Structural Design Of Insulating Fibre Section Based On Strange Attractor And Reinforcement Learning

The structure of the material affects the properties of the material,and the geometric shape of the artificial design of the material can make the material as a whole have properties that surpass the base material itself.Traditional material geometric design methods include periodic unit-based design methods,including European geometric coding methods,mathematical surface coding methods,flexible material coding methods based on lattice plus functions,fractal self-similar flexible material coding methods,and Kirigami encoding methods,and methods based on topology optimization.Periodic coding generally requires a lot of pre-processing,and the coding structure design space is small;while the topology optimization method has a large design space,and searching in the design space requires a lot of computing power.This paper proposes an optimization framework for structural parametric design and reinforcement learning based on fractal geometry for the above problems,and verifies the feasibility of the method under the condition of fiber section insulation.Main contents are included as follows:The research is based on strange attractor multi-level hole structure coding method and reinforcement learning optimization algorithm.First,a multi-level pore structure encoding method based on the strange attractor generated by the iterative function system is proposed.The corresponding structure has a regular arrangement of the multi-level pore structure,and the pore structure forms a kind of network structure as a whole through the regular arrangement.Considering that the porous structure mode encoded by the iterative function is too strong,this paper tries to disassemble the iterative function and extend the rotation angle and translation function respectively.Based on the above-mentioned parametrically encoded multi-level pore structure and the geometric information extracted by the Alpha shapes algorithm,two physical parameters affecting the thermal conductivity of the structure are proposed,namely the porosity of the structure and the equivalent radius of the structure pores.The porosity of the structure affects the thermal insulation performance of the structure by affecting the proportion of air in the structure,and the equivalent radius of the structural pores affects the thermal insulation performance of the structure by affecting the distribution of pores in the structure.The finite element method is used to verify the influence of the structural porosity on the thermal conductivity of the structure,and the effect of the equivalent radius of the pore on the pore distribution is given through mathematical proofs,and then the influence of the equivalent radius of the pore on the thermal insulation performance of the structure is established.Finally,a parametric design framework based on the singular attractor is established,and the design parameters are optimized for the proposed thermal insulation physical parameters.Considering that the singular attractor generated by the nonlinear iterative function has strong nonlinearity relative to its design parameters,and the design parameter space is in the continuous domain,the DDPG algorithm in reinforcement learning is used to optimize the structure,and finally the pore distribution of the optimized structure and the porosity verifies the feasibility of the method.