Scale, capacity and feedstock: which chemical plants get built: Integration, uncertainty, safety and sustainability: what decides what gets built

5.1 Integration and clusters cut capital, storage, freight and risk

Brennan distinguishes horizontal integration, which joins similar operations such as several ethylene plants, from vertical integration, which moves a company upstream to its raw materials or downstream to its markets: an ethylene producer buying a vinyl chloride plant, for example (pdf p. 253).

The bigger gains come from putting linked plants on one site, so the product of plant A goes by pipe to plant B. Less storage of intermediates saves fixed and working capital. Transport costs fall. Surplus energy and waste streams can be swapped, and skilled staff shared. Safety improves too: intermediates stay inside the site instead of travelling by road or rail, and fewer containers are filled and emptied (pdf pp. 253-255). In Australian sites Brennan found storage of about 15% of annual production for stand-alone plants against 2% for integrated ones. The costs are tighter reliability demands, since one outage stops the neighbours, and the risk of domino effects in an accident.

Integration at large scale produces clusters. BASF's Ludwigshafen site has some 39,000 employees and about 200 production units on 10 km2; Kwinana near Perth had output of A$15 billion a year in 2008 (pdf pp. 255-257). Plants locate near raw materials or markets, rarely both; Japan, with few raw materials, competes through a large local market, technology and well-organised bulk buying (pdf p. 256).

The IEA sees the same logic with refineries: refinery naphtha feeds crackers, and cracker hydrogen feeds refinery desulphurisation. More than three-quarters of China's HVC capacity sits next to refineries and about 40% of European naphtha cracking is integrated, but under 20% of US naphtha cracking, where cheap ethane favours stand-alone crackers (pdf pp. 38-42).

• Brennan (2020), Ch 9 Industry planning and structure · pdf pp. 253-257
• IEA (2018), Ch 2 · pdf pp. 38-42

5.2 Uncertainty is handled with sensitivity tests, scenarios, staged options and portfolios

Every plant decision rests on forecasts that will be wrong. Brennan's first tool is sensitivity analysis: vary each input by 10% and see what happens. In his illustrative project, a 10% lower selling price cuts the return from 22.1% to 16.2%, 10% lower sales volume to 18.5%, and 10% higher capital to 19.6%, while 10% higher operating costs matter much less (pdf p. 156). This dominance of market factors is common in chemical investment. 'What if' tests then explore events such as a six-month delay, lost market share or an exchange rate swing.

The next step attaches probabilities. Weighting each outcome's NPV by its probability gives an expected value; an investment can be laid out as a decision tree of choices and uncertain events; and a Monte Carlo simulation draws many random combinations of inputs to give a distribution of NPV (pdf pp. 156-157). Few companies use these methods, and people committed to a project tend to be optimistic.

Three further defences matter. Staging, as in Example 6.6, keeps the option to defer or shrink later capacity. Competitive analysis checks that a project's cash cost would sit in the lowest 25% of world producers, so it survives a downturn (pdf p. 158). And a portfolio spreads risk across projects (pdf p. 157). Sunk costs should be ignored; each new decision is judged on its own incremental cash flows.

The IEA stresses that its scenarios are not forecasts (pdf p. 69). Brennan's list of why projects fail is a useful checklist: weak markets, excessive competition, feedstock shortfalls, technology limits, weak teams, rushed planning and delays (pdf pp. 241-242).

• Brennan (2020), Ch 5 Evaluation of project profitability, 5.9-5.11 · pdf pp. 156-158
• Brennan (2020), Ch 8, 8.8.4 and 8.9 · pdf pp. 237-242
• IEA (2018), Ch 4 Box 4.1 · pdf p. 69

5.3 Safety rules and accident costs shape plant size, inventory and siting

Kent's safety chapter starts from one observation: loss of containment, through mechanical failure or wrong operation, is the leading cause of property loss in the hydrocarbon and chemical industries (pdf p. 62). Process safety management, which treats safety as something managed like any other part of the business, took hold after Bhopal and became US law in 1992.

