BIM and GIS construction data transmission method and system

Abstract

The invention relates to the technical field of construction data transmission, and discloses a BIM and GIS construction data transmission method and system, and the data transmission process comprises the steps: S1, obtaining construction data items in the construction data process, and uniformly outputting the construction data items as behavior vectors; s2, performing behavior prediction on the behavior vectors of the construction data items by using an autoregressive moving average (ARIMA) model to obtain a behavior mode and a behavior trend; s3, establishing a baseline model by using the behavior vector, the behavior mode and the behavior trend, taking the behavior vector as input, and obtaining an emergency degree score based on the baseline model; and S4, based on the emergency degree score, setting a transmission threshold, when the emergency degree score is less than the transmission threshold, not transmitting the construction data item, and when the emergency degree score is greater than the transmission threshold, performing transmission caching on the construction data item.

Description

Technical Field

[0001] The present invention relates to the technical field of construction data transmission, and in particular to a BIM and GIS construction data transmission method and system. Background Art

[0002] With the growing demand for spatial information in the construction industry, the integration of BIM and GIS has become an inevitable trend. BIM models provide detailed building structure information, while GIS provides rich geographic spatial information. The combination of the two can create a more comprehensive and accurate digital twin model that not only covers the details of the building's interior, but also includes the geographical features of the building's environment, such as topography, surrounding facilities, and traffic conditions.

[0003] However, in very crowded construction areas, especially in scenarios where multiple large projects are being constructed in parallel, a large number of cached data items will appear. The edge server responsible for processing BIM and GIS data has limited computing and storage resources. The accumulation of a large number of cached data items will cause resource constraints and make it impossible to efficiently process all urgent tasks, resulting in delays in some data processing and affecting the processing efficiency of construction data. Therefore, a BIM and GIS construction data transmission method and system are proposed herein, which can effectively solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art that the edge server has limited storage and computing resources and cannot process a large number of data items cached by BIM and GIS in the construction project in the first time, and to propose a BIM and GIS construction data transmission method and system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for transmitting BIM and GIS construction data, the data transmission process is as follows:

[0007] S1: Obtain construction data items during the construction data process, and uniformly output the construction data items as behavior vectors;

[0008] S2: Use the autoregressive moving average model (ARIMA) model to predict the behavior vector of the construction data item and obtain the behavior pattern and behavior trend;

[0009] S3: Use the behavior vector, behavior pattern, and behavior trend to establish a baseline model, use the behavior vector as input, and obtain an urgency score based on the baseline model;

[0010] S4: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission.

[0011] The construction data items include BIM three-dimensional model data and GIS geographic information data.

[0012] The construction data item acquisition process is as follows:

[0013] The construction data item is defined to be composed of a spatial behavior feature vector and a resource behavior feature vector;

[0014] Based on GIS geographic information data, spatial behavior characteristics are obtained. The spatial behavior characteristics include distance and azimuth. Based on the spatial behavior characteristics, spatial behavior characteristic vectors are obtained. Based on BIM three-dimensional model data, resource behavior characteristics are obtained. Resource behavior characteristics include material consumption rate. Based on resource behavior characteristics, resource behavior characteristic vectors are obtained.

[0015] The process of acquiring the spatial behavior characteristics is as follows:

[0016] Define the distance between the coordinate points of two GIS objects in the GIS geographic information data as d, and use the Euclidean distance to calculate the distance on the plane coordinate system:

[0017] Where r is the radius of the Earth, μ 1 , μ is the latitude of the two points, λ 1 ,λ 2 is the longitude of two points;

[0018] Define the azimuth of the coordinate points of two GIS objects in GIS geographic information data as α, then: α = arctan2(sin(Δλ)cos(μ 2 ),cos(μ 1 )sin(μ 2 )-sin(μ 1 )cos(μ 2 )cos(Δλ)), where Δλ is the longitude λ of the two points 1 ,λ 2 The difference between 2 -λ 1 , get the spatial behavior characteristics;

[0019] The process of obtaining the spatial behavior feature vector is as follows:

[0020] Define the spatial behavior eigenvector as D ij , based on the distance matrix to represent spatial relationships: D ij =(x i -xj ) 2 +(y i -y j ) 2 +(z i -z j ) 2 , where i and j represent the coordinate points of two GIS objects, x i ,y i , z i and x j ,y j , z j Respectively represent the coordinate values ​​of the coordinate points of two GIS objects in the three-dimensional coordinate system;

[0021] The process of obtaining the resource behavior characteristics is as follows:

[0022] Define the resource types in the BIM 3D model data to include manpower, equipment, materials and funds. Obtain the type and quantity of each resource type based on the BIM model, divide them to get the material consumption rate per unit volume, and express it as the resource behavior characteristics;

[0023] The process of obtaining the resource behavior feature vector is as follows:

