Construction engineering quality evaluation method and system

Abstract

The invention relates to the technical field of constructional engineering quality evaluation, in particular to a constructional engineering quality evaluation method and system, and the method comprises the following steps: building an auto-encoder model based on original constructional engineering data, adding an input layer, an encoding layer, a variation layer and a decoding layer, and obtaining a self-encoder model; primary features are captured through an auto-encoder model, Gaussian noise of input data is simulated, basic quality feature codes are generated, feature learning is carried out through the basic quality feature codes, and by adding an attention mechanism model and aiming at key quality features, recognition of key indexes is enhanced, and enhanced quality feature representation is obtained. According to the method, the automation level and the intelligence level in the evaluation process are remarkably improved through the self-learning ability of the self-encoder model in the process of processing complex data and the effectiveness of the attention mechanism in the aspect of strengthening key feature recognition. The introduction of the mutation algorithm enables the model to consider the influence of environmental factors on the building quality, and the application of the clustering algorithm achieves the precise division of risk levels.

Description

Technical Field

[0001] The present invention relates to the technical field of construction engineering quality assessment, and in particular to a construction engineering quality assessment method and system. Background Art

[0002] Inspired by the structure of the human brain, neural networks process data by connecting a large number of neurons to form a network, and can learn patterns and features from complex, unstructured data. This technology is widely used in image recognition, speech recognition, natural language processing and other fields. In recent years, it has also been increasingly used in the field of construction engineering, especially in quality assessment, structural health monitoring and safety evaluation.

[0003] Among them, the purpose of the construction project quality assessment method is to accurately assess the quality level of construction projects through automated means to ensure the safety, reliability and durability of the building. This method is particularly aimed at common problems in construction projects, such as cracks, structural defects, etc. Through early identification and assessment of these problems, potential safety accidents can be avoided and the service life of the building can be extended. Its purpose is to improve the accuracy and efficiency of the assessment through scientific methods, reduce the impact of human factors, and thus improve the overall quality of construction projects.

[0004] Traditional assessment methods are often inefficient when processing complex, unstructured data, and have difficulty capturing small but critical quality defects. In particular, in terms of the assessment of the impact of environmental factors and the classification of risk levels, there is a lack of effective automated tools, making it difficult to achieve refined management. In this case, the assessment results often lack accuracy and timeliness, resulting in the identification and response to potential safety hazards not being timely enough, increasing the risk of safety accidents, and also increasing the cost of subsequent maintenance and repair. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a construction engineering quality assessment method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a construction project quality assessment method, comprising the following steps:

[0007] S1: Based on the original construction project data, an autoencoder model is constructed, and input layer, encoding layer, mutation layer and decoding layer are added. The autoencoder model is used to capture primary features and simulate the Gaussian noise of the input data to generate basic quality feature encoding;

[0008] S2: Using the basic quality feature encoding, feature learning is performed, and by adding an attention mechanism model, the key quality features are targeted, the recognition of key indicators is strengthened, and enhanced quality feature representation is obtained;

[0009] S3: applying a mutation algorithm to perform feature diffusion on the enhanced quality feature representation to simulate performance changes under the influence of environmental factors, and obtaining simulated evolution features by multiplying a random mutation matrix with the feature representation;

[0010] S4: Based on the simulated evolution characteristics, a clustering algorithm is executed to perform risk classification, cluster centers are assigned to data points based on distance, differentiated quality risk levels are identified, and quality risk classification and improvement measures are generated;

[0011] S5: Using the quality risk classification and improvement measures, adopt a dynamic adaptive learning strategy, process real-time monitoring data, dynamically adjust learning parameters, perform special optimization on the detected risk points, and obtain a quality control strategy after special optimization;

[0012] S6: Apply the specially optimized quality control strategy to the quality management of construction projects, refine and adjust the model parameters through the feedback loop of simulation and actual application, and continuously optimize the model to obtain the building quality assessment model.

[0013] As a further solution of the present invention, the basic quality feature coding includes dimensional deviation coding, material strength coding and processing accuracy coding, the enhanced quality feature representation includes weighted representation of key structure bearing capacity, weighted representation of safety factor and weighted representation of construction quality, the simulation evolution characteristics include environmental corrosion impact coding, service life impact coding and maintenance frequency impact coding, the quality risk grading and improvement measures include high-risk level identification and reinforcement plan, medium-risk level monitoring plan and maintenance plan and low-risk level routine inspection plan, the quality control strategy after special optimization includes learning parameters adjusted by real-time monitoring data feedback, targeted defect repair technology selection and improved quality prevention measures, and the building quality assessment model includes refined model parameters, quality control strategy updates, and optimized risk assessment logic.

[0014] As a further solution of the present invention, based on the original construction project data, an autoencoder model is constructed, an input layer, an encoding layer, a mutation layer, and a decoding layer are added, and the steps of capturing primary features and simulating Gaussian noise of the input data through the autoencoder model to generate basic quality feature encoding are specifically as follows:

[0015] S101: Based on the original construction engineering data, build the autoencoder network framework, initialize the model and set the number of input layer matching data features to generate the initialization network framework;

[0016] S102: Based on the initialized network framework, a coding layer is added to compress the data representation, reduce the dimension, and use an activation function to capture nonlinear features to obtain a compressed data representation;

[0017] S103: Based on the compressed data representation, a variation layer is introduced to simulate Gaussian noise, random noise is added through a custom function, generalization capability is optimized, and noise optimized coding is obtained;

[0018] S104: Based on the noise optimization coding, a decoding layer is added to reconstruct data, and network parameters are adjusted to match the original data range to obtain basic quality feature coding.