The most reliable protection is inherent safety, Trevor Kletz's idea of avoiding hazards rather than controlling them. Its principles include intensification (hold less hazardous material), substitution (use safer materials), attenuation (milder conditions), limiting effects through layout, and simplification (pdf p. 71). These choices are cheapest at the laboratory and concept stage; later they become expensive retrofits.

Major accidents drove the rules. Flixborough in 1974 released about 30 t of cyclohexane and killed 28; its lesson was to treat limiting inventory as a design objective (Brennan, pdf p. 200; Kent, pdf p. 114). Seveso (1976), Bhopal (1984) and Piper Alpha (1988) followed, and Longford in 1998, which cut Victoria's gas for two weeks, led to major hazard facility law in Australia (Brennan, pdf pp. 199-201).

The economics run both ways. Accidents bring compensation, lost business, rebuilding and higher insurance: Buncefield in 2005 led to an estimated £900 million in claims plus £70 million for rebuilding (pdf p. 199). Prevention costs money too, in safety studies, controls, layout and materials, but smaller inventories also cut capital, working capital and insurance (pdf p. 219). Brennan treats safety standards as a precondition of a project, not an extra, and notes that bigger plants hold bigger inventories (pdf pp. 192, 220).

• Kent (2012), Ch 2 Safety Considerations in the Chemical Process Industries · pdf pp. 62, 71, 114-115
• Brennan (2020), Ch 7, 7.9.2; Ch 8, 8.5 and 8.8.1 · pdf pp. 192, 199-201, 219-220, 234-235

5.4 Green chemistry metrics and carbon costs are entering the cost sheet

Green chemistry, as Kent and Wittcoff describe it, is the design of products and processes that reduce or eliminate hazardous substances. Its twelve principles include preventing waste rather than treating it, maximising atom economy, using catalysts instead of reagents consumed in the reaction, safer solvents, energy efficiency and renewable feedstocks (Kent, pdf pp. 170-172; Wittcoff, pdf p. 628). Atom economy is the share of the atoms fed to a reaction that end up in the product. The E-factor is the mass of waste per unit mass of product (Kent, pdf p. 172).

The numbers vary hugely. Pollak puts fine chemicals typically at an E-factor of 5-50 (pdf p. 102). Methanol carbonylation to Acetic acid, CH₃OH + CO → CH₃COOH, puts every reactant atom into the product; Brennan's cost sheet uses 0.54 t of methanol and 0.52 t of carbon monoxide per tonne of acid, so only about 6% of input mass is lost (pdf p. 126). Acetylene from calcium carbide, by contrast, produced about 28 t of calcium hydroxide slurry per tonne (Wittcoff, pdf p. 628), and classical Caprolactam makes 5 kg of ammonium sulphate per kg, against 1.8 kg in DSM's process (Kent, pdf p. 908).

Brennan treats environmental performance as a cost-benefit trade-off. Extra conversion stages in a sulphuric acid plant cut SO₂ emissions but add capital and energy, and because limits keep tightening, a design that just meets today's rule may need rework (pdf p. 218). Regulation can end technologies: European mercury cells fell from 60% of chlor-alkali production to near zero in 2001-17 (pdf p. 198). Carbon prices are the newest cost line; the UK set £16 per tonne of CO₂ in 2013 (pdf p. 198), and the chemical sector emits about 1.5 Gt of CO₂ a year (IEA, pdf p. 12).

• Kent (2012), Ch 5 Green Engineering · pdf pp. 170-172
• Wittcoff (2004), Ch 17 Sustainability and Green Chemistry · pdf pp. 627-628
• Pollak (2011), Ch 6 Research and development; Ch 7 Cost calculation · pdf pp. 92, 102
• Brennan (2020), Ch 7 and Ch 8 · pdf pp. 198, 218, 237
• IEA (2018), Executive summary · pdf p. 12

See it in the table

Check your understanding

Brennan found storage of about 15% of annual production for stand-alone plants. What did he find for integrated sites?

In Brennan's sensitivity example, which 10% adverse change hurts the return most?

What is the core idea of inherently safer design?