Method and apparatus for predicting ultimate load capacity of composite panel assemblies

Abstract

The application discloses a method for predicting the ultimate bearing capacity of a composite wallboard assembly, and belongs to the technical field of material prediction. The method comprises the following steps: collecting multiple stress pictures of the composite wallboard assembly by a collecting device to obtain a picture dataset, each stress picture corresponding to a time step, and preprocessing each stress picture in the picture dataset to obtain a target picture set; inputting the target picture set into a trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set; obtaining an external force-time change curve based on the external force prediction value corresponding to each picture; training an ultimate bearing capacity prediction model based on the external force-time change curve, iteratively predicting by the ultimate bearing capacity prediction model, and obtaining a prediction value of the ultimate bearing capacity of the composite wallboard assembly. The method improves the accuracy and reliability of the prediction of the ultimate bearing capacity of the composite wallboard assembly.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of material prediction, and particularly relates to a method and device for predicting the ultimate bearing capacity of a composite material wall plate assembly. BACKGROUND

[0002] Composite materials are the product of multidisciplinary integration, which are prepared by physical compounding or chemical combination of two or more different materials, aiming to integrate the advantages of each component material, make up for the performance short board of single material, and thus meet the stringent requirements of modern industry on high-performance materials. Through micro-level structural design and performance complementation, the material can simultaneously have multiple excellent properties such as high strength, high stiffness, excellent corrosion resistance and thermal stability, and become a key basic material for promoting the development of high-end technical fields. With the continuous expansion of the application scale of composite materials in various industries, issues such as ensuring the safety of composite material structures, optimizing material design schemes, and prolonging the service life of materials have attracted increasing attention. Among them, in-depth study of the real-time stress state and ultimate bearing capacity of composite materials is the core of solving the above key problems, and has important significance for promoting the sustainable development of the composite material industry.

[0003] The existing prediction of the ultimate bearing capacity of the composite material wall plate assembly mainly relies on the classical laminated plate theory combined with linear or nonlinear finite element analysis, and mostly relies on direct measurement by sensors, which requires deploying a large number of sensors on the wall plate, and the operation is complex and may affect the original performance of the structure. The model based on the macroscopic homogenization assumption is difficult to accurately characterize the unique progressive damage behavior (such as fiber fracture, matrix cracking, interlaminar delamination, and local buckling) of real composite materials and their complex interactions, resulting in prediction deviation of the failure initiation, propagation path, and final failure mode; at the same time, the calculation cost of high-precision three-dimensional explicit damage model is extremely high, which is difficult to meet the demand of rapid iteration in engineering design. Although the methods based on the cohesive force model or the continuum damage mechanics can simulate delamination or matrix damage, their characterization ability for complex failure modes (such as delamination coupled with fiber micro-buckling) is limited, and the model parameters are highly dependent on tedious experimental calibration. The traditional semi-empirical design method based on safety margin is too conservative and cannot fully tap the potential of the material. The existing methods have the problems of low prediction accuracy, large calculation amount, and complex operation in predicting the ultimate bearing capacity of the composite material wall plate assembly. SUMMARY

[0004] The present application aims to at least solve one of the problems in the prior art. To this end, the present application proposes a method and device for predicting the ultimate bearing capacity of a composite material wall plate assembly, which improves the accuracy and reliability of the prediction of the ultimate bearing capacity of the composite material wall plate assembly.

[0005] In a first aspect, the present application provides a method for predicting the ultimate bearing capacity of a composite wallboard assembly, the method comprising:

[0006] collecting a plurality of stress pictures of the composite wallboard assembly by a collection device to obtain a picture dataset, each stress picture corresponding to a time step, and preprocessing each stress picture in the picture dataset to obtain a target picture set;

[0007] inputting the target picture set into a trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set;

[0008] obtaining an external force-time curve based on the external force prediction value corresponding to each picture;

[0009] training an ultimate bearing capacity prediction model based on the external force-time curve, and iteratively predicting the ultimate bearing capacity prediction value of the composite wallboard assembly by the ultimate bearing capacity prediction model.

[0010] According to an embodiment of the present application, the preprocessing of each stress picture in the picture dataset to obtain a target picture set comprises:

[0011] creating a mask based on the most obvious part of the material wallboard deformation with external force in each stress picture, and setting all picture parts outside the mask to black;

[0012] cropping the black part in each stress picture to obtain a cropped stress picture;

[0013] smoothing the boundary of each cropped stress picture based on an inflation operation, so that the edge of each stress picture is smoother and the picture is more continuous, to obtain a target picture set.

[0014] According to an embodiment of the present application, the inputting of the target picture set into the trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set comprises:

[0015] inputting the target picture set into a CNN module to obtain the local features of each picture in the target picture set;

[0016] inputting the target picture set into a Transformer module to obtain the global features of each picture in the target picture set;

[0017] fusing the local features and the global features of each picture based on a gating mechanism to obtain the fusion features corresponding to each picture;

[0018] inputting the fusion features corresponding to each picture into a fully connected network to obtain the external force prediction value corresponding to each picture in the target picture set;

[0019] The external force prediction model is a CNN-Transformer model, and the CNN-Transformer model comprises a CNN module and a Transformer module.

[0020] According to an embodiment of the present application, the iterative prediction by the ultimate bearing capacity prediction model obtains a predicted value of the ultimate bearing capacity of the composite wallboard assembly.

[0021] The external force data of 10 units of time are obtained as an input sequence, the input sequence is input into the ultimate bearing capacity prediction model, and a predicted value of external force of the next unit of time is obtained;

[0022] The predicted value of external force of the next unit of time is added to the end of the original input sequence, and the external force data of the first unit of time at the beginning of the original input sequence is deleted, to obtain an updated input sequence;

[0023] The updated input sequence is input into the ultimate bearing capacity prediction model for iterative prediction, to obtain a predicted value of the ultimate bearing capacity of the composite wallboard assembly.

[0024] According to an embodiment of the present application, the iterative prediction by the ultimate bearing capacity prediction model obtains a predicted value of the ultimate bearing capacity of the composite wallboard assembly.