[0019] As a further solution of the present invention, the basic quality feature encoding is used to perform feature learning, and by adding an attention mechanism model, the key quality features are targeted to strengthen the recognition of key indicators, and the steps of obtaining enhanced quality feature representation are specifically as follows:

[0020] S201: Based on the basic quality feature coding, perform data analysis and feature extraction, identify and distinguish key and non-key quality features in the data, and generate a basic quality feature set;

[0021] S202: using the basic quality feature set, assigning weights and highlighting features, enhancing the key quality features, improving the influence of the key quality features in the model, and generating a weighted set of key quality features;

[0022] S203: Optimizing the feature representation for the weighted set of key quality features, by strengthening the identification of key indicators, iteratively refining the expression of data features, improving the model's ability to identify key features of data quality, and generating enhanced quality feature representation.

[0023] As a further solution of the present invention, the enhanced quality feature representation is subjected to feature diffusion by applying a mutation algorithm to simulate performance changes under the influence of environmental factors, and the steps of obtaining simulated evolution features by multiplying a random mutation matrix with the feature representation are specifically as follows:

[0024] S301: Based on the enhanced quality feature representation, a random variation matrix is ​​generated to simulate the influence of environmental factors, and the matrix and the feature representation are adjusted to the same dimension to obtain a random variation matrix;

[0025] S302: performing standardization processing on the random variation matrix to constrain the range of variation effects and avoid excessive variation or reduction of features, thereby obtaining a standardized variation matrix;

[0026] S303: Using the standardized variation matrix, simulating the performance change of the feature under the influence of environmental factors through matrix multiplication to generate a multiplication variation feature;

[0027] S304: analyzing and processing the multiplication variation features, using the PCA algorithm to identify key variation features, enhancing the quality and availability of evolution features, and obtaining simulated evolution features;

[0028] The PCA algorithm uses the formula

[0029]

[0030] Among them, Σ adjusted is the improved covariance matrix, X is the standardized data matrix, is the average value of the data, n is the number of samples, δ is the smoothing parameter to avoid division by zero errors, S is the slope matrix of the samples, α is the weight coefficient of the slope, J is the weight matrix of the kurtosis, β is the weight coefficient of the kurtosis, K is the custom adjustment matrix, and γ is the weight coefficient of the adjustment matrix.

[0031] As a further solution of the present invention, based on the simulated evolution characteristics, a clustering algorithm is executed to perform risk classification, cluster centers are assigned to data points based on distance, differentiated quality risk levels are identified, and the steps of generating quality risk classification and improvement measures are specifically as follows:

[0032] S401: Based on the simulated evolution characteristics, perform cluster analysis, divide the data points according to the preset risk level number, optimize the cluster center position through iteration, until the convergence condition or the maximum number of iterations is met, and generate a risk indicator set;

[0033] S402: executing a decision tree classification method on the risk indicator set, classifying the data points into corresponding risk levels according to the risk indicators, until a stop condition is met, thereby obtaining a risk level classification;

[0034] S403: Utilizing the risk level classification, performing priority sorting, sorting in descending order according to the risk index, identifying the priority of improvement measures, and customizing quality risk classification and improvement measures;

[0035] The decision tree classification method uses the formula

[0036]

[0037] Where D represents the dataset, m is the number of categories, and p i is the relative frequency of the ith category in the dataset D, Depth-β represents the depth of the current node, α and β are parameters used to adjust the sensitivity of the depth adjustment factor, γ is a parameter used to weight the variability of the dataset, Var(D) is the variability indicator of the dataset, and e is the base of the natural logarithm.

[0038] As a further solution of the present invention, the steps of using the quality risk classification and improvement measures, taking a dynamic adaptive learning strategy, processing real-time monitoring data, dynamically adjusting learning parameters, and performing special optimization for the detected risk points to obtain the quality control strategy after special optimization are as follows:

[0039] S501: Using the quality risk classification and improvement measures, dynamically and adaptively process the real-time monitoring data, adjust the learning rate response risk by cyclically detecting data changes, and generate a preliminary risk response strategy;

[0040] S502: Based on the preliminary risk response strategy, using a decision tree model, identifying and optimizing the currently detected risk points, recalculating learning parameters, and obtaining optimized learning parameters;

[0041] S503: Based on the optimized learning parameters, the support vector machine model is used to adjust the control strategy for all monitoring data, verify the effectiveness of the parameter settings, and generate a special optimized quality control strategy.

[0042] As a further solution of the present invention, the quality control strategy after the special optimization is applied to the quality management of construction projects. Through the feedback loop of simulation and actual application, the model parameters are refined and adjusted, and the model is continuously optimized to obtain the steps of the building quality assessment model:

[0043] S601: Based on the specially optimized quality control strategy, the quality management of the construction project is applied to analyze the preliminary quality data, screen the key quality control points, mark the areas below the threshold as priority optimization areas, and generate a key quality control point analysis table;

[0044] S602: Based on the critical quality control point analysis table, simulate the construction method, material selection, and construction sequence, improve the quality through iterative cycles, adjust the construction parameters, and generate a simulation optimization parameter table;

[0045] S603: Based on the simulation optimization parameter table, feedback optimization is performed on the parameters used in the actual construction project, and the model parameters are fine-tuned until the quality assessment model is stable, thereby obtaining a building quality assessment model.

