Shaving board mechanical property prediction analysis method based on simulation and deep learning

Abstract

The invention discloses a shaving board mechanical property prediction analysis method based on simulation and deep learning, and the method comprises the steps: building an equivalent single-layer board mechanical model of an anisotropic small-layer splicing structure of a unit-thickness shaving board through finite element simulation and theory, and obtaining a shaving board geometrical morphology and stress-strain distribution data set; and then an equivalent laminated plate mechanical model of the shaving board vertical direction stacking structure is constructed, and a shaving board structure parameter and mechanical property data set is obtained. The artificial intelligence model is trained through the data set, and based on a wood shaving piece real-time paving structure obtained in the shaving board machining process, rapid intelligent prediction and quality control of the mechanical property of the shaving board are achieved through the trained artificial intelligence model. The scheme has the advantages of simplicity in operation, high universality, high detection efficiency, high processing precision, low technical cost, high quality controllability and the like; the method has important significance in the fields of structure-activity relationship analysis, mechanical property prediction, production online quality control and the like of the particle board small sheet layer pavement structure.

Description

Technical Field

[0001] The present invention relates to the technical field of particleboard mechanical property prediction and analysis, and in particular to a particleboard mechanical property prediction and analysis method based on simulation and deep learning. Background Art

[0002] Particleboard is a board made from wood, bamboo and other shavings. It has a wide range of applications in the construction, furniture, packaging and transportation industries due to its good mechanical properties, durability, environmental friendliness and low cost. Clarifying the mechanism of the mechanical properties of particleboard is an important fundamental issue for improving the performance of the board and conducting non-destructive testing of the board. However, the random structure of the board formed by the splicing and stacking of a large number of anisotropic small layers is complex, and the structure-activity relationship between the structure and the mechanical properties is still unclear.

[0003] At present, the preparation and optimization of various anisotropic small-layer paving structures such as particleboards face many challenges. Due to the large discreteness of particleboard performance and geometric parameters, the complex splicing and stacking structure of particleboards, and the non-continuous load transfer path between particleboards, the structure-function relationship between the complex structure and mechanical properties of particleboard is still unclear. The design and optimization of particleboard paving and gluing processes lack theoretical guidance, making it difficult to give full play to the inherent mechanical properties of particleboards. For example, the random assembly structure of a large number of small layers leads to stress concentration in the splicing structure of non-continuous anisotropic particleboards in the horizontal direction, and the stacking structure in the vertical direction is prone to complex mechanical behaviors such as interlayer stress difference and interlayer failure. How to study the structure-function relationship between the anisotropic small-layer paving structure (including the splicing structure of various anisotropic layers in the horizontal direction and the stacking structure of various anisotropic layers in the vertical direction) and the macroscopic mechanical properties of the board is still an important and challenging issue.

[0004] The current methods for regulating the performance of particleboard include regulating the mechanical properties of particleboard through surface densification (patent number: CN202110669498.0), gradient paving structure (patent number: CN202122101122.8), optimizing the particle selection range (patent number: CN202211559697.7), etc.; improving the moisture, humidity and antibacterial properties of particleboard through surface chemical modification (patent number: CN202110408694.2), etc. At present, the methods for analyzing the mechanical properties of particleboard include reliability and importance analysis of composite laminates (patent numbers: CN202011144951.8, CN202011144952.2), finite element simulation of variable stiffness composite laminates (patent number: CN202010662044.6), microscopic finite element modeling method of fiber reinforced composites based on CT scanning images (patent number: CN202011201949.X), testing the compression properties of thermoplastic composite flat plates by mechanical testing machine (patent number: CN202210851152.7), automated testing system and method for mechanical properties of composite materials (patent number: CN202211214484.0), experimental device and method for tensile mechanical properties of unidirectional composite materials (patent number: CN201910073695.9), etc. The above research progress shows that the experimental technology for the preparation and performance regulation of artificial boards has been well developed. However, the testing, simulation and analysis methods for the mechanical properties of artificial boards are mainly based on the laminate model. Theoretical analysis, numerical simulation and machine learning research on the structural characteristics of particleboards "composed of anisotropic small particle pieces spliced ​​and stacked" are still lacking, resulting in unclear structure-activity relationship between the structure and mechanical properties of particleboards, and lack of theoretical guidance for board preparation.

