crack length (such as exceeding 3 m) and a relatively high one-time pouring height (such as exceeding 5 m).

[0045] Based on expert evaluation suggestions and domain knowledge, the association rules generated using strong association rules need to be mutually verified with expert evaluation and domain knowledge to promote the improvement of the physical interpretability of the model and the accuracy of model prediction.

[0046] S4: Construct a deep learning model, use the influencing factor data and crack image feature data to train the deep learning model, and optimize the deep learning model in combination with the potential associated factors of crack leakage to improve the model performance;

[0047] In one embodiment, constructing a deep learning model includes at least CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network) for processing image and time series data; CNN is a feedforward neural network, and its artificial neurons can respond to the surrounding units within a part of the coverage range and are affected by adjacent units, regardless of the network structure. RNN is a time-recursive neural network, and its internal state can show the processing of time series of arbitrary lengths and can be applied to process sequences and time series data; the convolutional neural network is used to process crack image feature data, the recurrent neural network is used to process influencing factor data, the outputs of the convolutional neural network and the recurrent neural network are spliced, and further processed through a fully connected layer to output the crack leakage risk probability.

[0048] Train the deep learning model using the influencing factor data and the crack image feature data, optimize the deep learning model by combining the potential associated factors of crack leakage to improve the model performance. In one embodiment, train the deep learning model using the historical data collected previously and optimize the feature selection and model structure; the steps include deleting or modifying the features that affect the prediction results to improve the interpretability and robustness of the model. The deleted features include, but are not limited to, the features with little or no impact on the results, the highly correlated or redundant features, and the features that cannot be accurately measured or interpreted; adjust the model parameters according to the domain knowledge or cross-validation results, including the number of hidden layers, activation function, learning rate, regularization, etc., to improve the model prediction performance. The model structure can be optimized by increasing or decreasing the number of hidden layers, modifying the activation function or learning rate, and adjusting the regularization parameters to improve the model performance and the accuracy and comprehensiveness of the model output results.

[0049] As Figure 2 shown, in one embodiment, the deep learning model architecture is as follows:

[0050] Input layer

[0051] Image input: 224×224×3

[0052] Time series data input: Input time series data such as temperature, humidity, groundwater level, etc. The length of the time series is T, and the feature dimension is D

[0053] CNN part

[0054] Convolutional layer 1

[0055] Number of filters: 64

[0056] Kernel size: 3x3

[0057] Activation function: ReLU

[0058] Padding: same

[0059] Output size: 224×224×64

[0060] Batch normalization layer: Used to accelerate the training process and improve the stability and performance of the model.

[0061] Max pooling layer 1: 2x2 pooling

[0062] Output size: 112×112×64

[0063] Convolutional layer 2

[0064] Number of filters: 128

[0065] Kernel size: 3x3

[0066] Activation function: ReLU

[0067] Padding: same

[0068] Output size: 112×112×128

[0069] Batch normalization layer: Continuously accelerates the training process and improves the stability and performance of the model.

[0070] Max pooling layer 2: 2x2 pooling

[0071] Output size: 56×56×128

[0072] Convolutional layer 3

[0073] Number of filters: 256

[0074] Kernel size: 3x3

[0075] Activation function: ReLU

[0076] Padding: same

[0077] Output size: 56×56×256

[0078] Batch normalization layer: Further accelerates the training process and improves the stability and performance of the model.

[0079] Global average pooling layer

[0080] Output size: 256

[0081] RNN part

[0082] LSTM layer 1:

[0083] Number of units: 128

[0084] Return sequence: True

[0085] Output size: T×128

[0086] Batch normalization layer: Used to accelerate the training process and improve the stability and performance of the model.

[0087] Dropout layer: dropout rate is 0.5, used to prevent overfitting.

[0088] LSTM layer 2:

[0089] Number of units: 128

[0090] Return sequence: False

[0091] Output size: 128

[0092] Dropout layer: The dropout rate is 0.5, which is used to prevent overfitting.

[0093] Coupling part

[0094] Concatenation layer

[0095] Concatenate the outputs of CNN (256) and RNN (128).

