Building construction management method and system based on BIM building data

Abstract

The invention relates to the field of building construction management, in particular to a building construction management method and system based on BIM building data, and the method comprises the steps: obtaining and preprocessing quality monitoring data in historical BIM building data; and classifying the data by using the decision tree to form a plurality of categories. And calculating the fitting precision of each data point by analyzing the number of overlapped data of each category and an adjacent category in the data of each dimension, evaluating the similarity of the data points in the categories, and calculating the overall fusion degree. And taking the ratio of the fitting precision to the fusion degree as the overfitting degree, performing pruning operation on the decision tree, optimizing the classification model to obtain a corrected decision tree, and evaluating the real-time quality monitoring data to complete construction quality management. According to the method, fitting precision and fusion degree calculation are introduced, pruning operation is combined, the overfitting problem caused by limited historical data is effectively solved, and the accuracy of building quality evaluation and the generalization ability of the model are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of building construction management. In particular, it relates to a building construction management method and system based on BIM building data. BACKGROUND

[0002] Building construction is the process of transforming design drawings into actual buildings, involving multiple links and complex technical management requirements. Traditional management modes often face problems such as information silos, resource waste, and quality control lag, which seriously affect construction efficiency and engineering quality.

[0003] BIM technology integrates geometry, materials, progress, cost, and other multi-dimensional data by constructing a full-life-cycle three-dimensional visualization model, providing dynamic decision support for construction management, and is a key technology for breaking through traditional management modes in the construction industry. It can effectively solve the problem of information silos, optimize resource allocation, and improve the timeliness and accuracy of quality control, thereby significantly improving the efficiency and quality of construction management.

[0004] However, in practical applications, BIM technology also faces some challenges. Typically, decision trees are used to classify BIM data for building quality assessment. Due to the difficulty of historical data collection, the historical data set capacity is limited. In this case, using limited historical data for decision tree classification can easily lead to overfitting, resulting in inaccurate classification evaluation results. This not only affects the reliability of BIM-based quality assessment, but also may cause misjudgment in construction management. SUMMARY

[0005] To solve the problem that in practical applications, BIM technology has limited data sets due to the difficulty of historical data collection, making classification evaluation based on decision trees prone to overfitting, which affects the reliability of BIM-based quality assessment and may cause misjudgment in construction management, the present application provides solutions in the following aspects.

[0006] In a first aspect, a building construction management method based on BIM building data includes: obtaining quality monitoring data in historical BIM building data and preprocessing; performing decision tree classification on the preprocessed quality monitoring data to form several categories, calculating the fitting accuracy of each data point based on analyzing the number of overlapping data in each dimension between each category and adjacent categories, evaluating the similarity between data points within each category, calculating the overall fusion degree of each category, and taking the ratio of fitting accuracy and overall fusion degree as the overfitting degree of the category; pruning the decision tree according to the overfitting degree to optimize the classification model and obtain the corrected decision tree; using the corrected decision tree to perform quality assessment on real-time quality monitoring data to complete building construction quality management.

[0007] It effectively solves the overfitting problem of decision trees caused by limited historical data in existing technologies, improves the accuracy and reliability of quality assessment, enhances the generalization ability of the model, and thus provides more accurate decision support for construction management, improving the efficiency and scientific nature of construction management.

[0008] Preferably, the method for calculating the fitting accuracy includes: Taking any category as the target category, obtain the number of overlapping data in each dimension of the target category and the data in each dimension of the left and right adjacent categories. Sum the ratios of the number of overlapping data in each dimension to the total number of data points in the target category and adjacent categories, and take the average to obtain the overlap coefficient of each dimension of the left and right adjacent categories. Sum the overlap coefficients of all dimensions and take the average to obtain the comprehensive overlap coefficient of the target category. The number of data points in the target category is exponentially mapped using a negative exponential function. The mapping result is then multiplied by the comprehensive overlap coefficient to obtain the fitting accuracy of the target category.

[0009] By quantifying the overlap between the target category and adjacent categories across various dimensions of data, a comprehensive overlap coefficient is calculated. This coefficient is then used to perform an exponential mapping of the number of data points in the target category using a negative exponential function, yielding the fitting accuracy. This method can accurately assess the fitting effect of the target category and effectively identify overfitting. The comprehensive overlap coefficient reflects the degree of data overlap between the target category and adjacent categories; a higher overlap indicates lower class differentiation, potentially suggesting overfitting.

