Aircraft structure finite element calculation result data noise reduction method

Abstract

The invention belongs to the technical field of airplane structure strength digital twinborn model construction, and particularly relates to an airplane structure finite element calculation result data noise reduction method, which comprises the following steps: step 1, preliminarily identifying deformity units including triangular units and high slenderness ratio units; identifying the triangular units based on the continuity of the node numbers; identifying a high slenderness ratio unit based on the side length; 2, defining a stress concentration area, and removing units in the stress concentration area from the deformed units; defining a stress concentration region including a hole edge, a fillet, a boundary constraint position and a region where a concentrated load applying position is located; 3, calculating the gradient of the deformity units, and removing units in a real physical gradient field from the deformity units; step 4, carrying out noise reduction processing on the calculated value of the malformation unit; and smoothing the calculated value of the malformed unit, or removing the calculated value of the malformed unit.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aircraft structure strength digital twin model construction, and particularly relates to an aircraft structure finite element calculation result data denoising method. BACKGROUND

[0002] Training a machine learning model using finite element calculation result data is an important link in constructing an aircraft structure strength digital twin model such as a wing lower panel and a wing skin. Constructing a high-precision aircraft structure strength digital twin model fundamentally relies on high-fidelity finite element calculation result data.

[0003] Due to the complexity of the aircraft structure, during the finite element meshing process, some low-quality elements such as triangular elements and high-aspect-ratio elements will inevitably be introduced. These elements will produce unreasonable stress / strain abnormal values in finite element calculation. Such abnormal values belong to finite element calculation artifacts, not real stress concentration phenomena. That is, in the finite element calculation result data of the aircraft structure, there are non-physical numerical noises, and abnormal values are mixed with real stress concentration values. If directly used for training of a machine learning model, it will seriously mislead the machine learning algorithm, and the aircraft structure strength digital twin model constructed therefrom will have poor precision and will have serious prediction distortion problems.

[0004] Therefore, before being applied to training of a machine learning model, the finite element calculation result data of the aircraft structure needs to be denoised to remove abnormal values therefrom. Currently, a threshold method is usually used to denoise the finite element calculation result data of the aircraft structure, and data greater than a preset threshold is directly removed as an abnormal value. This method is simple and easy to implement, but it also removes real stress concentration values when removing abnormal values, and cannot effectively ensure high fidelity of the finite element calculation result data of the aircraft structure, and cannot meet the need of constructing a high-precision aircraft structure strength digital twin model.

[0005] The present application is proposed in view of the above technical defects. SUMMARY

[0006] The purpose of the present application is to provide an aircraft structure finite element calculation result data denoising method to effectively remove abnormal values from the finite element calculation result data of the aircraft structure and meet the need of constructing a high-precision aircraft structure strength digital twin model.

[0007] The technical solution of the present application is as follows:

[0008] An aircraft structure finite element calculation result data denoising method comprises the following steps:

[0009] Step 1: preliminary identification of abnormal elements, including triangular elements and high-aspect-ratio elements;

[0010] Identify triangle element based on node number continuity;

[0011] Identify high aspect ratio element based on edge length;

[0012] Step two, define stress concentration area, remove element in stress concentration area from distorted element;

[0013] Define stress concentration area including hole edge, round corner, boundary constraint position, concentrated load application position;

[0014] Step three, calculate gradient of distorted element, remove element in real physical gradient field from distorted element;

[0015] Step four, denoise the calculated value of distorted element;

[0016] Smooth the calculated value of distorted element, or remove the calculated value of distorted element.

[0017] Optionally, in the above aircraft structure finite element calculation result data denoising method, in step one, the triangle element is identified based on node number continuity, specifically:

[0018] The fourth node number is 0, or -1, or the same as the third node number, or the same as the fourth node number of the previous element, and the element is identified as a triangle element.

