• Open Access

Stretch: how innovation continues once investment is made



Process industries are characterised by large fixed items of capital equipment. This paper asks how innovation takes place once these plants are built. Stretch is the mechanism by which established plants incorporate improvements in process and product technology, and make higher output and new products as a result. A taxonomy of ‘stretch’ is proposed looking at five interrelated features: improved intensity of hardware use through experience and better maintenance; system-wide effects of improvements in feedstock and downstream processing; ‘bolt-on goodies’ and physical reconstruction of existing plants; and quality enhancement and new products. Intensity of use encompasses familiar learning effects, but also enhanced maintenance. Evidence for stretch is given for wide strip mills built under Marshall Aid, but the concept has wider application across process industries, manufacturing, transport and services. From a theoretical point of view, stretch is the expression of evolutionary problem solving. The practical implication is that research and development management should encourage cross-functional collaboration. Creative and unorthodox personnel need to work in routine areas such as maintenance. Results-orientated production managers need to accept risky interventions in production schedules to allow continuing innovation.

Stretch, v. fig. ‘to enlarge or amplify beyond proper or natural limits; to extend unduly the scope, application or meaning of’ 1553.

Shorter Oxford English Dictionary, vol.2, Oxford UP, 1973

1. Introduction – the reality of stretch

In 1966 a wide hot strip rolling mill was commissioned in Ghent, Belgium for an initial capacity of 1.75 million tonnes of steel strip per year. During the years 2007–2009, the same rolling mill produced 6.5 million tonnes of steel per annum – approaching four times its rated capacity. Recently the plant has run for a number of weeks at a rate equivalent to 7 million tonnes per annum. The Sidmar wide strip mill at Ghent is an example of ‘stretch’ – the ability of a process plant to produce output beyond its initial design capacity and produce a wider range of products than originally envisaged because of improvements in operating practice, developments in ‘bolt-on’ equipment and changes to feedstock after it has been commissioned.

Here we attempt to answer the question: how does innovation take place in process industries characterised by large items of fixed capital equipment once those plants are in place? As Thomas Lager (2002) points out, incremental development and refinement of existing products and processes is the most common form of innovation in manufacturing industries such as metals, building materials, glass, paper and chemicals. Yet how does innovation continue once the commitment has been made to build a large capital item? How does existing kit get upgraded once it is in production?

We define stretch as the continual modification of a plant, system or service beyond its initial design specification, with the main aim of increasing capacity, but also to increase product range and quality and use new inputs. There may be incidental advantages from stretch such as better environmental performance or improved working conditions. But the focus here is on modifying and upgrading existing equipment or services to make more output.

Stretch was traditionally seen as a way of increasing throughput in chemical production (Coe, 1962). Maxwell and Teubal (1980) suggest that stretch is often pursued when capital investment is not available. Here we use ‘stretch’ in a wider sense as the potential for further expansion, development and upgrading of existing plant in the process industries, including capital spending on reconstruction. Stretch is the mechanism by which established plants incorporate subsequent improvements in process and product technology and organisational innovation. It is making the most of the material immediately to hand – existing equipment (Miner et al., 2001).

The paper considers post-Second World War-wide strip mills supplied under the terms of Marshall Aid. This has the advantage of longevity, giving a clear perspective on retrofitting of technology developed over the subsequent 60 years. We show that obsolescence is not an automatic process of decay and replacement. Instead, old technologies remain in widespread use, upgraded and shaped to modern uses (Campbell, 2009).

Admittedly, there is occasional and severe discontinuous technical change in process industries, which requires complete new plant items and premature scrapping of old technology. Oxygen steelmaking, continuous casting of steel, tunnel kilns for brickmaking, float glass and dry process cement plants are all examples of radical innovations that diffused rapidly, supplanting existing techniques (Lacci et al., 1974; Aylen, 1980; Utterback, 1994; Alsop et al., 2007). In these circumstances investment in new plant brings substantial productivity gains (Sterner, 1990).

These breakthrough innovations themselves open up possibilities for further development. For the most part, process plant technology displays great longevity once built. The fluid catalytic cracking process for oil refining is a case in point. Indeed Yin (1994) argues returns from follow-on incremental improvements to a petroleum-refining innovation exceed those from the initial radical process innovation itself.

Existing technologies also evolve as successive new plant is built. These new techniques can be retrofitted to earlier equipment. Continuous wide strip rolling technology for steel moved through five generations of development from its inception in 1926 (Aylen, 2001; Aylen, 2010). Mill designs reflect growth prospects and relative prices of their time. The story of hot strip mill development is one of gradual evolution, with occasional step changes of direction – evolution as the engineers pursued the next logical technical improvement, changes of direction as economic imperatives forced a re-evaluation of the path of technical development. Each generation of mill design accumulated technical developments of the past, but added more in response to changed circumstance. The first three generations of wide strip mill pursued economies of scale in growing markets. High-capacity Generation III strip mills were a product of optimistic growth prospects, abundant capital and cheap energy before the oil crisis of 1974. The trajectory shifted towards compact layouts with Generation IV and V. Generation V mills cut capital outlays and energy usage further by casting thinner steel slabs, an approach that is viable on a small scale. The latest small mill, Cremona 2 (Arvedi et al., 2010) is migrating towards a continuous flow process for rolling, thereby expanding capacity by increasing throughput, notably by higher casting speeds to increase mass flow to the rolling train. One theme has been consistent, the move towards heavier piece weights as coil sizes have steadily increased over time. Innovations such as heavier coil weights can be retrofitted to earlier generations of equipment.

Stretch is important for competitiveness. Evidence suggests that plants that increase output over time and keep up with new technology are those that survive. A study of 17,000 manufacturing plants in Portugal finds current size and past growth are key determinants of plant survival (Mata et al., 1995). Again, a survey of 6,090 US manufacturing plants during the 1987–1991 period shows that capital-intensive plants and plants employing the latest technology are less likely to fail (Doms et al., 1995). Baldwin and Rafiquzzaman (1995) look at survival rates of new plants over two decades in Canadian censuses of production. They discover that new plants that were smaller and less efficient did not survive, and those that did survive this selection had improved their performance relative to established firms through a learning process.

