Dynamic reliability analysis method of double-veneered hollow glass under wind load

Abstract

This invention provides a dynamic reliability analysis method for double-laminated insulated glass under wind loads, belonging to the field of stress analysis technology for building curtain wall glass. This method proposes a dynamic reliability analysis approach for double-laminated insulated glass under wind loads. First, a finite element analysis model of the double-laminated insulated glass is established, considering the coupling problem between the interlayer structure and the flow field within the cavity. The differences between the finite element model and experimental results, as well as the differences between the sub-parts of the finite element model and the numerical algorithm, are compared. Then, the structural response under two common wind loads is calculated, and a dynamic reliability analysis process is constructed. This solves the problem that current double-laminated insulated glass design relies entirely on empirical formulas, and the finite element model is interpretable. The accuracy of the finite element model is verified from both theoretical and experimental perspectives. A comprehensive wind load damage model is established, comprehensively considering the influence of typhoons and natural winds.

Description

Technical Field

[0001] This invention belongs to the field of stress analysis of building curtain wall glass, specifically relating to a dynamic reliability analysis method for double-laminated insulated glass under wind load. Background Technology

[0002] Curtain wall glass is widely used on the surfaces of modern buildings. Based on the different connection methods between the glass panes, glass is broadly classified into three categories: single-pane glass, laminated glass, and insulated glass. Laminated glass is typically formed by bonding two panes of glass with a polymer, effectively reducing the fragmentation upon breakage. Insulated glass is formed by connecting two panes of glass with a spacer and a gas cavity, improving stress distribution and preventing fogging. Existing research indicates that wind-induced damage is the primary cause of failure for all types of curtain wall glass. The Northwest Pacific region is frequently affected by typhoons, leading to interest in double-laminated insulated glass units that combine the advantages of both laminated and insulated glass, reducing damage caused by typhoons.

[0003] In engineering projects, the failure of curtain wall glass can lead to serious safety problems and high repair costs. Among these factors, the change in failure probability over time is particularly important. However, estimating the failure probability of curtain wall glass relies heavily on engineers' experience and is difficult to quantify.

[0004] There is a lack of explicit numerical and finite element models for double-laminated insulated glass. Laminated glass and insulated glass can be considered as two sub-components of double-laminated insulated glass, and previous research has provided references and explanations for models of double-laminated insulated glass. In recent years, scholars both domestically and internationally have studied the mechanical response of glass under wind loads. Previous studies mainly focused on numerical models and experimental tests of laminated glass, providing explanations for the rationality of the models. Some scholars have constructed a sandwich model to analyze the mechanical response of each layer of laminated glass and given the assumptions used in the numerical model. Considering the influence of polymer interlayers on the strength of laminated glass, some scholars have proposed a method for calculating the effective thickness and compared it with classical methods. Some scholars have studied the load-sharing effect of the gas in the cavity of insulated glass, believing that this structure improves the stress distribution of the glass, and proposed an approximate method for the equivalent thickness of insulated glass considering the load-sharing effect based on the finite element method. However, merging the numerical models of laminated glass and insulated glass from previous studies is challenging because it leads to complex derivation processes.

[0005] Existing research indicates that wind load is a significant cause of failure in various types of curtain wall glass. Brownian integral was initially used to describe the cumulative damage caused by wind load on glass. Researchers have tested glass panels under time-varying wind loads and slope loads, verifying the rationale behind Brownian integral. Other researchers have analyzed the impact of wind pressure on building exterior surfaces using equivalent 1-minute loads, proposing a method for calculating the optimal glass thickness using simulated wind fields, while also considering the uncertainty of glass resistance.

[0006] Although some scholars have studied the wind-induced damage and failure behavior of glass, they have neglected the influence of random factors. Glass parameters are uncertain due to various random factors during manufacturing and processing. For wind loads, both strong winds (short-term high wind speeds) and natural winds (long-term low wind speeds) are uncertain in terms of both wind speed and frequency. Dynamic reliability analysis can comprehensively consider structural and load uncertainties. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this application proposes a dynamic reliability analysis method for double-laminated insulated glass under wind loads, which enables accurate prediction of glass stress and provides a theoretical basis for reducing the probability of glass failure under wind loads and mitigating safety hazards in building curtain wall glass.

[0008] This application proposes a dynamic reliability analysis method for double-laminated insulated glass under wind loads. First, a finite element analysis model of the double-laminated insulated glass is established, considering the coupling between the interlayer structure and the flow field within the cavity. The differences between the finite element model and experimental results, as well as the differences between sub-parts of the finite element model and the numerical algorithm, are compared. Then, the structural response under two common wind loads is calculated, and a dynamic reliability analysis process is constructed.

[0009] A method for dynamic reliability analysis of double-laminated insulated glass under wind load includes the following steps:

[0010] Step S1: Based on the structural form, size, and constraint method of double-laminated insulated glass, a finite element analysis model of double-laminated insulated glass is established using finite element simulation software; the finite element analysis model is divided into a laminated glass module and an insulated glass module.

[0011] The double-laminated insulated glass includes laminated glass and insulated glass, wherein the laminated glass and insulated glass correspond to the laminated glass module and the insulated glass module in the finite element analysis model, respectively.

[0012] Step S2: Verify the accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load through experiments, and optimize the finite element model based on the experimental results;

[0013] The accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load was verified by experiments, specifically using a 3D scanning system, a displacement measurement system, and a strain measurement system.

[0014] The 3D scanning system measures the installation status of the glass; based on the scanning results of the corner points of the double-laminated insulated glass, the glass position is adjusted to minimize the deviation.

[0015] The displacement measurement system measures the displacement value of double-laminated insulated glass. Specifically, the displacement sensor is installed on a rigid support independent of the double-laminated insulated glass and connected to a computer through a fiber optic adapter box.

[0016] The strain measurement system includes a fiber Bragg grating sensor and a grating demodulator. The fiber Bragg grating sensor is attached to the experimental measurement point at 45° intervals. The fiber Bragg grating sensor is connected to the grating demodulator and a computer via wires. Temperature compensation and the initial wavelength of the fiber Bragg grating sensor are set to zero the initial strain value. The first principal stress σ1 at the measurement point is expressed as:

[0017] ;

[0018] Where E is the elastic modulus of the glass, μ is the Poisson's ratio of the glass, and ε0, ε 45 ,ε 90 These are strain measurements taken by a fiber Bragg grating sensor in three directions spaced 45° apart.

[0019] During the experiment, the building's interior was connected to a smart pressure sensor, an electric air blower, and a central control room via pipelines. Air was extracted from the building to simulate wind load. Double-laminated insulated glass samples were mounted on an aluminum alloy sealing structure on the building's surface using aluminum panels, ensuring consistency with the boundary conditions of the finite element model. The power of the electric air blower was adjusted to simulate the effect of wind load. Once the smart pressure sensor reading reached the required level, the power of the electric air blower was kept constant, and the displacement and strain values ​​were recorded. The stress changes caused by the wind load under these conditions were then calculated.

