Fabricated building intelligent construction method and system

Abstract

The application discloses an intelligent construction method and system for fabricated buildings, and relates to the technical field of building construction.The method comprises the following steps: collecting on-site construction data of the fabricated building, obtaining environmental influence characteristics and component state characteristics, inputting a component sway posture physical model and a component damage risk probability model, obtaining a theoretical maximum sway amplitude, a theoretical sway speed and an initial damage probability of the prefabricated component, training a hoisting safety decision model using the on-site construction data, outputting the damage probability, issuing an operation instruction, establishing a post-construction database, and graphically rendering a hoisting track of the prefabricated component.The application builds a self-learning decision system combining a physical model and a data model, predicts dynamic risks in the hoisting process, and outputs graded operation instructions, thereby solving the problem of decision blindness caused by reliance on subjective experience and macro data, achieving a balance between avoiding hidden damage to components and unnecessary downtime, and improving the safety, quality and efficiency of the construction of fabricated buildings.

Description

Technical Field

[0001] This invention relates to the field of building construction technology, specifically to intelligent construction methods and systems for prefabricated buildings. Background Technology

[0002] Prefabricated construction is a new type of building construction method that involves prefabricating some building components in a factory. These prefabricated components are manufactured in the factory through standardized and mechanized production processes with quality control. Subsequently, the prefabricated components are transported to the construction site and assembled into a complete building through reliable connection methods, such as bolting, welding, and post-cast concrete connection. Compared with traditional cast-in-place construction, prefabricated construction has the advantages of faster construction speed, reduced labor demand, less impact from weather conditions, and less construction waste generation. It effectively improves construction efficiency and quality, and promotes the development of the construction industry towards industrialization and greening.

[0003] To address the challenge of accurately assessing hoisting risks in prefabricated building construction due to the dynamic and unpredictable construction environment and the inability to quantify the internal state of components in real time, existing technologies rely on fixed weather monitoring points at the construction site and on the experience of on-site personnel for operational decisions. However, this approach suffers from discrepancies between the monitored environmental data and the actual microenvironment at high altitudes, along with vague decision-making criteria and a failure to correlate real-time stress on the components. Consequently, it becomes impossible to quantify and predict hoisting safety, leading to issues such as overly conservative work stoppages to ensure safety, resulting in inefficiency, or continued work due to misjudgments, causing hidden damage to components or even triggering safety accidents. Therefore, a smart construction method and system for prefabricated buildings is proposed to solve these problems. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent construction method and system for prefabricated buildings to solve the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: Firstly, an intelligent construction method and system for prefabricated buildings, comprising the following steps 1-6:

[0006] Step 1: Collect and preprocess wind speed data, wind direction data, and internal strain data during the hoisting process of the prefabricated components of the prefabricated building in the corresponding hoisting operation area to obtain preprocessed wind speed data, wind direction data, and internal strain data.

[0007] Step 2: Obtain environmental impact characteristics and component state characteristics from the preprocessed wind speed data, wind direction data, and internal strain data;

[0008] Step 3, input the environmental influence characteristics and the component state characteristics into the component sway posture physical model and the component damage risk probability model, respectively output the theoretical maximum sway amplitude, the theoretical sway speed and the initial damage probability of the prefabricated component;

[0009] Step 4, combine the component state characteristics, the theoretical maximum sway amplitude, the theoretical sway speed and the initial damage probability, input into the hoisting safety decision model, and use the wind speed data, the wind direction data and the internal strain data to train the hoisting safety decision model;

[0010] Step 5, input the environmental influence characteristics and the component state characteristics into the hoisting safety decision model, output the damage probability, and based on the damage probability, issue operation instructions of different risk levels;

[0011] Step 6, establish a construction posterior database, associate the hoisting actual results with the output results of the hoisting safety decision model, and graphically render the hoisting trajectory of the prefabricated component.

[0012] The further improvement of the technical scheme of the present application is that the wind speed data, the wind direction data and the internal strain data of the prefabricated component of the fabricated building in the hoisting operation area are collected and preprocessed, and the process of obtaining the preprocessed wind speed data, wind direction data and internal strain data includes:

[0013] A three-dimensional ultrasonic wind speed and direction instrument is fixedly arranged on the hook trolley of the hoisting arm of the tower crane for performing hoisting operation, and is used for obtaining the wind speed data and the wind direction data of the prefabricated component at the real-time operation height in the hoisting process, wherein the three-dimensional ultrasonic wind speed and direction instrument calculates the wind speed data and the wind direction data by measuring the time difference of ultrasonic wave passing through the air.

[0014] During the production of the prefabricated component template pouring, the optical fiber Bragg grating sensing cable is embedded in the prefabricated component along the preset stress transmission path, the leading end of the optical fiber Bragg grating sensing cable is integrated into the optical fiber adapter interface, the optical fiber adapter interface is fixedly arranged on the surface of the prefabricated component, before hoisting operation, the optical fiber grating demodulator is connected with the optical fiber Bragg grating sensing cable through the optical fiber adapter interface, and the optical fiber grating demodulator calculates the internal strain data in real time by monitoring the center wavelength drift, wherein the center wavelength drift is generated by the optical fiber grating demodulator emitting broadband light to the optical fiber Bragg grating sensing cable, and the grating structure in the optical fiber Bragg grating sensing cable is physically deformed due to the influence of the internal strain of the prefabricated component;

[0015] The wind speed data and the wind direction data are processed by using a sliding time window average algorithm to smooth the instantaneous extreme fluctuation and obtain the preprocessed wind speed data and wind direction data;

[0016] The Kalman filtering algorithm is used to process internal strain data, so as to filter out measurement noise and obtain preprocessed internal strain data.