Everyday experience in process industries stands in marked contrast to economic theory on capital vintages (Salter, 1966). Economics quite simply assumes that new plants incorporate the latest technology. Subsequently, equipment deteriorates with age and use, and becomes relatively obsolete as new vintages of technology emerge (Fisher, 1963; Gort and Boddy, 1967; Shone, 1975; Gomulka, 1978). Kenneth Arrow (1962, p. 157) caricatures this putty-clay model: ‘at any moment of time, the new capital goods incorporate all the knowledge then available, but once built their productive efficiency cannot be altered by subsequent learning’. Change is further precluded because one item of sunk investment is interdependent on another (Frankel, 1955). Evidence from a range of steel-production processes (Wolf and Van der Rijst, 1982) suggests that technical advance has progressively raised the output and cut the capital costs per tonne of capacity of new equipment. But there is a clear assumption that all technical change is embodied in new equipment (Ruth and Amato, 2002), quite simply overlooking the extent to which existing equipment is both reorganised and modified in use.

Moreover, a new plant does not always adopt best practice technology. Gregory and James (1973) found a wide dispersion of productivity across new Australian factories. The productivity of new plants was no different from existing factories. Productivity growth reflected the whole past history of the capital stock, not the performance of the latest plants. So, the idea that successive vintages of capital equipment incorporate the best practice of their day and inevitably supplant existing equipment is not borne out by the evidence. Yet, as Edgerton (2006, p. 8) points out, ‘the assumption that the new is much superior to older methods is widespread’.

Instead, this paper develops a taxonomy of ‘stretch’, unpicking the sources of performance improvement that enhance output and increase the product range of established items of process plant. Stretch is related to key innovations in the process industries. Some of these innovations are quite generic, such as replacement of mechanical actuators by hydraulic devices, digital control systems or high-power microprocessor management of variable speed alternating current motors. Other innovations are highly specific to a particular installation or process, such as the quick roll change rigs and coilbox used on wide strip mills. The taxonomy of stretch covers interlocking changes in hardware, software, experience, intensity of use (including scheduling and maintenance), improvements in preparation of raw material feedstock and improvements in specification of finished products.

Future stretch is often anticipated at the design stage, before capital investment begins. Initial designs can provide for future staged expansion (Earnshaw, 1978), allowing capital spending to be staged in line with future market growth. Unrealised potential for expansion may be a credible threat deterring new entry by competitors (Dixit, 1980). A likely competitor would think twice before entering a sector where established firms can readily expand existing plants, raise output and stage a price war backed by latent new capacity.

Other sources of stretch are unforeseen by initial designers, builders and operators. Some improvements occur by plant level invention, such as the predictive software developed by Voest on their Linz wide strip mill (Luger and Hubmer, 2001). Others arise from generic developments outside the industry.

2. A taxonomy of stretch

Our taxonomy of stretch is based on a five-way classification (Table 1). This distinguishes between improved management and organisation, whether capital spending is involved, and considers upstream and downstream system-wide effects as factors influencing the central plant unit under study. Finally, we focus on quality improvements and production of new products unforeseen when a plant was built. So this classification looks at five interrelated features: firstly, improved intensity of hardware use through experience and better maintenance. Secondly, system-wide effects of improvements in material feedstock and downstream processing. Thirdly, we consider ‘bolt-on goodies’, including improved instrumentation or control systems. Fourthly, we consider physical reconstruction of existing plants. Finally, these improvements often enhance quality and offer scope for making novel products. These features of stretch are connected. Capital spending may be needed to accommodate heavier piece weights on rolling mills, for instance. New roll technology, such as high-speed steel rolls, are needed to make high-yield strength low-alloy steels.

Table 1. A taxonomy of stretch
Management and operator skill
 1. Pure learningOperating experienceEnhanced throughput when operating and wider product range
 2. Changes in maintenanceMostly experience, better monitoring and schedulingShorter downtime, enhanced plant availability and better product quality
Embodied in extra capital
 3. ‘Bolt-on goodies’Additional modules of equipment (e.g. instruments and controls)Higher output, better quality and reduced downtime
 4. Reconstruction of existing equipmentReplacement and upgrading of core facilities (e.g. furnaces and motors)Often higher output through removing bottlenecks
System-wide improvement – upstream
 5. Better availability of feedstockBetter scheduling of materials supply, more consistent supplyMore regular operation
 6. New feedstocksNew materials, different dimensionsHigher yield and heavier piece weights
System-wide improvement – downstream
 7. Better ‘take’ of materialsBetter scheduling, warehousing and product logisticsLonger production runs
 8. New capabilitiesLarger capacity items for downstream processingHigher yield and heavier piece weights
New products
 9. New qualitiesImproved quality of existing productse.g. Better surface finish, gauge, shape and dimensional accuracy
10. New productsComplete new productse.g. High-strength low-alloy steels, transformation-induced plasticity steels

2.1. Pure learning

One aspect of stretch, the notion of ‘learning’, has been widely studied. Learning is the ability to reduce unit costs or raise output with cumulative operating experience (Argote and Epple, 1990). But the concept of pure learning does not capture the heterogeneous mix of control equipment, capital spending, upstream innovation, additional equipment, feedstock changes and alterations in product specification that explain the steady expansion in processing capacity and product range of existing plants. The literature on learning curves or experience curves conflate a number of factors that enhance productivity over time (Hart, 1983). For example, Van der Rijst et al. (1978) suggest that maximum productivity of blast furnaces for ironmaking is directly related to production experience and claim that there is a constant rate of learning across Europe. This overlooks frequent upgrading of blast furnaces that takes place during every periodic re-line. New refractories, new stoves, better top equipment, higher power blowers, more efficient water cooling, new clay guns, better cast house equipment, new instrumentation and computer control are routinely adopted during these regular and prolonged stoppages for reconstruction. For instance, blast furnace 3 at SSAB, Luleå, Sweden was drastically rebuilt in 75 days during 2000 using prefabricated segments to give a modern 11.4-m hearth furnace. Regular upgrading has been spurred on by tighter environmental regulation. In short, experience is just one part of overall ‘stretch’. Other technologies are upgraded at regular intervals during rebuilds. Cement kilns or oil refineries typically run for 330 days between stoppages. Float glass tanks are relined every 7 years or so. These regular outages for prolonged maintenance give an opportunity for retrofitting equipment or judicious expansion of individual plant items.