[0024] Based on the resource behavior characteristics, the resource demand matrix is ​​used for representation. All resource types involved in the project, such as manpower, equipment, materials and funds, are clarified. The geometric information (length, width, height) and material information of each component (such as wall, beam, plate) are extracted from the BIM model. Based on the material information of each component in the BIM model, the material consumption rate of each resource type is summarized. At the same time, all activities or tasks in the project are determined. These resources and activities form the basis of the resource demand matrix. Create a resource demand matrix, which is a two-dimensional table in which rows represent the number of project activities and columns represent resource types. Each cell in the matrix represents the material consumption rate of a specific activity for a specific resource. The demand for it in the resource demand matrix constitutes a vector. The dimension of this vector is equal to the number of resource types. If the project involves three resource types, the resource behavior characteristic vector of each activity will be a three-dimensional vector, and each element represents the demand for that resource type. The resource behavior characteristic vector is obtained.

[0025] Based on the spatial behavior feature vector and the resource behavior feature vector, the construction data item is obtained.

[0026] The process of obtaining the behavior vector is as follows:

[0027] The spatial behavior feature vector and the resource behavior feature vector are normalized and a unified vector representation is output to obtain a behavior vector.

[0028] The process of obtaining the behavior pattern and behavior trend is as follows:

[0029] Use the behavior vector to draw the autocorrelation function ACF diagram and the partial autocorrelation function PACF diagram, and define the behavior vector as Y = (Y 1 Y 2 ,...Y T ), where T is the total number of observations of the behavior sequence in the behavior vector, and the process of drawing the ACF graph and the PACF graph is:

[0030] Define the autocorrelation coefficient as r, and the process of calculating the autocorrelation coefficient ACF is: Among them, r k represents the autocorrelation coefficient of lag k, Y t represents the observed value of the behavior sequence Y at time point t, represents the average value of the behavior sequence Y, Y t-k Represents the observed value of the behavior sequence Y at time point tk, based on the autocorrelation coefficient r k The values ​​are plotted statistically to obtain the autocorrelation function ACF graph;

[0031] The process of calculating the partial autocorrelation coefficient PACF is: Based on the regression model formula: Y t =α+β 1 Y t-1 +β 2 Y t-2 +…+β k Y t-k +∈ t , where α is the intercept term of the model, β 1 , β 2 , β k is the autoregression coefficient, Y t-1 , Y t-2 , Y t-k is the lag value of the vector sequence, ∈ t is the error term at time point t, based on the observed value Y of the behavior sequence Y output by the regression model at time point t t Draw a statistical graph to obtain the partial autocorrelation function PACF graph;

[0032] Based on the ACF and PACF graphs, the parameters p, d, and q of the autoregressive moving average model ARIMA model are determined, where p is the autoregressive term, d is the difference order, and q is the moving average term. The determination process of the parameters p, d, and q of the autoregressive moving average model ARIMA model is:

[0033] The stage where the bar graph in the PACF diagram is zero is the autoregressive term p, and the stage where the bar graph in the ACF diagram becomes zero is the moving average term q. The ACF diagram shows that the data becomes stable after d times of difference, so the difference order is d;

[0034] Based on the determined parameters p, d, q, the autoregressive moving average model ARIMA model is fitted. The autoregressive moving average model ARIMA model is expressed as: φ(B)(1-B) d yt=θ(B)∈ t Among them, φ(B) represents the autoregressive polynomial, θ(B) represents the moving average polynomial, B represents the backshift operator, and the process of obtaining the autoregressive polynomial is: φ(B) = 1-φ 1 B-φ 2 B2-...-φ p B q , the process of obtaining the moving average polynomial is: θ(B)=1+θ 1 B+θ 2 B 2 +...+θ q B q , the process of obtaining the backshift operator is: Input the behavior vector into the ARIMA model, output the seasonal and periodic sequence data appearing in the behavior sequence of the vector, obtain the behavior pattern, output the sequence data of the long-term trend of the behavior sequence over time, and obtain the behavior trend.

[0035] The baseline model establishment process is as follows:

[0036] Convolution calculation is performed on the sequence data in the behavior pattern and behavior trend. The convolution calculation process is:

[0037] The convolutional neural network CNN one-dimensional convolution layer, pooling layer and fully connected layer are used to process the sequence data in the behavior pattern and behavior trend. The output sequence of the one-dimensional convolution layer is defined as S. The sequence data is input into the one-dimensional convolution layer. The one-dimensional convolution layer slides on the time series through the filter and performs convolution operation on the sequence data in each window: Among them, S(t) is the output value of the convolution operation at time point t, w i is the weight of the filter at position i, X(ti) is the value of the input sequence data at time point (ti), b is the bias term, k-1 is the upper limit index of the filter window, and the output sequence S is obtained;

