Product status
This page documents the methodology and validation approach in advance of release. Pricing and signed installers will be announced when the product is ready. To register interest, use the Contact page.
Methodology & Validation
The Trader-side blend calculator uses the same blending methods that have underpinned refinery and bunker operations for decades. We treat the mathematics as a first-class engineering artefact: every method is documented in-code with its source equation, every release is validated against textbook and hand-calculated reference cases, and any regression in the pure blending math breaks the build.
At a glance
The blending methods
MaeWest uses the blending methods established across the petroleum industry. Each is implemented in a dedicated, testable calculation module with the source equation quoted in the code header, so a marine chemist or vetting inspector can trace the arithmetic without reverse-engineering a spreadsheet.
Sulphur, density at 15 °C, MCR, ash, water, and metals (V, Ni, Na, Al+Si, Fe, Ca, Zn, Pb) blend linearly on a mass basis, per API MPMS Ch. 11.1 conventions. Results are mathematically exact and match paper arithmetic to full floating-point precision.
Kinematic viscosity is blended via the Refutas Viscosity Blend Number, based on ASTM D341 (MacCoull–Walther). Component viscosities are mapped to VBN, mass-weighted, and back-transformed to cSt. The offset term handles very light distillates; all inputs must be at the same reference temperature (50 °C for residual grades, 40 °C for distillates per ISO 8217).
Flash point is combined via the Chevron flash blending index (Nelson form). Volume-weighted on the index; the closed-form inverse recovers the blend flash in °C. Used across refinery product-blending desks for cutter/residual combinations.
Pour point is combined via an ASTM-family pour blending index. This is the industry-standard non-linear method for pour prediction; wax chemistry means it inherits a wider accuracy envelope than viscosity or flash, and the on-screen breakdown labels it as such.
Every blend preview is scored against the target ISO 8217 grade (RMG 380, RMK 500, DMA, VLSFO 0.50 % S, ULSFO 0.10 % S, and the rest) with per-property pass/fail flags and the margin on each parameter. A live spec badge shows the result before the instruction is issued.
Calculated Carbon Aromaticity Index and related correlations follow the ISO 8217 Annex F closed-form definitions. Used to flag ignition-quality concerns on residual blends before the fuel reaches the receiving-end engine.
How we validate every release
MaeWest ships with a dedicated blend-validation test project — separate from the application itself — that runs on every build. It has two layers:
Hand-calculated and textbook worked examples for the pure blending mathematics — mass-weighting, Refutas index round-trip across the 3 cSt to 3500 cSt range, and independent-arithmetic checks of the residual + cutter case. Any drift outside the ISO 4259 reproducibility band fails the build. A coefficient regression cannot ship.
A growing library of real-world component + blend pairs, evaluated on every build and written to a Markdown delta report showing calculated vs measured per property. This tracks real-world accuracy across residual, distillate, and mixed regimes without artificially blocking releases when a case reveals a method-envelope limit.
The report uses the ISO 4259 reproducibility bands — the value below which two independent labs may disagree on the same sample with 95 % probability — as the “in agreement with the lab” threshold. Baked into the harness as constants so every case is scored the same way.
The Refutas validation cases include the worked example from Riazi, Characterization and Properties of Petroleum Fractions (2005), Ch. 3, alongside independent hand-calculations. Both are enforced on every build so a change to the blending code cannot silently move the answer.
Honest about the method envelope
The Refutas, Chevron, and ASTM pour-point indices are the standard methods used across the refining and bunker industries. They are also the reason every serious blend desk works alongside a receiving-end CoQ rather than in place of one: the method envelope on non-linear property prediction is typically ±10–20 %, widest in mid-range residual blends and narrowest at the extremes.
MaeWest treats that as an engineering fact rather than something to hide. The Blender header carries a small “Prediction accuracy” note reminding the trader that viscosity, flash and pour are index-blended and that the receiving-end lab CoQ is the source of truth. The signed audit chain records both the predicted and the delivered CoQ, so any drift is captured and reviewable — which is the correct posture for a QC-audited operation.
What that means in practice: sulphur, density, MCR and metals will match a lab CoQ to within instrument noise; viscosity, flash and pour will land inside the industry method envelope. Traders leave headroom on the non-linear properties (the Blender exposes explicit headroom controls) and confirm on the receiving CoQ. That is exactly how a well-run blend desk operates.
Bring your own data
The blend-validation harness is designed to accept new cases without touching the application code. If you share historical component + blend CoQ pairs from your operation — anonymised as needed — we will add them to the harness and publish a delta report comparing MaeWest predictions to your actual lab results across the properties you care about, scored against the ISO 4259 reproducibility bands.
This works particularly well for:
Trader / blend planner comparing tank CoQs to the delivered blend CoQ across a season of loadings. The delta report becomes a defensible QC record for procurement and class-society audits.
Component CoQ + barge CoQ + receiving vessel CoQ chains. Useful for demonstrating consistency of blending practice to charterers and P&I clubs.
A ready-to-use, method-labelled, ISO-4259-scored second opinion on a customer’s blend calculations. The Markdown report format is dropped straight into an audit file.
The whole point
Every blend instruction issued by the Trader is content-hashed and signed. Every execution and receiving-end CoQ returned by the FSO is content-hashed and signed. The full chain — predicted properties, delivered properties, and any downstream correction — is independently re-verifiable at charter close-out or during a dispute. The blend calculator is one input into that chain, and the CoQ is another; MaeWest’s value is that they end up in the same tamper-evident record instead of in three different spreadsheets on three different desks.