Costing Methods

The math from equipment size to total module cost, and from annual flows to OPEX. Every step pulls its numbers from the cost database with full provenance.

Correlation forms

The database stores six functional forms. Each correlation row declares its form, coefficients, valid size range, currency, and base index/year.

Turton log-polynomial (turton_log10)

Purchased cost at the correlation's base year (2001, CEPCI 397):

log₁₀(Cp⁰) = K₁ + K₂·log₁₀(S) + K₃·(log₁₀ S)²

Pressure factor (piecewise, pressure in barg):

log₁₀(Fp) = C₁ + C₂·log₁₀(P) + C₃·(log₁₀ P)²

Bare-module cost (installation, piping, instrumentation):

C_BM = Cp⁰ · (B₁ + B₂·F_M·F_P)        two-factor (exchangers, vessels, pumps)
C_BM = Cp⁰ · F_BM·F_M·F_P             single-factor (heaters, some reactors)

Seider/SSLW log-natural (sslw_ln)

Cp = exp(a₁ + a₂·ln(S) + a₃·(ln S)²)         base year 2006 (CE 500)
Fp = c₀ + c₁·(P/P_ref) + c₂·(P/P_ref)²
C_BM = F_M · F_P · Cp · F_install             F_install default 2.19

SSLW publishes purchased costs without Turton's bare-module split, so an installation factor reconciles the two bases — this is why multi-source comparisons are possible at the bare-module level.

Vessels by weight (sslw_vessel)

Volume → cylindrical geometry (L/D = 4) → shell weight → cost:

W = π·(D + t)·(L + 0.8·D)·t·ρ          shell weight, lb
Cp = exp(a₁ + a₂·ln W + a₃·(ln W)²)

Pumps by size factor (sslw_pump)

Power → flow at an assumed head (50 m, η = 0.7) → size factor:

S = Q_gpm · √(H_ft)
Cp = exp(a₁ + a₂·ln S + a₃·(ln S)²)

Column internals (sslw_trays, packing)

Trays: cost per tray from diameter, with a count factor and type/material factors:

C = N · N_factor · F_type · F_material(D) · base(D) · escalation · F_install

Packing: bed volume × unit cost by packing type/material, plus a distributor allowance. Columns are costed as shell (vessel correlation) + internals, each with its own provenance.

From purchased cost to CAPEX

Worked example — floating-head heat exchanger, 100 m², 25 barg, carbon steel, US-GC, Class 5:

Step	Calculation	Value
Purchased cost (2001)	10^(4.8306 − 0.8509·2 + 0.3187·4)	$54.5K
Pressure factor	Turton C-coefficients @ 25 barg	1.137
Material factor	carbon steel	1.0
Bare-module (2001)	54.5K · (1.63 + 1.66·1.0·1.137)	$192K
Escalation	Jasper Index 2001 → today (≈ 2.3×)	$445K
Location factor	US-GC (base)	× 1.0
Contingency	Class 5, +18%	$525K total module

Escalation

All costs escalate by the ratio of the Jasper Cost Index at the estimate date to its value at the correlation's base year. Correlations with different base years (2001, 2006) are brought to the same basis before assembly.

Location factors

Location is a composed model, not a country constant:

factor = 0.6 · (material_index · fx · freight_duty) + 0.4 · (labor_rate / productivity)

Region	Factor	Basis
US-GC (Gulf Coast)	1.000	definitional base
US-MW (Midwest)	1.032	representative drivers
EU	1.176	real ECB FX (1.082) + representative labor/freight

Because the factor is composed from drivers, editing a single driver (say, FX) propagates correctly instead of requiring a new hand-tuned constant.

Contingency

Applied by AACE class on located bare-module cost: Class 5 → 18%, Class 4 → 15% (Turton grassroots recommendation). Selecting the class in the UI changes only this step.

OPEX and revenue

estimate_project composes the operating side from annual flows × effective prices:

Item	Formula
Feed cost /yr	Σ annual kg × chemical price (override-aware)
Utility cost /yr	Σ annual kWh / GJ × utility price; steam uses a two-factor model that re-prices when gas price changes
Revenue /yr	Σ product annual kg × price, for Sinks marked product
Gross margin /yr	revenue − feeds − utilities (sign reported, never assumed)

Any item that can't be priced becomes a gap row, excluded from totals and listed in the result.

Multi-source blending

Where two independent sources cover the same equipment class (heat exchangers, compressors, vessels, pumps), the engine can return a blended estimate with the inter-source range — a built-in honesty check on the single-source numbers. Observed spreads: 3.3% (heat exchangers) to ~146% (vessels, method-sensitive), which empirically justifies the Class 4–5 accuracy band.