CN-121999892-A - Method for predicting adhesion work of steel slag asphalt interface based on physical neural network

CN121999892ACN 121999892 ACN121999892 ACN 121999892ACN-121999892-A

Abstract

The invention provides a steel slag asphalt interface adhesion work prediction method based on a physical neural network, which relates to the technical field of deep learning and comprises the steps of obtaining chemical component parameters and environment influence factors of a steel slag asphalt mixture interface sample, creating an original neural network prediction model based on diffusion items, reaction source items, multiple mechanical dissipation items and state variables corresponding to the space-time evolution of adhesion work, inputting the chemical component parameters and the environment influence factors into the original neural network prediction model, generating interface adhesion work prediction parameters by combining physical constraint items, training the original neural network prediction model by utilizing the interface adhesion work prediction parameters and a gradient descent algorithm to obtain a target neural network prediction model, and inputting the chemical component parameters and the environment influence factors of a material to be detected into the target neural network prediction model to obtain the target prediction parameters.

Inventors

REN MINDA
MA LINHUA
WU YUNHAO
REN QINGQING
HAN QINGHAO
HE ZIYI
Pei Zian

Assignees

内蒙古工业大学

Dates

Publication Date: 20260508
Application Date: 20260119

Claims (10)

1. The method for predicting the adhesion work of the steel slag asphalt interface based on the physical neural network is characterized by comprising the following steps of: The method comprises the steps of obtaining chemical component parameters and environmental impact factors of a steel slag asphalt mixture interface sample, wherein the chemical component parameters comprise various oxide parameters, and the environmental impact factors comprise load cycle times, stress amplitude, temperature field, moisture content and space-time parameters related to material performance evolution; creating an original neural network prediction model based on diffusion items, reaction source items, multi-mechanism dissipation items and state variables corresponding to the space-time evolution of the adhesion work; inputting the chemical component parameters and the environmental impact factors into the original neural network prediction model, and generating an interface adhesion work prediction parameter by combining a physical constraint item, wherein the physical constraint item comprises data loss, PDE residual loss, state equation loss and boundary condition loss; And inputting chemical component parameters and environmental impact factors of the material to be detected into the target neural network prediction model to obtain target prediction parameters.
2. The method for predicting the adhesion work of a steel slag asphalt interface based on a physical neural network according to claim 1, wherein the step of obtaining the chemical composition parameters and the environmental impact factors of the steel slag asphalt mixture interface sample comprises the steps of: for the first oxide with the mass ratio larger than a first threshold value, acquiring a corresponding first oxide parameter by adopting normal distribution simulation; For a second oxide with the mass ratio not larger than a first threshold value, acquiring a corresponding second oxide parameter by adopting gamma distribution simulation; and normalizing the first oxide parameter and the second oxide parameter to obtain the chemical component parameter.
3. The method for predicting the adhesion work of a steel slag asphalt interface based on a physical neural network according to claim 1, wherein the step of obtaining the chemical composition parameters and the environmental impact factors of the steel slag asphalt mixture interface sample comprises the steps of: acquiring the load cycle times by adopting lognormal distribution simulation, acquiring the stress amplitude by adopting Weibull distribution simulation, acquiring the temperature field by adopting sine function superposition random noise, and acquiring the water content by adopting Beta distribution; And acquiring time parameters in the space-time parameters by adopting an exponential distribution model, and acquiring the space parameters in the space-time parameters based on a crack propagation theory of fracture mechanics.
4. The method for predicting the interfacial adhesion work of steel slag asphalt based on a physical neural network according to claim 1, wherein the original neural network prediction model comprises a main prediction network and an auxiliary state network; the physical equation corresponding to the main prediction network comprises a first mapping relation of interface adhesion work, the diffusion term, the reaction source term, the multi-mechanism dissipation term and evolution time; The physical equation corresponding to the auxiliary state network comprises a second mapping relation between the state variable and evolution time, wherein the state variable comprises interface coverage and crack density, the evolution of the interface coverage represents dynamic balance between reaction generation and environmental attenuation, and the change of the crack density represents dynamic balance between crack expansion and product crack stopping.
5. The method for predicting the adhesion work of a steel slag asphalt interface based on a physical neural network according to claim 4, wherein the creating an original neural network prediction model based on a diffusion term, a reaction source term, a multi-mechanism dissipation term and a state variable corresponding to the space-time evolution of the adhesion work comprises: And limiting the state variable in the first mapping relation by using the second mapping relation to obtain an original neural network prediction model.
6. The method for predicting the interfacial work of adhesion of steel slag asphalt based on a physical neural network according to claim 1, wherein the step of inputting the chemical component parameters and the environmental impact factors into the original neural network prediction model and generating the interfacial work of adhesion prediction parameters in combination with physical constraint terms comprises the steps of: determining weights of all constraint items in the physical constraint items based on the current physical background; Weighting the corresponding constraint items based on the weights to obtain total constraint items; And optimizing the original neural network prediction model by using the total constraint term, and inputting the chemical component parameters and the environmental impact factors into the optimized original neural network prediction model to generate interface adhesion work prediction parameters.
7. The method for predicting the interfacial adhesion work of steel slag asphalt based on a physical neural network according to claim 6, wherein training the original neural network prediction model by using the interfacial adhesion work prediction parameter and a gradient descent algorithm to obtain a target neural network prediction model comprises: And in the training process, gradually reducing the weight of the data loss, and increasing the PDE residual loss and the state equation loss until the loss function of the original neural network prediction model is smaller than a second threshold value, so as to obtain a target neural network prediction model.
8. A steel slag asphalt interface adhesion work prediction system based on a physical neural network is characterized by comprising a parameter acquisition module, a model creation module, a model training module and a parameter prediction module, wherein, The parameter acquisition module is configured to acquire chemical component parameters and environmental impact factors of the steel slag asphalt mixture interface sample, wherein the chemical component parameters comprise various oxide parameters, and the environmental impact factors comprise load cycle times, stress amplitude, temperature field, moisture content and space-time parameters related to material performance evolution; The model creation module is configured to create an original neural network prediction model based on diffusion items, reaction source items, multi-mechanism dissipation items and state variables corresponding to the space-time evolution of the adhesion work; The model training module is configured to input the chemical component parameters and the environmental impact factors into the original neural network prediction model, and generate interface adhesion work prediction parameters by combining physical constraint items; training the original neural network prediction model by using the interface adhesion work prediction parameter and the gradient descent algorithm to obtain a target neural network prediction model, wherein the physical constraint terms comprise data loss, PDE residual loss, state equation loss and boundary condition loss; the parameter prediction module is configured to input chemical component parameters and environmental impact factors of the material to be detected into the target neural network prediction model to obtain target prediction parameters.
9. An electronic device comprising a processor and a memory, the memory having a stored computer program, wherein the computer program when executed by the processor implements the physical neural network-based steel slag asphalt interface work adhesion prediction method of any one of claims 1 to 7.
10. A non-transitory computer storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the physical neural network-based steel slag asphalt interface work of adhesion prediction method according to any one of claims 1 to 7.

