Search

CN-121659798-B - Bamboo-wood composite material creep prediction method based on rheological topological mapping

CN121659798BCN 121659798 BCN121659798 BCN 121659798BCN-121659798-B

Abstract

The application relates to a bamboo-wood composite creep prediction method based on rheological topology mapping, which comprises the following steps of S1, obtaining microstructure scanning data and environment parameters of a bamboo-wood composite, constructing a structure tensor field containing fiber direction information based on the microstructure scanning data and the environment parameters, S2, constructing a rheological topology mapping network, wherein the input of the rheological topology mapping network is the structure tensor field, the output layer of the rheological topology mapping network is configured to output specific rheological model parameters, and a thermodynamic monotonicity constraint term and a parameter non-negative constraint term are introduced to construct a loss function, the loss function is used for training the rheological topology mapping network, S3, obtaining monitoring data of sparse measuring points of an engineering site, reversely correcting the rheological model parameters through a Markov chain Monte Carlo sampling algorithm, and S4, substituting the corrected rheological model parameters into a rheological formula to obtain creep strain.

Inventors

  • WANG QIAN
  • YANG YING
  • LI KANGXIN
  • CHEN CHUPENG

Assignees

  • 湖南工商大学

Dates

Publication Date
20260508
Application Date
20260203

Claims (8)

  1. 1. A bamboo-wood composite creep prediction method based on rheological topological mapping is characterized by comprising the following steps: S1, acquiring microstructure scanning data and environmental parameters of a bamboo-wood composite material, and constructing a structure tensor field containing fiber direction information based on the microstructure scanning data and the environmental parameters; S2, constructing a rheological topology mapping network, wherein the input of the rheological topology mapping network is the structure tensor field, the output layer of the rheological topology mapping network is configured to output rheological model parameters, and a thermodynamic monotonicity constraint term and a parameter non-negative constraint term are introduced to construct a loss function, the loss function is used for training the rheological topology mapping network, the rheological model parameters comprise modulus and viscosity, and the expression of the loss function is as follows: ; ; Wherein, the The loss function is represented by a function of the loss, Representing the mean square error between the predicted creep strain and the true strain, A first weight is indicated and a second weight is indicated, Representing thermodynamic monotonicity constraints, creep being irreversible under constant load, i.e. strain rate , Representing a ReLU activation function, employing the function to punish points in time that violate thermodynamic monotonicity constraints, if the predicted strain rate is less than 0, the ReLU activation function will produce a positive punishment value, forcing network modification, Indicating the creep strain of the steel sheet, The time point of t is indicated as the time point, Indicating the deviation of the deflection of the beam, A second weight is indicated as being indicative of a second weight, Representing a parametric non-negative constraint term that is used to constrain modulus and viscosity to have positive values; S3, acquiring monitoring data of sparse measuring points of an engineering site, and reversely correcting the rheological model parameters through a Markov chain Monte Carlo sampling algorithm; s4, substituting the corrected rheological model parameters into a rheological formula to obtain creep strain.
  2. 2. The method for predicting creep of a bamboo-wood composite based on rheological topological mapping according to claim 1, wherein in S1, the microstructure scan data comprises a local gray scale field of the bamboo-wood composite, and the environmental parameters comprise stress tensor, temperature, and humidity; the structural tensor is calculated based on the local gray field of the bamboo-wood composite material, and the calculation formula is as follows: ; Wherein, the Representing the structure tensor of the i-th point in voxel space, A local gray field representing the i-th point in voxel space, Representing the local density gradient or gray field gradient of the i-th point in voxel space, Representing a transpose; performing eigenvalue decomposition on the structure tensor, and extracting the main fiber direction, radial constraint direction and tangential direction of the bamboo-wood composite material; And combining the stress tensor, the temperature, the humidity, the fiber main direction, the radial constraint direction and the tangential direction to obtain the structural tensor field.
  3. 3. The bamboo-wood composite creep prediction method based on rheological topology map of claim 1, wherein the rheological topology map network comprises: An input layer for acquiring the structure tensor field; a graphic neural network or a multi-layer perceptron is adopted as a backbone network and is used for mapping the structure tensor field into rheological model parameters; The output layer is used for outputting rheological model parameters; and the analysis and calculation layer is used for setting a rheological formula.
  4. 4. The method for predicting creep of a bamboo-wood composite based on a rheological topological map of claim 1, wherein the rheological model comprises a generalized Kelvin-Voigt model.
  5. 5. The method for predicting creep of a bamboo-wood composite based on rheological topological mapping according to claim 4, wherein the expression of the rheological formula is: ; Wherein, the The creep strain at time t is indicated, The stress tensor is represented by the number of stress tensors, Representing the modulus of the first Kelvin-Voigt unit, Representing the modulus of the kth Kelvin-Voigt unit, Representing the number of Kelvin-Voigt units, The viscosity of the kth Kelvin-Voigt unit is shown.
  6. 6. The bamboo-wood composite creep prediction method based on rheological topology mapping according to claim 1, wherein a gradient descent method is adopted to train the rheological topology mapping network based on the loss function, and network parameters of the trained rheological topology mapping network are updated.
  7. 7. The method for predicting creep of a bamboo-wood composite based on rheological topological mapping according to claim 1, wherein S3 comprises: acquiring monitoring data of sparse measuring points on an engineering site, wherein the monitoring data is displacement of the sparse measuring points on the bamboo-wood composite material; constructing a forward model based on the rheological model parameters, wherein the expression of the forward model is as follows: ; Wherein, the Representing the actual rheological model parameters of the kth Kelvin-Voigt unit, Indicating the local defect correction factor(s), Representing predicted rheological model parameters of the kth Kelvin-Voigt unit; constructing a likelihood function based on the monitoring data and rheological model parameters of the corresponding point positions; solving the likelihood function by adopting a Markov chain Monte Carlo sampling algorithm to obtain a parameter sample obeying posterior distribution, and calculating statistics of the parameter sample; substituting the statistic of the parameter sample into the actual rheological model parameter in the forward model to obtain the corrected rheological model parameter.
  8. 8. The method for predicting creep of a bamboo-wood composite based on rheological topological mapping of claim 7, wherein the likelihood function is expressed as: ; Wherein, the Representing the displacement of sparse measurement points And rheological model parameters A likelihood function between the two, Indicating that the ratio is proportional to the ratio, A predicted value representing the displacement under the rheological model parameters B of the corresponding point location, The L2 norm is represented by the number, Representing the measurement noise variance.

