Multivariate data fusion masonry strength correction system and method

Abstract

The invention relates to the field of constructional engineering, and discloses a multivariate data fusion masonry strength correction system which comprises a data collection module, a data fusion module, a dynamic correction module, a deep learning optimization module, an abnormal data detection and repair module and an output module, and further discloses a multivariate data fusion masonry strength correction method. Comprising the following steps: data collection: collecting multivariate data from on-site in-situ testing, masonry material performance, construction records and environment monitoring, and carrying out standardization processing on the multivariate data; data fusion: fusing the multivariate data by adopting a weighted correction algorithm to generate a preliminary strength estimation value; and dynamically correcting. According to the method, through deep learning optimization and adjacent data interpolation restoration, in combination with multi-source data fusion and a dynamic feedback mechanism, more accurate and reliable masonry strength prediction is realized, and the accuracy of a prediction result and the operability of a system are improved.

Description

Technical Field

[0001] The present invention relates to the field of construction engineering, and in particular to a multivariate data fusion masonry strength correction system and method. Background Art

[0002] In construction projects, accurate assessment of masonry strength is crucial for ensuring structural safety. With the continuous advancement of building materials and construction techniques, traditional masonry strength testing methods are gradually revealing their limitations. To achieve more accurate strength predictions during construction and mitigate potential structural risks, an increasing number of projects are relying on data analysis-based technologies.

[0003] In existing technologies, masonry strength prediction is typically achieved through statistical models and experimental data. These methods can provide basic strength estimates, but often rely on static data and traditional linear analysis. Their advantages lie in their ease of use, minimal computational effort, suitability for highly standardized situations, and rapid ability to generate strength predictions. However, these methods are limited in their applicability and are unable to address the diverse data challenges inherent in complex and dynamically changing real-world construction environments.

[0004] However, existing technologies also have some obvious shortcomings in practical applications. First, traditional strength prediction methods often assume that the data is complete and accurate. However, in reality, data can easily become abnormal or missing due to external factors or sensor problems. These abnormal data cannot be effectively repaired, resulting in biased prediction results. Second, existing abnormal data repair methods mostly rely on simple interpolation algorithms, which cannot cope with the nonlinear relationships and complex multidimensional features in the data, and the accuracy of the repair results is low. Furthermore, existing technologies mainly focus on single data sources or simple models, and cannot fully consider multiple factors such as the environment and materials. The prediction results are often lacking in comprehensiveness. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a multivariate data fusion masonry strength correction system and method, which solves the problem of strength prediction deviation caused by data anomalies or missing in the existing technology and overcomes the limitations of simple interpolation repair in traditional methods.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multivariate data fusion masonry strength correction system, comprising: a data collection module for collecting multivariate data from on-site in-situ testing, masonry material properties, construction records, and environmental monitoring, and providing the multivariate data to a data fusion module; a data fusion module, receiving multivariate data from the data collection module, fusing the multivariate data using a weighted correction algorithm to generate a fused intensity estimation value, and transmitting the fused intensity estimation value to the dynamic correction module; The dynamic correction module adjusts the key parameters of the estimation method through a dynamic correction algorithm based on the fusion intensity estimation value output by the data fusion module and the intensity estimation value of the traditional estimation method, generates a corrected intensity evaluation value, and passes the corrected evaluation value to the deep learning optimization module; A deep learning optimization module receives the corrected intensity evaluation value from the dynamic correction module, further processes the multivariate data using a convolutional neural network and a long short-term memory network to extract features of the image data and time series data, and outputs an optimized intensity evaluation value based on the extracted features; The abnormal data detection and repair module is connected to the data collection module and the deep learning optimization module respectively, monitoring abnormal data in multivariate data in real time and performing interpolation and repair based on neighboring data; The output module is used to receive the optimization evaluation results of the deep learning optimization module and output the final masonry strength prediction value.

[0007] Preferably, the data fusion module includes: A weighted correction unit is used to weight and sum the field test data, material performance data, construction record data and environmental monitoring data according to a preset weight coefficient to generate a preliminary masonry strength assessment value; The dynamic correction unit is used to generate a corrected intensity evaluation value based on the output of the data fusion module and the estimated value of the inference method.

[0008] Preferably, the deep learning optimization module includes: A convolutional neural network unit is used to process crack images and extract information about crack morphology, distribution characteristics, and damage degree to assess their impact on masonry strength; Long-short-term memory network units are used to analyze time series data on ambient temperature, humidity, wind speed, and climate change, and to predict the impact of environmental factors on masonry strength based on historical data; The fusion analysis unit is used to combine the extracted features with the output of the data fusion module to optimize the final intensity evaluation value.

[0009] Preferably, the abnormal data detection and repair module includes: Data consistency detection unit, used to monitor the rationality of various input data and trigger anomaly detection when the data exceeds the preset range; The data repair unit is used to repair abnormal data based on time series interpolation and adjacent data trends after detecting abnormal data.

[0010] Preferably, the data collection module includes: The sensor unit is used to collect on-site temperature and humidity, stress state and material property data, and transmit the data to the data fusion module in real time; The database unit is used to store and manage masonry strength data of different time periods to support subsequent dynamic correction and deep learning optimization.

[0011] Preferably, the database unit includes: Historical data storage unit, used to store masonry strength data, environmental monitoring data, and construction record data for different time periods, and supports strength trend analysis over time; A data retrieval unit is used to extract data that matches the current test environment, material type, and construction conditions from the historical data storage unit based on a query request, providing a reference for subsequent data fusion and dynamic correction; The data annotation unit is used to classify and annotate the stored data based on expert experience or existing research results.