Learning how to operate is the most widely discussed aspect of increasing the intensity of use, although there is little discussion of how it actually happens (von Hippel and Tyre, 1995). At its heart is the growing confidence in operation and assembly, which improves the pace of operation. As a rule of thumb, doubling output seems to be associated with a 20–30% unit cost reduction. Lieberman (1984) estimates learning curves for 37 different commodity chemicals and metals across six broad product categories using price as a proxy for cost. He finds that cost reductions are related to cumulative output and the capital intensity of the process concerned. He concludes that research and development (R&D) spending accelerates the learning process. Higher R&D intensity interacts with cumulative output. The cost advantage was shared by batch and continuous processes alike. Evidence for 221 special chemicals within one large company suggests that growth in past output is a strong signal of future returns to R&D expenditure, and so reductions in unit cost may actually reflect incentives to invest in process research (Sinclair et al., 2000).

Learning continues over the life of a plant. One factor is better scheduling (e.g. mixed rolling of electrical steels and carbon steels on wide strip mills), which is enhanced by the development of new computer programmes to optimise plant utilisation (Cowling, 2003).

Attempts have been made to widen the notion of learning towards a progress function. A wide ranging survey by Dutton and Thomas (1984) disaggregates the performance of firms into pure learning, technical progress, local industry and firm characteristics, and scale effects. But these broad categories do not capture the nitty-gritty of how plants actually get modified over time.

2.2. Changes in maintenance practice

Maintenance is a neglected topic. As Edgerton (2006) argues, ‘although central to our relationship with things, maintenance and repair are matters we would rather not think about’. Yet, maintenance is the third largest item in a steelmaker's budget, after raw materials and manpower (Aylen, 1989). Annual maintenance costs for chemical plants are typically 5–15% of fixed capital costs, divided evenly between labour and materials (Sinnott, 2005, p. 262).

The military were among the first to appreciate the output gains brought by maintenance planning. Switching Royal Air Force Coastal Command aircraft to ‘planned maintenance’ overcame a shortage in skilled labour, increased their flying hours per maintenance worker by 43% and raised their operating time from 140,383 flying hours in 1942 to 200,558 flying hours in 1943 (Air Publication, 1954, pp. 171–176).

Reduction in scheduled maintenance time from once a week to once a month typically releases an extra 12% potential capacity (assuming a cut from 49 days to 11 days with a 2-week annual shutdown). This is usually achieved by better planning and scheduling techniques. A rare comparative study on maintenance practice from the International Iron and Steel Institute (IISI, 1989, 3.23–3.24) argues that ‘detailed planning is recognised as having more effect on maintenance productivity than any other single factor’, highlighting the way in which organisational innovation and technological innovation interact in process industries.

Online maintenance while a plant is still in operation reduces down time. In the cement industry, the kiln is the critical feature. Hot alignment techniques allow corrections to kiln alignment while it is operating (Saxena, 2009, p. 19).

Reductions in unscheduled delays are more important because unplanned stoppages destroy the ‘rhythm’ of a working plant. Detailed results are available for Kimitsu hot strip mill in Japan over an 18-year period between commissioning in 1969 and 1987 (IISI, 1989, 3.22–3.23), which show a continuing reduction in unexpected stoppages from 74 hr a month in the first year of operation down to 13 hr in the 19th year of operation. Production over the same period grew by 24%. Most of the changes were attributed to training, changing approaches to work (e.g. quality circles), eliminating demarcations and reorganising maintenance. A regular pace of rolling mill operation brings a higher yield of finished steel from raw material, for instance through lower slag losses on reheated slabs (IISI, 1992).

2.3. ‘Bolt-on goodies’ – improvement of existing plant

Incremental process improvements are colloquially known as ‘bolt-on goodies’. These are modular items that can be attached to existing equipment often developed in collaboration with a specialised equipment supplier, electrical engineering company or bespoke software house (Hutcheson et al., 1995; Ozman, 2011). They are usually trialled on existing plants before being specified on new equipment. They are often supplied as ‘vendor packages’ (Hutcheson et al., 1996) from an equipment supplier who is responsible for the integration of the new equipment into the existing power supply, process control and instrumentation. The development of ‘bolt-on goodies’ highlights the importance of machinery suppliers as sources of innovation (von Hippel, 1977; van Rooij, 2005; Reichstein and Salter, 2006; Lager and Frishammar, 2010). A high level of retrofitting explains why Salvanes and Tveteras (2004) find that there are marked differences between the age of a works and the age of its equipment in a survey of Norwegian manufacturing plants (Dunne, 1994).

There are at least 30 readily identifiable bolt-on developments used to enhance existing wide hot strip mills over the past 50 years (Table 2). These include modifications at every stage of the production line from reheat furnaces (notably large walking beam furnaces that lift and carry slabs through heating zones) right through to coilers. Some rely on generic technologies, such as the use of hydraulics to manipulate roll screw-downs, edger rolls and coiler wrapper rolls. Some are highly specific to rolling wide strip such as gauge metres, shape metres and width metres for hot material. Roll developments might add another 10 incremental innovations. There have been numerous product developments, too, with perhaps 80% of the product range rolled on a modern wide strip mill having been developed in the last 20 years. This helps explain the predominance of quality-enhancing investments among the list in Table 2. But an incidental side effect of these quality-enhancing innovations has been an increase in mill output through faster rolling, higher yield of prime quality material and lower downtime because of maintenance.