[0046] A construction project quality assessment system, the construction project quality assessment system is used to implement the construction project quality assessment method, the system comprises:

[0047] The data preprocessing module deletes missing values, replaces outliers, performs standardization, and splits data sets based on the original construction engineering data to generate preprocessed construction engineering data;

[0048] The feature encoding module constructs an autoencoder model based on the preprocessed construction engineering data, sets the input layer dimension, adds the encoding layer, the mutation layer, the decoding layer, compiles the model, and generates the basic quality feature encoding;

[0049] The feature enhancement module identifies key features based on the basic quality feature coding, assigns weights, highlights features, reorganizes data features, and generates enhanced quality feature representation;

[0050] The simulated evolution module generates a random variation matrix based on the enhanced quality feature representation, simulates the influence of environmental factors, performs standardization processing and matrix multiplication, identifies key variation features, and generates simulated evolution features;

[0051] The risk grading module performs cluster analysis, decision tree classification method, data point division, risk level priority ranking based on the simulated evolution characteristics, and generates quality risk grading and improvement measures;

[0052] The dynamic learning optimization module uses the quality risk classification and improvement measures to dynamically and adaptively process the real-time monitoring data, generate preliminary risk response strategies, adjust the control strategies of the monitoring data, and generate specially optimized quality control strategies;

[0053] The quality control strategy module conducts simulation experiments on construction project quality management, improves construction methods, material selection, construction sequence, fine-tunes model parameters, and generates a building quality assessment model based on the quality control strategy optimized by the special project.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] In this invention, the self-learning ability of the autoencoder model in processing complex data and the effectiveness of the attention mechanism in strengthening the recognition of key features significantly improve the automation and intelligence level of the assessment process. The introduction of the mutation algorithm enables the model to consider the impact of environmental factors on building quality, while the application of the clustering algorithm achieves accurate classification of risk levels. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0057] Figure 2 This is a detailed flow chart of S1 of the present invention;

[0058] Figure 3 This is a detailed flow chart of S2 of the present invention;

[0059] Figure 4 This is a detailed flow chart of S3 of the present invention;

[0060] Figure 5 This is a detailed flow chart of S4 of the present invention;

[0061] Figure 6 This is a detailed flow chart of S5 of the present invention;

[0062] Figure 7 This is a detailed flow chart of S6 of the present invention;

[0063] Figure 8 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0065] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0066] Embodiment 1

[0067] See also Figure 1 The present invention provides a technical solution: a construction engineering quality assessment method, comprising the following steps:

[0068] S1: Based on the original construction project data, an autoencoder model is constructed, and input layer, encoding layer, mutation layer and decoding layer are added. The autoencoder model is used to capture primary features and simulate the Gaussian noise of the input data to generate basic quality feature encoding;

[0069] S2: Use basic quality feature encoding to perform feature learning. By adding an attention mechanism model, we can strengthen the recognition of key indicators for key quality features and obtain enhanced quality feature representation.

[0070] S3: For the enhanced quality feature representation, the mutation algorithm is applied to perform feature diffusion to simulate the performance changes under the influence of environmental factors. The simulated evolution feature is obtained by multiplying the random mutation matrix with the feature representation;

[0071] S4: Based on the simulated evolution characteristics, the clustering algorithm is executed to perform risk classification, the cluster centers are assigned to the data points based on the distance, the differentiated quality risk levels are identified, and the quality risk classification and improvement measures are generated;

[0072] S5: Using quality risk classification and improvement measures, adopt dynamic adaptive learning strategies, process real-time monitoring data, dynamically adjust learning parameters, perform special optimization for detected risk points, and obtain quality control strategies after special optimization;

[0073] S6: Apply the specially optimized quality control strategy to the quality management of construction projects. Through the feedback loop between simulation and actual application, refine and adjust the model parameters, continuously optimize the model, and obtain the building quality assessment model.

[0074] The basic quality feature coding includes dimensional deviation coding, material strength coding and processing accuracy coding. The enhanced quality feature representation includes weighted representation of key structure bearing capacity, weighted representation of safety factor and weighted representation of construction quality. The simulation evolution characteristics include environmental corrosion impact coding, service life impact coding and maintenance frequency impact coding. The quality risk classification and improvement measures include high-risk level identification and reinforcement plan, medium-risk level monitoring plan and maintenance plan, and low-risk level routine inspection plan. The quality control strategy after special optimization includes learning parameters adjusted by real-time monitoring data feedback, targeted defect repair technology selection and improved quality prevention measures. The building quality assessment model includes refined model parameters, quality control strategy updates, and optimized risk assessment logic.

[0075] See also Figure 2 Based on the original construction engineering data, an autoencoder model is constructed, and the input layer, encoding layer, mutation layer, and decoding layer are added. The autoencoder model is used to capture primary features and simulate the Gaussian noise of the input data. The specific steps for generating basic quality feature encoding are as follows:

[0076] S101: Based on the original construction engineering data, build the autoencoder network framework, initialize the model and set the number of input layer matching data features to generate the initialization network framework;

[0077] Based on the original construction project data, the autoencoder network framework is built using the TensorFlow and Keras libraries. The model is initialized and the number of input layer nodes is set to match the number of data features. The input layer is added through the Sequential model, and the input_shape parameter is set equal to the number of original data features to ensure that the network framework can receive all data features and generate an initialized network framework.

