基于Gram-Schmidt回归的军械器材储存期限测算方法

Shelf Life Prediction Modeling for Ordnance Equipment Based on Gram-Schmidt Regression

针对军械器材储存期限测算任务量大、时间紧迫,批量建模时与环境因素相关的回归项尤其是非线性项难以确定,因素间的多重共线性难以消除等问题,提出了一种基于Gram-Schmidt回归的军械器材储存期限测算方法。该方法可在众多备选非线性项中依次找到关键的影响因素,并利用消减投影分量的方法消除多重共线性的不良影响;利用该方法可批量构建储存期限测算模型,且测算流程规范、统一,易于编程实现。最后以橡胶类军械器材为例对该方法进行了验证,结果表明:批量建模选出的非线性项与Dakin寿命方程相符,测算的常温下天然橡胶的储存期限值与出厂值相符。

In order to overcome the difficulty of selecting the suitable regression items especially the non-linear regression items and eliminating the influence of multicollinearity when conducting mass modeling tasks,an ordnance equipment shelf life calculation method using Gram-Schmidt regression is put forward. This method can select the critical modeling items in all the available items in turn,and the influence of multicollinearity can be eliminated by deducting the orthogonal components.Mass of shelf life prediction models can be developed in this method,and the process is very standardized and easy programming.Fi-nally,an example of predicting rubber type of ordnance equipment is taken to test and verify the method. Results show that the nonlinear terms of regression model selected by modeling process is consistent with Dakin equation,and the predicted values of natural rubber shelf life are coordinated with factory settings.