Glass curtain wall monitoring system and method based on machine learning

Abstract

The invention discloses a glass curtain wall monitoring system and method based on machine learning, and the system comprises an acquisition module which comprises a first acquisition unit for acquiring the real-time wind pressure value of a glass curtain wall and a second acquisition unit for acquiring the strain value of the glass curtain wall; the modeling module is connected to the acquisition module, and the modeling module establishes an agent model so as to output predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and the strain value; the time sequence analysis module is connected to the acquisition module, and the time sequence analysis module enables the real-time wind pressure values to form a time sequence and adopts a dynamic mode to decompose a dynamic change rule of modeling wind pressure so as to obtain a predicted wind pressure value; and the early warning module comprises a first early warning unit, a second early warning unit and a third early warning unit, the first early warning unit is connected to the acquisition module, the second early warning unit is connected to the modeling module, and the third early warning unit is connected to the time sequence analysis module. The problem that the cost of monitoring curtain wall glass deformation by adopting methods such as machine vision and the like of a glass curtain wall is high is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building construction, and particularly to a glass curtain wall monitoring system and method based on machine learning. Background Art

[0002] With the increasing popularity of glass curtain walls in the fields of architecture and the like, in order to ensure their safe and reliable operation, real-time monitoring of glass curtain walls has become a new technical requirement. Among the more critical monitoring parameters are wind pressure, strain, and displacement deformation. How to obtain safety monitoring data such as wind pressure, strain, and deformation displacement more economically and simply has become a technical requirement point. Compared with monitoring quantities such as wind pressure and strain, the detection of its deformation displacement has a certain degree of difficulty. Detecting glass deformation by methods such as machine vision is costly, easily damaged, and not easy to maintain. Summary of the Invention

[0003] To overcome the defects of the prior art, there is provided a glass curtain wall monitoring system and method based on machine learning to solve the problem of high cost of monitoring the deformation of curtain wall glass by methods such as machine vision.

[0004] To achieve the above object, there is provided a glass curtain wall monitoring system based on machine learning, including:

[0005] An acquisition module, including a first acquisition unit for acquiring the real-time wind pressure value of the glass curtain wall and a second acquisition unit for acquiring the strain value of the glass curtain wall;

[0006] A modeling module, connected to the acquisition module, the modeling module uses a physics-informed neural network to establish a surrogate model, and through training and optimization, the surrogate model is made to output the predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and the strain value;

[0007] A time series analysis module, connected to the acquisition module, the time series analysis module forms a time series of the real-time wind pressure values and uses dynamic mode decomposition to model the dynamic change law of the wind pressure to obtain the predicted wind pressure value;

[0008] An early warning module, including a first early warning unit for generating a first alarm signal based on the comparison of the real-time wind pressure value with a wind pressure threshold, a second early warning unit for generating a second alarm signal based on the comparison of the predicted deformation displacement with a deformation threshold, and a third early warning unit for generating a third alarm signal based on the comparison of the predicted wind pressure value with the wind pressure threshold. The first early warning unit is connected to the acquisition module, the second early warning unit is connected to the modeling module, and the third early warning unit is connected to the time series analysis module.

[0009] Further, it further includes a preprocessing module for cleaning, missing value processing, and normalization processing of the real-time wind pressure value and the strain value, and the preprocessing module is connected to the acquisition module and the modeling module.

[0010] Further, the acquisition module further includes a storage unit, and the storage unit is connected to the first acquisition unit and the second acquisition unit.

[0011] Further, the warning module further includes a signal pushing unit, and the signal pushing unit is connected to the first warning unit, the second warning unit, and the third warning unit.

[0012] The present invention provides a glass curtain wall monitoring method for a glass curtain wall monitoring system based on machine learning, including the following steps:

[0013] The first acquisition unit of the acquisition module acquires the real-time wind pressure value of the glass curtain wall. Meanwhile, the second acquisition unit of the acquisition module acquires the strain value of the glass curtain wall;

[0014] The modeling module uses a physics-informed neural network to establish a surrogate model, and through training and optimization, the surrogate model is made to output the predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and the strain value;

[0015] The time series analysis module acquires the actual wind pressure value, forms the real-time wind pressure value into a time series, and uses dynamic mode decomposition to model the dynamic change law of the wind pressure to obtain the predicted wind pressure value;

[0016] The first warning unit of the warning module generates a first alarm signal based on the comparison of the real-time wind pressure value with the wind pressure threshold, the second warning unit of the warning module generates a second alarm signal based on the comparison of the predicted deformation displacement with the deformation threshold, and the third warning unit of the warning module generates a third alarm signal based on the comparison of the predicted wind pressure value with the wind pressure threshold.

