ADIPEC 2026

See Orbital live on the KBR stand · Hall 7, booth 7254 · 2–5 Nov, Abu Dhabi

ADIPEC 2026

See Orbital live on the KBR stand · Booth 7254, 2–5 Nov

ADIPEC 2026

Orbital live · Booth 7254

Physics-grounded AI, verified by the engineers who run the plant

Indorama Ventures · Port Neches, Texas

F5 PO/MTBE unit

31 Aug 2026

Cover of the Indorama Ventures F5 PO/MTBE case study

01

01

Results

Results

6/6
6/6
6/6

Process sections verified and signed off

16
16
16

Quality parameters within tolerance

4+
4+
4+

Years of plant history modelled

02

02

Goals

Goals

Applied Computing built Orbital for Indorama Ventures against more than 4 years of the F5 unit’s own DCS and laboratory history. This engagement aimed to:

Verify Orbital’s understanding of all 6 process sections of the F5 PO/MTBE unit, section by section and prompt by prompt.

Predict 16 quality parameters across the PO, TBHP and MTBE streams within an agreed accuracy tolerance, using DCS data continuously calibrated against lab results.

03

03

Process background

Process background

Indorama Ventures Limited’s F5 PO/MTBE unit is one of the most complex units within its Port Neches Operations (PNO) plant: six tightly coupled process sections across two production chains, more than 15,000 tags and instruments, and decades of operating knowledge spread across documents, historians and long-tenured engineers. IVL put Orbital on this complex unit to stress-test its full capability.

Orbital runs as three layers on top of that: a physics-grounded foundation that learns the unit’s own behaviour, a prediction layer that continuously estimates the 16 quality parameters above, and an access layer that lets engineers ask plain-language questions and see the reasoning behind each answer. It operates in an advisory capacity only. It never writes setpoints or trips equipment, and engineers implement any change themselves under full operator control.

See the reasoning behind a number