Position-Agnostic Aeroengine Intershaft Bearing Fault Diagnosis via Condition-Guided Multitask Learning

Real-time monitoring and diagnosis of bearing fault conditions are crucial for the stable operation of aero-engines. Recently, bearing fault diagnosis methods based on deep learning have greatly improved the accuracy and efficiency of fault diagnosis. However, they also face some challenges. Firstly, existing fault diagnosis models typically require an input of one or more vibration signals from fixed positions, lacking flexibility in practical applications. Moreover, auxiliary information such as engine operating conditions and sensor positions is often over-looked, especially for aero-engines with more complex operating conditions. To address the aforementioned challenges, this paper investigates inter-shaft bearing fault data collected from aero-engines, verifying the limitations of using fixed-position signal inputs for fault diagnosis through comprehensive experimental analysis. Based on this foundation, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">position-agnostic fault diagnosis</i> task is proposed, which inputs vibration signals from any position into the fault model, providing greater flexibility for practical applications. Furthermore, we designed a new condition-guided multi-task supervision model called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MTS-Net</i>, which fully utilizes auxiliary information such as rotational speed and sensor positions contained in the dataset, enhancing the generalization ability and prediction accuracy of the fault diagnosis model. Extensive experiments validated the effectiveness of the proposed method, particularly in limited-sample scenarios, where the model’s fault diagnosis classification accuracy can be improved by 6.5%.