Version classiqueVersion mobile

Aquam in altum exprimere

 | 
Alain Bouet

The technology of the 1st – and 2nd – century roman bucket chains from London: from excavation to reconstruction

Ian Blair, Robert Spain et Tony Taylor

Texte intégral

1. Introduction

1The problem of how efficiently to extract and distribute fresh water from deep wells, storage cisterns and sources on the surface, such as rivers and streams, has required ingenuity on the part of man since the earliest times. Although many of the solutions have remained virtually unchanged for thousands of years and are relatively simple, such as the shādūf, most water-lifting devices had limitations in terms of the volume of water that they could effectively lift and the depth from which they could raise water.

2Bucket chains or pot garlands evolved in Classical times as a means of lifting water from wells where the depth of the water source precluded the use of larger waterwheels, or norias, which collected water in compartments around their rims as the wheel was turned through the water. Although norias were capable of raising large volumes of water, their enormous size meant that an equally large pit had to be built to house the wheel and this would only be a viable option when the source of the water lay close to the surface. Lighter-weight norias or drainage wheels, which were turned by a man treading on the slatted outer rim of the wheel, were used extensively in mines across the Roman world. The most famous of these were the multiple wheels from the gold mines at Rio Tinto in Spain, where a succession of wheels worked in tandem at different levels to drain water from the deepest galleries and discharge it at the surface. The Romans freely used or adapted pre-existing technology. Whilst there is little doubt that bucket chains were used as a means of lifting water from wells or cisterns throughout the Roman world, there is a surprising dearth of the remains of these devices found on archaeological sites. This is due to a combination of factors, foremost being the absence of the right anaerobic waterlogged environment on most sites to preserve organic remains such as wood and rope and conversely inorganic remains such as metal. Equally significant, perhaps, is that bucket chains had to be easy to dismantle to replace parts and were designed to be shortened or lengthened depending on cyclical water levels in the wells. This meant that at the end of the functioning life of a well the component parts of the bucket chain could be salvaged and reused elsewhere. Because of these factors it is likely that they will remain under-represented in the archaeological record.

  • 1 Oleson 1987, 98-128.

3Prior to excavations during 2001 by the Museum of London Archaeology Service (MoLAS) at two sites in London, the only remains of a Roman bucket chain and its drive system to have been found was excavated at Cosa in Italy in 19841.

2. The bucket chains from Roman Londinium

4The discovery of bucket chains at 20-30 Gresham Street (site 1), and at 12 Arthur Street (site 2), in the City of London shed new light on the machines employed to supply water to the Roman city (fig. 1). Prior to the archaeological excavations on these sites, no evidence had ever been found to suggest that water-lifting machines or bucket chains could have played an important role in supplying fresh water in Londinium during the 1st and 2nd centuries AD. It is extraordinary, therefore, that by the end of 2001 the remains of four bucket chain systems had been found in the bottom of three massive wells, with two distinct types of device (hereafter referred to as type A and type B) being identified. These discoveries led to the retrospective identification of a previously unidentified type A device which had been found nearly 50 years earlier at the Cheapside baths (site 3), immediately to the south of site 1 in 1955.

2.1.20-30 Gresham Street (site 1)

5During the 2001 excavation of site 1, two deep wells and a third much shallower well were excavated. The earliest well had an oak lining and measured 2.60 m square in plan, and at least 4.50 m in depth. The lining was constructed of tiers of close fitting, rebated planks, which were set behind four corner posts and four centrally placed intermediate posts. The intermediate posts were braced by three sets of oak cross-braces, and dendrochronological dating of these timbers showed that the well had been constructed around AD 63 (fig. 2). This construction date indicates that the well was built in the years following the Boudican revolt of AD 60, and that it was most likely part of a well-planned enterprise, carried out during the rebuilding of the infrastructure of the Roman city. Interestingly, this renewal of infrastructure such as roads and water supply preceded the general reestablishment of roadside properties and new buildings. Supplementary dendrochronological dating AD 63-AD 69 of timbers thrown into the well when it was no longer in use has shown that the structure was in use for less then ten years, prior to being abandoned and partly dismantled following a major collapse of a section of its upper lining.

