01 // Industrial AI · Internship 2026 · Private Project

Compressor Condition Monitoring

A physics-grounded, three-plant condition-monitoring program that turns reconciled historian data into explainable fouling evidence, plant-specific instability review queues, and carefully gated operator diagnostics.

3Plant-Specific Contracts
4,000+Longest Daily History
3,879Reproduced DFI Rows
17/17Latest Validation Checks
Private source · De-identified case study See Engineering Decisions
Overview

The original brief sounded simple: show compressor degradation and predict maintenance. The difficult part was deciding which conclusions the data could actually support. I worked across reconciled thermodynamics, auxiliary mechanical channels, maintenance references, Excel lineage, research scripts, APIs, and operator-facing views.

Instead of optimizing a single opaque score, I built an evidence hierarchy. Fouling uses variables with direct thermodynamic meaning. Surge-oriented work uses different evidence contracts for each plant. Forecasts, thresholds, and maintenance resets are promoted only when their validation gates pass.

  • My role: research design, data forensics, feature engineering, validation, backend integration, and dashboard implementation.
  • Scientific boundary: daily historian exports can rank review candidates; they do not confirm aerodynamic surge without controller, map, or high-frequency evidence.
  • Privacy boundary: this page contains aggregate methodology only. Raw plant data, operational tags, and internal source artifacts are not published.
System Architecture

From Historian Data to Reviewable Evidence

01
Historian Exports
Process · mechanical · maintenance context
02
Reconciliation
Thermodynamic variables · quality gates
03
Evidence Engines
DFI · review queues · event context
04
Promotion Gates
Holdout · causality · regression tests
05
Operator UI
Trace · evidence · limitations · actions
DFI
Physics-grounded fouling

Plant 2 Candidate B combines reconciled polytropic-efficiency loss and temperature-rise penalties across all four stages, with prior-only operating envelopes and fixed development scaling.

QA
Claims are part of the code contract

Forecast enablement, maintenance labels, threshold use, and surge wording are protected by configuration, frozen artifacts, validators, and regression tests.

Research Decisions

The Most Useful Results Were Not Always Promotions

01

Rebuilt Plant 2 fouling from first principles

Audited each reconciled variable for monotonic degradation, stability, noise, and maintenance recovery. The selected shadow formulation reproduced 3,879 historical rows with maximum absolute difference below 1e-10.

Candidate B = stage efficiency loss + temperature-rise penalty
02

Rejected forecasts that looked plausible but failed gates

Rolling-origin tests covered classical, tree-based, physics-guided, and KNN approaches. Holt damped trend remained the numerical winner, but the evidence was too weak for maintenance decisions because the holdout contained only one effective maintenance event.

Decision: preserve monitoring; keep forecast policy disabled
03

Separated anomaly context from surge confirmation

A full historian audit found different observability at each plant. I replaced one shared detector with three plant-specific review contracts and renamed the continuous context layer so it did not imply physical distance to an OEM surge line.

Review candidate ≠ confirmed surge event
04

Cross-verified newly supplied wash-oil histories

Causal 21-day baselines, multi-wheel consensus, persistent restoration, and independent DFI recovery identified three complete dosing cycles. Only one cycle had sufficient thermodynamic corroboration, so it was promoted as event context without triggering an automatic DFI reset.

3 complete cycles → 1 corroborated context event → 0 automatic resets
De-identified Evidence

The Interface Shows the Signal and Its Scientific Boundary

These retained research figures show the two different analytical roles. Candidate B visualizes thermodynamic fouling evidence and recovery behavior. The instability index ranks dates for engineering review; it does not confirm aerodynamic surge.

Candidate B fouling history with watch and high thresholds, recorded references, and independent recovery boundaries
Fouling evidence. Candidate B is a continuous physics-grounded shadow index. Recorded maintenance references and independently detected recovery boundaries remain separate evidence layers.
Daily plant-specific instability review index with moderate and strong review floors
Instability review evidence. Daily scores identify dates that deserve review. The plot deliberately avoids presenting those dates as confirmed surge events.
Plant-Specific Evidence

One Platform, Three Scientific Contracts

Plant Available evidence Legitimate output Claim boundary
Plant 1 Flow, pressure ratio, speed, four vibration channels, two thrust probes Mechanical-aerodynamic review ranking Engineering review only
Plant 2 Stage flow and pressure, twelve compressor vibration channels, four displacement probes Stage-consensus review ranking Units and repaired lineage remain explicit
Plant 3 Flow, pressure, temperature, power Thermodynamic operating excursion No mechanical surge claim
Engineering Outcomes

What This Project Demonstrates

Risk

Optimizing against unreliable labels

Maintenance dates, thermodynamic recovery boundaries, and index responses did not consistently align.

Response

Separated three layers of truth

Recorded references, physical recoveries, and verified maintenance methods are retained independently, preventing label noise from silently steering the model.

Risk

One model across unlike plants

Sensor availability, units, operating ranges, and mechanical coverage differ materially.

Response

Plant-specific contracts

No shared cross-plant threshold is allowed. Each output is named and scored according to the evidence that plant can actually provide.

Risk

A polished graph can overstate certainty

Forecast and anomaly curves can look authoritative even when validation evidence is sparse.

Response

UI-level scientific boundaries

Research-only labels, disabled decision coupling, rejection reasons, and evidence details remain visible next to the visualization.

Current Status

A validated shadow platform, not a claim machine

01

Candidate B remains the preferred Plant 2 shadow fouling formulation.

02

Forecasting remains separated from alerts and blocked from maintenance decisions.

03

Surge-oriented outputs rank evidence for review and never confirm aerodynamic surge.

04

New maintenance evidence can improve provenance without automatically rewriting model state.