[0020] Step S3: Analyze the strength degradation and cumulative damage caused by typhoon insulated glass;

[0021] The intensity degradation is achieved through the intensity degradation function R. 60 (n) is determined, specifically:

[0022] ;

[0023] Among them, R 60 R represents the initial load capacity. 60 (n) represents the load capacity in year n, and α(n,k1) is a deterministic function that varies with time. TY represents the design life of the glass, and the proportional parameter k1 and the exponential parameter γ are determined experimentally.

[0024] The cumulative damage caused by the typhoon is specifically as follows:

[0025] ;

[0026] In the formula, Let S be the cumulative damage caused by the typhoon in year i, j = 1, 2, ..., i, and D1(S) be the glass damage caused by typhoon wind speed S, calculated using wind-induced glass damage; S = {S1, S2, ..., S...} N} T This represents a series of samples representing the distribution of typhoon wind speeds; the typhoon wind speeds S are sorted by occurrence time and represented using A. j and n j Group and name them as follows: to The following are the wind-induced glass damages: T represents the duration of glass stress, s represents the wind speed, σ(•,t) represents the stress as a function of time t, 16 is a constant related to fracture mechanics, and D(s) is the cumulative damage amount.

[0027] Step S4: Based on the Dalbert spectrum, consider the glass damage caused by natural wind to obtain the total cumulative damage from wind load;

[0028] The annual damage caused by natural wind in year t, calculated using Bernoulli's principle and Brownian integral, is represented as D2. t (u(z)), where the cumulative total damage from natural wind in year i is represented by CD2. i :

[0029] ;

[0030] In the formula, u(z) is the wind speed under natural wind load;

[0031] Taking into account the combined effects of typhoons and natural winds, the cumulative total wind load damage in year i is represented as CD. i= CD1 i +CD2 i ;

[0032] Step S5: Perform dynamic reliability analysis on double-laminated insulated glass;

[0033] The main random variables X considered include:

[0034] ;

[0035] Where L is the length of the glass, W is the width of the glass, E is the elastic modulus of the glass, μ is the Poisson's ratio of the glass, h is the thickness of the glass, and s t It is the cavity thickness;

[0036] The dynamic failure probability of the curtain wall glass is expressed as:

[0037] ;

[0038] ;

[0039] Among them, t s Here, is the design life of the curtain wall glass, CD(X,t) is the cumulative damage to the glass under wind load at time t, RD(t) is the degraded strength, G(X,t) is the limit state function, and Prob represents the probability; the failure region and the safety region are respectively represented as:

[0040] ;

[0041] The dynamic failure probability is solved by dividing the region using the composite limit state method, specifically including the following steps:

[0042] Step A1: Based on the typhoon wind speed distribution probability, a randomly arranged wind speed sample vector V is obtained using SOBO sequence sampling. The design life of the curtain wall glass is t. s .

[0043] Step A2: Simulate N random processes using the quasi-Monte Carlo method. The result is represented by matrix YN, which represents the number of typhoons that occur each year in the N simulations.

[0044] Step A3: Extract vector V sequentially based on the values ​​of matrix YN. Augment the missing typhoon counts compared to the maximum annual typhoon count with 0 to obtain matrix YNS. Therefore, matrix YNS simultaneously represents the frequency and intensity of typhoons. In the i-th simulation, t... j The typhoon wind speed vector occurring within a time period is represented by S. i Similarly, the number of missing values ​​is augmented with 0, and the sum of the damage corresponding to each column of matrix S is t. j The cumulative damage over time.

[0045] Step A4: Sample the random variable X according to the probability model, for a total of N. x Group;

[0046] Step A5: Calculate the wind pressure corresponding to the wind speed according to Bernoulli's principle;

[0047] Step A6: Considering the influence of the introduced random variable, the sample size is N. k The sample {X k}, k=1,…, N k N k < <N x ;

[0048] Step A7: Use the inner layer to build an adaptive proxy model G k (X * ,t), where X * ∈{X k},t∈[0,t s The calculation of the function G refers to the formula for cumulative total damage from wind load;

[0049] Step A7.1: Time region [0, t] s Discretized into discrete time points {t} * The training sample points are represented as S. k ={t * ,G(X * ,t * )}.

[0050] Step A7.2: Establish the agent model G k (X * ,t);

[0051] Step A7.3: Calculate the learning function E[I(t)] based on the predicted values ​​of the surrogate model;

[0052] ;

[0053] Where Ф(.) and φ(.) represent the standard normal cumulative distribution function and probability density function, respectively;

[0054] Step A7.4: The point of maximum improvement is added to the training samples, and the surrogate model is retrained; the point of maximum improvement t new Represented as

[0055] ;

[0056] Step A7.5: Learn the convergence criterion of the function:

[0057] ;

[0058] Step A8: In the middle layer, calculate the minimum value G of the function. e (X * ):

[0059] ;

[0060] Step A9: The training sample points of the middle layer are represented as S e ={X k G e (X k )}.

[0061] Step A10: Establish an adaptive agent model;

[0062] Step A11: Calculate the learning function C based on the predictions from the adaptive surrogate model. A (X);

[0063] ;

[0064] Where f X(x) is the joint probability density function of random variables;

[0065] Step A12: The point of maximum improvement is added to the training samples, and the adaptive agent model is retrained. Maximum improvement point X new Represented as:

[0066] ;

[0067] Step A13: The convergence criterion for the learning function is:

[0068] ;

[0069] Among them, C R This is the confidence level, set to 0.9999;

[0070] Step A14: Calculate the failure probability in the outer layer, the failure probability P of the i-th simulation. tf (P t |n=i) is represented as:

[0071] ;

[0072] in:

[0073] ;

[0074] Step A15: According to the conditional probability formula, the failure probability P at time t. tf Represented as:

[0075] ;

[0076] Where the probability of the i-th simulation occurring is P u .

[0077] Beneficial effects:

[0078] This invention provides a dynamic reliability analysis method for double-laminated insulated glass under wind loads. Compared with existing technologies, it first proposes a novel finite element model for double-laminated insulated glass. This addresses the current limitation of double-laminated insulated glass design relying entirely on empirical formulas, and the finite element model is interpretable. Furthermore, the accuracy of the finite element model is verified from both theoretical and experimental perspectives. A comprehensive wind load damage model is established, comprehensively considering the effects of typhoons and natural winds. Finally, a dynamic reliability assessment method for double-laminated insulated glass is proposed. This method can assist designers in comprehensively considering the impact of installation height and service life on the glass failure probability. Attached Figure Description

[0079] Figure 1 Overall flowchart of the reliability analysis method in this embodiment of the invention;

[0080] Figure 2 A schematic diagram illustrating the mechanical principle of double-laminated insulated glass in this invention embodiment;

[0081] Figure 3 Finite element analysis flowchart in an embodiment of the present invention;