[0078] S102: Based on the initialized network framework, a coding layer is added to compress the data representation, reduce the dimension, and use the activation function to capture nonlinear features to obtain the compressed data representation;

[0079] Based on the initialization network framework, an encoding layer is added to compress the data representation and reduce the dimension. The ReLU activation function is used to capture nonlinear features. A Dense layer is added through the Sequential model. The units parameter is set to be smaller than the number of nodes in the input layer to achieve data compression. The activation parameter is set to 'relu' to capture nonlinear relationships and obtain compressed data representation.

[0080] S103: Based on the compressed data representation, a mutation layer is introduced to simulate Gaussian noise, random noise is added through a custom function, generalization ability is optimized, and noise optimized coding is obtained;

[0081] Based on the compressed data representation, a mutation layer is introduced to simulate Gaussian noise. Random noise is added through the Lambda layer and the tf.random.normal function in TensorFlow. The stddev parameter is set to control the noise intensity. In this way, noise is introduced during the encoding process to optimize the generalization ability of the model and obtain noise-optimized encoding.

[0082] S104: Based on the noise optimization coding, a decoding layer is added to reconstruct the data, the network parameters are adjusted to match the original data range, and the basic quality feature coding is obtained;

[0083] Based on the noise optimization encoding, add a decoding layer to reconstruct the data, adjust the network parameters to match the original data range, add a Dense layer as the decoding layer through the Sequential model, set the units parameter equal to the number of original data features, and select 'sigmoid' or other suitable functions for the activation parameter to ensure that the reconstructed data matches the original data range, and obtain the basic quality feature encoding.

[0084] See also Figure 3 , using basic quality feature encoding, feature learning, by adding an attention mechanism model, for key quality features, strengthen the identification of key indicators, and obtain enhanced quality feature representation steps are as follows:

[0085] S201: Based on the basic quality feature coding, perform data analysis and feature extraction, identify and distinguish key and non-key quality features in the data, and generate a basic quality feature set;

[0086] Based on the basic quality feature encoding, the decision tree algorithm is adopted. The DecisionTreeClassifier function is called using Python's Scikit-learn library. The criterion parameter is set to 'gini' to measure the split quality. The max_depth parameter is set to 5 to limit the maximum depth of the tree. Data analysis and feature extraction are performed. The fit method is used to train the data. The feature_importances_ attribute is used to extract the importance score of each feature. According to the score, the key and non-key quality features in the data are identified and distinguished, and the basic quality feature set is generated.

[0087] S202: Using the basic quality feature set, assigning weights and highlighting features, enhancing the key quality features, improving the influence of the key quality features in the model, and generating a weighted set of key quality features;

[0088] Using the basic quality feature set and the weighted average method, the average function is called using Python's NumPy library. The weights parameter is set to assign values ​​according to the importance scores of the key quality features, the key quality features are enhanced, the influence of the key quality features in the model is improved, and a weighted set of key quality features is generated.

[0089] S203: Optimize the feature representation for the weighted set of key quality features, iteratively refine the expression of data features by strengthening the identification of key indicators, improve the model's ability to identify key features of data quality, and generate enhanced quality feature representation;

[0090] For the weighted set of key quality features, the support vector machine (SVM) algorithm is used. The SVC function is called using Python's Scikit-learn library. The kernel parameter is set to 'linear' to select the linear kernel function, and the C parameter is set to 1.0 to define the penalty coefficient for the error term. The feature representation is optimized, and the identification of key indicators is strengthened by calling the fit method and applying the decision_function method. The expression of data features is iteratively refined to improve the model's ability to identify key features of data quality and generate enhanced quality feature representation.

[0091] See also Figure 4 , for the enhanced quality feature representation, the mutation algorithm is applied to perform feature diffusion to simulate the performance changes under the influence of environmental factors. The steps of obtaining the simulated evolution feature by multiplying the random mutation matrix with the feature representation are as follows:

[0092] S301: Based on the enhanced quality feature representation, a random variation matrix is ​​generated to simulate the influence of environmental factors, and the matrix and the feature representation are adjusted to the same dimension to obtain a random variation matrix;

[0093] Based on the enhanced quality feature representation, the NumPy library is used to generate random numbers as elements, and a random mutation matrix is ​​created to simulate the influence of environmental factors. The reshape function is used to adjust the matrix and the enhanced quality feature representation to the same dimension to obtain the random mutation matrix. In this process, the numpy.random.rand function is used to generate element values. The dimension parameter is determined by the dimension of the enhanced quality feature representation to ensure the consistency of the random mutation matrix and the feature representation, and generate a random mutation matrix.

[0094] S302: Performing standardization processing on the random variation matrix to constrain the range of variation effects and avoid excessive variation or reduction of features, thereby obtaining a standardized variation matrix;

[0095] The random variation matrix is ​​standardized, and the StandardScaler function in the Sklearn.preprocessing library is used to constrain the range of variation effects to avoid excessive variation or reduction of features. The fit_transform method is applied to the random variation matrix to ensure that the mean value of each feature is 0 and the variance is 1, generating a standardized variation matrix.