[0017] Further improvement of the technical scheme of the present application is that the process of obtaining environmental influence features and component state features from the preprocessed wind speed data, wind direction data and internal strain data comprises:

[0018] The environmental influence features include a gust coefficient and a wind direction deviation angle, and the component state features include a maximum stress gradient and a stress fluctuation frequency;

[0019] In a preset time window, the maximum wind speed value and the average wind speed value are extracted from the preprocessed wind speed data, and the gust coefficient is obtained by dividing the maximum wind speed value by the average wind speed value;

[0020] The normal direction of the windward surface of the prefabricated component is set as a reference direction, and the wind direction deviation angle is the included angle between the real-time wind direction shown by the preprocessed wind direction data and the reference direction;

[0021] The preprocessed internal strain data is converted into stress data through a preset material elastic modulus of the prefabricated component , and the time variation rate of the stress data is calculated in the time window, and the maximum value of the absolute value of the time variation rate is extracted as the maximum stress gradient;

[0022] In the time window, the converted stress data is applied to the fast Fourier transform algorithm to obtain the frequency domain spectrum of the stress data, and the frequency point with the highest amplitude in the frequency domain spectrum is extracted as the stress fluctuation frequency.

[0023] Further improvement of the technical scheme of the present application is that the process of inputting the environmental influence features into the component swing posture physical model and outputting the theoretical maximum swing amplitude and the theoretical swing speed of the prefabricated component comprises:

[0024] The prefabricated component and the hoisting cable thereof in the hoisting process are simplified into a single pendulum dynamics model with the hook suspension point as the fixed rotation shaft and the prefabricated component as the pendulum bob, and the inherent property parameters of the single pendulum dynamics model include the mass of the prefabricated component, the projection area of the windward surface of the prefabricated component , the length of the hoisting cable and the air drag coefficient ;

[0025] The environmental influence features are input as external forcing terms of the single pendulum dynamics model architecture, and the component swing posture physical model is established to solve the component of wind force on the prefabricated component in the swing direction , and the calculation process is as follows:

[0026] ;

[0027] wherein, is the air density, is the real-time wind speed calculated from the gust factor, is the wind direction deviation angle;

[0028] The calculated wind force is substituted into the motion differential equation of the single-pendulum dynamics model, and the motion differential equation is solved by a numerical method to obtain a swing angle time sequence of the prefabricated component under the action of the wind force;

[0029] The maximum angle value is extracted from the swing angle time sequence, and the maximum angle value is multiplied by the length of the hoisting cable to obtain a theoretical maximum swing amplitude;

[0030] The swing angle time sequence is differentiated once to obtain a swing angular velocity sequence of the prefabricated component, and the maximum value in the swing angular velocity sequence is extracted as a theoretical swing speed.

[0031] Further improvement of the technical scheme of the present application is that the process of inputting the component state features into the component damage risk probability model and outputting the initial damage probability of the prefabricated component includes:

[0032] A component damage risk probability model based on a gradient boosting decision tree algorithm is constructed, the component damage risk probability model is composed of decision trees established in sequence, and a labeled data set containing historical component state features and corresponding real damage results is used for training in advance;

[0033] In the training stage of the component damage risk probability model, a basic decision tree is initialized, the difference between the historical damage result predicted by the component damage risk probability model and the real damage result is taken as a residual error, each subsequent decision tree is established, the residual error of the prediction result of the established decision tree combination is taken as a learning target, and the decision tree establishment process is iterated to optimize the component damage risk probability model in the gradient direction of reducing the residual error;

[0034] The real-time acquired maximum stress gradient and stress fluctuation frequency are input into the trained component damage risk probability model, the input maximum stress gradient and stress fluctuation frequency pass through the established decision trees in the component damage risk probability model in sequence, each decision tree generates a prediction value, and the weighted sum of each prediction value is taken as the initial damage probability.

[0035] Further improvement of the technical scheme of the present application is that the component state features, the theoretical maximum swing amplitude, the theoretical swing speed and the initial damage probability are combined and input into a hoisting safety decision model, and the process of training the hoisting safety decision model using wind speed data, wind direction data and internal strain data includes:

[0036] The component state feature, the theoretical maximum sway amplitude, the theoretical sway speed and the initial damage probability are combined into a multi-dimensional input feature vector;

[0037] A historical data set for training of the hoisting safety decision model is constructed, the historical data set including a multi-dimensional input feature vector derived from historical wind speed data, wind direction data and internal strain data, and a true safety result category corresponding to each hoisting operation, the safety result category including a component intact, a component damage and an operation interruption;

[0038] The hoisting safety decision model is supervised trained based on a multi-layer feedforward neural network architecture, internal parameters of the multi-layer feedforward neural network architecture including connection weights between neurons and biases of the neurons, the multi-layer feedforward neural network architecture including an input layer, a hidden layer and an output layer, a number of nodes of the input layer being the same as a dimension of the multi-dimensional input feature vector, a number of nodes of the output layer being the same as a number of preset safety result categories ;

[0039] The multi-dimensional input feature vector is input into the hoisting safety decision model, the hoisting safety decision model performing forward calculation according to the internal parameters and outputting a predicted safety result;

