Airbag state reconstruction method, device and computer equipment

Abstract

The application relates to an air bag state reconstruction method and device and computer equipment. The method comprises the following steps: acquiring node displacement strain corresponding to air bag node displacement; the air bag node displacement is determined according to inverse shell elements obtained by grid division on the air bag; a functional model of the inverse shell element is obtained according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the air bag, and the accuracy of the first strain is higher than that of the second strain; a pseudo-rigidity matrix and a pseudo-load vector of the inverse shell element are determined based on the functional model; a node displacement model of the air bag is obtained according to the pseudo-rigidity matrix and the pseudo-load vector; target node displacement of the air bag is determined based on the node displacement model, and the state of the air bag is reconstructed according to the target node displacement. The method can improve the state reconstruction accuracy of the air bag.

Description

Technical Field

[0001] This application relates to the field of underwater flexible gas storage technology, and in particular to a method, apparatus and computer equipment for reconstructing the state of an airbag. Background Technology

[0002] Underwater flexible airbags are an important component of underwater compressed air energy storage systems based on flexible air storage. During the system's energy storage and release processes, the airbags are excited by the internal flow of compressed gas and the external flow of water fluid, and coupled with the gas delivery pipeline and ballast system, they will repeatedly undergo large-scale deformation movements. Local deformation phenomena such as wrinkling may occur, and stress concentration is prone to occur at the connection points between the underwater flexible airbags and their gas delivery pipelines and ballast system. All of these factors can lead to certain fatigue damage risks for the airbags. Accurate monitoring of the strain during the deformation process of the airbags can reconstruct their deformation state, provide data for damage assessment of key nodes of the airbags, and enable timely response to fatigue damage.

[0003] The rapid development of sensors and information technology has enabled them to play a crucial role in structural monitoring. Structural monitoring systems are key to public safety, providing early warnings of unsafe conditions through accurate measurement and condition monitoring, supporting proactive maintenance plans, and helping to reduce inspection burdens and maintenance costs. However, in practical applications, the surface area of ​​underwater flexible air-storage bladders is usually quite large, making the deployment of numerous high-precision sensors costly and even difficult. While low-precision sensors are inexpensive and easy to deploy in large quantities, their measurement accuracy and results are low.

[0004] Therefore, current underwater flexible gas storage airbag state reconstruction technology suffers from low state reconstruction accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide an airbag state reconstruction method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve the accuracy of state reconstruction in response to the above-mentioned technical problems.

[0006] Firstly, this application provides a method for reconstructing the state of an airbag, including:

[0007] Obtain the nodal displacement strain corresponding to the nodal displacement of the airbag; the nodal displacement of the airbag is determined based on the inverse shell element obtained by meshing the airbag.

[0008] The functional model of the inverse shell element is obtained based on the nodal displacement strain and the strain estimate corresponding to the nodal displacement strain; the strain estimate is obtained by fusing the first strain and the second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain.

[0009] determine a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model;

[0010] obtain a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector;

[0011] determine a target node displacement of the airbag based on the node displacement model, and reconstruct a state of the airbag according to the target node displacement.

[0012] In a second aspect, the present application further provides an airbag state reconstruction device, comprising:

[0013] a strain obtaining module configured to obtain a node displacement strain corresponding to a node displacement of an airbag, wherein the node displacement of the airbag is determined according to an inverse shell element obtained by grid division of the airbag;

[0014] a functional model module configured to obtain a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain, wherein the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the first strain has a higher accuracy than the second strain;

[0015] a matrix determining module configured to determine a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model;

[0016] a displacement model module configured to obtain a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector;

[0017] a state reconstruction module configured to determine a target node displacement of the airbag based on the node displacement model, and reconstruct a state of the airbag according to the target node displacement.

[0018] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0019] obtain a node displacement strain corresponding to a node displacement of an airbag, wherein the node displacement of the airbag is determined according to an inverse shell element obtained by grid division of the airbag;

[0020] obtain a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain, wherein the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the first strain has a higher accuracy than the second strain;

[0021] determine a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model;

[0022] obtaining a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector;

[0023] determining a target node displacement of the airbag based on the node displacement model, and reconstructing a state of the airbag according to the target node displacement.

[0024] In a fourth aspect, the present application further provides a computer readable storage medium, which has a computer program stored thereon, and the computer program is executed by a processor to implement the following steps:

[0025] obtaining a node displacement strain corresponding to a node displacement of an airbag; the node displacement of the airbag is determined according to an inverse shell element obtained by grid division of the airbag;

[0026] obtaining a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain;

[0027] determining a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model;

[0028] obtaining a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector;

[0029] determining a target node displacement of the airbag based on the node displacement model, and reconstructing a state of the airbag according to the target node displacement.

[0030] In a fifth aspect, the present application further provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the following steps:

[0031] obtaining a node displacement strain corresponding to a node displacement of an airbag; the node displacement of the airbag is determined according to an inverse shell element obtained by grid division of the airbag;

[0032] obtaining a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain;

[0033] determining a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model;

[0034] obtaining a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector;

[0035] determine a target node displacement of the airbag based on the node displacement model, and reconstruct a state of the airbag according to the target node displacement.

[0036] The airbag state reconstruction method, device, computer equipment, computer readable storage medium and computer program product described above, by obtaining a node displacement strain corresponding to an airbag node displacement, the airbag node displacement being determined according to an inverse shell element obtained by grid division of the airbag, obtaining a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain, the strain estimation value being obtained by fusing a first strain and a second strain of the airbag, the accuracy of the first strain being higher than that of the second strain, determining a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model, obtaining a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector, determining a target node displacement of the airbag based on the node displacement model, and reconstructing a state of the airbag according to the target node displacement; the first strain of the airbag can be collected by a high-precision sensor, the second strain of the airbag can be collected by a low-precision sensor, the strain estimation value can be obtained by data fusion of the first strain and the second strain, the estimation accuracy of the strain observation point is improved, and the strain estimation value can be used for state reconstruction of the airbag, which can improve the state reconstruction accuracy of the underwater flexible gas storage airbag. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without creating any inventive labor.

[0038] Figure 1 A flowchart of an airbag state reconstruction method in one embodiment;

[0039] Figure 2 A schematic diagram of an underwater flexible airbag membrane structure in one embodiment;

[0040] Figure 3 A schematic diagram of a distributed strain sensor layer in one embodiment;

[0041] Figure 4 A schematic diagram of point and distributed strain sensor layout in one embodiment;

[0042] Figure 5 A schematic diagram of the distance from the point and distributed strain sensors to the airbag surface in one embodiment;

[0043] Figure 6 A schematic diagram of a planar four-node element in the local coordinate system xoy in one embodiment;

[0044] Figure 7 A schematic diagram of a planar four-node isoparametric element using bilinear interpolation in an embodiment;

[0045] Figure 8 A flowchart of an airbag state reconstruction method in another embodiment;

[0046] Figure 9 A block diagram of an airbag state reconstruction device in an embodiment;

[0047] Figure 10 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0048] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0049] It should be noted that the terms "first", "second", and the like used in the present application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "include" and "have" and any variations thereof used in the present application are intended to cover non-exclusive inclusion. The term "multiple" used in the present application refers to two or more. The term "and / or" used in the present application refers to one of the options or any combination of multiple options.

