Astrext Intelligence

Machine learning that estimates remaining life and compares strategies —
with engineers in the loop where liability lives.

Astrext runs degradation models, risk scoring, and 30-year cost forecasts in minutes. Chartered engineers validate assumptions, override model outputs, and sign the pack before it reaches the board.

OPERATING PRINCIPLE

AI proposes. Engineers dispose.

Models draft RUL, condition scores, strategy rankings, and cost series. Humans set assumptions, accept or reject recommendations, lock the study, and own the signature. Every AI output is editable, reversible, and logged.

STEP 1

Model

Degradation, RUL, strategy rankings, and cost series drafted from your inputs.

STEP 2

Review

Engineer inspects assumptions, overrides outputs, and records rationale.

STEP 3

Lock

Study is signed, versioned, and exported — reproducible and auditable.

12 tags flagged by model · 3 awaiting engineer review
Heat Exchanger E-301
RUL P50 10 yr · Model
A · WHAT THE MODELS DO

Models that draft the engineering work

Physics-informed and data-driven hybrids — not a black box inventing numbers.

P10 / P50 / P90

Remaining Useful Life (RUL)

P10 / P50 / P90 estimates from age, inspection history, operating severity, and degradation-physics hybrids — not a single point guess.

0–100

Condition Scoring

0–100 condition index decomposed by subsystem factors, so you can see what is driving the number.

11 strategies

Strategy Recommendation

11 default execution strategies (Run to Failure → Decommissioning) ranked on NPV, uptime, and risk.

10–100+ yr

CAPEX / OPEX Forecast

Year-by-year cost series over a 10–100+ year horizon with inflation and discount rate applied.

Portfolio

Portfolio Heatmaps

Criticality × probability matrices with obsolescence flags across the asset portfolio.

Ranked

Anomaly / Attention List

“N tags need action” — ranked interventions the model surfaces for engineer review.

B · HUMAN-IN-THE-LOOP

Engineers dispose — visibly and accountably

Equal weight to the AI section. This is where liability lives, and where the paid products sit.

Assumption Lock

Engineer sets discount rate, inflation, unit rates, and which coefficients are inspected vs. default. Nothing runs on hidden defaults.

Override Any Score

Condition, RUL band, criticality — any model output is editable. A reason is required and becomes part of the audit trail.

Strategy Veto / Promote

Accept the model’s pick or force another strategy. The rationale is stored with the study, not lost.

Validated Study

A senior reviewer signs off methodology and assumptions. $4,900 tier includes a written assurance letter.

Full Study

A multi-discipline team produces a FID-grade Life Extension Management Plan. AI accelerates — it does not replace the team.

C · TRUST & AUDIT

Every output is traceable, reproducible, and signed

Source Tags on Every Input

Each value is marked Inspected, Default, User override, or Model-inferred — visible at the field level.

Reproducible Runs

Same inputs → same P10 / P50 / P90. Runs are deterministic and versioned, not stochastic mood swings.

Assumption Audit Trail

Who changed what, when, and why — exportable for investment committees and lenders.

Engineer Owns the Signature

The model never signs. A chartered engineer locks the study and owns the decision.

Astrext augments engineering judgment. It does not replace professional responsibility or regulatory approval.

Grounded in recognized standards

Models are grounded in RBI (API 580/581), Fitness-for-Service (API 579), and ISO 55001 asset management concepts — physics-informed and data-driven hybrids, not a generative LLM inventing numbers.

Explore the Knowledge Hub
DemoWorked example — illustrative only
Platform Bravo
Heat Exchanger E-301
Condition 72RUL P10/P50/P90 = 6/10/14 yr
Tags scoped
45
12 need action
Recommended
Reliability-Driven
ranked by NPV / uptime
30-yr cost
$74M
$42M CAPEX + $32M OPEX
Uptime / NPV
97% / −$45M
@ 10% discount

Figures are demo values. Real outputs depend on your asset data and engineer-locked assumptions.

Models draft. Engineers sign.

Run unlimited scenarios free. Add human validation when the decision carries liability.