A horizontal brick laying method based on a decision tree model

Abstract

The application discloses a horizontal brick arrangement method based on a decision tree model, and comprises the following steps: (1) inputting brick arrangement parameters; (2) generating a horizontal brick arrangement test scheme; (3) constructing a horizontal arrangement decision tree model; and (4) selecting a horizontal brick arrangement scheme and outputting a brick arrangement result. The application can generate multiple brick arrangement schemes through the construction of a brick arrangement model, and can obtain an optimal brick arrangement scheme by using an automatic evaluation method, so that a brick arrangement scheme meeting the specification requirements, having low material loss and being beautiful can be automatically formed.

Description

Technical Field

[0001] This invention belongs to the field of building information modeling, and specifically relates to a horizontal bricklaying method based on a decision tree model, which is particularly suitable for bricklaying in secondary structures. Background Technology

[0002] As a major component of building construction, secondary structures have numerous detailed nodes. In particular, when building walls, it is necessary to satisfy both aesthetic requirements and brick-saving considerations. The bricklaying patterns are varied, and the accuracy of bricklaying is required, making the arrangement difficult.

[0003] Bricklaying is a crucial aspect of secondary structural construction in buildings. A well-arranged brickwork combination can enhance the aesthetics of the facade and reduce construction waste. Bricklaying includes both horizontal and vertical aspects. Traditionally, horizontal bricklaying requires workers to conduct trial runs and then adjust based on experience; this method is inefficient, and the final result depends heavily on the worker's experience.

[0004] Currently, there are BIM-based secondary structure bricklaying tools that automatically generate bricklaying models, but these models suffer from problems such as uneven joints and non-compliant brick selection, resulting in the generated bricklaying schemes being unable to guide on-site masonry.

[0005] To address the aforementioned issues, no publicly available solution has been found among the various plugins tested so far. Therefore, a two-structure horizontal bricklaying method is needed to analyze the merits of different bricklaying schemes and formulate a preferred two-structure horizontal bricklaying method. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art. By constructing a brick layout model, iteratively generating multiple brick layout schemes, and then using an automatic evaluation method to obtain the preferred brick layout scheme, a brick layout scheme that meets the specifications, has low material loss, and is aesthetically pleasing is automatically formed.

[0007] To achieve the above-mentioned objectives, this invention provides a horizontal bricklaying method based on a decision tree model, comprising the following steps: (1) inputting bricklaying parameters; (2) generating a horizontal bricklaying test scheme; (3) constructing a horizontal layout decision tree model; and (4) selecting a horizontal bricklaying scheme and outputting the bricklaying result.

[0008] Furthermore, the brick arrangement parameters mentioned in step (1) include the standard brick length s, the mortar joint width t, the toothed joint advance / retreat value d, and the wall length L to be bricked.

[0009] Furthermore, when the toothed joint moves forward and backward, the length L of the wall to be laid with bricks is the length of the bottom layer of bricks.

[0010] Further, step (2) includes the following steps: (21) Divide the wall length L to be laid by the sum of the length of a single brick and the length of a single mortar joint to obtain the number of bricks n and the remainder interval m; (22) Construct a horizontal brick laying model: Based on the input parameters in step (1), establish a two-dimensional array a[2][2n+1]={{x1,t,x2,t,…,x n-1 ,t,x n ,t,m1},{y1,t,y2,t,…,y n-1 ,t,y n ,t,m2}}, define the first row as odd-numbered bricks and the second row as even-numbered bricks, and initialize the array according to the input parameters in step (1), limiting the sum of the elements in each row to the wall length L to be laid out; (23) generate multiple brick laying schemes by adjusting the brick size: change the brick a ij The value of is used to iteratively generate multiple brick-laying test schemes; (24) the multiple brick-laying test schemes are stored in the database.

[0011] Furthermore, during the iterative generation process described in step (23), the odd-numbered bricks x1, x2, and x3 are adjusted respectively. n-1 ,x n and even-numbered bricks y1, y2, y n-1 ,y n The values ​​of 6 out of the 8 blocks are determined by subtracting the total length L from the values ​​of the remaining 2 blocks.

[0012] Furthermore, a whole brick can be split into two non-whole bricks, or the remainder can be added to the last brick m1, such as m1 > a. ij If so, a non-whole brick is added to the horizontal brick arrangement model.