[0010] Preferably, the calculation method for the overall degree of integration includes: Taking any category as the target category, calculate the Euclidean distance between each data point in the target category and other data points in turn. Normalize the Euclidean distances and calculate the sum of the normalized distances of all data points. Use a negative exponential function to map the sum of distances to obtain the degree of fusion for each data point. Sum the fusion degrees of all data points and take the average to obtain the overall fusion degree of the target category.

[0011] Preferably, the calculation method for the overall degree of integration further includes: Use any category as the target category Get the target category Number of data points in Take any data point in the target category as the target data point. Select target data points In all categories of history nearest The number of data points and the relationship with the target category are obtained. Number of data points belonging to the same category As the number of local consistency; the ratio between the number of local consistency and the number of data points of the target category is taken as the local consistency ratio; The Euclidean distance between each data point in the target category and other data points is calculated in turn, the Euclidean distance is normalized, the sum of the normalized distances of all data points is calculated, and the sum of the distances is mapped using a negative exponential function to obtain the fusion degree of each data point. The product sum of the local consistency ratio of all data points and the fusion degree of all data points is averaged to obtain the overall fusion degree of the target category.

[0012] Preferably, the pruning operation of the decision tree according to the overfitting degree comprises: A preset overfitting threshold is used to mark the categories with overfitting degrees greater than the overfitting threshold as overfitting categories, the overfitting categories are pruned, the similarity between the pruned data points and adjacent categories is calculated, and the data points are reclassified into the adjacent category with the highest similarity.

[0013] Preferably, the quality evaluation of the real-time quality monitoring data using the modified decision tree comprises: obtaining and preprocessing the real-time quality monitoring data, inputting the preprocessed real-time quality monitoring data into the decision tree model, classifying and evaluating the input feature data, and one data point corresponds to one quality category, wherein the quality category includes: qualified, unqualified and attention required.

[0014] Preferably, the preprocessing step comprises: The quality monitoring data is cleaned to remove repeated, abnormal and error data, and the data format is unified to ensure the consistency of data from different sources. The missing values of the unified quality monitoring data are processed and filled using the interpolation method to ensure data integrity. The complete quality monitoring data is normalized to eliminate the influence of dimension.

[0015] Preferably, the decision tree classification of the preprocessed quality monitoring data adopts the CART algorithm, and the Gini index is used as the impurity measure to recursively divide the data and construct the classification tree according to the preprocessed quality monitoring data.

[0016] In a second aspect, a building construction management system based on BIM building data comprises: a processor and a memory, the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned building construction management method based on BIM building data is realized.

[0017] The present application has the following effects: 1、The application introduces the calculation of fitting accuracy and overall fusion degree, and combines the pruning operation of the decision tree according to the overfitting degree, effectively solves the overfitting problem of the decision tree caused by limited historical data in the prior art, can more accurately classify and evaluate the building quality, reduces the misjudgment caused by overfitting, thereby improving the reliability and accuracy of building construction quality management.

[0018] 2、The application adopts the CART algorithm combined with the Gini index for recursive division, constructs a classification tree, and optimizes the classification model through pruning operation. Not only improves the fitting effect of the model on historical data, but also significantly enhances the generalization ability of the model on new data. The modified decision tree can more effectively process real-time quality monitoring data and provide more accurate decision support for building construction management. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a method flowchart of steps S1-S3 in the building construction management method based on BIM building data according to an embodiment of the application.

[0020] Figure 2 is a structural block diagram of the building construction management system based on BIM building data according to an embodiment of the application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments of the application.

[0022] Referring to Figure 1 , the building construction management method based on BIM building data includes steps S1-S3, as follows: S1: Obtain quality monitoring data in historical BIM building data and perform preprocessing.

[0023] Obtain the BIM data of historical buildings, extract each item of data related to building quality as quality monitoring data, including but not limited to steel quality, concrete strength, and strength of various materials, etc.

[0024] Standardize the quality monitoring data to ensure that the BIM data format of each building is uniform. Buildings with missing data items are excluded. The first item of BIM data of the first building is denoted as , where represents the total number of buildings collected in the BIM data set, represents the total number of quality monitoring data of each building, and all data items of each building form a data point.

[0025] The preprocessing steps include: Conduct a preliminary review of the quality monitoring data to remove duplicate data, outliers, and obviously erroneous data records. For example, remove values ​​that exceed reasonable ranges for steel reinforcement quality or concrete strength.