Fig. 1 : Plan of the western half of Roman Londinium in the 2nd century AD showing the location of sites 1-3.

Fig. 2 : Cruciform cross-bracing structure in the massive 1st-century well on site 1, dated by dendrochronology to AD 63.

6To build a structure of this size and complexity in unstable and wet ground would have necessitated prefabrication of the well frame in the upper levels of the construction pit, following which the structure could have settled gradually under its own weight as the excavation of the shaft continued, with additional upper lining planks added from close to the surface.

7Set into a cut in the base of the well, beneath the lower cross frame, was a perfectly preserved softwood half barrel, bound with hardwood hoops (fig. 3). The barrel was made from alpine softwood, probably larch or silver fir, with the binding hoops possibly made of hazel. The function of this barrel is somewhat enigmatic, although it has been suggested that it might have acted as a silt trap at the bottom of the well. An alternative and more plausible explanation is that the barrel was the base of a lined well or sump excavated to test the aquifer and assist in the drainage of the well shaft during its construction.

Fig. 3 : Softwood barrel at the bottom of the 1st-century well on site 1 ; one of the type A water boxes can be seen in situ in the left-hand side of the barrel.

Fig. 4 : Near-complete water box found in the 1 st-century well on site 1 ; the lower face of the box is missing revealing the inner water chamber.

8The most significant finds from the well were recovered from its lower levels, including many from the barrel at its base, which contained a number of parts of a mechanical bucket chain. These included parts of twelve rectangular timber water boxes or buckets, a number of associated iron rivets or connecting pins and a small wooden roller. Each of the water boxes had been carefully made from hollowed-out blocks of oak, with each of the containers covered by a board lid, which was nailed over the water chamber (fig. 4). At each end of the boxes were two parallel slots, which originally housed a pair of pivoting iron links held in place by an iron rivet or pin to join the buckets together in a continuous loop or chain. Each of the boxes could hold about 1.75 litres of water, which entered and exited the water chamber via a small rectangular port cut into the upper side of each of the boxes. The presence of ports on opposite faces of two of the water boxes suggests that there may have been two bucket chains operating in the well, effectively doubling its output. Surviving on the surfaces of some of the boxes were traces of pine resin, which had been applied as a sealant to make the joints watertight.

  • 2 Marsden 1976, 40 figs 16 and 26.

9Two almost identical, but fragmentary, wooden boxes had been found in a wood-lined well or water tank on the Cheapside bathhouse on site 3, immediately to the south of site 1 in 1955, although they were not identified as being parts of a bucket chain2.

10The second deep well on site 1 was part of an improved and larger structure, provisionally dated to the early 2nd century AD, precluding its having been a direct replacement for the earlier well. The well was also made entirely of oak, and was seen in plan to be 3.60 m square, and 5 m in depth. The lining was composed of tiers of close-fitting planks set behind four corner posts and eight centrally placed intermediate posts. Nails fastened the planking to the reverse of the posts. The intermediate posts were originally braced by at least three sets of cross-braces, which were found collapsed over one another in the lower levels of the well, with only the lowest frame likely to have remained in situ. These cross-braces effectively sub-divided the well into nine square compartments. This well, like the adjoining 1st-century well, was partly prefabricated, with the lower timber frame being assembled in the part-excavated construction pit before being undermined ; the upper lining planks were added sequentially as the excavation of the well shaft continued, forcing the entire structure to sink gradually under its increasing weight.

11The lower levels of this well included several large semi-articulated parts of a contrasting type B bucket chain (fig. 5). The mechanism was made of wrought iron, forming a heavy double chain supporting boarded oak box buckets held between flat links alternating between pairs of cranked iron links (fig. 6). In this mechanism the iron links bore the entire weight of the bucket chain, in contrast to the type A device where the tension and load were transferred directly through the body of the oak water boxes.

