针对高可靠性、长寿命复杂产品的可靠性评估过程, 在加速寿命退化试验数据的基础上, 提出了一种基于试验数据驱动的自适应智能方法, 并对某型LED灯管的寿命与可靠性进行预测分析。首先, 通过指数模型拟合性能退化曲线, 推算出各组应力条件下的伪失效寿命值; 再将蚁群算法结合BP神经网络等智能算法应用于寿命预测模型的建立, 根据试验证明寿命服从对数正态分布, 且检验寿命必须满足置信度区间范围内; 最后, 预测出正常应力条件下LED灯管的工作寿命。结果表明, 基于蚁群神经网络预测LED灯管寿命的方法, 预测误差较小, 收敛速度快, 能够满足工程要求。

To evaluate the reliability of long life, high reliable and complex products, proposed was an adaptive intelligent method based on the connection of accelerated life test data, and it was used to analyze the life and reliability of LEDs. First, a index model was built to fit the trend of performance degradation, and pseudo failure life was calculated under each stress condition. Then, the ant colony algorithm combined with BP neural network was applied to establish the life predicted model, proving the life obeyes the lognormal distribution, and the test life must meet the range of confidence interval. Finally, the working life of the LED lamp was predicted under the condition of normal stress. The results show that the method for forecasting LED lamps life based on ant colony neural network presents small prediction errors and faster convergence rate, satisfying the engineering requirement.