Description

Method for predicting adhesion work of steel slag asphalt interface based on physical neural network Technical Field The invention relates to the technical field of deep learning, in particular to a steel slag asphalt interface adhesion work prediction method based on a physical neural network. Background The steel slag asphalt mixture is used as a green building material, is beneficial to reducing solid waste in the metallurgical industry when used in road engineering, and improves the skid resistance and durability of the road surface. However, in the actual use process, under the long-term effects of running load and environmental factors, diseases such as stripping and loosening between an asphalt film and steel slag aggregates easily occur, and the integrity and the service life of a pavement structure are seriously affected. The interface bonding area of the steel slag and the asphalt is the weakest link in the mixture system, and the strength of the interface adhesion performance is directly related to the overall stability and the damage resistance of the mixture. Therefore, the method accurately predicts and scientifically evaluates the adhesion performance of the steel slag-asphalt interface, becomes a key technical link for optimizing the material design, preventing early damage and guaranteeing the long-term service performance of the pavement, and has important research value and engineering application significance. Currently, the evaluation of the interfacial work of adhesion between steel slag and asphalt mainly depends on empirical formulas or direct laboratory measurements based on specific conditions. The method has the advantages of high economic cost and long test period, and has limited prediction precision due to simplifying assumptions and experimental errors, so that the real interface behavior under the action of multi-field coupling is difficult to comprehensively reflect. Although data-driven neural networks are used for prediction, the model is usually a pure 'black box' model, extrapolation and generalization capability are insufficient, physical mechanisms of multifactor coupling influences such as complex chemical components of steel slag, temperature time-varying effects and the like are difficult to accurately represent, and the prediction accuracy of adhesion performance is limited. Disclosure of Invention In view of the above, the invention provides a steel slag asphalt interface adhesion work prediction method based on a physical neural network. The technical scheme of the invention is realized in such a way that the first aspect of the invention provides a steel slag asphalt interface adhesion work prediction method based on a physical neural network, which comprises the following steps: The method comprises the steps of obtaining chemical component parameters and environmental impact factors of a steel slag asphalt mixture interface sample, wherein the chemical component parameters comprise various oxide parameters, and the environmental impact factors comprise load cycle times, stress amplitude, temperature field, moisture content and space-time parameters related to material performance evolution; creating an original neural network prediction model based on diffusion items, reaction source items, multi-mechanism dissipation items and state variables corresponding to the space-time evolution of the adhesion work; inputting the chemical component parameters and the environmental impact factors into the original neural network prediction model, and generating an interface adhesion work prediction parameter by combining a physical constraint item, wherein the physical constraint item comprises data loss, PDE residual loss, state equation loss and boundary condition loss; And inputting chemical component parameters and environmental impact factors of the material to be detected into the target neural network prediction model to obtain target prediction parameters. On the basis of the technical scheme, preferably, the method for acquiring the chemical component parameters and the environmental impact factors of the steel slag asphalt mixture interface sample comprises the following steps: for the first oxide with the mass ratio larger than a first threshold value, acquiring a corresponding first oxide parameter by adopting normal distribution simulation; For a second oxide with the mass ratio not larger than a first threshold value, acquiring a corresponding second oxide parameter by adopting gamma distribution simulation; and normalizing the first oxide parameter and the second oxide parameter to obtain the chemical component parameter. On the basis of the technical scheme, preferably, the method for acquiring the chemical component parameters and the environmental impact factors of the steel slag asphalt mixture interface sample comprises the following steps: acquiring the load cycle times by adopting lognormal distribution simulation, acquiring the stress amplitude by