[0005] On the other hand, the randomness of the particleboard paving structure on the production line is large, and the stability of the mechanical properties quality control of particleboard is insufficient. It is of great significance to conduct real-time performance prediction and quality control of the mechanical properties of particleboard during the production process. At present, the detection and analysis methods of the mechanical properties of particleboard include surface defect detection based on neural network and color camera (patent number: CN202310238055.5), image defect classification detection (patent number: CN202110133237.7), adaptive board thickness surface defect detection (patent number: CN202111680878.0), visual Transformer surface defect detection (patent number: CN202111680499.1), non-destructive testing of bamboo-wood composite container bottom plate using image processing based on machine learning algorithm (patent number: CN202010699496.1), particleboard surface defect detection based on deep learning algorithm (paper: https: / / doi.org / 10.3390 / s22207733), and prediction of particleboard mechanical properties based on principal component regression-random forest (paper: https: / / www.doi.org / 10.15376 / biores.16.2.2448-2471). Among them, a class of methods for detecting surface and internal structural defects of particleboard has the advantages of obtaining particleboard structural information, fast detection speed, and continuous automatic detection. However, this type of method cannot obtain stress distribution information based on the structure. The method of mechanical testing of board specimens through mechanical experiments can obtain the mechanical properties of the board. However, this type of method usually requires experimental testing after the board is processed, and requires a large amount of time and economic cost, and it is difficult to achieve real-time online rapid detection and analysis of the mechanical properties of the board. The above research progress shows that current technology still finds it difficult to achieve rapid, high-precision, and low-cost particleboard mechanical property analysis and prediction of "obtaining stress distribution based on geometric images and obtaining mechanical properties based on structural parameters."

[0006] With the continuous advancement of science and technology, there are endless methods for predicting the mechanical properties of materials through artificial intelligence and other means. For example, the cross-scale simulation method for predicting the mechanical properties of carbon fiber reinforced composite materials (patent number: CN201811116128.9) and the material performance prediction method based on deep learning (patent number: CN202010415084.0). Its cross-scale simulation method for predicting the mechanical properties of carbon fiber reinforced composite materials realizes variable parameter modeling based on the uniform stiffness method, reduces the difficulty of mesh division, and improves calculation efficiency, but it is not applicable to the mechanical properties of artificial boards in the form of stacked structures of various anisotropic small layers with large discreteness. The material performance prediction method of deep learning simplifies the calculation and improves the calculation speed, but does not simplify the structure. It faces difficulties such as large RVE volume of the overall mechanical properties of composite materials and difficulty in capturing key mechanisms.

[0007] In summary, there is an urgent need to develop an intelligent prediction and analysis method for the mechanical properties of particleboard that can simplify the complex structure of particleboard, comprehensively analyze the influence of horizontal splicing structure and vertical stacking structure, have a fast detection speed, high processing accuracy, and low technical cost. Summary of the invention

[0008] In view of this, the purpose of the present invention is to propose a particleboard mechanical properties prediction and analysis method based on simulation and deep learning, which has a simplified structure, strong versatility, fast detection speed and strong quality control.

[0009] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0010] A prediction and analysis method for mechanical properties of particleboard based on simulation and deep learning, comprising:

[0011] S01. Constructing a finite element model of a particleboard structure with unit thickness according to the particleboard paving structure data in the particleboard preparation process;

[0012] S02. Based on the finite element model, an equivalent single-layer board model of particleboard is constructed according to the finite element simulation method and equivalent modulus theory, and then an equivalent laminated board model is established according to the classical laminated board theory;

[0013] S03. Constructing a data set of particleboard structure and performance based on the constructed finite element model, equivalent single-layer board model and / or equivalent laminated board model, and then using the obtained data set to train an artificial intelligence model until it converges;

[0014] S04. Collect parameters during the particleboard installation process and input them into a trained artificial intelligence model. The artificial intelligence model predicts and analyzes the mechanical properties of the particleboard and outputs the prediction and analysis results.

[0015] As a possible implementation, further, in this solution S03, the artificial intelligence model includes a deep learning model and a machine learning model; both the deep learning model and the machine learning model are trained using the data set to meet the preset requirements;

[0016] Among them, the deep learning model is used to predict one or more data of the stress distribution of the splicing structure during particleboard paving, the equivalent laying angle of the equivalent single-layer board, and the equivalent material engineering constant; the machine learning model is used to predict the mechanical properties of the particleboard.

[0017] As a preferred implementation option, preferably, this solution S04 includes:

[0018] S041. Collect the geometric morphology data of the particleboard per unit thickness during the particleboard laying process, use it as an input parameter, and use the trained deep learning model to predict the stress distribution of the particleboard splicing structure, as well as one or more of the equivalent ply angle and equivalent material engineering constant of the equivalent single-layer board, to obtain the equivalent single-layer board data;

[0019] S042. Based on the data of equivalent single-layer boards, the equivalent single-layer boards are vertically stacked into equivalent laminates of particleboard to deduce the structural parameters of the equivalent laminates, which are then used as input parameters to predict the mechanical properties of particleboard using the trained machine learning model.