[0082] Among them, (a) - simplified finite element model, (b) - setting contact and constraints, (c) - activating element display, and (d) - the effect of mesh density;

[0083] Figure 4 Schematic diagram of wind load test principle in this embodiment of the invention;

[0084] Figure 5 Wind load experiment diagram in an embodiment of the present invention;

[0085] Wherein (a) - experimental setup inside the building, (b) - experimental setup outside the building, (c) - experimental scheme, (d) - strain sensor, (e) - displacement sensor;

[0086] Figure 6 Three-dimensional scanning results in the embodiments of the present invention;

[0087] Figure 7 Comparison chart of test results under three working conditions in embodiments of the present invention;

[0088] Among them, (a) - displacement comparison results, (b) - stress comparison results;

[0089] Figure 8 A schematic diagram of the load and constraints of the laminated glass structure in an embodiment of the present invention;

[0090] Figure 9 A schematic diagram of the load and constraints of the insulated glass structure in an embodiment of the present invention;

[0091] Figure 10 A comparison diagram of the finite element model of sub-part 1 in this embodiment of the invention and the existing technical method;

[0092] Among them, (a) -2990Pa working condition, (b) -3990Pa working condition, and (c) -4990Pa working condition;

[0093] Figure 11 A comparison diagram of the finite element model in sub-part 2 of this invention and the existing technical method;

[0094] Among them, (a) -1230Pa working condition, (b) -1730Pa working condition, and (c) -2230Pa working condition;

[0095] Figure 12 The principles of single-instance damage and cumulative damage in the embodiments of the present invention;

[0096] Figure 13 A graph showing the number of typhoons occurring annually in Wanning City, as described in this invention.

[0097] Among them, (a) - 1951-1987, (b) - 1988-2024;

[0098] Figure 14 The results of simulating a stochastic process using the quasi-Monte Carlo method in this embodiment of the invention are shown in the figure.

[0099] Figure 15 Typhoon wind speed sample diagram in an embodiment of the present invention;

[0100] Where (a) - statistical sample values, (b) - fitted sample values;

[0101] Figure 16 Natural wind calculation process in this embodiment of the invention;

[0102] Figure 17 Natural wind speed signal diagram in an embodiment of the present invention;

[0103] Wherein, (a) - time domain results, (b) - frequency domain results;

[0104] Figure 18 The simulated wind speed matrix diagram in this embodiment of the invention;

[0105] Figure 19 Flowchart of dynamic reliability calculation in this embodiment of the invention;

[0106] Figure 20 The first six fundamental frequencies and mode shapes of the model in this embodiment of the invention;

[0107] Among them, (a) -11.520HZ calculation result, (b) -13.133HZ calculation result, (c) -17.404HZ calculation result, (d) -31.875HZ calculation result, (e) -32.744HZ calculation result, and (f) -42.407HZ calculation result;

[0108] Figure 21 A diagram showing the dynamic reliability calculation results in an embodiment of the present invention. Detailed Implementation

[0109] In this embodiment, a finite element model for double-laminated insulated glass is first proposed, comprehensively considering the coupling problem between the laminated structure and the flow field within the cavity. This model exhibits high accuracy and stability, and can be used for both static and dynamic analyses. The displacement and stress responses of the double-laminated insulated glass finite element model and experimental results are compared, demonstrating the accuracy of the finite element model. To provide interpretability of the finite element model, the differences between the finite element model of two sub-parts under the same constraints and similar load conditions and existing theories are compared. Taking Wanning City, Hainan Province as an example, wind loads are classified as a combination of natural wind and typhoon, considering the randomness of the loads. Simultaneously considering the randomness of structural parameters, the cumulative damage to the glass under wind loads over a certain period is calculated. The dynamic reliability of the curtain wall glass failure probability as a function of height and time is analyzed.

[0110] A method for dynamic reliability analysis of double-laminated insulated glass under wind load, such as Figure 1 As shown, it includes:

[0111] Step S1: Based on the structural form, size, and constraint method of double-laminated insulated glass, a finite element analysis model of double-laminated insulated glass is established using finite element simulation software; the finite element analysis model is divided into a laminated glass module and an insulated glass module.

[0112] The double-laminated insulated glass includes laminated glass and insulated glass, wherein the laminated glass and insulated glass correspond to the laminated glass module and the insulated glass module in the finite element analysis model, respectively.

[0113] The mechanical model of glass damage caused by wind loads is usually simplified to a simply supported plate under uniform pressure. In engineering practice, laminated glass is used to improve safety. This sandwich structure consists of a soft core and a rigid outer shell. Insulating glass is used to improve sound insulation and thermal insulation, and to prevent fogging. Both the core layer and the gas can improve the magnitude and distribution of glass stress, reducing damage. In some areas frequently hit by typhoons, builders also use double-laminated insulating glass. In engineering practice, curtain wall glass is more likely to conform to the constraint of simply supported on four sides rather than simply supported on two sides. Therefore, this embodiment uses plate theory instead of the simplified beam theory. This embodiment uses ANSYS to construct the finite element model. Figure 2 The diagram illustrates the stress principle of double-laminated insulated glass. Other types of curtain wall glass are considered sub-parts of this model in this embodiment.

[0114] like Figure 2As shown, the double-laminated insulated glass in this embodiment comprises two sub-parts: sub-part 1 is considered laminated glass, and sub-part 2 is considered insulated glass; the four edges of glass 4 are fixed to the building structure. The outer surface of glass 1 is subjected to positive wind pressure. In sub-part 1, glass 1, PVB, and glass 2 form a sandwich structure. In the finite element analysis, SHELL181 elements are used for layered simulation. It is assumed that there is no relative displacement between the layers. In previous experiments and studies, no slippage was observed, demonstrating the rationality of the assumption. In sub-part 2, after the deformation of glass 2 and glass 3, they exert forces on the seal and the cavity gas. The phenomenon of force transmission by the cavity gas is called load sharing. In the finite element analysis, a fluid-structure interaction model of the enclosed space is used to analyze load sharing. This model uses HSFLD242 elements, including pressure nodes N and surface elements Ei on the inner wall of the cavity. The entire cone formed by surface elements Ei and pressure nodes N represents the fluid in the enclosed space. The volume change and pressure change of this fluid are coupled together. This model assumes a uniform fluid pressure distribution and that fluid disturbances and viscosity are negligible. Previous studies have demonstrated the validity of these assumptions.

[0115] like Figure 3 (a) shows a simplified finite element model. The material parameters were provided by the manufacturer. The glass is considered isotropic and remains linearly elastic before fracture. There is no relative motion between the glass plate and the seal; they are connected via nodal coupling.

[0116] like Figure 3 (b) shows the setting of contact and constraints, applying constraints and loads to the finite element model. The four sides of the inner glass surface are simply supported, and three translational degrees of freedom are constrained. The outer glass surface is subjected to a uniform compressive load.

[0117] like Figure 3 (c) shows the finite element elements displayed as solid shapes. To provide a clearer view of the model, the sealed side and the connected cavity gas elements have been removed.

