Intelligent construction process monitoring method and system based on BIM

Abstract

The invention discloses a BIM-based intelligent construction process monitoring method and system, and belongs to the technical field of intelligent construction, and the method comprises the steps: constructing a BIM corresponding to a subway station conversion section underground excavation construction site, carrying out the finite element analysis of the subway station conversion section underground excavation construction site based on the BIM, and determining a stress concentration region, the method comprises the following steps: acquiring a stress concentration area, acquiring safety influence factors corresponding to the stress concentration area, identifying the safety influence factors by adopting a pre-deployed machine learning model, determining support structure optimization data, and finally transmitting the support structure optimization data to field staff. According to the method, the support structure parameters can be dynamically optimized, the method can adapt to the surrounding rock stress redistribution process, and the construction safety is effectively guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent construction, and particularly relates to a BIM-based intelligent construction process monitoring method and system. Background Art

[0002] The mined construction of the transfer section of a subway station is a construction method that involves excavation operations underground, mainly used for the construction of the transfer section of a subway station. Since this construction method requires excavation underground, a series of safety measures need to be taken to ensure the safety of construction workers and the surrounding environment. Mined construction usually uses a shield machine for excavation, and the shield machine can perform precise excavation and propulsion underground. During the excavation process, it is necessary to continuously monitor the geological conditions and the groundwater level to ensure construction safety. At the same time, a series of support measures, such as shotcrete spraying and steel frame installation, need to be taken to prevent soil collapse. The advantage of mined construction is that it can reduce the impact on surface traffic and the surrounding environment, and at the same time improve construction efficiency. However, mined construction also has certain risks, such as changes in the groundwater level and complex geological conditions. Therefore, strict safety measures and monitoring measures need to be taken to ensure construction safety. In short, the mined construction of the transfer section of a subway station is a complex underground engineering construction method that requires strict safety measures and monitoring measures to ensure construction safety. Existing technologies mostly set support parameters (such as bolt length and steel frame spacing) based on experience or static geological reports, and do not adjust them in real time according to excavation disturbances. For example, during the construction of bolts, the pre-tightening force of the bolts is fixed and cannot adapt to the stress redistribution process of the surrounding rock, resulting in the failure of bolts in local stress concentration areas (such as breakage or slippage), and the inability to achieve a stiffness gradient transition in the bedding bias pressure formation, exacerbating the stress mutation at the connection section of the cross passage. Summary of the Invention

[0003] The present invention provides a BIM-based intelligent construction process monitoring method and system to solve the problem of low safety caused by setting support parameters based on experience or static geological reports in the prior art.

[0004] On the one hand, the present invention provides a BIM-based intelligent construction process monitoring method, including: Construct a BIM corresponding to the mined construction site of the transfer section of the subway station, and perform finite element analysis on the mined construction site of the transfer section of the subway station based on the BIM to determine the stress concentration area; Collect the safety influencing factors corresponding to the stress concentration area, and use a pre-deployed machine learning model to identify the safety influencing factors to determine the support structure optimization data; Transmit the support structure optimization data to the on-site staff so that the on-site staff can perform construction according to the support structure optimization data to complete the BIM-based intelligent construction process monitoring.

[0005] In a possible implementation, a BIM corresponding to the cut-and-cover construction site of the subway station transition section is constructed, and a finite element analysis is performed on the cut-and-cover construction site of the subway station transition section based on the BIM to determine the stress concentration area, including: Construct a BIM corresponding to the cut-and-cover construction site of the subway station transition section, and perform a finite element analysis on the construction site based on the BIM to obtain the stress data corresponding to the cut-and-cover construction site of the subway station transition section; The BIM corresponding to the cut-and-cover construction site of the subway station transition section is evenly divided into multiple different areas, and according to the stress data corresponding to the cut-and-cover construction site of the subway station transition section, the area where the stress data exceeding the preset stress threshold is greater than the preset percentage is determined as the stress concentration area.

[0006] In a possible implementation, collect the safety influencing factors corresponding to the stress concentration area, including: Divide the stress concentration area into several sub-stress concentration areas, and for each sub-stress concentration area, use the stress, strain, displacement, and lithology corresponding to the center of the sub-stress concentration area to obtain the safety influencing factors corresponding to the sub-stress concentration area; The safety influencing factors corresponding to all sub-stress concentration areas together constitute the safety influencing factors corresponding to the stress concentration area.

[0007] In a possible implementation, use a pre-deployed machine learning model to identify the safety influencing factors and determine the support structure optimization data, including: The safety influencing factors corresponding to the stress concentration area are composed into a data matrix, and the data matrix is input into a pre-deployed machine learning model to obtain the support structure optimization data.