[0020] As a preferred implementation option, preferably, in the present solution S01, the wood chip paving structure data is an optical image of the wood chip paving structure, which is obtained by an industrial camera instrument to provide one or more structural information of wood chip length, width, distribution density, and fiber angle.

[0021] As a preferred implementation option, preferably, in this scheme S01, the finite element model of the unit thickness particleboard structure is based on the optical diagram of the particleboard paving structure during the particleboard preparation process, and one or more structures of the particleboard length, width, distribution density, and fiber angle are characterized through finite element modeling. It also sets the particleboard anisotropic material engineering constants and ply angles according to the fiber angles of the particleboards in each area, and performs a general static analysis on the established unit thickness bottom splicing structure model, so as to calculate the stress-strain distribution and / or constraint reaction force of the unit thickness particleboard structure under tensile and shear displacement conditions.

[0022] As a preferred implementation option, preferably, in this scheme S02, the equivalent single-layer board model of the particleboard is a simulation result output by a finite element model based on a unit thickness particleboard structure. By solving the principal stress direction, equivalent stress, equivalent strain, and equivalent anisotropic material parameters of the two-dimensional shell unit model, the unit thickness particle board splicing structure is equivalently simplified to an equivalent single-layer board with a continuous anisotropic layer angle and a consistent ply angle, and the equivalent material engineering constants and equivalent ply angles of the equivalent single-layer board are obtained.

[0023] As a preferred implementation option, preferably, in this scheme S02, the equivalent laminate model is established according to the classical laminate theory, that is, the equivalent single-layer boards of horizontally spliced ​​structures of unit thickness particle sheets of different heights are stacked in sequence according to the vertical paving order of the particleboard to form an equivalent laminate model of the particleboard, and then the equivalent laminate of the particleboard is calculated based on the equivalent material engineering constants and equivalent ply angles of each equivalent single-layer board, and the mechanical properties of the particleboard are obtained by simulation based on the classical laminate theory and the equivalent laminate model.

[0024] As a preferred implementation option, preferably, in this solution S03, the data set of particleboard structure and performance includes the following two categories:

[0025] (1) A data set with the geometric structure of unit thickness wood chips as input and the stress and strain distribution on the structure as output;

[0026] (2) A data set with particleboard ply structure parameters as input and particleboard mechanical properties as output.

[0027] As a preferred implementation option, preferably, the deep learning model described in this solution uses the aforementioned experimental, simulation and theoretical methods to construct a dataset of geometric morphology of unit thickness particle board splicing structure and stress distribution under unit displacement; it uses the unit thickness particle board splicing geometric morphology as input and stress distribution as output, and establishes the association between geometric image and stress distribution by training the hidden layer and nonlinear activation function of the deep learning model. That is, this solution uses the current unit thickness particle board paving geometric morphology on the bamboo particle board production line captured by an industrial camera as input parameters, and uses the trained deep learning model to quickly predict the stress distribution of the splicing structure.

[0028] As a better implementation option, preferably, the machine learning model described in this scheme uses the aforementioned experimental, simulation and theoretical methods to construct a data set of particleboard structural parameters and macro-mechanical properties. It uses structural parameters as independent variables and macro-mechanical properties as target variables to train the weights and biases of the machine learning model, and establish the association between structural parameters and mechanical properties. The machine learning model records the equivalent ply angles and equivalent material engineering constants of the "equivalent single-layer board" at different positions in the vertical direction in sequence according to the paving order of the particleboard on the bamboo particleboard production line, and stacks the "equivalent single-layer board" vertically into a particleboard "equivalent laminate". The current "equivalent laminate" structural parameters (number of pavements, ply thickness, ply angle ratio, ply order, etc.) are used as input parameters, and the trained machine learning model is used to quickly predict the macro-mechanical properties of the board.

[0029] In this scheme, the equivalent laminate model will simplify the complex structure of the particleboard and reflect the random splicing structure of the particleboard in the horizontal direction and the layered stacking structure of the particleboard in the vertical direction.

[0030] Preferably, the machine learning model is one or more of a support vector machine, a decision tree, a random forest, and a Bayesian learner.

[0031] As a better implementation option, preferably, the deep learning model described in this scheme is a fully convolutional neural network deep learning model; the machine learning model is an integrated machine learning model based on a decision tree.