[0118] like Figure 3 (d) illustrates the impact of mesh density on simulation results; convergence is typically achieved when each side exceeds 10 elements. In this embodiment, a mapping mesh strategy with a custom element size is employed. Figure 3 As shown, the finite element results failed to converge when using a mesh density of 8×16. However, higher mesh densities increased the required computational cost and time. This embodiment uses a mesh density of 40×80, with a relative deviation of less than 0.3%, reaching 0.07%.

[0119] Step S2: Verify the accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load through experiments, and optimize the finite element model based on the experimental results;

[0120] The accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load was verified by experiments, specifically using a 3D scanning system, a displacement measurement system, and a strain measurement system.

[0121] The 3D scanning system measures the installation status of the glass; based on the scanning results of the corner points of the insulated laminated glass, it adjusts the glass position to minimize deviation; such as... Figure 6 The deviation between the scanning results and the theoretical results is within 1mm, indicating that the installation deviation is within a reasonable range.

[0122] The displacement measurement system measures the displacement value of the double-laminated insulating glass. Specifically, the displacement sensor is installed on a rigid support independent of the double-laminated insulating glass and connected to a computer through a fiber optic adapter box. In this embodiment, the fiber optic adapter box specifically uses the XYXJ-I adapter box.

[0123] The strain measurement system includes a fiber Bragg grating sensor and a grating demodulator; in this embodiment, the grating demodulator specifically uses a ZX-FP / SL-C FBG demodulator; the fiber Bragg grating sensor is attached to the experimental measurement point at 45° intervals, as shown below. Figure 5 As shown in the image. Figure 5 (a) shows the experimental setup inside the building. Figure 5 (b) shows the experimental setup on the exterior of the building. Figure 5 (c) in the diagram represents the experimental design. Figure 5 (d) in the figure represents a strain sensor. Figure 5 (e) in the figure represents a displacement sensor; the fiber optic grating sensor is connected to the grating demodulator and the computer via wires. Temperature compensation and the initial wavelength of the fiber optic grating sensor are set to zero the initial strain value. The first principal stress σ1 at the measuring point is expressed as:

[0124] ;

[0125] Where E is the elastic modulus of the glass, μ is the Poisson's ratio of the glass, and ε0, ε 45 ,ε 90 These are strain measurements taken by a fiber Bragg grating sensor in three directions spaced 45° apart.

[0126] During the experiment, the building's interior was connected to a smart pressure sensor, an electric air blower, and a central control room via pipelines. Air was extracted from the building to simulate wind load. Double-laminated insulated glass samples were mounted on an aluminum alloy sealing structure on the building's surface using aluminum panels, ensuring consistency with the boundary conditions of the finite element model. The power of the electric air blower was adjusted to simulate the effect of wind load. Once the smart pressure sensor reading reached the required level, the power of the electric air blower was kept constant for one minute, and the displacement and strain values ​​were recorded. The stress changes caused by the wind load under these conditions were then calculated.

[0127] Existing research has not proposed a complete theoretical model for insulated laminated glass; therefore, this embodiment experimentally verifies the accuracy of the finite element model. The glass used in the experiment was provided by the manufacturer. The experimental principle is as follows: Figure 4 As shown. The test consists of three systems (3D scanning system, displacement measurement system, and strain measurement system), which measure the displacement and strain of the insulated laminated glass under initial and loaded conditions, respectively. Before the experiment, the geometry of the glass sample was measured and damage was checked.

[0128] In this embodiment, a ZX-FP / SL-C FBG demodulator is specifically used in the strain measurement system. Since the direction of the maximum stress on the surface of the hollow laminated glass is unknown, fiber Bragg grating sensors are arranged at 45° intervals to calculate the first principal stress at the measuring points. The fiber Bragg grating sensors are connected to the demodulator and the computer via wires. Flanges are used to ensure a reliable connection. Temperature compensation and the initial wavelength of the fiber Bragg grating sensors are set to zero the initial strain value.

[0129] The experimental results and finite element results under different wind load conditions in this embodiment are compared as follows: Figure 7 As shown in Table 1, Figure 7 (a) in the figure represents the displacement comparison result. Figure 7 (b) in the figure represents the stress comparison results.

[0130] Table 1: Relative errors of displacement and stress under three types of wind loads:

[0131] Note: Nodes are sorted in ascending order of coordinates, with the x-coordinate considered first, followed by the y-coordinate.

[0132] This embodiment compares a theoretical model of laminated glass (LGT model) with the laminated portion (sub-part 1) of the finite element model. The LGT model has higher accuracy than commonly used semi-empirical formulas and can provide verification, interpretation, and optimization for the finite element model. The constraints and loads of the LGT model are the same as those of double-laminated insulated glass, such as... Figure 8 As shown.

[0133] The solution to the partial differential equations of the LGT model is equivalent to a linear system of equations, including the coefficient matrix, the unknown displacement function vector β, and the right-hand side vector q. Finally, the displacement w(x,y) of the glass at the (x,y) coordinates is solved:

[0134] ;

[0135] ;

[0136] ;

[0137] Among them, w mn It is an unknown variable, β 11 It is the first element of the unknown displacement function vector β, β 21 It is the second element of the unknown displacement function vector β, β 31 It is the third element of the unknown displacement function vector β.

[0138] ;

[0139] ;

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] ;

[0145] ;

[0146] ;

[0147] ;

[0148] ;

[0149] Where L is the length of the laminated glass, B is the width of the laminated glass, and E is the length of the laminated glass. g It is the elastic modulus of glass, v g It is the Poisson's ratio of glass, G t It is the shear modulus of the sandwich layer, h g It is the thickness of the glass plate, h t ρ is the interlayer thickness, and p is the pressure load. The final displacement at (x, y) coordinates is expressed as:

[0150] ;

[0151] In the LGT model, the displacement of point (x,y) is w(x,y). Compare the displacements of points at the same position on the L1 line in plane A1 and the L2 line in plane A2 of the two models. Figure 10 The comparison results show that sub-part 1 of the finite element model achieves the accuracy of the LGT model. Both models conform to the Kirchhoff plate assumption without loss of accuracy. Figure 10 (a) represents the 2990Pa operating condition. Figure 10 (b) represents the 3990Pa operating condition. Figure 10 (c) is the working condition of 4990Pa;

[0152] This embodiment compares a Betty Analysis Method (BAM) with a finite element model of the insulated glass portion (sub-part 2). BAM offers higher accuracy than commonly used semi-empirical formulas and can provide verification, interpretation, and optimization for the finite element model. The constraints and loads of BAM are the same as those for double-laminated insulated glass, such as... Figure 9 .