[0038] The output sequence S is subjected to the maximum pooling value output feature operation through the pooling layer, and the maximum value is selected from each pooling window as the representative of the window. The pooling window size is defined as p, and the maximum pooling process is: Z(t)=max{S(t),S(t+1),…,S(t+p-1)}, where Z(t) is the output feature of the pooling layer at time point t, and the maximum pooling value feature Z of the output sequence S after the pooling layer is obtained;

[0039] The output sequence S is averaged and pooled to calculate the average value in the pooling window at each time t as the output feature: Among them, M(t) is the output feature of the average layer at time point t, It means the sum of the values ​​of p consecutive time points starting from time point t and ending at t+p-1, and the average value feature M of the output sequence S after the averaging layer is obtained;

[0040] The fully connected layer takes the maximum pooling value feature Z and the average value feature M of the average layer as input, and defines the maximum number of neurons in the fully connected layer, convolutional layer and pooling layer to be j, and the output of the fully connected layer is H: Among them, H j is the output of the jth neuron in the fully connected layer, v ij is the weight between the input feature at the ith position and the jth neuron, z i is the output of the neuron at position i, c j is the bias term of the jth neuron, σ is the activation function, and the output prediction value H of the fully connected layer is obtained;

[0041] The fully connected layer output prediction value H is combined with the ARIMA model to obtain the baseline model, which is expressed as: (1-φ 1 B-φ 2 B2-...-φ p B q )(1-B) d Y t =(1+θ 1 B+θ 2 B 2 +...+θ q B q )∈ t +βH, β is the coefficient of the fully connected layer output prediction value H, and the baseline model is obtained.

[0042] The process of obtaining the urgency score is as follows:

[0043] Input the behavior vector Y corresponding to the data item into the baseline model to obtain the predicted value H Y , predicted value H Y Define the transmission threshold V, H as the urgency score YWhen it is >V, for H Y Perform transmission caching on the corresponding data item, H Y When it is <V, for H Y Do not transmit the corresponding data item.

[0044] A BIM and GIS construction data transmission system includes a BIM+GIS data integration module for obtaining construction data and status indicators during the process, and the module for construction data transmission includes:

[0045] Data preprocessing module: Obtain construction data items during the process of obtaining construction data, and uniformly output the construction data items as behavior vectors;

[0046] Data planning module: Use the autoregressive integrated moving average model ARIMA to perform behavior prediction on the behavior vectors of construction data items, and obtain behavior patterns and behavior trends based on the prediction results;

[0047] Data urgency assignment module: Establish a baseline model using the behavior vector and behavior patterns and trends, use the behavior vector as input, and obtain an urgency score based on the baseline model;

[0048] Data transmission strategy module: Based on the urgency score, set a transmission threshold. When the urgency score is less than the transmission threshold, do not transmit the construction data item. When the urgency score is greater than the transmission threshold, perform transmission caching on the construction data item.

[0049] The present invention has the following beneficial effects:

[0050] 1. By combining the behavior vectors, behavior patterns, and behavior trends of data items during the transmission process, and establishing a baseline model through the ARIMA model to calculate the urgency score, all safety-related data can be automatically sorted to ensure that urgent and critical data is processed first. In this way, construction managers and safety personnel can quickly focus on the most urgent issues instead of wasting time on low-priority events. Description of the Drawings

[0051] Figure 1 It is a method step diagram of a BIM and GIS construction data transmission method and system proposed by the present invention.

[0052] Figure 2 It is a system block diagram of a BIM and GIS construction data transmission method and system proposed by the present invention

[0053] S1: Obtain construction data items during the process of obtaining construction data, and uniformly output the construction data items as behavior vectors;

[0054] S2: Use the autoregressive moving average model (ARIMA) model to predict the behavior vector of the construction data item and obtain the behavior pattern and behavior trend;

[0055] S3: Use the behavior vector, behavior pattern, and behavior trend to establish a baseline model, use the behavior vector as input, and obtain an urgency score based on the baseline model;

[0056] S4: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Embodiment 1

[0059] like Figure 1 As shown, the present invention proposes a construction data transmission method based on BIM+GIS, and the data transmission process is as follows:

[0060] S1: Obtain construction data items during the construction data process, and uniformly output the construction data items as behavior vectors;

[0061] S2: Use the autoregressive moving average model (ARIMA) model to predict the behavior vector of the construction data item and obtain the behavior pattern and behavior trend;

[0062] S3: Use the behavior vector, behavior pattern, and behavior trend to establish a baseline model, use the behavior vector as input, and obtain an urgency score based on the baseline model;

[0063] S4: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission.

[0064] Construction data items include BIM three-dimensional model data and GIS geographic information data.

[0065] The process of obtaining construction data items is as follows:

[0066] The construction data item is defined to be composed of a spatial behavior feature vector and a resource behavior feature vector;

[0067] Based on GIS geographic information data, spatial behavior characteristics are obtained. The spatial behavior characteristics include distance and azimuth. Based on the spatial behavior characteristics, spatial behavior characteristic vectors are obtained. Based on BIM three-dimensional model data, resource behavior characteristics are obtained. Resource behavior characteristics include material consumption rate. Based on resource behavior characteristics, resource behavior characteristic vectors are obtained.