Table 2. ‘Bolt-on goodies’ used to modify wide hot strip mills since 1960
  1. Omitting false-start technologies, local developments such as six high hot mills in Japan, roll improvements, a wide range of new products and major changes in mill configuration such as thin slab casting.
 1. Walking beam reheat furnacesOutput, energy saving and quality gain
 2. Direct hot chargingEnergy saving
 3. Computer control of reheat furnacesEnergy saving
 4. New burners (e.g. pulsed, atmosphere control)Energy saving and yield gain
 5. Waste heat recuperationEnergy saving
 6. Lift and carry furnace dischargeQuality gain
 7. Slab skid mark compensationQuality gain for slabs from pusher furnace
 8. Hydraulically controlled vertical edgers for slab shapingYield gain
 9. Hydraulic screw downQuality gain and yield gain at coil ends
10. High pressure descalersQuality gain
11. Delay table covers (e.g. ENCOPANELS)Energy saving
12. CoilboxOutput, quality and product range
13. Strip edge heaters (not common)Roll wear, so output
14. Quick roll change rigsOutput and obviate mill damage
15. Zoom-rollingOutput-offsetting thermal rundown of coil
16. Back-up roll bending and work roll bending for crown controlQuality
17. CVC mills for crown controlQuality
18. Interstand coolingOutput and product range
19. Hydraulic loopersMaintenance and quality
20. Work roll side shifting to distribute edge wearRoll wear, so output
21. AC motors under high-power digital controlEnergy saving
22. Automatic gauge controlQuality, but also output
23. Computer control (successive generations)Savings all round and output gain
24. Laminar flow coolingQuality
25. Width metersQuality
26. Shape metersQuality
27. Automatic crop controlOptimising yield, so output
28. Revolving mandrel coilersQuality and yield
29. Hydraulic wrapper rollsQuality and yield
30. Lap detectionQuality
31. Close-coupled coilers (not common)Product range

We illustrate three of these modular improvements to wide strip mill technology in more detail in Boxes 1-3. These innovations have been retrofitted to existing equipment to enhance capacity, improve product quality, cut energy use or more generally to obviate the need to invest in a complete new installation. Two of these examples require substantial mechanical engineering enhancements; the third is a software improvement.

Box 1. The Stelco Coilbox – increasing piece weights

The Stelco Coilbox was prompted by the conceptual planning of a new greenfield works on the north shore of Lake Erie, near Dover, Ontario. Stelco wanted to build the mill as cheaply as possible. The mill was laid out for in-line rolling directly from the continuous caster to save energy. Overall mill length was reduced by the use of a novel Stelco Coilbox in place of a lengthy delay table between the roughing stand and finishing train. This improved energy efficiency and reduced the capital costs of the mill. Obviating the delay table cuts, perhaps 75 m off mill length. Lake Erie was a logical way to develop conventional and long-established North American hot strip mill technology in order to pare initial capital costs to the bare minimum (Carroll and MacNeil, 1985).

The coilbox simply receives the hot breakdown strip in the final pass from the roughing train as a large, open coil. In effect, coiling is a temporary storage device. Coiling the strip obviates a long, horizontal delay table. Coiling also promotes heat transfer so that the bar is at a more uniform temperature.

Coilbox operation starts with the transfer bar from the roughing train being captured and deflected through bending rolls so that the hot material forms an open, loose coil on a simple cradle of rolls. The hot coil is then paid-off into the finishing train in the reverse direction. The cooler tail end of the breakdown strip now becomes the leading end as it enters the finisher, making for a more even distribution of temperature along the length of the strip. Finally, the coil is moved sideways once it has started to feed into the finishing train so that the cradle is free to receive the next transfer bar from the roughing train.

The coilbox was conceived by William Smith. It is a tribute to the creative pressure imposed on good engineers by shortage of resources. It was trialled as a sequence of prototypes on Stelco's existing Hilton Works hot strip mill in Hamilton to the point where it was suitable for commercial use. Commissioning of Lake Erie works was delayed for financial reasons. Les Gore of steelmaker John Lysaghts Australia saw the prototype at work and adopted the idea for their new wide strip mill in Australia (Gore and Shegog, 1979). Thereafter, the Stelco Coilbox was widely adopted in Europe and Canada both as a technique for saving energy and improving strip quality through more uniform heating. It is an ideal way of rebuilding old roughing trains as the original delay table can be cut and the space used for two widely spaced roughing stands rolling a heavy slab. By the mid-1980s Stelco had made $20 million in royalty on an initial $1 million outlay on R&D. By 2000 some 43 coilboxes had been installed – half since 1990.


US Patents 3803891, filed 13 November 1972 and awarded 16 April 1974, and US Patent 3805570, filed 13 November 1972 and awarded 23 April 1974, both invented by William Smith and assigned to the Steel Company of Canada Limited, Toronto, ‘Method for rolling hot metal workpieces’ and ‘method and apparatus for rolling hot metal workpieces and coiler for use in coiling hot metal workpieces’.

Box 2. Quick roll change rigs – reducing mill down time and damage

The original turntable roll change rig patent filed in 1960 was an output-enhancing innovation (US patent 3208260). As the preamble states:

Not only is it extremely desirable to provide a roll changing system that will not require the use of any cranes and one that will reduce the manual assistance necessary, but even of more importance is the need for keeping at a minimum the downtime of the mill incident to roll changing. In the case of a six stand tandem mill, it is not uncommon to experience more than a forty-five minute delay in changing work rolls thereof. Moreover, this operation usually is performed as frequently as every two to four operating hours so that the aggregate of the lost production time is most costly.

As with many technical developments, the quick roll change rig was prompted by the needs of a user, in this case Inland Steel of Chicago, who wanted to change rolls in 10 min. Inland approached the plant supplier United who made little progress on a solution for 2 years until they came up with the idea of using a turntable. (Characteristically, Jim Adair said the idea came to him sitting on a revolving bar stool.) The quick roll change device met its brief. Ess (1970, pp.82–86) reports that a roll change on one finishing stand of a hot mill could be accomplished in 2 min and a full set changed in 12–15 min. The advent of quick roll change rigs was said to increase mill output by 5–8% compared with a precarious conventional C hook or porter bar roll change.