[0040] The forward calculation process includes, from the input layer, each neuron of a next layer performing weighted summation on output signals of neurons of a previous layer, inputting a weighted summation result into a preset nonlinear activation function for processing, a processing result of the nonlinear activation function being taken as an output signal of each neuron, and being transmitted to a next layer of neurons until the hidden layer, and after output signals of the hidden layer are transmitted to the output layer, the first weighted summation result of the output layer neurons is converted into a damage probability distribution of the first safety result category by a normalization exponential function, and a safety result category corresponding to a highest damage probability value is taken as the predicted safety result output by the hoisting safety decision model; ;

[0041] A loss function is defined to quantify an error between the predicted safety result and a true safety result, and a gradient descent method is adopted to reversely adjust the internal parameters according to the error calculated by the loss function, so that the predicted safety result tends to be close to the true safety result;

[0042] The step of supervised training is repeatedly performed until a value of the loss function converges below a preset loss threshold, and after the process is completed, a trained hoisting safety decision model is obtained.

[0043] ​​The further improvement of the technical scheme of the present application is that the environmental influence features and the component state features are input into the hoisting safety decision model, the damage probability is output, and the process of issuing operation instructions of different risk levels based on the damage probability includes:

[0044] In the actual hoisting operation, the real-time acquired environmental influence features and component state features are input into the trained hoisting safety decision model, the hoisting safety decision model performs forward calculation process by using internal parameters, and outputs the damage probability corresponding to the predicted safety result;

[0045] Based on the damage probability, operation instructions of different risk levels are issued through preset judgment logic, the judgment logic includes preset high-risk threshold and low-risk threshold, wherein the high-risk threshold is greater than the low-risk threshold, the damage probability is compared with the high-risk threshold and the low-risk threshold, and the operation instruction includes stopping operation, adjusting operation parameters and continuing operation according to the current parameters;

[0046] When the damage probability is greater than the high-risk threshold, it is determined as a high-risk state, and the operation is stopped, when the damage probability is lower than the high-risk threshold and higher than the low-risk threshold, it is determined as a medium-risk state, and the operation parameters are adjusted, and when the damage probability is lower than the low-risk threshold, it is determined as a low-risk state, and the operation is continued according to the current parameters.

[0047] The further improvement of the technical scheme of the present application is that the construction posterior database is established, and the process of associating the hoisting actual result with the output result of the hoisting safety decision model includes:

[0048] After each independent hoisting operation is completed, the multi-dimensional input feature vector, the damage probability, the operation instruction and the real safety result related to the hoisting operation are associated and combined as a posterior data record, the posterior data record is added to the historical data set to form the construction posterior database.

[0049] The further improvement of the technical scheme of the present application is that the process of graphically rendering the hoisting trajectory of the prefabricated component includes:

[0050] A digital twin scene including the prefabricated building construction site, the tower crane and the prefabricated component is constructed in advance;

[0051] In the hoisting operation, the three-dimensional space coordinates of the hook are acquired from the control system of the tower crane, and the position of the prefabricated component three-dimensional model in the digital twin scene is updated in real time according to the three-dimensional space coordinates, so that the hoisting trajectory of the prefabricated component is dynamically drawn;

[0052] The real-time acquired environmental influence features are rendered as dynamic vector arrows with direction, length and color change around the prefabricated component three-dimensional model to visualize the wind field environment;

[0053] Map the real-time calculated component state features to the preset color gradient, and apply the color gradient in the form of texture to the surface of the prefabricated component three-dimensional model in real time to visualize the internal stress distribution of the prefabricated component.

[0054] Map the generated operation instructions to state identifiers with different styles in the digital twin scene.

[0055] Continuously perform the graphical rendering process to keep the virtual state in the digital twin scene synchronized with the actual working state of the physical world.

[0056] In a second aspect, the prefabricated building intelligent construction system is used to implement the prefabricated building intelligent construction method, and includes a construction data acquisition module, a working state analysis module, a component state prediction module, a hoisting decision training module, an operation instruction issuing module, and a post-facto feedback twin module. The modules are connected by electrical signals.

[0057] The construction data acquisition module is used to acquire and preprocess wind speed data, wind direction data, and internal strain data of the prefabricated component corresponding to the hoisting working area of the prefabricated building, to obtain preprocessed wind speed data, wind direction data, and internal strain data.

[0058] The working state analysis module is used to obtain environmental influence features and component state features from the preprocessed wind speed data, wind direction data, and internal strain data.

[0059] The component state prediction module is used to input the environmental influence features and the component state features into a component sway posture physical model and a component damage risk probability model, to respectively output a theoretical maximum sway amplitude, a theoretical sway speed, and an initial damage probability of the prefabricated component.

[0060] The hoisting decision training module is used to combine the component state features, the theoretical maximum sway amplitude, the theoretical sway speed, and the initial damage probability, input them into a hoisting safety decision model, and train the hoisting safety decision model using the wind speed data, the wind direction data, and the internal strain data.

[0061] The operation instruction issuing module is used to input the environmental influence features and the component state features into the hoisting safety decision model, output a damage probability, and issue operation instructions of different risk levels based on the damage probability.

[0062] The post-facto feedback twin module is used to establish a post-facto construction database, associate hoisting actual results with output results of the hoisting safety decision model, and graphically render a hoisting trajectory of the prefabricated component.