[0025] The updated input sequence is input into the ultimate bearing capacity prediction model, to obtain a predicted value of external force of the next unit of time;

[0026] The increase of the predicted value of external force of the next unit of time and the external force value at the current time is calculated;

[0027] When the increase is less than a preset threshold, the predicted value of external force of the next unit of time is taken as the predicted value of the ultimate bearing capacity of the composite wallboard assembly;

[0028] When the increase is greater than or equal to the preset threshold, the predicted value of external force of the next unit of time is added to the end of the original input sequence, and the external force data of the first unit of time at the beginning of the original input sequence is deleted, to obtain an updated input sequence, and the step of inputting the updated input sequence into the ultimate bearing capacity prediction model to obtain a predicted value of external force of the next unit of time is returned to.

[0029] According to an embodiment of the present application, the training process of the external force prediction model comprises:

[0030] A preset external force prediction model is constructed;

[0031] A plurality of stress pictures of the composite wallboard assembly are obtained and preprocessed, to obtain a first data set;

[0032] performing a sliding window operation on the first data set to obtain a sliding window data set;

[0033] training the preset external force prediction model through a mean square error loss function according to the sliding window data set, to obtain a trained external force prediction model.

[0034] According to an embodiment of the present application, the training of the limit bearing capacity prediction model based on the external force-time change curve comprises:

[0035] constructing a preset limit bearing capacity prediction model, the preset limit bearing capacity prediction model being an LSTM prediction model;

[0036] obtaining a plurality of data points based on the external force-time change curve, and cleaning the plurality of data points to obtain cleaned data points;

[0037] identifying and removing the cleaned data points through a 3σ principle to obtain removed data points;

[0038] filling the removed data points through linear interpolation to obtain a second data set;

[0039] training the preset limit bearing capacity prediction model through a mean square error loss function according to the second data set, to obtain a trained limit bearing capacity prediction model.

[0040] In a second aspect, the present application provides a device for predicting limit bearing capacity of a composite material wall panel assembly, the device comprising:

[0041] an acquisition module configured to acquire a plurality of stress pictures of the composite material wall panel assembly through an acquisition device to obtain a picture data set, each stress picture corresponding to a time step, and to preprocess each stress picture in the picture data set to obtain a target picture set;

[0042] a first processing module configured to input the target picture set into the trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set;

[0043] a second processing module configured to obtain an external force-time change curve based on the external force prediction value corresponding to each picture;

[0044] a prediction module configured to train a limit bearing capacity prediction model based on the external force-time change curve, and to perform iterative prediction through the limit bearing capacity prediction model to obtain a prediction value of the limit bearing capacity of the composite material wall panel assembly.

[0045] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for predicting the ultimate bearing capacity of the composite wallboard assembly according to the first aspect.

[0046] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for predicting the ultimate bearing capacity of the composite wallboard assembly according to the first aspect.

[0047] In a fifth aspect, the present application provides a chip comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is configured to execute a program or an instruction to implement the method for predicting the ultimate bearing capacity of the composite wallboard assembly according to the first aspect.

[0048] In a sixth aspect, the present application provides a computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements the method for predicting the ultimate bearing capacity of the composite wallboard assembly according to the first aspect.

[0049] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter.

[0050] The present application provides a method for predicting the ultimate bearing capacity of a composite wallboard assembly, which has the following advantages over the prior art:

[0051] (1) The present application can quickly obtain overall deformation information of the wallboard by collecting multiple stress pictures of the composite wallboard assembly using a collecting device, thereby reducing interference with the structure and improving data collection efficiency and convenience, and achieving non-contact collection. The local and global feature depth of the tensile picture of the composite wallboard assembly is extracted and analyzed by an external force prediction model, and the external force prediction value corresponding to each picture in the target picture set is obtained. The external force prediction value is then input into the ultimate bearing capacity prediction model, and the ultimate bearing capacity is obtained through iterative prediction, thereby achieving efficient conversion from image data to mechanical performance indicators and improving the accuracy and reliability of the ultimate bearing capacity prediction of the composite wallboard assembly.

[0052] (2) The present application can dynamically adjust the input of the ultimate bearing capacity prediction model by inputting the updated input sequence into the ultimate bearing capacity prediction model and calculating the increase in the predicted value of the external force of the next unit time and the current time, and determining whether to update the input sequence according to the increase, thereby effectively improving the adaptability and accuracy of the ultimate bearing capacity prediction model in the ultimate bearing capacity prediction process.

[0053] (3) The present application can predict the external force change of the composite wall panel assembly in the next time unit by acquiring the external force data in a certain time period and inputting it into the ultimate bearing capacity prediction model. By continuously updating the input sequence and iteratively predicting, the predicted value of the bearing capacity can be updated in real time, obtaining more accurate ultimate bearing capacity prediction value, having better applicability, and being able to more comprehensively mine data characteristics, improving the accuracy of the ultimate bearing capacity prediction of the composite wall panel assembly. BRIEF DESCRIPTION OF DRAWINGS

[0054] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings in which:

[0055] Figure 1 is one of the flowcharts of the method for predicting the ultimate bearing capacity of the composite wall panel assembly provided by the embodiments of the present application;

[0056] Figure 2 is a basic structure diagram of the wall panel assembly provided by the embodiments of the present application;

[0057] Figure 3 is a schematic diagram of creating a mask provided by the embodiments of the present application;

[0058] Figure 4 is the second flowchart of the method for predicting the ultimate bearing capacity of the composite wall panel assembly provided by the embodiments of the present application;

[0059] Figure 5 is a structural schematic diagram of the device for predicting the ultimate bearing capacity of the composite wall panel assembly provided by the embodiments of the present application;

[0060] Figure 6 is a structural schematic diagram of the electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be described clearly below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0062] The terms "first", "second", and the like in the description and in the claims of the present application are used for distinguishing between similar objects and not necessarily for describing a specific sequential or chronological order. It is to be understood that the use of such terms is interchangeable under appropriate circumstances such that the embodiments of the present application described herein are capable of accomplishing functionalities using, for example, either the first or the second approach. Other sequences of objects' described herein can be equally applicable to other alternative or combinational methodologies. Moreover, the terms "first", "second", and the like, are used herein to distinguish between two independent (individually) objects that can or can not be in proximity to, contact, or in other manners relatively near, each other, unless otherwise specifically indicated. Additionally, the term "and / or" as used herein refers to at least one of the objects, and / or means that the objects are either used separately or together, unless otherwise specifically indicated.

