SeaChem: Predictive Modeling for Marine Corrosion & Environmental Monitoring

Project Overview

SeaChem is an interdisciplinary research initiative aimed at developing predictive models for maritime corrosion in combination with environmental monitoring and mitigation strategies. The project brings together oceanographers, materials scientists, corrosion engineers, and data scientists to understand and forecast corrosion risks for marine infrastructure, vessels, and submerged structures.

Key aims:

  • Monitor key environmental variables (salinity, temperature, pH, dissolved oxygen, pollutant concentrations) in coastal and open sea zones.

  • Collect corrosion metrics from test coupons and sensor‐instrumented structures.

  • Build data pipelines and analytics to fuse multi-source observations.

  • Develop machine learning / statistical models to predict corrosion progression under varying marine conditions.

  • Provide actionable insights and decision support for maintenance, inspection scheduling, and material choice.

Research Questions

  1. Which environmental factors drive corrosion rates in marine environments, and how do they interact?

  2. Can we forecast future corrosion progression at specific locations and depths with useful lead time?

  3. How accurate are predictive models across different materials and environmental regimes?

  4. How can uncertainty in predictions be quantified and communicated for engineering decisions?

Methodology & Approach

Data Acquisition & Monitoring

  • Deploy sensor arrays near test structures (e.g. steel coupons) to collect time series of salinity, temperature, pH, dissolved oxygen, chlorides, etc.

  • Periodic retrieval of the test coupons to measure weight loss, surface degradation, electrochemical measurements, microscopy, etc.

  • Use auxiliary data sources (marine weather forecasting, oceanographic databases, remote sensing) to enrich the dataset.

Data Integration & Preprocessing

  • Time alignment, interpolation, gap filling, and normalization across heterogeneous sources.

  • Feature engineering: moving averages, lagged variables, rate-of-change, derived indices (e.g. corrosion potential indices).

  • Outlier detection and filtering to eliminate noise or sensor errors.

Predictive Modeling

  • Use regression, time-series forecasting (e.g. ARIMA, state-space models), and machine learning methods (e.g. random forest, gradient boosting, neural networks) to model corrosion rate (e.g. mass loss per time) as a function of environmental predictors.

  • Explore hybrid physics-informed models where expert corrosion theory constrains the learned models.

  • Perform cross-validation, hold-out testing, and spatial/temporal generalization experiments.

  • Quantify prediction uncertainty (prediction intervals, Bayesian methods, ensemble variance).

Evaluation & Validation

  • Compare predicted corrosion progression curves vs. measured ones from withheld test coupons.

  • Evaluate performance metrics (RMSE, MAE, correlation, bias) across materials, depths, and environmental regimes.

  • Conduct sensitivity and ablation studies: which predictors matter most; how does removal of certain data streams affect forecasts.

Key Deliverables

  • A curated multi-source marine corrosion dataset, ready for sharing with the research community.

  • Trained predictive models of corrosion progression, along with code and a reproducible forecasting pipeline.

  • A web dashboard or decision-support tool for stakeholders (e.g. marine infrastructure operators), to visualize predicted corrosion risk and schedule inspections.

  • Recommendations for best practices in monitoring design, sensor deployment, and data usage.

  • Publications, open-source code, and documentation.

Project Impact & Innovation

  • Provides proactive maintenance strategies for marine structures, reducing unscheduled downtime and costly repairs.

  • Advances the novel integration of environmental monitoring and corrosion science via predictive modeling.

  • Bridges data science and materials engineering in a marine context.

  • Fills a gap: while corrosion has been well-studied in static lab conditions, real-world predictive forecasting in dynamic marine environments remains underdeveloped.

  • Supports sustainable maritime infrastructure and resilience.