[0153] BAM first calculated the fluid pressure change Δp caused by pressure p:

[0154] ;

[0155] In the formula, L is the length of the insulating glass unit, and B is the width of the insulating glass unit. g For the glass thickness, s t E represents the cavity thickness. g It is the elastic modulus of glass, v g is Poisson's ratio of the glass. p is the pressure load, and p0 is the initial pressure of the gas in the cavity. The mean value of the shape function is expressed as:

[0156] ;

[0157] In BAM, the displacement of point (x,y) is w2(x,y);

[0158] ;

[0159] D2 represents the bending stiffness of the glass plate, Ψ A Represents a shape function.

[0160] ;

[0161] ;

[0162] The displacements of the same points on the L1 line in the A1 plane and the L2 line in the A2 plane were compared between the two models and the professional software MEPLA ISO. The comparison results show that sub-part 2 of the finite element model achieves the accuracy of BAM and MEPLA ISO.

[0163] This section provides a case study to demonstrate the effectiveness of the proposed model. All numerical examples were tested on a personal computer using an i9-14900 CPU and 16GB of RAM.

[0164] This embodiment uses a glass plate with dimensions of 2000 mm × 1000 mm and a thickness of 10 mm. It also includes a 1.52 mm thick PVB interlayer and a 12 mm thick square cross-section spacer. The glass type used in this embodiment is widely used in the building curtain wall industry. It is worth noting that the method proposed in this embodiment is also applicable to other glass types. Material parameters provided by the manufacturer are shown in Table 2.

[0165] Table 2 Material Parameters:

[0166] Note: (1) Elastic modulus (E), (2) Poisson's ratio (μ), shear modulus (G);

[0167] This embodiment verifies the effectiveness of the two main sub-parts of the proposed finite element model by comparing it with some advanced theoretical models. Although the theoretical solution is in the form of an infinite series, a finite number of terms are used in the numerical simulation. The theoretical model series has been verified to have good numerical stability, with the first 31 terms achieving rapid convergence. For sub-part 1, the displacements at positions A1-L1 and A2-L2 under the same loads (2990Pa, 3990Pa, 4990Pa) were calculated using the finite element model and the LDT model. To illustrate that the model is applicable to various working conditions, the comparison results under the three load conditions are as follows: Figure 10 The positive validation results demonstrate that the assumptions used in the theoretical model can be used to describe the finite element model. For the finite element model, these assumptions provide interpretability and possibilities for further optimization.

[0168] For sub-part 2, the displacements at locations A1-L1 and A2-L2 under the same loads (1230Pa, 1730Pa, 2230Pa) were calculated using the finite element model, the BAM model, and the specialized software MEPLA ISO. To demonstrate the model's applicability to various working conditions, comparison results under the three load conditions are shown below. Figure 11 .in Figure 11 (a) represents the 1230Pa operating condition. Figure 11 (b) represents the 1730Pa operating condition. Figure 11 (c) represents the 2230Pa working condition; the good verification results indicate that the finite element model accurately simulates the load sharing of gas in the insulating glass structure. Similarly, the assumptions of the BAM model provide interpretability and the possibility for further optimization.

[0169] Step S3: Analyze the strength degradation and cumulative damage caused by typhoon insulated glass;

[0170] The intensity degradation is achieved through the intensity degradation function R. 60 (n) is determined, specifically:

[0171] ;

[0172] Among them, R 60 R represents the initial load capacity. 60 (n) represents the load capacity in year n, and α(n,k1) is a deterministic function that varies with time. TY represents the design life of the glass, and the proportional parameter k1 and the exponential parameter γ are determined experimentally.

[0173] The cumulative damage caused by the typhoon is specifically as follows:

[0174] ;

[0175] In the formula, Let S be the cumulative damage caused by the typhoon in year i, j = 1, 2, ..., i, and D1(S) be the glass damage caused by typhoon wind speed S, calculated using wind-induced glass damage; S = {S1, S2, ..., S...} N} T This represents a series of samples representing the distribution of typhoon wind speeds; the typhoon wind speeds S are sorted by occurrence time and represented using A. j and n j Group and name them as follows: to The following are the wind-induced glass damages: T represents the duration of glass stress, s represents the wind speed, σ(•,t) represents the stress as a function of time t, 16 is a constant related to fracture mechanics, and D(s) is the cumulative damage amount.

[0176] Microcracks inevitably exist on the surface of glass, varying in geometry and orientation. The strength of curtain wall glass depends on the size of these cracks. Wind loads cause crack propagation, becoming a major load in the design of many building curtain wall glass systems. Wind loads include strong winds (short-term high wind speeds) and natural winds (long-term low wind speeds). The cumulative damage function D(S) and the strength degradation function R... 60 (n) together describe the failure behavior of glass under wind load.

[0177] Due to the influence of the surface environment, wind speeds at different altitudes exhibit a gradient distribution pattern. This embodiment uses an exponential average wind speed profile, where the wind speed at point z is represented as u(z).

[0178] ;

[0179] Among them, z r For reference height, u r The wind speed is used as a reference height. Chinese national standards and North American standards typically use 10m as the reference height. The wind speed profile index α is selected based on the surface roughness type of the city. This formula reflects the relationship between wind speeds at different heights within the wind field.

[0180] According to Bernoulli's principle, wind speed u exerts pressure p on the building surface:

[0181] ;

[0182] Where ρ is the air density.

[0183] Surface defects in glass propagate under tensile stress. When this stress accumulates to a certain point, the glass breaks. This tensile stress is typically represented by the first principal stress σ1. Brown used the integral of stress over time to describe the damage process of glass. Brown's integral successfully explained many glass failure behaviors. In studies of wind-induced glass damage, the stress σ1 is caused by wind pressure p.

[0184] ;

[0185] T represents the duration of glass stress, s represents the wind speed. σ(t) represents the stress as a function of time, which can be obtained through experiments, finite element method, finite difference method, and Galerkin decomposition. 16 is a constant related to fracture mechanics. D represents the cumulative damage to the glass. This formula shows that, for the same cumulative damage limit, both small stress over a long period and large stress over a short period can lead to glass breakage. Previous studies have shown that cumulative damage is not affected by the frequency of stress fluctuations.

[0186] In practical engineering, the strength degradation of glass over long periods is unavoidable. Wear, corrosion, and material aging are the causes of glass strength degradation. Previous studies have shown that neglecting glass strength degradation would significantly overestimate the reliability of curtain wall glass. In North America, the equivalent 60-second stress R... 60 This is used to describe the load-bearing capacity of glass. That is, the glass breaks after 60 seconds of stress. The deterministic law of glass strength degradation is expressed as follows.

[0187] ;

[0188] Among them, R 60 R represents the initial load capacity. 60 (n) represents the load capacity in year n, and α(n,k1) is a deterministic function that changes over time; in this embodiment, the parameters are k1=0.3 and γ=1. In engineering, there is a long construction period from glass fabrication to installation. Although the properties of the interlayer adhesive also degrade, all interlayer adhesive parameters used in this embodiment are taken from a stable stage, and subsequent performance degradation of the interlayer adhesive is not considered.

