Predictive maintenance is now a pressing necessity in defense and aerospace systems to maintain reliability, safety and efficiency of operations. The conventional maintenance plans are not always effective in responding to the real-time failures because they are usually reactive or periodic. The current paper suggests a predictive maintenance implementation using an AI-based cloud-edge hybrid architecture, which combines edge computing (low-latency processing) with cloud computing (advanced analytics). It uses machine learning methods to make anomaly detection, failure prediction, and Remaining Useful Life (RUL) forecasts with sensor-generated time-series data. An active task scheduler is used to optimize the distribution of workload between edge and cloud layers. Experimental analysis with simulated data show that the proposed model is more accurate, has lower latency, high scalability, and lower computational costs than the conventional models of LSTM, SVM, Random Forest, and CNN. The findings point to the suitability of the hybrid architecture in facilitating real-time, scalable, and smart maintenance solution to mission-dependent aerospace and defense systems.
Pawan Kalyan Jonnalagadda (Mon,) studied this question.