WellWatch

What it is

A Python system that watches produced-water chemistry and injection-well time series for scale, corrosion, souring and injectivity anomalies. A deterministic chemistry engine computes saturation indices, an Isolation Forest flags injection anomalies, and an LLM layer turns both into field recommendations.

The problem

Operators find scale and injectivity problems after equipment damage because water chemistry and injection data are not watched together.

How it works

1. Layer 1: deterministic chemistry engine for saturation indices and corrosion, with no ML; the build plan swaps the Davies model for Pitzer to handle high-TDS brines
2. Layer 2: Isolation Forest anomaly detector on injection monitoring series, exercised on synthetic wells
3. Layer 3: Claude interpreter merges both layers into deduplicated recommended actions, with a template fallback
4. FastAPI service, CLI, Texas RRC data fetch/prepare scripts and a pytest suite

Stack

Disciplines

For your business

Industrial monitoring that pairs deterministic domain physics with anomaly detection and LLM-written field guidance.