组合导航系统软件可靠性的预测模型

针对组合导航系统软件可靠性长期预测的复杂性,建模时在引入神经网络技术的基础上,提出了滚动式预测方法.该方法考虑了多种不确定因素对预测结果的影响,不断加入新预测的故障数据,克服了单纯依靠神经网络技术预测偏差较大的缺陷.仿真结果表明,该模型预测精度高,可用于实践中指导软件测试.

Long time prediction of software reliability in integrated navigation system

According to the complexity of software reliability long-term forecast in the integrated navigation, neural network technology is applied to the module. The paper offers a method of rolling prediction. The method involves the comprehensive consideration of a variety of factors of uncertainty impacting on the outcome. New forecast data as a sample model are joined constantly with the change of time. The method overcomes the defects of large prediction error due to simply relying on neural network technology. Simulation results show that the model established by the method shows a higher accuracy and can find a better use in guiding practical software test.