A Reliability Analysis Method and System for Sky Building Machines Based on Active Learning

Abstract

This invention discloses a method and system for structural reliability analysis of a high-rise building machine based on active learning. The method includes: establishing a stochastic pulsating wind field to simulate pulsating wind loads on a high-rise building and creating a sample pool; establishing a finite element model of the high-rise building machine and assigning material properties; training an AI agent model based on the sample pool and the finite element model, using an improved QBDC active learning approach; predicting the time-history response of all samples in the sample pool using the trained AI agent model; and determining the structural dynamic reliability based on the prediction results and the PDEM algorithm. Addressing the limitations of traditional reliability analysis methods, such as narrow applicability, large workload, and time-consuming computation, this invention combines artificial intelligence algorithms, finite element method (FEM), and probability density evolution method (PDEM) to achieve rapid construction of a stochastic excitation-structural response AI agent model, thereby overcoming the shortcomings of existing methods in structural reliability assessment.

Description

Technical Field

[0001] This invention belongs to the field of structural random vibration response and reliability assessment technology, and more specifically, relates to a method and system for structural reliability analysis of a building-in-the-sky machine based on active learning. Background Technology

[0002] With the rapid urbanization of the past two decades, high-rise buildings have gained widespread popularity and use worldwide due to their aesthetic design, comfortable environment, and land-saving advantages. However, high-rise building construction presents a series of challenges, including high difficulty, high risk, and complex organization, necessitating advanced construction equipment to improve the on-site working environment and enhance construction efficiency. Based on these needs, aerial building machines (i.e., integrated construction equipment platforms), with their high equipment integration, factory-like operation, fast construction speed, and guaranteed construction safety, have played a significant role in the construction of numerous landmark buildings in China, greatly improving the mechanization and intelligence level of high-rise building construction and propelling my country's high-rise building construction technology to new heights. However, excessive building height and extreme weather conditions can lead to excessive wind vibration responses in the auxiliary structures or equipment of high-rise buildings, potentially causing safety accidents and resulting in significant loss of life and property damage. Therefore, proposing a practical and feasible method for structural reliability analysis of aerial building machines under extreme wind loads is of significant practical importance.

[0003] Traditional methods commonly used for structural reliability calculations include the first-order second-moment method, the second-order second-moment method, and the Monte Carlo method (MCS). These traditional methods are easy to understand and operate, and are typically used for simple structures. However, when dealing with large and complex structures where the function is implicitly expressed, they often reveal shortcomings such as inaccurate calculation results and cumbersome and redundant workload. Furthermore, currently popular surrogate model methods, such as Response Surface Method (RSM), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Kriging, while capable of quickly solving structural reliability problems while maintaining computational accuracy, still cannot solve multivariable, high-dimensional, and complex problems, and are difficult to evaluate the random vibration response of structures, failing to fully and comprehensively express the response characteristics of structures under random excitation. In recent years, the rapidly developing Probability Density Evolution (PDEM) method has obtained the probabilistic evolution characteristics of structural response by solving the generalized probability density equation (GDEE), and has received widespread attention in the fields of structural random vibration and dynamic reliability analysis. However, this method requires multiple simulations of the time-varying response process of the structure in advance to consider the randomness of external excitation, which leads to excessive computational cost and time consumption in the overall analysis process, thus hindering the application and promotion of the method.

[0004] Therefore, how to use intelligent algorithms to construct a complete structural reliability analysis system and achieve accurate and rapid acquisition of the structural response and dynamic reliability of the aerial building machine under random excitation is a key issue. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a method and system for structural reliability analysis of a building-in-the-sky machine based on active learning. Addressing the problems of narrow applicability, large workload, and time-consuming calculations associated with traditional reliability analysis methods, this invention combines artificial intelligence algorithms, finite element method (FEM), and probability density evolution method (PDEM) to achieve rapid construction of a stochastic excitation-structural response AI proxy model, thereby overcoming the shortcomings of existing methods in structural reliability assessment.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for structural reliability analysis of a building-in-the-sky machine based on active learning is provided, comprising:

[0007] Establish a random pulsating wind field to simulate the pulsating wind load on high-rise buildings, and establish a sample pool;

[0008] Establish a finite element model of the aerial building machine and assign material properties to it;

[0009] Based on the sample pool and finite element model, and using improved QBDC active learning to train the AI ​​agent model;

[0010] The trained AI agent model is used to predict the time history response of all samples in the sample pool. Based on the prediction results and the PDEM algorithm, the structural dynamic reliability is determined.

[0011] Furthermore, the establishment of a random fluctuating wind field to simulate the fluctuating wind load on high-rise buildings and the establishment of a sample pool include:

[0012] The time history of horizontal fluctuating wind speed in high-rise buildings is established using spectral representation methods.

[0013] The number of random variables used for simulating fluctuating wind loads was reduced to two by using a dimensionality reduction method based on orthogonal random functions.

[0014] The number theory-based point selection method is used to select a set of sample points from a two-dimensional probability space as a sample pool based on the range of values ​​of the random variable.

[0015] Furthermore, the establishment of the finite element model of the aerial building machine and the assignment of material properties include:

[0016] Finite element modeling was performed on the main structure of the aerial building machine, using beam and shell elements for simulation;

[0017] The cross-sectional properties of beam and shell elements in each part are assigned based on the structural material data of the aerial building machine.

[0018] Furthermore, the step of training the AI ​​agent model based on the sample pool and the finite element model, and on the basis of improved QBDC active learning, includes:

[0019] KMeans clustering is performed on the sample pool, and the k samples closest to the cluster center are selected. The input and output datasets calculated by the finite element model are used as the initial training set for the AI ​​agent model to train the deep learning model.

[0020] The trained AI agent model is used to predict the unlabeled samples in the representative sample set one by one. The sample point with the largest divergence index is selected and added to the dataset used for training the AI ​​agent model. The training process is repeated until all samples have been selected or the fit meets the requirements.

[0021] Furthermore, the input-output dataset calculated by the finite element model is used as the initial training set for training the AI ​​agent model, which includes: using implicit kinematics to calculate the real structural response of the training set samples by the finite element model, with the input being the spatial pulsating wind load corresponding to the sample point and the output being the time-varying response curve of the specific location of the structure corresponding to the sample point, and using the input-output dataset calculated by the finite element model as the training set for training the deep learning model.

[0022] Furthermore, the divergence index is:

[0023]

[0024] i = 1, 2, ..., n sel

[0025] Where i is the i-th sample in the sample pool, and n sel n is the number of samples in the overall sample pool. sel y represents the number of samples in the overall sample pool. nt This represents the predicted value of the nth committee member at time t. Δt represents the average prediction of all committee members at time t, N represents the number of committee members, T is the total time of the response curve, and Δt represents the time interval.

[0026] Furthermore, determining the structural dynamic reliability based on the prediction results and the PDEM algorithm includes:

[0027] The prediction results are substituted into the generalized probability density evolution equation and solved using the finite difference method of the TVD scheme to obtain the structural response probability density function, wherein the prediction results are time-varying response curves.

[0028] The cumulative density function of the structural response is obtained from the probability density function of the structural response, and the structural dynamic reliability is obtained from the cumulative density function of the structural response.

[0029] According to a second aspect of the present invention, a structural reliability analysis system for an aerial building machine based on active learning is provided, comprising:

[0030] The first main module is used to establish a random pulsating wind field, simulate the pulsating wind load on high-rise buildings, and establish a sample pool.

[0031] The second main module is used to build the finite element model of the aerial building machine and assign material properties;

[0032] The third main module is used to train the AI ​​agent model based on the sample pool and finite element model, and on the improved QBDC active learning.

[0033] The fourth main module is used to predict the time history response of all samples in the sample pool using the trained AI agent model, and to determine the structural dynamic reliability based on the prediction results and the PDEM algorithm.

[0034] According to a third aspect of the present invention, an electronic terminal is provided, characterized in that it comprises:

[0035] At least one processor, at least one memory, a communication interface, and a bus; wherein,

[0036] The processor, memory, and communication interface communicate with each other through the bus;

[0037] The memory stores program instructions that can be executed by the processor, which calls the program instructions to implement the method.

[0038] According to a fourth aspect of the present invention, a non-transitory computer-readable storage medium is provided, characterized in that the non-transitory computer-readable storage medium stores computer instructions that cause the computer to implement the method described herein.

[0039] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects:

[0040] 1. The active learning-based structural reliability analysis method of the aerial building machine of the present invention addresses the problems of narrow applicability, large workload and time-consuming calculation of traditional reliability analysis methods. It combines artificial intelligence algorithms (AI), finite element technology (FEM) and probability density evolution method (PDEM) to realize the rapid construction process of stochastic excitation-structural response AI proxy model, thereby overcoming the shortcomings of existing methods in structural reliability assessment.