Fig. 5 : The type B wrought-iron bucket chain during excavation.

  • 3 Wilkes 2002, 6.

12All the water containers had been heavily burnt, and much of the ironwork distorted by heat, indicating that the upper sections of the bucket chain and the overlying well house had been destroyed by fire. Laboratory analysis of the metalwork has shown that the iron had a quenched structure, that is it had been heated up until it was red-hot and then plunged into water3. Given that the bucket chain was intended to be under tension, it is thought that quenching, which would have hardened the metal but at the same time made it more brittle, was unlikely to have been part of the original manufacturing process. It is assumed, therefore, that quenching was unintentional and that it had taken place when the bucket chain collapsed into the well water during the fire that destroyed the structure. Because all the recovered elements appear to be heat damaged, and large sections of the bucket chain were clearly missing, it is evident that the undamaged (below water) section of the bucket chain had been salvaged after the fire. The salvaging of the undamaged section of the bucket chain from the lower level of the well clearly illustrates the value and importance of the device to the Romans, who would presumably have repaired it and reused it elsewhere. Following the fire that destroyed the well house, the well was abandoned and its shaft gradually filled during the mid to late 2nd century.

Fig. 6 : Articulated section of type B wrought-iron bucket chain from the 2nd-century well on site 1.

Fig. 7 : The reconstructed box buckets of the 2nd-century type B bucket chain from site 1.

Fig. 8 : Close-up showing domed head of the linkage pin of the type B bucket chain.

13Although the box buckets were badly damaged, enough fragments survived to reconstruct how they were made and how they articulated with the iron bucket chain (fig. 7). The straight links of the chain were nailed onto, and in isolated cases had been recessed into, the outer sides of the boxes and were joined at each end to the intermediate cranked links by a large pin (fig. 8), held in place by a split ring (fig. 9). The base of the box was recessed beneath the water chamber to allow the lower pin to pass through the body of the box. This system of linked elements meant that the bucket chain could easily be shortened or lengthened dependent on the relative cyclical or seasonal water level in the well. Unlike the side discharge system used in the much smaller type A water boxes from the adjoining well, and those found on sites 2 and 3, the containers were open at the top and had a forward discharge, with the cranked links located in the open mouth of the containers. In order to stop the cranked links from moving inwards, two guide pins or nails had been driven through the recessed base to restrict any lateral movement across the end of the boxes. It is interesting that only one box bucket was found to have holes for retaining pins at the opposite open mouth end, indicating that the cranked links only required anchorage at one end of each box.

Fig. 9 : Close-up showing the iron connecting ring through the linkage pin of the type B bucket chain.

  • 4 Blair & Hall 2003, 43.

14It is calculated that in order to lift water from the lower levels of the well, the bucket chain would have required a chain of 30 box buckets. Each of the containers held approximately 6 litres of water when full, or three times the amount of one of the water boxes from the smaller type A bucket chains described above. It has been calculated that using a geared capstan drive to power the bucket chain, it would be capable of raising 72,000 litres (15,838 gallons) of water over a ten hour operating day. To put this output figure into perspective, this volume of water would be capable of providing a subsistence level of household supply to around 8000 people, or approximately one third of the estimated population of Roman London in the 2nd century AD4. Water may also have been supplied to trades and industries or used to flush out drainage systems.

2.2. 12 Arthur Street (site 2)

15Although the unusual rectangular shape of the well at 12 Arthur Street contrasts with the square well structures from Gresham Street, the type A bucket chain elements it contained were of identical form and dimensions to the boxes from the earlier well at Gresham Street and the Cheapside baths site. In plan, the oak timber superstructure of the well measured 5.85 m x 2.50 m and consisted of horizontally-set timber base plates into which were jointed timber posts that retained the plank lining of the well. To provide horizontal support within the structure, cross-bracing timbers were jointed into the intermediate posts spanning the aperture of the well. Dendrochronological and finds dating imply that the well was constructed c. AD 90 and that it remained in use for at least 30 years.