[0096] Output size: 384

[0097] Fully connected layer 1:

[0098] Number of neurons: 256

[0099] Activation function: ReLU

[0100] Output size: 256

[0101] Dropout layer: The dropout rate is 0.5, which is used to prevent overfitting.

[0102] Fully connected layer 2:

[0103] Number of neurons: 128

[0104] Activation function: ReLU

[0105] Output size: 128

[0106] Dropout layer: 0.5

[0107] Output layer

[0108] Number of neurons: 1

[0109] Activation function: Sigmoid

[0110] Output size: 1

[0111] Output content: The probability of crack leakage risk (between 0 and 1).

[0112] During the model training and optimization process, by randomly initializing or using the parameters of a pre-trained model, inputting training data, calculating the output of the model, using a loss function (such as cross-entropy loss, mean squared error, etc.) to calculate the difference between the model output and the true labels, calculating the gradient of the loss function with respect to the model parameters, and using an optimization algorithm (such as SGD, Adam, etc.) to update the model parameters until the performance of the model on the validation set no longer improves or reaches a predetermined number of training epochs. Evaluate the model performance through accuracy, precision, recall, F1-score, and AUC value to help us understand the performance of the model on the training data and ensure the generalization ability of the model on unknown data.

[0113] Through the above steps and evaluation metrics, the deep learning model can be systematically trained and optimized, improving the adaptability and robustness of the model. By deleting or modifying features, the model's dependence on irrelevant or redundant data is reduced, ensuring the accuracy and generalization ability of the model in the task of predicting the risk of cracks and leakage in the basement exterior wall, thereby providing reliable decision-making support for building managers.

[0114] S5: Input the monitoring data of the basement exterior wall to be evaluated into the trained deep learning model to output the prediction results. The result output includes the probability of crack leakage risk, estimated crack characteristics, maintenance suggestions, etc.

[0115] In one embodiment, data collection is performed on the basement exterior wall to be evaluated. The data includes structural monitoring data, environmental parameters, material properties, etc. The data collection can be carried out through monitoring devices or sensors to collect data in real time. After converting the data format, it is input into the deep learning model. Through the model for data analysis and prediction, the probability of crack leakage risk and the generation location and distribution density are output. The estimated crack characteristics such as the development trend of crack length, crack width, and crack depth are output. Maintenance suggestions such as concrete structure reinforcement, material repair, environmental optimization, and maintenance are output, providing decision-making support for users, reducing safety problems and economic losses caused by crack leakage, and enabling the building to reach the best state during use. By combining real-time data collection with the deep learning model, rapid and real-time prediction result output can be achieved, realizing the risk prediction of the crack development trend, providing a comprehensive risk decision-making reference, providing an earlier and more accurate basis for taking maintenance measures in advance, helping building managers formulate maintenance plans, improving building safety, and contributing to reducing failure and maintenance costs.

[0116] In an actual project case application, multiple cracks and leakage problems occurred in the exterior wall of a large underground garage during use, seriously affecting the normal use and safety of the garage. Traditional crack leakage detection methods rely on manual inspections and empirical judgments, which are not only inefficient but also difficult to predict potential risks in advance. By adopting the intelligent prediction method for crack leakage risk in the basement exterior wall based on the combination of strong association rules and deep learning provided by the present invention, it helps managers discover potential risks in a timely manner, take preventive measures, improve the safety of the underground garage, and at the same time, by predicting and preventing crack leakage in advance, the economic losses caused by crack leakage are reduced. It is estimated that the maintenance cost can be saved by about 30% per year.

[0117] One embodiment of the present invention relates to a computer program storage medium, on which a computer program is stored. When the computer program is executed by a processor, it can implement any of the above intelligent prediction methods for crack leakage risk in the basement exterior wall.

[0118] One embodiment of the present invention relates to a computer program device, including a computer program which, when processed and executed, can implement any of the above-mentioned intelligent prediction methods for the risk of basement exterior wall crack leakage.

