The economics of a chemical plant: costs, prices and profit: Capital cost: what the plant costs to build

3.1 Capital comes in four kinds, and estimates come in classes of accuracy

Brennan separates four kinds of capital. Land is usually small and recoverable at the end of the project. Fixed capital pays for plant and buildings and has little value at the end. Start-up capital covers commissioning; it is reported at 1-10% of fixed capital, more for new technology. Working capital funds stocks and customer credit and is recovered when the plant closes. It is typically 10-20% of fixed capital, but it is all spent in a short burst at start-up.

Fixed capital is divided at the battery limits, a boundary, real or notional, around the process plant itself. Inside battery limits (IBL) costs depend on capacity, technology and the quality of raw materials and products. Outside battery limits (OBL) costs, or off-sites, cover storage, utilities such as steam, power, cooling water and effluent treatment, and buildings. OBL can range from 5% to 100% of IBL depending on the site.

Estimates are graded by purpose. Brennan cites AACE International's five classes, from Class 5 concept screening to Class 1 check or tender estimates. In his rough figures, a feasibility estimate of about 30% accuracy costs some 0.2% of project cost to prepare. A sanction estimate of about 10% needs a chosen site, flowsheets, specified and costed equipment and preliminary layouts, and costs 2-5%. A contractor's 5% estimate costs more than 5%. Contingency, the allowance for the unforeseen, might be 30% at feasibility and 10% in a detailed estimate.

Errors are not symmetrical. Unforeseen details nearly always add cost, so Brennan says a realistic range is wider on the high side than the low.

• Brennan (2020), Ch 3 Capital cost estimation · pdf pp. 59-65, 91-96

3.2 The six-tenths rule scales cost with capacity, and the exponent depends on the flowsheet

The quickest estimate starts from a plant already built. Brennan's general equation is Ip/Ir = (Qp/Qr)^b x (Fp/Fr) x L, where I is fixed capital, Q capacity, F a plant cost index, L a location factor and b a scale exponent; p is the proposed plant and r the reference. Typical accuracy is about 50%. It works best for IBL cost, since off-sites vary so much between projects.

The exponent b carries the economy of scale. If capacity came only from adding identical trains in parallel, cost would rise in proportion and b would be 1. Single large trains do better. Brennan gives b of 0.8-0.9 for mainly parallel-stream plants and 0.5-0.6 for single-stream plants. Mixed plants can be weighted. An ethylene plant is about 30% parallel furnaces and waste heat boilers and 70% single-stream purification, so b = 0.3 x 0.9 + 0.7 x 0.6 = 0.69. A chlor-alkali plant, half parallel cells and half single-stream brine treatment, gives 0.75.

For single items of equipment, a study of 662 cases found 71% of exponents between 0.4 and 0.8, averaging 0.6, which is the origin of the six-tenths rule. Wittcoff's square-cube reasoning gives about 0.67 for vessels, 1 for heat exchangers, whose duty scales with surface area, and 0 for control systems, which cost the same whatever the size.

Two cautions. Exponents are not constant. UK data showed single-stream plants at about 0.3 for 500-1,000 t a year, rising to 0.6 at 100,000-200,000 t a year. Small plants, common for new technology or small domestic markets, need particular care.

• Brennan (2020), Ch 3 Capital cost estimation · pdf pp. 65-67, 75
• Brennan (2020), Ch 6 Some examples in process economics · pdf pp. 174-177
• Wittcoff (2004), Ch 1 The chemical industry · pdf p. 58

3.3 Cost indices and location factors move an estimate through time and across borders

Reference costs are always old and usually from somewhere else, most often the United States. Two corrections are needed.

Cost indices update for inflation. The Chemical Engineering Plant Cost Index weights equipment, construction and other components; the Nelson Refinery Cost Index ran in the Oil and Gas Journal until 2017. Where no such index exists, a composite of a wage index and a materials price index is used; Brennan uses 50% of each for Australia. Indices ignore changes in technology, labour productivity and plant scope. During the 1970s actual plant costs on UK projects rose about 25% more than the indices implied, because designs became more elaborate for safety, environment and energy efficiency while site labour productivity stalled.

Location factors compare the cost of the same plant in two places, in one currency. They reflect labour rates and productivity, freight, imported components and climate, and they drift with exchange rates. Saudi Arabia relative to the US Gulf Coast ranged from 1.48 to 1.98 between 1974 and 1985. A 1994 study of a US$35 million fine chemicals plant put the United Kingdom at 1.13, India at 0.81 and Indonesia at 0.72, with the United States at 1.0. Factors also vary widely within countries.

The IEA adds a point about equipment. Because few vendors supply each technology, process equipment prices are fairly uniform worldwide. The big differences are between routes: for ammonia, equipment for coal costs more than twice that for natural gas per unit of output, and for biomass nearly seven times.

• Brennan (2020), Ch 3 Capital cost estimation · pdf pp. 67-71
• IEA (2018), The Future of Petrochemicals, Ch 2 · pdf p. 31

3.4 Factoring from purchased equipment builds the plant cost from its flowsheet

Once a flowsheet exists, the engineer lists the main plant items (reactors, columns, heat exchangers, pumps and compressors, with their motors or turbines), prices each one and sums them as the purchased equipment cost (PEC). Process plant cost is then PEC times a Lang factor, which for most plants lies between 3 and 4. The factor covers everything that turns delivered equipment into a working plant: site preparation, foundations and structures, erection, piping, instruments, electrics, insulation and painting, plus design, procurement and project management. Brennan gives a potential accuracy of about 30%.

Textbook multipliers that also include off-sites are about 4.5-6. The scope behind any factor must be checked, or off-sites get counted twice or not at all.

Brennan's data from 11 Australian projects, mostly oil and gas from the late 1980s to mid-1990s, show the breakdown. Piping is the largest addition at 0.60 times PEC, followed by instruments, civil works and installation. Overall factors vary much less between projects than individual items: the ratio of standard deviation to mean was 0.19 for direct plant cost and 0.20 for total IBL cost, but 1.58 for buildings.

Factors depend on size. The IChemE tables give larger factors for cheap items, because a controller costs the same on a small column as on a large one. Lang factors may reach 10 for pilot plants and fall to 2.5 for very large plants. Materials matter too: a Type 316 stainless steel vessel costs 2.9-3.5 times one in carbon steel.

Equipment may be only 30% of IBL cost, but the method ties capital to design choices.

• Brennan (2020), Ch 3 Capital cost estimation · pdf pp. 63, 76, 79-81, 85, 89

See it in the table

Check your understanding

A single-stream plant of 150,000 t a year costs $47.7 million. Using an exponent of 0.6, roughly what does a 600,000 t a year plant cost?

Why does an ethylene plant have a higher capacity exponent (about 0.69) than a typical single-stream plant (0.5-0.6)?

Which statement about Lang factors matches Brennan's account?