[0008] In a possible implementation, the method for pre-deploying the machine learning model includes: Construct a machine learning model with the ability to identify data matrices, and perform initialization encoding on the hyperparameters of the machine learning model to obtain multiple hyperparameter encodings; Obtain the pre-stored historical safety influencing factors and the support structure optimization data labels corresponding to the historical safety influencing factors, and according to the historical safety influencing factors and the support structure optimization data labels corresponding to the historical safety influencing factors, obtain the fitness corresponding to each hyperparameter encoding; According to the fitness corresponding to the hyperparameter encoding, determine the hyperparameter encoding with the maximum fitness as the optimal encoding; Adopt a random chain search strategy to perform a joint chain search on the hyperparameter encodings to obtain the hyperparameter encodings after the joint chain search; Based on the optimal coding, a probability decision-based equilibrium search strategy is used to perform global and local equilibrium searches on the hyperparameter coding after the joint chain search, and the hyperparameter coding after the equilibrium search is obtained; An adaptive random chaotic search strategy is used to perform a global search on the hyperparameter coding after the equilibrium search, and the hyperparameter coding after the global search is obtained; The random chain search strategy, the probability decision-based equilibrium search strategy, and the adaptive random chaotic search strategy are repeatedly executed until the training end condition is satisfied, and the optimal coding is re-obtained; Based on the re-obtained optimal coding, the machine learning model is deployed.

[0009] In a possible implementation manner, the hyperparameters of the machine learning model are initialized and coded, and multiple hyperparameter codings are obtained, including: Random initialization is performed between the upper and lower limits corresponding to the hyperparameters of the machine learning model, and the initialized hyperparameters are coded as vectors to obtain the initial hyperparameter coding; Based on the initial hyperparameter coding, multiple other hyperparameter codings are obtained as:

[0010] Among them, represents the k th hyperparameter coding, and when k = 1, is the initial hyperparameter coding, represents the k +1 th hyperparameter coding, represents the remainder function, represents pi, sin represents the sine function, represents the first constant term and is set to 0.5; b represents the second constant term and is set to 0.2.

[0011] In a possible implementation manner, a random chain search strategy is used to perform a joint chain search on the hyperparameter coding, and the hyperparameter coding after the joint chain search is obtained, including: For any hyperparameter coding, a random other hyperparameter coding is matched for the hyperparameter coding to obtain the joint search coding corresponding to each hyperparameter coding; According to the fitness corresponding to all hyperparameter codings, the hyperparameter codings are arranged in descending order of fitness to obtain the arranged hyperparameter codings; For the arranged hyperparameter codings, according to the joint search coding corresponding to the hyperparameter coding, a joint chain search is performed on the hyperparameter coding, and the hyperparameter coding after the joint chain search is obtained as:

[0012] Among them, represents the t th training, the i th sorted hyperparameter encoding's d -dimensional hyperparameter, i = 1, 2,.., M, where M represents the total number of hyperparameter encodings, d = 1, 2,.., D, where D represents the total dimension of hyperparameters in the hyperparameter encoding, represents the i th hyperparameter encoding's d -dimensional hyperparameter after joint chain search, sin represents the sine function, cos represents the cosine function, represents pi, represents the first random number between (0, 1), represents the second random number between (0, 1), represents the i th hyperparameter encoding's corresponding joint search encoding's d -dimensional hyperparameter.

[0013] In a possible implementation manner, based on the optimal encoding, a global and local balanced search is performed on the hyperparameter encoding after joint chain search by using a probability decision balanced search strategy to obtain the hyperparameter encoding after balanced search, including: Based on the current training times, obtain the adaptive decision probability as:

[0014]

[0015] Among them, represents the adaptive decision probability, represents the preset probability basic value, represents the Weibull distribution control factor, W represents the Weibull distribution function, T represents the preset maximum training times, t represents the current training times; represents the shape parameter of the Weibull distribution, represents the scale parameter of the Weibull distribution, e represents the natural constant; Generate a third random number between (0, 1), and determine whether the third random number is less than the adaptive decision probability. If so, perform a Lévy flight search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after balanced search. Otherwise, based on the optimal encoding, perform an optimal direction search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after balanced search; Perform Levy flight search on the hyperparameter encoding after joint chained search, and obtain the hyperparameter encoding after balanced search as:

[0016]

[0017]

[0018]

[0019]

[0020] Among them, represents the t -th dimension hyperparameter of the hyperparameter encoding after the j -th joint chained search during the d -th training, represents the j -th dimension hyperparameter of the hyperparameter encoding after balanced search, d j = 1, 2,.., M, where M represents the total number of hyperparameter encodings, represents the t +1-th element in the j -th dimension of the update speed corresponding to the hyperparameter encoding after the d -th joint chained search during the represents the first inertia weight, represents the t -th element in the j -th dimension of the update speed corresponding to the hyperparameter encoding after the d -th joint chained search during the represents the first learning factor, represents the second learning factor, represents the search range control factor set between (1, 3), represents the j -th hyperparameter encoding after the joint chained search and the Euclidean distance between the first random hyperparameter encoding , represents the j -th hyperparameter encoding after the joint chained search and the Euclidean distance between the second random hyperparameter encoding , represents the first Levy flight factor between (0, 1), represents the second Levy flight factor between (0, 1); ​Based on the optimal coding, perform an optimal direction search on the hyperparameter coding after the joint chain search to obtain the hyperparameter coding after the balanced search as follows:

[0021]

[0022] Among them, represents the d -dimensional hyperparameter of the optimal coding.