[0063] Thanks to the above technical solutions, the present application has the following technical progress compared with the prior art:

[0064] 1. The application provides an intelligent construction method and system for fabricated buildings, which collects internal stress data of components and high-altitude micro-environment data of hoisting points in real time, conducts deep analysis by combining physical simulation and data models, and converts the previous fuzzy judgment relying on subjective experience into accurate quantitative prediction of hoisting risks, thereby fundamentally improving the data basis and scientificity of decision-making.

[0065] 2. The application provides an intelligent construction method and system for fabricated buildings, which can output graded operation instructions based on accurate risk probability, instead of simple binary alarm, thereby solving the contradiction between over-conservative impact on construction period and component damage caused by risky operation, and realizing dynamic balance and active control of safety and efficiency.

[0066] 3. The application provides an intelligent construction method and system for fabricated buildings, which feeds back actual results of each hoisting to the hoisting safety decision-making model by establishing a construction a posteriori database, realizes self-learning and continuous evolution of the system, and makes complex construction status and risks intuitive and visual by combining digital twin visual rendering, thereby comprehensively improving the traceability, management and cognitive level of the construction process. BRIEF DESCRIPTION OF DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings described in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0068] Figure 1 The flowchart of the intelligent construction method for fabricated buildings provided by the present application.

[0069] Figure 2 The structural block diagram of the intelligent construction system for fabricated buildings provided by the present application. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0071] Embodiment 1, as shown in the present application provides an intelligent construction method and system for fabricated buildings, which includes the following steps 1-6: Figure 1

[0072] ​Step 1, collect and pre-process the wind speed data, wind direction data and internal strain data of the precast component of the prefabricated building corresponding to the hoisting operation area, to obtain the pre-processed wind speed data, wind direction data and internal strain data.

[0073] In some embodiments, a three-dimensional ultrasonic wind speed and direction instrument is fixed on the boom of the tower crane performing the hoisting operation, near the hook trolley of the boom, to obtain the wind speed data and wind direction data of the precast component at the real-time operation height during the hoisting process. The three-dimensional ultrasonic wind speed and direction instrument calculates the wind speed data and wind direction data by measuring the time difference of ultrasonic wave passing through the air.

[0074] In some embodiments, during the production of the precast component prototype, an optical fiber Bragg grating sensing cable is embedded in the precast component along a pre-set stress transmission path. The lead end of the optical fiber Bragg grating sensing cable is integrated into an optical fiber adapter interface, which is fixed to the surface of the precast component. Before the hoisting operation, the optical fiber grating demodulator is connected to the optical fiber Bragg grating sensing cable through the optical fiber adapter interface. The optical fiber grating demodulator calculates the internal strain data in real time by monitoring the center wavelength shift. The center wavelength shift is generated by the optical fiber grating demodulator emitting broadband light to the optical fiber Bragg grating sensing cable, and the grating structure in the optical fiber Bragg grating sensing cable is physically deformed due to the internal strain of the precast component.

[0075] In some embodiments, the wind speed data and wind direction data are processed using a sliding time window average algorithm to smooth out instantaneous extreme fluctuations and obtain pre-processed wind speed data and wind direction data.

[0076] In some embodiments, the internal strain data is processed using a Kalman filter algorithm to filter out measurement noise and obtain pre-processed internal strain data.

[0077] Step 2, obtain the environmental influence features and component state features from the pre-processed wind speed data, wind direction data and internal strain data.

[0078] In some embodiments, the environmental influence features include the gust coefficient and the wind direction deviation angle, and the component state features include the maximum stress gradient and the stress fluctuation frequency.

[0079] In some embodiments, within a pre-set time window, the maximum wind speed value and the average wind speed value are extracted from the pre-processed wind speed data, and the gust coefficient is obtained by dividing the maximum wind speed value by the average wind speed value.

[0080] In some embodiments, the normal direction of the windward surface of the precast component is pre-set as the reference direction, and the wind direction deviation angle is the angle between the real-time wind direction shown by the pre-processed wind direction data and the reference direction.

[0081] In some embodiments, the pre-processed internal strain data is converted to stress data by presetting the material elastic modulus of the precast component The time rate of change of the stress data is calculated within a time window, and the maximum value of the absolute value of the time rate of change is extracted as the maximum stress gradient.

[0082] In some embodiments, a fast Fourier transform algorithm is applied to the converted stress data within a time window to obtain a frequency domain spectrum of the stress data, and the frequency point with the highest amplitude in the frequency domain spectrum is extracted as the stress fluctuation frequency.

[0083] Step 3, input the environmental influence characteristics and component state characteristics into the component sway posture physical model and the component damage risk probability model, respectively output the theoretical maximum sway amplitude, the theoretical sway speed and the initial damage probability of the precast component.

[0084] In some embodiments, the precast component and its hoisting cable during hoisting are simplified into a single pendulum dynamics model with the hook suspension point as the fixed rotation axis and the precast component as the pendulum bob. The inherent property parameters of the single pendulum dynamics model include the mass of the precast component, the projected area of the windward surface of the precast component , the length of the hoisting cable and the air drag coefficient .

[0085] In some embodiments, the environmental influence characteristics are input as external forcing terms of the single pendulum dynamics model architecture, and a component sway posture physical model is established to solve the component of wind force on the precast component in the swing direction , the specific calculation formula is as follows:

[0086]

[0087] Wherein, is the air density, is the real-time wind speed calculated from the gust factor, is the wind direction deviation angle.