[0026] Convert quality monitoring data from different sources or in different formats into a unified format to ensure data consistency; handle missing parts of the data by deleting records with missing values ​​or filling missing values ​​(such as with the mean, median, or by interpolation) to ensure data integrity; normalize the quality monitoring data to scale it to a fixed range to eliminate the influence of different data units on subsequent analysis, improve the efficiency and accuracy of data processing, and eliminate the influence of units.

[0027] S2: Perform decision tree classification on the preprocessed quality monitoring data to obtain several categories. Based on the number of overlapping data points between each category and its neighboring categories in each dimension, calculate the fitting accuracy of each data point, analyze the similarity between data points within each category, calculate the overall fusion degree of each category, and use the ratio of fitting accuracy to overall fusion degree as the overfit degree of the category.

[0028] It should be noted that the Classification and Regression Tree (CART) algorithm was used when classifying the preprocessed quality monitoring data using a decision tree. Specifically, the Gini index was used as a measure of impurity to recursively divide the data to construct a classification tree for classification.

[0029] For any target category, calculate the number of overlaps between its data in each dimension and the corresponding data in the left and right adjacent categories. Then, sum the ratios of the number of overlapping data in each dimension to the sum of the total number of data points in the target category and the adjacent categories, and take the average to obtain the overlap coefficient for each dimension. The summation and average of the overlap coefficients across all dimensions yield the comprehensive overlap coefficient of the target category. A negative exponential function is used to perform an exponential mapping on the number of data points in the target category, and the mapping result is multiplied by the comprehensive overlap coefficient to obtain the fitting accuracy of the target category.

[0030] Specifically, the fitting accuracy satisfies the following relationship: ; In the formula, Indicates the first Fitting accuracy for each category, Indicates the first The number of data points contained in each category Indicates the total number of categories. Indicates the first and the the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, the number of coincident data in the category, represents an exponential function with base natural number That is, the greater the exponent of the number of data points in each category, the fewer the number of data points in the category, indicating that the segmentation accuracy is too high and needs to be adjusted.

[0031] represents the proportion of the number of coincident data in the target category and the total number of data points in the two adjacent categories on the left and right, which reflects the degree of overlap between the target category and the adjacent categories on the left and right in a certain dimension by taking the average value, i.e., the overlap coefficient. In decision tree classification, the division of categories is based on the distribution characteristics of data. By calculating the data range of two adjacent categories in each dimension, it can be evaluated whether the current classification is reasonable. If there is a large overlap area between the two categories in multiple dimensions, it means that their data distribution in these dimensions is similar, which may indicate that the current classification is too fine and there is a risk of overfitting.

[0032] It should be noted that, taking the target category

[0033] as an example, the left category of the target category is , and the right category of the target category is , the two adjacent categories and the target category are analyzed respectively, specifically, the maximum and minimum values in the two categories are obtained, the data range of each category in each dimension is determined, by comparing the data range of the two categories, the overlap area in that dimension can be identified, specifically, if the data range of one category partially or completely covers the data range of another category, it means that the two categories overlap in that dimension. Therefore, by calculating the number of coincident data in the overlap area, the overlap degree of the two categories in that dimension can be quantified. And by identifying and quantifying the overlap area between categories, it provides a basis for pruning operation of decision tree, thereby optimizing the classification model and improving the generalization ability of classification results.

[0034]

[0035] ​It can effectively quantify the degree of overlap between categories and adjacent categories, thereby accurately evaluating the rationality of clustering results, providing a scientific basis for subsequent decision tree pruning operations, thus avoiding overfitting and improving the accuracy and generalization ability of classification models.

[0036] In addition to considering the overlap coefficient between a category and its neighboring categories, it is also necessary to pay attention to the similarity of data points within a category when evaluating overfitting. If a data point in a category has low similarity to other data points, it can be considered an outlier. Due to their uniqueness, these data points are more likely to be misclassified as overfitting results during the classification process. Therefore, it is necessary to assess the degree of category fusion by calculating the similarity between data points within a category, thereby further measuring the risk of overfitting. The specific steps are as follows: Taking any category as the target category, calculate the Euclidean distance between each data point in the target category and other data points in turn. Normalize the Euclidean distances and calculate the sum of the normalized distances of all data points. Use a negative exponential function to map the sum of distances to obtain the degree of fusion for each data point. Sum the fusion degrees of all data points and take the average to obtain the overall fusion degree of the target category.