[0068] Furthermore, the distance and azimuth include the spatial distance between objects (such as equipment, personnel, and material stacking points), which can be a straight-line distance or a distance based on a road network, and the directional relationship between objects, that is, the directional angle from one point to another. The change in the position of the object over time can be used to track the movement path of equipment or personnel, obtain characteristic data that reflects the spatial behavior characteristics, and form a vector. The material consumption rate is the amount of material used per unit volume or unit area, such as concrete, steel, wood, etc., which can be used for cost estimation and resource planning, collect relevant cost data, and form a resource behavior vector;

[0069] The process of acquiring spatial behavior features is as follows:

[0070] Define the distance between the coordinate points of two GIS objects in the GIS geographic information data as d, and use the Euclidean distance to calculate the distance on the plane coordinate system:

[0071] Where r is the radius of the Earth, μ 1 , μ is the latitude of the two points, λ 1 ,λ 2 is the longitude of two points;

[0072] Define the azimuth of the coordinate points of two GIS objects in GIS geographic information data as α, then: α = arctan2(sin(Δλ)cos(μ 2 ),cos(μ 1 )sin(μ 2 )-sin(μ 1 )cos(μ 2 )cos(Δλ)), where Δλ is the longitude λ of the two points 1 ,λ 2 The difference between 2 -λ 1 , get the spatial behavior characteristics;

[0073] The process of obtaining the spatial behavior feature vector is as follows:

[0074] Define the spatial behavior eigenvector as D ij , based on the distance matrix to represent spatial relationships: D ij =(x i -x j ) 2 +(yi -y j ) 2 +(z i -z j ) 2 , where i and j represent the coordinate points of two GIS objects, x i ,y i , z i and x j ,y j , z j Respectively represent the coordinate values ​​of the coordinate points of two GIS objects in the three-dimensional coordinate system;

[0075] The process of obtaining resource behavior characteristics is as follows:

[0076] Define the resource types in the BIM 3D model data to include manpower, equipment, materials and funds. Obtain the type and quantity of each resource type based on the BIM model, divide them to get the material consumption rate per unit volume, and express it as the resource behavior characteristics;

[0077] The process of obtaining resource behavior feature vector is as follows:

[0078] Based on the resource behavior characteristics, the resource demand matrix is ​​used to represent and obtain the resource behavior characteristic vector;

[0079] Furthermore, the resource feature vector is represented by the resource requirement matrix as follows: obtain all resource types involved in the project, such as manpower, equipment, materials and funds. At the same time, identify all specific activities or tasks in the project, which are the carriers of resource requirements, and then construct a resource requirement matrix. Create a resource requirement matrix, the rows of which represent the activities in the project, and the columns represent the resource types. For each activity in the project, collect and record its specific requirements for each resource, and fill in the resource requirement matrix. The resource requirement matrix represents the resource behavior feature vector;

[0080] Based on the spatial behavior feature vector and the resource behavior feature vector, the construction data item is obtained.

[0081] The process of obtaining the behavior vector is:

[0082] The spatial behavior feature vector and resource behavior feature vector are normalized, and the minimum-maximum normalization is selected to scale the original spatial behavior feature vector and resource behavior feature vector to a uniform range of [0,1]. Among them, min i and max i are the maximum and minimum values ​​in the eigenvector, r i For the element corresponding to position i in the feature vector, calculate the normalized value, output a unified vector representation, and obtain the behavior vector.

[0083] The process of obtaining behavior patterns and behavior trends is as follows:

[0084] Use the behavior vector to draw the autocorrelation function ACF diagram and the partial autocorrelation function PACF diagram, and define the behavior vector as Y = (Y 1 Y 2 ,...Y T ), where T is the total number of observations of the behavior sequence in the behavior vector, and the process of drawing the ACF graph and the PACF graph is:

[0085] Define the autocorrelation coefficient as r, and the process of calculating the autocorrelation coefficient ACF is: Among them, r k represents the autocorrelation coefficient of lag k, Y t represents the observed value of the behavior sequence Y at time point t, represents the average value of the behavior sequence Y, Y t-k Represents the observed value of the behavior sequence Y at time point tk, based on the autocorrelation coefficient r k The values ​​are plotted statistically to obtain the autocorrelation function ACF graph;

[0086] Furthermore, the ACF plot shows the autocorrelation coefficient r at different lags k. k In the ACF diagram, the horizontal axis represents the value of lag k, and the vertical axis represents the corresponding autocorrelation coefficient r k Through the ACF graph, we can intuitively see how the autocorrelation of the sequence changes with the lag time. If the ACF graph shows a high autocorrelation coefficient, especially at a small lag value, this indicates that the sequence has a strong autoregressive characteristic. If the ACF graph shows a significant peak at certain specific lag values, this may indicate that there is a periodic pattern in the sequence. If the ACF graph still shows a high autocorrelation coefficient when the lag k is large, this may indicate that there is a long-range dependency in the sequence.

