Experimental and computational
Work emphasizing problems where a result changes practice rather than only the literature.
Research
Forty-five studies, each published as a self-contained page that states its finding, its method and its limits. Most return a negative or qualified result, and we publish those in full, because the qualification is usually the finding.
- Studies
- 45 published
- Period
- 2021 to 2026
5
Steam and Cogeneration Dispatch Testbed
One modeled chemical site with boilers, turbines, pressure headers and a chlor-alkali cell, followed in five parts from its first verified balances to its price sensitivity. The site is modeled, not measured.
- 1FoundationsEvery balance in a modeled steam network closes to round-off, and that is not the same as accuracyApril 14, 2026
- 2DispatchOnsite condensing power is cheaper than the grid but carries four times its carbonMay 26, 2026
- 3Electrolysis loadA chlor-alkali cell house coupled to a cogeneration system draws every extra megawatt from the gridJuly 14, 2026
- 4ScenariosOn a modeled cogeneration site, electrified steam raises total emissions and steam savings show up as electricitySeptember 15, 2026
- 5Price sensitivityA cheaper-gas year flips the optimal steam dispatch into a regime that emits a fifth moreOctober 9, 2026
- 6Acceptance criteria and independent review caught the computational faults of an autonomously built engineering model, but not its modeling errorsOctober 11, 2026 | How it was built
- 7A read-only record server is not read-only by default, and five ways a mirror's own configuration could subvert a readOctober 13, 2026 | How its record is served
7
Independent steam studies
Each tests one thing the testbed relies on. They cover the federal energy tables, the steam properties, the DOE tip sheets, the commitment under uncertain scenarios, the grid emission factor, the steady-state headers and the code it is made of. None changes a testbed result.
- 8DOE steam tip sheet examples reproduce to their printed rounding, and each assumes that saved energy displaces boiler productionDecember 5, 2022 | Verification
- 9A minimax cogeneration commitment chosen on training scenarios had the largest regret on held-out scenariosJune 14, 2023 | Process systems
- 10In every hour of 2023, California's marginal emission rate exceeded the annual inventory emission factorNovember 12, 2024 | Life-cycle assessment
- 11Fewer than a third of the cells in the federal manufacturing-energy tables are ordinary published numbersMarch 20, 2025 | Data analysis
- 12A test corpus can be wrong too, and an independent steam-property verifier found three invalid states that were validSeptember 28, 2025 | Verification
- 13A steam demand step that a steady-state dispatch treats as instantaneous lowers header pressure by 37 kPa before control restores itOctober 2, 2026 | Process control
- 14A verifier for autonomous scientific code caught every seeded defect, after two of its own tests were made able to failOctober 4, 2026 | Verification
6
Hydrogen, fuels and carbon
What low-carbon hydrogen, methanol, ethanol and ammonia fuels cost, and what sets their footprint.
- 15What it costs to make green hydrogen by PEM electrolysisOctober 2, 2024 | Techno-economics
- 16Anhydrous ethanol separation takes 57.6 % of the heating value at 23.9 g/L, the rule-limited beer of a yeast fermenter modelNovember 24, 2025 | Process design
- 17Across four kinetic mechanisms, ammonia/hydrogen flames reach the burning velocity of methane at hydrogen fractions of 0.443 to 0.528December 15, 2025 | Combustion
- 18Grid carbon intensity, not electrolyser efficiency, determines the well-to-gate footprint of PEM hydrogenSeptember 9, 2026 | Life-cycle assessment
- 19Hydrogen sets 94 % of the cost of methanol made from captured CO2 and PEM electrolysisSeptember 29, 2026 | Techno-economics
- 20Syngas CO2 capture on a steam methane reformer matches grey hydrogen at a carbon price of 54 USD/t, and flue-gas capture at 164 USD/tOctober 19, 2026 | Techno-economics
12
Process design, control and thermodynamics
Flowsheets, unit operations, property data, and the solvers and controllers that run them.