[0050] In an exemplary embodiment, as shown in Figure 1 An airbag state reconstruction method is provided, which is described by taking a terminal as an example, and includes the following steps S101 to S105. Wherein:

[0051] Step S101, obtaining a node displacement strain corresponding to an airbag node displacement; the airbag node displacement is determined according to an inverse shell element obtained by meshing the airbag.

[0052] Wherein, the airbag can be but is not limited to an underwater flexible gas storage airbag. The inverse shell element is an element type for structure displacement field reconstruction, which inversely solves the displacement field through the strain-displacement relationship. The airbag node displacement can be a displacement vector of the inverse shell element node, which can be understood as an unknown variable. The node displacement strain can be a strain caused by the inverse shell element node displacement, including but not limited to a membrane strain, a bending strain and a transverse shear strain, which can be understood as a theoretical strain value corresponding to the unknown variable.

[0053] Optionally, the terminal can mesh the air bag to obtain an inverse shell unit, take the node displacement vector of the inverse shell unit after deformation as the air bag node displacement, determine the membrane strain, bending strain and transverse shear strain according to the air bag node displacement, and obtain the node displacement strain according to the membrane strain, bending strain and transverse shear strain.

[0054] Exemplarily, the inverse finite element method can be used to reconstruct the deformation state of the underwater flexible gas storage air bag. The underwater flexible gas storage air bag is meshed to obtain a four-node inverse shell unit, each inverse shell unit node has three translation coordinates and three rotation coordinates , the node displacement vector of the four-node inverse shell unit after deformation is denoted as , and the strain of the four-node inverse shell unit is as follows:

[0055] .

[0056] wherein, is an unknown variable, and represents the air bag node displacement, ( ), is a theoretical strain value corresponding to the unknown variable, that is, the node displacement strain, wherein, represents the membrane strain, represents the bending strain, represents the transverse shear strain, is the normal strain, is the weight.

[0057] wherein, , and are strain displacement relationship matrices of the inverse shell unit, ( ), and:

[0058] .

[0059] .

[0060] .

[0061] wherein, is a four-node isoparametric unit shape function, and a bilinear isoparametric shape function is used, so:

[0062] .

[0063] .

[0064] .

[0065] .

[0066] where is the normalized nodal coordinate of the isoparametric element, the isoparametric coordinate ; the inverse shell element coordinate is and , , and is the nodal coordinate index vector.

[0067] In step S102, a functional model of the inverse shell element is obtained according to the nodal displacement strain and the strain estimation value corresponding to the nodal displacement strain. The strain estimation value is obtained by fusing the first strain and the second strain of the airbag. The accuracy of the first strain is higher than that of the second strain.

[0068] wherein the strain estimation value can be the fusion result of the first strain and the second strain. The functional model can be a functional of a four-node inverse shell element considering the film strain, the bending strain and the transverse shear strain, which is constructed by a weighted least square function. The first strain and the second strain can be strain measurement values of the airbag. The first strain is a high-precision strain measurement value, and the second strain is a low-precision strain measurement value.

[0069] Optionally, a point strain sensor and a distributed strain sensor can be arranged on the airbag. The measurement accuracy of the point strain sensor is higher than that of the distributed strain sensor, but the number of points that can be monitored by the point strain sensor is less than that of the distributed strain sensor. Therefore, the terminal can collect high-precision but low-spatial-resolution strain measurement values as the first strain through the point strain sensor, and collect low-precision but high-spatial-resolution strain measurement values as the second strain through the distributed strain sensor. The terminal can then perform data fusion on the first strain and the second strain to obtain a strain estimation value that takes into account both accuracy and spatial resolution, and then construct a functional model according to the nodal displacement strain and the strain estimation value corresponding thereto.

[0070] Exemplarily, assuming that the strain estimation value obtained by data fusion has a film strain estimation value , a bending strain estimation value , and a transverse shear strain estimation value , the film strain estimation value can be ignored due to the small thickness of the airbag. Considering the film strain, the bending strain and the transverse shear strain, the functional of the four-node inverse shell element constructed by the weighted least square function is:

[0071] .

[0072] wherein ,​​ and are weights; , and are strain measurements, or strain estimates obtained from a data fusion method; denotes the L2 norm symbol; denotes the area of the current inverse shell element.

[0073] Step S103, determining the pseudo stiffness matrix and the pseudo load vector of the inverse shell element based on the functional model.

[0074] where the pseudo stiffness matrix is a matrix similar to the stiffness matrix in finite elements. The pseudo load vector is a vector similar to the load vector in finite elements.

[0075] Optionally, the terminal can extremize the functional model to obtain an equation representation of the inverse shell element, and solve the equation representation to obtain the pseudo stiffness matrix and the pseudo load vector of the inverse shell element.

[0076] Exemplarily, considering that the strain theoretical value is as close as possible to the strain estimate value, a functional about the element node displacement vector can be minimized to obtain the inverse shell element equation:

[0077] .

[0078] Solving obtains:

[0079] .

[0080] .

[0081] wherein, is the element pseudo stiffness matrix, is the element pseudo load vector, is the strain matrix corresponding to the element membrane strain, is the strain matrix corresponding to the element bending strain, is the strain matrix corresponding to the element energy shear strain, denotes the transposition operation.

[0082] Step S104, obtaining the node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector.

[0083] wherein the node displacement model refers to a system equation representing the node displacement of all four-node inverse shell elements of the airbag.

[0084] Optionally, the terminal can assemble and integrate all the inverse shell elements according to the pseudo stiffness matrix, the pseudo load vector and the transformation matrix of the inverse shell element to obtain the node displacement model of the entire airbag.

[0085] For example, based on the pseudo-stiffness matrix of the inverse shell element and pseudo-load vector and the transformation matrix of the inverse shell element. By assembling and integrating all four-node inverse shell elements, the system equations in the global coordinate system are obtained:

[0086] .

[0087] .

[0088] in, This is the total number of four-node inverse shell elements obtained by meshing an underwater flexible gas storage bladder. The overall pseudo-stiffness matrix, Let be the total nodal displacement vector. This is the overall pseudo-load vector. The system equations... As a nodal displacement model for airbags.

[0089] Step S105: Determine the target node displacement of the airbag based on the node displacement model, and reconstruct the state of the airbag according to the target node displacement.

[0090] The target nodal displacement can be the nodal displacement when the theoretical strain value and the strain estimate are close.

[0091] Optionally, the terminal can obtain the target node displacement of the airbag by solving the node displacement model, sum the target node displacement with the original node coordinates to obtain the deformed node coordinates, and use the deformed node coordinates to reconstruct the state of the airbag.