[0023] In a possible implementation manner, adopt an adaptive random chaos search strategy to perform a global search on the hyperparameter coding after the balanced search to obtain the hyperparameter coding after the global search, including: Obtain the chaos search control factor as:

[0024] Among them, represents the chaos search control factor at the t -th training, and at the initial moment, the chaos search control factor is randomly generated between (0, 1), represents the chaos search control factor at the t +1-th training, represents the remainder function, represents the pi; According to the chaos search control factor, perform a global search on the hyperparameter coding after the balanced search to obtain the hyperparameter coding after the global search as:

[0025]

[0026]

[0027]

[0028] Among them, represents the t -th training, the n -th d -dimensional hyperparameter of the n -th hyperparameter coding after the balanced search, = 1, 2,.., M, M represents the total number of hyperparameter codings, n represents the d -dimensional hyperparameter of the -th hyperparameter coding after the global search, t +1-th training, the n -th ddimensional element, indicating the t th training, the n dimensional element in the update speed corresponding to the hyperparameter encoding after the d th equilibrium search, indicating the second inertia weight, indicating the third learning factor, indicating the d dimensional hyperparameter of the newly randomly generated hyperparameter encoding, indicating the fourth random number uniformly distributed between (-1, 1), indicating the preset maximum value of the second inertia weight, indicating the preset minimum value of the second inertia weight, indicating the preset maximum value of the third learning factor, indicating the preset minimum value of the third learning factor, indicating the fifth random number uniformly distributed between (0, 1).

[0029] On the other hand, the present invention provides an intelligent construction process monitoring system based on BIM, including: a BIM data analysis module, a machine learning module, and a data feedback module; The BIM data analysis module is used to construct a BIM corresponding to the cut-and-cover construction site of the subway station transition section, and perform finite element analysis on the cut-and-cover construction site of the subway station transition section based on the BIM to determine the stress concentration area; The machine learning module is used to collect the safety influencing factors corresponding to the stress concentration area, and identify the safety influencing factors by using a pre-deployed machine learning model to determine the support structure optimization data; The data feedback module is used to transmit the support structure optimization data to the on-site staff, so that the on-site staff can perform construction according to the support structure optimization data to complete the intelligent construction process monitoring based on BIM.

[0030] An intelligent construction process monitoring method and system provided by the present invention, by constructing a BIM corresponding to the cut-and-cover construction site of the subway station transition section, performing finite element analysis on the cut-and-cover construction site of the subway station transition section based on the BIM to determine the stress concentration area, then collecting the safety influencing factors corresponding to the stress concentration area, and identifying the safety influencing factors by using a pre-deployed machine learning model to determine the support structure optimization data, and finally transmitting the support structure optimization data to the on-site staff, so that the on-site staff can perform construction according to the support structure optimization data, realizes the dynamic optimization of the support structure parameters, can adapt to the surrounding rock stress redistribution process, and effectively ensures the construction safety. Description of the Drawings

[0031] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0032] Figure 1 It is a flowchart of an intelligent construction process monitoring method based on BIM provided for an embodiment of the present invention.

[0033] Figure 2 It is a schematic structural diagram of an intelligent construction process monitoring system based on BIM provided for an embodiment of the present invention.

[0034] Among them, 201 - BIM data analysis module, 202 - machine learning module, 203 - data feedback module.

[0035] Through the above - mentioned accompanying drawings, specific embodiments of the present invention have been shown, and there will be more detailed descriptions hereinafter. These drawings and textual descriptions are not intended to limit the scope of the inventive concept in any way, but to illustrate the concept of the present invention to those skilled in the art by referring to specific embodiments. Specific Embodiments

[0036] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0037] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] As Figure 1 shown, the embodiments of the present invention provide an intelligent construction process monitoring method based on BIM, including: S101. Construct a BIM (Building Information Modeling) corresponding to the cut - and - cover construction site of the subway station transition section, and perform finite - element analysis on the cut - and - cover construction site of the subway station transition section based on the BIM to determine the stress - concentration area; BIM is an overall working method based on intelligent models and digital technologies. By integrating data from multiple fields, it realizes information management throughout the entire life cycle of building assets. It not only includes the geometric shape, location information, physical properties, material information, etc. of the building, but also supports all participating parties to share data based on a unified model, improving collaboration efficiency. Therefore, the embodiment of the present invention constructs a BIM corresponding to the cut-and-cover construction site of the subway station transition section, which can achieve comprehensive monitoring of the cut-and-cover construction site of the subway station transition section. Then, based on the BIM, a finite element analysis is performed on the cut-and-cover construction site of the subway station transition section, and stress analysis can be realized, thereby determining the stress concentration area and providing data support for optimizing the support structure parameters.

[0039] S102. Collect the safety influencing factors corresponding to the stress concentration area, and use a pre-deployed machine learning model to identify the safety influencing factors to determine the support structure optimization data; The safety influencing factors can be parameters such as stress, lithology, strain, and displacement at the cut-and-cover construction site of the subway station transition section that affect safety. These parameters are very important for optimizing the support structure parameters. Therefore, a pre-deployed machine learning model can be used to identify the safety influencing factors, thereby determining the support structure optimization data. The support structure optimization data here refers to the optimized support structure parameters.