[0012] Preferably, the weighted correction unit includes: The data feature extraction unit is used to perform dimensionality reduction, normalization, and denoising on the raw data from the data collection module. The weight calculation unit is used to calculate the weights of various types of data in the masonry strength assessment based on statistical analysis, machine learning training, and expert experience, and dynamically adjust the weight coefficients to optimize the data fusion effect. The fusion calculation unit is used to perform weighted summation on the strength evaluation values of different data sources according to the weight parameters generated by the weight calculation unit, generate a preliminary masonry strength estimation value, and transmit it to the dynamic correction unit.

[0013] Preferably, the fusion analysis unit includes: A feature matching unit, which matches the image features extracted by the convolutional neural network unit with the time series features analyzed by the long short-term memory network unit to identify potential correlations; The comprehensive calculation unit is used to perform comprehensive calculations on various feature data based on the results of the feature matching unit, using multivariate regression and deep neural networks, and output optimized strength evaluation values; The evaluation and correction unit is used to perform error analysis on the evaluation results after the comprehensive calculation unit completes the calculation, and to perform secondary corrections based on the data from the historical data storage unit.

[0014] Preferably, the data repair unit includes: The anomaly recognition unit is used to identify outliers in the data and determine the degree of anomaly based on statistical analysis and machine learning. The interpolation repair unit is used to interpolate and repair abnormal data based on time series interpolation, moving average and deep learning prediction methods when abnormal data appears. A data integrity verification unit, used to verify the repair results after the data repair is completed, including comparison with the historical data storage unit; The adaptive optimization unit is used to analyze the effect of data repair and adjust the repair strategy during multiple data repair processes.

[0015] The present invention also provides a multivariate data fusion masonry strength correction method, comprising the following steps: Data collection: Collect and standardize multivariate data from in-situ field testing, masonry material properties, construction records, and environmental monitoring; Data fusion: A weighted correction algorithm is used to fuse multivariate data to generate preliminary intensity estimates; Dynamic correction: Based on the results of data fusion and the intensity estimation value of the traditional estimation method, the key parameters in the estimation method are adjusted to generate a revised intensity assessment value; Deep learning optimization: Utilizes convolutional neural networks and long short-term memory networks to analyze crack images and time series data in multivariate data and optimize masonry strength assessment results; Abnormal data detection and repair: Monitor multivariate data input for abnormalities and perform repairs based on adjacent time periods if abnormalities occur. Strength prediction output: Combined with the evaluation values optimized by deep learning, the final masonry strength prediction results are generated and visually presented.

[0016] The present invention provides a multivariate data fusion masonry strength correction system and method. It has the following beneficial effects: 1. This invention combines a deep learning optimization algorithm with neighboring data interpolation and restoration to achieve more accurate masonry strength prediction. Compared to existing prediction schemes that rely solely on traditional statistical methods, this invention effectively addresses strength estimation errors caused by abnormal or missing data, improving the accuracy and reliability of prediction results.

[0017] 2. This invention ensures the integrity and consistency of collected data throughout the entire process by monitoring and repairing abnormal data in real time. Compared to existing solutions that involve delayed data repair processes, this invention can quickly respond to and repair abnormal data, avoiding the negative impact of data issues on subsequent analysis results and ensuring the real-time performance and stability of the system.

[0018] 3. This invention combines multiple data sources with a deep learning optimization module to ultimately output a comprehensive strength prediction, providing a more comprehensive assessment result. Compared with existing assessment solutions that rely solely on a single data source or traditional algorithms, this multi-source data fusion reduces assessment bias, improves the reliability of masonry strength predictions, and addresses the problem that a single data source cannot fully reflect the actual masonry strength.

[0019] 4. Through the efficient feedback mechanism of the output module, this invention not only provides numerical strength prediction results but also generates graphical reports and risk warnings, effectively supporting project decision-making. Compared with existing solutions that only provide static numerical reports, this invention's dynamic feedback and intelligent reporting help engineers promptly identify risk points, optimize construction management, and improve the safety and accuracy of the entire construction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic diagram of the system construction of the present invention; Figure 2 This is a data fusion module framework diagram of the present invention; Figure 3 This is a framework diagram of the deep learning optimization module of the present invention; Figure 4 This is a framework diagram of the abnormal data detection and repair module of the present invention; Figure 5 This is a data collection module framework diagram of the present invention; Figure 6 Schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] Please see the attached Figure 1 -Attached Figure 5 ,An embodiment of the present invention provides a multivariate data fusion masonry strength correction system, comprising; The data collection module is used to collect multivariate data from on-site in-situ tests, masonry material properties, construction records, and environmental monitoring, and provide the multivariate data to the data fusion module; As the starting point of the system, the data collection module's main task is to obtain multi-source data for masonry strength assessment and ensure data integrity, timeliness, and accuracy. Generally speaking, the strength assessment of masonry structures requires consideration of multiple factors, including on-site in-situ test results, physical performance parameters of masonry materials, construction records, environmental monitoring data, etc. These data often come from different sources, have different distribution forms, and have certain differences in measurement methods. To ensure the effectiveness of subsequent data fusion and correction processes, the data collection module needs to adopt a reasonable data acquisition strategy, combined with sensor technology, database storage mechanisms, and data synchronization strategies to ensure data consistency and availability.

[0023] In this embodiment, the data collection module mainly includes a sensor unit and a database unit, which work together to complete data collection, storage and management, and provide the collected multi-source data to the data fusion module to ensure the continuity and integrity of the data processing chain.