[0037] Specifically, the overall degree of integration satisfies the following relationship: ; In the formula, Indicates the first The first in the category The overall degree of integration between each data point and other data points Indicates the first The number of data points contained in each category Indicates the first The first in the category The data point and the first Euclidean distance between data points This represents the maximum Euclidean distance between all data points. This indicates the number of categories in the decision tree. Represented by natural numbers An exponential function with base 0.

[0038] The closer the data points are, the higher the degree of fusion, indicating that the data distribution within the category is compact and the features are consistent; the farther apart the data points are, the lower the degree of fusion, indicating that there are discrete points or outliers within the category, and the classification boundary may be over-divided.

[0039] By transforming abstract intra-category consistency into a measurable numerical metric, an objective basis is provided for identifying anomalous categories caused by overfitting. Categories with low fusion are more likely to contain noise or boundary errors and should be pruned first.

[0040] In addition, another embodiment includes: Use any category as the target category Get the target category Number of data points in Take any data point in the target category as the target data point. Select target data points In all categories of history nearest The number of data points and the relationship with the target category are obtained. Number of data points belonging to the same category The number of locally consistent data points is used as the local consistency ratio; the ratio between the number of locally consistent data points and the number of data points in the target category is used as the local consistency ratio. Calculate the Euclidean distance between each data point in the target category and other data points in turn. Normalize the Euclidean distance and calculate the sum of the normalized distances of all data points. Use a negative exponential function to map the sum of distances to obtain the degree of fusion for each data point. The overall fusion degree of the target category is obtained by summing the products of the local consistency ratio of all data points and the fusion degree of all data points, and taking the average.

[0041] Specifically, the overall degree of integration satisfies the subordinate relationship: ; In the formula, Indicates the first The first in the category The degree of integration between each data point and other data points Indicates the relationship with the first The first in the category Data points in all categories The data point and the first The data points belong to the same category. The number of data points in each category Indicates the first The number of data points contained in each category Indicates the first The first in the category The data point and the first Euclidean distance between data points This represents the maximum Euclidean distance between all data points. This indicates the number of categories in the decision tree. Represented by natural numbers An exponential function with base 0.

[0042] In other words, Reflects the proportion of data points belonging to the same cluster in the nearest neighbor data points of the target data point. The higher the proportion, the more consistent the classification of the target data point with the classification of its surrounding data points, and the more reliable the classification result; on the contrary, the lower the proportion, the less consistent the classification of the target data point with the classification of its surrounding data points, and there may be overfitting phenomenon.

[0043] After introducing the local consistency proportion, the fusion degree not only considers the spatial distance between data points, but also considers the classification consistency of data points. This makes the fusion degree more comprehensively reflect the rationality of the data points in the class. If most of the data points in the local neighborhood of a data point belong to the same class, the classification confidence of the data point is higher, and the fusion degree will also increase accordingly.

[0044] When the local consistency proportion is low, even if the data point is close in space to other data points, its classification may not be consistent. In this case, the fusion degree will be reduced, thereby more effectively identifying data points that may be misclassified due to overfitting. In this way, embodiment two can more accurately identify overfitting phenomenon and provide more reliable basis for subsequent decision tree pruning.

[0045] S3: pruning the decision tree based on the overfitting degree to obtain a modified decision tree, using the modified decision tree to perform quality assessment on real-time quality monitoring data, and completing construction quality management.

[0046] A preset overfitting threshold, for example 0.8, is used to mark the classes with overfitting degree greater than the threshold as overfitting classes. Subsequently, pruning operation is performed on these overfitting classes to remove part of the data points therein. Then, the similarity of these removed data points with adjacent classes is calculated, and they are reclassified into the adjacent class with the highest similarity, thereby completing the update of the classes.

[0047] Real-time quality monitoring data is obtained and preprocessed; the preprocessed real-time quality monitoring data is input into the decision tree model to perform classification and evaluation operation on the input feature data, wherein each data point corresponds to a quality class, and the quality class covers: qualified, unqualified and attention required.

[0048] The application also provides a construction management system based on BIM building data. As shown in Figure 2 The system includes a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the construction management method based on BIM building data according to the first aspect of the application. The system also includes a communication bus and a communication interface and other components familiar to those skilled in the art, the arrangement and function of which are known in the art, so they will not be described here.

