论文部分内容阅读
针对神经网络逆动态补偿策略中基于向后差分的微分估计算法在噪声环境下存在较大的微分估计误差,且提高估计精度会增加硬件成本等缺陷,提出了改进的微分估计算法,并以加速度传感器为例,设计了动态补偿实验装置,给出了基于改进的微分估计算法的动态补偿算法的实现步骤。改进的微分估计算法,采用低通滤波器以去除高频噪声对微分的影响。实验结果验证该方法能明显改进和提高加速度传感器的动态性能,实现对脉冲力的高精度重构。
Aiming at the drawbacks of differential estimation algorithm based on backward difference in inverse compensation of neural network, such as large differential estimation error under noisy environment and increasing the accuracy of estimation, the hardware cost will be increased. An improved differential estimation algorithm is proposed, Sensor as an example, a dynamic compensation experimental device is designed and the steps of the dynamic compensation algorithm based on the improved differential estimation algorithm are given. Improved differential estimation algorithm, the use of low-pass filter to remove the impact of high-frequency noise on the differential. Experimental results show that this method can significantly improve and improve the dynamic performance of the acceleration sensor to achieve high-precision reconstruction of impulsive force.