[0189] Typhoons are destructive storms that occur in the western Pacific Ocean. The area with the highest wind speeds and greatest destructive power is called the eyewall. For buildings, the duration of the eyewall effect can be conservatively taken as 2 hours. As a complex climate phenomenon, the frequency and wind speed of typhoons are uncertain. Stochastic processes are used to describe this uncertainty over time. For curtain wall glass, occasional typhoons are considered as an impact load, leading to relatively large stress values ​​in a short period. Composite Poisson processes are used to describe the cumulative effect of this stress. Typhoon occurrences are independent of each other and have independent incremental properties. A year is often chosen as the time scale for measuring glass failure. The number of typhoons in year t is expressed using a Poisson process as follows:

[0190] ;

[0191] The distribution of the number of typhoons within year t is expressed as follows:

[0192] ;

[0193] Where λ is a parameter representing the average rate of event occurrence. Because the Poisson process has a stationary incremental property, the distribution of typhoon occurrences is independent of the starting time of the statistics. Based on meteorological statistics, t... * Expected occurrence of a typhoon within a given time period (n) * According to the principle of maximum likelihood estimation, λ should make t * Typhoon n occurred within a time period * The probability P[N(t)] times * )=n * ;λ] is the maximum.

[0194] ;

[0195] Solving for the parameter λ, we obtain the following:

[0196] ;

[0197] Event A j Indicates the number of typhoons in year j; n j This represents the total number of typhoons that occurred before year j:

[0198] ;

[0199] S={S1,S2,…,S N} T This represents a series of samples indicating the wind speed distribution of typhoons. Samples S are sorted by occurrence time and use A... j and n jGrouping and naming. Typically, the typhoon wind speed distribution in a given area can be obtained through statistical methods. The glass damage D1(S) caused by typhoon wind speed S can be obtained through simulation or experimentation. CD1 i This represents the total cumulative damage caused by the typhoon in year i. Figure 12 It demonstrates the naming conventions for single-instance and cumulative typhoon damage.

[0200] ;

[0201] Taking Wanning City as an example, this embodiment verifies the calculation method for wind-induced damage to curtain wall glass. Based on data from the China Typhoon Network, this embodiment summarizes the number of typhoons occurring annually in Wanning City from 1951 to 2024, such as... Figure 13 .in Figure 13 (a) represents 1951-1987. Figure 13 (b) represents the period from 1988 to 2024. It is worth noting that losses caused by typhoons with wind speeds below 10 m / s are not considered. According to the maximum likelihood estimation formula, the Poisson process parameter λ = 0.6351. The stochastic process is simulated 2000 times using the quasi-Monte Carlo method, such as... Figure 14 As can be seen, the annual average values ​​of all simulation results fluctuate around the expected values, but remain within a reasonable range. A sample of typhoon wind speeds is shown below. Figure 15 Use a truncated exponential distribution for best fit. Figure 15 The typhoon wind speed sample V from the numerical simulation was obtained by sampling from the SOBO sequence. Figure 15 (a) represents the statistical sample value. Figure 15 (b) represents the fitted sample value.

[0202] Step S4: Based on the Dalbert spectrum, consider glass damage caused by natural wind:

[0203] The natural wind load wind speed u(z) is expressed as the average wind speed u m (z) and pulsating wind speed u t The combination of (z) represents the average wind, which indicates long periodicity, and the fluctuating wind, which indicates short periodicity. Fluctuating wind is typically represented as a stationary, ergodic, normally distributed random process. Although the stress caused by natural wind is relatively small, its duration is long. To comprehensively consider glass damage, this embodiment takes into account the influence of natural wind fluctuations. The annual damage caused by natural wind in year t, calculated according to Bernoulli's principle and Brownian integral, is represented as D2. t (u(z)), where the cumulative total damage from natural wind in year i is represented by CD2. i :

[0204] ;

[0205] In the formula, u(z) is the wind speed under natural wind load;

[0206] Dalbert summarized the average results of multiple wind load records measured in the field at different locations and altitudes, and gave the following empirical expression for the power spectrum of fluctuating wind:

[0207] ;

[0208] ;

[0209] Where k is the geomorphic coefficient, u m (z) represents the average wind speed at height z, u r The wind speed is at the reference altitude, and f is the frequency range;

[0210] The harmonic method is used to simulate the time history of fluctuating wind. The calculation process in this embodiment is as follows: Figure 16 Finally, the fluctuating wind is added to the average wind to obtain the total wind speed, where the spectral density matrix S P Represented as:

[0211] ;

[0212] ;

[0213] ;

[0214] ;

[0215] ;

[0216] Among them, S v Represents the self-power spectrum, spectral density matrix S P The Choreski decomposition matrix H is expressed as:

[0217] ;

[0218] Taking into account the combined effects of typhoons and natural winds, the cumulative total wind load damage in year i is represented as CD. i= CD1 i +CD2 i ;

[0219] The historical average maximum wind speed in Wanning City is 4.1 m / s. For conservative estimates, this embodiment considers the damage caused by this wind speed. Based on the above formula, the simulated natural wind speed signal is as follows: Figure 17 The power spectral density calculated using the Welch periodogram method is as follows: Figure 17 The natural wind speed signal was imported into the finite element model of the double-laminated insulated glass, and the (maximum stress) response results are as follows: Figure 17 ,in Figure 17 (a) represents the time-domain result. Figure 17 (b) represents the frequency domain result.

[0220] Step S5: Perform dynamic reliability analysis on double-laminated insulated glass;

[0221] Previous studies have typically neglected the influence of uncertainties in failure analysis, focusing instead on the failure behavior of a single pane of glass under a given load. The numerical value of this given load warrants further discussion. In practical engineering, curtain wall glass is inevitably affected by manufacturing and load uncertainties. Therefore, this paper presents a failure model based on dynamic reliability analysis.

[0222] In reliability analysis, the system describes a deterministic relationship between input variables and output response. This means that uncertainty in the input variables will cause uncertainty in the output response. Reliability analysis uses random variables to model uncertainty. The distribution of input random variables is described by probability and statistical models, and the probability of system failure is described by output random variables. Uncertainties such as temperature and vibration lead to uncertainties in glass parameters, even though they are within specified tolerances. The mechanical properties of glass are affected by composition and heat treatment processes. In addition, the geometric parameters of glass are affected by processing and manufacturing. Therefore, the main random variable X considered in this embodiment includes:

[0223] ;

[0224] Where L is the length of the glass, W is the width of the glass, E is the elastic modulus of the glass, μ is the Poisson's ratio of the glass, h is the thickness of the glass, and s t It is the cavity thickness;

[0225] The dynamic failure probability of the curtain wall glass is expressed as:

[0226] ;

[0227] ;

[0228] Among them, t s is the design life of the curtain wall glass, CD(X,t) is the cumulative damage to the glass under wind load at time t, considering the effects of load and random variables in the structure. RD(t) is the degraded strength, and G(X,t) is the limit state function. The failure domain and safety domain are respectively expressed as:

[0229] ;

[0230] The dynamic failure probability is solved by dividing the region using the composite limit state method, specifically including the following steps:

[0231] Step A1: Based on the typhoon wind speed distribution probability, a randomly arranged wind speed sample vector V is obtained using SOBO sequence sampling. The design life of the curtain wall glass is t. s .