Quick Roll Change Rigs

US Patent 3208260, filed 18 August 1960 and awarded 28 September 1965, inventors Maurice Paul Sieger and James R. Adair, assigned to United Engineering and Foundry Company, Pittsburgh, ‘Rolling mill’.

Box 3. Real time prediction of metallurgical transformation – more uniform quality and higher production rates

Direct digital control of a complete wide strip mill began at Llanwern in South Wales in 1963 (Aylen, 2004). This computer controlled the whole process in real time, tracked material through the mill and logged the results. By 1970, direct control of strip mills was widespread in the United Kingdom and Germany. Attention turned to simulation and control of product quality (e.g. Van Ditzhuijzen, 1993). During the 1990s, large-scale computer models were developed that allow the integration of scheduling and process control with quality assessment. This required a real-time model of metallurgical transformation during the rolling process. Developments by Voest-Alpine Stahl and VAI at Linz allow immediate predictions of quality for the whole length of a rolled coil (Luger and Hubmer, 2001). Linz makes a range of demanding products to high standards and sells profitably to sophisticated customers, especially for automotive applications (Marsh, 2011).

The starting point is a physical-metallurgical model to predict strip quality in terms of tensile strength, yield strength and elongation. This modelling project began in November 1995. Once it was established that an off-line model gave accurate predictions of the mechanical properties of hot-rolled coil, the model was used to actually control set-up and cooling on the finishing train from January 2000 onwards. Strip varies throughout its length. So the model tracks each segment of the strip so that microstructure can be predicted and modified during final rolling and cooling. Optimum rolling and coiling temperatures are crucial for high-strength low-alloy steels, for example.

There are commercial advantages from being able to predict the quality of each coil straightaway. The coil is passed on for further processing immediately. The whole length of the coil is ‘checked’ by inference, whereas conventional measurement is restricted to samples from the head and tail end, which may be untypical. There are raw material savings as it is possible to predict mechanical properties such as tensile strength, rather than relying on expensive alloy additions at the steelmaking stage.

The examples include quick roll change rigs, which were almost universally adopted as an output-enhancing innovation between the mid-1960s and the mid-1980s. Coilboxes require more substantial reconstruction and changes in operating procedures, but they offer the chance to make much heavier coils on existing mills with short delay tables. Predictive modelling of microstructure is one way in which steelmakers supplying the car industry have been able to satisfy more demanding quality requirements in terms of yield strength and surface finish while maintaining the pace of mill operation. Predictive modelling is a highly R&D-intensive retrofit.

Retrofitting of modular items is not unique to steelmaking. Cement plants have upgraded clinker coolers with higher pressure static coolers, added heat exchangers to cooler dust collection to enhance capacity and switched to new materials for dust collection in bag houses (Alsop et al., 2007, pp. 190–192).

2.4. Reconstruction of existing plant

A more drastic way of improving plant is reconstruction. A number of European hot wide strip mills have been converted from continuous layout to more efficient three-fourths continuous or semi-continuous layout, which brings greater utilisation of the finishing train by better scheduling of the roughing train and use of heavier incoming slabs (Aylen, 2001). Reversing roughing stands improve heat distribution in the breakdown bar. This modification in layout is often associated with the construction of larger walking beam reheating furnaces and, sometimes, the use of a Stelco Coilbox on the delay table. Reconstruction typically involves mechanical preparation and civil engineering underneath the roughing train while the plant is still in operation followed by an extended summer shutdown. This approach has seldom been used in Japan or the United States, where hot strip mills remain as built, suggesting that stretch is more widely practiced in Europe.

2.5– 8. System-wide improvement – upstream and downstream

One of the easiest ways to stretch a plant is to increase the unit size or ‘piece weight’ of the inputs used by the plant. The Sidmar mill discussed in the opening sentence receives thicker slabs than it was designed for, resulting in longer and heavier coils produced by greater reduction along the mill. Aylen (1982) finds striking differences in capacity between UK and German wide hot strip mills, partly reflecting higher drive powers and heavier coil weights of German mills. Heavier coils also make for cheaper processing both within the works and on customers' facilities. One variant is to roll thicker gauge material on the hot rolling mill and rely on heavier reduction in subsequent downstream cold rolling facilities.

A key factor behind stretch in the case of the Marshall Aid mills turns out to be heavier piece weights. The Linz wide strip mill was built in 1952 for 10.7-tonne slabs. The mill now rolls coils up to 32 tonnes maximum, a threefold increase. Here again stretch conceals a sequence of technical innovations retrofitted since the plant was built: adoption of continuous casting, higher capacity reheat furnaces, more powerful edging rolls on roughing mills, extra finishing stands, increases in motor power, higher capacity coilers and more robust coil handling facilities.

Use of larger inputs to an existing process system is not confined to mechanical equipment. A study of Heathrow airport shows that existing runways, taxiways and terminals were stretched by airlines using larger planes to overcome capacity constraints (Tether and Metcalfe, 2003). In this way, slots were utilised more intensively through system-wide modification of operating procedures. Mixed approach paths were used by traffic control to overcome the turbulence problem brought by larger, more powerful engines of bigger jets, in the same way that wide strip mill operators mix rolling of silicon steels at ultra-high temperatures with tough stainless steels as alternate slabs.

A rolling mill designer (Weiss, 1978) sees the process of technical improvement as one of breaking successive bottlenecks. Once one constraint on throughput has been overcome, attention shifts to the next. So the process of stretch is one of stepwise continual improvement to release the potential of latent capacity upstream or downstream from a problem area. Each scheme to enhance output then reveals another constraint elsewhere. Constraints on technical change themselves throw up new opportunities. While the nature of improvement is piecemeal, the overall effect over a long period of time is a substantial enhancement of the productive capacity of the overall plant. In the same vein, Alsop et al. (2007) write:

De-bottlenecking: this repulsive but descriptive term is a mandatory focus for cement plant engineers. Every process has one or more capacity limitations. Identification of the limiting equipment is the first step.