[0088] In some embodiments, the calculated wind force is substituted into the motion differential equation of the single pendulum dynamics model, and the motion differential equation is solved by numerical method to obtain the swing angle time series of the precast component under the action of wind force.

[0089] In some embodiments, the maximum angle value is extracted from the swing angle time series, and the maximum angle value is multiplied by the length of the hoisting cable to obtain the theoretical maximum sway amplitude.

[0090] ​In some embodiments, the swing angle time series is once differentiated to obtain a swing angle velocity sequence of the prefabricated component, and the maximum value in the swing angle velocity sequence is extracted as the theoretical sway speed.

[0091] In some embodiments, a component damage risk probability model based on a gradient boosting decision tree algorithm is constructed, and the component damage risk probability model is composed of decision trees established in sequence. The component damage risk probability model is trained in advance using a labeled data set containing historical component state features and corresponding real damage results.

[0092] In some embodiments, in the component damage risk probability model training stage, a basic decision tree is initialized, and the difference between the historical damage results predicted by the component damage risk probability model and the real damage results is taken as the residual error. Each subsequent decision tree is established, and the residual error of the prediction results of the established decision tree combination is taken as the learning target. The decision tree establishment process is iterated to optimize the component damage risk probability model in the gradient direction of reducing the residual error.

[0093] In some embodiments, the maximum stress gradient and stress fluctuation frequency obtained in real time are taken as inputs and input into the trained component damage risk probability model. The input maximum stress gradient and stress fluctuation frequency pass through the established decision trees in the component damage risk probability model in sequence, so that each decision tree generates a prediction value. The weighted sum of each prediction value is taken as the initial damage probability.

[0094] Step 4: Merge the component state features, the theoretical maximum sway amplitude, the theoretical sway speed, and the initial damage probability, and input them into the hoisting safety decision model. The hoisting safety decision model is trained using wind speed data, wind direction data, and internal strain data.

[0095] In some embodiments, the component state features, the theoretical maximum sway amplitude, the theoretical sway speed, and the initial damage probability are merged into a multi-dimensional input feature vector.

[0096] In some embodiments, a historical data set for training the hoisting safety decision model is constructed. The historical data set contains multi-dimensional input feature vectors derived from historical wind speed data, wind direction data, and internal strain data, as well as real safety result categories corresponding to each hoisting operation. The safety result categories include a component intact, a component damaged, and an operation interrupted.

[0097] In some embodiments, the hoisting safety decision model is supervised trained based on a multi-layer feedforward neural network architecture. The internal parameters of the multi-layer feedforward neural network architecture include connection weights between neurons and biases of neurons. The multi-layer feedforward neural network structure includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is the same as the dimension of the multi-dimensional input feature vector, and the number of nodes in the output layer is the same as the number of preset safety result categories same.

[0098] In some embodiments, a multidimensional input feature vector is input into the hoisting safety decision model, which performs forward calculations based on internal parameters and outputs a predicted safety result.

[0099] In some embodiments, the forward computation process includes, starting from the input layer, each neuron in the subsequent layer performing a weighted summation on the output signals of the neurons in the previous layer, inputting the weighted summation result into a preset nonlinear activation function for processing, and using the processing result of the nonlinear activation function as the output signal of each neuron, which is then passed to the next layer of neurons, up to the hidden layer. After the output signal of the hidden layer is passed to the output layer, it is processed by a normalized exponential function to determine the output layer neuron's output signal. The weighted summation results Convert to the first Damage probability distribution for each safety outcome category ,in, For the first neuron in the output layer The weighted summation results are used, and the safety outcome category corresponding to the highest damage probability value is taken as the predicted safety outcome output by the hoisting safety decision model.

[0100] In some embodiments, a loss function is defined to quantify the error between the predicted security outcome and the actual security outcome. Gradient descent is used to adjust the internal parameters in reverse based on the error calculated by the loss function, so that the predicted security outcome approaches the actual security outcome.

[0101] In some embodiments, the supervised training steps are repeated until the value of the loss function converges to below a preset loss threshold. After this process is completed, a trained hoisting safety decision model is obtained.

[0102] Step 5: Input the environmental impact characteristics and component status characteristics into the hoisting safety decision model, output the damage probability, and issue operation instructions with different risk levels based on the damage probability.

[0103] In some embodiments, when performing actual hoisting operations, the environmental impact characteristics and component state characteristics acquired in real time are input into the trained hoisting safety decision model. The hoisting safety decision model uses internal parameters to perform a forward calculation process and outputs the damage probability corresponding to the predicted safety result.

[0104] In some embodiments, based on the probability of damage, operation instructions of different risk levels are issued through a preset judgment logic. The judgment logic includes a preset high-risk threshold and a low-risk threshold, wherein the high-risk threshold is greater than the low-risk threshold. The probability of damage is compared with the high-risk threshold and the low-risk threshold. The operation instructions include stopping the operation, adjusting the operation parameters, and continuing the operation with the current parameters.

[0105] In some embodiments, when the injury probability is greater than the high-risk threshold, the high-risk state is determined, the operation is stopped, when the injury probability is lower than the high-risk threshold and higher than the low-risk threshold, the medium-risk state is determined, the operation parameter is adjusted, and when the injury probability is lower than the low-risk threshold, the low-risk state is determined, and the operation is continued according to the current parameter.