[0096] S303: using the standardized variation matrix, simulating the performance change of the feature under the influence of environmental factors through matrix multiplication, and generating a multiplication variation feature;

[0097] The standardized variation matrix is ​​used to perform matrix multiplication using the NumPy library. The numpy.dot function is used to simulate the performance changes of features under the influence of environmental factors. This function multiplies the standardized variation matrix with the original feature matrix to simulate the impact of environmental factors on feature performance and generate multiplication variation features.

[0098] S304: Analyze and process multiplicative variation features, use PCA algorithm to identify key variation features, enhance the quality and availability of evolution features, and obtain simulated evolution features;

[0099] Analyze and process multiplicative variation features, use the PCA algorithm, identify key variation features through the Sklearn.decomposition.PCA library, and enhance the quality and availability of evolution features. The PCA method is used for dimensionality reduction. The number of principal components is determined by setting the n_components parameter, retaining the most important variation features, and obtaining simulated evolution features.

[0100] The PCA algorithm uses the formula

[0101]

[0102] Among them, Σ adjusted is the improved covariance matrix, X is the standardized data matrix, is the average value of the data, n is the number of samples, δ is the smoothing parameter to avoid division by zero errors, S is the slope matrix of the samples, α is the weight coefficient of the slope, J is the weight matrix of the kurtosis, β is the weight coefficient of the kurtosis, K is the custom adjustment matrix, and γ is the weight coefficient of the adjustment matrix.

[0103] First, the original data is standardized to ensure the fairness of each feature in the analysis. Then the improved covariance matrix is ​​calculated, which is not only based on the original data, but also takes into account the slope and kurtosis of the data and the custom adjustment matrix. These are weighted by the corresponding weight coefficients to increase the sensitivity and adaptability of the model to the details of the data. The smoothing parameter δ is introduced to increase the stability of the calculation and avoid mathematical errors when the number of samples is small. The determination of the weight coefficient can be based on cross-validation to maximize the prediction accuracy of the model for unseen data. Finally, the improved covariance matrix is ​​eigendecomposed to determine the principal components, and the dimensionality reduction is performed according to the set number of principal components n. The introduction of the slope matrix S and the kurtosis matrix J, as well as the addition of the custom adjustment matrix K, is to allow the PCA algorithm to not only consider the linear correlation between features, but also capture more complex patterns of data distribution, such as skewness and peaks, thereby improving the quality and availability of evolutionary features.

[0104] See also Figure 5 ,Based on the simulated evolution characteristics, the clustering algorithm is executed to perform risk classification, the cluster centers are assigned to the data points based on the distance, the differentiated quality risk levels are identified, and the steps of generating quality risk classification and improvement measures are as follows:

[0105] S401: Based on the simulated evolution characteristics, cluster analysis is performed, data points are divided according to the number of preset risk levels, and the location of cluster centers is optimized through iteration until the convergence condition or the maximum number of iterations is met, thereby generating a risk indicator set;

[0106] Based on the simulated evolutionary characteristics, the K-means clustering algorithm is used to perform cluster analysis. The KMeans class in the Scikit-learn library is used to divide the data points according to the preset number of risk levels. The algorithm is applied through the fit method and the n_clusters parameter is set equal to the preset number of risk levels. The init parameter selects 'k-means++' to optimize the selection of the initial cluster centers. The max_iter parameter is set to 300 to define the maximum number of iterations. The iterative optimization of the cluster center positions is performed until the convergence condition is met or the maximum number of iterations is reached, and the risk indicator set is generated.

[0107] S402: executing a decision tree classification method on the risk indicator set, classifying the data points into corresponding risk levels according to the risk indicators, until a stop condition is met, thereby obtaining a risk level classification;

[0108] For the risk indicator set, the decision tree classification method is performed. The DecisionTreeClassifier class in the Scikit-learn library is used to divide the data points into corresponding risk levels according to the risk indicators. The decision tree model is applied through the fit method. The criterion parameter is set to 'gini' to use the Gini coefficient as the splitting criterion. The max_depth parameter is set to a preset value to limit the maximum depth of the tree to prevent overfitting until the stopping condition is met, such as reaching the tree depth defined by max_depth, and the risk level division is obtained;

[0109] The decision tree classification method uses the formula

[0110]

[0111] Where D represents the dataset, m is the number of categories, and p i is the relative frequency of the ith category in the dataset D, Depth-β represents the depth of the current node, α and β are parameters used to adjust the sensitivity of the depth adjustment factor, γ is a parameter used to weight the variability of the dataset, Var(D) is the variability indicator of the dataset, and e is the base of the natural logarithm.

[0112] First, the Gini coefficient is calculated according to the original formula, and then the depth adjustment factor is introduced. The factor is adjusted by the depth Depth-β and the parameters α and β. This factor is used to reduce the influence of deep nodes on the Gini coefficient to avoid overfitting. Then the data set variability weight is introduced. The weight is adjusted by the parameter γ and the variability of the data set Var(D). This weight is used to increase the influence of the internal variability of the data set on the Gini coefficient to improve the model's sensitivity to differences within the data set. Finally, these factors are multiplied to obtain the improved Gini coefficient.