[0013] Further, step (3) includes the following steps: (31) Collect publicly available atlases, excellent bricklaying cases, and typical bricklaying problems in the industry, and select the following feature attribute data based on engineering experience: boundary recognition, odd / even number of bricks laid, joint position, and edge brick size; statistically analyze the feature attribute data, label the degree of excellence of the scheme as excellent, average, and poor, integrate and establish a model training dataset, and divide it into 70% as the training set and 30% as the test set; (32) Construct a decision tree: construct a horizontal layout decision tree class = DecisionTree(DB) from the given training set DB; (33) Train the horizontal layout decision tree. Decision tree: Filter all possible decision conditions and select the decision condition that maximizes the information gain of the child nodes; repeat the above steps until each child node belongs to the same type of brick arrangement scheme; (34) Decision tree pruning optimization: use the test set described in step (31) to test the training model in step (33). If the accuracy of the detection result is greater than or equal to 95%, it meets the actual application requirements, and the model training and testing are completed; if the accuracy of the detection result is less than 95%, it does not meet the actual application requirements. Iterative optimization is carried out through at least one of the following methods: dataset expansion, data augmentation, and adjustment of the maximum leaf node limit parameter, until the actual application requirements are met.

[0014] Further, step (32) includes the following steps: (321) using the C4.5 algorithm to calculate the information gain rate of each feature after its introduction, starting from the root node, based on the feature attributes established in step (31), and selecting the splitting node of the feature decision tree that brings the maximum information gain rate to the dataset. (322) Classify the continuous features such as joint location, broken brick size, and edge non-whole brick size, divide the joint location into different intervals, and compare the information gain of each split point.

[0015] Further, step (4) includes the following steps: (41) Calculate feature value: input the brick laying test scheme in step (24) and calculate the feature value according to the feature attribute; (42) Classification decision: divide the classification attribute of the decision tree layer by layer until the leaf node, and obtain the attribute classification result y = DecisionTree(x);

[0016] (43) Filter out the brick arrangement schemes with the label "excellent"; (44) Select the array with the largest number of bricks n from the obtained preferred feasible schemes as the horizontal brick arrangement scheme.

[0017] Furthermore, the feature attribute described in step (41) includes the seam position (a 21 -d) / a 11 , Dimensions of characteristic edge bricks.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0019] By constructing a mathematical model for brick layout, multiple brick layout schemes are generated iteratively. Then, an automatic evaluation method is used to obtain the optimal brick layout scheme, and finally, a brick layout scheme that meets the specifications, has low material waste, and is aesthetically pleasing is automatically formed. Attached Figure Description

[0020] Figure 1 This is a flowchart of one embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the brick wall to be laid in one embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of a decision tree in one embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of the brick arrangement result in one embodiment of the present invention. Detailed Implementation

[0024] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0025] like Figures 1-4 As shown, an embodiment of the horizontal brick-laying method based on the decision tree model of the present invention includes the following steps:

[0026] Step 1: Inputting the bricklaying parameters for the second structure

[0027] Input the standard brick length s (i.e.) Figure 2 and Figure 4 S), mortar joint width t, toothed joint advance / retreat value d. Wall length L to be laid (Note: When there is toothed joint advance / retreat, the wall length for odd and even numbered brick courses is different, L is the length of the bottom brick course).

[0028] Step 2: Generate a horizontal bricklaying test plan

[0029] 2.1 Divide the bricklaying length L by the sum of the brick length and mortar joint length to obtain the remainder interval m, and the theoretical maximum number of bricks to be laid out n. Figure 2 Taking walls as an example

[0030]

[0031] 2.2 Construct a horizontal brick-laying model. Based on the input parameters mentioned above, create a 2*2n+1 two-dimensional array to represent odd-numbered and even-numbered bricks. Define the first row as odd-numbered bricks and the second row as even-numbered bricks, and initialize the array according to the input parameters in step 1. The sum of each row is limited to the brick-laying length.

[0032] a[2][2n+1]={{x1,t,x2,t,…,x n-1 ,t,xn ,t,m1},{y1,t,y2,t,…,y n-1 ,t,y n ,t,m2}}

[0033] 2.3 Generate a horizontal bricklaying test plan

[0034] By adjusting the brick size, various brick layout schemes can be generated, changing brick a ij The values ​​are selected and iteratively generated to produce multiple brick layout schemes.

[0035] Generation principle: The position of the joints at both ends is affected by the two bricks in the upper and lower layers, therefore, the odd-numbered layers of bricks x1, x2, and x3 are adjusted respectively. n-1 ,x n , even-numbered bricks y1,y2,y n-1 ,y n Six of the values ​​are taken, and the remaining two are obtained by subtracting the total length.

