Petrochemicals

Petrochemical assets carry enormous production value per unit, so small improvements in runtime and availability compound into large returns. Causal models make those improvements findable and defensible.

Energy recycling asset runtime: five years of progress replicated in eight weeks

Global petrochemical plant operator

$65M annual saving across a 10-unit fleet, delivered in 8 weeks against a 5-year baseline improvement cycle

Challenge

Ten energy recycling assets had been improved from 10-day to 80-day average runtimes over a five-year operations program. With each unit representing $6.5M of annual production value, a further 10% improvement would generate $65M per year, but no clear path existed to close the gap.

Approach

A causal AI model analyzed interdependencies between operating parameters, maintenance actions, feedstock conditions and runtime outcomes. Asset-specific maintenance plans and optimal runtime parameter envelopes were derived for each unit individually rather than applying fleet-wide averages.

Outcome

Average runtime extended to 88 days across the fleet, delivering the 10% improvement target in eight weeks compared to the five years it took to reach the previous benchmark.

Causal AIML/DLKnowledge GraphDeterministic Code

Oxy reactor predictive maintenance: reducing unplanned shutdowns

Global petrochemical plant operator

Challenge

Oxy reactors were subject to unplanned shutdowns caused by degradation modes that threshold monitoring could not reliably detect in advance. Interdependencies between feedstock conditions, operating temperatures, pressure profiles and catalyst behavior made early warning difficult.

Approach

A causal degradation model was built using operating sensor data, feedstock quality records, maintenance history and shutdown data. It identified leading causal indicators of developing reactor risk, so operating conditions and maintenance timing could be evaluated against their expected impact on availability.

Outcome

Unplanned shutdown frequency reduced materially. Operations teams gained earlier, more reliable warning of developing conditions and a clearer framework for deciding when to intervene, improving both production continuity and maintenance planning efficiency.

Causal AIML/DLDeterministic Code

Asphaltene deposition management: preventing valve clogging in deepwater production

Global oil and gas operator

Challenge

Asphaltene deposition in deepwater production equipment was causing valve clogging, unplanned intervention and production disruption. Predicting which pressure, temperature and flow conditions would trigger deposition events was not achievable with existing monitoring tools.

Approach

A causal model characterized the operating conditions under which deposition risk was highest, drawing on wellhead pressure, temperature, flow rates, crude composition and historical intervention records. It supported operating parameter recommendations, chemical injection timing and risk-based inspection scheduling.

Outcome

Unplanned clogging events reduced significantly. Chemical injection programs were rationalized based on causal risk rather than fixed schedules, reducing both intervention costs and production disruption across the affected assets.

Causal AIML/DLDeterministic CodeKnowledge Graph

The right problem. The right AI. Working in weeks.

If you recognize your challenge here, or want to discuss one that is not covered, we would like to talk.

Book a discovery session See a demo