[0119] The intelligent prediction method, storage medium and device for the risk of basement exterior wall crack leakage provided by the present invention, through the combination of strong association rules and deep learning models, provide an intelligent prediction method for the risk of basement exterior wall crack leakage, which can realize the automatic and accurate risk prediction of basement exterior wall crack leakage. Through data collection, construction of a deep learning model, and training of the deep learning model, the intelligent prediction of the risk of basement exterior wall crack leakage can be realized, providing a basis for the early identification of basement exterior wall crack leakage problems, helping building managers take maintenance measures in advance, improving the timeliness of the risk prediction of basement exterior wall crack leakage, increasing the accuracy and comprehensiveness of the prediction results, being able to provide decision-making support for building managers, being able to prevent and handle the risk of basement crack leakage in a timely and effective manner, enhancing the safety of the building, reducing subsequent maintenance and repair costs, and reducing economic losses.

[0120] Those of ordinary skill in the art can understand that the above embodiments are specific examples for implementing the present invention, and in actual applications, various changes can be made in form and details without departing from the spirit and scope of the present invention.

Claims

1. A method for intelligently predicting the risk of leakage from cracks in basement exterior walls, characterized in that: The steps include: S1: Collecting data on influencing factors related to crack leakage in basement exterior walls, including material factors, construction factors and environmental factors; S2: extract crack image feature data; S3: Use strong association rules to analyze the influencing factor data and crack image feature data, generate association rules, and identify potential associated factors of crack leakage; S4: construct a deep learning model, use the influencing factor data and the crack image feature data to train the deep learning model, and optimize the deep learning model in combination with the potential correlation factors of crack leakage to improve the model performance; S5: Input the basement exterior wall monitoring data that needs to be evaluated into the trained deep learning model to output the prediction results.

2. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 1 is characterized in that: The material factors include the elastic modulus, linear expansion coefficient and slump of concrete; the construction factors include the vertical main reinforcement spacing, horizontal stirrup spacing, construction joint setting, single pouring height, curing time and curing quality; the environmental factors include temperature difference, shrinkage difference and groundwater.

3. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 1 is characterized in that: Step S2 includes shooting continuous crack images, preprocessing the crack images to reduce image noise and correct distortion, performing semantic segmentation on the preprocessed images based on a deep neural network algorithm, and extracting image feature data.

4. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 1 is characterized in that: Step S3 includes generating item sets based on the combination of influencing factor data and crack image feature data, setting the minimum support and minimum confidence according to the experience in this field, using strong association rule mining technology to find frequent item sets, generating association rules based on frequent item sets, and identifying potential associated factors of crack leakage.

5. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 4 is characterized in that: Based on expert evaluation suggestions and domain knowledge, the potential correlation factors of fracture leakage are optimized and filtered to improve the physical interpretability of the model.

6. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 1 is characterized in that: In step S4, the deep learning model includes at least a convolutional neural network and a recurrent neural network, the convolutional neural network is used to process crack image feature data, and the recurrent neural network is used to process influencing factor data. The outputs of the convolutional neural network and the recurrent neural network are spliced, further processed through a fully connected layer, and the crack leakage risk probability is output.

7. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 6 is characterized in that: The deep learning model is optimized by combining potential correlation factors of crack leakage, including deleting features that have little effect or no effect on the output results, highly correlated or redundant features, and features that cannot be accurately measured or explained, adjusting model parameters according to domain knowledge or cross-validation results, increasing or decreasing the number of hidden layers, modifying activation functions or learning rates, and adjusting regularization parameters.

8. The intelligent prediction method for basement exterior wall crack leakage risk according to claim 1 is characterized in that: In step S5, the output results include crack leakage risk probability, estimated crack characteristics, and maintenance recommendations.

9. A computer program storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it can implement the intelligent prediction method for basement exterior wall crack leakage risk as described in any one of claims 1-8.

10. A computer program device comprising a computer program, characterized in that When the computer program is processed and executed, it can implement the intelligent prediction method for basement exterior wall crack leakage risk as described in any one of claims 1-8.

Images