In this paper, for the first time, the deep learning technique of the artificial neural network method is used to determine the free vibration parameters of the rectangular bistable composite plates. For this purpose, first, the static and free vibration behaviours of a cross-ply bistable composite plate are studied using analytical, finite element and experimental methods. By comparing them, it is turned out that there is a considerable difference among obtained natural frequencies so that the analytical method is only able to determine the fourth natural frequency and cannot estimate the first three natural frequencies. To solve this problem, the deep neural network is employed to model the modal parameters of the bistable laminate as an explicit mathematical relationship that can be generalized to the other bistable composite plates. This mathematical relation makes it possible to obtain the natural frequencies in each of the stable configurations based on the geometric dimensions of the plate. In the following that, the inverse problem method is considered and the mode management capability is investigated. A fast swarm intelligence algorithm called firefly algorithm is used to optimize the optimization function of the mode management problem. Mode management using the evolutionary algorithm provides the appropriate physical dimensions of the plate according to the scenarios for natural frequencies arrangement. The results are validated by comparing them with those obtained by the finite element method and experiment test, which results show that this method estimates the modal parameters with high accuracy. The method used in this paper can also be applied to determine the modal parameters of other morphing structures.

本文首次将人工神经网络方法的深度学习技术用于确定矩形双稳态复合材料板的自由振动参数。为此，首先，使用解析，有限元和实验方法研究了交叉双稳态复合板的静态和自由振动行为。通过比较它们，结果发现所获得的固有频率之间存在相当大的差异，使得该分析方法仅能够确定第四固有频率，而不能估计前三个固有频率。为了解决这个问题，采用深层神经网络将双稳态层压板的模态参数建模为一个明确的数学关系，该关系可以推广到其他双稳态复合材料板。该数学关系使得可以基于板的几何尺寸获得每个稳定配置中的固有频率。在下文中，考虑了反问题方法并研究了模式管理能力。一种快速群智能算法，称为萤火虫算法，用于优化模式管理问题的优化功能。使用进化算法的模式管理可根据自然频率安排的方案，为板提供适当的物理尺寸。通过与有限元方法和实验测试获得的结果进行比较，验证了结果，结果表明该方法可以高精度估计模态参数。