[0092] For example, the system equations By applying boundary conditions, a reduced equation can be obtained. Solving the reduced equation yields ,in, This is the reduced pseudo-stiffness matrix. This is the reduced nodal displacement vector. This is the reduced pseudo-load vector. Assume that when meshing the underwater flexible air-storage bladder, the nodes... The initial coordinates are Solving for the nodes The displacement is , If the total number of nodes is [number], then the coordinates of the nodes after deformation are [coordinates]. ( ).

[0093] The air bag state reconstruction method described above, by obtaining the node displacement strain corresponding to the air bag node displacement, the air bag node displacement is determined according to the inverse shell element obtained by meshing the air bag, and the functional model of the inverse shell element is obtained according to the node displacement strain and the strain estimation value corresponding to the node displacement strain. The strain estimation value is obtained by fusing the first strain and the second strain of the air bag. The accuracy of the first strain is higher than that of the second strain. The pseudo stiffness matrix and the pseudo load vector of the inverse shell element are determined based on the functional model. The node displacement model of the air bag is obtained according to the pseudo stiffness matrix and the pseudo load vector. The target node displacement of the air bag is determined based on the node displacement model. The state of the air bag is reconstructed according to the target node displacement; the first strain of the air bag can be collected by a high-precision sensor, the second strain of the air bag can be collected by a low-precision sensor, and the strain estimation value can be obtained by fusing the first strain and the second strain. The estimation accuracy of the strain observation point is improved. The strain estimation value is used for state reconstruction of the air bag, which can improve the state reconstruction accuracy of the underwater flexible gas storage air bag.

[0094] In one exemplary embodiment, the air bag state reconstruction method described above can further include:

[0095] Step S201, obtaining a first data set and a second data set of the air bag; the accuracy of the first data set is higher than that of the second data set;

[0096] Step S202, determining the first strain of the air bag according to the first data set, and determining the second strain of the air bag according to the second data set; the first strain includes the first bending strain and the first film strain, and the second strain includes the second bending strain and the second film strain;

[0097] Step S203, determining the strain estimation value corresponding to the node displacement strain according to the first strain and the second strain; the strain estimation value includes the bending strain estimation value and the film strain estimation value.

[0098] Wherein, the first data set can be the strain value collected by a high-precision sensor. The second data set can be the strain value collected by a low-precision sensor. The first strain can be a high-precision strain measurement value. The second strain can be a low-precision strain measurement value. The first bending strain can be a high-precision bending strain measurement value. The first film strain can be a high-precision film strain measurement value. The second bending strain can be a low-precision bending strain measurement value. The second film strain can be a low-precision film strain measurement value.

[0099] Optionally, point strain sensors can be provided on the air bag. The point strain sensors can be installed on the inner surface and the outer surface of the air bag, and the positions of the point strain sensors installed on the inner surface and the outer surface of the air bag in the air bag membrane structure are one-to-one corresponding. The point strain sensors collect the first data set , wherein the position of the projection of the membrane structure in the airbag, high-precision strain data collected by point strain sensors mounted on the inner surface of the airbag, high-precision strain data collected by point strain sensors mounted on the outer surface of the airbag.

[0100] A distributed strain sensor can also be provided on the airbag, which can be deployed on two layers in the membrane structure, the two layers being located on two sides of the membrane structure respectively, and the positions of the projections of the two layers of distributed strain sensors in the membrane structure correspond one-to-one, and the distributed strain sensor collects a second data set , wherein the position of the projection of the membrane structure in the airbag, low-precision strain data collected by the distributed strain sensor mounted on one side of the membrane structure, low-precision strain data collected by the distributed strain sensor mounted on the other side of the membrane structure.

[0101] According to the high-precision first data set the high-precision bending strain and the membrane strain are calculated to obtain the first bending strain and the first membrane strain , according to the low-precision second data set the low-precision bending strain and the membrane strain are calculated to obtain the second bending strain and the second membrane strain , the specific formula is:

[0102] .

[0103] .

[0104] .

[0105] .

[0106] The first bending strain and the second bending strain are fused to obtain the bending strain estimate , the first membrane strain and the second membrane strain are fused to obtain the membrane strain estimate .

[0107] In the embodiment, by acquiring the first data set and the second data set of the air bag, determining the first strain of the air bag according to the first data set, determining the second strain of the air bag according to the second data set, and determining the strain estimation value corresponding to the node displacement strain according to the first strain and the second strain, the air bag strain estimation value considering the accuracy and spatial resolution can be constructed, and the accuracy of air bag state reconstruction is improved.

[0108] In an exemplary embodiment, the air bag comprises an inner liner layer, a first distributed strain sensing layer, a bearing layer, a second distributed strain sensing layer and an outer protective layer; the first distributed strain sensing layer is provided with first distributed strain sensors, the second distributed strain sensing layer is provided with second distributed strain sensors, and the arrangement positions of the first distributed strain sensors correspond to the arrangement positions of the second distributed strain sensors; the inner liner layer is pasted with first point strain sensors, and the outer protective layer is pasted with second point strain sensors, and the arrangement positions of the first point strain sensors correspond to the arrangement positions of the second point strain sensors; the above step S201 can specifically include: acquiring the first data set collected by the first point strain sensors and the second point strain sensors, and acquiring the second data set collected by the first distributed strain sensors and the second distributed strain sensors.

[0109] Optionally, the membrane structure of the underwater flexible gas storage air bag can be sequentially from inside to outside of the air bag: an inner liner layer, a first distributed strain sensing layer, a bearing layer, a second distributed strain sensing layer and an outer protective layer. The inner liner layer and the outer protective layer can be made of polyvinyl chloride material. The bearing layer can be made of polyester fiber material. The distributed strain sensing layer includes distributed strain sensors and their substrates, wherein the substrate can be made of polyimide material, and the distributed strain sensors can form a quadrature two-dimensional sensing network and be encapsulated inside the substrate material. The inner liner layer can be coated on the inner side of the first distributed strain sensing layer, the first distributed strain sensing layer and the second distributed strain sensing layer can be bonded on the inner side and the outer side of the bearing layer respectively, and the outer protective layer can be coated on the outer side of the second distributed strain sensing layer.

[0110] The first distributed strain sensors can be arranged in the first distributed strain sensing layer, and the second distributed strain sensors can be arranged in the second distributed strain sensing layer. The positions of the first distributed strain sensors projected on the membrane structure of the air bag can correspond to the positions of the second distributed strain sensors projected on the membrane structure of the air bag one by one.

[0111] The point strain sensors can be arranged on the inner surface and the outer surface of the airbag, the first point strain sensor can be arranged on the inner side of the inner liner in a surface pasting manner, and the second point strain sensor can be arranged on the outer side of the outer protective layer in a surface pasting manner.

[0112] The point strain sensors and the distributed strain sensors can measure the normal strain and the shear strain in two orthogonal directions at the positions where the point strain sensors and the distributed strain sensors are arranged, and are used to monitor the strain in real time. The first point strain sensor and the second point strain sensor can collect the first data set and transmit the first data set to the terminal. The first distributed strain sensor and the second distributed strain sensor can collect the second data set and transmit the second data set to the terminal. It can be understood that the measurement accuracy of the point strain sensor is higher than the measurement accuracy of the distributed strain sensor, and therefore the accuracy of the first data set is higher than the accuracy of the second data set. However, the number of points that can be monitored by the point strain sensor is less than the number of points that can be monitored by the distributed strain sensor, that is, the spatial resolution of the first data set is less than the spatial resolution of the second data set.

