response-surface candidates generated in Docker
Docker-Generated Response Surface
Optimization Lab
A 3D bench-design workspace for optimizing sewage-sludge HTL temperature, residence time, solids loading, catalyst family, and aqueous-phase recirculation before the next physical campaign.
bench runs selected for the first optimization portfolio
top modeled catalyst family in the balanced objective
modeled yield for the top balanced run; best yield ceiling is 49.11%
Frontier run
Ni-hydrochar catalyst at 350 C, 60 min, 15 wt% solids, 1x AP recycle
Fuel quality condition with 60.9% modeled energy recovery, 36.51 MJ/kg HHV,4.78% oil nitrogen, and 38.2 g H2/kg oil estimated upgrading burden.
- Oil C
- 74.58%
- Oil O
- 9.32%
- Aqueous COD
- 79.2 g/L
- NH3
- 4.83 g/L
Aqueous recycle is treated as a real tradeoff: the literature-backed branch can lift yield from 32.1% toward 40.2%, but the model also penalizes lower oil carbon, higher oxygen, COD, ammonia, and biological-return risk.
The Docker selector picked 8 runs from 864 modeled candidates. Portfolio average yield is 42.78%, average quality index is 92.1, and average aqueous risk is 71.9. Pyomo-formulated eight-run design-of-experiments portfolio with baseline, catalyst, recirculation, yield-ceiling, continuous-stress, and lower-temperature-quality gates; selected by deterministic category enumeration.
The current top run is Ni-hydrochar catalyst at 350 C, 60 min, 15 wt% solids, 1x AP recycle. It is a bench priority, not a recipe, until real product separation, hydrotreating, catalyst carryover, aqueous treatment, and LCA data are measured.
RDKit marker molecules keep the response surface tied to lipid, protein, carbohydrate, aqueous COD, oil nitrogen, oxygenate, fuel proxy, and oxidant chemistry.
Current optimization decision
Run a measured portfolio around Ni-hydrochar, aqueous recycle, and a no-catalyst control
The response surface ranks Ni-hydrochar at 350 C, 60 min, 15 wt% solids, and one aqueous-phase recycle as the strongest balanced run. The selected portfolio deliberately includes a baseline, a 3x and 6x recycle tradeoff, acid hydrochar, CuSO4 reference chemistry, a 20 wt% solids stress condition, and a lower-temperature quality check so the team can learn why the optimum works instead of only testing the single best-looking condition.
Published optimized sludge-composition work is carried as a high-performance benchmark with 69% energy recovery.
AP recycle can raise yield, but the same branch is penalized for carbon dilution, oxygen increase, COD, ammonia, and return-load risk.
ML-aided TEA/LCA is treated as a planning aid; measured data still controls product, economics, and climate claims.