[0087] The process of calculating the partial autocorrelation coefficient PACF is: Based on the regression model formula: Y t =α+β 1 Y t-1 +β 2 Y t-2 +…+β k Y t-k +∈ t , where α is the intercept term of the model, β 1 , β 2 , β k is the autoregression coefficient, Y t-1 , Y t-2 , Y t-kis the lag value of the vector sequence, where i = 1, 2, ..., k, Y t-1 represents the observed value of the previous period at time point t, Y t-2 Represents the observed value two periods before time point t, and so on, ∈ t is the error term at time point t;

[0088] Furthermore, by estimating α and β i We can understand the dynamic behavior of the sequence by looking at the value of , and then calculate the observed value Y at time point t based on the behavior sequence Y output by the regression model. t Draw a statistical graph to obtain the partial autocorrelation function PACF graph;

[0089] Based on the ACF and PACF graphs, the use of data to draw ACF and PACF graphs has different effects. The ACF graph shows the degree of autocorrelation between the time series and its own different lag orders, and the PACF graph shows the degree of autocorrelation between the time series and its own different lag orders, but excludes the influence of the intermediate lag. The parameters p, d, and q of the autoregressive moving average model ARIMA model are determined, where p is the autoregressive term, d is the difference order, and q is the moving average term. The determination process of the parameters p, d, and q of the autoregressive moving average model ARIMA model is:

[0090] The stage where the bar graph in the PACF diagram is zero is the autoregressive term p, and the stage where the bar graph in the ACF diagram becomes zero is the moving average term q. The ACF diagram shows that the data becomes stable after d times of difference, so the difference order is d;

[0091] Further, the autoregressive term p is determined. In the PACF graph, if the bar chart after a certain order suddenly becomes zero or close to zero, then the order is the value of p. Usually, the lag order corresponding to the significant peak of PACF indicates the order of the autoregressive model.

[0092] Further, the moving average term q is determined. In the ACF graph, if the bar chart after a certain order suddenly becomes zero or close to zero, then the order is the value of q. The lag order corresponding to the significant peak of the ACF usually indicates the order of the moving average model.

[0093] Furthermore, the order of difference d is determined. If the ACF plot shows that the data becomes stable after multiple differences, then the value of d is the number of differences required to achieve stability.

[0094] Based on the determined parameters p, d, q, the autoregressive moving average model ARIMA model is fitted. The autoregressive moving average model ARIMA model is expressed as: φ(B)(1-B) d yt=θ(B)∈ tAmong them, φ(B) represents the autoregressive polynomial, θ(B) represents the moving average polynomial, B represents the backshift operator, and the process of obtaining the autoregressive polynomial is: φ(B) = 1-φ 1 B-φ 2 B2-...-φ p B q , the process of obtaining the moving average polynomial is: θ(B)=1+θ 1 B+θ 2 B 2 +...+θ q B q , the process of obtaining the backshift operator is: , input the behavior vector into the autoregressive moving average model ARIMA model, output the seasonal and periodic sequence data appearing in the behavior sequence of the vector, obtain the behavior pattern, output the sequence data of the long-term trend of the behavior sequence over time, and obtain the behavior trend.

[0095] The baseline model building process is:

[0096] Convolution calculation is performed on the sequence data in the behavior pattern and behavior trend. The convolution calculation process is:

[0097] Use the one-dimensional convolutional layer, pooling layer and fully connected layer of the convolutional neural network (CNN) to process the sequence data in the behavior patterns and behavior trends;

[0098] Furthermore, the one-dimensional convolutional layer slides over the time series through a small window called a "filter" and performs a convolution operation on the data within each window. The filter is usually composed of a set of learnable weights, which produces an output by performing a dot product operation with the time series data;

[0099] Define the output sequence of the one-dimensional convolution layer as S, input the sequence data into the one-dimensional convolution layer, and the one-dimensional convolution layer slides on the time series through the filter to perform convolution operation on the sequence data in each window: Among them, S(t) is the output value of the convolution operation at time point t, w i is the weight of the filter at position i, X(ti) is the value of the input sequence data at time point (ti), b is the bias term, k-1 is the upper limit index of the filter window, and the output sequence S is obtained;

[0100] Furthermore, k-1 defines the range of input sequence elements considered when the filter is applied. The size of the filter is k, which means that the filter will consider the data at the current time point t and the previous k-1 time points, that is, the data from t to t-(k-1). Specifically, when i=0, the filter is applied to the data X(t) at the current time point t, multiplied by the weight w 0, when i=1, the filter is applied to the data x(t-1) at the previous time point t-1, multiplied by the weight w 1 , when i = k-1, the filter is applied to the data x(t-(k-1)) at t-(k-1) before k-1 time points, multiplied by the weight w k-1 , therefore, k-1 determines the span of the filter;