[0113] In the embodiment, the first data set collected by the first point strain sensor and the second point strain sensor and the second data set collected by the first distributed strain sensor and the second distributed strain sensor are obtained, and high-precision sensors with high cost and a large number of low-precision sensors with low cost are arranged to obtain high-precision measurement results, and long-term reliable structure health monitoring is realized.

[0114] In an exemplary embodiment, the step S203 can specifically include: training the second strain to obtain a trained deep neural network; determining a first strain estimation value corresponding to the first strain according to the deep neural network; determining a transfer learning model of the first strain according to the first strain estimation value; inputting the second strain estimation value of the airbag node determined according to the deep neural network into the transfer learning model to obtain a third strain estimation value of the airbag node; and taking the third strain estimation value as a strain estimation value corresponding to the node displacement strain.

[0115] The first strain estimation value can be a low-precision strain estimation value corresponding to a high-precision strain estimated by the deep neural network. The airbag node can be a specified node on the airbag corresponding to the airbag node displacement and the node displacement variable. The second strain estimation value can be a low-precision strain estimation value obtained by inputting the coordinates of the airbag node into the deep neural network. The third strain estimation value can be a high-precision strain estimation value obtained by transfer learning on the second strain estimation value.

[0116] Optionally, a multi-fidelity data fusion method of deep neural network can be used to construct the bending strain estimation value of the airbag with consideration of the accuracy and spatial resolution according to the high-precision bending strain data set and the low-precision bending strain data set . Specifically, the method can include the following steps:

[0117] Step S301, according to the low-precision bending strain data set , train a deep neural network (DNN) and its parameters, and input the high-precision position information into the deep neural network to obtain the low-precision estimation value corresponding to the high-precision bending strain , that is, the first strain estimation value.

[0118] Step S302, define the transfer learning model of the high-precision bending strain data set :

[0119] .

[0120] .

[0121] .

[0122] wherein, is the order of the polynomial, is the transfer learning parameter, is a Gaussian Process (GP) with mean 0, and the covariance function is a Gaussian kernel function:

[0123] .

[0124] wherein, and are the hyperparameters of the Gaussian kernel.

[0125] Step S303, define the optimization problem:

[0126] .

[0127] wherein, is the number of data points in the high-precision bending strain data set. By minimizing the residual function using a linear regression algorithm, the estimation value of the transfer learning parameter is obtained, and the high-precision residual data set .

[0128] Step S304, through the high-precision residual data set training a Bayesian neural network (BNN) and parameters thereof to obtain an estimate of the residual .

[0129] Step S305, according to the DNN model of the low-precision bending strain data set , the estimate of the transfer learning parameter , the estimate of the residual , the bending strain estimate value at the given point is calculated , wherein is the position coordinate of the balloon node, is the second strain estimate value, is the third strain estimate value.

[0130] By the same method as the above steps, the film strain estimate value at the given point is obtained , and and are taken as strain estimate values corresponding to the node displacement strain to construct a functional model.

[0131] In this embodiment, the second strain is trained to obtain a trained deep neural network, the first strain estimate value corresponding to the first strain is determined according to the deep neural network, the transfer learning model of the first strain is determined according to the first strain estimate value, the second strain estimate value of the balloon node determined according to the deep neural network is input into the transfer learning model, and the third strain estimate value of the balloon node is obtained. The third strain estimate value is taken as the strain estimate value corresponding to the node displacement strain, the strain estimate value considering the accuracy and spatial resolution can be constructed, and the accuracy of the balloon state reconstruction is improved.

[0132] In an exemplary embodiment, the above step S103 can specifically include: determining a target cube containing the balloon; recursively dividing the target cube using an octree to obtain a plurality of sub-cubes of the target cube; the plurality of sub-cubes correspond to a plurality of sub-layers, and the plurality of sub-layers contain a target sub-layer; determining the number of inverse shell units in each sub-cube of the target sub-layer; and determining the pseudo load vector of the inverse shell unit according to the number of inverse shell units and whether the inverse shell unit contains the first strain of the balloon.

[0133] Wherein, the target cube refers to the smallest cube containing the balloon. The target sub-layer refers to the sub-layer obtained by the last division.

[0134] Optionally, the pseudo load vector of the inverse shell unit can be calculated by the following steps:

[0135] Step S401, assuming that the node coordinates at the initial time or the previous reconstruction time are ( ), is the total number of nodes, and the minimum cube containing the underwater flexible gas storage air bag is obtained , that is, the target cube.

[0136] Step S402, the minimum cube is recursively divided by using an octree, and the division level is , represents rounding up, is the total number of inverse shell units. The first division obtains eight cubes in the first sub-layer, the second division obtains 64 cubes in the second sub-layer, and so on. The th division obtains cubes in the th sub-layer, ,

[0137] Step S403, the number of inverse shell units in each cube ( ) in the th sub-layer (target sub-layer) is counted . If the center point of the inverse shell unit is in the cube , the inverse shell unit is in the cube .

[0138] Step S404, according to the value of and whether there is a high-precision strain measurement value in the inverse shell unit, the pseudo load vector of the inverse shell unit is obtained by calculation.

[0139] In this embodiment, by determining the target cube containing the air bag, the target cube is recursively divided by using an octree, a plurality of sub-cubes of the target cube are obtained, the number of inverse shell units in each sub-cube of the target sub-layer is determined, and the pseudo load vector of the inverse shell unit is determined according to the number of inverse shell units and whether the first strain of the inverse shell unit contains the air bag. The pseudo load vector of the inverse shell unit can be quickly determined, and the efficiency of air bag state reconstruction is improved.

[0140] In an exemplary embodiment, the step of determining the pseudo-load vector of the reverse shell unit based on the number of reverse shell units and whether the reverse shell unit contains the first strain of the airbag can specifically include: if the number of reverse shell units is one and the reverse shell unit contains the first strain, then a pseudo-load vector is obtained based on the first strain; if the number of reverse shell units is one and the reverse shell unit does not contain the first strain, then a pseudo-load vector is obtained based on the strain estimate; the weight of the reverse shell unit meets a first condition; if the number of reverse shell units is more than one and the reverse shell unit contains the first strain, then a pseudo-load vector is obtained based on the first strain; the weight of the reverse shell unit meets a first condition; if the number of reverse shell units is more than one and the reverse shell unit does not contain the first strain, then a pseudo-load vector is obtained based on the strain estimate; the weight of the reverse shell unit meets a second condition.

[0141] The first condition can be used to calculate the pseudo-load vector. weight The second condition can be used to calculate the pseudo-load vector. The weight is part of Another part is much smaller than 1, for example, .