[0019] Optionally, in the above aircraft structure finite element calculation result data denoising method, in step one, the high aspect ratio element is identified based on edge length, specifically:

[0020] Calculate the aspect ratio of the element :

[0021] ;

[0022] Or,

[0023] ;

[0024] Wherein,

[0025] The four edge lengths of the element, The maximum value of the four edge lengths of the element, The minimum value of the four edge lengths of the element;

[0026] The area of the element is calculated by dividing the element into two triangles and summing the areas;

[0027] The element with is identified as a high aspect ratio element, wherein, The threshold value for the high slenderness ratio unit is 10-20.

[0028] Optionally, in the aircraft structure finite element calculation result data denoising method, the step three comprises:

[0029] The gradient of the abnormal unit and the adjacent unit is calculated.

[0030] ;

[0031] The gradient of the abnormal unit and the adjacent unit is calculated.

[0032] The gradient of the abnormal unit and the adjacent unit is calculated.

[0033] The calculated value of the abnormal unit is calculated.

[0034] The calculated value of the jth adjacent unit is calculated.

[0035] The Euclidean distance between the abnormal unit and the jth adjacent unit is calculated.

[0036] The adjacent unit of the abnormal unit is K-level, wherein the first-level adjacent unit is the unit sharing the node with the abnormal unit, the K-level adjacent unit is the unit sharing the node with the K-1-level adjacent unit, and K is 3 or 4.

[0037] The gradient of the abnormal unit is calculated.

[0038] ;

[0039] Alternatively,

[0040] ;

[0041] The gradient of the abnormal unit is calculated.

[0042] The gradient of the abnormal unit is calculated.

[0043] The number of adjacent units is calculated. The abnormal unit is identified as being in the real physical gradient field, and then removed from the abnormal unit, wherein:

[0044] The abnormal unit is identified as being in the real physical gradient field, and then removed from the abnormal unit, wherein:

[0045] The mean value of the calculated values of the adjacent units is calculated.

[0046] The setting coefficient is 3-5.

[0047] ​ The standard deviation of the calculated values ​​for the neighboring cells;

[0048] is the standard deviation of the gradient between the malformed unit and the i-th neighboring unit;

[0049] The set coefficient has a value range of 1.5 to 2.

[0050] Optionally, the above-mentioned method for denoising the finite element calculation results of aircraft structures also includes:

[0051] Step 5: Denoise the extreme points of the calculated values ​​of the elements in the non-key areas of interest. The non-key areas of interest are deformed elements, stress concentration areas, and areas outside the real physical gradient field.

[0052] Optionally, in the above-mentioned denoising method for finite element calculation results of aircraft structures, in step five, if there are extreme points in the element calculation values ​​within the non-key focus area... Then, noise reduction is performed on the extreme points, specifically as follows:

[0053] Replace the calculated value of the extreme point with Alternatively, it can be replaced with a distance-weighted average of the calculated values ​​from neighboring cells to ensure the continuity of changes in cell calculated values, or it can be deleted.

[0054] It is an extreme point;

[0055] This is the average value calculated from the neighboring units, where the neighboring units are of level L, and L is 3 or 4.

[0056] The threshold value is set to a range of 2.5 to 3.5.

[0057] This represents the standard deviation of the calculated values ​​for the neighboring cells.

[0058] Optionally, in the above-mentioned method for denoising the finite element calculation results of aircraft structures, step five defines the non-key areas of interest as local areas far from the connection points and force transmission paths.

[0059] This application has at least the following beneficial technical effects:

[0060] This paper presents a method for denoising the finite element calculation results of aircraft structures. It integrates multiple methods such as geometric feature recognition and physical gradient field analysis to identify low-quality deformed elements and real stress concentration elements. This method can remove outliers while retaining real stress concentration values, effectively ensuring the high fidelity of the finite element calculation results of aircraft structures and meeting the needs of constructing high-precision digital twin models of aircraft structural strength. Attached Figure Description

[0061] Figure 1 This is a flowchart of the method for denoising the finite element calculation results of aircraft structures provided in the embodiments of this application.

[0062] To better illustrate this embodiment, some content in the accompanying drawings may be omitted. They are for illustrative purposes only and should not be construed as limiting the scope of this application. Detailed Implementation

[0063] To make the technical solution and advantages of this application clearer, the technical solution of this application will be described in a clearer and more complete manner below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some embodiments of this application, and are only used to explain this application, not to limit this application. It should be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, and other related parts can be referred to the general design.