[0101] The output sequence S is passed through the pooling layer to perform the maximum pooling value output feature operation, and the maximum value in each pooling window is selected as the representative of the window;

[0102] Define the pooling window size as p, and the maximum pooling process is: Z(t) = max{S(t), S(t+1), ..., S(t+p-1)}, where Z(t) is the output feature of the pooling layer at time point t, and the maximum pooling value feature Z of the output sequence S after the pooling layer is obtained;

[0103] The output sequence S is averaged and pooled to calculate the average value in the pooling window at each time t as the output feature: Among them, M(t) is the output feature of the average layer at time point t, It means the sum of the values ​​of p consecutive time points starting from time point t and ending at t+p-1, and the average value feature M of the output sequence S after the averaging layer is obtained;

[0104] The fully connected layer takes the maximum pooled value feature Z and the average value feature M of the average layer as input. The function of the fully connected layer is to integrate all the features extracted from the convolutional layer and the pooling layer to form the final prediction or classification. In a fully connected layer, each neuron is connected to all neurons in the previous layer.

[0105] Define the maximum number of neurons in the fully connected layer, convolutional layer and pooling layer to be j;

[0106] Furthermore, in a fully connected layer, each neuron is connected to all neurons in the previous layer, which can be expressed as:

[0107] For each neuron j in a fully connected layer, it receives the output z from all neurons in the previous layer i , and compare them with the corresponding weights v ij Multiply and then sum, and add the bias term c to the result of this weighted sum j , forming the net input of the neuron. Finally, this net input is transformed by the activation function σ to produce the final output H j ;

[0108] The output of the fully connected layer is H: Among them, H jis the output of the j-th neuron in the fully connected layer, v ij is the weight between the input feature at the i-th position and the j-th neuron, z i is the output of the neuron at the i-th position, c j is the bias term of the j-th neuron, σ is the activation function, and the output prediction value H of the fully connected layer is obtained;

[0109] Combining the output prediction value H of the fully connected layer into the ARIMA model to obtain the baseline model, which is expressed as: (1 - φ 1 B - φ 2 B2 -... - φ p B q )(1 - B) d Y t =(1 + θ 1 B + θ 2 B 2 +... + θ q B q )∈ t + βH, where β is the coefficient of the output prediction value H of the fully connected layer.

[0110] Furthermore, the model formed by combining the convolutional neural network CNN and the autoregressive integrated moving average model ARIMA can be regarded as a composite baseline model. In this composite model, CNN is responsible for automatically learning and extracting complex local features from time series data, while the ARIMA model is responsible for dealing with the trend and seasonal components in the data. The combination of the two can more comprehensively capture the dynamic characteristics of the time series.

[0111] The process of obtaining the urgency score is as follows:

[0112] Input the behavior vector Y corresponding to the data item into the baseline model to obtain the prediction value H Y , and the prediction value H Y is the urgency score. Define the transmission threshold V. When H Y > V, cache the data item corresponding to H Y . When H Y < V, do not transmit the data item corresponding to H Y .

[0113] In this embodiment, by collecting construction safety-related data items from BIM and GIS systems, including but not limited to BIM three-dimensional model data and GIS geographic information data, a behavior vector is constructed. First, the characteristics are defined to determine which variables can reflect the characteristics of the behavior, and the spatial behavior feature vector and the resource behavior feature vector are obtained. These features are converted into numerical forms to form a behavior vector, and then the ARIMA model is selected for behavior prediction to capture the trend and seasonal fluctuation of data over time. The ARIMA model is trained using the behavior vector, and the model parameters, p autoregressive term, d difference order, q sliding average term, are adjusted to obtain the behavior pattern and behavior characteristics. Then, the baseline is defined. The baseline model reflects the expected value and range of variation of the behavior vector under normal construction conditions. The sequence data in the behavior pattern and behavior trend are convoluted to obtain the large pooling value feature Z of the most pooled layer and the average value feature M of the average layer, and then the full connection layer outputs H j ,The fully connected layer output prediction value H is combined into the ARIMA model to obtain the baseline model.,Based on the output result of the baseline model, it is regarded as an urgency score.,A threshold for the urgency score is defined. Data exceeding the threshold is regarded as high priority and given,priority transmission.

[0114] Embodiment 2

[0115] like Figure 2 As shown, based on the first embodiment, please refer to Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, and a construction data transmission system based on BIM+GIS is provided, including a BIM+GIS data integration module for acquiring construction data and process status indicators, and the modules for construction data transmission include:

[0116] Data preprocessing module: obtains construction data items in the process of construction data, and uniformly outputs the construction data items as behavior vectors;

[0117] Data planning module: Use the autoregressive moving average model ARIMA to predict the behavior vector of the construction data item, and obtain the behavior pattern and behavior trend based on the prediction results;

[0118] Data urgency allocation module: Use behavior vectors, behavior patterns, and behavior trends to establish a baseline model, use the behavior vectors as input, and obtain urgency scores based on the baseline model;

[0119] Data transmission strategy module: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission.