[0032] By adopting the above-mentioned technical scheme, the present invention has the following beneficial effects compared with the prior art: This scheme analyzes and predicts the mechanical properties of particleboard by combining finite element and artificial intelligence. According to the optical image of the particleboard paving structure in the particleboard preparation process, a unit thickness particle structure model is constructed, and finite element analysis is performed. Finally, an "equivalent single-layer board" model suitable for particleboard splicing structure analysis is constructed to analyze the impact of horizontal splicing on performance and an "equivalent laminated board" model for stacking structure analysis is constructed to analyze the impact of vertical stacking on performance, providing a theoretical basis for the "structure-performance integrated design" of this type of artificial board; based on the above-mentioned method, a data set of particleboard structure and performance, particleboard splicing appearance and stress distribution is established, and machine learning and deep learning models are constructed and trained. The stress distribution of the unit thickness splicing structure is predicted through a deep learning model based on the paving structure of the particleboard. Thus, the unit thickness "equivalent single-layer board" model is equivalent, and then stacked into an "equivalent laminated board" model. The "equivalent laminated board" model is used to quickly predict the mechanical properties of the particleboard through a machine learning model, and finally the particleboard quality control information is output, providing a method for online intelligent detection of product quality. In addition, this method also has the advantages of simplified structure, strong versatility, fast detection speed, and strong quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a simplified flow chart in one embodiment of the present invention; in the figure, (a) is a method for predicting stress distribution based on the geometric morphology of bamboo particle board paving using a deep learning model, and (b) is a method for predicting mechanical properties based on the structural parameters of the equivalent laminate of bamboo particle board using a machine learning model;

[0035] Figure 2 The technical roadmap of one embodiment of the present invention is as follows. The structure-activity relationship and intelligent prediction of the mechanical properties of bamboo particleboard are studied by finite element numerical simulation, theoretical model construction, artificial intelligence prediction and other methods.

[0036] Figure 3 The main process diagram of preparing bamboo particle board in one embodiment of the present invention is shown in the figure; in the figure, (a) drying of bamboo particle board, (b) gluing, (c) forming, (d) paving, (e) hot pressing, and (f) post-processing;

[0037] Figure 4Schematic diagram of constructing an equivalent single-layer board model of a bamboo shavings horizontal splicing structure in an embodiment of the present invention; in the figure, (a) the bamboo shavings splicing structure is modeled, (b) the bamboo shavings fiber orientation and anisotropic material parameters are set, (c) the structural stress distribution under uniaxial tensile displacement and shear displacement, (d) the structural equivalent stress analysis through the equivalent modulus theory, (e) the equivalent single-layer board of the bamboo shavings splicing structure with unit thickness;

[0038] Figure 5 Schematic diagram of constructing an equivalent laminate model of a vertically stacked structure of bamboo shavings in one embodiment of the present invention; in the figure, (a) an optical picture of a bamboo shavings paving structure, (b) an intelligent identification method for bamboo shavings, (c) a bamboo shavings splicing structure model constructed according to the geometric parameter rules of bamboo shavings, (d) a bamboo shavings splicing structure model and grid division, (e) stress and strain distribution on anisotropic bamboo shavings splicing structures, (f) a vertically stacked structure of bamboo shavings and its corresponding (g) equivalent single-layer board stacking structure; (h) an equivalent laminate structure of bamboo shavings;

[0039] Figure 6 Schematic diagram of the analysis of the "equivalent laminate" model of the vertical stacking structure of bamboo particle boards in one embodiment of the present invention; in the figure, (a) the equivalent laminate of bamboo particle board is constructed by stacking equivalent single-layer boards, (b) the mechanical properties of the equivalent laminate of bamboo particle board are analyzed by the classical laminate theory, (c) the finite element ply setting of the equivalent laminate of bamboo particle board and (d) the mechanical simulation;

[0040] Figure 7 This is a structural diagram of a convolutional neural network model composed of an encoder and a decoder used in this embodiment;

[0041] Figure 8 Graph showing the changes in model indicators during model training in this embodiment; (a) is the mean absolute error MAE, (b) is the mean square error MSE, and (c) is the correlation coefficient R 2 Graph of changes in . DETAILED DESCRIPTION

[0042] The present invention will be further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present invention, but are not intended to limit the scope of the present invention. Similarly, the following examples are only partial embodiments of the present invention rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0043] Combination Figures 1 to 6 As shown in one of the embodiments, a particleboard mechanical property prediction and analysis method based on simulation and deep learning is provided, which includes:

[0044] S01. Constructing a finite element model of a particleboard structure with unit thickness according to the particleboard paving structure data in the particleboard preparation process;

[0045] S02. Based on the finite element model, an equivalent single-layer board model of particleboard is constructed according to the finite element simulation method and equivalent modulus theory, and then an equivalent laminated board model is established according to the classical laminated board theory;

[0046] S03. Constructing a data set of particleboard structure and performance based on the constructed finite element model, equivalent single-layer board model and / or equivalent laminated board model, and then using the obtained data set to train an artificial intelligence model until it converges;

[0047] S04. Collect parameters during the particleboard installation process and input them into a trained artificial intelligence model. The artificial intelligence model predicts and analyzes the mechanical properties of the particleboard and outputs the prediction and analysis results.