[0040] S103. Transmit the support structure optimization data to the on-site staff so that the on-site staff can construct according to the support structure optimization data to complete the intelligent construction process monitoring based on BIM.

[0041] For example, after transmitting the support structure optimization data to the on-site staff, the staff can optimize the support structure parameters on-site according to the support structure optimization data (such as optimizing the bolt length and / or the spacing of steel frames).

[0042] It should be noted that in order to improve the optimization accuracy, the support structure optimization data is a kind of parameter. For different types of parameters, the above technical solutions are all executed to achieve the optimization of individual parameters.

[0043] In a possible implementation manner, constructing a BIM corresponding to the cut-and-cover construction site of the subway station transition section and performing a finite element analysis on the cut-and-cover construction site of the subway station transition section based on the BIM to determine the stress concentration area includes: Construct a BIM corresponding to the cut-and-cover construction site of the subway station transition section, and perform a finite element analysis on the construction site based on the BIM to obtain the stress data corresponding to the cut-and-cover construction site of the subway station transition section; The BIM corresponding to the underground excavation construction site of the subway station conversion section is evenly divided into multiple different regions, and according to the stress data corresponding to the underground excavation construction site of the subway station conversion section, the region where the stress data exceeding the preset stress threshold is greater than the preset percentage is determined as the stress concentration region.

[0044] For the stress concentration region with relatively large stress, stronger support is required. Therefore, in the embodiment of the present invention, it is necessary to optimize the support structure parameters of the stress concentration region.

[0045] In a possible implementation manner, the safety influencing factors corresponding to the stress concentration region are collected, including: The stress concentration region is divided into several sub-stress concentration regions (which can be evenly divided), and for each sub-stress concentration region, the stress, strain, displacement, and lithology corresponding to the center of the sub-stress concentration region are used to obtain the safety influencing factors corresponding to the sub-stress concentration region; The safety influencing factors corresponding to all sub-stress concentration regions together constitute the safety influencing factors corresponding to the stress concentration region.

[0046] By dividing the sub-stress concentration regions, the acquisition of the data matrix can be realized, which is more conducive to subsequent data identification.

[0047] In a possible implementation manner, a pre-deployed machine learning model is used to identify the safety influencing factors to determine the support structure optimization data, including: The safety influencing factors corresponding to the stress concentration region are formed into a data matrix, and the data matrix is input into the pre-deployed machine learning model to obtain the support structure optimization data.

[0048] Optionally, the data matrix can also be normalized first, and then the data matrix is input into the pre-deployed machine learning model to reduce the data complexity and improve the data identification efficiency. However, it should be noted that if the data is to be input after normalization, the training data also needs to be normalized during the pre-deployment process of the machine learning model.

[0049] In a possible implementation manner, the method for pre-deploying the machine learning model includes: A1. Construct a machine learning model with the ability to identify data matrices, and perform initialization coding on the hyperparameters of the machine learning model to obtain multiple hyperparameter codings; For example, a convolutional neural network can be used to construct a machine learning model, and then the hyperparameters of the machine learning model are initialized and coded by the method of random initialization.

[0050] A2. Obtain the pre-stored historical safety influencing factors and the support structure optimization data labels corresponding to the historical safety influencing factors, and based on the historical safety influencing factors and the support structure optimization data labels corresponding to the historical safety influencing factors, obtain the fitness corresponding to each hyperparameter encoding; The support structure optimization data labels corresponding to the historical safety influencing factors can be labels preset by the staff after professional analysis. Then, after data learning, the staff do not need to perform complex and repetitive analysis processes, and can quickly adjust the support structure parameters to ensure construction safety.

[0051] After applying the hyperparameter encoding to the machine learning model, using the historical safety influencing factors as the input data and the support structure optimization data labels corresponding to the historical safety influencing factors as the expected output data, then obtain the loss function value corresponding to the hyperparameter encoding, and after taking the negative of the loss function value, obtain the fitness corresponding to the hyperparameter encoding.

[0052] A3. Determine the hyperparameter encoding with the maximum fitness as the optimal encoding according to the fitness corresponding to the hyperparameter encoding; A4. Perform a joint chain search on the hyperparameter encoding using a random chain search strategy to obtain the hyperparameter encoding after the joint chain search; A5. Based on the optimal encoding, perform a global and local balanced search on the hyperparameter encoding after the joint chain search using a probability decision balanced search strategy to obtain the hyperparameter encoding after the balanced search; A6. Perform a global search on the hyperparameter encoding after the balanced search using an adaptive random chaos search strategy to obtain the hyperparameter encoding after the global search; A7. Repeat the execution of the random chain search strategy, the probability decision balanced search strategy, and the adaptive random chaos search strategy until the training end condition is met, and re-obtain the optimal encoding; A8. Deploy the machine learning model according to the re-obtained optimal encoding, that is, use the hyperparameters included in the optimal encoding as the final hyperparameters of the machine learning model, and deploy the machine learning model.

