BRENNEN

Industries / Mining and materials / Chemicals and petrochemicals / Refining

Distillation

Distillation heats crude oil and splits it by boiling point into gases, gasoline, kerosene, diesel, heavier gas oils and a residue, and it is the first step at every refinery.

4

Milestones achieved, from 1856 to 2015.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1856

    Kerosene from a still

    Ignacy Łukasiewicz, a pharmacist in Lwów, purified crude oil into lamp kerosene in 1852–1853, and lamps made to his design were first lit in a Lwów hospital on 31 July 1853. In 1856 he opened an oil distillery at Ulaszowice, near Jasło, which the Bóbrka oil museum calls the first on Polish land. Early refineries worked in batches: a kettle of crude was heated over a fire, the kerosene boiled off, and the tar left behind was cleared out before the next batch, according to Penn State's refining course.

    bobrka.plcourses.ems.psu.edu

    Credit

    Ignacy Łukasiewicz, who did the first work with Jan Zehn

  2. 1911

    Electric fields dry crude

    Crude oil arrives carrying droplets of salty water, and refineries remove the salt before distillation because chloride corrodes the equipment downstream, as Penn State's course explains. A US patent granted on 21 March 1911 described applying electricity to crude petroleum so that small water droplets gather and settle out. A 2016 patent application by Cameron Solutions describes an electrostatic desalter that Petreco designed in 1980 as used by many of the world's refiners.

    patents.google.compatents.google.comcourses.ems.psu.edu

    Credit

    Frederick Gardner Cottrell and James Buckner Speed, for the Petroleum Rectifying Company

  3. 1976

    Packing made of sheet metal

    Sulzer showed Mellapak at the ACHEMA trade fair in 1976. It is structured packing, corrugated metal sheets stacked inside a column in place of trays. Sulzer had sold packing made of wire gauze since 1964, and the sheet-metal kind succeeded because it cost far less. In the 1980s refiners took it up for vacuum towers, which distill the heaviest oil at low pressure, because vapor loses little pressure as it rises through it. As columns grew, spreading liquid evenly over the packing became a design problem of its own, according to a 2018 review by Lothar Spiegel.

    LessonA larger column needs as much care in feeding liquid evenly as in choosing the packing.

    aidic.it

    Credit

    Sulzer

  4. 2015

    Light oil, heavy-oil plants

    In 2015 more than 70% of the crude produced in the US Lower 48 states was light oil, above 35 degrees on the API gravity scale, on which higher numbers mean lighter oil. Ninety percent of imported crude was heavier, and most US refineries were designed for medium and heavy crude, according to the US Energy Information Administration (EIA). Refiners took in the new oil mainly by importing less light crude, running their plants harder and blending light with heavy crude, so the average input stayed at about 32 degrees.

    LessonA refinery is built around one kind of crude, and when supply changes it adapts first by changing what it buys and blends.

    eia.gov

    Credit

    US Energy Information Administration, for the analysis

Watch for

  1. Dangote at full rate

    The Dangote refinery near Lagos, Nigeria, said its crude distillation unit and gasoline-making units had reached their design rate of 650,000 barrels a day. It called itself the first refinery of that size with a single train, meaning one crude unit rather than several side by side, to do so. It began 72 hours of performance test runs with UOP, its technology licensor, and said the remaining units would be tested in a second phase.

    Achieved whenUOP, a Nigerian regulator or another outside party publishes measured crude runs at or near 650,000 barrels a day held for at least a month.

    nairametrics.com

    Claimed by

    Dangote Petroleum Refinery, February 2026

Open

  1. 01

    Which crudes fit which refinery

    An open screening model that takes a public crude assay and one US refinery's published unit capacities, and reports whether that refinery could run the crude without overloading any unit, such as the vacuum tower, coker, hydrotreaters or sulfur plant. It is run for every US refinery against a library of public assays. An assay is the laboratory breakdown of a crude into boiling-range cuts, with the yield and sulfur of each.

    The brief
    Credit · unclaimed
  2. 02

    A shared test for assays

    A public benchmark that asks a method to predict a crude oil's full assay from quick measurements. A full assay takes days of laboratory distillation. The quick inputs would be density, sulfur and a simulated distillation curve measured by gas chromatography. A fixed set of held-out crudes would let every method be scored the same way.

    The brief
    Credit · unclaimed

Open 01 · the brief

Which crudes fit which refinery

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open screening model that takes a public crude assay and one US refinery's published unit capacities, and reports whether that refinery could run the crude without overloading any unit, such as the vacuum tower, coker, hydrotreaters or sulfur plant. It is run for every US refinery against a library of public assays. An assay is the laboratory breakdown of a crude into boiling-range cuts, with the yield and sulfur of each.

Why it is still open

Refiners choose crudes with their own private planning models. The University of Calgary's free PRELIM model runs public assays through eleven standard refinery configurations to estimate energy use and emissions, not through a named refinery's capacities. A 2025 paper released what it calls the first open-source production planning benchmark for refinery-petrochemical complexes, built on three cases rather than on public refineries. EIA reports each refinery's crude imports in five grades, from light sweet to heavy sour, but not which crudes each refinery could run. A search found no open model of this kind.

ucalgary.caarxiv.orgcatalog.data.gov

Built with what exists

EIA's refinery capacity report for the size of each unit; public assays from Equinor, ExxonMobil and NOAA's ADIOS database; PRELIM's documented unit relations as a starting point; Python with Pyomo and the open-source HiGHS solver; a coding agent that tests its own work, such as one built on Inkling 2 Small (see the Note "Checked without being asked").

eia.gov

  1. First step

    Pick five refineries with different unit mixes from EIA's capacity report and work out by hand the heaviest and most sulfurous crude each could run.

  2. Done looks like

    Every US refinery with a crude unit screened against at least 50 public assays. A test against EIA's company-level import files, which give the gravity and sulfur of each cargo and the refinery that received it, in which at least 90% of the cargoes each refinery imported in 2025 and 2026 fall inside the range the model calls feasible for it. The share of the assay library ruled out for each refinery published, with the code.

    eia.gov

Open 02 · the brief

A shared test for assays

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A public benchmark that asks a method to predict a crude oil's full assay from quick measurements. A full assay takes days of laboratory distillation. The quick inputs would be density, sulfur and a simulated distillation curve measured by gas chromatography. A fixed set of held-out crudes would let every method be scored the same way.

Why it is still open

NOAA's ADIOS Oil Database already gathers thousands of public oil records, including ExxonMobil assays, in one JSON format, but for oil-spill response rather than for testing predictions. A patent application describes making a "virtual assay" from a simulated distillation and the oil's density without fractionating the sample, but gives no shared test set. No benchmark with a fixed test set was found. A GitHub search found small projects, the largest a student internship at IIT Bombay that modeled 28 crude samples.

response.restoration.noaa.govipiq.iogithub.com

Built with what exists

The ADIOS data on GitHub, which holds 58 ExxonMobil assay records with eight cuts each alongside records from Environment and Climate Change Canada and Norway; assays that producers such as Equinor publish free; Python with scikit-learn; a published boiling-curve correlation as the baseline; a language model that reads PDF-only assays into the same fields and quotes the line each number came from, with a person checking each record (see the Note "Findings that cite notes").

github.com

  1. First step

    Download the ADIOS JSON files and count how many records give the yield and sulfur for at least five cuts.

  2. Done looks like

    At least 300 crudes in the set; test crudes held out by producer, so no model sees the same company's lab; the error of the baseline and of at least one learned model published for each cut, in percentage points of yield; and results from at least two outside teams.