[0142] Optionally, pseudo-load vector The calculation may specifically include the following steps:

[0143] Step S501, when At that time, in the cube There are no inverse shell elements in the middle;

[0144] Step S502, when At that time, if the reverse shell unit contains High-precision strain measurement value and Then, based on the high-precision strain measurement values, the pseudo-load vector is calculated, and the element's The calculation is as follows:

[0145] .

[0146] If there are no high-precision strain measurements within the inverse shell element, then the pseudo-load vector is calculated based on the strain estimate, and the strain estimate is used. and As a strain measurement value, the unit's The calculation is as follows:

[0147] .

[0148] Among them, weight The transverse shear strain weights are adjusted based on the convergence of the calculations. In the interval are adjusted appropriately.

[0149] Step S503, when there are high-precision strain measurement values and for the inverse shell element, a pseudo load vector is calculated according to the high-precision strain measurement values, and the of the element is calculated as follows:

[0150] .

[0151] wherein the weight is adjusted appropriately according to the convergence of the calculation of the transverse shear strain weight in the interval .

[0152] For the inverse shell elements without high-precision strain measurement values, let the number thereof be , of which are determined by a random algorithm, a pseudo load vector is calculated according to the strain estimate values, and the estimate values and are used as the strain measurement values, and the is calculated as follows:

[0153] .

[0154] wherein the weight is adjusted appropriately according to the convergence of the calculation of the transverse shear strain weight in the interval .

[0155] For the remaining inverse shell elements, a pseudo load vector is also calculated according to the strain estimate values, and the estimate values and are used as the interpolated strain values, and the is calculated as follows:

[0156] .

[0157] wherein the weight is adjusted appropriately according to the convergence of the calculation of the transverse shear strain weight in the interval .

[0158] In this embodiment, the pseudo load vector is determined by considering different numbers of inverse shell units and whether the first strain is included in the inverse shell unit. The weight parameter of part of the inverse shell unit is set to 1, so that the constraint of the strain measurement data is enhanced. The weight parameter of part of the inverse shell unit is far less than 1, so that the airbag shape at the dense distribution of the inverse shell unit can be better reconstructed, and the complex deformation mode such as the wrinkle can be captured.

[0159] In order for those skilled in the art to have a deep understanding of the embodiments of the present application, the following will be described in combination with a specific example.

[0160] In view of the great influence of strain measurement accuracy on the reconstruction accuracy of the structural deformation state, the present application proposes a sensor arrangement and state reconstruction method for an underwater flexible gas storage airbag. A high-precision point strain sensor monitoring scheme and a low-precision distributed strain sensor monitoring scheme are designed. A data fusion algorithm is used to improve the estimation accuracy of the strain observation points for state reconstruction. An octree algorithm is used for grouping and classifying inverse shell units. Different weight strategies are used to calculate the pseudo stiffness matrix and pseudo load vector of the inverse shell unit according to the different inverse shell unit density in the unit space and whether a high-precision point strain sensor is arranged, so as to improve the state reconstruction accuracy of the underwater flexible gas storage airbag.

[0161] In an exemplary embodiment, a membrane structure of an underwater flexible gas storage airbag is provided, as shown in Figure 2 The membrane structure is sequentially provided with an inner liner, a first distributed strain sensor layer, a load bearing layer, a second distributed strain sensor layer, and an outer protective layer from inside to outside of the airbag. The inner liner and the outer protective layer are made of polyvinyl chloride material, and the load bearing layer is made of polyester fiber material, as shown in Figure 3 The distributed strain sensor layer includes distributed strain sensors and a substrate. The substrate is made of polyimide material, and the distributed strain sensors form a quadrature two-dimensional sensing network and are encapsulated inside the substrate material. The inner liner is coated on the inner side of the first distributed strain sensor layer. The first distributed strain sensor layer and the second distributed strain sensor layer are respectively bonded on the inner side and the outer side of the load bearing layer. The outer protective layer is coated on the outer side of the second distributed strain sensor layer.

[0162] Referring to Figure 4 The first distributed strain sensors are arranged in the first distributed strain sensor layer, and the second distributed strain sensors are arranged in the second distributed strain sensor layer, as shown in Figure 5 The positions of the first distributed strain sensors in the airbag membrane structure are in one-to-one correspondence with the positions of the second distributed strain sensors in the airbag membrane structure.

[0163] Point strain sensors are installed on the inner and outer surfaces of the airbag, a first point strain sensor is pasted on the inner side of the inner liner in a surface pasting manner, and a second point strain sensor is pasted on the outer side of the outer protective layer in a surface pasting manner, as shown in Figure 5 The position of the first point strain sensor in the airbag membrane structure in planar projection corresponds to the position of the second point strain sensor in the airbag membrane structure in planar projection.

[0164] Both the distributed strain sensor and the point strain sensor can measure the normal strain and shear strain in two orthogonal directions at the installation position thereof, for real-time monitoring of the strain size.

[0165] The measurement accuracy of the point strain sensor is higher than that of the distributed strain sensor, but the number of points that can be monitored by the point strain sensor is less than that of the distributed strain sensor.

[0166] In an exemplary embodiment, a state reconstruction method of an underwater flexible gas storage airbag is provided, the underwater flexible gas storage airbag is manufactured according to the foregoing underwater flexible gas storage airbag membrane structure, strain monitoring is performed during service, and data fusion is performed according to the following method:

[0167] Step S610, constructing an airbag strain high-precision data set by the first point strain sensor constructing an airbag strain high-precision data set by the second point strain sensor constructing an airbag strain low-precision data set by the first distributed strain sensor constructing an airbag strain low-precision data set by the second distributed strain sensor .

[0168] Step S620, the low-precision bending strain and the high-precision bending strain of the airbag are calculated as follows:

[0169] .

[0170] .

[0171] Step S630, the low-precision membrane strain and the high-precision membrane strain of the airbag are calculated as follows:

[0172] .

[0173] .

[0174] Step S640, using the multi-fidelity data fusion method of deep neural network, according to the high-precision data set and the low-precision data set , the bending strain of the airbag considering the accuracy and spatial resolution is constructed , the specific steps are as follows:

[0175] Step S641, according to the low-precision data set , the deep neural network (DNN) and its parameters are trained to obtain the estimated value of the low-precision bending strain .

[0176] Step S642, define the transfer learning model of the high-precision data set .

[0177] .

[0178] .

[0179] .

[0180] wherein, is the polynomial order, is the transfer learning parameter, is a Gaussian process with mean 0, and the covariance function uses a Gaussian kernel function:

[0181] .

[0182] wherein, and are the hyperparameters of the Gaussian kernel.

[0183] Step S643, define the optimization problem:

[0184] .

[0185] By linear regression algorithm to minimize the residual function , the estimated value of the transfer learning parameter , and the high-precision residual data set .

[0186] wherein, is the number of data points of the high-precision data set.

[0187] Step S644, train the Bayesian neural network (BNN) and its parameters through the high-precision residual data set , to obtain the estimated value of the residual .