[0053] Optionally, after executing the random chain search strategy, the probability decision balanced search strategy, and the adaptive random chaos search strategy, perform out-of-limit processing on the hyperparameters to ensure that the hyperparameters are always within their corresponding upper and lower limits.

[0054] In the process of training a machine learning model in the prior art, there are problems such as poor training effect and difficulty in finding the global optimal solution, resulting in the inability to accurately optimize the support structure parameters. Therefore, an embodiment of the present invention proposes an algorithm for optimizing a machine learning model to improve the training speed, training effect, and global training ability, so as to find the global optimal solution, improve the output accuracy of the machine learning model, achieve effective optimization of the support structure parameters, and improve construction safety.

[0055] In a possible implementation manner, initialize the encoding of the hyperparameters of the machine learning model to obtain multiple hyperparameter encodings, including: Perform random initialization between the upper and lower limits corresponding to the hyperparameters of the machine learning model, and encode the initialized hyperparameters into vectors to obtain the initial hyperparameter encoding; Based on the initial hyperparameter encoding, obtain multiple other hyperparameter encodings as:

[0056] Among them, represents the k th hyperparameter encoding, and when k = 1, is the initial hyperparameter encoding, represents the k +1 th hyperparameter encoding, represents the remainder function, represents pi, sin represents the sine function, represents the first constant term and is set to 0.5; b represents the second constant term and is set to 0.2.

[0057] The initialization encoding process provided by the embodiment of the present invention enables multiple hyperparameter encodings to be more evenly distributed in the high-dimensional solution space at the beginning stage of the algorithm, which not only helps to improve the training speed of the algorithm but also assists the algorithm in finding the global optimal solution.

[0058] In a possible implementation manner, adopt a random chain search strategy to perform a joint chain search on the hyperparameter encodings to obtain the hyperparameter encodings after the joint chain search, including: For any one hyperparameter encoding, randomly match an other hyperparameter encoding for the hyperparameter encoding to obtain the joint search encoding corresponding to each hyperparameter encoding; According to the fitness corresponding to all hyperparameter encodings, arrange the hyperparameter encodings in descending order of fitness to obtain the arranged hyperparameter encodings; For the arranged hyperparameter encodings, perform a joint chain search on the hyperparameter encodings according to the joint search encoding corresponding to the hyperparameter encoding to obtain the hyperparameter encodings after the joint chain search as:

[0059] Among them, represents the t th i dimensional hyperparameter of the d th sorted hyperparameter encoding during the i th training, where d = 1, 2,.., M, and M represents the total number of hyperparameter encodings, = 1, 2,.., D, and D represents the total dimension of hyperparameters in the hyperparameter encoding; i represents the d dimensional hyperparameter of the hyperparameter encoding after the th joint chain search, where sin represents the sine function and cos represents the cosine function; represents pi, represents the first random number between (0, 1), represents the i dimensional hyperparameter of the joint search encoding corresponding to the d th sorted hyperparameter encoding.

[0060] In the embodiment of the present invention, the random chain search strategy is adopted, which can enable all hyperparameter encodings to be searched in a chain in the solution space while using the sine-cosine route for search. This not only helps to jump out of the local optimal solution but also improves the search fineness in the later stage of the algorithm.

[0061] In a possible implementation manner, based on the optimal encoding, a probability decision-based balanced search strategy is used to perform global and local balanced searches on the hyperparameter encoding after the joint chain search, and the hyperparameter encoding after the balanced search is obtained, including: Based on the current training times, the adaptive decision probability is obtained as:

[0062]

[0063] Among them, represents the adaptive decision probability, represents the preset probability basic value, represents the Weibull distribution control factor, and W represents the Weibull distribution function; T represents the preset maximum training times, t represents the current training times; represents the shape parameter of the Weibull distribution, represents the scale parameter of the Weibull distribution, and e represents the natural constant; In the initial stage of the algorithm, When the value is relatively large, the algorithm mainly conducts global search, which enhances the global search ability and makes the hyperparameter encoding in the population closer to the optimal solution. As the number of iterations increases, the value gradually decreases, and the algorithm mainly conducts local search to implement refined search, improving the optimization accuracy. In the late stage of algorithm iteration, the value increases again, enabling the algorithm to jump out of the local optimum.