[0041] 2. The structural reliability analysis method for a high-rise building machine based on active learning, as described in this invention, simulates pulsating wind loads on high-rise buildings. It employs a dimensionality reduction method and NTM to establish a representative sample point set of wind loads, and combines FEM, active learning, and PDEM methods to construct an AI surrogate model and solve the GDEE, thereby rapidly obtaining the structure's stochastic vibration response and dynamic reliability. The error in dynamic reliability calculation obtained by this invention compared to traditional PDEM is within 0.5%, but the calculation time of this invention is reduced by 77.08% compared to traditional PDEM. This invention's method exhibits excellent accuracy and speed.

[0042] 3. The active learning-based structural reliability analysis method for aerial building machines of the present invention provides a new approach and implementation method for the calculation and analysis of structural vibration response and dynamic reliability under external random excitation.

[0043] 4. The active learning-based structural reliability analysis method of the aerial building machine of the present invention constructs an AI proxy model to replace finite element calculation through active learning technology, which greatly improves the calculation speed of representative sample points in the PDEM algorithm and realizes the rapid and accurate acquisition of structural vibration response and dynamic reliability under external random excitation. Attached Figure Description

[0044] Figure 1 This is a flowchart of the sample divergence determination and quantification process according to an embodiment of the present invention;

[0045] Figure 2 This is a flowchart of the active learning method based on improved QBDC according to an embodiment of the present invention;

[0046] Figure 3 This is a flowchart of the random response and reliability analysis of the aerial building machine combining QBDC and PDEM according to an embodiment of the present invention.

[0047] Figure 4 This is a schematic diagram of the finite element model of the aerial building machine according to an embodiment of the present invention;

[0048] Figure 5 This is a schematic diagram of the GRU deep neural network architecture according to an embodiment of the present invention;

[0049] Figure 6 The graph shows a comparison of metrics during iterative training using different methods in this embodiment of the invention (where (a) R2; (b) RMSE; (c) MAE).

[0050] Figure 7 The following is a diagram illustrating the prediction results of the surrogate model in this embodiment of the invention (where (a) sample 50; (b) sample 100; (c) all samples R). 2 distributed);

[0051] Figure 8 The structural stress random response curves of an embodiment of the present invention are shown (where (a) 3D-PDF; (b) PDF curve at a typical moment; and (c) CDF curve at a critical moment).

[0052] Figure 9 The graph shows a comparison of the accuracy and time of the calculation results in the embodiments of the present invention (where (a) is the reliability convergence curve with the number of sampling points; and (b) is a comparison of calculation time).

[0053] Figure 10 This is a flowchart of the structural reliability analysis method for an aerial building machine based on active learning, according to an embodiment of the present invention.

[0054] Figure 11 This is a schematic diagram of the structural reliability analysis system for an aerial building machine based on active learning, according to an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0056] In the description of this invention, unless otherwise explicitly specified and limited, the terms "connected," "linked," and "fixed" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0057] Those skilled in the art will understand that, unless otherwise stated, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the word “comprising” as used in the specification of this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or combinations thereof.

[0058] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as in the embodiments of this application.

[0059] The method of this invention can be applied to fields such as structural random vibration response and reliability assessment, and can accurately and quickly obtain the structural response and dynamic reliability of a building-in-the-sky machine under random excitation.

[0060] The structural reliability analysis method for aerial building machines based on active learning provided in this application can be executed by a server or a terminal. The server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be a smartphone, tablet, laptop, desktop computer, smartwatch, smart terminal, smart TV, etc., and the specific device can be determined based on the actual application scenario requirements, without limitation here.

[0061] Figure 10 A flowchart of a structural reliability analysis method for an aerial building machine based on active learning, provided as an embodiment of the present invention, is shown below. Figure 10 As shown, the structural reliability analysis method of the aerial building machine based on active learning includes the following steps S100 to S400.

[0062] Step S100: Establish a random pulsating wind field to simulate the pulsating wind load on high-rise buildings and establish a sample pool;

[0063] In structural wind engineering, fluctuating wind loads acting on structures exhibit significant dynamic and spatiotemporal distribution characteristics, typically described using spatiotemporal random fields. Simulating random fluctuating wind fields is a primary task in the wind-resistant design of engineering structures. Spectral representation methods are widely used for simulating random wind fields due to their well-developed theory, simple algorithms, and high computational accuracy. To achieve structural reliability analysis of high-rise buildings, it is necessary to simulate the fluctuating wind field of high-rise buildings; therefore, it is essential to establish a random fluctuating wind field to simulate the fluctuating wind loads on high-rise buildings.