[0017] The beneficial effect of the present invention is that the glass curtain wall monitoring system based on machine learning of the present invention uses sensors to extract the strain distribution characteristics on the surface of the curtain wall, and combines with a physics-informed neural network to establish a surrogate model of the elastic mechanics control equation for the stress and deformation of the glass plate, and uses this as the surrogate model (the input quantities are the wind pressure value and the strain value, and the output quantity is the deformation displacement). The glass curtain wall monitoring system based on machine learning of the present invention establishes a surrogate physical model with the wind pressure and strain as input quantities and the deformation displacement as the output quantity through machine learning, and finally realizes obtaining the key parameters of the wind pressure, strain, and deformation displacement of the glass curtain wall through wind pressure and strain measurement, which can reduce the additional measurement system required for deformation displacement measurement and can greatly reduce the installation cost and maintenance cost. Description of the Drawings

[0018] Other features, objectives, and advantages of the present application will become more apparent by reading the following detailed description of non - restrictive embodiments with reference to the accompanying drawings:

[0019] Figure 1 It is a schematic structural diagram of a machine - learning - based glass curtain wall monitoring system according to an embodiment of the present invention. Detailed implementation manners

[0020] The present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention, rather than limiting the invention. Additionally, it should be noted that, for the sake of convenience of description, only the parts related to the invention are shown in the drawings.

[0021] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0022] Referring to Figure 1 As shown, the present invention provides a machine - learning - based glass curtain wall monitoring system, including: a collection module 1, a modeling module 2, a time - series analysis module 3, and an early - warning module 4.

[0023] In this embodiment, the collection module 1 includes a first collection unit 11 and a second collection unit 12. Among them, the first collection unit 11 is used to collect the real - time wind pressure value of the glass curtain wall. The second collection unit 12 is used to collect the strain value of the glass curtain wall.

[0024] The first collection unit 11 is a wind pressure sensor. The second collection unit 12 is a strain sensor.

[0025] As a preferred implementation manner, the collection module 1 further includes a storage unit 13, and the storage unit is connected to the first collection unit 11 and the second collection unit 12.

[0026] The first collection unit 11 and the second collection unit 12 are deployed on the surface of the glass curtain wall, and they collect the strain distribution and wind pressure data on the surface of the curtain wall in real - time and store them in the storage unit.

[0027] The modeling module 2 is connected to the collection module 1. The modeling module 2 uses a physics - informed neural network to establish a surrogate model, and through training and optimization, the surrogate model outputs the predicted deformation displacement of the glass curtain wall based on the real - time wind pressure value and strain value.

[0028] The modeling module takes the measured wind pressure value and the strain value of the glass curtain wall distribution as the input of the model, uses a physics - informed neural network (PINN) to establish a surrogate model, and through training and optimization, the model can accurately predict the deformation displacement from the pressure distribution.

[0029] As a preferred embodiment, the machine learning-based glass curtain wall monitoring system of the present invention further includes a preprocessing module 5. The preprocessing module is connected to the acquisition module 1 and the modeling module 2. The preprocessing module 5 is used to clean, handle missing values, and normalize the real-time wind pressure value and strain value.

[0030] The time series analysis module 3 is connected to the acquisition module 1. The time series analysis module 3 forms a time series with the real-time wind pressure values and uses dynamic mode decomposition to model the dynamic change law of the wind pressure to obtain the predicted wind pressure value.

[0031] Specifically, the time series analysis module forms a time series with the obtained real-time wind pressure values. By using dynamic mode decomposition (DMD), the wind pressure value after a short future time interval can be predicted relatively reliably. Then, it is input into the trained physics-informed neural network (PINN) to obtain the predicted wind pressure value, completing the prediction of the wind pressure and deformation displacement in the short future.