[0232] Step A2: Simulate N random processes using the quasi-Monte Carlo method. The result is represented by a matrix YN, which represents the number of typhoons occurring each year in the N simulations. For example, the element YN in the matrix... ji This represents the number of typhoons that occurred in year j during the i-th simulation.

[0233] Step A3: Extract vector V sequentially based on the values ​​of matrix YN. Augment the missing typhoon counts compared to the maximum annual typhoon count with 0 to obtain matrix YNS. Therefore, matrix YNS simultaneously represents the frequency and intensity of typhoons. In the i-th simulation, t... j The typhoon wind speed vector occurring within a time period is represented by S. i Similarly, the number of missing occurrences is augmented with 0. For example... Figure 18 As shown, the sum of the damages corresponding to each column of matrix S is t. j The cumulative damage over time.

[0234] Step A4: Sample the random variable X according to the probability model, for a total of N. x Group;

[0235] Step A5: Calculate the wind pressure corresponding to the wind speed according to Bernoulli's principle;

[0236] Step A6: Considering the influence of the introduced random variable, the sample size is N. k The sample {X k}, k=1,…, N k N k < <N x ;

[0237] Step A7: Use the inner layer to build an adaptive proxy model G k (X * ,t), where X * ∈{X k},t∈[0,T s ], T s The maximum time of concern; the calculation of the function G refers to the formula for cumulative total damage from wind load;

[0238] Step A7.1: Time region [0, t] s Discretized into discrete time points {t} * The training sample points are represented as S. k ={t * ,G(X * ,t * )}.

[0239] Step A7.2: Establish the agent model G k (X * ,t);

[0240] Step A7.3: Calculate the learning function E[I(t)] based on the predicted values ​​of the surrogate model;

[0241] ;

[0242] Where Ф(.) and φ(.) represent the standard normal cumulative distribution function and probability density function, respectively;

[0243] Step A7.4: The point of maximum improvement is added to the training samples, and the surrogate model is retrained; the point of maximum improvement t new Represented as:

[0244] ;

[0245] Step A7.5: Learn the convergence criterion of the function:

[0246] ;

[0247] Step A8: In the middle layer, calculate the minimum value G of the function. e (X * ):

[0248] ;

[0249] Step A9: The training sample points of the middle layer are represented as S e ={X k G e (X k )}.

[0250] Step A10: Establish an adaptive agent model;

[0251] Step A11: Calculate the learning function C based on the predictions from the adaptive surrogate model. A (X):

[0252] ;

[0253] Where f X (x) is the joint probability density function of random variables;

[0254] Step A12: The point of maximum improvement is added to the training samples, and the adaptive agent model is retrained. Maximum improvement point X new Represented as:

[0255] ;

[0256] Step A13: The convergence criterion for the learning function is:

[0257] ;

[0258] Among them, C R This is the confidence level, set to 0.9999;

[0259] Step A14: Calculate the failure probability in the outer layer, the failure probability P of the i-th simulation. tf (P t |n=i) is represented as:

[0260] ;

[0261] in:

[0262] ;

[0263] Step A15: According to the conditional probability formula, the failure probability P at time t. tf Represented as:

[0264] ;

[0265] Where the probability of the i-th simulation occurring is P u ;

[0266] The dynamic reliability calculation process is as follows: Figure 19 Based on the finite element model of insulated laminated glass, the first six frequencies and mode shapes are calculated using the subspace iteration method. Figure 20 ,in Figure 20 (a) represents the calculation result at 11.520 Hz. Figure 20 (b) represents the calculation result at 13.133 Hz. Figure 20 (c) represents the calculation result at 17.404 Hz. Figure 20 (d) represents the calculation result at 31.875 Hz. Figure 20 (e) represents the calculation result at 32.744 Hz. Figure 20 (f) represents the calculation result at 42.407 Hz; the frequency of the insulated laminated glass is greater than 3 Hz, indicating that wind load will not cause resonance failure. Cumulative damage can well explain the failure behavior of the glass.

[0267] The degradation of glass load-bearing capacity requires long-term observation. This embodiment references the glass strength degradation law of Li and Li. Recommended parameters are k1=0.3, γ=1. The damage limit of the glass material can be determined experimentally. For conservative considerations, experimental failure data of 21 new glass pieces under an equivalent 1-minute wind pressure of 3.53 kPa were referenced from Taylor et al.'s study. This embodiment uses six random variables to quantify the influence of randomness on the time-varying reliability of insulated laminated glass. The distribution of glass product parameters provided by the manufacturer is shown in the table. Where E represents the elastic modulus of the glass, μ represents the Poisson's ratio of the glass, L represents the length of the glass, and B represents the width of the glass. g The thickness of the glass is represented by s. The cavity thickness is represented by s. Other parameters used in the finite element model are shown in Table 3.

[0268] Table 3 Distribution types and parameters of random variables:

[0269] Finally, taking Wanning City, Hainan Province as an example, the dynamic reliability of this type of curtain wall glass was calculated, such as... Figure 21 The combined effects of two wind loads were considered. The reliability variations caused by load randomness and product parameter randomness were calculated. The changes in glass failure probability due to glass height and usage time were plotted. This provides a quantified safety standard for the use of curtain wall glass in the city.

[0270] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0271] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of the methods disclosed herein and their equivalents, then the intent of this disclosure also includes such modifications and variations.

Claims

1. A method for dynamic reliability analysis of double-laminated insulated glass under wind load, characterized in that, Includes the following steps: Step S1: Based on the structural form, size and constraint method of double-laminated insulated glass, establish a finite element analysis model of double-laminated insulated glass using finite element simulation software; The finite element analysis model includes a laminated glass module and an insulated glass module; Step S2: Verify the accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load through experiments, and optimize the finite element model based on the experimental results; Step S3: Analyze the strength degradation and cumulative damage caused by typhoon insulated glass; Step S4: Based on the Dalbert spectrum, consider the glass damage caused by natural wind to obtain the total cumulative damage from wind load; Step S5: Perform dynamic reliability analysis on double-laminated insulated glass.

2. The method for dynamic reliability analysis of double-laminated insulated glass under wind load as described in claim 1, characterized in that, The double-laminated insulated glass in step S1 includes laminated glass and insulated glass, wherein the laminated glass and insulated glass correspond to the laminated glass module and the insulated glass module in the finite element analysis model, respectively.