[0064] Furthermore, unless otherwise defined, the technical or scientific terms used in this application description shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The word "comprising" as used in this application description indicates that the concept preceding the word encompasses the concepts listed following the word and their equivalents, without excluding other related concepts.

[0065] A method for denoising aircraft structural finite element calculation results, such as Figure 1 As shown.

[0066] Step 1: Preliminary identification of abnormal units.

[0067] The aim is to quickly identify low-quality deformed elements introduced by mesh generation from all elements, including triangular elements and high aspect ratio elements.

[0068] Read the finite element calculation results data, parse and store the information of each element, including the element ID, type, node list, node coordinates, center point coordinates, and calculated values ​​(stress or strain values).

[0069] For a list of elements arranged in order of element ID, the node numbers of normal quadrilateral elements are usually unique and incremental. In finite element analysis results, to maintain data format consistency, triangular elements are typically configured with four nodes, and their fourth node number is filled with 0 or -1, or taken from the third node number, or the fourth node number of the previous element. This leads to anomalies in the node list. Therefore, triangular elements can be identified based on the continuity of node numbers, specifically:

[0070] The element whose fourth node number is 0, -1, the same as the third node number, or the same as the fourth node number of the previous element is identified as a triangular element.

[0071] After identifying triangular elements from all elements, height-to-slenderness ratio elements can be identified based on their side lengths, specifically:

[0072] Calculation unit slenderness ratio :

[0073] ;

[0074] or,

[0075] ;

[0076] in,

[0077] The four side lengths of the unit, It is the maximum value among the four side lengths of the unit. It is the minimum value among the four side lengths of the unit;

[0078] To find the area of ​​a unit, the unit can be divided into two triangles, and the areas of each triangle can be calculated separately and then summed.

[0079] Will The elements are identified as high aspect ratio elements, where, The threshold for identifying high aspect ratio units is typically set to 10 to 20.

[0080] Step 2: Define the stress concentration region and remove the elements within the stress concentration region from the malformed elements.

[0081] The aim is to remove real stress concentration elements from malformed elements without denoising the calculated values ​​of stress concentration elements.

[0082] Based on the geometric characteristics and mechanical knowledge of the aircraft structure, stress concentration regions are defined, which are areas where stress concentration is expected to occur. These typically include areas such as hole edges, fillets, boundary constraint locations, and areas where concentrated loads are applied. Elements within these regions are removed from the deformed elements, i.e., the corresponding element numbers are removed.

[0083] Stress concentration regions can be described as:

[0084] All regions where the location coordinates are greater than Y;

[0085] A circular region with center O and radius R.

[0086] Step 3: Calculate the gradient of the malformed cells and remove the cells that are in the real physical gradient field from the malformed cells.

[0087] The aim is to remove cells that are located within the real physical gradient field from the malformed cells, without performing noise reduction processing on the calculated values ​​of cells located within the real physical gradient field.

[0088] Calculate the gradient between the malformed cell and its neighboring cells:

[0089] ;

[0090] in,

[0091] The gradient between the deformed unit and the j-th neighboring unit;

[0092] The calculated value for the malformed unit;

[0093] The calculated value for the j-th neighborhood cell;

[0094] Let be the Euclidean distance between the malformed cell and its j-th neighboring cell.

[0095] The neighboring units of the malformed unit are of level K. The first-level neighboring units are units that share nodes with the malformed unit, and the K-level neighboring units are units that share nodes with the (K-1)-level neighboring units. K can usually be 3 or 4, which is enough to cover a sufficient range of units.

[0096] Calculate the gradient of the malformed unit:

[0097] ;

[0098] or,

[0099] ;

[0100] in,

[0101] The gradient of the malformed unit;

[0102] This represents the number of neighboring units.