[0049] It should be noted that, for the person of ordinary skill in the art, several modifications and improvements can be made without departing from the inventive concept, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

Claims

1. A construction management method based on BIM building data, characterized in that, include: Acquire quality monitoring data from historical BIM building data and preprocess it; The preprocessed quality monitoring data is classified using decision trees to form several categories. Based on the analysis of the number of overlapping data points between each category and its neighboring categories in each dimension, the fitting accuracy of each data point is calculated, the similarity between data points within each category is evaluated, and the overall fusion degree of each category is calculated. The ratio of fitting accuracy to overall fusion degree is used as the overfit degree of the category. The decision tree is pruned based on the degree of overfitting to optimize the classification model and obtain a corrected decision tree. The corrected decision tree is then used to assess the quality of real-time quality monitoring data and complete the construction quality management.

2. The construction management method based on BIM building data according to claim 1, characterized in that, The calculation method for the fitting accuracy includes: Taking any category as the target category, obtain the number of overlapping data in each dimension of the target category and the data in each dimension of the left and right adjacent categories. Sum the ratios of the number of overlapping data in each dimension to the total number of data points in the target category and adjacent categories, and take the average to obtain the overlap coefficient of each dimension of the left and right adjacent categories. Sum the overlap coefficients of all dimensions and take the average to obtain the comprehensive overlap coefficient of the target category. The number of data points in the target category is exponentially mapped using a negative exponential function. The mapping result is then multiplied by the comprehensive overlap coefficient to obtain the fitting accuracy of the target category.

3. The construction management method based on BIM building data according to claim 1, characterized in that, The calculation method for the overall degree of integration includes: Taking any category as the target category, calculate the Euclidean distance between each data point in the target category and other data points in turn. Normalize the Euclidean distances and calculate the sum of the normalized distances of all data points. Use a negative exponential function to map the sum of distances to obtain the degree of fusion for each data point. Sum the fusion degrees of all data points and take the average to obtain the overall fusion degree of the target category.

4. The construction management method based on BIM building data according to claim 1, characterized in that, The calculation method for the overall degree of integration also includes: Use any category as the target category Get the target category Number of data points in Take any data point in the target category as the target data point. Select target data points In all categories of history nearest The number of data points and the relationship with the target category are obtained. Number of data points belonging to the same category The number of locally consistent data points is used as the local consistency ratio; the ratio between the number of locally consistent data points and the number of data points in the target category is used as the local consistency ratio. Calculate the Euclidean distance between each data point in the target category and other data points in turn. Normalize the Euclidean distance and calculate the sum of the normalized distances of all data points. Use a negative exponential function to map the sum of distances to obtain the degree of fusion for each data point. The overall fusion degree of the target category is obtained by summing the products of the local consistency ratio of all data points and the fusion degree of all data points, and taking the average.

5. The construction management method based on BIM building data according to claim 1, characterized in that, The pruning operation of the decision tree based on the degree of overfitting includes: A preset overfitting threshold is set, and categories with overfitting degree greater than the overfitting threshold are marked as overfitting categories. The overfitting categories are pruned, and the similarity between the pruned data points and the adjacent categories is calculated. The data points are then reclassified into the adjacent category with the highest similarity.

6. The construction management method based on BIM building data according to claim 1, characterized in that, The process of using the modified decision tree to assess the quality of real-time quality monitoring data includes: acquiring real-time quality monitoring data and preprocessing it; inputting the preprocessed real-time quality monitoring data into the decision tree model; classifying and evaluating the input feature data; with each data point corresponding to a quality category; and the quality categories including: qualified, unqualified, and requiring attention.

7. The construction management method based on BIM building data according to claim 1, characterized in that, The preprocessing steps include: The quality monitoring data is cleaned to remove duplicate, abnormal, and erroneous data, and the data format is standardized to ensure consistency between data from different sources. Missing values ​​in the standardized quality monitoring data are filled using interpolation to ensure data integrity. The complete quality monitoring data is then normalized to eliminate the influence of units.

8. The construction management method based on BIM building data according to claim 1, characterized in that, The decision tree classification of the preprocessed quality monitoring data uses the CART algorithm. Based on the preprocessed quality monitoring data, the Gini index is used as a measure of impurity to recursively divide the data and construct a classification tree.

9. A building construction management system based on BIM building data, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the building construction management method based on BIM building data according to any one of claims 1-8.

Images