3. The method for dynamic reliability analysis of double-laminated insulated glass under wind load as described in claim 1, characterized in that, Step S2 describes the experimental verification of the accuracy of the finite element model of the double-laminated insulated glass module under uniformly distributed load, specifically using a 3D scanning system, a displacement measurement system, and a strain measurement system. The 3D scanning system measures the installation status of the glass; based on the scanning results of the corner points of the double-laminated insulated glass, the glass position is adjusted to minimize the deviation. The displacement measurement system measures the displacement value of double-laminated insulated glass. Specifically, the displacement sensor is installed on a rigid support independent of the double-laminated insulated glass and connected to a computer through a fiber optic adapter box. The strain measurement system includes a fiber Bragg grating sensor and a grating demodulator. The fiber Bragg grating sensor is attached to the experimental measurement point at 45° intervals. The fiber Bragg grating sensor is connected to the grating demodulator and a computer via wires. Temperature compensation and the initial wavelength of the fiber Bragg grating sensor are set to zero the initial strain value. The first principal stress σ1 at the measurement point is expressed as: ； wherein, wherein E is the elastic modulus of the glass, μ is the Poisson's ratio of the glass, ε0,ε 45 ,ε 90 are the strain measurements of the fiber Bragg grating sensor in three directions at intervals of 45°; During the experiment, the building's interior was connected to a smart pressure sensor, an electric air blower, and a central control room via pipelines. Air was extracted from the building to simulate wind load. Double-laminated insulated glass samples were mounted on an aluminum alloy sealing structure on the building's surface using aluminum panels, ensuring consistency with the boundary conditions of the finite element model. The power of the electric air blower was adjusted to simulate the effect of wind load. Once the smart pressure sensor reading reached the required level, the power of the electric air blower was kept constant, and the displacement and strain values ​​were recorded. The stress changes caused by the wind load under these conditions were then calculated.

4. The method for dynamic reliability analysis of double-laminated insulated glass under wind load according to claim 3, characterized in that, The strength degradation in step S3 is by a strength degradation function R 60 (n) determining, in particular: ； Among them, R 60 R represents the initial load capacity. 60 (n) represents the load capacity in year n, and α(n,k1) is a deterministic function that varies with time. TY represents the design life of the glass, and the proportional parameter k1 and the exponential parameter γ are determined experimentally. The cumulative damage caused by the typhoon is specifically as follows: ； In the formula, Let S be the cumulative damage caused by the typhoon in year i, j = 1, 2, ..., i, and D1(S) be the glass damage caused by typhoon wind speed S, calculated using wind-induced glass damage; S = {S1, S2, ..., S...} N } T This represents a series of samples representing the distribution of typhoon wind speeds; the typhoon wind speeds S are sorted by occurrence time and represented using A. j and n j Group and name them as follows: to The following are the wind-induced glass damages: T represents the duration of glass stress, s represents the wind speed, σ(•,t) represents the stress as a function of time t, 16 is a constant related to fracture mechanics, and D(s) is the cumulative damage.

5. The method for dynamic reliability analysis of double-laminated insulated glass under wind load according to claim 4, characterized in that, Step S4 is specifically: the single-year damage caused by natural wind in the tth year is expressed as D2 according to Bernoulli's principle and Brown integral calculation t (u(z)), the cumulative total damage of natural wind in the ith year is expressed as CD2 i : ； In the formula, u(z) is the wind speed under natural wind load; The cumulative total damage of wind load in the ith year considering the influence of typhoon and natural wind is expressed as CD i= CD1 i +CD2 i .

6. The method for dynamic reliability analysis of double-laminated insulated glass under wind load according to claim 5, characterized in that, Step S5 specifically involves considering the following main random variables X: X=[L,W,E,μ,h,s t ] ; where L is the length of the glass, W is the width of the glass, E is the modulus of elasticity of the glass, μ is the Poisson's ratio of the glass, h is the thickness of the glass, s t is the cavity thickness; The dynamic failure probability of the curtain wall glass is expressed as: ； ； Among them, t s Here, is the design life of the curtain wall glass, CD(X,t) is the cumulative damage to the glass under wind load at time t, RD(t) is the degraded strength, G(X,t) is the limit state function, and Prob represents the probability; the failure region and the safety region are respectively represented as: 。 7. The method for dynamic reliability analysis of double-laminated insulated glass under wind load according to claim 6, characterized in that, The dynamic failure probability is solved by dividing the region using the composite limit state method, specifically including the following steps: Step A1: Based on the typhoon wind speed distribution probability, a randomly arranged wind speed sample vector V is obtained using SOBO sequence sampling. The design life of the curtain wall glass is t. s ; Step A2: Simulate N random processes using the quasi-Monte Carlo method. The result is represented by matrix YN, which represents the number of typhoons that occur each year in the N simulations. Step A3: Extract vector V sequentially based on the values ​​of matrix YN. Augment the missing typhoon counts compared to the maximum annual typhoon count with 0 to obtain matrix YNS. Therefore, matrix YNS simultaneously represents the frequency and intensity of typhoons. In the i-th simulation, t... j The typhoon wind speed vector occurring within a time period is represented by S. i Similarly, the number of missing values ​​is augmented with 0, and the sum of the damage corresponding to each column of matrix S is t. j Cumulative damage over time; Step A4: Sample the random variable X according to the probability model, for a total of N. x Group; Step A5: Calculate the wind pressure corresponding to the wind speed according to Bernoulli's principle; Step A6: Considering the influence of the introduced random variable, the sample size is N. k The sample {X k }, k=1,…, N k N k < <N x ; Step A7: Use the inner layer to build an adaptive proxy model G k (X * ,t), where X * ∈{X k },t∈[0,t s The calculation of the function G refers to the formula for cumulative total damage from wind load; Step A8: In the middle layer, calculate the minimum value G of the function. e (X * ): ； Step A9: The training sample points of the middle layer are represented as S e ={X k G e (X k )}; Step A10: Establish an adaptive agent model; Step A11: Calculate the learning function C based on the predictions from the adaptive surrogate model. A (X): ； Where f X (x) is the joint probability density function of random variables; Step A12: The point of maximum improvement is added to the training samples, and the adaptive agent model is retrained; the point of maximum improvement X new Represented as: ； Step A13: The convergence criterion for the learning function is: ； Among them, C R This is the confidence level, set to 0.9999. Step A14: Calculate the failure probability in the outer layer, the failure probability P of the i-th simulation. tf (P t |n=i) is represented as: ； in: ； Step A15: According to the conditional probability formula, the failure probability P at time t. tf Represented as: ； Where the probability of the i-th simulation occurring is P u .

8. The method for dynamic reliability analysis of double-laminated insulated glass under wind load according to claim 7, characterized in that, Step A7 specifically includes the following steps: Step A7.1: Time region [0, t] s Discretized into discrete time points {t} * The training sample points are represented as S. k ={t * ,G(X * ,t * )}; Step A7.2: Establish the agent model G k (X * ,t); Step A7.3: Calculate the learning function E[I(t)] based on the predicted values ​​of the surrogate model; ； in, φ(.) represent the standard normal cumulative distribution function and probability density function, respectively; Step A7.4: The point of maximum improvement is added to the training samples, and the surrogate model is retrained; the point of maximum improvement t new Represented as: ； Step A7.5: Learn the convergence criterion of the function: 。

Images