基于蚁群BP神经网路算法的六维力传感器解耦研究

针对传统BP神经网络在六维力传感器解耦训练过程中，由于其初始参数的选取不确定性导致神经网络出现震 荡、收敛速度缓慢和陷入局部极值等问题，提出一种基于蚁群BP神经网络算法并应用于六维力传感器解耦研究。该算 法利用蚁群算法在种群寻优方面的优越性，通过局部和全局信息素更新相结合的方式更新信息素，提高蚁群算法搜索的 效率，为BP神经网络提供一组最优的训练初始参数，网络收敛速度得到很大地提高，同时局部极值和震荡等缺点也有一 定的改善。实验仿真结果表明，在六维力传感器神经网络模型训练过程中，达到同样的目标误差，基于蚁群BP神经网络 算法的迭代次数Ⅳ比传统算法少50%，运行时间r快60%。这说明蚁群BP种经网络算法在六维力传感器解耦研究中 有着很好的应用效果。

Study on Decoupling Algorithms for Six-Axis Force Sensor

ln view of the traditional BP neural network in the process of six-axis force sensor decoupling training,due to the selection of initial parameters uncertainty prompted a concussion. slow convergence speed and neural network into local extreme problems. This paper proposed an ant algorithm to optimize BP neural network and was applied to study six- axis force sensor decoupling. The algorithm took advantage of ant colony algorithm in terms of population optimization, through the combination of local and global pheromone way to update pheromone ,improved the efficiency of ant colony algorithm to search for a set of optimal initial parameters for the BP neural network.the network convergence speed was greatly improved, while the local extreme and shock had certain improvement. The experimental simulation results show that in the training process of six-axis force sensor. to achieve the same target error, ant colony BP neural network algorithm is 50% less than traditional BP neural network.the running time is 60% faster. This shows that BP neural network based on ant colony algorithm in the study of six-axis force sensor decoupling has a good application effect.