一种基于神经网络与粒子群算法辨识针刺炭/炭复合材料弹性常数的方法

Inverse method based on neural network and particle swarm optimization for characterizing the C/C material elastic constants

针对传统方法测量炭/炭复合材料弹性常数精度差的问题,提出了一种利用神经网络与粒子群算法结合模态试验辨识材料弹性常数的新方法。针对板件材料,利用模态与弹性常数的关系,确定材料弹性常数范围,进而采用神经网络建立两者之间的关系,最后以试验振动频率为目标函数,利用粒子群算法寻找弹性常数的最优解。分别利用各向同性薄壁板件及正交各向异性薄壁板件对此方法进行验证,获得了理想结果。最后,将其应用于针刺炭/炭复合材料,获得其弹性常数。仿真及试验结果均表明,该方法较传统方法效率大为提高,且结果更具准确性。另外,针对板件的模态试验及优化计算解决了传统方法数据结果离散度大、局部特性明显的问题,使结果更具全局性。

In order to solve the problem of poor precision for measuring C / C composite elastic constant by traditional test method,a new method for obtaining the main elastic constants from modal test was proposed. For plate material,according to the modal test data and the relationship between vibration frequencies and elastic constant,the scope of the elastic constants were estimated preliminarily. Then the neural network model about elastic constants and vibration frequencies was established. At last,use particle swarm optimization to find the optimal solution of the elastic constants. The isotropic and orthotropic thin plates were used to verify this method. and the perfect results were obtained. At last,the elastic constant of needled C / C composite was obtained through the present method. The simulation and experimental results prove that the method is nondestructive,efficient and accurate,and the data of discrete degree is smaller.