【摘要】 自动供输弹装置是火炮上故障发生率较高的子系统之一,因而有必要进行自动供输弹装置故障预测研究。自动供输弹装置是一种机-电-液一体化的机器人系统,既有实现动作的机械部分,即弹药装填和补给装置,也有控制机构动作顺序的控制部分。根据供输弹装置的组成,其故障诊断的BP网络输入层采用9个神经元,隐含层节点数有8个,隐函数选取sigmoid函数,输出层有9个单元。故障预测网络的训练包括火炮自动供输弹装置特征样本输入方式、测试方法、最大允许迭代次数和训练终止的允许误差等的确定。实时测取的样本验证,该预测模型合理。更多还原

【Abstract】 Automatic loader is one of subsystems that have higher failure rate,it is necessary to perform fault prediction research of automatic loader. Automatic loader is a kind of mechanic, electric and hydraulic integrated device, it is not only equipped with mechanic component, such as ammunition loading and feeding devices, but also equipped with control component that controls action order of mechanism. According to automatic loader composition, BP network input layer of fault diagnosis makes use of 9 neural elements, and has 9 units. Training of fault prediction network includes input mode of feature example, test method, maximum permission iterative number of times and permission error at training end etc. Example verification in real time test showed that the prediction model is a kind of reasonable proposal.更多还原