16In the partially excavated lower fills of the well were found three well-preserved complete rectangular oak water boxes and five less complete examples of very similar dimensions to those found on sites 1 and 3, but slightly heavier and more crudely made than the examples found on site 1. Uniquely, one of the boxes still retained its original linkage, an important detail missing from all of the previously found examples (fig. 10). The short wrought-iron links, which connected the boxes together, were held in place and pivoted on a rigid iron pin or nail bent over into the outer face of the container. It is thought that these pairs of links would have connected directly to the water box in front, but it is conceivable that each box had two sets of links. Such a configuration would have increased the space between the containers and provided a mobile pivot point where the links joined one another, effectively creating a cranked link. Given the difficulties inherent in removing the linkage pin from the water box as found, such a configuration would have some advantages in that it would have been easier to lengthen or shorten the chain if required.

Fig. 10 : Complete type A water box with linkage from site 2 ; the narrow entry port can be seen in the upper right-hand side of the box.

17Interestingly, the other complete water boxes were notably smaller then the box described above and the linkage slots were much closer together. These small, but important, differences in design clearly indicate that more than one bucket chain had been used in the well, as it would have been impossible for water boxes of different dimensions and linkage configurations to work effectively as part of a single bucket chain.

18Considering the unusual rectangular plan of the well, it is conceivable that the structure was specifically designed to accommodate two bucket chains working simultaneously, possibly turning around a common axle.

3. The reconstruction programme

  • 5 Blair & Hall 2003, 6.

19Soon after the completion of the excavations on site 1, the Museum of London decided to build a full-scale replica of the type B water-lifting machine that had existed in the early 2nd-century AD well.5 The Museum wanted the replica to be a public demonstration of Roman technology and allow visitors, including families and children, to power the machine. The building of the replica would provide historians of technology with the opportunity of exploring and learning from its design and operation. A valuable and kind offer of financial sponsorship allowed the theoretical analysis and design to be developed by two engineers.

20When the theoretical reconstruction began, the first and most important problem to solve was how the water was transferred from the ascending wooden buckets to the trough taking the water away from the machine. No remains were found of the machinery that carried and drove the bucket chain above the well, although this structure was presumably affected by the lire that partially destroyed the bucket chain.

21In the type A bucket chains that were made up of hollow boxes carved out of solid oak blocks, covered with a lid, the water was admitted and discharged via a port in the side of the box (fig. 10). Technical analysis has shown that these hollow boxes, joined together with straight iron links, were probably carried and driven by an octagonal faceted wheel (fig. 11). Clearly these could discharge sideways straight into the take-away trough, very similar in function to the compartmented rims of Roman drainage wheels.

22With the type B bucket chain, the first method of water transfer that the engineers considered was for the water to be thrown forward from the buckets and collected by a trough immediately in front of the wheel carrying the chain. Experimentation and theoretical analysis showed that this arrangement, which required a fairly fast and steady velocity using a relatively small diameter wheel, generated much splashing and had the disadvantage of filling the descending hollow bases of the preceding buckets. It was considered that the weight, shape and thickness of the wooden buckets was not such an effective design for throwing water forward for collection, when compared with thin-walled metal or wooden buckets.

Fig. 11 : Schematic design for an octagonal faceted drive wheel and type A water boxes.

23An alternative method, considered but also rejected, was that the bucket chain could be carried by an ‘open cage’ wheel, where the chain is supported by cantilevered arms and the water is collected in a trough in the ‘hollow’ side of the wheel (fig. 12). Although this design was used in recent centuries and can still be found in parts of North Africa and the Middle East for example, it was considered to be most unlikely to have existed at this site because of the great weight of the bucket chain. It was calculated that the original bucket chain weighed 500 kg including the water being lifted.

24The engineers realised that the bucket chain required a two-stage transfer, whereby the water is first discharged from the bucket into a compartment and second discharged from the compartment through a port into the take-away trough positioned alongside the wheel.