[0024] Generally speaking, in-situ testing of masonry strength is the core means of obtaining structural strength information. This unit is mainly composed of mechanical sensors, ultrasonic measuring devices, infrared thermal imagers, strain monitoring equipment, environmental monitoring sensors, etc.

[0025] Alternatively, in some embodiments, the mechanical sensors may include devices such as a rebound hammer, a pressure sensor, and a displacement sensor. These devices are used to measure the compressive strength, rebound value, stress distribution, and displacement change of the masonry, and transmit the data to the data fusion module via wired or wireless means.

[0026] Specifically, the rebound hammer is used to measure the rebound value of the masonry surface, which is related to the surface strength of the masonry and is suitable for non-destructive testing; the pressure sensor can measure the stress response of the masonry under load, providing data support for strength assessment; the displacement sensor can be used to monitor the deformation of the masonry structure under different working conditions to assist in evaluating the strength change trend.

[0027] In one possible implementation, an ultrasonic measuring device is used to measure internal defects and homogeneity of masonry materials. By analyzing the propagation velocity of ultrasonic waves within the material, the density and internal damage of the masonry can be calculated. Assuming that the propagation velocity of ultrasonic waves in the material is vvv, the calculation formula is as follows: Where v is the propagation speed of ultrasound in masonry (m / s); d is the distance between sensors (m); and t is the time from ultrasound emission to reception (s).

[0028] Alternatively, an infrared thermal imager can be used to monitor the temperature distribution within the masonry and, based on changes in the temperature field, determine the extent of damage. Under certain conditions, such as extreme cold or high temperatures, the thermal expansion and contraction of masonry structures can affect their strength. This data is crucial for revising assessment results.

[0029] In some embodiments, environmental monitoring sensors are used to obtain environmental parameters such as temperature, humidity, wind speed, and air pressure in the construction area. These factors can affect the long-term durability and mechanical properties of masonry. For example, the effect of ambient humidity H on masonry strength can be expressed by the empirical formula: S = S0 × (1-kH); Among them, S is the masonry strength after humidity correction (MPa); S0 is the masonry strength under standard conditions (MPa); k is the empirical correction coefficient, which is related to the type of masonry material and construction method; H is the ambient humidity (0-1, representing 0% to 100%).

[0030] In this embodiment, the database unit is used to store, manage and provide historical data support for subsequent data fusion and dynamic correction. Generally, the unit includes submodules such as a historical data storage unit, a data retrieval unit, and a data annotation unit.

[0031] In one possible implementation, the historical data storage unit is used to store masonry strength data over different time periods and under different construction conditions, including rebound test records, compressive strength test results, environmental monitoring data, construction logs, material ratio information, etc. This data can be used to construct long-term trend analysis and provide a reference for dynamic corrections.

[0032] As an option, the data retrieval unit provides multi-dimensional data query capabilities based on material type, construction method, environmental conditions, and other factors. For example, if a specific mortar and brick ratio is currently being used at a construction site, the system can query historical data with similar material ratios to adjust the weights of various parameters during the data fusion process.

[0033] Specifically, the data retrieval unit uses an index-based query optimization algorithm to quickly match the closest data set based on conditions such as masonry material type, environmental factors, and construction records, thereby improving the efficiency of data retrieval. For example, when querying historical data, the retrieval function can be defined as: Among them, D q The optimal dataset for query matching; D i is the historical data set in the database; X ij is the value of the jth attribute in the data set; X qj is the jth attribute value of the current query data; w jis the weight coefficient of each attribute, n is the number of attributes in the data set, and argmin means minimizing the parameters of the given function.

[0034] In some embodiments, the data annotation unit is used to classify and annotate the stored data to improve data quality. For example, the data can be classified according to labels such as masonry material type, construction method, and testing method to improve data readability and usability.

[0035] As an option, the database unit also features a data synchronization and update mechanism, automatically updating the latest masonry test data and backtracking historical data in the event of anomalies to ensure data timeliness and consistency. For example, if a new batch of masonry test data is added to the construction site, the database unit can automatically detect the data source and determine whether there are any data redundancies or outliers, ensuring the reliability of the stored data.

[0036] In this invention, data collection modules and data fusion modules are transmitted via wired or wireless communication to ensure the real-time and integrity of the data. Generally, data transmission can adopt 5G communication, LoRa wireless protocol, industrial Ethernet and other methods to meet the needs of different construction environments.

[0037] In one possible implementation, data encryption and error detection mechanisms are used during data transmission to improve the security and accuracy of transmission. For example, a CRC (cyclic redundancy check) is used to detect errors during data transmission and ensure data integrity.

[0038] Specifically, the CRC check code calculation formula is as follows: C(x)=P(x)mod G(x); Where C(x) is the generated CRC check code; P(x) is the original data bit sequence to be transmitted; G(x) is the predefined generating polynomial, and mod represents the modulo operation, which is usually used for polynomial division in CRC calculation.

[0039] As an option, in some embodiments, the system can configure a data transmission priority strategy, giving priority to the transmission of critical data (such as in-situ test data and real-time monitoring data), while a batch processing mode can be used for auxiliary data (such as historical construction records) to improve transmission efficiency.

[0040] The data fusion module receives multivariate data from the data collection module, fuses the multivariate data using a weighted correction algorithm, generates a fused intensity estimation value, and passes the fused intensity estimation value to the dynamic correction module; The data fusion module is the core component of the entire system. It receives multivariate data from the data collection module and effectively fuses this data using a weighted correction algorithm to generate fused strength estimates. These strength estimates play a vital role in the subsequent dynamic correction module, ensuring dynamic correction of masonry strength and the accuracy of assessment results. The data fusion module not only integrates multivariate data from different sources but also quantifies the importance of each data point by assigning weights to improve the accuracy of the final strength assessment.