Description

Bamboo-wood composite material creep prediction method based on rheological topological mapping Technical Field The application relates to the technical field of bamboo-wood composite material creep prediction, in particular to a bamboo-wood composite material creep prediction method based on rheological topological mapping. Background The bamboo-wood composite material (such as inorganic glue composite bamboo beams and recombined bamboo) is taken as a biomass viscoelastic material, can generate remarkable creep deformation under the action of long-term load, and is highly sensitive to environmental temperature and humidity changes. The prior art has the following main defects when predicting the creep of bamboo materials: 1. The Findley, burgers model and other pure mechanism models have limitations, namely the pure mechanism model depends on an idealized assumption, so that the complex stress field caused by initial defects (knots and cracks) and non-uniformity inside the bamboo material is difficult to accurately represent, and the parameters of the model are fixed and cannot adapt to environmental changes. 2. The pure data driving models such as the long-short-term memory network LSTM, the convolutional neural network CNN and the like have physical failure risks, namely the existing deep learning method lacks physical mechanism constraint. Under small sample training data, predictions that violate thermodynamic laws (e.g., spontaneous reduction of strain under constant load) are very likely to occur, and errors accumulate over time in long-period extrapolation predictions. 3. The environment self-adaptation capability is poor, the temperature and humidity are changeable in actual engineering, once the training of the existing model is completed, parameters are solidified, and the model cannot be dynamically corrected by using sparse monitoring data of the engineering site to adapt to a new environment, so that the prediction precision is greatly reduced. Therefore, a high-precision creep prediction method capable of ensuring physical reliability and adaptively updating by using sparse data is needed. Disclosure of Invention Based on this, there is a need to provide a method for predicting creep of a bamboo-wood composite based on rheological topological mapping, comprising: S1, acquiring microstructure scanning data and environmental parameters of a bamboo-wood composite material, and constructing a structure tensor field containing fiber direction information based on the microstructure scanning data and the environmental parameters; S2, constructing a rheological topology mapping network, wherein the input of the rheological topology mapping network is the structure tensor field, and the output layer of the rheological topology mapping network is configured to output specific rheological model parameters; S3, acquiring monitoring data of sparse measuring points of an engineering site, and reversely correcting the rheological model parameters through a Markov chain Monte Carlo sampling algorithm; s4, substituting the corrected rheological model parameters into a rheological formula to obtain creep strain. Preferably, in S1, the microstructure scanning data includes a local gray field of the bamboo-wood composite material, and the environmental parameters include stress tensor, temperature and humidity; the structural tensor is calculated based on the local gray field of the bamboo-wood composite material, and the calculation formula is as follows: ; Wherein, the Representing the structure tensor of the i-th point in voxel space,A local gray field representing the i-th point in voxel space,Representing the local density gradient or gray field gradient of the i-th point in voxel space,Representing a transpose; performing eigenvalue decomposition on the structure tensor, and extracting the main fiber direction, radial constraint direction and tangential direction of the bamboo-wood composite material; And combining the stress tensor, the temperature, the humidity, the fiber main direction, the radial constraint direction and the tangential direction to obtain the structural tensor field. Preferably, the rheological topology map network comprises: An input layer for acquiring the structure tensor field; a graphic neural network or a multi-layer perceptron is adopted as a backbone network and is used for mapping the structure tensor field into rheological model parameters; an output layer for outputting specific rheological model parameters; and the analysis and calculation layer is used for setting a rheological formula. Preferably, the rheological model comprises a generalized Kelvin-Voigt model. Preferably, the rheological model parameters include modulus and viscosity. Preferably, the expression of the rheological formula is: ; Wherein, the The creep strain at time t is indicated,The stress tensor is represented by the number of stress tensors,Representing the modulus of the first Kelvin-Voigt unit,Re