【摘要】 弹体侵彻目标钢板实验是用来测试侵彻过程中弹体在出炮口,飞行过程以及侵彻过程的数据,从而为提高弹体的性能提供有效依据。在实验中会产生较高谐振频率,需要高g值加速度传感器固有频率较高才可无失真复现信号。校准与试验发现传感器动态工作频带与静态频带标注有一定差距,导致现有传感器不能满足测试系统测量侵彻硬目标信号的需要。为提高传感器动态性能,展宽传感器动态工作频带,改善测试系统精度,提出一种改进型径向基函数(RBF)神经网络算法对传感器动态带宽进行补偿。运用将梯度下降法与改进型变尺度法(L-BFGS)融合的方式,对RBF神经网络权值训练。实验表明,上述算法能够有效实现对传感器工作频带的动态补偿,从而准确预测对侵彻目标的冲击力。更多还原

【Abstract】 The project is to test the data of the projectile in the flight process and the penetration process,and provides the effective basis for improving the performance of the projectile. The experiment can produce a higher resonant frequency,this time need an acceleration sensor with high g value and high inherent frequency without distortion signal. Thought calibration and test,it was found that the dynamic frequency band of the sensor has a certain gap with the static frequency band,which leads to the fact that the existing sensor cannot meet the test system to measure the penetration of the hard target signal. In order to improve the dynamic performance of the sensor,widen the dynamic working frequency band of the sensor and improve the accuracy of the test system,an improved RBF neural network algorithm was proposed to compensate the dynamic bandwidth of the sensor. The training of RBF neural network weights was realized by combining the gradient descent method with the improved variable scale method( L-BFGS). The experimental results show that the proposed algorithm can effectively realize the dynamic motion of the sensor make up.更多还原