[0041] In this embodiment, the data fusion module primarily utilizes a weighted correction algorithm. This algorithm assigns a weight coefficient to each data type based on its importance and reliability, then performs a weighted summation of the results from different data sources to produce a fused strength estimate. This estimate is then passed as input to the dynamic correction module for further strength correction and optimization. This approach enables the system to provide a more accurate masonry strength assessment based on multiple dimensions of information, including real-time testing, historical data, and environmental factors.

[0042] Generally, the primary purpose of a weighted correction algorithm is to combine multiple data sources and dynamically adjust the weights assigned to each. Alternatively, this weighted correction algorithm first considers measurement results from multiple sources, including in-situ test data, masonry material performance data, construction records, and environmental monitoring data. Specifically, data from different sources are assigned different weights based on their reliability, time sensitivity, and relevance to masonry strength.

[0043] In some embodiments, assume that the data sources x1, x2, ..., x n Representing in-situ test results, masonry material performance data, construction records, and environmental monitoring data, respectively, the algorithm calculates the combined strength estimate using the following formula: Among them, S fused is the estimated intensity after fusion; x i is the intensity value of the i-th data source; w i is the weight coefficient of the data source, and n is the number of attributes in the data set.

[0044] In the embodiment of the present invention, the weight coefficient w i Dynamic adjustments are made based on the contribution, credibility, and measurement error of each data source. For example, in-situ test data may be given a higher weight because it directly reflects the actual strength of the masonry; whereas environmental monitoring data, despite its impact on strength assessment, may be given a relatively lower weight.

[0045] Specifically, the weight coefficient w iIt can be calculated as follows: in, is the reciprocal square of the accuracy of the i-th data source, that is, the reciprocal of the measurement error; The sum of the reciprocal squares of the precision of all data sources, ensuring that the sum of all weights is 1; x i is the intensity value of the i-th data source.

[0046] In one possible implementation, the precision coefficient σ i This coefficient is given by the standard deviation of historical data or the estimated error of real-time measurement, indicating the reliability of each data source under current conditions. This coefficient can be calculated based on the fitting results of experimental data or through sensor calibration.

[0047] To ensure the data fusion module's timely and responsive performance in practical applications, the algorithm in this embodiment employs an incremental calculation mechanism. Generally, the data collection module continuously collects field data, while the data fusion module dynamically updates and calculates based on real-time data. Therefore, the algorithm requires efficient computing capabilities and support for real-time updates of data source weights.

[0048] In some embodiments, to achieve incremental updates, the data fusion module recalculates the weights of each data source periodically or at set intervals to ensure that the weight coefficients accurately reflect changes and importance of the data sources over time. For example, as the amount of field test data increases, the system dynamically adjusts the weight of the in-situ test data and gradually reduces the weight of the environmental data, thereby preventing outdated data from significantly influencing the final estimate.

[0049] In the embodiment of the present invention, the fused strength estimation value S fused It is the intensity prediction result obtained after weighted correction of multivariate data. This estimate will be passed as input to the dynamic correction module for subsequent intensity correction and optimization. Specifically, the intensity estimate S fused It will be further revised according to different construction conditions and environmental changes to achieve accurate dynamic assessment of masonry strength.

[0050] Alternatively, the transmission of intensity estimates can be secured and efficient through encryption, compression, or other means. During transmission, a data structure similar to JSON may be used to encapsulate the fused results for efficient processing by subsequent modules.

[0051] In some embodiments, the dynamic correction module receives the fused strength estimate S fusedFactoring in construction progress, environmental monitoring data, and construction crew experience, the strength assessment is further refined to provide a dynamically updated masonry strength value. This value serves as real-time decision support, helping construction crews take timely countermeasures.

[0052] The dynamic correction module adjusts the key parameters of the estimation method through a dynamic correction algorithm based on the fusion intensity estimation value output by the data fusion module and the intensity estimation value of the traditional estimation method, generates a corrected intensity evaluation value, and passes the corrected evaluation value to the deep learning optimization module; The dynamic correction module combines and adjusts the fused intensity estimate output from the data fusion module with the intensity estimate from the traditional estimation method. Specifically, based on these two intensity estimates, the dynamic correction algorithm dynamically adjusts key parameters in the traditional estimation method to generate a corrected intensity estimate. This estimate is then passed as input to the deep learning optimization module, which further optimizes the system model and provides the final intensity estimate. This process enables the system to continuously optimize the accuracy and reliability of intensity predictions based on the evolving real-time and historical data.

[0053] In this embodiment, the dynamic correction module employs a dynamic adjustment mechanism based on a weighted correction algorithm. This mechanism continuously refines the parameters used in the estimation method to improve the accuracy of the estimated value. The core of this mechanism is to combine the output of the data fusion module with the results of the traditional estimation method, eliminating potential errors through dynamic adjustments. The resulting corrected strength assessment provides accurate baseline data for the subsequent deep learning optimization module, thereby enhancing the system's ability to assess masonry strength.

[0054] Traditional estimation methods typically rely on historical data and predetermined formulas for strength estimation. This method can be affected by factors such as environmental changes and construction conditions, resulting in certain deviations in the results. Alternatively, in this embodiment, the dynamic correction module combines the fused strength estimate from the data fusion module with the strength estimate from the traditional estimation method using a weighted correction algorithm. This allows the correction process to adaptively adjust key parameters in the estimation method, thereby ensuring a more accurate final strength assessment.