[0063] The prediction method of the ultimate load capacity of the composite wall panel assembly, the prediction device of the ultimate load capacity of the composite wall panel assembly, the electronic device, and the readable storage medium provided by the embodiments of the present application will be described in detail below with reference to the specific embodiments and application scenarios.

[0064] The prediction method of the ultimate load capacity of the composite wall panel assembly can be applied to a terminal, and can be executed by hardware or software in the terminal.

[0065] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or a tablet computer having a touch-sensitive surface (for example, a touchscreen display and / or a touchpad). It should also be understood that, in some embodiments, the terminal can not be a portable communication device, but a desktop computer having a touch-sensitive surface (for example, a touchscreen display and / or a touchpad).

[0066] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal can include one or more other physical user interface devices such as physical keyboards, mice, and joysticks.

[0067] The prediction method of the ultimate load capacity of the composite wall panel assembly provided by the embodiments of the present application, the execution subject of the prediction method of the ultimate load capacity of the composite wall panel assembly can be an electronic device or a functional module or functional entity capable of realizing the prediction method of the ultimate load capacity of the composite wall panel assembly in the electronic device. The electronic device mentioned in the embodiments of the present application includes, but is not limited to, a mobile phone, a tablet computer, a computer, a camera, and a wearable device, etc. The prediction method of the ultimate load capacity of the composite wall panel assembly provided by the embodiments of the present application will be described below taking the electronic device as an execution subject.

[0068] In the aerospace, automotive, new energy and other frontier fields, with the continuous rise of the performance requirements of lightweight, high strength, high temperature resistance and corrosion resistance of materials, the application of composite materials is becoming more and more widespread. In the aerospace field, represented by carbon fiber reinforced composite materials, with its excellent specific strength and specific modulus characteristics, it is widely used in the manufacture of aircraft fuselage and key structural components, not only effectively reducing the weight of the aircraft, but also significantly improving the fuel efficiency and load capacity. In response to the development needs of energy saving and emission reduction and improving fuel economy, the automotive manufacturing industry actively applies composite materials to vehicle body structures, seat systems, interior components and engine components, while reducing the weight of the vehicle, significantly enhancing the safety performance and anti-collision capability of the vehicle. In the field of wind power generation, glass fiber or carbon fiber composite materials are often used to manufacture large-size wind turbine blades, which effectively improve the strength and durability of the blades on the basis of reducing the weight of the blades, thereby improving the wind energy conversion efficiency.

[0069] Figure 1 is one of the flowcharts of the method for predicting the ultimate bearing capacity of the composite wallboard assembly provided by the embodiments of the present application, as shown in Figure 1 The method for predicting the ultimate bearing capacity of the composite wallboard assembly comprises steps 110, 120, 130 and 140.

[0070] Step 110, acquiring a plurality of stress pictures of the composite wallboard assembly by a collection device to obtain a picture data set, each stress picture corresponding to a time step, and pre-processing each stress picture in the picture data set to obtain a target picture set;

[0071] In some embodiments, the pre-processing of each stress picture in the picture data set to obtain a target picture set comprises:

[0072] Creating a mask based on the most obvious part of the material wallboard deformation in each stress picture, and setting all the picture parts outside the mask to black;

[0073] Cutting the black part in each stress picture to obtain a cut stress picture;

[0074] Based on the expansion operation, the boundaries of each cut stress picture are smoothed, the edges of each stress picture are more smooth, and the picture is more continuous, and a target picture set is obtained.

[0075] For example, the collection device is an industrial camera, and a plurality of pictures of the composite wallboard assembly under stress are obtained by the industrial camera to obtain a picture data set, each stress picture corresponding to a time step.

[0076] Figure 2 is a basic structure diagram of the wallboard assembly provided by the embodiments of the present application, as shown in Figure 2As shown, the material is a polymer-based composite material, the length L of the composite wallboard assembly is 135 mm, the width w is 36 mm, the thickness h is 3 mm, the hole diameter is 6 mm, the end distance e is 18 mm, s is the length of the pad, and d is the diameter of the fastener.

[0077] Figure 3 is a schematic diagram for creating a mask provided by the embodiments of the present application, as Figure 3 shown, the middle bolt part of the composite wallboard assembly changes most obviously with the increase of the tensile force during the stretching process. The part of the deformation that changes obviously with the external force in each stress picture is extracted by the method of creating a mask. The picture taken by the industrial camera is converted into a gray matrix. The blue part of the wallboard assembly is extracted by the mask extraction method, that is, a mask is set, the value of the position in the gray matrix whose value is in the blue range is set to 1, and the value of the remaining position is set to 0. Multiply the mask and the original gray matrix to obtain a picture with only the blue area, and the remaining part is all black.

[0078] It should be noted that the mask is essentially a binary picture (only two pixel values of 0 and 1), and the area in the composite wallboard assembly picture where the deformation changes obviously with the external force is identified by a specific algorithm. The area where the deformation is not obvious is set to 0, and the other areas are set to 1.

[0079] Multiply the mask and the original picture taken by the industrial camera, that is, the pixel value of the original picture corresponding to the position where the value of the mask is 1 is retained, and the pixel value of the position where the value is 0 is set to 0. The part where the deformation changes obviously with the external force is extracted.

[0080] Further, the black part in each stress picture is cropped to reduce the size of the picture and delete useless information. The boundary smoothing and noise elimination of the cropped stress picture are performed by using the dilation operation in morphology. A circular disk structure element is used, and its radius is set to 2 pixel units. In the process of traversing the pixels of the picture, the center of the structure element is aligned with each pixel point in turn. When there is a pixel with a value of 1 in the covered area of the structure element, the center pixel point is forcibly updated to 1. Taking the jagged noise on the edge of the blue area in the picture as an example, these noise points usually present as discrete pixel points. Through the dilation operation, the small gaps in the edge can be effectively filled, and the originally uneven edge can be smoothly expanded outward, so that the blue area contour is closer to the actual physical boundary. At the same time, for the isolated black noise points that may exist in the picture, since most of the surrounding pixel values are 0, they will be covered by the normal pixel values in the dilation process, and finally the noise in the picture is effectively eliminated, providing a clearer and more accurate data basis for subsequent picture analysis, making the edge of each stress picture smoother and the picture more continuous, and obtaining a target picture set.