[0106] Step 6, a construction posterior database is established, the actual hoisting result is associated with the output result of the hoisting safety decision model, and the hoisting trajectory of the prefabricated component is rendered graphically.

[0107] In some embodiments, after each independent hoisting operation is completed, the multi-dimensional input feature vector, the injury probability, the operation instruction and the true safety result related to the hoisting operation are associated and combined to form a posterior data record, and the posterior data record is added to the historical data set to form a construction posterior database.

[0108] In some embodiments, a digital twin scene including the fabricated building construction site, the tower crane and the prefabricated component is constructed in advance.

[0109] In some embodiments, when the hoisting operation is performed, the three-dimensional space coordinates of the hook are obtained from the control system of the tower crane, and the position of the prefabricated component three-dimensional model in the digital twin scene is updated in real time according to the three-dimensional space coordinates, so that the hoisting trajectory of the prefabricated component is dynamically drawn.

[0110] In some embodiments, the real-time obtained environmental influence features are rendered as dynamic vector arrows with direction, length and color change around the prefabricated component three-dimensional model to visualize the wind field environment.

[0111] In some embodiments, the real-time calculated component state features are mapped to a preset color gradient, and the color gradient is applied to the surface of the prefabricated component three-dimensional model in the form of texture in real time to visualize the internal stress distribution of the prefabricated component.

[0112] In some embodiments, the generated operation instruction is mapped to a state identifier with different styles in the digital twin scene.

[0113] In some embodiments, the graphical rendering process is continuously performed, so that the virtual state in the digital twin scene is kept synchronized with the actual operation state of the physical world.

[0114] Embodiment 2, as Figure 2As shown, on the basis of embodiment 1, the application also provides a technical solution: an intelligent construction system for fabricated buildings, which is used to realize an intelligent construction method for fabricated buildings, and comprises a construction data acquisition module, an operation state analysis module, a component state prediction module, a hoisting decision training module, an operation instruction issuing module and a posteriori feedback twin module, wherein the modules are connected by electrical signals;

[0115] The construction data acquisition module is used to acquire and preprocess wind speed data, wind direction data and internal strain data of the hoisting operation area of the prefabricated component of the fabricated building during hoisting, so as to obtain preprocessed wind speed data, wind direction data and internal strain data;

[0116] The operation state analysis module is used to obtain environmental influence features and component state features from the preprocessed wind speed data, wind direction data and internal strain data;

[0117] The component state prediction module is used to input the environmental influence features and the component state features into a component sway posture physical model and a component damage risk probability model, and respectively output a theoretical maximum sway amplitude, a theoretical sway speed and an initial damage probability of the prefabricated component;

[0118] The hoisting decision training module is used to combine the component state features, the theoretical maximum sway amplitude, the theoretical sway speed and the initial damage probability, input them into a hoisting safety decision model, and train the hoisting safety decision model using the wind speed data, the wind direction data and the internal strain data;

[0119] The operation instruction issuing module is used to input the environmental influence features and the component state features into the hoisting safety decision model, output a damage probability, and issue operation instructions of different risk levels based on the damage probability;

[0120] The posteriori feedback twin module is used to establish a construction posteriori database, associate hoisting actual results with output results of the hoisting safety decision model, and graphically render a hoisting trajectory of the prefabricated component.

[0121] It should be noted that the explanation and effects of the foregoing Figure 1 The explanation and effects of the method embodiment are also applicable to the method of the present embodiment, and the principles are the same, which will not be limited in the present embodiment.

[0122] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A smart construction method for prefabricated buildings, characterized in that, Includes the following steps: Collect and preprocess wind speed data, wind direction data, and internal strain data during the hoisting process of the prefabricated components of the prefabricated building in the corresponding hoisting operation area to obtain the preprocessed wind speed data, wind direction data, and internal strain data; Environmental impact characteristics and component state characteristics are obtained from the preprocessed wind speed data, wind direction data, and internal strain data; The environmental impact characteristics and the component state characteristics are input into the component swaying posture physical model and the component damage risk probability model, respectively, and the theoretical maximum swaying amplitude, theoretical swaying velocity and initial damage probability of the prefabricated component are output. The component state characteristics, the theoretical maximum sway amplitude, the theoretical sway velocity, and the initial damage probability are combined and input into the hoisting safety decision model. The hoisting safety decision model is then trained using the wind speed data, the wind direction data, and the internal strain data. The environmental impact characteristics and the component state characteristics are input into the hoisting safety decision model, the damage probability is output, and operation instructions of different risk levels are issued based on the damage probability. A post-construction verification database is established, linking the actual hoisting results with the output of the hoisting safety decision model, and the hoisting trajectory of the precast components is graphically rendered.

2. The intelligent construction method for prefabricated buildings according to claim 1, characterized in that: The process of collecting and preprocessing wind speed data, wind direction data, and internal strain data during the hoisting process of prefabricated components of prefabricated buildings in the corresponding hoisting operation area to obtain the preprocessed wind speed data, wind direction data, and internal strain data includes: A three-dimensional ultrasonic anemometer is fixed on the boom of the tower crane performing the hoisting operation, near the hook trolley of the boom, to obtain real-time wind speed and wind direction data of the prefabricated component at the working height during the hoisting process. During the casting and production of precast component templates, fiber Bragg grating sensing cables are embedded inside the precast components along a preset stress transmission path. The lead-out end of the fiber Bragg grating sensing cable is integrated into a fiber optic adapter interface, which is fixed to the surface of the precast component. Before hoisting, a fiber Bragg grating demodulator is connected to the fiber Bragg grating sensing cable through the fiber optic adapter interface. The fiber Bragg grating demodulator calculates the internal strain data in real time by monitoring the center wavelength drift. The wind speed data and wind direction data are processed using a sliding time window averaging algorithm to smooth out instantaneous extreme fluctuations and obtain preprocessed wind speed data and wind direction data; The internal strain data is processed using a Kalman filter algorithm to filter out measurement noise and obtain preprocessed internal strain data.