[0055] Specifically, the core formula of the dynamic correction algorithm is as follows: S corrected =α·S fused +(1-α)·S traditional ; Among them, S corrected is the corrected strength assessment value; S fused is the fusion strength estimation value output by the data fusion module; S traditionalis the strength estimation value of the traditional inference method; α is the weight coefficient, which indicates the relative importance of the data fusion estimation value and the traditional inference method estimation value, and its value range is [0,1].

[0056] In some embodiments, the weight coefficient α is dynamically adjusted based on the reliability and time sensitivity of the real-time data and the accuracy of the prediction. For example, when the system detects that the data fusion result is highly accurate, the value of α will increase, thereby relying more on the fusion strength estimate; when the data fusion result is uncertain, the system will increase its reliance on the estimate of the traditional inference method.

[0057] Specifically, the weight coefficient α can be dynamically adjusted according to the following formula: in, is the data fusion estimate S fused The reciprocal square of the precision; S is the estimated value by traditional estimation method traditional The reciprocal square of the precision; α is the weight coefficient.

[0058] This formula ensures that when the data fusion result is more accurate, the weight coefficient α is higher, otherwise it increases the reliance on traditional inference methods.

[0059] In a possible implementation, the corrected strength evaluation value S corrected The data is then passed to the Deep Learning Optimization module, which uses artificial intelligence algorithms to further refine the prediction model. By inputting the evaluation values from the Dynamic Correction Module, the Deep Learning Optimization module conducts in-depth analysis based on a large amount of historical and real-time data, further improving the system's prediction accuracy and stability in various construction environments.

[0060] As an option, the revised evaluation value S corrected It can also be used in real-time monitoring systems to provide on-site construction personnel with immediate feedback, helping them adjust construction strategies based on the real-time strength assessment results. For example, if the revised assessment value shows a downward trend in masonry strength, the system will automatically alert the construction personnel to take appropriate repair measures.

[0061] In general, the dynamic correction module's algorithm must ensure real-time and high efficiency. Because masonry materials and external environmental factors in the construction environment are constantly changing, the correction algorithm must be able to quickly respond to new data inputs. To this end, this embodiment introduces incremental calculation and adaptive adjustment mechanisms, allowing the system to quickly adjust correction parameters based on each data update, avoiding algorithm lag.

[0062] In some embodiments, the dynamic correction module also incorporates an error monitoring and feedback mechanism to further adjust the correction algorithm based on the changing trends in masonry strength during actual construction. For example, if significant fluctuations in the corrected strength assessment are detected, the system will automatically adjust the calculation of the weighting coefficient to ensure the stability of the assessment.

[0063] A deep learning optimization module receives the corrected intensity evaluation value from the dynamic correction module, further processes the multivariate data using a convolutional neural network and a long short-term memory network to extract features of the image data and time series data, and outputs an optimized intensity evaluation value based on the extracted features; The Deep Learning Optimization Module further improves the accuracy and reliability of masonry strength assessments by receiving revised strength assessments from the Dynamic Correction Module. This module utilizes deep learning techniques such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks to further process multivariate data. Specifically, CNNs are used to extract spatial features from image data, while LSTMs are used to analyze temporal features in time series data. Through deep learning processing of multivariate data, the Optimization Module can output more accurate strength assessments based on the extracted features.

[0064] In this embodiment, the deep learning optimization module receives inputs consisting of revised intensity estimates, which undergo multiple rounds of correction and optimization by the data fusion module and the dynamic correction module. The deep learning optimization module's core task is to utilize CNN and LSTM to extract and optimize features from the input data, ultimately generating the final intensity assessment results. This process fully leverages the multidimensional information from image and time series data, making intensity assessments more predictive and accurate.

[0065] Image data plays an important role in masonry strength assessment in general, particularly during in-situ testing and environmental monitoring. Image data acquired using infrared cameras or other imaging techniques can provide information about the internal condition of masonry structures. Convolutional neural networks can automatically identify and extract spatial features from these images, such as material defects, cracks, and heat distribution. These image features can provide important clues for strength assessment, especially in the absence of direct measurement data.

[0066] Specifically, a convolutional neural network gradually extracts features from an image through multiple convolutional layers, pooling layers, and fully connected layers. Assuming the input image is I, the convolution operation of the CNN generates a feature map F, which is calculated as follows: F = Conv(I, K); Among them, I is the input image; K is the convolution kernel, which is usually automatically learned through training; F is the feature map after convolution.

[0067] In this embodiment, a CNN automatically identifies and extracts multiple feature layers from an image, ultimately outputting a feature vector containing spatial information for further intensity assessment. The convolutional neural network training process continuously adjusts the convolution kernel K to minimize the error between the network's predicted output and the true intensity value, thereby improving assessment accuracy.

[0068] In one possible implementation, long short-term memory networks are used to process time series data, such as environmental monitoring data and construction records. This data is typically continuous and fluctuates dynamically over time. LSTM networks can effectively capture and learn this temporal dependency, enabling more accurate predictions of masonry strength.

[0069] Specifically, LSTM is a variant of recurrent neural network (RNN) that has the ability to retain long-term memories and can capture long-term dependencies in time series data. Assume that the input time series data is X = {x1, x2, ..., x t}, the LSTM network recursively updates its internal memory through the following state equation: h t =f(W h x t +b h ); Among them, h t is the hidden state of LSTM, representing the memory of the network; x t is the input data at time t; W h and b h are weights and bias terms; f is the activation function, usually tanh or sigmoid function.