Stretch also involves shifts in inputs in response to availability of materials, change in prices and downstream quality requirements. The British chemicals firm Imperial Chemical Industries offers an extreme case of changing feedstock where ammonia production at the company shifted from water gas and producer gas made using hot coke, to steam reforming of by-product hydrocarbons, to steam reforming of cheap naptha, all within 30 years (Gard, 1966; van Rooij, 2004, chapter 9). Papermakers have retrofitted process lines to de-ink and pulp waste paper as a supplement to conventional wood pulp (Engstrand and Johansson, 2009). Adoption of supplementary processes in existing plants is not always successful. As part of an expansion plan begun in the year 2000, the Terni wide strip mill of ThyssenKrupp Stainless supplemented the rolling of traditional thick continuously cast stainless slabs with thin slabs cast on a new machine feeding a long tunnel furnace (Brascugli et al., 2002). This required mixed rolling of alternate thick and thin slabs, but the interleaving did not work and the new machine was dismantled.

2.9– 10. Quality and product range

Stretch can also focus on changing product mix to meet shifts in demand. The petrochemical industry responded to rising demand for propylene, usually made as a by-product of ethylene manufacture, by developing ‘on purpose propylene’ technologies, for instance using Olefins metathesis (the metal catalysed redistribution of carbon–carbon double bonds). ABB Lummus' Olefins Conversion Technology uses this approach by adding an extra reactor to an existing cracker (Plotkin, 2005). The addition of an extra process stage is analogous to the near universal adoption of ladle steelmaking since 1980, which provides a holding buffer and refining vessel between steel melting and continuous casting, thereby allowing more efficient scheduling and production of a much wider range of steel products (e.g. Price, 2007, pp. 392–395). Ladle furnaces provide a consistent flow of batches of molten steel at the precise temperature and metallurgical composition required for continuous sequence casting.

Stretch also expands product range. Making ultra-low carbon electrical steels is a step towards making the more recent interstitial-free ultra-low carbon steels for car body sheet. The spread of product range reflects a broader point made by Malerba (1992) that learning adds to the stock of knowledge and technical capabilities of the firm, which then opens up a range of opportunities for technical advance, not just cost reduction. The Linz wide strip mill considered here opened up the development of high-strength formable coated steels for Voestalpine, which puts the firm at the forefront of steel strip supply for auto bodies (Marsh, 2011).

3. The Marshall Aid mills – hot strip and cold war

The continuous wide hot strip mill was an American invention, pioneered by Columbia Steel, Pennsylvania in 1926 (Aylen, 2010). By 1953 there were 55 continuous wide strip mills and Steckel mills, and three continuous plate mills in existence or under construction worldwide (BISF, 1953). All but three mills worldwide were of American construction.

American process plant builders emerged from World War II with a technical and commercial lead in many fields, notably steel and aluminium, cement and petrochemical plant. Basic equipment was needed to support post-war reconstruction and growth. The European Recovery Programme – colloquially known as Marshall Aid – funded the acquisition of US technology for reconstruction and defence. The programme had twin objectives of defending Europe from communism and tied-Aid easing the transition of American heavy industry to peace time production.

The European Recovery Programme led to seven major hot strip mill orders for US plant suppliers United and Mesta (Ranieri, 2012). These were Generation I mills, a design that remained unchanged to 1960 (Ess, 1941). Some of these mills were built with low capacity, such as IJmuiden, Linz and Cornigliano, while Sollac and Port Talbot were standard US style continuous wide hot strip mills supplied by United.

Three of these rolling mills are still operating 60 years later (Table 3). The remaining four closed after an operating life of over 30 years. The median increase in capacity over their working lives was 80%. Typical stretch in capacity across all the Marshall Aid mills is actually less than 2–a median of 1.8. One reason for the relatively low growth is the subsequent construction of high-output Generation II mills by the same company with first claims on orders and steel supply. Another reason is the maturity of the European steel market after the first oil crisis and recession of the mid 1970s. It is easier to energise changes when there is pressure on capacity.

Table 3. The Marshall Aid wide strip mills in Europe
MillCommissioned (US builder)StatusRated capacity when built (ktpa)Capacity at closure/now (ktpa)RatioAge (years)
Usinor, DenainMarch 1951 (United)Closed March 19851,3002,0001.534
Port TalbotJune 1951 (United)Operating, semi-half continuous1,8003,000 (estimate)1.761
Linz, AustriaJuly 1952 (Mesta)Operating, semi-half continuous6005,000 (in 2010)8.360
Breedband, IJmuidenOctober 1952 (United)Closed November 19856001,1001.833
Sollac, SérémangeJanuary 1953 (United)Operating, M-stand added 19831,8003,5001.959
Cornigliano, Genoaend 1953 (Mesta)Closed 19849001,700 (peak output)1.831
Ougree-Seraing1954 (Mesta)Closed 1980s8001,4001.830–35

The Port Talbot mill operated in standard form from commissioning in June 1951 for over 30 years. Ingot casting and slab rolling were replaced by continuous casting of slabs. There was a major reconstruction during the mid-1980s to a semi-continuous layout (Cook, 1979; Bryant and Dimblebee, 1986). Two walking beam reheat furnaces replaced the pusher reheat furnaces. A single, large, reversing rougher was installed in place of the continuous roughing train, a Stelco Coilbox was installed ahead of the finishing train and two new 34-tonne coilers put in behind the existing coilers, more than doubling the coil weight (Kidd and Dimblebee, 1987). The finishing train was little altered apart from an extra stand and new motors. The mill production manager was also the project manager to ensure continued mill operation during radical reconstruction.

The outstanding performance of the Linz mill is striking (Box 4, Figure 1). Linz is the only wide hot strip mill in Austria. It has a commanding market among the car manufacturers and white goods makers of southern Germany and northern Italy. The mill continually developed through the gradual accretion of additional innovations, which permitted heavier piece weights brought by improvements in upstream slab supply and downstream cold rolling, and finishing facilities and higher quality products. Box 4 lists over 30 significant performance-enhancing modifications for the Linz mill.

Figure 1.

Stages of modernisation of the Linz wide strip mill.

