AI and machine learning drive predictive maintenance in power and manufacturing

Artificial intelligence and machine learning are becoming foundational to predictive maintenance (PdM) within the power and manufacturing sectors. In the power industry, companies such as Ørsted, Florida Power & Light, and National Grid are utilizing AI and ML to analyze sensor data and operational history to detect anomalies and predict failure probabilities. This approach helps utilities maintain grid stability and optimize maintenance scheduling amidst the variability of renewable energy.

In manufacturing contexts, research into predictive maintenance under limited data conditions compares the effectiveness of deep learning (DL) architectures, such as Long Short-Term Memory (LSTM) models, against classical machine learning (ML) models like Random Forest (RF). Studies involving CNC machines indicate that while LSTM models may offer higher precision and fewer false alarms, Random Forest models can outperform LSTM in terms of AUC, recall, and stable cross-machine generalization when data is scarce. This suggests that classical ML remains a robust, computationally efficient alternative for industrial environments with limited failure data.