[0064] Generate a third random number between (0, 1), and determine whether the third random number is less than the adaptive decision probability. If so, conduct Levy flight search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after balanced search; otherwise, based on the optimal encoding, conduct optimal direction search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after balanced search; Conduct Levy flight search on the hyperparameter encoding after joint chain search, and the hyperparameter encoding obtained after balanced search is:

[0065]

[0066]

[0067]

[0068]

[0069] where, represents the t -th dimension hyperparameter of the j -th hyperparameter encoding after joint chain search during the d -th training, represents the j -th dimension hyperparameter of the d -th hyperparameter encoding after balanced search, j = 1, 2,.., M, where M represents the total number of hyperparameter encodings, represents the t -th element in the update speed corresponding to the j -th hyperparameter encoding after joint chain search during the d -th training, represents the first inertia weight, represents the t -th element in the update speed corresponding to the j -th hyperparameter encoding after joint chain search during the d -th training, represents the first learning factor, represents the second learning factor, represents the search range control factor set between (1, 3), represents the j hyperparameter encoding after the th combined chained search and the Euclidean distance between the first random hyperparameter encoding represents the j th combined chained search and the Euclidean distance between the hyperparameter encoding and the second random hyperparameter encoding ; represents the first Levy flight factor between (0, 1), represents the second Levy flight factor between (0, 1); Based on the optimal encoding, perform an optimal direction search on the hyperparameter encoding after the combined chained search to obtain the hyperparameter encoding after the balanced search as:

[0070]

[0071] wherein, represents the d -dimensional hyperparameter of the optimal encoding.

[0072] In the embodiment of the present invention, a probability decision-based balanced search strategy is adopted, which can perform searches in two different search methods. In the early stage of the algorithm, Levy flight search is mainly used for adaptive learning, and Levy flight is also used to provide a certain global search ability. In the later stage of the algorithm, optimal direction search is mainly used, which can make the algorithm converge quickly and improve the convergence accuracy.

[0073] In a possible implementation manner, an adaptive random chaos search strategy is used to perform a global search on the hyperparameter encoding after the balanced search to obtain the hyperparameter encoding after the global search, including: Obtain the chaos search control factor as:

[0074] wherein, represents the chaos search control factor at the t th training, and at the initial moment, the chaos search control factor is randomly generated between (0, 1), represents the chaos search control factor at the t +1th training, represents the remainder function, represents the pi; According to the chaos search control factor, perform a global search on the hyperparameter encoding after the balanced search to obtain the hyperparameter encoding after the global search as:

[0075]

[0076]

[0077]

[0078] Among them, represents the t -th training, the n -th hyperparameter encoding after the d -th dimensional hyperparameter, n = 1, 2,.., M, where M represents the total number of hyperparameter encodings, represents the n -th dimensional hyperparameter of the hyperparameter encoding after the d -th global search, represents the t +1-th training, the n -th dimensional element in the update speed corresponding to the hyperparameter encoding after the d -th balanced search, represents the t -th training, the n -th dimensional element in the update speed corresponding to the hyperparameter encoding after the d -th balanced search, represents the second inertia weight, represents the third learning factor, represents the d -th dimensional hyperparameter of the newly randomly generated hyperparameter encoding, represents the fourth random number uniformly distributed between (-1, 1), represents the preset maximum value of the second inertia weight, represents the preset minimum value of the second inertia weight, represents the preset maximum value of the third learning factor, represents the preset minimum value of the third learning factor, represents the fifth random number uniformly distributed between (0, 1).

[0079] The adaptive random chaos search strategy provided by the embodiments of the present invention enables the algorithm to be in a chaotic state during search, which is more conducive to jumping out of local optimal solutions and enhances the powerful global search ability. Moreover, adaptive search parameters are set, and the algorithm will not overly affect the convergence of the algorithm in the later stage.

[0080] Optionally, an annealing simulation algorithm or a greedy strategy can also be used to control the search process of the adaptive random chaos search strategy to improve the training speed of the algorithm.

[0081] The present invention provides an intelligent construction process monitoring method and system based on BIM. By constructing a BIM corresponding to the cut-and-cover construction site of the transfer section of a subway station and performing finite element analysis on the cut-and-cover construction site of the transfer section of the subway station based on the BIM to determine the stress concentration area, then collecting the safety influencing factors corresponding to the stress concentration area, and using a pre-deployed machine learning model to identify the safety influencing factors to determine the optimized data of the support structure, and finally transmitting the optimized data of the support structure to the on-site staff, so that the on-site staff can perform construction according to the optimized data of the support structure, realizing the dynamic optimization of the support structure parameters, being able to adapt to the process of surrounding rock stress redistribution, and effectively ensuring the construction safety.

[0082] As Figure 2 shown, the present invention provides an intelligent construction process monitoring system based on BIM, including: a BIM data analysis module 201, a machine learning module 202, and a data feedback module 203; The BIM data analysis module 201 is used to construct a BIM corresponding to the cut-and-cover construction site of the transfer section of a subway station and perform finite element analysis on the cut-and-cover construction site of the transfer section of the subway station based on the BIM to determine the stress concentration area; The machine learning module 202 is used to collect the safety influencing factors corresponding to the stress concentration area and use a pre-deployed machine learning model to identify the safety influencing factors to determine the optimized data of the support structure; The data feedback module 203 is used to transmit the optimized data of the support structure to the on-site staff, so that the on-site staff can perform construction according to the optimized data of the support structure to complete the intelligent construction process monitoring based on BIM.

[0083] The intelligent construction process monitoring system based on BIM provided by the embodiment of the present invention can execute the above method technical solution, and its principle and beneficial effects are similar, so details are not described here again.