[0064] Specifically, firstly, the spectral representation method (SRM) is used to establish the time history of horizontal fluctuating wind speeds in high-rise buildings, and the number of random variables used for fluctuating wind load simulation is reduced to two using a dimension reduction method based on orthogonal random functions. Then, the number theory selection method (NTM) is used to select a representative and uniformly distributed set of sample points from the two-dimensional probability space based on the range of values ​​of the random variables, which serves as the sample pool for subsequent active learning and PDEM calculation.

[0065] After the number of random variables is reduced to two, their value range is generally (0, 2π). A set of sample points is selected in the two-dimensional probability space within this range as the sample pool.

[0066] Step S200: Establish a finite element model of the aerial building machine and assign material properties;

[0067] The "Sky Building Machine," a new industrialized intelligent construction technology, is an equipment platform and supporting construction technology independently developed in my country. The Sky Building Machine and its construction technology utilize mechanical operations and intelligent control to achieve industrialized intelligent construction of high-rise residential buildings using cast-in-place reinforced concrete. A key feature is that the entire process is completed centrally and layer by layer in the air. Therefore, it is also known as the "Sky Building Machine."

[0068] Specifically, step S200 includes the following steps:

[0069] Finite element modeling was performed on the main structure of the aerial building machine, using beam and shell elements for simulation;

[0070] The cross-sectional properties of beam and shell elements in each part are assigned based on the structural material data of the aerial building machine.

[0071] Specifically, the aerial building machine is a complex prefabricated steel structure system. To simplify the modeling process, only the main structure of the machine, namely the steel platform and supporting columns, is modeled using finite element methods. Considering both computational efficiency and accuracy, the steel trusses, reinforcements, and columns of the aerial building machine are simulated using beam and shell elements. For each part of the aerial building machine, beam and shell elements are used for simulation. Based on the structural material data of each part, the beam and shell elements are assigned corresponding section properties, thus achieving finite element simulation of each part of the aerial building machine.

[0072] To establish a random fluctuating wind field, the sample pool obtained by simulating the fluctuating wind load on a high-rise building is input into the finite element model. Implicit dynamics can be used to solve the structural response and obtain the time-varying response curve.

[0073] After completing the finite element model construction of the random pulsating wind field, sample pool, and aerial building machine, the training and test sets of the AI ​​agent model are generated based on the sample pool and the aerial building machine's finite element model, and the agent model is trained. Details are as follows:

[0074] Step S300: Based on the sample pool and finite element model, and using the improved QBDC active learning to train the AI ​​agent model;

[0075] Query-by-committee (QBC) is a classic active learning method commonly used for classification problems. This method trains multiple machine learning models (each model acting as a committee member) to predict all samples in a sample pool, using the divergence between the prediction models as a criterion for the information content of a specific sample. The prediction accuracy of the surrogate model is improved by including the sample with the highest divergence in the training set. To use a deep learning model as a surrogate model estimator and improve its training speed, the QBC method utilizes Monte Carlo Dropout technology, enabling multiple prediction results to be obtained from a single deep learning model. This is equivalent to training multiple deep learning models simultaneously, significantly improving the training efficiency of the surrogate model. To apply the QBC method to prediction and regression problems, this invention proposes an improved QBC method. For a given sample, the standard deviation of multiple time-history prediction curves obtained from the surrogate model is processed at each time step to obtain the standard deviation time-history curve for that sample. Integrating this time-history curve yields the model prediction divergence index for that specific sample. The flowchart is illustrated below. Figure 1 As shown.

[0076] In this invention, the prediction divergence of a sample is represented by DD (Degree of Disagreement), and the expression is shown in Equation (1):

[0077]

[0078] Where i is the i-th sample in the sample pool, and n sel y represents the number of samples in the overall sample pool. nt This represents the predicted value of the nth committee member at time t. Let N represent the average prediction of all committee members at time t, and let N represent the number of committee members. T Δt represents the total time of the response curve.

[0079] The process of training an AI agent model based on a sample pool and a finite element model, and using improved QBDC active learning, includes:

[0080] KMeans clustering is performed on the sample pool, and the k samples closest to the cluster center are selected. The input and output datasets calculated by the finite element model are used as the initial training set for the AI ​​agent model to train the deep learning model.

