基于机器学习的动力学环境预测方法

A new dynamical environment predication method based on machine learning

为了研究建立同一结构在两个不同系统中振动响应之间的映射关系模型的基本问题,以两种边界条件组成的两个系统为例,从其动力学方程出发,建立了基于时域、频域描述的映射关系的力学模型。为求解两个系统的映射关系模型,引入以统计学习理论为基础的机器学习方法——支持向量机法,借助于决策函数可以从一种边界条件预测同一结构在另一种边界条件下结构的响应,并通过实验证明了该方法的有效性。

Some basic problems of the mapping models for structural response in different dynamic environments are discussed in this paper.For a structure under different boundary conditions,the mechanical models of the mapping relationship in time and frequency fields are established.To solve the mapping models,the machine learning method based on statistical learning theory,namely,support vector machine method,is introduced.Using the decision-making function,the structural response under one boundary condition can be predicated with that under the other boundary condition.Furthermore,the validation results are presented.