【摘要】 剩余寿命作为预测性维修的重要支撑,智能预测能及时识别出复杂系统寿命变化规律,准确反映不同工况对剩余寿命的影响。针对复杂系统运行工况与高维度、多尺度时序数据,构建多层LSTM预测模型,防止了梯度消失,能够提取不同工况下时序数据的深层次关联性抽象特征。利用Dropout方法减少预测模型过拟合,针对预测模型的不确定性,设计误差得分函数以评估超前与滞后预测。以航空发动机为例,验证了所提出的智能预测模型的有效性。更多还原

【Abstract】 Remaining useful life is an important support for predictive maintenance. Intelligent prediction can identify the life change law of complex systems in time. The impact of different working conditions RUL was reflected. For complex system operating conditions and high-dimensional, multi-scale time series data, a multi-layer LSTM prediction model was constructed. It can prevent vanishing or exploding gradient. The correlation features of time-series data under different working conditions were extracted. Using Dropout method can reduce overfitting of the predictive model. Aiming at the uncertainty of the prediction model, an error score function was designed to evaluate the leading and lagging predictions. Taking an aero-engine as an example, the effectiveness of the proposed intelligent prediction model was verified.更多还原