[0188] Step S645, the estimated value of the DNN model according to the low-precision data set , the estimated value of the migration learning parameter , the estimated value of the residual error , the bending strain estimate value at the given point . .

[0189] Step S650, the same as the data fusion process steps S641 to S645 in step S640, the film strain of the air bag considering the accuracy and spatial resolution is constructed according to the high-precision data set and the low-precision data set . .

[0190] In an exemplary embodiment, a method for reconstructing the state of an underwater flexible gas storage air bag is provided, which reconstructs the deformation state of the underwater flexible gas storage air bag by using the inverse finite element method according to the film strain and the bending strain obtained after data fusion. Since the thickness of the air bag is much smaller than other dimensions, the transverse shear strain is negligible. The method specifically includes the following steps:

[0191] Step S701, the underwater flexible gas storage air bag is meshed to obtain a four-node inverse shell element, each inverse shell element node has three translation coordinates and three rotation coordinates ;

[0192] Step S702, the node displacement vector of the four-node inverse shell element after deformation is denoted as , then the strain of the four-node inverse shell element is as follows:

[0193] .

[0194] wherein, , ( ), , and is the strain displacement relationship matrix of the element, ( ), and:

[0195] .

[0196] .

[0197] .

[0198] Reference Figure 6 and Figure 7Using bilinear isoparametric functions, then:

[0199] .

[0200] .

[0201] .

[0202] .

[0203] in ( () are the standardized nodal coordinates of the isoparametric element. ; as well as ( , ), , .

[0204] Step S703: Considering membrane strain, bending strain, and transverse shear strain, construct the functional of the four-node inverse shell element using a weighted least squares function:

[0205] .

[0206] in, , and It is weight; , and It is either a strain measurement or a strain estimate given by the data fusion method described above. It is difficult to measure, and therefore there is no estimate.

[0207] Step S704, minimize the unknown variables functionals The unit equations are obtained as follows:

[0208] .

[0209] .

[0210] .

[0211] Step S705, based on the pseudo-stiffness matrix of the inverse shell element and pseudo-load vector and unit transformation matrix By assembling and integrating all four-node inverse shell elements, the system equations in the global coordinate system are obtained:

[0212] .

[0213] .

[0214] wherein, is the total number of elements divided in step S701.

[0215] In step S706, boundary conditions are applied to the system equation to obtain the reduced equation , and the solution is obtained as .

[0216] In step S707, the initial coordinates of the nodes are denoted as , the displacement of the nodes is obtained as from the solution in step S706, and the coordinates of the nodes after deformation are ( ), , wherein n is the total number of nodes.

[0217] In an exemplary embodiment, in step S701, the grid division is performed so that the areas of the elements at the initial time are as equal as possible, and the ratio of the largest element area to the smallest element area is not more than 2.

[0218] In an exemplary embodiment, in step S704, the of the inverse shell element is calculated as follows:

[0219] In step S810, the coordinates of the nodes at the initial time or at the previous reconstruction time are denoted as ( ), and the smallest cube containing the underwater flexible gas storage bag is obtained therefrom. .

[0220] In step S820, the cube is recursively divided by using an octree, and the division level is , , wherein the ceiling function is used. The first division obtains 8 cubes in the first sub-layer, the second division obtains 64 cubes in the second sub-layer, and so on. The th division obtains cubes in the th sub-layer. .

[0221] In step S830, the number of cubes in the th sub-layer is counted. the number of inverse shell elements in the cube If the center point of an inverse shell element is in the cube , the inverse shell element is in the cube .

[0222] Step S840, according to the value of and whether there is a high-precision strain measurement value in the inverse shell element, the calculation is as follows:

[0223] Step S841, when , there is no inverse shell element in the cube .

[0224] Step S842, when , if there are high-precision strain measurement values and in the inverse shell element, the calculation of is as follows:

[0225] .

[0226] If there is no high-precision strain measurement value in the inverse shell element, the estimated value and are used as the strain measurement value, and the calculation of is as follows:

[0227] .

[0228] wherein the weight , according to the calculation convergence, the transverse shear strain weight is properly adjusted within the interval .

[0229] Step S843, when , for the inverse shell element with high-precision strain measurement values and , the calculation of is as follows:

[0230] .

[0231] wherein the weight , according to the calculation convergence, the transverse shear strain weight is properly adjusted within the interval .

[0232] For the inverse shell element without high-precision strain measurement value, let the number be , and an inverse shell element, the estimated value and as a strain measurement value, the is calculated as follows:

[0233] .

[0234] wherein the weight , according to the calculation convergence of the transverse shear strain weight is appropriately adjusted within the interval .

[0235] For the remaining inverse shell element, the estimated value and as an interpolated strain value, the is calculated as follows:

[0236] .

[0237] wherein the weight , according to the calculation convergence of the transverse shear strain weight is appropriately adjusted within the interval .

[0238] The underwater flexible gas storage air bag sensor arrangement and state reconstruction method designs the underwater flexible gas storage air bag membrane structure and its point type and distributed strain sensing monitoring scheme, improves the precision of strain data based on the high-precision strain and low-precision strain monitoring data fusion method of DNN and BNN, and applies it to inverse finite element to reconstruct the deformation state of the underwater flexible gas storage air bag. In addition, the octree algorithm is used for inverse shell element grouping and classification. Different inverse shell element density degrees in a unit space are combined with or without high-precision point strain sensors arranged, different weight strategies are used to calculate the pseudo stiffness matrix and pseudo load vector of the inverse shell element.

[0239] Because the arrangement of the sensors and the mesh division of the inverse shell element are not consistent, strain sensors can be arranged in each inverse shell element or can not be arranged in each inverse shell element. In the traditional inverse finite element method, the strain values obtained by direct measurement are used in a small part of the inverse shell elements, and the strain data obtained by interpolation are used in most of the inverse shell elements, and there is a deviation from the true strain data. Arrangement of fewer strain measuring points will result in lower accuracy of the reconstructed deformation field, and it is necessary to increase the measuring points to improve the accuracy and reduce the reconstruction error. However, high-precision point strain sensors are high in cost. In addition, in the traditional inverse finite element method, if no strain measuring point is arranged in the inverse shell element, all the weight parameters are much smaller than 1. The sensor arrangement and state reconstruction method for the underwater flexible gas storage air bag provided in the application, on the one hand, based on as few high-precision point strain sensors as possible, a large number of low-cost low-precision sensors are deployed, and the accuracy of the strain data is improved through the data fusion method based on DNN and BNN, and then the deformation state reconstruction accuracy is improved; on the other hand, the octree algorithm is used for grouping and classifying the inverse shell elements, different weight strategies are used according to the different density of the inverse shell elements in the unit space and combined with the arrangement of the high-precision point strain sensors, because the accuracy of the strain data is improved through the data fusion method, the weight parameter of part of the inverse shell elements is 1, so that the constraint of the strain measurement data is enhanced, and the weight parameter of part of the inverse shell elements is much smaller than 1, so that the air bag shape at the dense distribution of the inverse shell elements can be better reconstructed, and the complex deformation mode such as wrinkle can be captured.

[0240] In one embodiment, as shown in Figure 8 a gas bag state reconstruction method is provided, comprising the following steps:

[0241] Step S901, obtaining a first data set and a second data set of the gas bag; the accuracy of the first data set is higher than the accuracy of the second data set;

[0242] Step S902, determining a first strain of the gas bag according to the first data set, and determining a second strain of the gas bag according to the second data set; the first strain includes a first bending strain and a first film strain, and the second strain includes a second bending strain and a second film strain;

[0243] Step S903, determining a strain estimation value corresponding to a node displacement strain according to the first strain and the second strain; the strain estimation value includes a bending strain estimation value and a film strain estimation value;

[0244] Step S904, obtaining a node displacement strain corresponding to a node displacement of the gas bag; the node displacement of the gas bag is determined according to the inverse shell element obtained by mesh division of the gas bag;

[0245] Step S905: Based on the nodal displacement strain and the strain estimate corresponding to the nodal displacement strain, the functional model of the inverse shell element is obtained; the strain estimate is obtained by fusing the first strain and the second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain.

[0246] Step S906: Determine the pseudo stiffness matrix and pseudo load vector of the inverse shell element based on the functional model.

[0247] Step S907: Obtain the nodal displacement model of the airbag based on the pseudo stiffness matrix and pseudo load vector.

[0248] Step S908: Determine the target node displacement of the airbag based on the node displacement model, and reconstruct the state of the airbag according to the target node displacement.

[0249] In practice, the terminal can collect high-precision data of the airbag using high-precision sensors and low-precision data of the airbag using low-precision sensors. It determines the high-precision strain of the airbag based on the high-precision data and the low-precision strain based on the low-precision data. The high-precision and low-precision strains are then fused to obtain strain estimates for each node on the airbag. Transverse shear strain Negligible. The terminal can also set the displacement of each node on the airbag. Based on the nodal displacement strain corresponding to the nodal displacement of the airbag nodes Each node on the airbag is determined based on the inverse shell elements obtained by meshing the airbag. A functional is constructed using the nodal displacement, strain, and strain estimates:

[0250] .

[0251] The pseudo-stiffness matrix is ​​obtained by solving. and pseudo-load vector , where the pseudo-load vector The octree algorithm can be used to group and classify inverse shell elements. Based on the different densities of inverse shell elements within a unit space and the presence or absence of high-precision point strain sensors, different weighting strategies are used to calculate the inverse shell elements. Then, a nodal displacement model is constructed.

[0252] .

[0253] .

[0254] Solving for the nodes The displacement is Then the coordinates of the deformed nodes are ( Based on this, the state of the airbag is reconstructed.

[0255] The air bag state reconstruction method can improve the estimation accuracy of the strain observation point by collecting the first strain of the air bag through a high-precision sensor, collecting the second strain of the air bag through a low-precision sensor, fusing the first strain and the second strain to obtain a strain estimation value, and using the strain estimation value for air bag state reconstruction, thereby improving the state reconstruction accuracy of the underwater flexible gas storage air bag.

[0256] It should be understood that, although each step in the flowchart involved in each embodiment as described above is shown in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps. It can be understood that the steps in different embodiments can be freely combined as needed, and various non-contradictory schemes formed by the combination are within the scope of protection of the present application.

[0257] Based on the same inventive concept, the embodiments of the present application also provide an air bag state reconstruction device for implementing the air bag state reconstruction method described above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more air bag state reconstruction device embodiments provided below can refer to the limitations of the air bag state reconstruction method described above, which will not be repeated here.

[0258] In one exemplary embodiment, as shown in Figure 9 An air bag state reconstruction device is provided, comprising: a strain acquisition module 1001, a functional model module 1002, a matrix determination module 1003, a displacement model module 1004, and a state reconstruction module 1005, wherein:

[0259] The strain acquisition module 1001 is configured to acquire a node displacement strain corresponding to a node displacement of an air bag; the node displacement of the air bag is determined according to an inverse shell element obtained by meshing the air bag;

[0260] The functional model module 1002 is configured to obtain a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the air bag, and the accuracy of the first strain is higher than that of the second strain;

[0261] The matrix determination module 1003 is configured to determine a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model.

[0262] The displacement model module 1004 is configured to obtain a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector.

[0263] The state reconstruction module 1005 is configured to determine a target node displacement of the airbag based on the node displacement model, and reconstruct a state of the airbag according to the target node displacement.

[0264] In an exemplary embodiment, the airbag state reconstruction device further comprises:

[0265] The data acquisition module is configured to acquire a first data set and a second data set of the airbag, wherein the first data set has a higher accuracy than the second data set.

[0266] The strain determination module is configured to determine the first strain of the airbag according to the first data set, and determine the second strain of the airbag according to the second data set, wherein the first strain comprises a first bending strain and a first membrane strain, and the second strain comprises a second bending strain and a second membrane strain.

[0267] The strain estimation module is configured to determine the strain estimation value corresponding to the node displacement strain according to the first strain and the second strain, wherein the strain estimation value comprises a bending strain estimation value and a membrane strain estimation value.

[0268] In an exemplary embodiment, the data acquisition module is further configured to acquire the first data set collected by the first point strain sensor and the second point strain sensor, and acquire the second data set collected by the first distributed strain sensor and the second distributed strain sensor.

[0269] In an exemplary embodiment, the strain estimation module is further configured to train the second strain to obtain a trained deep neural network, determine a first strain estimation value corresponding to the first strain according to the deep neural network, determine a transfer learning model of the first strain according to the first strain estimation value, input a second strain estimation value of an airbag node determined according to the deep neural network into the transfer learning model to obtain a third strain estimation value of the airbag node, and take the third strain estimation value as the strain estimation value corresponding to the node displacement strain.

[0270] In an example embodiment, the matrix determining module 1003 is further configured to determine a target cube containing the airbag; recursively divide the target cube using an octree to obtain a plurality of sub-cubes of the target cube; the plurality of sub-cubes correspond to a plurality of sub-layers, and the plurality of sub-layers contain a target sub-layer; determine a number of inverse shell units in each sub-cube of the target sub-layer; and determine the pseudo load vector of the inverse shell units according to the number of the inverse shell units and whether the inverse shell units contain a first strain of the airbag.

[0271] In an example embodiment, the matrix determining module 1003 is further configured to, if the number of the inverse shell units is one and the inverse shell units contain the first strain, determine the pseudo load vector according to the first strain; if the number of the inverse shell units is one and the inverse shell units do not contain the first strain, determine the pseudo load vector according to the strain estimation value; the weight of the inverse shell units meets a first condition; if the number of the inverse shell units is more than one and the inverse shell units contain the first strain, determine the pseudo load vector according to the first strain; the weight of the inverse shell units meets the first condition; if the number of the inverse shell units is more than one and the inverse shell units do not contain the first strain, determine the pseudo load vector according to the strain estimation value; and the weight of the inverse shell units meets a second condition.

[0272] The modules in the airbag state reconstruction apparatus can be implemented by software, hardware, or a combination thereof. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in a computer device in software form, so as to be invoked and executed by a processor to perform operations corresponding to the modules.

