Cooler WHRS Optimization Kiln Elbow Jam Prediction Performance Reporting & Visualization
Key Tools: TensorFlow, Transformers, Time Series Modeling, Root Cause Analysis
- Led a 4-member team to optimize the Cooler Waste Heat Recovery System (WHRS), boosting power generation by 1.5 MW through advanced time series analysis.
- Applied transformer-based deep learning models using TensorFlow to forecast temperature and pressure trends, identifying latent inefficiencies in heat recovery.
- Conducted root cause analysis of multi-source sensor data to uncover operational bottlenecks and recommend actionable improvements.
- Collaborated with domain experts to deploy insights, ensuring measurable gains in system efficiency and stability.
Key Tools: Python, Pandas, Scikit-learn, Time Series Forecasting
- Developed and maintained a predictive time series model for early detection of coating formation in kiln elbows, using historical process data.
- Reduced operational disruptions and maintenance costs by forecasting precursors to equipment failures.
- Integrated model predictions into operations through real-time alerts and preventive actions, improving kiln availability and planning efficiency.
- Aligned closely with business stakeholders to translate operational needs into data science solutions.
Key Tools: Pandas, Matplotlib, Seaborn, Fault Tree Analysis
- Incorporated fault tree analysis to systematically trace root causes of failures and visualize critical paths.
- Preprocessed and transformed high-volume time series data into actionable insights for technical and management teams.
- Enhanced cross-departmental decision-making by delivering clear, data-backed visualizations and summaries.