[0081] In this embodiment, by creating a mask and cutting the part of the material wallboard in each stress picture that deforms most obviously under external force, the size of the picture is reduced, irrelevant information is removed, and the computing efficiency is improved. The quality of the picture is effectively improved through the inflation operation, and a more smooth and continuous target picture set is obtained, which optimizes the quality of the input data and improves the design efficiency and safety of the wallboard assembly.

[0082] Step 120, inputting the target picture set into the trained external force prediction model to obtain the external force prediction value corresponding to each picture in the target picture set;

[0083] In some embodiments, the inputting the target picture set into the trained external force prediction model to obtain the external force prediction value corresponding to each picture in the target picture set comprises:

[0084] inputting the target picture set into a CNN module to obtain the local features of each picture in the target picture set;

[0085] inputting the target picture set into a Transformer module to obtain the global features of each picture in the target picture set;

[0086] fusing the local features and the global features of each picture based on a gating mechanism to obtain the fusion features corresponding to each picture;

[0087] inputting the fusion features corresponding to each picture into a fully connected network to obtain the external force prediction value corresponding to each picture in the target picture set;

[0088] The external force prediction model is a CNN-Transformer model, and the CNN-Transformer model comprises a CNN module and a Transformer module.

[0089] It can be understood that the target picture set comprises multiple pictures at multiple time steps, and each picture contains a large number of features reflecting detailed information. These features are detailed descriptions of specific positions and regions, and can reflect the microscopic changes of the material under stress. These features belong to local features, and usually have high resolution and accuracy. In the picture sequence at multiple time steps, the overall shape change trend of the wallboard assembly, the relative position relationship between different parts, and the overall deformation mode during the stress process all belong to global features.

[0090] For example, the small deformation of the bolt connection and the texture change of the local stress concentration area of the material all belong to local features. With the continuous action of external force, the change of the bolt position cannot be directly obtained from a single local feature, which belongs to global features.

[0091] In some embodiments, the external force prediction model is a CNN-Transformer model, which includes a CNN module and a Transformer module. The CNN module uses different size convolution kernels to slide on the picture to finely extract local features such as surface cracks and stress concentration areas from the composite wall panel assembly picture by virtue of the local connection and weight sharing characteristics of convolution operation. The Transformer module breaks the local neighborhood restriction and calculates the correlation weight of the features at different positions in the picture by virtue of the self-attention mechanism and multi-head attention mechanism, so as to realize long-range dependence modeling of global features such as the overall deformation trend of the wall panel and the cooperative relationship between components. Finally, the local features and the global features are spliced, and the weights of the two features are dynamically adjusted by using a gating mechanism, so that the advantages of the two features are complementary. After nonlinear transformation by multiple fully connected layers, the external force prediction value corresponding to each picture in the target picture set is obtained.

[0092] The steps of inputting the target picture set into the trained external force prediction model to obtain the external force prediction value corresponding to each picture in the target picture set include the following steps:

[0093] (1) Input the target picture set into the CNN module to extract the local features of the wall panel assembly picture by using the CNN.

[0094] (2) Input the target picture set into the Transformer module to extract the global features of the wall panel assembly picture by using the Transformer.

[0095] (3) Fuse the local features and the global features by using the gating mechanism, so as to realize effective integration of local detailed information and global structural information.

[0096] The formula of the gating mechanism is as follows:

[0097]

[0098]

[0099]

[0100] wherein, are the fused features, the local features extracted by the CNN, and the global features extracted by the Transformer, respectively, is the spliced local features and global features, W is a weight matrix, and b is a bias term, is a Sigmoid function, is the gating mechanism, is the spliced features.

[0101] (4) The fused features are input into a fully connected network to obtain the external force prediction value corresponding to each picture.

[0102] In this embodiment, by obtaining a target picture set and inputting it into the trained CNN-Transformer model, the CNN-Transformer can extract spatial features in the image, capture local deformation details and global structure information of the wallboard, and optimize and fuse the local features and global features through the gating mechanism, thereby improving the accuracy and safety of the prediction of the ultimate bearing capacity of the composite wallboard assembly.

[0103] Step 130, based on the external force prediction value corresponding to each picture, an external force-time curve is obtained;

[0104] Step 140, training an ultimate bearing capacity prediction model based on the external force-time curve, and iteratively predicting the ultimate bearing capacity prediction model to obtain the prediction value of the ultimate bearing capacity of the composite wallboard assembly.

[0105] For example, the external force-time curve is used as a training set to construct an ultimate bearing capacity prediction model. Based on the external force data of the first ten unit times, the external force of the eleventh unit time is predicted, and then the input is updated according to the value predicted by the ultimate bearing capacity prediction model. The iteration is stopped when the increment of the next time is less than 10%, thereby obtaining the ultimate bearing capacity. Specifically, the following steps are included:

[0106] (1) The external force data obtained by the external force prediction model is used as input (continuous 10 unit time external force) to obtain the external force of the 11th unit time.

[0107] (2) Remove the data of the first unit time in the input data, add the predicted external force to the end of the input data, obtain a new input, and predict the external force of the 12th unit time by the ultimate bearing capacity prediction model.

[0108] (3) Repeat (2) continuously until the external force of the next unit time is not increasing or the increment is less than 10%, and the prediction is stopped, thereby obtaining the prediction value of the ultimate bearing capacity of the composite wallboard assembly.

[0109] According to the method for predicting the ultimate bearing capacity of the composite wallboard assembly provided in the embodiments of the present application, the stress pictures of the composite wallboard assembly are collected by the collection device, the overall deformation information of the wallboard can be quickly obtained, the interference to the structure is reduced, the data collection efficiency and convenience are improved, the non-contact collection is realized, the local and global feature depth extraction and analysis of the tensile picture of the composite wallboard assembly are performed by the external force prediction model, the external force prediction value corresponding to each picture in the target picture set is obtained, and then the external force prediction value is input into the ultimate bearing capacity prediction model to obtain the ultimate bearing capacity through iterative prediction, so that the efficient conversion from the image data to the mechanical performance index is realized, and the accuracy and reliability of the prediction of the ultimate bearing capacity of the composite wallboard assembly are improved.

[0110] In some embodiments, the iterative prediction by the ultimate bearing capacity prediction model to obtain the prediction value of the ultimate bearing capacity of the composite wallboard assembly comprises:

[0111] The external force data of 10 unit times is obtained as an input sequence, the input sequence is input into the ultimate bearing capacity prediction model to obtain the prediction value of the external force of the next unit time;

[0112] The prediction value of the external force of the next unit time is added to the end of the original input sequence, and the external force data of the first unit time at the head of the original input sequence is deleted to obtain an updated input sequence;

[0113] The updated input sequence is input into the ultimate bearing capacity prediction model for iterative prediction to obtain the prediction value of the ultimate bearing capacity of the composite wallboard assembly.

[0114] In some embodiments, the iterative prediction by the ultimate bearing capacity prediction model to obtain the prediction value of the ultimate bearing capacity of the composite wallboard assembly comprises:

[0115] The updated input sequence is input into the ultimate bearing capacity prediction model to obtain the prediction value of the external force of the next unit time;

[0116] The increase amplitude of the prediction value of the external force of the next unit time and the current external force value is calculated;

[0117] When the increase amplitude is less than a preset threshold, the prediction value of the external force of the next unit time is taken as the prediction value of the ultimate bearing capacity of the composite wallboard assembly;

[0118] When the increase amplitude is greater than or equal to the preset threshold, the prediction value of the external force of the next unit time is added to the end of the original input sequence, and the external force data of the first unit time at the head of the original input sequence is deleted to obtain an updated input sequence, and the step of inputting the updated input sequence into the ultimate bearing capacity prediction model to obtain the prediction value of the external force of the next unit time is returned to.

[0119] For example, 9 groups of data on the change of tension over time of the composite wall panel assembly under tension were obtained under different working conditions, with data collected every 0.1 seconds. In the data preprocessing stage, the original tension-time curve was denoised by using the Gaussian smoothing method. The Gaussian smoothing algorithm is based on the Gaussian kernel function, which effectively suppresses noise interference and retains the true trend of the signal by weighting the average of the data points and their neighborhood. The window size of Gaussian smoothing is set to 61.

[0120] The unit time is set to 0.5 seconds. This setting is based on the consideration of the characteristics of the composite wall panel assembly under force and the frequency of data collection. Because in the actual tension process, the mechanical properties of the composite wall panel assembly change gradually over a certain time scale, setting the unit time to 0.5 seconds can capture the key features of the material under force, without making the data too fragmented, thus ensuring the continuity and integrity of the data.

[0121] Based on this unit time setting, the originally continuous data set is reasonably divided into 5 groups. This grouping method is to prepare for subsequent model training and verification, following the common training set and validation set division principle in machine learning, to ensure the good generalization ability of the model.

[0122] Next, one of the five groups of data is selected as the validation set, and the other four groups are selected as the training set. The validation set is used to evaluate the trained model and check the model's performance on unseen data, to judge the accuracy and stability of the model. The training set is used for the learning process of the model, so that the model can learn the law of the change of tension over time of the composite wall panel assembly under tension from these data.

[0123] In order to let the model learn the time series features in the data better, the sliding window operation is used for the training set and the validation set respectively, and the window size is set to 10. The sliding window operation is a common technique in time series data processing, which is like a moving window that slides over the data sequence, taking the data within the window as a sample each time. Specifically, for the training set and the validation set, the window starts from the beginning of the data and moves one unit of time backward each time, taking the data of 10 unit times (5 seconds) as a sample. In this way, through the sliding window operation, multiple samples can be generated from the original data sequence, each containing 10 unit time of tension data. These samples will be used as the input of the model, so that the model can learn the pattern and trend of the change of tension in different time periods, and the tension at the eleventh unit time point as the output.

[0124] For the trained LSTM prediction model, the ultimate bearing capacity prediction work is carried out. The specific operation process is: from the tensile test data of 9 groups of composite wall panel assemblies under different working conditions, the verification set data is selected, and the first ten unit time tension data of each verification set is selected as the initial input and input into the trained LSTM prediction model. The model outputs the predicted value of the tension of the 11th unit time according to the input data.

[0125] Subsequently, the input data is updated. The predicted tension data of the 11th unit time is added to the end of the original input data sequence, and the first value of the original input data sequence is removed, so as to form a new input sequence containing ten unit time tension data. The new input sequence is input into the LSTM prediction model again to perform the next round of tension prediction.

[0126] Stopping condition judgment: after each iteration prediction, the increment of the current predicted value compared with the previous predicted value is calculated. The increment calculation formula is:

[0127]

[0128] The above data updating and prediction process is continuously repeated to monitor the change of the tension increment in real time. When the increment is less than 10%, it is considered that the external force growth tends to be stable, reaches the limit state, and the iteration prediction is stopped. At this time, the external force value obtained by the last prediction is the ultimate bearing capacity prediction value of the composite wall panel assembly.

[0129] The relative error of the ultimate bearing capacity on the verification set of nine different working conditions is shown in Table 1.

[0130] Table 1

[0131]

[0132] As shown in Table 1, the LSTM prediction model performs well on the verification set, and the maximum error of the 5th group of working conditions is 0.0232, and the average relative error on the verification set of 9 different working conditions is 0.0123.

[0133] In this embodiment, by inputting the updated input sequence into the ultimate bearing capacity prediction model and calculating the increment of the predicted value of the external force of the next unit time and the current time external force value, the input of the ultimate bearing capacity prediction model is dynamically adjusted according to the increment to judge whether to update the input sequence, which effectively improves the adaptability and accuracy of the ultimate bearing capacity prediction model in the ultimate bearing capacity prediction process.