[0070] Through multi-layer LSTM training, the system can extract trend characteristics of strength changes from time series data, providing a more accurate reference for strength assessment. In some embodiments, the LSTM network not only considers current data but also predicts future strength trends based on historical data, thereby predicting future construction environments or masonry changes.

[0071] In this embodiment, the deep learning optimization module combines the spatial features extracted by CNN and the temporal features extracted by LSTM, and further fuses these features to obtain the optimized strength evaluation value. Generally, the feature extraction and optimization process goes through a fully connected layer.

[0072] (fullyconnectedlayer) to complete, combining the different feature information extracted by CNN and LSTM to calculate the optimized strength evaluation value S optimized The calculation formula for this process is as follows: S optimized=FC(h cnn ,h lstm ); Among them, h cnn is the spatial feature extracted by CNN; h lstm It is the temporal feature extracted by LSTM; FC is the linear transformation operation of the fully connected layer, which can output the optimized strength evaluation value based on the feature information.

[0073] Specifically, the fully connected layer connects the feature vectors from CNN and LSTM, and maps these features to the intensity evaluation value space through linear transformation, and finally obtains the optimized intensity estimation value S optimized .

[0074] In some embodiments, the optimized strength evaluation value S optimized The final masonry strength estimate is then output by the deep learning optimization module. This output can be directly provided to on-site construction personnel and engineers as construction decision support, helping them adjust construction plans or monitor the health of the masonry based on the latest assessment results.

[0075] As an option, the optimized strength assessment value can also be fed back to other modules of the system for further data updates and feedback adjustments to improve the system's adaptability.

[0076] The abnormal data detection and repair module is connected to the data collection module and the deep learning optimization module respectively, monitoring abnormal data in multivariate data in real time and performing interpolation and repair based on neighboring data; The Abnormal Data Detection and Repair Module monitors the multi-dimensional data output by the Data Collection Module and the Deep Learning Optimization Module in real time, detecting and repairing any abnormal data. Abnormal data can be caused by a variety of factors, such as sensor failure, external interference, and environmental fluctuations. If not promptly repaired, it can seriously impact the system's strength assessment results. Therefore, this module monitors data outliers in real time and repairs them using an interpolation algorithm to ensure data continuity and accuracy, thereby improving the reliability of the final strength assessment results.

[0077] In this embodiment, the Abnormal Data Detection and Repair Module, connected to the Data Collection Module and the Deep Learning Optimization Module, performs real-time detection and repair of data from different sources. Specifically, the module repairs abnormal data by interpolating neighboring data, effectively avoiding estimation bias caused by missing or abnormal data. This repair process not only ensures data timeliness but also avoids inaccurate intensity estimates caused by missing data, thereby improving overall system performance.

[0078] Typically, the Abnormal Data Detection and Repair module receives raw multivariate data from the Data Collection Module, including in-situ test data, masonry material performance data, construction records, and environmental monitoring data. This data can often be affected by external interference or sensor errors during the collection process, resulting in anomalies or missing data. Therefore, the Abnormal Data Detection and Repair module must first monitor this data, identify outliers, and address them promptly.

[0079] Alternatively, in some embodiments, abnormal data detection uses a detection algorithm based on statistical methods, such as the mean-standard deviation method, the Z-score method, etc. These methods can identify data points that deviate from the normal range and mark them as abnormal data. For example, by calculating the Z-score of each data point: Among them, Z i is the Z-score of the i-th data point; X i is the value of the i-th data point; μ is the mean of the data; σ is the standard deviation of the data.

[0080] When | Z i When |>θ, it indicates that the data point is abnormal data, where θ is the set threshold, usually 2 or 3. When abnormal data is detected, the abnormal data detection and repair module will automatically start the repair mechanism.

[0081] Specifically, in this embodiment, anomaly data repair uses an interpolation algorithm to perform repair using adjacent data, ensuring that the repaired data has a high degree of similarity to the original data. Common interpolation methods include linear interpolation, spline interpolation, and Lagrange interpolation. These methods can infer and repair the anomaly data based on the data points before and after it.

[0082] For example, in linear interpolation, suppose there is an abnormal data point X in the time series i , and the normal data points before and after are X i-1 and X i+1 , it can be repaired by the following linear interpolation formula: in, is the data value after repair; X i-1 and X i+1 are the data values before and after the abnormal data point, respectively.

[0083] In some cases, if the data varies significantly, more complex interpolation methods, such as spline interpolation, are used to obtain smoother repair results that conform to the actual data distribution. Spline interpolation uses piecewise functions to represent the changing trend of the data, which can provide more accurate repair results than linear interpolation. Specifically, spline interpolation can be expressed using the following formula: Among them, S(x) is the interpolation function; x i is the intensity value of the i-th data source; c i ,d i are the coefficients of the spline interpolation, which are obtained by solving a system of linear equations, and n is the number of attributes in the dataset.

[0084] As an option, the abnormal data detection and repair module is also connected with the deep learning optimization module to monitor the corrected data input by the module. If the data processed by the deep learning optimization module is abnormal, the abnormal data detection and repair module will monitor and repair it in real time. For example, the strength evaluation value S output by the deep learning optimization module optimized It may be affected by abnormal data, causing the strength assessment results to deviate from the actual value.

[0085] Specifically, the Outlier Data Detection and Repair Module uses a similar interpolation algorithm to repair the time series data output by the Deep Learning Optimization Module. By interpolating outliers, the module ensures that there are no missing or inconsistent data in the feature data delivered by the Deep Learning Optimization Module, thereby ensuring the reliability of the strength assessment.