[0032] The time series analysis module includes a data processing unit and a time series prediction unit.

[0033] The data processing unit forms time series data of the real-time wind pressure values collected by the acquisition module in chronological order.

[0034] The time series prediction unit uses dynamic mode decomposition (DMD) technology to model the wind pressure time series and estimates the model parameters by using methods such as singular value decomposition and least squares method.

[0035] The early warning module 4 includes a first early warning unit 41, a second early warning unit 42, and a third early warning unit 43. The first early warning unit 41 is connected to the acquisition module 1, the second early warning unit 42 is connected to the modeling module 2, and the third early warning unit 43 is connected to the time series analysis module 3.

[0036] Based on the aforementioned trained machine learning model, predicting the future wind pressure change of the curtain wall in real time, when the predicted wind pressure value exceeds the safety threshold, the early warning module timely issues an early warning signal.

[0037] Specifically, the first early warning unit 41 generates a first alarm signal based on the comparison between the real-time wind pressure value and the wind pressure threshold.

[0038] The second early warning unit 42 generates a second alarm signal based on the comparison between the predicted deformation displacement and the deformation threshold.

[0039] The third early warning unit 43 generates a third alarm signal based on the comparison between the predicted wind pressure value and the wind pressure threshold.

[0040] As a preferred embodiment, the early warning module 4 further includes a signal pushing unit 44. The signal pushing unit is connected to the first early warning unit 41, the second early warning unit 42, and the third early warning unit 43.

[0041] An information push unit that promptly pushes the alarm signal to the glass curtain wall management personnel and takes corresponding emergency measures. The glass curtain wall management personnel take corresponding response measures for the glass curtain wall according to the alarm signal.

[0042] The present invention provides a glass curtain wall monitoring method for a glass curtain wall monitoring system based on machine learning, including the following steps:

[0043] S1. The first acquisition unit of acquisition module 1 acquires the real-time wind pressure value of the glass curtain wall. Meanwhile, the second acquisition unit 12 of acquisition module 1 acquires the strain value of the glass curtain wall.

[0044] S2. Modeling module 2 uses a physics-informed neural network to establish a surrogate model, and through training and optimization, makes the surrogate model output the predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and strain value.

[0045] The data processing module performs necessary cleaning, missing value processing, and normalization operations on the data collected by the acquisition module.

[0046] The modeling module uses the data of the preprocessing module to establish a neural network surrogate physical model.

[0047] S3. The time series analysis module 3 obtains the actual wind pressure value, forms the real-time wind pressure value into a time series, and uses dynamic mode decomposition to model the dynamic change law of the wind pressure to obtain the predicted wind pressure value.

[0048] S4. The first warning unit of warning module 4 generates a first alarm signal based on the comparison of the real-time wind pressure value and the wind pressure threshold. The second warning unit 42 of warning module 4 generates a second alarm signal based on the comparison of the predicted deformation displacement and the deformation threshold. The third warning unit 43 of warning module 4 generates a third alarm signal based on the comparison of the predicted wind pressure value and the wind pressure threshold.

[0049] The glass curtain wall monitoring system based on machine learning of the present invention uses sensors to extract the strain distribution characteristics on the surface of the curtain wall, combines with a physics-informed neural network, establishes a surrogate model of the elastic mechanics control equation for the force deformation of the glass plate, and uses this as the surrogate model (the input quantities are the wind pressure value and the strain value, and the output quantity is the deformation displacement). The glass curtain wall monitoring system based on machine learning of the present invention establishes a surrogate physical model with the wind pressure and strain as input quantities and the deformation displacement as the output quantity through machine learning, and finally realizes obtaining the key parameters of the wind pressure, strain, and deformation displacement of the glass curtain wall by measuring the wind pressure and strain, which can reduce the additional measurement system required for measuring the deformation displacement and can greatly reduce the installation cost and maintenance cost.

[0050] The machine learning-based glass curtain wall monitoring system of the present invention can predict the deformation displacement of the glass curtain wall only through wind pressure and strain measurements, greatly simplifying the measurement system, reducing the generation and maintenance costs, and also enabling short-term prediction and early warning of future wind pressure.