[0134] In this embodiment, by obtaining the external force data in a certain period of time and inputting it into the ultimate bearing capacity prediction model, the external force change of the composite wall panel assembly in the next time unit can be predicted. By continuously updating the input sequence and iteratively predicting, the predicted value of the bearing capacity can be updated in real time, obtaining a more accurate ultimate bearing capacity prediction value, having better applicability, being able to more comprehensively mine data features, and improving the accuracy of the ultimate bearing capacity prediction of the composite wall panel assembly.

[0135] In some embodiments, the training process of the external force prediction model comprises:

[0136] constructing a preset external force prediction model;

[0137] obtaining multiple stress pictures of the composite wall panel assembly and preprocessing to obtain a first data set;

[0138] performing a sliding window operation on the first data set to obtain a sliding window data set;

[0139] According to the sliding window data set, the preset external force prediction model is trained by a mean square error loss function to obtain a trained external force prediction model.

[0140] It is easy to understand that the CNN-Transformer prediction model is built, multiple stress pictures of the composite wall panel assembly are obtained and preprocessed to obtain a first data set, wherein the label is external force, the first data set includes photos and load data of the wall panel assembly, the load data includes external force corresponding to each photo shooting time, and the specific time of each photo shooting.

[0141] The pictures in the first data set are subjected to a sliding window operation to obtain a sliding window data set, and the size of the window is 4, that is, 4 adjacent pictures at different times are taken as input, and the external force at the last time is taken as output.

[0142] For example, the CNN-Transformer model is trained by using the sliding window data set containing 2868 groups of different working conditions to obtain a trained CNN-Transformer prediction model, and the prediction of the external force corresponding to each picture is realized. The optimizer uses Adam optimizer, the learning rate is set to 0.001, the decay rate of first moment estimation is set to 0.9, the decay rate of second moment estimation is set to 0.99, the weight decay is set to 0.0005, the loss function is set to mean square error loss function, and the maximum number of early stopping mechanism is 30.

[0143] In some embodiments, a CNN network is trained with the same training settings and training set, and the average relative error of the CNN model and the CNN-Transformer model is verified by 9 groups of different working conditions as shown in Table 2. Among them, the average relative error of the CNN model in the first group of working conditions is the largest, which is 0.4283. The average relative error in the ninth group of working conditions is the smallest, which is 0.0379. The average relative error of the CNN model on the validation set of the nine groups of different working conditions is 0.1242; the error of the CNN-Transformer model in the third group of working conditions is the largest, which is 0.1512. The average relative error in the eighth group of working conditions is the smallest, which is 0.0306. The average relative error of the CNN-Transformer model on the validation set of the nine groups of different working conditions is 0.0779.

[0144] Table 2

[0145]

[0146] As can be seen from Table 2, the CNN-Transformer model has a better prediction effect than the CNN model on the validation set of the nine groups of different working conditions, and the average relative error is improved by 37.3% compared with the CNN model.

[0147] In this embodiment, by obtaining a plurality of stress pictures of the composite wall panel assembly and preprocessing to obtain a first data set, and generating a sliding window data set through a sliding window operation, the quality of the training data is effectively enhanced. The preset external force prediction model is trained by using the mean square error loss function, which can realize the prediction of the stress condition of the composite wall panel assembly, and improve the accuracy and reliability of the external force prediction.

[0148] In some embodiments, the limit bearing capacity prediction model is trained based on the external force-time change curve, comprising:

[0149] The preset limit bearing capacity prediction model is an LSTM prediction model;

[0150] Based on the external force-time change curve, a plurality of data points are obtained, and the plurality of data points are cleaned to obtain cleaned data points;

[0151] The cleaned data points are identified and removed by using the 3σ principle to obtain removed data points;

[0152] The removed data points are filled by using linear interpolation to obtain a second data set;

[0153] According to the second data set, the preset ultimate bearing capacity prediction model is trained through a mean square error loss function to obtain a trained ultimate bearing capacity prediction model.

[0154] The training process of the ultimate bearing capacity prediction model includes the following steps:

[0155] (1) A preset ultimate bearing capacity prediction model is constructed, and the preset ultimate bearing capacity prediction model is an LSTM prediction model;

[0156] (2) Based on the external force-time change curve, a plurality of data points are obtained, the plurality of data points are cleaned to remove outliers and noise, a statistical analysis method such as the 3σ principle is used to identify and eliminate data points that obviously deviate from the normal range; for missing data, linear interpolation method can be used to fill in according to the trend of the previous and subsequent data to obtain a second data set;

[0157] (3) The number of network layers, the number of hidden layer neurons and other key parameters are determined. For example, 2-3 layers of LSTM layers can better capture the long-term dependence of time series data; the number of hidden layer neurons can be adjusted according to the complexity of the data and the computing resources, and is generally between 64-256. The mean square error is selected as the loss function for measuring the difference between the model prediction value and the true value. The Adam optimizer is selected as the optimizer, and its adaptive learning rate adjustment mechanism can automatically adjust the learning rate during the training process to speed up the convergence speed of the model. The initial learning rate is set to 0.001, and is adjusted in time according to the training situation, the decay rate of the first moment estimation is set to 0.9, the decay rate of the second moment estimation is set to 0.99, the weight decay is set to 0.0005, the maximum number of early stopping mechanism is 10, and the preset ultimate bearing capacity prediction model is trained through a mean square error loss function according to the second data set to obtain a trained ultimate bearing capacity prediction model.

[0158] Figure 4 is a flowchart of the second embodiment of the method for predicting the ultimate bearing capacity of the composite wall panel assembly provided by the present application, as shown in Figure 4 , the method includes the following steps:

[0159] A plurality of stress pictures of the composite wall panel assembly are collected by a collection device to obtain a picture data set, and the picture data set is preprocessed to obtain a target picture set;

[0160] The target picture set is input into the trained CNN-Transformer model to obtain the external force prediction value corresponding to each picture in the target picture set, and the external force-time change curve is obtained based on the external force prediction value corresponding to each picture.