[0086] In this embodiment, the repaired data is fed back to the data collection module and the deep learning optimization module to ensure that downstream modules can continue to process complete and accurate multivariate data. In some embodiments, the repaired data may also need to undergo a verification mechanism, that is, a secondary verification by comparing it with other data sources or historical data to ensure the effectiveness and accuracy of the repair process.

[0087] The output module is used to receive the optimization evaluation results of the deep learning optimization module and output the final masonry strength prediction value.

[0088] The output module primarily receives the optimization evaluation results from the deep learning optimization module and outputs the final masonry strength prediction based on these results. Through processing by the deep learning optimization module, the system can produce more accurate masonry strength predictions based on the collected multivariate data and optimized strength evaluation information. As the final step in the system, the output module's primary function is to present the optimized evaluation results, providing reliable prediction data for engineers or decision support systems.

[0089] In this embodiment, the output module receives the optimization assessment results output by the deep learning optimization module. These results have been adjusted and optimized through deep learning algorithms and interpolation repair processes. The output module converts these results into easily understandable and actionable formats, such as numerical strength predictions, graphical reports, or warning notifications, ensuring that engineering personnel can promptly understand the health status and strength prediction of the masonry. This module typically feeds the predicted values back to the user interface through a graphical interface or data interface to support subsequent decision-making and construction operations.

[0090] Typically, the output module connects to the deep learning optimization module via an interface, ensuring real-time transmission and display of evaluation results. After multiple rounds of data processing, repair, and optimization, the deep learning optimization module outputs a final optimized strength evaluation value. This value represents the strength prediction after restoration using the deep learning algorithm and interpolation. These optimized evaluation results typically include the strength estimate, the predicted strength trend, and the predicted error range.

[0091] As an option, in some embodiments, the output of the deep learning optimization module can adopt a floating evaluation value, that is, different strength evaluation values are output according to different environmental conditions or construction status. For example, if the evaluation value is S optimized , it may include multiple levels of output, indicating the range of strength values under different assumptions. After receiving the optimized evaluation value, the output module will further process and format it to generate the final masonry strength prediction value.

[0092] Specifically, the output module further processes the optimized evaluation value from the deep learning optimization module to obtain the final masonry strength prediction value. The calculation process of the output module can be based on the optimized evaluation results and combined with factors such as construction environment, material properties, historical data, etc. to further refine the prediction value. For example, if the evaluation value output by the deep learning optimization module is S optimized , the output module adjusts according to some additional factors and outputs the final prediction value S final :S final =f(S optimized , environment, materials, historical data); Among them, S final is the predicted value of the final masonry strength; S optimized is the strength evaluation value output by the deep learning optimization module; environment is the construction environment parameters; materials is the material properties; historicaldata is the historical construction data.

[0093] The output module will calculate the final masonry strength prediction value based on the combined influence of these factors, and output it in various forms such as numerical, graphical or report types to facilitate users to conduct further analysis and decision-making.

[0094] In some embodiments, the output module not only provides numerical strength prediction results but may also generate corresponding graphical reports based on different needs. For example, the output module can generate a strength prediction trend chart, showing the changing trend of masonry strength over time or construction progress, to help engineers predict future strength development and potential risk points. Furthermore, the output module can trigger warning mechanisms based on the prediction results. For example, if the strength falls below a safety threshold, the system will automatically alert the engineer to take necessary measures.

[0095] As an option, the output module can display the final strength prediction value through the user interface. The interface can display information including numerical results, risk assessment, strength trend, etc., to help decision makers make more scientific and reasonable adjustments to the construction process.

[0096] In this embodiment, the output module's final strength predictions not only serve as a reference for engineers but can also be used by automated decision-making systems, such as intelligent construction monitoring platforms, to provide real-time feedback and adjustment suggestions. When the system predicts a masonry strength issue, the output module automatically feeds this information back to other modules, such as the dynamic correction module and the abnormal data detection and repair module, enabling appropriate repairs and adjustments, thus achieving closed-loop control.

[0097] The multivariate data fusion masonry strength correction method described below and the multivariate data fusion masonry strength correction system described above can be referenced to each other.

[0098] Please see the attached Figure 6 The present invention also provides a multivariate data fusion masonry strength correction method, comprising the following steps: S1. Data collection: Collect multivariate data from on-site in-situ testing, masonry material properties, construction records, and environmental monitoring, and standardize the multivariate data; S2, data fusion: using a weighted correction algorithm to fuse the multivariate data and generate a preliminary intensity estimate; S3. Dynamic correction: Based on the results of data fusion and the strength estimation value of the traditional estimation method, the key parameters in the estimation method are adjusted to generate a revised strength assessment value; S4. Deep Learning Optimization: Analyze crack images and time series data in multivariate data using convolutional neural networks and long short-term memory networks to optimize masonry strength assessment results. S5. Abnormal data detection and repair: Monitor multivariate data input for abnormalities. If the multivariate data is abnormal, repair it based on the adjacent time period; S6. Strength prediction output: Combined with the evaluation value after deep learning optimization, the final masonry strength prediction result is generated and a visual presentation is provided.