[0084] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can be implemented in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can be implemented in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0086] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction means that implements the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0087] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operating steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0088] Those of ordinary skill in the art can understand that all or part of the steps in implementing the above-mentioned facts and methods can be completed by instructing relevant hardware through a program. The involved program or the program described above can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: At this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0089] The specific embodiments described above further elaborate on the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent construction process monitoring method based on BIM, characterized in that, Including: Construct a BIM corresponding to the excavation construction site of the subway station transition section, and perform finite element analysis on the excavation construction site of the subway station transition section based on the BIM to determine the stress concentration area; Collect the safety influencing factors corresponding to the stress concentration area, and use a pre-deployed machine learning model to identify the safety influencing factors to determine the support structure optimization data; Transmit the support structure optimization data to the on-site staff, so that the on-site staff can construct according to the support structure optimization data to complete the intelligent construction process monitoring based on BIM.

2. The intelligent construction process monitoring method based on BIM according to claim 1, wherein, Construct a BIM corresponding to the excavation construction site of the subway station transition section, and perform finite element analysis on the excavation construction site of the subway station transition section based on the BIM to determine the stress concentration area, including: Construct a BIM corresponding to the excavation construction site of the subway station transition section, and perform finite element analysis on the construction site based on the BIM to obtain the stress data corresponding to the excavation construction site of the subway station transition section; Divide the BIM corresponding to the excavation construction site of the subway station transition section into multiple different areas evenly, and determine the area where the stress data exceeding the preset stress threshold is greater than the preset percentage as the stress concentration area according to the stress data corresponding to the excavation construction site of the subway station transition section.

3. The BIM-based intelligent construction process monitoring method according to claim 1, wherein, Collect the safety influencing factors corresponding to the stress concentration area, including: Divide the stress concentration area into several sub-stress concentration areas, and for each sub-stress concentration area, use the stress, strain, displacement and lithology corresponding to the center of the sub-stress concentration area to obtain the safety influencing factors corresponding to the sub-stress concentration area; Combine the safety influencing factors corresponding to all sub-stress concentration areas to form the safety influencing factors corresponding to the stress concentration area.

4. The BIM-based intelligent construction process monitoring method according to claim 3, characterized in that, Use a pre-deployed machine learning model to identify the safety influencing factors to determine the support structure optimization data, including: Form a data matrix with the safety influencing factors corresponding to the stress concentration area, and input the data matrix into the pre-deployed machine learning model to obtain the support structure optimization data.

5. The BIM-based intelligent construction process monitoring method according to any one of claims 1 to 4, characterized in that The pre-deployment method of the machine learning model includes: Construct a machine learning model with the ability to identify data matrices, and perform initialization coding on the hyperparameters of the machine learning model to obtain multiple hyperparameter codings; Obtain the historical safety influencing factors stored in advance and the support structure optimization data labels corresponding to the historical safety influencing factors, and obtain the fitness corresponding to each hyperparameter coding according to the historical safety influencing factors and the support structure optimization data labels corresponding to the historical safety influencing factors; Determine the hyperparameter coding with the maximum fitness as the optimal coding according to the fitness corresponding to the hyperparameter coding; Perform joint chain search on the hyperparameter coding using a random chain search strategy to obtain the hyperparameter coding after joint chain search; Based on the optimal coding, perform global and local balanced search on the hyperparameter coding after joint chain search using a probability decision balanced search strategy to obtain the hyperparameter coding after balanced search; Use an adaptive random chaos search strategy to globally search the hyperparameter encoding after equilibrium search to obtain the hyperparameter encoding after global search; Repeat the execution of the random chain search strategy, the probability decision equilibrium search strategy, and the adaptive random chaos search strategy until the training end condition is met, and re-obtain the optimal encoding; Deploy the machine learning model according to the re-obtained optimal encoding.

6. The BIM-based intelligent construction process monitoring method according to claim 5, characterized in that Initialize the encoding of the hyperparameters of the machine learning model to obtain multiple hyperparameter encodings, including: Perform random initialization between the upper and lower limits corresponding to the hyperparameters of the machine learning model, and encode the initialized hyperparameters into a vector to obtain the initial hyperparameter encoding; Based on the initial hyperparameter encoding, obtain multiple other hyperparameter encodings as: Among them, represents the k th hyperparameter encoding, and when k = 1, it is the initial hyperparameter encoding, represents the k +(k + 1)th hyperparameter encoding, represents the remainder function, represents pi, sin represents the sine function, represents the first constant term, and is set to 0.5; b represents the second constant term, and is set to 0.

2. It should be noted that there may be some inaccuracies in the content you provided, especially in the description of "the (k + 1)th hyperparameter encoding" which may need to be further adjusted according to the specific context. The above translation is based on the existing text.