Box 4. Marshall Aid mill – the wide hot strip mill at Voest, Linz

The semi-continuous wide strip mill at Linz was built as part of the European Recovery Programme. The mill was supplied by the Mesta Machine Company of Pittsburgh under the terms of Marshall Aid. The wide strip mill for Linz was a revival of a pre-war Mesta contract for a strip mill at the Hermann Goering Works in Linz. A wartime armour plate mill was available for roughing down feedstock and a temporary sheet mill was installed after the war. But Voest was still keen to acquire the American continuous wide strip mill planned before the war. The contract was controversial because of concerns that the site beside the Danube was close to the Russian sector in Austria and might fall into Soviet hands in the event of conflict (Tweraser, 2000, pp. 312–313). The works itself is famous because it was the first location worldwide to use the now dominant basic oxygen steelmaking process.

Only Manfred Wirth, a salesman with Voest, could get a visa to visit the United States to progress the order for the wide hot strip mill in March 1949 (Wirth, 2001), as other nominated individuals were on the US ‘watch list’. Wirth took advice from Stelco in Canada and ordered a wider finishing train from Mesta than envisaged pre-war – 66 inches – and one fewer finishing stand, making five in all, but with more powerful motors than those specified in the pre-war order. He also bought a slabbing mill from Mesta, one 100-tonne per hour reheat furnace, a flying shear from Wean and a single downcoiler. Wirth played a key role in persuading the Organisation for European Economic Cooperation to give Marshall Aid support to the mill in August 1949, reversing an earlier decision to exclude the Austrian mill from funding. The 66-inch semi-continuous hot strip mill was completed in July 1952, with an initial output of 280,000 tonnes of plate and sheet per year (Lovay, 1957). The same rolling train now makes 5 million tonnes.

Linz runs 60 years on, ostensibly little changed as a semi-continuous wide strip mill. It is a leading supplier of high-quality car body sheet to the German and Italian carmakers. Marked increase in output and quality over these intervening years are due to small changes in plant configuration, upstream product supply and downstream processing. Major gains in throughput have been realised through heavier piece weights – larger slabs and coils. Advances in software and small enhancements to equipment coupled to higher plant availability have helped increase output and product quality.

Modernisation stages Linz hot wide strip mill
1953Basic semi-continuous mill with five finishers and one fixed mandrel coiler (10.7 tonnes)
1956Second 100-tonne/hr pusher reheat furnace
1957Sixth finishing stand added
1958Edger ahead of roughing stand
1959Third pusher furnace 100 tonne/hr
1968Two more pusher furnaces
1974/6Two large pusher furnaces 350 tonne/hr, leaving four in total
Extra finishing stand at F0
Roughing mill with attached edger
Down coilers 3 and 4, three wrapper roll on extended run-out table
Process computer for the finishing train
1981Down coiler number 5
1984Hot storage boxes for slabs
1985Roughing mill computer
1987New edger
Skid mark compensation
Reheat furnace automation
1989Profile control
1991Water descaler
1992Furnace riders
Low-NOx Burners for 350-tonne pusher furnaces
1994/5New cooling lines on run-out table
1995Basic automation
1996Two new thickness gauges
New fume collection for stands F4 to F6
New crop shear – two blade rotary shear
New entry guide in front of finishing train
Cobble pusher on delay table
Pinion stand for rougher and F1
1997Water treatment plant
2010Coiler revamp

This is a partial list, omitting quick roll change rigs. Over time, roll use has changed and the process control software has been upgraded.

4. Discussion

Detailed examples are drawn here from rolling mills for steel strip, but similar instances are found in a wide range of process industries such as cement, paper or petrochemicals. The principles hold across a wide range of sectors, such as transport systems, the service sector where hotels and shops are routinely refurbished, and defence where weapons systems are regularly upgraded.

There are other explanations for the phenomena we observe here. Plant items may have been overengineered in the first place with generous safety margins to help ensure that buyer performance guarantees were easily met, thereby realising the prompt release of final payments for equipment. The extreme durability of mechanical items such as wide strip mills and chemical processes such as oil refineries suggests that they were substantially built in the first place. The Linz mill had a low initial rolling capacity but was installed in a huge building, which allowed for a significant upstream spread to incorporate more and larger reheat furnaces and a downstream extension to the run-out tables. So the overall length of the rolling line within the building has increased by just under 40% (Figure 1). The Scunthorpe rod mill in the United Kingdom has been physically extended upstream in similar fashion with six new horizontal/vertical roughing stands to roll larger billets ahead of the existing mill (Price, 2007, p. 400). This is quite a literal ‘stretch’, but it requires a generously proportioned building.

So far we have ignored the costs and revenues from systematic innovation. Mills built at the outset for staged expansion can usually be extended quite cheaply for a small outlay on extra mechanical and electrical equipment, coilers and reheat furnaces. Doubling the capacity might cost an extra 20% of the initial outlay. This is a clear reflection of the fact that mechanical and electrical equipment would only take up one third of the cost of the initial contract, with fixed items such as site preparation, civil engineering and building costs, water supply, pipe and power runs, and cranes accounting for the rest. Once a plant is built, the marginal cost of additional capacity from stretching is very low. Presumably there are diminishing returns from stretch determined by the basic initial design, but Linz suggests that these limits have not been reached yet.

We focus here on a batch process with production along a continuous flow line. It is possible that processes made up of discrete, individual stages such as a machine shop have different stretch potential, with scope for modifying individual items without disrupting the rest. Continuous flow processes such as paper mills, petrochemical plant and float glass lines may be restricted to occasional opportunities for modification. The example of blast furnaces for ironmaking shows that there is considerable opportunity for carefully planned stretch during periodic relines. However, a dedicated single-train chemical plant may be harder to stretch than a cracker making multiple products.

Stretch also treats the factory or chemical plant as a closed system. But in truth customers and suppliers are often closely involved in process improvement, for instance the way in which iron ore supplier LKAB develops pellets to enhance blast furnace productivity. Customers too benefit from expansion and quality upgrading.