[0036] a[2][2n+1]={{x1,t,x2,t,…,x n-1 ,t,x n ,t,m1},{y1,t,y2,t,…,y n-1 ,t,y n ,t,m2}}

[0037] by Figure 2 For example,

[0038] a={{600,10.600.10,600,10,600,10,m1},{600,10.600.10,600,10,600,10,600,10,m2}}

[0039] A whole brick can be split into two non-whole bricks or the remainder can be added to the tail brick m1. If m1 > aij, then a broken brick will be added to the brick arrangement model.

[0040] 2.4 Store the generated test plans into the database.

[0041] by Figure 2 For example,

[0042] a1={{600,10.600.10,…600,10,m1},{600,10.600.10,…600,10,m2}}

[0043] a2={{600,10.600.10,…600,10,m1},[400,10.600.10,…600,10,m2+200-s}}

[0044] a3={{600,10.600.10,…600,10,m1},{300,10.600.10,…600,10,m2+300-s}}

[0045] a4={{600,10.600.10,…600,10,m1},{200,10.600.10,…600,10,m2+400-s}}

[0046] a5={{400,10.600.10,…600,10,m1,10,200},{600,10.600.10,…600,10,m2}}

[0047] a6={{400,10.600.10,…600,10,m1+200-s},{400,10.600.10,…600,10,m2+200-s}}

[0048] a7={{400,10.600.10,…600,10,m1+200-s},{300,10.600.10,…600,10,m2+300-s}}

[0049] a8={{400,10.600.10,…600,10,m1+200-s},{200,10.600.10,…600,10,m2+400-s}}

[0050] a9={{300,10.600.10,…600,10,m1+300-s},{200,10.600.10,…600,10,m2+400-s}}

[0051] a10={{100,10.600.10,…600,10,m1+500-s},{600,10.600.10,…600,10,m2}}

[0052] a11={{100,10.600.10,…600,10,m1+500-s},{400,10.600.10,…600,10,m2+200-s}}

[0053] ...

[0054] Step 3: Construct a horizontal layout decision model

[0055] 3.1 Establishing the training dataset

[0056] Collect publicly available industry diagrams, excellent bricklaying case studies, and typical bricklaying problems. Based on engineering experience, select the following feature attributes: 1. Boundary recognition; 2. Odd / even number of brick layers laid with whole bricks; 3. Joint position; 4. Edge brick dimensions. Statistically analyze these attribute data. Label the schemes according to their implementation: Excellent, Average, Poor. Merge and build a model training dataset, dividing it into a 70% training set and a 30% test set.

[0057] 3.2 Constructing a decision tree

[0058] Construct a decision tree from the given training set DB.

[0059] class = DecisionTree(DB)

[0060] 3.2.1 Using the C4.5 algorithm, starting from the root node, calculate the information gain rate of each feature after its introduction, and select the splitting node of the decision tree that brings the maximum information gain rate to the dataset.

[0061]

[0062]

[0063] like Figure 3 As shown, the root node of the wall represents the number of edge-identifying toothed joints, and the child nodes contain the initial brick arrangement and the dimensions of the edge bricks.

[0064] 3.2.2 The continuous characteristics such as joint location, broken brick size, and edge non-whole brick size are classified and processed. The joint location is divided into different intervals, and the information gain of each split point is compared.

[0065] 3.3 Training Level Layout Decision Tree

[0066] Filter all possible decision conditions and select the one that maximizes the information gain of the child nodes. Repeat the above steps until a bricklaying scheme in which each child node belongs to the same label class is found.

[0067] Set up the training environment using Python 3.6 and Sklearn, use the entropy metric, select the best feature splitting metric, set max_leaf__nodes=100, and then execute model training.

[0068] 3.4 Decision Tree Pruning

[0069] The model trained in the previous step is tested using the test set obtained in step 3.1. If the accuracy of the detection result is greater than or equal to 95%, it meets the requirements of the actual application, and the model training and testing are completed. If the accuracy of the detection result is less than 95%, it does not meet the requirements of the actual application. Iterative optimization is carried out through methods such as dataset expansion, data augmentation, and adjusting the parameter that limits the maximum number of leaf nodes until the requirements of the actual application are met.

[0070] Step 4: Select a horizontal bricklaying scheme

[0071] 4.1 Calculate eigenvalues

[0072] Input the brick-laying test scheme generated in 2.4 and calculate the feature values ​​based on the feature attributes.

[0073] like Figure 2 Seam position:

[0074] 4.2 Classification Decision Stage

[0075] The decision tree is divided layer by layer according to the classification attributes until the leaf node is reached to obtain the attribute classification result.