[0081] Using the AI ​​agent model that has been trained in the previous iteration, predict the unlabeled samples in the representative sample set one by one, and select the sample point with the largest divergence index to add to the dataset used for training the AI ​​agent model; repeat the training process until all samples have been selected or the fit meets the requirements.

[0082] Each sample point in the dataset used for training the AI ​​agent model includes the spatial pulsating wind load corresponding to the sample point and the actual structural response of the sample point calculated by the finite element model.

[0083] In the above process, when selecting the k samples closest to the cluster center, the number of samples can be adjusted based on the speed and accuracy of the model's first training, which will not be elaborated here.

[0084] The input-output dataset calculated by the finite element model is used as the initial training set for training the AI ​​agent model. This includes: using implicit kinematics to calculate the real structural response of the training set samples through the finite element model, with the input being the spatial pulsating wind load corresponding to the sample point and the output being the time-varying response curve of the structure at a specific location corresponding to the sample point. The input-output dataset calculated by the finite element model is used as the training set for training the deep learning model.

[0085] Step S400: Use the trained AI agent model to predict the time history response of all samples in the sample pool, and determine the structural dynamic reliability based on the prediction results and the PDEM method.

[0086] Specifically, a high-performance AI proxy model is used to perform fast time-history response prediction on all samples in the sample pool, obtaining the prediction results (time-varying response curves). The PDEM algorithm is then used to convert the prediction results (time-varying response curves) into probabilistic data (structural dynamic reliability). Based on the prediction results and the PDEM algorithm, the structural dynamic reliability is determined, including: substituting the prediction results into the generalized probability density evolution equation and solving it using the finite difference method in TVD format to obtain the structural response probability density function, where the prediction results are time-varying response curves; obtaining the cumulative density function of the structural response from the structural response probability density function; and obtaining the structural dynamic reliability from the cumulative density function of the structural response.

[0087] The implementation of the various embodiments of this invention is based on programmed processing through a device with a central processing unit (CPU). Therefore, in practical engineering, the technical solutions and functions of the various embodiments of this invention can be encapsulated into various modules. Based on this reality, and building upon the above embodiments, the embodiments of this invention provide an active learning-based aerial building machine structural reliability analysis system for executing the active learning-based aerial building machine structural reliability analysis method in the above method embodiments. For example... Figure 11 As shown, it includes:

[0088] The first main module is used to establish a random pulsating wind field, simulate the pulsating wind load on high-rise buildings, and establish a sample pool.

[0089] The second main module is used to build the finite element model of the aerial building machine and assign material properties;

[0090] The third main module is used to train the AI ​​agent model based on the sample pool and finite element model, and on the improved QBDC active learning.

[0091] The fourth main module is used to predict the time history response of all samples in the sample pool using the trained AI agent model, and to determine the structural dynamic reliability based on the prediction results and the PDEM algorithm.

[0092] It should be noted that the apparatus in the device embodiments provided by the present invention can be used not only to implement the methods in the above method embodiments, but also to implement the methods in other method embodiments provided by the present invention. The only difference is that corresponding functional modules are set. The principle is basically the same as that of the above device embodiments provided by the present invention. As long as those skilled in the art can improve the apparatus in the above device embodiments by referring to the specific technical solutions in other method embodiments and combining technical features to obtain corresponding technical means and technical solutions composed of these technical means, on the basis of the above device embodiments, under the premise of ensuring the practicality of the technical solutions, so as to obtain corresponding device-type embodiments for implementing the methods in other method-type embodiments.

[0093] The methods in the embodiments of the present invention are implemented using electronic devices; therefore, it is necessary to describe the relevant electronic devices. For this purpose, embodiments of the present invention provide an electronic device comprising: at least one central processor, a communications interface, at least one memory, and a communication bus, wherein the at least one central processor, the communications interface, and the at least one memory communicate with each other via the communication bus. The at least one central processor can invoke logical instructions stored in the at least one memory to execute all or part of the steps of the methods provided in the foregoing method embodiments.

