Distributed fault diagnosis for data reliability in operational status monitoring of large mining electric shovels

Large mining electric shovels are critical to open-pit mining, yet unreliable data hinder their operational monitoring in harsh environments and inefficient centralised fault diagnosis. This paper proposes a distributed fault diagnosis framework to enhance data reliability and diagnostic efficiency. The framework first deploys an edge computing-based architecture to enable local data processing and avoid single-point failures. It then incorporates a two-stage module at the edge to enhance data reliability through adaptive denoising and sensor fault self-checking. Finally, a collaborative diagnosis model is developed by integrating a lightweight convolutional neural network with federated learning, allowing multiple units to train a shared model without exchanging raw data. Field experiments demonstrate that the proposed method significantly outperforms traditional centralised approaches, increasing the signal-to-noise ratio by 28.3%, reducing diagnosis latency by 42.1%, and achieving a fault diagnosis accuracy of 96.7%.