7. The BIM-based intelligent construction process monitoring method according to claim 5, characterized in that, Use the random chain search strategy to perform a joint chain search on the hyperparameter encoding to obtain the hyperparameter encoding after joint chain search, including: For any hyperparameter encoding, randomly match an other hyperparameter encoding to the hyperparameter encoding to obtain the joint search encoding corresponding to each hyperparameter encoding; Arrange the hyperparameter encodings in descending order of fitness according to the fitness corresponding to all hyperparameter encodings to obtain the arranged hyperparameter encodings; For the arranged hyperparameter encodings, perform a joint chain search on the hyperparameter encodings according to the joint search encoding corresponding to the hyperparameter encodings to obtain the hyperparameter encoding after joint chain search as: Among them, represents the t -th training i -th sorted hyperparameter encoding's d -th dimensional hyperparameter, i = 1, 2,.., M, where M represents the total number of hyperparameter encodings, d = 1, 2,.., D, where D represents the total dimension of hyperparameters in the hyperparameter encoding, represents the i -th dimensional hyperparameter of the hyperparameter encoding after the d -th joint chain search, sin represents the sine function, cos represents the cosine function, represents pi, represents the first random number between (0, 1), represents the second random number between (0, 1), represents the i -th dimensional hyperparameter of the joint search encoding corresponding to the d -th sorted hyperparameter encoding.

8. The BIM-based intelligent construction process monitoring method according to claim 7, wherein Based on the optimal encoding, use the probability decision equilibrium search strategy to perform global and local equilibrium search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after equilibrium search, including: Based on the current training times, obtain the adaptive decision probability as: Among them, represents the adaptive decision probability, represents the preset basic probability value, represents the Weibull distribution control factor, and W represents the Weibull distribution function, T represents the preset maximum number of training times, t represents the current number of training times; represents the shape parameter of the Weibull distribution, represents the scale parameter of the Weibull distribution, and e represents the natural constant; Generate a third random number between (0, 1), and determine whether the third random number is less than the adaptive decision probability. If so, perform Levy flight search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after equilibrium search. Otherwise, based on the optimal encoding, perform optimal direction search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after equilibrium search; Perform Levy flight search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after equilibrium search as: Among them, represents the t -th training, the j -th hyperparameter encoding after the d -dimensional hyperparameter after the joint chain search, represents the j -th hyperparameter encoding after the d -dimensional hyperparameter after the equilibrium search, j = 1, 2,.., M, where M represents the total number of hyperparameter encodings, represents the t +1-th training, the j -th element in the update speed corresponding to the hyperparameter encoding after the d -dimensional hyperparameter after the joint chain search, represents the first inertia weight, represents the t -th training, the j -th element in the update speed corresponding to the hyperparameter encoding after the d -dimensional hyperparameter after the joint chain search, represents the first learning factor, represents the second learning factor, represents the search range control factor set between (1, 3), represents the j -th hyperparameter encoding after the joint chain search and the first random hyperparameter encoding the Euclidean distance between them, represents the j -th hyperparameter encoding after the joint chain search and the second random hyperparameter encoding the Euclidean distance between them, represents the first Levy flight factor between (0, 1), represents the second Levy flight factor between (0, 1); Based on the optimal encoding, perform optimal direction search on the hyperparameter encoding after joint chain search to obtain the hyperparameter encoding after equilibrium search as: Among them, represents the d -dimensional hyperparameter of the optimal encoding.

9. The BIM-based intelligent construction process monitoring method according to claim 8, wherein Use the adaptive random chaos search strategy to globally search the hyperparameter encoding after equilibrium search to obtain the hyperparameter encoding after global search, including: Obtain the chaos search control factor as: Among them, represents the chaotic search control factor at the t -th training. At the initial moment, the chaotic search control factor is randomly generated between (0, 1). represents the chaotic search control factor at the t +1-th training. represents the remainder function. represents the pi. According to the chaos search control factor, perform global search on the hyperparameter encoding after equilibrium search to obtain the hyperparameter encoding after global search as: Among them, represents the t th n hyperparameter encoding after the d -th balanced search during the n -th training, where = 1, 2,.., M, and M represents the total number of hyperparameter encodings, n represents the d -th hyperparameter encoding after the th global search, t represents the n -th element in the update speed corresponding to the hyperparameter encoding after the d -th balanced search during the + 1-th training, t represents the n -th element in the update speed corresponding to the hyperparameter encoding after the d -th balanced search during the -th training, represents the second inertia weight, represents the d -th hyperparameter of the newly generated hyperparameter encoding randomly, represents the fourth random number uniformly distributed between (-1, 1), represents the preset maximum value of the second inertia weight, represents the preset minimum value of the second inertia weight, represents the preset maximum value of the third learning factor, represents the preset minimum value of the third learning factor, represents the fifth random number uniformly distributed between (0, 1).

10. An intelligent construction process monitoring system based on BIM, characterized in that, Include: BIM data analysis module, machine learning module, and data feedback module; The BIM data analysis module is used to construct a BIM corresponding to the cut-and-cover construction site of the subway station transition section, and perform finite element analysis on the cut-and-cover construction site of the subway station transition section based on the BIM to determine the stress concentration area; The machine learning module is used to collect the safety influencing factors corresponding to the stress concentration area, and use a pre-deployed machine learning model to identify the safety influencing factors to determine the support structure optimization data; The data feedback module is used to transmit the support structure optimization data to on-site workers, so that the on-site workers can perform construction according to the support structure optimization data to complete the intelligent construction process monitoring based on BIM.

Images