[0094] Furthermore, when the logical instructions in at least one of the aforementioned memories can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0095] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0096] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0097] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Based on this understanding, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, or sometimes in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0098] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0099] like Figure 3 The illustration shows a structural reliability analysis method for a building-building machine based on improved QBDC active learning, provided by an embodiment of the present invention. Taking the application of a lightweight building-building machine in the construction of a 400m high-rise building as an example, the specific process of the method of the present invention is described in detail, including:

[0100] Step S100: Establish a random pulsating wind field to simulate the pulsating wind load on high-rise buildings and establish a sample pool;

[0101] The spectral representation method based on orthogonal random variables is shown in formula (2):

[0102]

[0103]

[0104]

[0105] s = 1, 2, ..., n, m = 1, 2, ..., N

[0106] Where i represents the i-th spatial simulation point in the pulsating wind field; N Indicates the number of frequency intervals; H ij (w k ) represents the lower triangular matrix obtained through Cholesky decomposition; Δw is the frequency increment; Represents an independent random phase angle sequence uniformly distributed within an interval of 0 to 2π; orthogonal random variables Let be the set of orthogonal functions with basic random variables {Θ1,Θ2}; n is the number of spatial simulation points in the fluctuating wind field; w jk Indicates the frequency of double indexing.

[0107] Since the original orthogonal random variable {R,I} only needs to satisfy the orthogonality condition, a two-dimensional random variable determined by the probability distribution can be used. To represent the original orthogonal random variable {R,I}, thus reducing the complexity of the simulation. The expressions are shown in equations (3) and (4), where the orthogonal random variable... Defined as a set of orthogonal functions with basic random variables {Θ1,Θ2}, followed by the completion of orthogonal random variables. A deterministic one-to-one mapping between the original orthogonal random variables {R,I} ensures that the orthogonal random variables are uniquely determined. A significant advantage of the dimensionality reduction method is that it can reduce the number of random variables involved in SRM to two. According to the "Code for Design of Building Structures" (GB50009-2012), the Davenport wind speed spectrum is usually selected for wind load simulation of high-rise buildings, and its wind spectrum expression is shown in (5) below:

[0108]

[0109] Where k0 is the surface roughness; ω represents the average wind speed at a height of 10 meters; ω is the angular frequency.

[0110] Based on the actual operation of aerial construction machines, construction should be halted and reinforcement measures implemented when wind speeds exceed 40 m / s. Therefore, the average wind speed at a height of 400 m is set at 40 m / s, and the relevant parameter values ​​are shown in Table 1.

[0111] Table 1. Simulation parameters of pulsating wind load

[0112]

[0113] The time history of horizontal fluctuating wind speed in high-rise buildings was established using spectral representation; the number of random variables used for fluctuating wind load simulation was reduced to two using a dimension reduction method based on orthogonal random functions; and a number theory-based point selection method was used to select a set of sample points from the two-dimensional probability space as a sample pool based on the range of values ​​of the random variables.

[0114] Step S200: Establish a finite element model of the aerial building machine and assign material properties; based on the actual structural conditions, establish a finite element model of the aerial building machine, such as... Figure 4 As shown, the model material properties and construction load data are presented in Tables 2 and 3. Each column of the building support system has three support points. The model boundary conditions are set as follows: the upper support point is subject to horizontal constraints, while the middle and lower support points are subject to lateral and longitudinal constraints. To fully consider the elastic-plastic properties of the material, an ideal bilinear constitutive model is introduced. This constitutive model assumes that when the steel yields, the plastic modulus is 0.1 times that of the elastic model, i.e., E... s =0.1E.

[0115] Table 2. Q355 Material Properties

[0116]

[0117] Table 3. Load Application Information for Finite Element Model

[0118]

[0119] A uniform representative sample set of 233 sample points was generated using the NTM method. Specifically, when the number of random variables in the system is 2, the sample point set can be calculated using the Fibonacci sequence, as shown in Equation (6). Subsequently, the KMeans clustering algorithm was used to classify the established sample set, and the 5 samples closest to the cluster center were selected as the initial training set and added to the finite element model to calculate the true response of the structure.

[0120]

[0121] Where int{·} represents taking the integer part of the value; k is the number of sample points; F n =F n-1 +F n-2 F1 = F2 = 1, n = 3, 4, ...;

[0122] Step S300: Based on the sample pool and finite element model, and using the improved QBDC active learning to train the AI ​​agent model;

[0123] The deep learning model GRU is selected as the surrogate model estimator, such as... Figure 5As shown, the selected network architecture has two GRU layers and two fully connected (FC) layers, with the FC layers following the GRU network. Additionally, a Monte Carlo Droput layer is added between every two layers to simulate the multi-committee effect. The network parameters are set as follows: input data dimension = 10, number of neurons in the GRU layer and the first FC layer = 100, and number of neurons in the second FC layer = 1. The training process of the surrogate model used for time-varying structural response analysis under random excitation in this case can be summarized as follows:

