摘要
火炮的身管寿命基本上决定了整门火炮的寿命,对身管寿命加以预测具有十分重要的意义。本文首先提出了一种基于测试数据的火炮身管寿命的递推预测方案,而后将目前应用较为广泛的BP神经网络用于这种递推预测,为避免网络的“过拟合”,将样本分为训练样本和检验样本提交网络训练,结果发现只有训练次数较大时才可以获得较好的预测效果。在分析了样本的特点后,将BP网络的检验样本思想融入多项式拟合后进行同样的递推预测,发现简单的多项式预测也可以达到较好的效果。
It is of significance to predict gun tube抯 lifespan, which can decide whole gun抯 lifespan basically. This paper first brings out a gun tube抯 lifespan stepwise predicting scheme, then applies BP neural network to this scheme. To avoid network抯 overfiting, this paper divides sampling data into training data and validating data, then submit these data to network, and finds that better predicting result can only be obtained when training steps are sufficient. After sampling data抯 characteristic was analyzed, polynomial fit imported BP network 憇 憊alidating data?thought was applied to same scheme, and which can also obtain better predicting results.
出处
《机械》
2004年第1期10-12,共3页
Machinery