[0113] S403: Use risk level classification to perform priority sorting, sort in descending order according to risk index, identify the priority of improvement measures, and customize quality risk classification and improvement measures;

[0114] Utilize risk level classification and perform priority sorting. Use the sorted function in the Python standard library to sort in descending order according to the risk index. Set the key parameter of the sorted function to the risk index field and the reverse parameter to True to achieve descending order, identify the priority of improvement measures, and customize quality risk grading and improvement measures.

[0115] See also Figure 6 , using quality risk classification and improvement measures, adopting dynamic adaptive learning strategies, processing real-time monitoring data, dynamically adjusting learning parameters, and performing special optimization for detected risk points. The specific steps for obtaining the specially optimized quality control strategy are as follows:

[0116] S501: Use quality risk classification and improvement measures to dynamically and adaptively process real-time monitoring data, adjust the learning rate to respond to risks through cyclic detection of data changes, and generate preliminary risk response strategies;

[0117] By utilizing quality risk grading and improvement measures, a dynamic learning rate adjustment method is adopted to dynamically and adaptively process real-time monitoring data. The tf.train.ExponentialDecay function of TensorFlow is used to loop and detect data changes, and the learning rate is adjusted to respond to risks. The initial learning rate initial_learning_rate parameter is set, and the decay rate decay_rate parameter is dynamically adjusted according to the risk level. The decay_steps parameter is set to update the learning rate once a certain number of data points are detected to generate a preliminary risk response strategy.

[0118] S502: Based on the preliminary risk response strategy, use the decision tree model to identify and optimize the currently detected risk points, recalculate the learning parameters, and obtain the optimized learning parameters;

[0119] Based on the preliminary risk response strategy, a decision tree model is used, and the DecisionTreeClassifier class of Scikit-learn is adopted to identify and optimize the currently detected risk points. The learning parameters are recalculated through the fit method, and the criterion parameter is set to 'entropy' to use information gain as the splitting criterion. The max_depth parameter is dynamically adjusted to control the model complexity and prevent overfitting, and the optimized learning parameters are obtained.

[0120] S503: Based on the optimized learning parameters, the support vector machine model is used to adjust the control strategy for all monitoring data, verify the effectiveness of the parameter settings, and generate a special optimized quality control strategy;

[0121] Based on the optimized learning parameters, the support vector machine model is applied, and the SVC class of Scikit-learn is used to adjust the control strategy for all monitoring data. The model is applied through the fit method and the C parameter is set. The penalty coefficient of the error term is adjusted using the optimized learning parameters, and the kernel parameter selects 'linear' to adapt to the data characteristics. The effectiveness of the parameter setting is verified, and the quality control strategy after special optimization is generated.

[0122] See also Figure 7 , the quality control strategy optimized by the special project is applied to the quality management of construction projects. Through the feedback loop between simulation and actual application, the model parameters are refined and adjusted, and the model is continuously optimized. The specific steps of the building quality assessment model are as follows:

[0123] S601: Apply the quality control strategy after special optimization to the quality management of construction projects, analyze the preliminary quality data, screen the key quality control points, mark the areas below the threshold as priority optimization areas, and generate the key quality control point analysis table;

[0124] The quality control strategy after special optimization is applied to the quality management of construction projects. Data analysis tools are used to analyze the preliminary quality data. Python's Pandas library is used to screen key quality control points. Conditional screening is performed through the DataFrame object. Areas below the threshold are marked as priority optimization areas. Threshold parameters are set according to actual quality management standards. The screening conditions are applied through the query method to generate a key quality control point analysis table.

[0125] S602: Based on the key quality control point analysis table, simulate the construction method, material selection, and construction sequence, improve the quality through iterative cycles, adjust the construction parameters, and generate a simulation optimization parameter table;

[0126] Based on the key quality control point analysis table, simulation experiments were carried out on the construction method, material selection, and construction sequence. Simulation software was used to perform iterative cycles to improve quality. The construction process was simulated using Python's SimPy library. By defining the simulation environment and parameters, the construction method, material selection, and construction sequence in the simulation model were set as variables. The for loop was used to iterate the parameters, and the construction parameters were adjusted to find the optimal solution, generating a simulation optimization parameter table.

[0127] S603: Based on the simulation optimization parameter table, feedback optimization is performed on the parameters used in the actual construction project, and the model parameters are fine-tuned until the quality assessment model is stable, thereby obtaining a building quality assessment model;

[0128] Based on the simulation optimization parameter table, the parameters used in actual construction projects are feedback optimized, the machine learning method is used to fine-tune the model parameters, and the GridSearchCV in the Scikit-learn library is used for parameter optimization. By defining the parameter grid and evaluation criteria, GridSearchCV systematically searches and evaluates the model parameters, and adjusts the parameters according to the performance feedback of the quality assessment model until the quality assessment model is stable, thus obtaining the building quality assessment model.