[0124] (1) 15% of the sample points (approximately 35 samples) were randomly selected from the NTM sample pool (233 samples) as the test set. During the training of the surrogate model, each sample point consisted of wind speed input and output as structural response;

[0125] (2) The dataset used for model training is randomly divided into a training set and a validation set in an 8:2 ratio;

[0126] (3) Based on the test set R after training is completed 2 The indicators determine whether further training is needed. If so, each member of the committee (i.e., the deep learning model) makes predictions for each unlabeled sample in the sample pool, selects the sample point with the highest divergence index DD, adds it to the dataset used for model training, and then repeats steps (2)-(3) until the test set R... 2 The termination conditions have been met.

[0127] The generated random fluctuating wind loads are used as input data for model training, and the time-varying stress response at the weakest points of the structure is used as output. The number of QBDC members is set to 10, and when R... 2 A value greater than 0.95 is the training termination condition. Figure 6 Shows R on the test set during active learning training 2 The changes in RMSE and MAE can be observed. It can be seen that the active learning strategy of this invention improves RMSE and MAE during the training process. 2 The RMSE and MAE curves of actively learned R are all superior to those of passive sampling. During the first sampling, the R... 2 It can reach 0.81, while the R of passive learning is... 2 The R-value is 0.65. When only 16 sample points are sampled, the active learning method achieves an R-value of 0.65. 2 It can reach 0.963, while the passive learning R... 2The accuracy is only 0.851. The results show that the learning efficiency and accuracy of the active learning strategy are higher than those of passive sampling. It is worth noting that fluctuations in the metric curve during iterations are reasonable, as the training and validation sets are shuffled in each iteration. Therefore, the training accuracy of GRU may not be ideal in a single iteration. However, as the number of iterations increases, the overall trend of the metric curve becomes increasingly better. Using the trained AI agent model to predict the responses of all samples in the sample pool, two samples are randomly selected, and the time-history curves of the AI ​​agent model's prediction results and the exact solutions are compared. Figure 7 As shown in (a) and (b), the prediction results showed good fitting performance. Figure 7 (c) shows the R values ​​for the total of 233 samples. 2 Distribution. It can be seen that for most predicted samples, the sample R... 2 It achieved a satisfactory value of over 0.9.

[0128] Step S400: Use the trained AI agent model to predict the time history response of all samples in the sample pool, and determine the structural dynamic reliability based on the prediction results and the probability density evolution method (PDEM).

[0129] First, an improved QBDC active learning method is used to predict the time history response curves of the weak points of the aerial building machine corresponding to all sample points in the sample pool. The response prediction results of 233 sample points can be quickly obtained through an AI proxy model. Then, the prediction results are substituted into GDEE to solve the stochastic response of the structure. The structural stress response PDF is shown below. Figure 8 As shown in (a) and (b). Finally, by integrating the portion of the PDF curve below the specification limit at each moment of the entire process, the minimum dynamic reliability of the structure and its corresponding CDF curve can be obtained, as shown in (a) and (b). Figure 8 As shown in (c).

[0130] Regarding the calculation results, under pulsating winds with an average wind speed of 40 m / s, if the material strength design value of 310 MPa is taken as the control threshold, the stress reliability of the aerial building machine is 0.932. In terms of calculation accuracy, using the traditional PDEM algorithm to calculate the structural reliability, the lowest dynamic reliability under stress control is 0.936, which differs from the result calculated by the method of this invention by only 0.43%. Furthermore, from... Figure 9As can be seen, with the increase of sampling points, the calculation results of this invention gradually converge to the precise value of 0.936 obtained by traditional PDEM calculation, indicating that the method of this invention has good accuracy and robustness. In terms of computation time, the experiment used a personal computer configured with a 12th generation Intel(R) Core(TM) i7-12700 CPU, NVIDIA GeForce RTX 3070 GPU card and 16GB RAM. It took 97.08 hours to calculate 233 sets of samples using the traditional PDEM method, while the method of this invention only took 22.25 hours, which shortened the computation and analysis time by 77.08% and greatly improved the efficiency of structural response analysis.

