Privacy-Aware CNC Predictive Maintenance

This repository contains the research paper “Federated Multi-Task Learning Framework for Privacy-Preserving Predictive Maintenance in CNC Machines.” The study proposes a novel Personalized Federated Multi-Task Learning (PFMTL) framework designed to enable privacy-aware predictive maintenance in distributed industrial environments. Predictive maintenance has become an essential component of modern smart manufacturing systems. However, traditional centralized machine learning approaches require aggregating operational data from multiple machines or factories into a central server, which raises concerns related to data privacy, communication overhead, and scalability. To address these challenges, this work integrates Federated Learning (FL) with Multi-Task Deep Learning (MTL) to create a decentralized learning framework that allows multiple clients to collaboratively train predictive models without sharing raw data. The proposed system enables CNC machines or manufacturing units to train local deep learning models using their private sensor data, while only sharing encrypted model parameters with a central server. A personalized federated aggregation strategy is implemented to improve model adaptability across heterogeneous industrial environments. The framework is evaluated using the Milling Machine Dataset (Matzka, 2020) containing 10,000 data samples and operational features such as temperature, torque, rotational speed, and tool wear indicators. Experimental results demonstrate that the proposed federated framework achieves 93.2% predictive accuracy, closely matching centralized learning performance (94.1%) while reducing communication costs by approximately 92%. The model also shows strong performance across multiple failure types including Tool Wear Failure, Heat Dissipation Failure, Power Failure, Overstrain Failure, and Random Failure. Additionally, explainable AI techniques such as SHAP analysis are used to interpret feature importance and support decision-making in industrial maintenance operations. Overall, this work presents a scalable, privacy-preserving, and industry-ready predictive maintenance solution suitable for distributed manufacturing systems and Industrial IoT environments. The framework contributes to advancing Industry 4.0 technologies by enabling collaborative machine learning while protecting sensitive operational data.