Corrosive Predictive Analytics System
Petronas
2024
Background
CPAS was developed as an internal predictive maintenance system for Petronas, designed to monitor, analyze, and forecast corrosion rates across critical assets in real time. The project aimed to help engineering teams detect early degradation, understand long term corrosion behavior, and prioritize maintenance actions based on data driven insights to support operational reliability.
Responsibilities
During my internship, I worked on two major data driven initiatives: the Discrete Acoustic Sensing (DAS) Dataset project and CPAS. For DAS, I processed large HDF5 datasets, applied a bounding box algorithm to detect and isolate triggered acoustic events, validated extracted signals, visualized waveform outputs, and automated the entire pipeline to improve processing efficiency. For CPAS, I focused on understanding the system’s ERD structure, validating data integrity, performing comprehensive preprocessing, and developing machine learning models to forecast corrosion patterns and potential failure points, with deployment integrated into a real time monitoring ecosystem using Python, ML libraries, and Grafana.
Key Features
The system delivered automated corrosion forecasting, anomaly detection, and real time monitoring capabilities, supported by a modeling workflow that transformed raw operational data into predictive indicators. Visual dashboards in Grafana showcased corrosion trends, risk levels, and asset conditions, enabling engineers to monitor system health and derive actionable insights through a clear and intuitive interface.
Result
The implementation strengthened predictive maintenance capabilities by improving accuracy in identifying early corrosion signals, reducing manual analysis time, and providing stakeholders with real time, data driven visibility of asset health. The combined work in DAS and CPAS supported faster decision making and contributed to a more reliable and scalable monitoring framework for critical infrastructure.
