Home / Current Issue / Paper 1718462
FactoryGuard AI: An IoT-Based Predictive Maintenance Engine for Industrial Robotic Arms Using XGBoost, SMOTE, and SHAP Explainability
Subject area: Science,Engineering and Technology · Area of research: Robotics, ML
DOI: https://doi.org/10.64388/IREV9I11-1718462
Abstract
Unplanned equipment failure in large-scale manufacturing facilities inflicts enormous operational and financial consequences, with downtime costs estimated at $10,000 per hour per production line. This paper presents FactoryGuard AI, a comprehensive IoT-driven predictive maintenance engine designed to forecast binary failure events of industrial robotic arms with a 24-hour lead time. The system ingests streaming sensor telemetry—vibration, temperature, and pressure—from a fleet of 500 critical robotic arms and transforms raw signals into rich temporal feature representations through sliding-window statistics and exponential moving averages. The inherent class imbalance of failure events (< 1% positive class prevalence) is addressed via Synthetic Minority Over-sampling Technique (SMOTE) combined with scale-position weighting. A Logistic Regression baseline is established for benchmarking, after which an optimised XGBoost gradient-boosted ensemble is trained, achieving an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 0.9847, a Precision of 0.923, a Recall of 0.917, and an F1-Score of 0.920 on held-out test data. SHapley Additive exPlanations (SHAP) are integrated to deliver local and global model interpretability, enabling maintenance engineers to understand and trust individual failure predictions. The system reduces false negatives by 78.4% compared to the baseline, translating to an estimated annual saving of $4.2 million in avoided downtime for a 500-robot facility.
Keywords
Predictive Maintenance, IoT Sensor Fusion, XGBoost, SMOTE, SHAP, Explainable AI, Rolling Window Features, Binary Classification, Industrial IoT, Robotic Arm Monitoring
How to cite this paper
@article{1718462,
author = {Pratik Joshi, Vrushali Shinde},
title = {FactoryGuard AI: An IoT-Based Predictive Maintenance Engine for Industrial Robotic Arms Using XGBoost, SMOTE, and SHAP Explainability},
journal = {Iconic Research And Engineering Journals},
year = {2026},
volume = {9},
number = {11},
pages = {4755-4766},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1718462.pdf},
abstract = {Unplanned equipment failure in large-scale manufacturing facilities inflicts enormous operational and financial consequences, with downtime costs estimated at $10,000 per hour per production line. This paper presents FactoryGuard AI, a comprehensive IoT-driven predictive maintenance engine designed to forecast binary failure events of industrial robotic arms with a 24-hour lead time. The system ingests streaming sensor telemetry—vibration, temperature, and pressure—from a fleet of 500 critical robotic arms and transforms raw signals into rich temporal feature representations through sliding-window statistics and exponential moving averages. The inherent class imbalance of failure events (< 1% positive class prevalence) is addressed via Synthetic Minority Over-sampling Technique (SMOTE) combined with scale-position weighting. A Logistic Regression baseline is established for benchmarking, after which an optimised XGBoost gradient-boosted ensemble is trained, achieving an Area Under the Receiver Operating Characteristic Curve (AUC-ROC) of 0.9847, a Precision of 0.923, a Recall of 0.917, and an F1-Score of 0.920 on held-out test data. SHapley Additive exPlanations (SHAP) are integrated to deliver local and global model interpretability, enabling maintenance engineers to understand and trust individual failure predictions. The system reduces false negatives by 78.4% compared to the baseline, translating to an estimated annual saving of $4.2 million in avoided downtime for a 500-robot facility.},
keywords = {Predictive Maintenance, IoT Sensor Fusion, XGBoost, SMOTE, SHAP, Explainable AI, Rolling Window Features, Binary Classification, Industrial IoT, Robotic Arm Monitoring},
month = {May},
doi = {https://doi.org/10.64388/IREV9I11-1718462}
}