[0129] See also Figure 8 , a construction project quality assessment system, the construction project quality assessment system is used to implement the above construction project quality assessment method, the system includes:

[0130] The data preprocessing module deletes missing values, replaces outliers, performs standardization, and splits data sets based on the original construction engineering data to generate preprocessed construction engineering data;

[0131] The feature encoding module builds an autoencoder model based on the preprocessed construction engineering data, sets the input layer dimension, adds encoding layer, mutation layer, decoding layer, compiles the model, and generates basic quality feature encoding;

[0132] The feature enhancement module is based on the basic quality feature coding, identifies key features, assigns weights, highlights features, reorganizes data features, and generates enhanced quality feature representation;

[0133] The simulated evolution module generates a random variation matrix based on enhanced quality feature representation, simulates the influence of environmental factors, performs standardization and matrix multiplication, identifies key variation features, and generates simulated evolution features;

[0134] The risk classification module performs cluster analysis, decision tree classification methods, data point division, risk level priority ranking, and generates quality risk classification and improvement measures based on simulated evolution characteristics;

[0135] The dynamic learning optimization module uses quality risk classification and improvement measures to dynamically and adaptively process real-time monitoring data, generate preliminary risk response strategies, adjust control strategies for monitoring data, and generate specially optimized quality control strategies;

[0136] The quality control strategy module conducts simulation experiments on construction project quality management based on the specially optimized quality control strategy, improves construction methods, material selection, construction sequence, fine-tunes model parameters, and generates a building quality assessment model.

[0137] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A construction project quality assessment method, characterized in that: The following steps are involved: Based on the original construction project data, an autoencoder model is constructed, and input layer, encoding layer, mutation layer and decoding layer are added. The autoencoder model is used to capture primary features and simulate the Gaussian noise of the input data to generate basic quality feature encoding; The basic quality feature encoding is used to perform feature learning, and by adding an attention mechanism model, the key quality features are targeted to strengthen the recognition of key indicators and obtain enhanced quality feature representation; Applying a mutation algorithm to the enhanced quality feature representation to perform feature diffusion, simulating performance changes under the influence of environmental factors, and obtaining simulated evolution features by multiplying a random mutation matrix with the feature representation; Based on the simulated evolution characteristics, a clustering algorithm is executed to perform risk classification, cluster centers are assigned to data points based on distance, differentiated quality risk levels are identified, and quality risk classification and improvement measures are generated; By using the quality risk classification and improvement measures, a dynamic adaptive learning strategy is adopted to process real-time monitoring data, dynamically adjust learning parameters, perform special optimization on detected risk points, and obtain a quality control strategy after special optimization; The specially optimized quality control strategy is applied to the quality management of construction projects. Through the feedback loop of simulation and actual application, the model parameters are refined and adjusted, and the model is continuously optimized to obtain a building quality assessment model.

2. The construction engineering quality assessment method according to claim 1, characterized in that: The basic quality feature coding includes dimensional deviation coding, material strength coding and processing accuracy coding, the enhanced quality feature representation includes weighted representation of key structure bearing capacity, weighted representation of safety factor and weighted representation of construction quality, the simulation evolution characteristics include environmental corrosion impact coding, service life impact coding and maintenance frequency impact coding, the quality risk classification and improvement measures include high-risk level identification and reinforcement plan, medium-risk level monitoring plan and maintenance plan and low-risk level routine inspection plan, the quality control strategy after special optimization includes learning parameters adjusted by real-time monitoring data feedback, targeted defect repair technology selection and improved quality prevention measures, and the building quality assessment model includes refined model parameters, quality control strategy updates, and optimized risk assessment logic.

3. The construction engineering quality assessment method according to claim 1, characterized in that: Based on the original construction project data, an autoencoder model is constructed, and the input layer, encoding layer, mutation layer, and decoding layer are added. The autoencoder model is used to capture primary features and simulate the Gaussian noise of the input data. The specific steps for generating basic quality feature encoding are as follows: Based on the original construction engineering data, the autoencoder network framework is constructed, the model is initialized and the number of input layer matching data features is set to generate the initialized network framework; Based on the initialization network framework, a coding layer is added to compress the data representation, reduce the dimension, and use the activation function to capture the nonlinear features to obtain the compressed data representation; Based on the compressed data representation, a mutation layer is introduced to simulate Gaussian noise, random noise is added through a custom function, generalization ability is optimized, and noise optimized coding is obtained; Based on the noise optimized coding, a decoding layer is added to reconstruct the data, and the network parameters are adjusted to match the original data range to obtain the basic quality feature coding.

4. The construction engineering quality assessment method according to claim 1, characterized in that: The basic quality feature encoding is used to perform feature learning. By adding an attention mechanism model, the key quality features are targeted to strengthen the recognition of key indicators. The specific steps for obtaining enhanced quality feature representation are as follows: Based on the basic quality feature coding, data analysis and feature extraction are performed to identify and distinguish key and non-key quality features in the data and generate a basic quality feature set; Using the basic quality feature set, weights are assigned and features are highlighted, key quality features are enhanced, the influence of key quality features in the model is enhanced, and a weighted set of key quality features is generated; The feature representation is optimized for the weighted set of key quality features. By strengthening the identification of key indicators and iteratively refining the expression of data features, the model's ability to identify key features of data quality is improved, and enhanced quality feature representation is generated.