[0076] y = DecisionTree(x)

[0077] 4.3 Based on the evaluation results in 3.4, filter out the brick layout schemes with the "excellent" label.

[0078] Taking test scheme 2.4 as an example, filter out a2, a3, a7, a9...

[0079] 4.4 such as Figure 4 As shown, select the brick array with the largest number of whole bricks (n) from the obtained preferred feasible schemes, and output the horizontal brick arrangement scheme.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A horizontal brick-laying method based on a decision tree model, characterized in that, Includes the following steps: (1) Input the brick layout parameters, which include the standard brick length s, mortar joint width t, toothed joint advance / retreat value d, and wall length L to be laid; (2) Generate a horizontal bricklaying test plan: (21) Divide the length L of the wall to be laid by the sum of the length of a single brick and the length of a single mortar joint to obtain the number of bricks to be laid n and the remainder interval m. (22) Constructing a horizontal bricklaying model: Establish a two-dimensional array based on the input parameters in step (1). Define the first row as odd-numbered bricks and the second row as even-numbered bricks, and initialize the array according to the input parameters in step (1), limiting the sum of the elements in each row to the length L of the wall to be laid. (23) Generate multiple brick layout schemes by adjusting the brick size: change the brick Values ​​are selected and iteratively generated to produce multiple brick-laying test schemes. (24) Store multiple brick-laying test schemes in the database; (3) Construct a horizontal arrangement decision tree model: (31) Collect publicly available atlases, excellent bricklaying cases, and typical bricklaying problems in the industry. Based on engineering experience, select the following characteristic attribute data: boundary recognition, odd / even number of brick layers laid with whole bricks, joint position, and edge brick size. Statistically analyze the aforementioned characteristic attribute data, label the scheme as excellent, average, or poor, integrate them to establish a model training dataset, and divide it into a training set of 70% and a test set of 30%. (32) Constructing a decision tree: Construct a horizontally arranged decision tree from the given training set DB. ， (33) Training the level of the decision tree: Filter all possible decision conditions and select the decision condition that maximizes the information gain of the child nodes; repeat the above steps until each child node belongs to the same type of label in the brick arrangement scheme. (34) Decision tree pruning optimization: The training model in step (33) is tested using the test set described in step (31). If the accuracy of the detection result is greater than or equal to 95%, it meets the actual application requirements, and the model training and testing are completed. If the accuracy of the detection result is less than 95%, it does not meet the actual application requirements. Iterative optimization is carried out through at least one of the following methods: dataset expansion, data augmentation, and adjustment of the maximum leaf node limit parameter, until the actual application requirements are met. (4) Select the horizontal brick layout scheme and output the brick layout result.

2. The horizontal brick-laying method based on a decision tree model according to claim 1, characterized in that, When the toothed joint moves forward and backward, the length L of the wall where bricks need to be laid is the length of the bottom layer of bricks.

3. The horizontal bricklaying method based on a decision tree model according to claim 1, characterized in that, During the iterative generation process described in step (23), the odd number of tiles are adjusted respectively. and even number of bricks The values ​​of 6 out of the 8 blocks are determined by subtracting the total length L from the values ​​of the remaining 2 blocks.

4. The horizontal bricklaying method based on a decision tree model according to claim 3, characterized in that, One whole brick can be split into two non-whole bricks, or the remainder can be added to the last brick m1, such as m1 > If so, a non-whole brick is added to the horizontal brick arrangement model.

5. The horizontal bricklaying method based on a decision tree model according to claim 1, characterized in that, Step (32) includes the following steps: (321) Using the C4.5 algorithm, starting from the root node, calculate the information gain ratio of each feature after its introduction, and select the splitting node of the decision tree that brings the maximum information gain ratio to the dataset. , ; (322) Classify the continuous features such as joint position, broken brick size, and edge non-whole brick size, divide the joint position into different intervals, and compare the information gain of each split point.

6. The horizontal bricklaying method based on a decision tree model according to claim 1, characterized in that, Step (4) includes the following steps: (41) Calculate feature values: Input the brick arrangement test plan in step (24) and calculate the feature values ​​based on the feature attributes; (42) Classification decision: Divide the decision tree layer by layer according to the classification attributes until the leaf node to obtain the attribute classification result. ; (43) Filter out the brick arrangement schemes with the best label; (44) Select the array with the largest number of bricks n from the obtained preferred feasible schemes as the horizontal brick arrangement scheme.

7. The horizontal bricklaying method based on a decision tree model according to claim 6, characterized in that, The feature attributes mentioned in step (41) include the seam position. , Dimensions of characteristic edge bricks.

Images