The imputation of missing data in aeroengines based on a new trajectory similarity evaluation method

Acquiring high-quality data is essential for aeroengine health management, ensuring engine safety and reliability. However, during data acquisition and storage, some data may be lost due to sensor damage, server failures, and human factors. A new trajectory similarity evaluation method is proposed to impute aeroengine missing data. First, a new algorithm called variable-weight parameter warping (VPW) is introduced to optimize cross-sectional alignment, along with a new filter called variable window filtering that accounts for parameter dispersion from individual differences. These two algorithms help identify the segment closest to the missing data on each engine's sample library to create a similar trajectory library. Next, a novel trajectory similarity evaluation metric is proposed, which comprehensively measures the similarity from multiple dimensions, including VPW value, window length and window position. Finally, the proposed method's effectiveness is validated using the turbofan engine simulation dataset. The results show that not only the imputed data accurately fit the original data, but also the RMSE and Score of the RUL prediction result with the imputed data as input are 12.71 and 325, closely matching the RMSE of the original data at 12.76 and Score at 324, showing the same functionality as the original data.