[0051] The machine learning-based glass curtain wall monitoring system of the present invention installs strain sensors on the windward side of the glass and differential pressure wind pressure sensors for measuring the differential wind pressure on both sides of the glass curtain wall. The machine learning-based glass curtain wall monitoring system of the present invention trains a surrogate physical equation (a neural network with wind pressure and strain as input quantities and strain and displacement as output quantities) based on the physics-informed neural network (PINN) of the glass elasticity mechanics equation. The deformation displacement of the glass curtain wall is given by this neural network, without the need to install other sensors for measuring deformation displacement, making the entire measurement device simple, low-cost, and easy to maintain.

[0052] The machine learning-based glass curtain wall monitoring system of the present invention can predict the wind pressure in the short term in the future: through the historical and current wind pressure values, the trend of the wind pressure in the short term in the future can be predicted, and thus the predicted wind pressure is used as the input quantity of the physical model obtained by machine learning to calculate the deformation displacement in the short term in the future.

[0053] The above description is only the preferred embodiment of the present application and the explanation of the applied technical principles. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the technical solutions formed by mutually replacing the above features with the (but not limited to) technical features with similar functions disclosed in the present application.

Claims

1. A glass curtain wall monitoring system based on machine learning, characterized in that: include: A collection module, comprising a first collection unit for collecting real-time wind pressure values ​​of the glass curtain wall and a second collection unit for collecting strain values ​​of the glass curtain wall; A modeling module, connected to the acquisition module, wherein the modeling module uses a physical information neural network to establish an agent model, and through training and optimization, the agent model is trained to output a predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and the strain value; A time series analysis module, connected to the acquisition module, which combines the real-time wind pressure values ​​into a time series and uses dynamic mode decomposition to model the dynamic change law of wind pressure to obtain a predicted wind pressure value; The early warning module includes a first early warning unit for generating a first alarm signal based on a comparison between the real-time wind pressure value and the wind pressure threshold, a second early warning unit for generating a second alarm signal based on a comparison between the predicted deformation displacement and the deformation threshold, and a third early warning unit for generating a third alarm signal based on a comparison between the predicted wind pressure value and the wind pressure threshold, wherein the first early warning unit is connected to the acquisition module, the second early warning unit is connected to the modeling module, and the third early warning unit is connected to the timing analysis module.

2. The glass curtain wall monitoring system based on machine learning according to claim 1 is characterized in that: It also includes a preprocessing module for cleaning, processing missing values ​​and normalizing the real-time wind pressure value and the strain value, and the preprocessing module is connected to the acquisition module and the modeling module.

3. The glass curtain wall monitoring system based on machine learning according to claim 1 is characterized in that: The acquisition module further includes a storage unit, and the storage unit is connected to the first acquisition unit and the second acquisition unit.

4. The glass curtain wall monitoring system based on machine learning according to claim 1, characterized in that: The early warning module further includes a signal pushing unit, and the signal pushing unit is connected to the first early warning unit, the second early warning unit and the third early warning unit.

5. A glass curtain wall monitoring method of a glass curtain wall monitoring system based on machine learning as claimed in any one of claims 1 to 4, characterized in that: The following steps are involved: The first acquisition unit of the acquisition module acquires the real-time wind pressure value of the glass curtain wall, and at the same time, the second acquisition unit of the acquisition module acquires the strain value of the glass curtain wall; The modeling module uses a physical information neural network to establish an agent model, and through training and optimization, the agent model is made to output the predicted deformation displacement of the glass curtain wall based on the real-time wind pressure value and the strain value; The time series analysis module obtains the implemented wind pressure value, and composes the real-time wind pressure value into a time series and uses dynamic mode decomposition to model the dynamic change law of wind pressure to obtain the predicted wind pressure value; The first warning unit of the warning module generates a first alarm signal based on the comparison between the real-time wind pressure value and the wind pressure threshold, the second warning unit of the warning module generates a second alarm signal based on the comparison between the predicted deformation displacement and the deformation threshold, and the third warning unit of the warning module generates a third alarm signal based on the comparison between the predicted wind pressure value and the wind pressure threshold.

Images