[0048] Among them, in scheme S03 of this embodiment, the artificial intelligence model includes a deep learning model and a machine learning model; both the deep learning model and the machine learning model are trained using the data set to meet preset requirements.

[0049] Accordingly, the deep learning model described in this scheme is used to predict one or more of the stress distribution of the splicing structure during particleboard paving, the equivalent laying angle of the equivalent single-layer board, and the equivalent material engineering constant; the machine learning model is used to predict the mechanical properties of the particleboard.

[0050] As a better implementation option, preferably, this solution S04 includes:

[0051] S041. Collect the geometric morphology data of the particleboard per unit thickness during the particleboard laying process, use it as an input parameter, and use the trained deep learning model to predict the stress distribution of the particleboard splicing structure, as well as one or more of the equivalent ply angle and equivalent material engineering constant of the equivalent single-layer board, to obtain the equivalent single-layer board data;

[0052] S042. Based on the data of equivalent single-layer boards, the equivalent single-layer boards are vertically stacked into equivalent laminates of particleboard to deduce the structural parameters of the equivalent laminates, which are then used as input parameters to predict the mechanical properties of particleboard using the trained machine learning model.

[0053] In brief, this approach includes experimental preparation and testing, finite element numerical simulation, theoretical model construction, and artificial intelligence prediction.

[0054] Among them, combined Figure 3As shown, the experimental preparation includes obtaining the size parameter information of bamboo shavings (length, width, distribution density, fiber angle, etc.), exploring the sizing scheme and directional paving method suitable for bamboo particleboard, constructing slabs with different bamboo shavings paving angles, and exploring the hot pressing conditions and post-processing conditions for the preparation of bamboo particleboard. In this scheme, the mechanical properties of bamboo particleboard can be quantified by bending stiffness and strength. Through the above work, the experimental results of the structural parameters and mechanical properties of bamboo particleboard were obtained, which provided experimental data for the subsequent theoretical model construction, finite element numerical simulation, machine learning and deep learning model training.

[0055] For finite element numerical simulation, in this scheme S01, the finite element model of the unit thickness particleboard structure is based on the optical diagram of the particleboard paving structure during the particleboard preparation process, and one or more structures of the particleboard length, width, distribution density, and fiber angle are characterized through finite element modeling. It also sets the anisotropic material engineering constants and ply angles of the particleboard according to the fiber angles of the particleboards in each area, and performs a general static analysis on the established unit thickness bottom splicing structure model, so as to calculate the stress-strain distribution and / or constraint reaction force of the unit thickness particleboard structure under tensile and shear displacement conditions.

[0056] As a possible implementation example, combining Figure 4 , Figure 5 As shown, in the present scheme S01, the wood chip paving structure data is an optical image of the wood chip paving structure, which is obtained by an industrial camera instrument to provide a statistical distribution of one or more structural parameters such as length, width, distribution density, fiber angle, etc. of the wood chip; according to the parameter statistical law, the ABAQUS pre-processing module based on MATLAB and Python is used to construct a two-dimensional shell unit model of anisotropic small pieces per unit thickness, and the anisotropic material engineering constants and ply angles of the bamboo chip are set according to the fiber angle of the bamboo chip in each area; the finite element method is used to simulate the stress-strain distribution and constraint reaction force of the two-dimensional model under uniaxial tension and shear displacement conditions.

[0057] As for the construction of the theoretical model, as a better implementation option, preferably, in this scheme S02, the equivalent single-layer board model of the particleboard is a simulation result output by a finite element model based on a unit thickness particleboard structure. By solving the principal stress direction, equivalent stress, equivalent strain, and equivalent anisotropic material parameters of the two-dimensional shell unit model, the unit thickness particle board splicing structure is equivalently simplified to an equivalent single-layer board with a continuous anisotropic layer angle and a consistent ply angle, and the equivalent material engineering constants and equivalent ply angles of the equivalent single-layer board are obtained.

[0058] In addition, in this solution S02, the equivalent laminate model is established according to the classical laminate theory. The equivalent single-layer boards of horizontally spliced ​​structures of unit thickness particle sheets of different heights are stacked in sequence according to the vertical paving order of the particle board to form an equivalent laminate model of the particle board. Then, based on the equivalent material engineering constants and equivalent ply angles of each equivalent single-layer board, the equivalent laminate of the particle board is calculated, and the mechanical properties of the particle board are obtained by simulation based on the classical laminate theory and the equivalent laminate model.