[0120] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for transmitting BIM and GIS construction data, characterized in that: The data transmission process is: S1: Obtain construction data items during the construction data process, and uniformly output the construction data items as behavior vectors; S2: Use the autoregressive moving average model (ARIMA) model to predict the behavior vector of the construction data item and obtain the behavior pattern and behavior trend; S3: Use the behavior vector, behavior pattern, and behavior trend to establish a baseline model, use the behavior vector as input, and obtain an urgency score based on the baseline model; S4: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission.

2. A BIM and GIS construction data transmission method according to claim 1, characterized in that: The construction data items include BIM three-dimensional model data and GIS geographic information data.

3. A BIM and GIS construction data transmission method according to claim 1, characterized in that: The construction data item acquisition process is as follows: The construction data item is defined to be composed of a spatial behavior feature vector and a resource behavior feature vector; Based on GIS geographic information data, spatial behavior characteristics are obtained. The spatial behavior characteristics include distance and azimuth. Based on the spatial behavior characteristics, spatial behavior characteristic vectors are obtained. Based on BIM three-dimensional model data, resource behavior characteristics are obtained. Resource behavior characteristics include material consumption rate. Based on resource behavior characteristics, resource behavior characteristic vectors are obtained. The process of acquiring the spatial behavior characteristics is as follows: Define the distance between the coordinate points of two GIS objects in the GIS geographic information data as d, and use the Euclidean distance to calculate the distance on the plane coordinate system: Among them, r is the radius of the earth, μ1, μ are the latitudes of the two points, λ1, λ2 are the longitudes of the two points; Define the azimuth of the coordinate points of two GIS objects in the GIS geographic information data as α, then: α = arctan2(sin(Δλ)cos(μ2),cos(μ1)sin(μ2)-sin(μ1)cos(μ2)cos(Δλ)), where Δλ is the difference between the longitudes λ1 and λ2 of the two points, expressed as Δλ = λ2-λ1; The process of obtaining the spatial behavior feature vector is as follows: Define the spatial behavior eigenvector as D ij , based on the distance matrix to represent spatial relationships: D ij =(x i -x j ) 2 +(y i -y j ) 2 +(z i -z j ) 2 , where i and j represent the coordinate points of two GIS objects, x i ,y i , z i and x j ,y j , z j Respectively represent the coordinate values ​​of the coordinate points of two GIS objects in the three-dimensional coordinate system; The process of obtaining the resource behavior characteristics is as follows: Define the resource types in the BIM 3D model data to include manpower, equipment, materials and funds. Obtain the type and quantity of each resource type based on the BIM model, divide them to get the material consumption rate per unit volume, and express it as the resource behavior characteristics; The process of obtaining the resource behavior feature vector is as follows: Based on the resource behavior characteristics, the resource demand matrix is ​​used to represent and obtain the resource behavior characteristic vector; Based on the spatial behavior feature vector and the resource behavior feature vector, the construction data item is obtained.

4. A BIM and GIS construction data transmission method according to claim 1, characterized in that: The process of obtaining the behavior vector is as follows: The spatial behavior feature vector and the resource behavior feature vector are normalized and a unified vector representation is output to obtain a behavior vector.

5. A BIM and GIS construction data transmission method according to claim 1, characterized in that: The process of obtaining the behavior pattern and behavior trend is as follows: Use the behavior vector to draw the autocorrelation function ACF diagram and the partial autocorrelation function PACF diagram, and define the behavior vector as Y = (Y1Y2, ...Y T ), where T is the total number of observations of the behavior sequence in the behavior vector. The process of drawing the ACF graph and the PACF graph is: Define the autocorrelation coefficient as r, and the process of calculating the autocorrelation coefficient ACF is: Among them, r k represents the autocorrelation coefficient of lag k, Y t represents the observed value of the behavior sequence Y at time point t, represents the average value of the behavior sequence Y, Y t-k Represents the observed value of the behavior sequence Y at time point tk, based on the autocorrelation coefficient r k The values ​​are plotted statistically to obtain the autocorrelation function ACF graph; The process of calculating the partial autocorrelation coefficient PACF is: Based on the regression model formula: Y t =α+β1Y t-1 +β2Y t-2 +…+β k Y t-k +∈ t , where α is the intercept term of the model, β1, β2, β k is the autoregression coefficient, Y t-k is the lag value of the vector sequence, k is a constant, ∈ t is the error term at time point t, based on the observed value Y of the behavior sequence Y output by the regression model at time point t t Draw a statistical graph to obtain the partial autocorrelation function PACF graph; Based on the ACF and PACF graphs, the parameters p, d, and q of the autoregressive moving average model ARIMA model are determined, where p is the autoregressive term, d is the difference order, and q is the moving average term. The determination process of the parameters p, d, and q of the autoregressive moving average model ARIMA is: The stage where the bar graph in the PACF diagram is zero is the autoregressive term p, and the stage where the bar graph in the ACF diagram becomes zero is the moving average term q. The ACF diagram shows that the data becomes stable after d times of difference, so the difference order is d; Based on the parameters p, d, and q, the autoregressive moving average model ARIMA model is fitted. The autoregressive moving average model ARIMA model is expressed as: φ(B)(1-B) d yt=θ(B)∈ t Among them, φ(B) represents the autoregressive polynomial, θ(B) represents the moving average polynomial, B represents the backshift operator, and the acquisition process of the autoregressive polynomial is: φ(B) = 1-φ1B-φ2B2-...-φ p B q , the process of obtaining the moving average polynomial is: θ(B)=1+θ1B+θ2B2+...+θ q B q , the process of obtaining the backshift operator is: B Yt =Y t-1 , input the behavior vector into the autoregressive moving average model ARIMA model, output the seasonal and periodic sequence data appearing in the behavior sequence of the vector, obtain the behavior pattern, output the sequence data of the long-term trend of the behavior sequence over time, and obtain the behavior trend.