[0161] (3) The external force-time change curve is arranged as a training set to ensure that the data is arranged in chronological order and the data format meets the model input requirements, thereby providing basic data for subsequent model training;

[0162] (4) An LSTM prediction model is constructed, and the number of hidden layers, the number of neurons, and other hyperparameters of the model are set reasonably according to the task requirements, thereby building a prediction framework capable of processing time series data;

[0163] (5) The external force data of the first ten unit times in the training set is input into the constructed LSTM prediction model, and the prediction value of the eleventh unit time is output through the calculation and learning of the model;

[0164] (6) The external force value predicted by the model in (5) is used as new input data, which is combined with the original time series data to form a new input sequence. The data of the first ten unit times in the new sequence is input into the model again to obtain the prediction value of the external force of the next unit time (i.e., the twelfth unit time). The process of data updating and model prediction is repeated;

[0165] (7) After each iteration, the increment of the external force value of the next time compared with the current time is calculated. When the increment is less than 10%, the iteration process is stopped, and the external force value obtained by the last prediction is the prediction value of the ultimate bearing capacity of the composite wall panel assembly.

[0166] In this embodiment, by constructing a preset ultimate bearing capacity prediction model and based on the external force-time change curve, a plurality of data points are obtained, the data is cleaned, removed, and interpolated to obtain a second data set, thereby effectively improving the quality of the model training data. By training the preset ultimate bearing capacity prediction model using the mean square error loss function, the ultimate bearing capacity of the composite wall panel assembly can be predicted, and the accuracy and reliability of the ultimate bearing capacity prediction model are improved.

[0167] The prediction method for the ultimate bearing capacity of the composite wall panel assembly provided in the embodiments of the present application can be executed by a prediction device for the ultimate bearing capacity of the composite wall panel assembly. In the embodiments of the present application, the prediction method for the ultimate bearing capacity of the composite wall panel assembly is executed by the prediction device for the ultimate bearing capacity of the composite wall panel assembly, which is taken as an example to illustrate the prediction device for the ultimate bearing capacity of the composite wall panel assembly provided in the embodiments of the present application.

[0168] The embodiments of the present application also provide a prediction device for the ultimate bearing capacity of the composite wall panel assembly, as shown in Figure 5 The prediction device for the ultimate bearing capacity of the composite wall panel assembly includes an acquisition module 510, a first processing module 520, a second processing module 530, and a prediction module 540.

[0169] The acquisition module 510 is configured to acquire multiple stress pictures of the composite wallboard assembly by using an acquisition device, to obtain a picture dataset, each stress picture corresponding to a time step, and to preprocess each stress picture in the picture dataset to obtain a target picture set;

[0170] The first processing module 520 is configured to input the target picture set into a trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set.

[0171] The second processing module 530 is configured to obtain an external force-time change curve based on the external force prediction value corresponding to each picture.

[0172] The prediction module 540 is configured to train a limit bearing capacity prediction model based on the external force-time change curve, and to obtain a prediction value of the limit bearing capacity of the composite wallboard assembly by iterative prediction of the limit bearing capacity prediction model.

[0173] According to the prediction method of the limit bearing capacity of the composite wallboard assembly provided in the embodiments of the present application, multiple stress pictures of the composite wallboard assembly are acquired by using an acquisition device, so that the overall deformation information of the wallboard can be quickly obtained, the interference with the structure is reduced, the data acquisition efficiency and convenience are improved, and non-contact acquisition is achieved. The external force prediction model is used to perform local and global feature depth extraction and analysis on the tensile pictures of the composite wallboard assembly, to obtain an external force prediction value corresponding to each picture in the target picture set, which is then input into the limit bearing capacity prediction model. The limit bearing capacity is obtained through iterative prediction, the efficient conversion from image data to mechanical performance indicators is achieved, and the accuracy and reliability of the prediction of the limit bearing capacity of the composite wallboard assembly are improved.

[0174] The prediction device for the limit bearing capacity of the composite wallboard assembly provided in the embodiments of the present application can implement the prediction method for the limit bearing capacity of the composite wallboard assembly Figures 1 to 5 The prediction device for the limit bearing capacity of the composite wallboard assembly provided in the embodiments of the present application can implement the prediction method for the limit bearing capacity of the composite wallboard assembly

[0175] In some embodiments, as shown in Figure 6 The present application also provides an electronic device 600, including a processor 601, a memory 602, and a computer program stored in the memory 602 and executable on the processor 601. The program is executed by the processor 601 to implement each process of the above-mentioned prediction method for the limit bearing capacity of the composite wallboard assembly, and can achieve the same technical effects. To avoid repetition, details are not repeated here.

[0176] It should be noted that the electronic device in the embodiments of the present application includes the mobile electronic device and the non-mobile electronic device described above.

[0177] The embodiment of the present application further provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement each process of the method for predicting the ultimate bearing capacity of the composite panel assembly and achieve the same technical effects. To avoid repetition, details are not described herein.

[0178] The processor is a processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0179] The embodiment of the present application further provides a computer program product, which includes a computer program, and the computer program is executed by a processor to implement the method for predicting the ultimate bearing capacity of the composite panel assembly.

[0180] The processor is a processor in the electronic device in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0181] The embodiment of the present application further provides a chip, which includes a processor and a communication interface. The communication interface is coupled with the processor. The processor is configured to execute a program or an instruction to implement each process of the method for predicting the ultimate bearing capacity of the composite panel assembly and achieve the same technical effects. To avoid repetition, details are not described herein.

[0182] It should be understood that the chip mentioned in the embodiment of the present application can also be referred to as a device-level chip, a device chip, a chip device or a system-on-chip device, etc.

[0183] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a", "comprising", or "includes a", does not, without more constraints, preclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element. Additionally, it should be noted that the scope of the methods and apparatus of the present embodiments are not limited by the order of the steps or the sequence for performing the steps, as some steps can occur simultaneously, in other steps can occur sequentially, or in between other steps can occur, as can be appreciated by one of ordinary skill in the art. Further, features described with respect to certain examples can be combined in other examples.

[0184] From the above description of the embodiments, it is apparent that the method of the above-mentioned embodiments can be realized by means of software plus necessary universal hardware platforms, of course, can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a computer software product in essence or the part of the prior art that contributes to the present application, which is stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc), and includes a plurality of instructions for making a terminal (which can be a mobile phone, computer, server, or network device, etc.) execute the method of predicting the limit bearing capacity of the composite wallboard assembly of each embodiment of the present application.