Fig. 12 : Schematic design for an open cage drive wheel.

25Using mechanical and hydraulic theory a number of options were developed for the size of the driving wheel and the way in which the water could be transferred. Theoretically, as the wheel was made larger, the transfer became smoother and the delivery rate of the machine increased, but more power was required. There was also the question of what the aquifer could yield. The engineers determined that the type A water boxes in the earlier 1st-century well probably delivered approximately 0.75 litres per second. Furthermore, two different designs of boxes were found which, in conjunction with scratches made by the chain against the wooden walls of the well, suggested the probable existence of two bucket chains operating simultaneously. Noting that the later well, which was built 30 years after the adjoining well was abandoned, had twice the water capacity and bucket capacity per unit length of chain, the engineers considered it prudent to adopt a provisional target of 2 litres per second for the replica machine.

26Although the type A bucket chain was very probably carried and driven by a faceted wheel, the engineers believe that the open buckets of the type B bucket chain were not engaging with a faceted wheel. The main reason for this conclusion is that the water discharging from the bucket requires space to allow its collection and re-direction via a port. The experiments showed that to achieve this transfer with the minimum of turbulence, the collection chamber needed to be positioned below the level of the bucket and be shaped to encourage rapid and smooth transfer. In the absence of a faceted wheel, it is believed that a positive drive could be obtained by using ‘registers’ that engage with the underside of the cranked links (fig. 13). This aspect of the design was much debated, because the angle of the cranked chain links showed no evidence of wear underneath. This meant that the ‘registers’ that engaged them and drove the wheel around could not have been made of iron, and it was decided that these would have to be made of hardwood and with a sacrificial bearing surface. The singular advantage that the cranked-link engagement gave the builders was that the register could be extended as a plate to form the end of the compartment, which could then be deeper to facilitate the efficient collection and re-direction of the water and the positioning of the transfer ports.

Fig. 13 : Early schematic diagram of type B box buckets showing how the bucket chain may have engaged with registers on drive wheel.

Fig. 14 : Preliminary reconstruction drawing of type B bucket chain powered by a treadwheel.

27At the commencement of the project alternative methods of powering the machine were considered and sketch drawings were produced of possible arrangements. The first option was a treadwheel mounted on the wheel-shaft that would allow one or more men walking on the rim to power the machine (fig. 14). This arrangement reflects the design of the man-powered drainage wheels used within mines in several provinces of the Roman empire. The second option was for the machine to be driven by a capstan using animal or human power driving through two gears (as in the saqqiyah or sāqiya). However, as the theoretical design developed, it was realized that the power needed to drive the machine was probably approaching the limit that one man could provide comfortably on a treadwheel. Although it was possible to design a treadwheel on which two men simultaneously worked, either side by side or one behind the other, the weight of the rotating elements and bearing loads would have been substantially increased. In the final design a capstan and gear drive was adopted because of its greater flexibility of providing power and the advantage that it gave to demonstrate early gearing in action (fig. 15). Additionally, having regard to the desire to have the public driving the machine, this was much safer and allowed several adults or children to take part.

28A considerable amount of theoretical design work was devoted to exploring the design of the two-stage transfer that mainly concentrated on the number of registers, that is the diameter of the wheel engaging the bucket chain. Designs varied considerably, ranging from a 5-register wheel the smallest possible arrangement, to a much larger 12-register wheel the largest possible arrangement produced by the cranked link geometry. The engineers sought for a solution that would provide a lighter machine having an efficient transfer and an output close to the adopted figure of 2 litres per second. It was essential that the two-stage water transfer worked efficiently, and this required development work on the collection compartment, and the transfer port design. The engineers realised that the programme of replication would need to include experimentation with a prototype wheel, which would help to decide the final design.