3. The intelligent construction method for prefabricated buildings according to claim 2, characterized in that: The process of obtaining environmental impact characteristics and component state characteristics from the preprocessed wind speed data, wind direction data, and internal strain data includes: The environmental impact characteristics include gust coefficient and wind direction angle; the component state characteristics include maximum stress gradient and stress fluctuation frequency. Within a preset time window, the maximum wind speed value and the average wind speed value are extracted from the preprocessed wind speed data, and the gust coefficient is obtained by dividing the maximum wind speed value by the average wind speed value. The normal direction of the windward surface of the prefabricated component is preset as the reference direction, and the wind direction deflection angle is the angle between the real-time wind direction shown in the preprocessed wind direction data and the reference direction; The preprocessed internal strain data is converted into stress data using the preset material elastic modulus of the precast component. Within the time window, the time change rate of the stress data is calculated, and the maximum absolute value of the time change rate is extracted as the maximum stress gradient. Within the time window, the Fast Fourier Transform algorithm is applied to the transformed stress data to obtain the frequency domain spectrum of the stress data, and the frequency point with the highest amplitude in the frequency domain spectrum is extracted as the stress fluctuation frequency.

4. The intelligent construction method for prefabricated buildings according to claim 3, characterized in that: The process of inputting the environmental impact characteristics into the physical model of the component's swaying posture and outputting the theoretical maximum swaying amplitude and theoretical swaying velocity of the prefabricated component includes: The prefabricated component and its hoisting cable during the hoisting process are simplified into a simple pendulum dynamic model with the hook suspension point as the fixed axis of rotation and the prefabricated component as the pendulum. The inherent property parameters of the simple pendulum dynamic model include the mass of the prefabricated component, the projected area of ​​the windward side of the prefabricated component, the length of the hoisting cable, and the air drag coefficient. The environmental impact characteristics are used as external forcing terms input to the pendulum dynamics model architecture to establish a physical model of the component swaying attitude, which is used to calculate the wind force component of the prefabricated component in the swaying direction. Substitute the calculated wind force into the motion differential equation of the simple pendulum dynamic model, and solve the motion differential equation by numerical method to obtain the time series of the swing angle of the prefabricated component under the action of the wind force. The maximum angle value is extracted from the swing angle time series, and the maximum angle value is multiplied by the length of the hoisting cable to obtain the theoretical maximum sway amplitude; The swing angle time series is differentiated once to obtain the swing angular velocity series of the precast component, and the maximum value in the swing angular velocity series is extracted as the theoretical swaying velocity.

5. The intelligent construction method for prefabricated buildings according to claim 3, characterized in that: The process of inputting the component state characteristics into the component damage risk probability model and outputting the initial damage probability of the prefabricated component includes: A component damage risk probability model based on gradient boosting decision tree algorithm is constructed. The component damage risk probability model consists of decision trees built in sequence and is trained in advance using a labeled dataset containing historical component state features and corresponding real damage results. During the training phase of the component damage risk probability model, a base decision tree is initialized. The difference between the historical damage results predicted by the component damage risk probability model and the actual damage results is used as the residual. The establishment of each subsequent decision tree takes the residual of the prediction results of the combination of established decision trees as the learning objective. The decision tree establishment process is iterated to optimize the component damage risk probability model in the direction of reducing the residual. The maximum stress gradient and the stress fluctuation frequency, which are acquired in real time, are used as inputs and fed into the trained component damage risk probability model. The input maximum stress gradient and the stress fluctuation frequency are passed sequentially through the established decision trees within the component damage risk probability model, so that each decision tree generates a prediction value. The weighted sum of each prediction value is the initial damage probability.

6. The intelligent construction method for prefabricated buildings according to claim 2, 3, 4 or 5, characterized in that: The process of combining the component state characteristics, the theoretical maximum sway amplitude, the theoretical sway velocity, and the initial damage probability, inputting them into the hoisting safety decision model, and training the hoisting safety decision model using the wind speed data, the wind direction data, and the internal strain data includes: The component state features, the theoretical maximum sway amplitude, the theoretical sway velocity, and the initial damage probability are combined into a multi-dimensional input feature vector; Construct a historical dataset for training the hoisting safety decision model. The historical dataset contains the multidimensional input feature vector derived from historical wind speed data, wind direction data, and internal strain data, as well as the actual safety result category corresponding to each hoisting operation. The safety result category includes component intact, component damaged, and operation interrupted. The hoisting safety decision model is trained under supervision based on a multi-layer feedforward neural network architecture. The internal parameters of the multi-layer feedforward neural network architecture include the connection weights between neurons and the bias of each neuron. The multi-layer feedforward neural network structure includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is the same as the dimension of the multi-dimensional input feature vector, and the number of nodes in the output layer is the same as the number of preset safety result categories. The multidimensional input feature vector is input into the hoisting safety decision model, and the hoisting safety decision model performs forward calculations based on the internal parameters and outputs the predicted safety result. The forward computation process includes, starting from the input layer, each neuron in the subsequent layer performs a weighted summation on the output signal of the neuron in the previous layer, inputs the weighted summation result into a preset nonlinear activation function for processing, and the processing result of the nonlinear activation function is used as the output signal of each neuron and passed to the next layer of neurons, up to the hidden layer. After the output signal of the hidden layer is passed to the output layer, it is processed by a normalized exponential function to convert the weighted summation result of the output layer neurons into the damage probability distribution of the safety result category, and the safety result category corresponding to the highest damage probability value is used as the predicted safety result output by the hoisting safety decision model. A loss function is defined to quantify the error between the predicted security result and the actual security result. The gradient descent method is used to adjust the internal parameters in reverse according to the error calculated by the loss function, so that the predicted security result approaches the actual security result. Repeat the supervised training steps until the value of the loss function converges to below the preset loss threshold. After this process is completed, the trained hoisting safety decision model is obtained.