[0059] As a specific example, Figure 5 , Figure 6 As shown in the figure, this scheme includes solving the principal stress direction, equivalent stress, equivalent strain, and equivalent anisotropic material parameters of the two-dimensional shell unit model based on the finite element simulation results, and then simplifying the unit thickness bamboo particle splicing structure to a continuous anisotropic "equivalent single-layer board" with consistent ply angles, and obtaining the equivalent material engineering constants and equivalent ply angles of the "equivalent single-layer board". According to the paving sequence of the bamboo particleboard preparation process, the "equivalent single-layer board" model of the horizontal splicing structure of bamboo particleboards at different paving moments is stacked in sequence to form the "equivalent laminated board" model of the bamboo particleboard. The "equivalent laminated board" model of the bamboo particleboard is constructed using ABAQUS software, and the equivalent material engineering constants and equivalent ply angles of each layer are set in turn; the laminated board stress-strain distribution and constraint reaction force of the bamboo particleboard under uniaxial tension and three-point bending loading are simulated; the stiffness matrix of the "equivalent laminated board" of the bamboo particleboard is calculated using the classical laminated board theory, and the mechanical properties of the bamboo particleboard are tested using uniaxial tension and three-point bending loading experiments.

[0060] As a preferred implementation option, preferably, in this solution S03, the data set of particleboard structure and performance includes the following two categories:

[0061] (1) A data set with the geometric structure of unit thickness wood chips as input and the stress and strain distribution on the structure as output;

[0062] (2) A data set with particleboard ply structure parameters as input and particleboard mechanical properties as output.

[0063] Combination Figure 1As shown, in terms of artificial intelligence prediction, the deep learning model described in this scheme uses the aforementioned experimental, simulation and theoretical methods to construct a dataset of geometric morphology of unit thickness particle splicing structure and stress distribution under unit displacement; it takes the unit thickness particle splicing geometric morphology as input and stress distribution as output, and establishes the association between geometric image and stress distribution by training the hidden layer and nonlinear activation function of the deep learning model. That is, this scheme uses the current unit thickness particle pavement geometry on the bamboo particle board production line captured by an industrial camera as input parameters, and uses the trained deep learning model to quickly predict the stress distribution of the splicing structure. Then, based on the stress distribution prediction results, the equivalent ply angle and equivalent material engineering constant of the "equivalent single-layer board" are determined.

[0064] As a better implementation option, preferably, the machine learning model described in this scheme uses the aforementioned experimental, simulation and theoretical methods to construct a data set of particleboard structural parameters and macro-mechanical properties. It uses structural parameters as independent variables and macro-mechanical properties as target variables to train the weights and biases of the machine learning model, and establish the association between structural parameters and mechanical properties. The machine learning model records the equivalent ply angles and equivalent material engineering constants of the "equivalent single-layer board" at different positions in the vertical direction in sequence according to the paving order of the particleboard on the bamboo particleboard production line, and stacks the "equivalent single-layer board" vertically into a particleboard "equivalent laminate". The current "equivalent laminate" structural parameters (number of pavements, ply thickness, ply angle ratio, ply order, etc.) are used as input parameters, and the trained machine learning model is used to quickly predict the macro-mechanical properties of the board. Based on the stress distribution prediction results, the equivalent ply angle and equivalent material engineering constant of the "equivalent single-layer board" are determined, and the "equivalent single-layer board" is stacked into an "equivalent laminate" according to the paving order and the laminate structure parameters are extracted; the machine learning model is used to predict the mechanical properties of the bamboo particleboard based on the laminate structure parameters. Rapid, high-precision, and low-cost prediction of the mechanical properties and quality control of bamboo particleboard are achieved.

[0065] In this scheme, the equivalent laminate model will simplify the complex structure of the particleboard and reflect the random splicing structure of the particleboard in the horizontal direction and the layered stacking structure of the particleboard in the vertical direction.

[0066] Preferably, the machine learning model is one or more of a support vector machine, a decision tree, a random forest, and a Bayesian learner.

[0067] As a better implementation option, preferably, the deep learning model described in this scheme is a fully convolutional neural network deep learning model; the machine learning model is an integrated machine learning model based on a decision tree.

[0068] In order to further illustrate the present embodiment, an example is given below to further illustrate the present embodiment:

[0069] Obtain relevant dimensional parameter information of bamboo particleboard. This includes measuring hundreds of bamboo shavings, obtaining the length and width of the shavings in the range of 40-90mm, the width in the range of 10-28mm, and measuring the number of shavings in a certain number of particleboards, with a distribution density of 0.68-0.75g / cm 3 , apply glue to the measured wood chips, the glue amount is 12%, and stir for 15 minutes in both the forward and reverse directions. Randomly lay the wood chips into the mold for pre-pressing to form a slab, let it stand for 3-5 minutes, and then use a hot press to press it at 140℃ for 15 minutes with a hot pressure of 20Mpa. Finally, the bamboo particle board laid at random angles is made. Figure 3 The pressed bamboo particleboard was subjected to tensile performance tests and other tests, and its material parameters were obtained for simulation in finite element.

