信息缺失的航空发动机传感器数据重构

为处理传感器数据缺失问题,利用子空间表示系统演化特征,提出了基于极化增量矩阵填充(PIMC)的航空发动机传感器数据的在线重构模型。该模型通过历史数据获得当前的数据特征表示,并用新增的数据不断更新子空间以跟踪并表示数据发展特征。将本文模型用于仿真数据进行验证,重构结果和无噪数据的归一化均方误差(MSE)均小于1×10-5,实验结果显示本文模型对于航空发动机传感器数据重构有很好的应用价值,对缺失数据和噪声是鲁棒的。 

Aeroengine sensor data reconstruction with missing data

Aiming at handling incomplete sensor data, we propose an online-reconstruction model based on the polar incremental matrix completion (PIMC) algorithm for aeroengine sensor data, which can represent the evolving features of system by subspace. The model extracts the current data feature from the history data and updates the subspace to track the evolving features via new data. The proposed model was validated and compared on two simulated datasets and the normalized mean square errors (MSE) between the reconstruction by PIMC and the ground truth are all less than 1×10^-5. The experimental results show that the proposed model is practical for aeroengine sensor data reconstruction, which is robust to missing data and noise.