[0099] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0100] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The multivariate data fusion masonry strength correction system is characterized by: include: a data collection module for collecting multivariate data from on-site in-situ testing, masonry material properties, construction records, and environmental monitoring, and providing the multivariate data to a data fusion module; a data fusion module, receiving multivariate data from the data collection module, fusing the multivariate data using a weighted correction algorithm to generate a fused intensity estimation value, and transmitting the fused intensity estimation value to the dynamic correction module; The dynamic correction module adjusts the key parameters of the estimation method through a dynamic correction algorithm based on the fusion intensity estimation value output by the data fusion module and the intensity estimation value of the traditional estimation method, generates a corrected intensity evaluation value, and passes the corrected evaluation value to the deep learning optimization module; A deep learning optimization module receives the corrected intensity evaluation value from the dynamic correction module, further processes the multivariate data using a convolutional neural network and a long short-term memory network to extract features of the image data and time series data, and outputs an optimized intensity evaluation value based on the extracted features; The abnormal data detection and repair module is connected to the data collection module and the deep learning optimization module respectively, monitoring abnormal data in multivariate data in real time and performing interpolation and repair based on neighboring data; The output module is used to receive the optimization evaluation results of the deep learning optimization module and output the final masonry strength prediction value.

2. The multivariate data fusion masonry strength correction system according to claim 1 is characterized in that: The data fusion module includes: A weighted correction unit is used to weight and sum the field test data, material performance data, construction record data and environmental monitoring data according to a preset weight coefficient to generate a preliminary masonry strength assessment value; The dynamic correction unit is used to generate a corrected intensity evaluation value based on the output of the data fusion module and the estimated value of the inference method.

3. The multivariate data fusion masonry strength correction system according to claim 1, characterized in that: The deep learning optimization module includes: A convolutional neural network unit is used to process crack images and extract information about crack morphology, distribution characteristics, and damage degree to assess their impact on masonry strength; Long-short-term memory network units are used to analyze time series data on ambient temperature, humidity, wind speed, and climate change, and to predict the impact of environmental factors on masonry strength based on historical data; The fusion analysis unit is used to combine the extracted features with the output of the data fusion module to optimize the final intensity evaluation value.

4. The multivariate data fusion masonry strength correction system according to claim 1, characterized in that: The abnormal data detection and repair module includes: Data consistency detection unit, used to monitor the rationality of various input data and trigger anomaly detection when the data exceeds the preset range; The data repair unit is used to repair abnormal data based on time series interpolation and adjacent data trends after detecting abnormal data.

5. The multivariate data fusion masonry strength correction system according to claim 1 is characterized in that: The data collection module includes: The sensor unit is used to collect on-site temperature and humidity, stress state and material property data, and transmit the data to the data fusion module in real time; The database unit is used to store and manage masonry strength data of different time periods to support subsequent dynamic correction and deep learning optimization.

6. The multivariate data fusion masonry strength correction system according to claim 5, characterized in that: The database unit includes: Historical data storage unit, used to store masonry strength data, environmental monitoring data, and construction record data for different time periods, and supports strength trend analysis over time; A data retrieval unit is used to extract data that matches the current test environment, material type, and construction conditions from the historical data storage unit based on a query request, providing a reference for subsequent data fusion and dynamic correction; The data annotation unit is used to classify and annotate the stored data based on expert experience or existing research results.

7. The multivariate data fusion masonry strength correction system according to claim 2, characterized in that: The weighted correction unit includes: The data feature extraction unit is used to perform dimensionality reduction, normalization and denoising on the raw data from the data collection module; The weight calculation unit is used to calculate the weight of various data in the masonry strength assessment based on statistical analysis, machine learning training and expert experience, and dynamically adjust the weight coefficient to optimize the data fusion effect; The fusion calculation unit is used to perform weighted summation on the strength evaluation values of different data sources according to the weight parameters generated by the weight calculation unit, generate a preliminary masonry strength estimation value, and transmit it to the dynamic correction unit.

8. The multivariate data fusion masonry strength correction system according to claim 3 is characterized in that: The fusion analysis unit includes: A feature matching unit, which matches the image features extracted by the convolutional neural network unit with the time series features analyzed by the long short-term memory network unit to identify potential correlations; The comprehensive calculation unit is used to perform comprehensive calculations on various feature data based on the results of the feature matching unit using multivariate regression and deep neural networks, and output optimized strength evaluation values; The evaluation and correction unit is used to perform error analysis on the evaluation results after the comprehensive calculation unit completes the calculation, and to perform secondary corrections based on the data from the historical data storage unit.

9. The multivariate data fusion masonry strength correction system according to claim 4, characterized in that: The data repair unit includes: Anomaly recognition unit, used to identify outliers in data and determine the degree of anomaly based on statistical analysis and machine learning; The interpolation repair unit is used to perform interpolation repair on abnormal data based on time series interpolation method, moving average method and deep learning prediction method when abnormal data appears; A data integrity verification unit, used to verify the repair results after the data repair is completed, including comparison with the historical data storage unit; The adaptive optimization unit is used to analyze the effect of data repair and adjust the repair strategy during multiple data repair processes.

10. A method for correcting masonry strength using multivariate data fusion, a system for correcting masonry strength using multivariate data fusion according to any one of claims 1 to 9, characterized in that: The following steps are involved: Data collection: Collect and standardize multivariate data from in-situ field testing, masonry material properties, construction records, and environmental monitoring; Data fusion: A weighted correction algorithm is used to fuse multivariate data to generate preliminary intensity estimates; Dynamic correction: Based on the results of data fusion and the intensity estimation value of the traditional estimation method, the key parameters in the estimation method are adjusted to generate a revised intensity assessment value; Deep learning optimization: Utilizes convolutional neural networks and long short-term memory networks to analyze crack images and time series data in multivariate data and optimize masonry strength assessment results; Abnormal data detection and repair: Monitor multivariate data input for abnormalities and perform repairs based on adjacent time periods if abnormalities occur. Strength prediction output: Combined with the evaluation values optimized by deep learning, the final masonry strength prediction results are generated and visually presented.

Images