[0070] Finite element model numerical simulation. The finite element model calculation and solution are all carried out in ABAQUS2020 2.0 software. According to the bamboo particle board experimental preparation measured bamboo particle board structural parameters, PYTHON secondary development was used to establish 200 150×150mm two-dimensional shell element random angle anisotropic small layer models in ABAQUS software. Figure 5 c. According to the area of ​​the bamboo shavings, the fiber angle and material engineering constants of the bamboo shavings are set as follows: Figure 4 As shown in b, its material engineering constants and fiber angle range are shown in Table 1. According to the consistent setting of material engineering constants, the assembly angle of each piece of wood shavings is randomly spliced ​​around the x-axis, which represents the discrete type of the material to the greatest extent, resulting in the difference in its mechanical properties. The mesh division adopts a four-node double-bent thin shell to reduce integration, hourglass control, and finite membrane strain. The load and boundary conditions are set to be fully fixed on the left and a 1mm displacement load is applied to the right (x-axis loading); the lower side is fully fixed and a 1mm displacement load is applied to the upper side (y-axis loading). After the calculation is completed, the Mises stress value of the node is exported from ABAQUS, and the finite element post-processing program based on PYTHON secondary development is used to extract the particleboard node coordinates from ABAQUS.

[0071] Theoretical analysis. ABAQUS is used to calculate the macroscopic response and material properties of the material by applying unidirectional displacement strain. Loading is performed in the 1 and 2 normal directions and the 12 tangent direction, and the stress response of the representative plane is calculated. The extracted node coordinates and stress values ​​are interpolated by MATLAB to calculate the average stress value. The macroscopic average stress of the representative volume unit is obtained by averaging the volume of the stress, so as to solve the following material constitutive stiffness matrix partial coefficients.

[0072]

[0073] After solving the constitutive stiffness matrix, the constitutive stiffness matrix analysis coefficients are input into the ABAQUS material property module for equivalent stress calculation, and then the equivalent single-layer plate material properties are obtained. Each equivalent single-layer plate is layered and input into the laminate model to obtain the equivalent laminate model material properties.

[0074] Table 1 Engineering constants of particleboard and random angle range of particle board

[0075]

[0076] A deep learning model was established. The two-dimensional image data exported from the ABAQUS software was used for model learning and training. The two-dimensional image data was sliced ​​into layers of 64×64 pixels. Data from different plane models were used to train and evaluate the model. The processed two-dimensional data was divided into a training set and a validation set. The training set and validation set account for approximately 93% and 7% of the entire data set size, respectively. In this embodiment, the training data set size is 595, and the validation set data size is 45. The corresponding two-dimensional material image, the corresponding tensile load value of the displacement loading in the x-direction and y-direction, and the two-dimensional stress distribution image were reshaped into a one-dimensional vector of length 8194 (64×64+2+64×64). Therefore, a model is represented as a two-dimensional matrix with a dimension of 64×8194.

[0077] The convolutional neural network model structure used is a kind of encoder and decoder composed of Figure 7 As shown in the figure, the encoder is used to extract image features, and the decoder restores the model's final output image size to the same size as the input image, which is 64×64×1. The three dimensions of the image are length, width, and channels. Each convolution layer has a convolution kernel of size 3×3 and stride 1×1. The third dimension of the convolution kernel is the same as the number of channels of the input image. After passing through the encoder in the SCSNet model, the output feature map is 8×8×128 in size, and the feature map is reshaped into a one-dimensional vector with a length of 1×1×8192 in the reshape layer E8 to meet the input requirements of the fully connected layer. After passing through the fully connected layers E9 and E10, the model learns to obtain the potential features of the microscopic geometric structure of the anisotropic small platelet layer. Before the encoder's result is input to the decoder, the different tensile load values ​​in the x and y directions are connected to the vector output by the fully connected layer E10 in the E11 layer. The decoder has the opposite process to the encoder. Through deconvolution and unpooling methods, the low-resolution feature map generated by the encoder is mapped and restored to a high-resolution image. By mean absolute error MAE, mean square error MSE, correlation coefficient R 2 Indicators are used to evaluate the prediction accuracy of the model. The formula is as follows.

[0078] Mean absolute error:

[0079] Mean Square Error:

[0080] Correlation coefficient:

[0081] In the formula, n represents the number of data samples, i represents the i-th data sample, and y i represents the true value of the i-th sample, represents the predicted value, Represents the average value of the true value of all samples. The data in the validation set are input into the stress distribution prediction model according to the geometric morphology. From the above indicators, it can be seen that both the MSE and MAE curves drop rapidly in the first fifty training iterations (epochs), then enter the convergence stage and gradually tend to a stable state. Figure 8 After 100 training iterations (epochs), the loss function is basically constant. The correlation coefficient finally stabilizes at 0.92, and the stress prediction distribution can meet the prediction requirements.