7. The intelligent construction method for prefabricated buildings according to claim 6, characterized in that: The process of inputting the environmental impact characteristics and the component state characteristics into the hoisting safety decision model, outputting the damage probability, and issuing operation instructions of different risk levels based on the damage probability includes: During actual hoisting operations, the environmental impact features and component state features acquired in real time are input into the hoisting safety decision model that has been trained. The hoisting safety decision model uses the internal parameters to perform the forward calculation process and outputs the damage probability corresponding to the predicted safety result. Based on the damage probability, operation instructions of different risk levels are issued through preset judgment logic. The judgment logic includes preset high-risk threshold and low-risk threshold, wherein the high-risk threshold is greater than the low-risk threshold. The damage probability is compared with the high-risk threshold and the low-risk threshold. The operation instructions include stopping the operation, adjusting the operation parameters, and continuing the operation with the current parameters. When the probability of damage is greater than the high-risk threshold, the operation is determined to be in a high-risk state and the operation is stopped. When the probability of damage is lower than the high-risk threshold but higher than the low-risk threshold, the operation is determined to be in a medium-risk state and the operation parameters are adjusted. When the probability of damage is lower than the low-risk threshold, the operation is determined to be in a low-risk state and the operation continues with the current parameters.

8. The intelligent construction method for prefabricated buildings according to claim 7, characterized in that: The process of establishing a post-construction verification database and linking the actual hoisting results with the output of the hoisting safety decision model includes: After each independent hoisting operation is completed, the multidimensional input feature vector, the damage probability, the operation command and the actual safety result related to that hoisting operation are associated and combined into a posterior data record. The posterior data record is added to the historical dataset to form the construction posterior database.

9. The intelligent construction method for prefabricated buildings according to claim 7, characterized in that: The process of graphically rendering the hoisting trajectory of the prefabricated component includes: A digital twin scenario is pre-constructed, including the prefabricated building construction site, the tower crane, and the prefabricated components. During hoisting operations, the three-dimensional spatial coordinates of the hook are obtained from the control system of the tower crane, and the position of the three-dimensional model of the prefabricated component in the digital twin scene is updated in real time based on the three-dimensional spatial coordinates, thereby dynamically drawing the hoisting trajectory of the prefabricated component. The environmental impact features acquired in real time are rendered as dynamic vector arrows with varying directions, lengths, and colors around the 3D model of the prefabricated component to visualize the wind field environment. The component state characteristics calculated in real time are mapped to a preset color gradient, and the color gradient is applied to the surface of the three-dimensional model of the prefabricated component in real time in the form of a texture to visualize the internal stress distribution of the prefabricated component. The generated operation instructions are mapped to state identifiers with different styles in the digital twin scenario; The graphical rendering process is executed continuously to keep the virtual state in the digital twin scene synchronized with the actual working state in the physical world.

10. An intelligent construction system for prefabricated buildings, used to implement the intelligent construction method for prefabricated buildings as described in any one of claims 1-9, characterized in that: It includes a construction data acquisition module, an operation status analysis module, a component status prediction module, a hoisting decision training module, an operation instruction issuance module, and a post-feedback twin module, wherein the modules are connected by electrical signals. The construction data acquisition module is used to collect and preprocess wind speed data, wind direction data, and internal strain data during the hoisting process of the prefabricated components of the prefabricated building corresponding hoisting operation area, so as to obtain the preprocessed wind speed data, wind direction data, and internal strain data. The operational status analysis module is used to obtain environmental impact characteristics and component status characteristics from the preprocessed wind speed data, wind direction data, and internal strain data; The component state prediction module is used to input the environmental impact characteristics and the component state characteristics into the component swaying posture physical model and the component damage risk probability model, and output the theoretical maximum swaying amplitude, theoretical swaying velocity and initial damage probability of the prefabricated component, respectively. The hoisting decision training module is used to combine the component state characteristics, the theoretical maximum sway amplitude, the theoretical sway velocity and the initial damage probability, input them into the hoisting safety decision model, and use the wind speed data, the wind direction data and the internal strain data to train the hoisting safety decision model; The operation instruction issuing module is used to input the environmental impact characteristics and the component status characteristics into the hoisting safety decision model, output the damage probability, and issue operation instructions of different risk levels based on the damage probability. The post-hoc feedback twin module is used to establish a post-construction database, link the actual hoisting results with the output results of the hoisting safety decision model, and graphically render the hoisting trajectory of the prefabricated components.

Images