4.1. Implications for R&D – the paradox of stretch

Stretch may be considered as an evolutionary process where a factory evolves through solving a sequence of problems. The idea of ‘problem sequences’ is familiar in the health sector where medical treatments ‘evolve along trajectories of change shaped by the search for solutions to interdependent problems’ (Ramlogan et al., 2007). Again, Helfat (1994) finds that individual US oil companies accumulate knowledge by persisting with individual lines of R&D, which evolve over time in divergent ways. She sees firms as learning organisations that search for solutions in the face of bounded rationality (Simon, 1979). They do not consider all possible outcomes but focus on existing lines of research where they are best informed.

Problem solving through development may be hard to implement in a process industry. Stable operation and strict adherence to routine are key features of process plant operation. In contrast, R&D requires creativity, project management skills and entrepreneurial implementation in the face of incomplete knowledge and changing circumstance. This poses a dilemma: how do you stretch a process plant through application of novel equipment, untested operating procedures and unfamiliar products when steady operation is at a premium?

This is a challenge for managerial organisation and personnel selection. Contingency theory suggests a firm's organisation adapts to the tasks it faces (Donaldson, 1996). Manufacturing operations are likely to be characterised by formal routines and hierarchical structures, while R&D departments are likely to be structured more organically in a way that allows for trial-and-error learning and fluid movement of personnel from task to task as problems evolve.

There is also a subtle danger of managerial selection. As Schneider (1987, p. 439) recognises, ‘humans, at least in Western societies, are not randomly assigned to settings. Humans select themselves into and out of settings’. People are attracted to like-minded individuals and leave organisations where their face does not fit. As a result, organisations evolve as they attract recruits, perhaps along a trajectory set by the founders who may select and ‘sort’ the type of people who work for a firm (Witt, 1998). Over time a process of attraction, selection and attrition produces a group of like-minded people who shape the organisation for which they work. An extreme case is Firestone Tire and Rubber where Sull (1999) reports by the early 1970s that all of Firestone's management team had spent their entire career with the company, two thirds were born and brought up in the company town Akron, Ohio, one third followed their fathers as Firestone executives and most of the leading executives lived within a five-block radius of one another. So, within a company we expect a science-focussed R&D department to have different characteristics to practical, technology-focussed production management as a result of the selection of different competencies for different roles.

One feature of stretch is the way that change is often initiated by search for solutions to practical operating problems. Innovation is driven by problem solving. Localised electrical heating of slabs to obviate skid marks caused by the pusher furnaces at Linz is a case in point. A customer requirement to obviate these regular imperfection marks in coils drove an ad hoc technical solution that resolved the problem of not having modern walking beam furnaces – a process of ‘resourceful improvisation’ (Hendry and Harborne, 2011)

So the central organisational problem facing R&D management is how do you encourage creative and unorthodox personnel to work in routine areas such as maintenance and persuade results-orientated production managers to accept risky interventions in their production schedules in the interest of continuing innovation?

One solution is to refocus engineering teams on continuing development with the support of R&D personnel. This approach to continuous improvement helps solve the frustration of gifted engineers working in specialised roles on routine and repetitive tasks where they are unable to use their full professional skills (Holt, 1974). A study of Scandinavian manufacturing finds that cross-functional collaboration between departments is one of the key factors explaining innovation performance (Frishammar and Hörte, 2005). This finding is supported by Love and Roper (2009) who find that cross-functional teams enhance technical elements in the innovation process.

There is systematic scope for exploiting potential stretch. This resolves the central problem facing operators of process plant of flexing production in the face of shifting market requirements (Edler et al., 2002; Larsson and Bergfors, 2006). There is a need to systematically plan innovation that increases product flexibility in established manufacturing plants. Process innovation needs to be planned in the same way as product innovation (Pisano, 1997; Lager et al., 2010). In truth, process and product innovations are intertwined with new process steps generating new products and vice versa (Reichstein and Salter, 2006).

The notion of stretch has implications for development (Maxwell and Teubal, 1980). China and India have purchased second-hand steel plant, machine tools and plastics equipment from the United States and Europe to operate with local, low-cost labour. The existence of stretch suggests that this machinery has a considerable future, providing wear parts, deteriorating components and electronic circuits are replaced during reconstruction. Rebuilding provides an opportune moment for upgrading and retrofitting key plant items. Data are not available on the performance of these plants subsequent to recommissioning.

5. Conclusion

We propose a taxonomy of stretch based on five aspects of plant enhancement: (1) learning to operate and maintain; (2) addition of novel equipment and reconstruction; (3) system-wide improvements upstream, notably heavier piece weights; 4) system-wide improvements downstream to improve the offtake of materials; and 5) quality improvements and development of new products. In the case of wide strip mills built under Marshal Aid in Europe, stretch enhanced their capacity by a factor of around 1.8 over 30 years, and in one case the increase is over 8 times the initial rated capacity.

Failure to recognise stretch reflects overemphasis on novelty in economics and innovation studies, and a neglect of the old. There is a corresponding emphasis in engineering training on new plant design rather than refurbishment. Conscious exploitation of stretch is one strategic option for competing in mature industries. Empirical evidence suggests that it is vital to plant survival. But promotion of stretch requires a re-engineering of R&D organisation and planning of future process innovation.


This paper benefited from advice at ‘Managing R&D, Technology and Innovation in the Process Industries’, Ecole de Management, Grenoble, May 2011 and workshops at Outokumpu Oy, Avesta and SSAB, Luleå, Sweden. Warmest thanks to Thomas Lager for encouraging me to write the paper, to Ruggero Ranieri for help on Marshall Aid mills and to Emilia Brodén, Chris Foster, Johan Frishammar, Lennart Gustavsson, Lotta Jakobsson, Phil Judkins, Stan Metcalfe, Peter Samuelsson, Mick Steeper and Arjan van Rooij for excellent discussions. Research was supported by the EPSRC grant EP/D032709/1, ‘Unlocking low carbon potential’.


  • Jonathan Aylen is Senior Lecturer and a Director of the Manchester Institute of Innovation Research, Manchester Business School, University of Manchester, England. He is a member of the rolling committee of the Institute of Materials, Minerals and Mining and chair of the Newcomen Society in north-west England. Recent publications focus on computer history, transfer of technology, forecasting wildfires and on foresight techniques.