[0185] In the description of the present application, "first feature" and "second feature" can include one or more of the features.

[0186] In the description of the present application, "a plurality of" means two or more.

[0187] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative, not restrictive, and those of ordinary skill in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the scope protected by the claims.

[0188] In the description of the application, reference has been made to descriptive terms such as "one embodiment", "some embodiments", "an embodiment", "example", "specific example" or "some examples" etc. It is emphasized that each of these terms refers to a specific feature, structure, material or characteristic described in connection with a particular embodiment or example. The descriptive terms are not necessarily referring to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0189] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since the scope of the application is defined with respect to the appended claims.

Claims

1. A method of predicting the ultimate load capacity of a composite wall panel assembly, characterized by, The method comprises: collecting a plurality of stress pictures of the composite wallboard assembly by a collection device to obtain a picture data set, each stress picture corresponding to a time step, and preprocessing each stress picture in the picture data set to obtain a target picture set; inputting the target picture set into a trained external force prediction model to obtain an external force prediction value corresponding to each picture in the target picture set; obtaining an external force-time change curve based on the external force prediction value corresponding to each picture; training a limit bearing capacity prediction model based on the external force-time change curve, and iteratively predicting by the limit bearing capacity prediction model to obtain a prediction value of the limit bearing capacity of the composite wallboard assembly; the preprocessing of each stress picture in the picture data set to obtain the target picture set comprises: creating a mask based on the most obvious part of the material wallboard deformation with external force in each stress picture, and setting all picture parts outside the mask to black; cropping the black part in each stress picture to obtain a cropped stress picture; smoothing the boundary of each cropped stress picture based on an expansion operation to make the edge of each stress picture smoother and the picture more continuous to obtain the target picture set; the inputting of the target picture set into the trained external force prediction model to obtain the external force prediction value corresponding to each picture in the target picture set comprises: inputting the target picture set into a CNN module to obtain local features of each picture in the target picture set; inputting the target picture set into a Transformer module to obtain global features of each picture in the target picture set; fusing the local features and the global features of each picture based on a gating mechanism to obtain a fusion feature corresponding to each picture; inputting the fusion feature corresponding to each picture into a fully connected network to obtain the external force prediction value corresponding to each picture in the target picture set; wherein the external force prediction model is a CNN-Transformer model, and the CNN-Transformer model comprises a CNN module and a Transformer module.

2. The method of predicting the ultimate load capacity of a composite wall panel assembly of claim 1, wherein, the iterative prediction by the limit bearing capacity prediction model to obtain the prediction value of the limit bearing capacity of the composite wallboard assembly comprises: obtaining external force data of 10 unit times as an input sequence, inputting the input sequence into the limit bearing capacity prediction model to obtain a prediction value of external force of the next unit time; adding the prediction value of external force of the next unit time to the end of the original input sequence, deleting the external force data of the first unit time at the head of the original input sequence to obtain an updated input sequence; inputting the updated input sequence into the limit bearing capacity prediction model for iterative prediction to obtain the prediction value of the limit bearing capacity of the composite wallboard assembly.

3. The method of predicting the ultimate load capacity of a composite wall panel assembly of claim 2, wherein, the inputting of the updated input sequence into the limit bearing capacity prediction model for iterative prediction to obtain the prediction value of the limit bearing capacity of the composite wallboard assembly comprises: inputting the updated input sequence into the limit bearing capacity prediction model to obtain a prediction value of external force of the next unit time; calculating an increment of the prediction value of external force of the next unit time and a current time external force value; When the increment is less than the preset threshold, the predicted value of the next unit time external force is taken as the predicted value of the ultimate bearing capacity of the composite panel assembly; When the increment is greater than or equal to the preset threshold, the predicted value of the next unit time external force is added to the end of the original input sequence, the first unit time external force data at the beginning of the original input sequence is deleted to obtain an updated input sequence, and the step of inputting the updated input sequence into the ultimate bearing capacity prediction model is returned to obtain the predicted value of the next unit time external force.

4. The method of predicting the ultimate load capacity of a composite wall panel assembly of claim 1, wherein, The training process of the external force prediction model comprises: constructing a preset external force prediction model; obtaining multiple stress pictures of the composite panel assembly and preprocessing to obtain a first data set; performing a sliding window operation on the first data set to obtain a sliding window data set; training the preset external force prediction model based on the sliding window data set through a mean square error loss function to obtain a trained external force prediction model.

5. The method of predicting the ultimate load capacity of a composite wall panel assembly of claim 1 wherein, The training of the ultimate bearing capacity prediction model based on the external force-time curve comprises: constructing a preset ultimate bearing capacity prediction model, which is an LSTM prediction model; obtaining multiple data points based on the external force-time curve, cleaning the multiple data points to obtain cleaned data points; identifying and removing the cleaned data points through the 3σ principle to obtain removed data points; filling the removed data points through linear interpolation to obtain a second data set; training the preset ultimate bearing capacity prediction model through a mean square error loss function based on the second data set to obtain a trained ultimate bearing capacity prediction model.

6. A device for predicting the ultimate load capacity of a composite panel assembly, implemented by the method for predicting the ultimate load capacity of a composite panel assembly according to any one of claims 1 to 5, characterized in that, The device comprises: an acquisition module configured to acquire multiple stress pictures of the composite panel assembly through a collection device to obtain a picture data set, each stress picture corresponding to a time step, and preprocess each stress picture in the picture data set to obtain a target picture set; a first processing module configured to input the target picture set into the trained external force prediction model to obtain a predicted value of external force corresponding to each picture in the target picture set; a second processing module configured to obtain an external force-time curve based on the predicted value of external force corresponding to each picture; a prediction module configured to train an ultimate bearing capacity prediction model based on the external force-time curve, and iteratively predict through the ultimate bearing capacity prediction model to obtain a predicted value of the ultimate bearing capacity of the composite panel assembly.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the prediction method of the ultimate bearing capacity of the composite panel assembly according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the prediction method of the ultimate bearing capacity of the composite panel assembly according to any one of claims 1 to 5.

Images