6. A BIM and GIS construction data transmission method according to claim 4, characterized in that: The baseline model establishment process is as follows: Convolution calculation is performed on the sequence data in the behavior pattern and behavior trend. The convolution calculation process is: The convolutional neural network CNN one-dimensional convolution layer, pooling layer and fully connected layer are used to process the sequence data in the behavior pattern and behavior trend. The output sequence of the one-dimensional convolution layer is defined as S. The sequence data is input into the one-dimensional convolution layer. The one-dimensional convolution layer slides on the time series through the filter and performs convolution operation on the sequence data in each window: Among them, S(t) is the output value of the convolution operation at time point t, w i is the weight of the filter at position i, X(ti) is the value of the input sequence data at time point (ti), b is the bias term, k-1 is the upper limit index of the filter window, and the output sequence S is obtained; The output sequence S is subjected to the maximum pooling value output feature operation through the pooling layer, and the maximum value is selected from each pooling window as the representative of the window. The pooling window size is defined as p, and the maximum pooling process is: Z(t)=max{S(t),S(t+1),…,S(t+p-1)}, where Z(t) is the output feature of the pooling layer at time point t, and the maximum pooling value feature Z of the output sequence S after the pooling layer is obtained; The output sequence S is averaged and pooled to calculate the average value in the pooling window at each time t as the output feature: Among them, M(t) is the output feature of the average layer at time point t, It means the sum of the values ​​of p consecutive time points starting from time point t and ending at t+p-1, and the average value feature M of the output sequence S after the averaging layer is obtained; The fully connected layer takes the maximum pooling value feature Z and the average value feature M of the average layer as input, and defines the maximum number of neurons in the fully connected layer, convolutional layer and pooling layer to be j, and the output of the fully connected layer is H: Among them, H j is the output of the jth neuron in the fully connected layer, v ij is the weight between the input feature at the ith position and the jth neuron, z i is the output of the neuron at position i, c j is the bias term of the jth neuron, σ is the activation function, and the output prediction value H of the fully connected layer is obtained; The fully connected layer output prediction value H is combined with the ARIMA model to obtain the baseline model, which is expressed as: (1-φ1B-φ2B2-...-φ p B q )(1-B) d Y t =(1+θ1B+θ2B2+...+θ q B q )∈ t +βH, β is the coefficient of the predicted value H output by the fully connected layer.

7. The BIM and GIS construction data transmission according to claim 1, characterized in that: The process of obtaining the urgency score is as follows: Input the behavior vector Y corresponding to the data item into the baseline model to obtain the predicted value H Y , the predicted value H Y is the urgency score. Define the transmission threshold V. When H Y > V, perform transmission caching on the data item corresponding to H Y . When H Y < V, do not transmit the data item corresponding to H Y .

8. A BIM and GIS construction data transmission system, used for a BIM+GIS-based construction data transmission method according to any one of claims 1 to 7, comprising a BIM+GIS data integration module for acquiring construction data and process status indicators, characterized in that: Modules for construction data transmission include: Data preprocessing module: obtains construction data items in the process of construction data, and uniformly outputs the construction data items as behavior vectors; Data planning module: Use the autoregressive moving average model ARIMA to predict the behavior vector of the construction data item, and obtain the behavior pattern and behavior trend based on the prediction results; Data urgency allocation module: Use behavior vectors, behavior patterns, and behavior trends to establish a baseline model, use the behavior vectors as input, and obtain urgency scores based on the baseline model; Data transmission strategy module: Based on the urgency score, a transmission threshold is set. When the urgency score is less than the transmission threshold, the construction data item is not transmitted. When the urgency score is greater than the transmission threshold, the construction data item is cached for transmission.

Images