29With preliminary designs of the machine and support frames finalised and a specification and drawings produced, tenders were invited from several organisations considered to have the necessary experience to undertake the work. A contractor was appointed and the works commenced in August 2002. One of the first things done was the building of a prototype wheel which allowed the engineers to experiment with the design of the driving wheel. Initially a 7-register wheel was built and the design of the water compartments and transfer ports were developed at the builder’s site. Numerous trials helped to determine the most efficient shape and speed for the two-stage transfer (fig. 16). It became clear however that the output might not reach 2 litres a second without incurring more losses than were acceptable and this led to the adoption of an 8-register wheel that gave a more efficient transfer and easier construction (fig. 17).

Fig. 15 : Preliminary reconstruction drawing of type B bucket chain powered by a capstan and gears.

Fig. 16 : Testing the water discharge sequence of the type B bucket chain using a 7-register prototype wheel.

Fig. 17 : The drive wheel (left) under construction alongside the prototype drive wheel (right) in the McCurdy workshop.

30At each end of the horizontal shaft iron journals (axles) were inserted and strengthened by wrought-iron bands shrunk on. Evidence of this method of strengthening the ends of wooden shafts has been found in the Roman world. The journals are supported by oak bearings and lubricated with animal fat. A similar arrangement exists at the top of the vertical shaft, but at the bottom of the shaft – called the footstep bearing – the lower end of the journal is rounded and supported by an iron plate having a shallow cavity.

  • 6 Jacobi 1912, 75-95.

31Very little is known about the design of gearing that the Romans used in their water mills and other machines. Although one remarkable specimen of a Roman lantern gear (a cylindrical gear with spokes fixed between two circular plates) has been found, this was clearly intended for an engagement having a high mechanical ratio of perhaps 4 or 6 to 16. The experiments showed that the two gears needed to be 1 to 1 and therefore approximately the same diameter. One of the gears was provided with one cog more than the other gear to ensure an extended cyclical wear pattern of cog engagement. The diameter of the cogs, 55 mm, was determined by the torque to be transferred, and their shape and pitch was developed from full-scale prototypes that ensured a smooth engagement and disengagement (fig. 18). With a gear diameter close to 1 m it was considered unnecessary to have a lantern gear and a contrate (teeth perpendicular to gear wheel) gear form was chosen with a simple design of cog, circular in section, driven tight into the body onto a shoulder, with the ends split and wedged (fig. 19).

Fig. 18 : One of the gear wheels being assembled.

32The machine was framed and supported to provide a delivery head of approximately 3 m above ground. This was determined in part by the overhead horizontal shaft and gearing (fig. 20) and (fig. 21). In the replica machine the water level is 1 m below platform level, but in the original well it is believed that the water level varied and in dry periods could have been up to 3 m or more below ground level. More power would therefore have been required than is needed to drive the replica machine.

33Preliminary tests have shown that an output of 2 litres per second can be achieved at a capstan speed of 2.72 rpm. At higher speeds the water losses due to splashing and less efficient transfer increased disproportionately and at 6 rpm it is estimated the net output is 3.7 litres per second (fig. 22). Torque measurements showed that at the lower speed of 2.72 rpm the power necessary to maintain constant speed was approximately 0.16 HP rising to 0.35 HP at 6 rpm. Assuming that the original machine had a lift twice that of the replica, these power requirements would double. Using a figure of 0.1 HP per man, this indicates that several men would have been used to work the original machine. However, it is more likely that it was powered by two animals, probably donkeys or mules with a walking circle somewhat larger than used in the replica.

Fig. 19 : The overhead shaft and gearing on the finished machine.

Fig. 20 : Preliminary engineering drawing showing the elevation of the proposed water-lifting machine.

34Working the machine shows that there is an optimum speed of operation : too fast and the trough fails to collect all of the water in the transfer compartment, while if the operators walk too slowly, the transfer potential is under-used. At the lower speed the chain velocity is 20 cm/second and delivers 20 buckets per minute (fig. 23). In the replica machine the water is purposely diverted from the collection trough and recycled down into the well within a transparent conduit so that operators can see the products of their labour. In the original machine the water would possibly have been distributed to other parts of Londinium using wooden or stone troughs or conduits, or even lead pipes, at levels that maintained a head of delivery.