5. The construction engineering quality assessment method according to claim 1, characterized in that: The steps of applying a mutation algorithm to the enhanced quality feature representation for feature diffusion to simulate performance changes under the influence of environmental factors and obtaining simulated evolution features by multiplying the random mutation matrix with the feature representation are as follows: Based on the enhanced quality feature representation, a random variation matrix is ​​generated to simulate the influence of environmental factors, and the matrix and the feature representation are adjusted to the same dimension to obtain a random variation matrix; Performing standardization processing on the random variation matrix to constrain the range of variation effects and avoid excessive variation or reduction of features, thereby obtaining a standardized variation matrix; Using the standardized variation matrix, simulating the performance change of the feature under the influence of environmental factors through matrix multiplication to generate a multiplication variation feature; Analyze and process the multiplicative variation features, use the PCA algorithm to identify key variation features, enhance the quality and availability of evolution features, and obtain simulated evolution features; The PCA algorithm uses the formula Among them, Σ adjusted is the improved covariance matrix, X is the standardized data matrix, is the average value of the data, n is the number of samples, δ is the smoothing parameter to avoid division by zero errors, S is the slope matrix of the samples, α is the weight coefficient of the slope, J is the weight matrix of the kurtosis, β is the weight coefficient of the kurtosis, K is the custom adjustment matrix, and γ is the weight coefficient of the adjustment matrix.

6. The construction engineering quality assessment method according to claim 1, characterized in that: Based on the simulated evolution characteristics, the clustering algorithm is executed to perform risk classification, the cluster centers are assigned to the data points based on the distance, the differentiated quality risk levels are identified, and the steps of generating quality risk classification and improvement measures are as follows: Based on the simulated evolution characteristics, cluster analysis is performed to divide the data points according to the preset risk level number, and the cluster center position is optimized by iteration until the convergence condition or the maximum number of iterations is met, thereby generating a risk indicator set; For the risk indicator set, a decision tree classification method is executed to classify the data points into corresponding risk levels according to the risk indicators until a stop condition is met, thereby obtaining a risk level classification; Using the risk level classification, perform priority sorting, sort in descending order according to the risk index, identify the priority of improvement measures, and customize quality risk classification and improvement measures; The decision tree classification method uses the formula Where D represents the dataset, m is the number of categories, and p i is the relative frequency of the ith category in the dataset D, Depth-β represents the depth of the current node, α and β are parameters used to adjust the sensitivity of the depth adjustment factor, γ is a parameter used to weight the variability of the dataset, Var(D) is the variability indicator of the dataset, and e is the base of the natural logarithm.

7. The construction engineering quality assessment method according to claim 1, characterized in that: Using the quality risk classification and improvement measures, a dynamic adaptive learning strategy is adopted to process real-time monitoring data, dynamically adjust learning parameters, and perform special optimization for the detected risk points. The specific steps for obtaining the specially optimized quality control strategy are as follows: Using the quality risk classification and improvement measures, real-time monitoring data is dynamically and adaptively processed, and a preliminary risk response strategy is generated by adjusting the learning rate response risk through cyclic detection of data changes; Based on the preliminary risk response strategy, using a decision tree model, identifying and optimizing currently detected risk points, recalculating learning parameters, and obtaining optimized learning parameters; Based on the optimized learning parameters, the support vector machine model is used to adjust the control strategy for all monitoring data in a detailed manner, verify the effectiveness of the parameter settings, and generate a specially optimized quality control strategy.

8. The construction engineering quality assessment method according to claim 1, characterized in that: The optimized quality control strategy is applied to the quality management of construction projects. Through the feedback loop between simulation and actual application, the model parameters are refined and adjusted, and the model is continuously optimized to obtain the steps of the building quality assessment model: Based on the application of the specially optimized quality control strategy to the quality management of construction projects, the preliminary quality data is analyzed, key quality control points are screened, areas below the threshold are marked as priority optimization areas, and a key quality control point analysis table is generated; Based on the critical quality control point analysis table, simulate the construction method, material selection, and construction sequence, improve the quality through iterative cycles, adjust the construction parameters, and generate a simulation optimization parameter table; Based on the simulation optimization parameter table, the parameters used in the actual construction project are feedback optimized and the model parameters are fine-tuned until the quality assessment model is stable, thereby obtaining a building quality assessment model.

9. A construction engineering quality assessment system, characterized in that: According to the construction project quality assessment method according to any one of claims 1 to 8, the system comprises: The data preprocessing module deletes missing values, replaces outliers, performs standardization, and splits data sets based on the original construction engineering data to generate preprocessed construction engineering data; The feature encoding module constructs an autoencoder model based on the preprocessed construction engineering data, sets the input layer dimension, adds the encoding layer, the mutation layer, the decoding layer, compiles the model, and generates the basic quality feature encoding; The feature enhancement module identifies key features based on the basic quality feature coding, assigns weights, highlights features, reorganizes data features, and generates enhanced quality feature representation; The simulated evolution module generates a random variation matrix based on the enhanced quality feature representation, simulates the influence of environmental factors, performs standardization processing and matrix multiplication, identifies key variation features, and generates simulated evolution features; The risk grading module performs cluster analysis, decision tree classification method, data point division, risk level priority ranking based on the simulated evolution characteristics, and generates quality risk grading and improvement measures; The dynamic learning optimization module uses the quality risk classification and improvement measures to dynamically and adaptively process the real-time monitoring data, generate preliminary risk response strategies, adjust the control strategies of the monitoring data, and generate specially optimized quality control strategies; The quality control strategy module conducts simulation experiments on construction project quality management, improves construction methods, material selection, construction sequence, fine-tunes model parameters, and generates a building quality assessment model based on the quality control strategy optimized by the special project.

Images