Selected bench portfolio
Eight-run portfolio to find the next real optimum
No catalyst baseline
350 C, 60 min, 15 wt% solids, 0x AP recycle. Yield 36.32%, quality 100, aqueous risk 52.7.
Ni-hydrochar catalyst
350 C, 60 min, 15 wt% solids, 1x AP recycle. Yield 43.6%, quality 100, aqueous risk 65.6.
Ni-hydrochar catalyst
350 C, 60 min, 15 wt% solids, 3x AP recycle. Yield 46.78%, quality 84.9, aqueous risk 90.3.
Ni-hydrochar catalyst
350 C, 40 min, 15 wt% solids, 6x AP recycle. Yield 48.45%, quality 52, aqueous risk 100.
Acid-activated hydrochar
350 C, 60 min, 15 wt% solids, 1x AP recycle. Yield 41.32%, quality 100, aqueous risk 65.
CuSO4 reference catalyst
350 C, 60 min, 15 wt% solids, 1x AP recycle. Yield 42.09%, quality 100, aqueous risk 66.7.
Ni-hydrochar catalyst
350 C, 60 min, 20 wt% solids, 1x AP recycle. Yield 42.46%, quality 100, aqueous risk 71.
Ni-hydrochar catalyst
325 C, 60 min, 15 wt% solids, 1x AP recycle. Yield 41.23%, quality 100, aqueous risk 64.2.
Response-surface averages
Temperature and aqueous recycle change yield and quality together
- Yield
- 23.6%
- Quality
- 100
- Energy
- 47.7%
- Water risk
- 51
- Yield
- 25.76%
- Quality
- 99.4
- Energy
- 47.6%
- Water risk
- 63.2
- Yield
- 29.09%
- Quality
- 80.6
- Energy
- 46.5%
- Water risk
- 88
- Yield
- 31.65%
- Quality
- 47.8
- Energy
- 43.2%
- Water risk
- 100
- Yield
- 34.14%
- Quality
- 100
- Energy
- 55.4%
- Water risk
- 52
- Yield
- 36.43%
- Quality
- 99.6
- Energy
- 55.3%
- Water risk
- 64.3
- Yield
- 39.6%
- Quality
- 82
- Energy
- 54.2%
- Water risk
- 89
- Yield
- 42.22%
- Quality
- 49.3
- Energy
- 50.9%
- Water risk
- 100
- Yield
- 36.56%
- Quality
- 100
- Energy
- 57.1%
- Water risk
- 53.1
- Yield
- 38.71%
- Quality
- 99.7
- Energy
- 57%
- Water risk
- 65.5
- Yield
- 42.09%
- Quality
- 81.9
- Energy
- 55.9%
- Water risk
- 90.2
- Yield
- 44.7%
- Quality
- 49.3
- Energy
- 52.7%
- Water risk
- 100
- Yield
- 35.28%
- Quality
- 100
- Energy
- 56%
- Water risk
- 53.4
- Yield
- 37.42%
- Quality
- 99.5
- Energy
- 55.8%
- Water risk
- 65.9
- Yield
- 40.76%
- Quality
- 80.7
- Energy
- 54.8%
- Water risk
- 90.6
- Yield
- 43.4%
- Quality
- 47.9
- Energy
- 51.5%
- Water risk
- 100
Catalyst ledger
Catalysts are ranked as bench candidates, not deployment decisions
Ni-hydrochar catalyst
- Best score
- 61.1
- Best yield
- 49.11%
- Mean quality
- 83.6
- Water risk
- 76.4
Acid-activated hydrochar
- Best score
- 57.5
- Best yield
- 47.75%
- Mean quality
- 83.2
- Water risk
- 76.8
CuSO4 reference catalyst
- Best score
- 57.1
- Best yield
- 48.32%
- Mean quality
- 82.5
- Water risk
- 77.5
KOH-activated hydrochar
- Best score
- 56.5
- Best yield
- 46.79%
- Mean quality
- 83.9
- Water risk
- 76.5
FeSO4 low-cost salt
- Best score
- 55.6
- Best yield
- 46.73%
- Mean quality
- 81.9
- Water risk
- 76.6
No catalyst baseline
- Best score
- 53.9
- Best yield
- 43.84%
- Mean quality
- 79
- Water risk
- 76
Molecular anchors
Representative chemistry behind the optimization surface
Citric acid C6H8O7
- Weight
- 73.6
- MW
- 192.123
- C atoms
- 6
- Hetero
- 7
Glucose C6H12O6
- Weight
- 66.6
- MW
- 180.156
- C atoms
- 6
- Hetero
- 6
Oleic acid C18H34O2
- Weight
- 51.8
- MW
- 282.468
- C atoms
- 18
- Hetero
- 2
Glycine C2H5NO2
- Weight
- 49.7
- MW
- 75.067
- C atoms
- 2
- Hetero
- 3
Pyridine C5H5N
- Weight
- 39
- MW
- 79.102
- C atoms
- 5
- Hetero
- 1
Ozone O3
- Weight
- 39
- MW
- 47.997
- C atoms
- 0
- Hetero
- 3
Acetic acid C2H4O2
- Weight
- 34.2
- MW
- 60.052
- C atoms
- 2
- Hetero
- 2
Ammonia H3N
- Weight
- 33.5
- MW
- 17.031
- C atoms
- 0
- Hetero
- 1
Carbon dioxide CO2
- Weight
- 33.1
- MW
- 44.009
- C atoms
- 1
- Hetero
- 2
Phenol C6H6O
- Weight
- 31.6
- MW
- 94.113
- C atoms
- 6
- Hetero
- 1
Dodecane C12H26
- Weight
- 31.2
- MW
- 170.34
- C atoms
- 12
- Hetero
- 0
Hydrogen H2
- Weight
- 18
- MW
- 2.016
- C atoms
- 0
- Hetero
- 0
opt-sludge-composition-2025
Municipal sludge composition and HTL performance optimization
Supports treating feedstock composition, biocrude elemental quality, HHV, and energy recovery as linked optimization outputs.
Open sourceopt-cuso4-optimization-2025
Municipal sludge HTL optimization with CuSO4 catalyst
Supports metal-salt catalyst and response-surface style optimization around operating conditions.
Open sourceopt-ni-hydrochar-2025
Ni-hydrochar catalytic sludge HTL
Supports recycled hydrochar catalyst screening at roughly 350 C, 60 min, and 5 wt% loading as a high-value bench branch.
Open sourceopt-aqueous-recirculation-2026
Aqueous phase recirculation in sewage-sludge HTL
Supports explicitly modeling recirculation as a yield/quality tradeoff rather than assuming recycled aqueous phase is free upside.
Open sourceopt-ml-tea-lca-2025
Machine-learning-aided LCA and TEA for sludge HTL
Supports using surrogate models as planning tools while preserving measured TEA/LCA validation gates.
Open sourceopt-continuous-pfr-2025
Continuous plug-flow sludge HTL modeling and optimization
Supports carrying batch optimization into continuous residence-time, heat-transfer, and operability questions.
Open sourceopt-ozone-aqueous-2025
Ozonation and anaerobic digestion of HTL aqueous phase
Supports testing aqueous-phase pretreatment near an optimum ozone dose before returning water to biological treatment.
Open sourceopt-aqueous-cotreatment-2025
Co-treatment of HTL wastewater in activated-sludge systems
Supports measuring biological compatibility, settling, and effluent quality when HTL water is blended into WRRF treatment.
Open sourceoptimization output is a bench-experiment plan, not a proven recipe, until every selected condition has measured oil, gas, char, aqueous, upgrading, TEA, and LCA data