[0131] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0132] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for structural reliability analysis of a building-in-the-sky machine based on active learning, characterized in that, include: Establish a random pulsating wind field to simulate the pulsating wind load on high-rise buildings, and establish a sample pool; Establish a finite element model of the aerial building machine and assign material properties to it; Based on the sample pool and finite element model, and using improved QBDC active learning to train the AI ​​agent model; The process of training an AI agent model based on a sample pool and a finite element model, and using improved QBDC active learning, includes: KMeans clustering is performed on the sample pool, and the k samples closest to the cluster center are selected. The input and output datasets calculated by the finite element model are used as the initial training set for the AI ​​agent model to train the deep learning model. The trained AI agent model is used to predict the unlabeled samples in the representative sample set one by one. The sample point with the largest divergence index is selected and added to the dataset used for training the AI ​​agent model. The training process is repeated until all samples have been selected or the fit meets the requirements. The trained AI agent model is used to predict the time history response of all samples in the sample pool. Based on the prediction results and the PDEM algorithm, the structural dynamic reliability is determined.

2. The structural reliability analysis method for an aerial building machine based on active learning according to claim 1, characterized in that, The establishment of a random fluctuating wind field to simulate the fluctuating wind load on high-rise buildings and the establishment of a sample pool include: The time history of horizontal fluctuating wind speed in high-rise buildings is established using spectral representation methods. The number of random variables used for simulating fluctuating wind loads was reduced to two by using a dimensionality reduction method based on orthogonal random functions. The number theory-based point selection method is used to select a set of sample points from a two-dimensional probability space as a sample pool based on the range of values ​​of the random variable.

3. The structural reliability analysis method for an aerial building machine based on active learning according to claim 1, characterized in that, The establishment of a finite element model of the aerial building machine and the assignment of material properties include: Finite element modeling was performed on the main structure of the aerial building machine, using beam and shell elements for simulation; The cross-sectional properties of beam and shell elements in each part are assigned based on the structural material data of the aerial building machine.

4. The structural reliability analysis method for an aerial building machine based on active learning according to claim 3, characterized in that, The input-output dataset calculated by the finite element model is used as the initial training set for training the AI ​​agent model. This includes: using implicit kinematics to calculate the real structural response of the training set samples through the finite element model, with the input being the spatial pulsating wind load corresponding to the sample point and the output being the time-varying response curve of the specific location of the structure corresponding to the sample point. The input-output dataset calculated by the finite element model is used as the training set for training the deep learning model.

5. The structural reliability analysis method for an aerial building machine based on active learning according to claim 3, characterized in that, The divergence index is: ， in, For the first in the sample pool One sample, This represents the number of samples in the overall sample pool. Indicates the first Each committee member Predicted time value, Indicates that all committee members are in The average predicted time. Indicates the number of committee members, It is the total time of the response curve. Indicates a time interval.

6. The structural reliability analysis method for an aerial building machine based on active learning according to claim 1, characterized in that, The determination of structural dynamic reliability based on prediction results and the PDEM algorithm includes: The prediction results are substituted into the generalized probability density evolution equation and solved using the finite difference method of the TVD scheme to obtain the structural response probability density function, wherein the prediction results are time-varying response curves. The cumulative density function of the structural response is obtained from the probability density function of the structural response, and the structural dynamic reliability is obtained from the cumulative density function of the structural response.

7. A structural reliability analysis system for an aerial building machine based on active learning, characterized in that, include: The first main module is used to establish a random pulsating wind field, simulate the pulsating wind load on high-rise buildings, and establish a sample pool. The second main module is used to build the finite element model of the aerial building machine and assign material properties; The third main module is used to train the AI ​​agent model based on the sample pool and finite element model, and on the improved QBDC active learning. The process of training an AI agent model based on a sample pool and a finite element model, and using improved QBDC active learning, includes: KMeans clustering is performed on the sample pool, and the k samples closest to the cluster center are selected. The input and output datasets calculated by the finite element model are used as the initial training set for the AI ​​agent model to train the deep learning model. The trained AI agent model is used to predict the unlabeled samples in the representative sample set one by one. The sample point with the largest divergence index is selected and added to the dataset used for training the AI ​​agent model. The training process is repeated until all samples have been selected or the fit meets the requirements. The fourth main module is used to predict the time history response of all samples in the sample pool using the trained AI agent model, and to determine the structural dynamic reliability based on the prediction results and the PDEM algorithm.

8. An electronic terminal, characterized in that, include: At least one processor, at least one memory, a communication interface, and a bus; wherein, The processor, memory, and communication interface communicate with each other through the bus; The memory stores program instructions that can be executed by the processor, which invokes the program instructions to implement the method as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1-6.

Images