[0273] In an example embodiment, a computer device is provided, which can be a terminal. An internal structure diagram of the computer device can be as shown in FIG. 8. Figure 10The computer device shown in the figure includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, mobile cellular network, near field communication (NFC) or other technologies. The computer program is executed by the processor to realize an airbag state reconstruction method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0274] Those skilled in the art can understand that, Figure 10 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0275] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to realize the steps in the above method embodiments.

[0276] In one exemplary embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to realize the steps in the above method embodiments.

[0277] In one exemplary embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to realize the steps in the above method embodiments.

[0278] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0279] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments of each method. In the embodiments provided in the present application, any reference to memory, database or other medium can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0280] Any technical features in the above embodiments can be combined, and for the sake of brevity, not all possible combinations are described above, however, any combination of these technical features is deemed to be within the scope of the present application.

[0281] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. An airbag state reconstruction method characterized by, The airbag comprises an inner liner, a first distributed strain sensing layer, a bearing layer, a second distributed strain sensing layer and an outer protective layer; the first distributed strain sensing layer is provided with a first distributed strain sensor, the second distributed strain sensing layer is provided with a second distributed strain sensor, the arrangement position of the first distributed strain sensor corresponds to the arrangement position of the second distributed strain sensor; the inner liner is pasted with a first point strain sensor, the outer protective layer is pasted with a second point strain sensor, the arrangement position of the first point strain sensor corresponds to the arrangement position of the second point strain sensor; the method comprises: obtaining a node displacement strain corresponding to an airbag node displacement; the airbag node displacement is determined according to an inverse shell element obtained by meshing the airbag; obtaining a functional model of the inverse shell element according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain; determining a pseudo stiffness matrix and a pseudo load vector of the inverse shell element based on the functional model, which comprises: determining a target cube containing the airbag; recursively dividing the target cube by using an octree to obtain a plurality of sub-cubes of the target cube; the plurality of sub-cubes correspond to a plurality of sub-layers, and the plurality of sub-layers contain a target sub-layer; determining the number of inverse shell elements in each sub-cube of the target sub-layer; determining the pseudo load vector of the inverse shell element according to the number of inverse shell elements and whether the inverse shell element contains the first strain of the airbag; obtaining a node displacement model of the airbag according to the pseudo stiffness matrix and the pseudo load vector; determining a target node displacement of the airbag based on the node displacement model, and reconstructing the state of the airbag according to the target node displacement.

2. The method of claim 1, wherein, The method further comprises: obtaining a first data set and a second data set of the airbag; the accuracy of the first data set is higher than that of the second data set; determining the first strain of the airbag according to the first data set, and determining the second strain of the airbag according to the second data set; the first strain comprises a first bending strain and a first membrane strain, and the second strain comprises a second bending strain and a second membrane strain; determining the strain estimation value corresponding to the node displacement strain according to the first strain and the second strain; the strain estimation value comprises a bending strain estimation value and a membrane strain estimation value.

3. The method of claim 2, wherein, The method further comprises: obtaining the first data set collected by the first point strain sensor and the second point strain sensor, and obtaining the second data set collected by the first distributed strain sensor and the second distributed strain sensor.

4. The method of claim 2, wherein, The method further comprises: training the second strain to obtain a trained deep neural network; determining a first strain estimation value corresponding to the first strain according to the deep neural network; determining a transfer learning model of the first strain according to the first strain estimation value; inputting a second strain estimation value of the airbag node determined according to the deep neural network into the transfer learning model to obtain a third strain estimation value of the airbag node; taking the third strain estimation value as the strain estimation value corresponding to the node displacement strain.

5. The method of claim 1, wherein, The determination of the pseudo load vector of the inverse shell unit according to the number of the inverse shell units and whether the inverse shell units contain the first strain of the airbag comprises: if the number of the inverse shell units is one and the inverse shell units contain the first strain, the pseudo load vector is obtained according to the first strain; if the number of the inverse shell units is one and the inverse shell units do not contain the first strain, the pseudo load vector is obtained according to the strain estimation value; the weight of the inverse shell unit meets a first condition; if the number of the inverse shell units is more than one and the inverse shell units contain the first strain, the pseudo load vector is obtained according to the first strain; the weight of the inverse shell unit meets the first condition; if the number of the inverse shell units is more than one and the inverse shell units do not contain the first strain, the pseudo load vector is obtained according to the strain estimation value; the weight of the inverse shell unit meets a second condition.

6. An airbag status reconstruction apparatus characterized by comprising: The airbag comprises an inner liner, a first distributed strain sensing layer, a bearing layer, a second distributed strain sensing layer and an outer protective layer; the first distributed strain sensing layer is provided with a first distributed strain sensor, the second distributed strain sensing layer is provided with a second distributed strain sensor, and the arrangement position of the first distributed strain sensor corresponds to the arrangement position of the second distributed strain sensor; the inner liner is pasted with a first point strain sensor, the outer protective layer is pasted with a second point strain sensor, and the arrangement position of the first point strain sensor corresponds to the arrangement position of the second point strain sensor; and the device comprises: a strain acquisition module configured to acquire a node displacement strain corresponding to a node displacement of an airbag; the node displacement of the airbag is determined according to an inverse shell unit obtained by grid division of the airbag; a functional model module configured to obtain a functional model of the inverse shell unit according to the node displacement strain and a strain estimation value corresponding to the node displacement strain; the strain estimation value is obtained by fusing a first strain and a second strain of the airbag, and the accuracy of the first strain is higher than that of the second strain; a matrix determination module configured to determine a pseudo stiffness matrix and a pseudo load vector of the inverse shell unit based on the functional model. The matrix determination module is further configured to: determine a target cube containing the airbag; recursively partition the target cube using an octree to obtain multiple sub-cubes of the target cube; the multiple sub-cubes correspond to multiple sub-layers, and the multiple sub-layers contain the target sub-layer; determine the number of inverse shell units in each sub-cube of the target sub-layer; and determine the pseudo-load vector of the inverse shell unit based on the number of inverse shell units and whether the inverse shell unit contains the first strain of the airbag. The displacement model module is used to obtain the nodal displacement model of the airbag based on the pseudo stiffness matrix and the pseudo load vector. The state reconstruction module is used to determine the target node displacement of the airbag based on the node displacement model, and to reconstruct the state of the airbag according to the target node displacement.

7. The apparatus of claim 6, wherein, The device further includes: The data acquisition module is used to acquire a first dataset and a second dataset of the airbag; the accuracy of the first dataset is higher than that of the second dataset. The strain determination module is used to determine the first strain of the airbag based on the first dataset and to determine the second strain of the airbag based on the second dataset; the first strain includes a first bending strain and a first membrane strain, and the second strain includes a second bending strain and a second membrane strain; The strain estimation module is used to determine the strain estimate corresponding to the nodal displacement strain based on the first strain and the second strain; the strain estimate includes a bending strain estimate and a membrane strain estimate.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images