- 21Ipopt misses the proven optimal water network in up to 44 % of random starts on six benchmarksOctober 12, 2021 | Optimization
- 22Linear MPC lowers tracking error against PID in 26 of 30 cells of a simulated yeast fermenter but moves the flows far moreNovember 29, 2021 | Process control
- 23Hard output constraints make a multi-fidelity reactor surrogate both data-efficient and safe to optimise againstDecember 8, 2021 | Machine learning
- 24A molded-fiber cushioning tray that does not passApril 28, 2023 | Packaging design
- 25Which reactor makes the most B, and what does it cost?January 30, 2024 | Process design
- 26Does a two-height BDST line predict the third height?April 22, 2024 | Process design
- 27Packed-column flooding against one measured operating pathJune 22, 2024 | Process design
- 28Aromatics lost to the purge set the cost optimum of the IDAES hydrodealkylation flowsheet under an assumed hydrogen limitOctober 18, 2024 | Process design
- 29A pinch analysis and heat integration explorerOctober 31, 2024 | Interactive tool
- 30Ammonia, high-temperature shift and methanol equilibria tolerate only 0.30-0.65 kJ/mol of reaction-enthalpy error per point of conversionNovember 3, 2025 | Thermodynamics
- 31Raoult's law misplaces the cost-optimal condenser temperature of a 300 bar ammonia synthesis loop by 46-54 °FNovember 10, 2025 | Process design
- 32Six in ten testable ThermoML vapour-liquid equilibrium data sets fail each of two consistency testsDecember 28, 2025 | Thermodynamics
10
Safety, monitoring and reliability
Runaway screening, incident causes, fault detection, life prediction and measurement quality.
- 33Which sensors carry real signal about lot failure?July 16, 2021 | Data analysis
- 34Can the measurement system tell good parts from bad data?January 8, 2022 | Quality engineering
- 35At the earliest deciding horizon, a nominal 90 % lower bound on early-predicted battery life covers only 17 of 40 held-out cellsNovember 14, 2022 | Reliability
- 36Reliability qualification by accelerated life testingFebruary 1, 2023 | Reliability
- 37The semi-batch MTSR matches dynamic simulation within 0.03 K when the feed stops at a cooling failureAugust 11, 2025 | Process safety
- 38On Gaussian AR(1) data, moving-range control charts have an in-control run length below 93 observations, not 370, once lag-1 autocorrelation reaches 0.4February 24, 2026 | Quality engineering
- 39Seven published C-MAPSS methods show no demonstrated RMSE gain over tuned gradient boosting in 28 comparisonsMay 19, 2026 | Reliability
- 40Published PCA fault-detection rates for the Tennessee Eastman process hold with held-out control limits, at 0.719 against a printed 0.720July 7, 2026 | Process monitoring
- 41Increasing hazards in 8 of 22 hard-drive models and field failure rates above ratings for 20 of 21 in Backblaze dataOctober 16, 2026 | Reliability
- 42Four management-system elements are the most frequent reliably coded causes in 119 CSB investigations, and recommendations to government take a median 876 days longer to closeOctober 20, 2026 | Process safety
3
Research methods and verification
How public data is qualified for reuse, and how published results are checked independently.
- 43Qualifying public data for reuse across five stages, three candidates and one passMarch 2, 2022 | Research methods
- 44A disagreement catalogue for licences that do not answer the questionMarch 9, 2022 | Research methods
- 45Independent verification of nine published studiesAugust 31, 2026 | Verification
02
AI ethics and applied research
Program 02. Tools and methods for anyone who needs to judge whether a claim, a citation, or a generated answer can be trusted.
Ethical AI is often discussed as principle and rarely delivered as something a person can actually use. Our work in this program is applied: tools and methods for anyone who needs to judge whether a claim, a citation, or a generated answer can be trusted.
That audience is deliberately broad. Students, journalists, clinicians, librarians, and researchers face the same underlying problem, and it does not take a technical background for it to matter to them.
Interactive teaching tools
These run in the browser, are free, and need no account.
The study reports above are fixed documents: they state what was found and the reader follows the argument. A tool is a different kind of teaching. It lets someone change a condition and watch the answer move, which is how the shape of a result becomes intuition rather than a figure to be taken on trust.
Vapor liquid equilibrium explorer
Bubble and dew points, activity coefficients, flash calculations and phase diagrams for 44 components. Choose a mixture, pick an activity model, and plot the Txy or Pxy diagram. Every number is computed from a physics core and traceable to its model, parameters and literature source.
It is built for a first course in separations, where the ethanol and water azeotrope is easier to believe once you have watched the curve reach it yourself.
03
Chemistry, engineering, and AI agent systems
Program 03. Research at the undergraduate and graduate level.
Agent systems are increasingly used to plan syntheses, screen candidates, and read literature at scale, which makes their failure modes a chemistry problem as much as a computing one.
Students and collaborators: we take on research partners.
Process and systems
Scaling what works at the bench into something that survives real production constraints.
Multi-agent architectures
How agent systems coordinate, where they fail, and what verification they need before a lab trusts their output.