[0082] The above descriptions are only some embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A particleboard mechanical properties prediction and analysis method based on simulation and deep learning, characterized in that: It includes: S01. Constructing a finite element model of a particleboard structure with unit thickness according to the particleboard paving structure data in the particleboard preparation process; S02. Based on the finite element model, an equivalent single-layer board model of particleboard is constructed according to the finite element simulation method and equivalent modulus theory, and then an equivalent laminated board model is established according to the classical laminated board theory; S03. Constructing a data set of particleboard structure and performance based on the constructed finite element model, equivalent single-layer board model and / or equivalent laminated board model, and then using the obtained data set to train an artificial intelligence model until it converges; S04. Collect parameters during the particleboard installation process and input them into a trained artificial intelligence model. The artificial intelligence model predicts and analyzes the mechanical properties of the particleboard and outputs the prediction and analysis results.

2. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 1, characterized in that: In S03, the artificial intelligence model includes a deep learning model and a machine learning model; both the deep learning model and the machine learning model are trained using the data set to meet preset requirements; The deep learning model is used to predict one or more of the following data: stress distribution of the splicing structure during particleboard paving, equivalent ply angle of equivalent single-layer board, and equivalent material engineering constant; the machine learning model is used to predict the mechanical properties of particleboard; S04 includes: S041. Collect the geometric morphology data of the particleboard per unit thickness during the particleboard laying process, use it as an input parameter, and use the trained deep learning model to predict the stress distribution of the particleboard splicing structure, as well as one or more of the equivalent ply angle and equivalent material engineering constant of the equivalent single-layer board, to obtain the equivalent single-layer board data; S042. Based on the data of equivalent single-layer boards, the equivalent single-layer boards are vertically stacked into equivalent laminates of particleboard to deduce the structural parameters of the equivalent laminates, which are then used as input parameters to predict the mechanical properties of particleboard using the trained machine learning model.

3. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 1 or 2, characterized in that: In S01, the wood chip paving structure data is an optical image of the wood chip paving structure, which is obtained by an industrial camera instrument to provide one or more structural information of the wood chip length, width, distribution density, and fiber angle.

4. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 1, characterized in that: In S01, the finite element model of the particleboard structure of unit thickness is used to characterize one or more structures of the length, width, distribution density, and fiber angle of the particle board. It also sets the anisotropic material engineering constants and ply angles of the particle board according to the fiber angle of the particle board in each area, so as to calculate the stress-strain distribution and / or constraint reaction force of the particleboard structure of unit thickness under tensile and shear displacement conditions.

5. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 4, characterized in that: In S02, the equivalent single-layer board model of the particleboard is a simulation result output by a finite element model based on a unit thickness particleboard structure. By solving the principal stress direction, equivalent stress, equivalent strain, and equivalent anisotropic material parameters of the two-dimensional shell unit model, the unit thickness particle board splicing structure is equivalently simplified to an equivalent single-layer board with a continuous anisotropic layer angle and a consistent ply angle, and the equivalent material engineering constants and equivalent ply angles of the equivalent single-layer board are obtained.

6. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 5, characterized in that: In S02, the equivalent laminate model is established according to the classical laminate theory. The equivalent single-layer boards with unit thickness of particleboards of different heights and horizontal splicing structures are stacked in sequence according to the vertical paving order of the particleboard to form an equivalent laminate model of the particleboard. Then, based on the equivalent material engineering constants and equivalent ply angles of each equivalent single-layer board, the equivalent laminate of the particleboard is calculated. The mechanical properties of the particleboard are obtained by simulation based on the classical laminate theory and the equivalent laminate model.

7. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 2, characterized in that: In S03, the particleboard structure and performance data sets include the following two categories: (1) A data set with the geometric structure of unit thickness wood chips as input and the stress and strain distribution on the structure as output; (2) A data set with particleboard ply structure parameters as input and particleboard mechanical properties as output.

8. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 2, characterized in that: The deep learning model uses the geometric shape of the splicing of wood chips with unit thickness as input and the stress distribution as output, and establishes the association between the geometric image and the stress distribution by training the hidden layer and nonlinear activation function of the deep learning model.

9. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 2, characterized in that: The machine learning model uses structural parameters as independent variables and macroscopic mechanical properties as target variables to train the weights and biases of the machine learning model and establish the relationship between structural parameters and mechanical properties; The machine learning model is one or more of a support vector machine, a decision tree, a random forest, and a Bayesian learner.

10. The particleboard mechanical properties prediction and analysis method based on simulation and deep learning as claimed in claim 2, characterized in that: The deep learning model is a fully convolutional neural network deep learning model; the machine learning model is an integrated machine learning model based on a decision tree.

Images