By Marthi Harmse
The youngest ORSSA SIG – History of OR in SA – was launched on 13 March 2024, just a day before Pi Day! We were privileged to visits behind the scenes of the early days of OR in the South African chemicals industry.
Our guide was Dave Evans, who spent the first 28 years of his career in the chemical industry. Dave grew up near York, England, obtained an MA in Biochemistry from Cambridge University and an MSc in Operational Research and Management Studies from Imperial College, London University, after which he joined ICI in England. In late 1970 Jim Buttery, OR manager at AECI in South Africa, was sent on a “technical” visit to ICI by his production director, Harry North, who was quite familiar with the use of OR in industry. Jim returned with among others John Hudson, Dave Bromley, Alan Hawcroft, Brian Fuller, Geoff Knights, James Grant, and Dave.
Some of the OR at AECI involved linear programming for monthly budgeting purposes, annual scenario planning, and capital investment decisions over twelve-year periods. The largest computer in Africa was utilised namely an IBM 360 at Anglo American’s Dedicated Data Centre with half a megabyte core memory. It could handle 800 variables and 600 equations, and completed a run in one hour. This required clever modelling since maintaining mass balances across 2 factories, 25 plants, 15 products and 6 sale regions potentially could involve 3 000 equations… Multi-time period models were built by hooking together up to twelve models through stock and cash flows. The objective was to maximise profit – ignoring pollution, energy consumption, global warming and similar complexities.
Discrete event simulation models were developed with the aid of IBM’s General Purpose Simulation Software (GPSS). Such models included a scheduling model that was developed to such a level of detail to include tea breaks, and when the first PCs became available another model was developed to determine the number of computers required by the company. Differential equation modelling also was done with the Continuous Systems Modelling Package (CSMP). These models decomposed into Fortran and then Assembler.
One example of an OR project – the “R100” project – involved meta-modeling of capital investment with uncertain returns. Effectively developing Excel in Fortran, a discounted cash flow model was constructed. Future market share was estimated for the current situation with most factories far away from customers, an expensive inflexible distribution infrastructure, weekly or fortnightly rail deliveries to customers, and each customer maintaining extensive storage facilities. This was compared with possible market share if an additional R105M (“R100”) was invested to construct new factories or distribution sites closer to customers with just-in-time deliveries. This would entail lower costs to customers and more difficulties for new entrants into the market. Legislative implications had to be considered for both situations. Iso-profit lines for the return on investment were developed and a best estimate was made. The investment went ahead and 40 years later the division is still in business. Although the optimality of the decision could never be proved, this is a good example of decision support where a decision without the model would have been largely guesswork.
The history of OR in the South African chemicals industry shows that operations researchers not only effectively support better decision-making. They also pragmatically support decision-making by intelligently utilising what is available. And they build long-term trust relationships with the decision makers. Thanks, Dave.
If you would like to share your story on the history of OR in South Africa, please contact Gavin le Roux, ORSSA Head of External Liaison. For further fascinating stories, join the SIG at here.