[0103] Will ,or The malformed units are identified as units located within the real physical gradient field and are thus removed from the malformed units, wherein:

[0104] The mean value calculated for the neighboring cells;

[0105] The set coefficient has a value range of 3 to 5;

[0106] The standard deviation of the calculated values ​​for the neighboring cells;

[0107] is the standard deviation of the gradient between the malformed unit and the i-th neighboring unit;

[0108] The set coefficient has a value range of 1.5 to 2.

[0109] Ultimately, only those units with abnormally high / low calculated values ​​and highly discrete gradients with neighboring units are retained in the malformed units. Units with relatively high calculated values ​​but small and stable gradients with neighboring units, showing good decay patterns, are identified as units within the real physical gradient field and are removed.

[0110] Step 4: Perform noise reduction processing on the calculated values ​​of the malformed cells.

[0111] Smoothing or removing the calculated values ​​of malformed units, as well as deleting a small amount of low-quality data, can benefit the training of machine learning models.

[0112] Step 5: Define non-key areas of interest and perform noise reduction on the extreme points of the unit calculation values ​​within the non-key areas of interest. The non-key areas of interest are local areas far away from connection points, force transmission paths, and other non-key areas of interest.

[0113] Extreme values ​​in unit computations, occurring in areas of non-critical interest, can negatively impact the training of machine learning models.

[0114] If the extreme points of the unit calculation values ​​in the non-key area are... If the extreme points are found to be denoised, then noise reduction is performed on them; otherwise, noise reduction is not required. The noise reduction strategy is as follows:

[0115] Replace the calculated value of the extreme point with Alternatively, it can be replaced with a distance-weighted average of the calculated values ​​from neighboring cells to ensure the continuity of changes in cell calculated values, or it can be deleted.

[0116] It is an extreme point;

[0117] This is the average value calculated from the neighboring units, where the neighboring units are of level L, and L can typically be 3 or 4.

[0118] The threshold value is set to a range of 2.5 to 3.5.

[0119] This represents the standard deviation of the calculated values ​​for the neighboring cells.

[0120] The denoising method for finite element calculation results of aircraft structures disclosed in the above embodiments constructs a three-level collaborative denoising process. First, to address the quality issues of the mesh generation itself, malformed elements are initially identified based on geometric features. Then, by defining stress concentration regions and judging physical gradients, elements with reasonable stress concentrations among the malformed elements are retained without denoising the calculated values. Finally, the extreme points of the calculated values ​​of elements in non-key areas are statistically analyzed and denoised. This achieves a layer-by-layer in-depth analysis from "geometric" to "physical" to "statistical," ensuring high accuracy in the denoising process.

[0121] The denoising method for finite element calculation results of aircraft structures disclosed in the above embodiments integrates multiple means such as geometric feature recognition and physical gradient field analysis to identify low-quality deformed elements and real stress concentration elements. This allows for the removal of outliers while retaining real stress concentration values, effectively ensuring high fidelity of finite element calculation results for aircraft structures and meeting the needs of constructing high-precision digital twin models of aircraft structural strength.

[0122] In one embodiment, based on the denoising method for the finite element calculation results of aircraft structures disclosed in the above embodiments, the stress data of the finite element calculation results of the lower wing panel is denoised. The stress concentration areas are defined at the boundary with the wing and at the corner of the beam rib. The triangular elements and high slenderness ratio elements are successfully identified. The stress values ​​of the triangular elements and high slenderness ratio elements are denoised, and the stress values ​​of the stress concentration elements are basically preserved, thus ensuring the high fidelity of the stress data of the finite element calculation results of the lower wing panel.

[0123] In another embodiment, based on the denoising method for the finite element calculation results of aircraft structures disclosed in the above embodiments, the stress data of the finite element calculation results of the wing skin are denoised. In this method, the transition area of ​​the skin boundary and the key connection is defined as the stress concentration area. Finally, the triangular elements and high slenderness ratio elements are successfully identified. The stress values ​​of the triangular elements and high slenderness ratio elements are denoised, and the stress values ​​of the stress concentration elements are basically preserved, thus ensuring the high fidelity of the stress data of the finite element calculation results of the wing skin.

[0124] The technical solution of this application has been described in conjunction with the preferred embodiments shown in the accompanying drawings. Those skilled in the art should understand that the scope of protection of this application is obviously not limited to these specific embodiments. Without departing from the principles of this application, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of this application.

Claims

1. A method for denoising aircraft structural finite element calculation results, characterized in that, include: Step 1: Preliminary identification of deformed units, including triangular units and units with a height-to-length ratio; Continuous identification of triangular units based on node numbering; High-slenderness ratio unit based on side length identification; Step 2: Define the stress concentration region and remove the elements within the stress concentration region from the malformed elements; The stress concentration region is defined to include the area where the hole edge, fillet, boundary constraint location, and concentrated load application location are located. Step 3: Calculate the gradient of the malformed cells and remove the cells that are within the real physical gradient field from the malformed cells; Step 4: Perform noise reduction processing on the calculated values ​​of the malformed cells; Smooth the calculated values ​​of malformed cells, or remove the calculated values ​​of malformed cells.

2. The method for denoising aircraft structural finite element calculation results according to claim 1, characterized in that, In step one, the triangular unit is identified based on the continuity of node numbers, specifically as follows: The element whose fourth node number is 0, -1, the same as the third node number, or the same as the fourth node number of the previous element is identified as a triangular element.

3. The method for denoising aircraft structural finite element calculation results according to claim 2, characterized in that, In step one, the height-to-slenderness ratio unit is identified based on the side length, specifically as follows: Calculation unit slenderness ratio : ； or, ； in, The four side lengths of the unit, It is the maximum value among the four side lengths of the unit. It is the minimum value among the four side lengths of the unit; The area of ​​a unit is obtained by dividing the unit into two triangles, calculating the areas of each triangle separately, and then summing the results. Will The elements are identified as high aspect ratio elements, where, The threshold value for identifying high aspect ratio units is 10 to 20.

4. The method for denoising aircraft structure finite element calculation results according to claim 3, characterized in that, Step three includes: Calculate the gradient between the malformed cell and its neighboring cells: ； in, The gradient between the deformed unit and the j-th neighboring unit; The calculated value for the malformed unit; The calculated value for the j-th neighborhood cell; Let be the Euclidean distance between the deformed cell and the j-th neighboring cell; The neighboring units of the malformed unit are of level K, where the first-level neighboring units are units that share nodes with the malformed unit, and the K-level neighboring units are units that share nodes with the (K-1)-level neighboring units. K is 3 or 4. Calculate the gradient of the malformed unit: ； or, ； in, The gradient of the malformed unit; The number of neighboring units; Will ,or The malformed units are identified as units located within the real physical gradient field and are thus removed from the malformed units, wherein: The mean value calculated for the neighboring cells; The set coefficient has a value range of 3 to 5; The standard deviation of the calculated values ​​for the neighboring cells; is the standard deviation of the gradient between the malformed unit and the i-th neighboring unit; The set coefficient has a value range of 1.5 to 2.

5. The method for denoising aircraft structural finite element calculation results according to claim 4, characterized in that, Also includes: Step 5: Define non-key interest regions and perform noise reduction processing on the extreme points of the calculated values ​​of cells within the non-key interest regions.

6. The method for denoising aircraft structural finite element calculation results according to claim 5, characterized in that, In step five, if the extreme points of the calculated values ​​of the cells within the non-key area are... Then, noise reduction is performed on the extreme points, specifically as follows: Replace the calculated value of the extreme point with Alternatively, it can be replaced with a distance-weighted average of the calculated values ​​from neighboring cells to ensure the continuity of changes in cell calculated values, or it can be deleted. It is an extreme point; This is the average value calculated from the neighboring units, where the neighboring units are of level L, and L is 3 or 4. The threshold value is set to a range of 2.5 to 3.

5. This represents the standard deviation of the calculated values ​​for the neighboring cells.

7. The method for denoising aircraft structure finite element calculation results according to claim 6, characterized in that, In step five, the non-key focus area is defined as a local area far away from the connection point and the force transmission path.

Images