The enterprise friction point
GulfStream's linear programming (LP) planning tools ran on weekly static cycles, failing to adapt to real-time tank stratification and crude assay delivery variances. As a result, product blenders over-added expensive reformate and alkylate to guarantee 87 and 93 octane specifications, giving away an estimated $0.28 per barrel in octane and Reid Vapor Pressure (RVP) margin.
Transformation objectives
- Ingest LIMS component assays and DCS crude distillation unit temperatures in real time.
- Predict non-linear blend property interactions (octane number, RVP, sulfur, distillation points).
- Calculate least-cost component recipe additions for finished gasoline and diesel tanks in under 10 minutes.
- Eliminate off-spec tank re-blending and reduce product component giveaway by >80%.
Discovery & architectural audit
Historical tank batch audits revealed that 74% of finished gasoline batches exceeded octane specs by 0.6–1.2 AKI points. Blenders operated conservatively because static laboratory assays were 12 to 24 hours old by the time recipes were pumped.
The Fortiv AI architecture
- 01
DCS & LIMS Ingestion Pipeline: Secure read-only OPC-UA bridge connecting OSIsoft PI historian tags with laboratory chromatography assays.
- 02
Non-Linear Thermodynamic Engine: Neural physics-informed algorithms modeling true boiling point (TBP) curves and component blending interactions.
- 03
Supervisory Blending Console: Control room advisory interface calculating exact component valve flow setpoints with operator verification checkpoints.
- 04
Closed-Loop Model Recalibration: Post-blend laboratory certification results automatically update tank stratification coefficients continuously.
Verified audit outcomes
Post-deployment impact
- +3.4% Margin Lift
- Representing $3.2M in unlocked annual refining value
- 4h → 8 mins
- Blend recipe calculation turnaround compressed by 96%
- -82% Giveaway
- In expensive high-octane reformate and alkylate blendstocks
- 0 Off-Spec Batches
- Across 400+ finished product blending runs
Architectural lessons learned
Displaying real-time margin cost per component gives control room board operators immediate economic clarity when adjusting recipe trims.
Read-only DCS integration satisfies refinery cybersecurity without requiring operational shutdown windows.
Future scale roadmap
Extending predictive yield analytics to fluid catalytic cracking (FCC) unit catalyst circulation and hydrocracker hydrogen consumption optimization.
Implementation FAQ
Related Architecture & Intelligence
Explore Fortiv AI Capabilities & Blueprints
39 Production Use Cases
Explore Enterprise Software Use Cases
Browse practical software blueprints across 12 sectors with zero hardware capex.
Assessment
AI Readiness & ROI Evaluation
Benchmark technical maturity and estimate payback timelines for your organization.
Core Solution Blueprints: