摘要
目前备件需求预测的研究在历史数据的选取和预测方法上存在诸多不合理,如缺少数据预处理及与忽视数据与设备的特性之间的关系,需要给予解决。在考虑不同备件之间需求相关性进行预处理的基础上,以某型大型空气压缩机为例,利用BP神经网络方法,对其备件历史需求数量的时间序列数据建立预测模型。最后将预处理后的历史数据输入到神经网络预测模型之中,并将模型的预测结果与未考虑备件之间需求相关性的预测结果进行比较,可以有效解决神经网络的"欠训练"问题,平均偏差率显著降低。
At present, there are some mistakes in choice, pretreatment and forecasting of time series datum of spare parts demand in some researches, such as improper data set, using raw datum indiscriminately and ignoring the relationship between the datum and the equipments＇ characters, which need to be improved. Taking the large-size air compressor as an example, its spare parts historical demand data series were pretreated. Based on this a forecast model of time series demand of spare parts was presented with BP neural networks. In the end, the processed demand time se- ries datum were input into neural networks forecasting model. The forecasting results between raw datum and processed datum, which were using neural networks, were compared . The phenomena of ＂lack-training＂ vanished, and the average deviation rate remarkably reduced.
出处
《机械设计与研究》
CSCD
北大核心
2010年第1期72-76,共5页
Machine Design And Research
基金
863计划资助项目(2007AA04Z189E)
关键词
备件需求
神经网络
设备特性
预测
spare parts demand
neural networks
equipment＇ s character
forecast