  • 7 Spain & Taylor in preparation.

35Further scientific experimentation and studies have still to take place on the replica, and will be included within a final report detailing the specialised engineering aspects of the project7.

36Another Roman site which has yielded evidence of a bucket chain is the Cosa Spring House, 100 km north-west of Rome. From the evidence found the archaeologists determined that spring water was lifted up to 13 m by wooden buckets suspended on a pair of ropes carried by a vertical bucket-chain wheel. This wheel, which probably had 7 or 8 registers, may have used troughs on the perimeter to funnel or re-direct the water to a collecting trough beside the wheel, in other words a two-stage transfer. This machine, dated to sometime between the 1st century BC and2nd century AD, was driven from a capstan via two gears and an elevated horizontal shaft.

Fig. 21 : The finished water-lifting machine assembled on site outside the Museum of London.

4. Summary and conclusions

37Given the small number of parts of Roman bucket chains and their drive systems that have been discovered prior to 2001, it is hardly surprising that this was an area that has to date remained largely shrouded in mystery and informed conjecture. One of the most pleasing aspects of the discoveries has been the dialogue that has opened up between archaeologists and experts in the field of ancient technology across the world. We were especially pleased to be invited to attend the study day at Bordeaux University and to be able to contribute to the on-going debate of the technology of water-lifting machines in the Roman world.

Fig. 22 : The two-stage water transfer system of the reconstructed type B bucket chain in operation.

38It was an odd archaeological coincidence that at the same time that MoLAS were excavating the bucket chains in London an excavation by Bordeaux University at Barzan had also recovered the well-preserved waterlogged remains of substantial parts of the drive system and possible gearing of a bucket chain. The Barzan device was originally used to lift water from a deep rock-cut well in the thermal baths complex, although in marked contrast to the London finds no parts of the actual suspended bucket chain appear to have been found.

39One of the most intriguing aspects of these devices is their diversity as reflected in the contrasting remains from England, France and Italy. Although it would have been convenient to have found the drive system we were missing in London at Barzan, it is clear that we are looking at two very different machines which would have required completely different bucket chains and drive configurations for them to have worked.

40The use of iron for the linkage in the London bucket chains was made practical because they were required to raise large volumes of water to a height of 6-8 m. The Roman engineers would have been able to prove empirically through ‘trial and error’ testing that the total weight factor was not a concern. This contrasts with the bucket chains from both Cosa and Barzan where the aquifer was much lower than in London, and the water had to be lifted 13-16 m to the surface. Because of weaknesses inherent in individual components of the London bucket chains, most notably the connecting pins of the type B device which would have bent under excessive load, such a machine without modification would not have been a viable option for raising water from the much deeper wells at Cosa and Barzan. In those cases the depth factor allied to the weaker strength of the rope would have limited the water volume to be lifted. The individual water boxes or buckets would probably have been linked together on the rope at larger intervals to both lessen the weight of the much longer suspended chain and aid a more practical connection to the drive wheel without slippage due to the out of balance forces over the drive wheel. The lighter weight of lifted water per metre of chain and perhaps slower speed would have resulted in more comfortable power requirements, possibly close to those demonstrated by the London bucket chain. The lighter total weight without an iron chain would simplify the suspension support.

Fig. 23 : The finished full-size reconstruction of the 2nd-century water-lifting machine being demonstrated outside the Museum of London in 2003.

41Rarely do we get the chance to move quickly and seamlessly in less than a year from a significant archaeological discovery through to designing and building a full-scale working reconstruction based upon the excavated finds. Engineers, archaeologists and timber construction specialists collaborated to create a unique working example of Roman technology using the evidence from the surviving remains and known examples of ancient engineering, supported by modern engineering principles. The completed water-lifting machine was finally erected outside the Museum of London in November 2002 and was demonstrated daily over the next 9 months.

42The specialists most closely involved in the project would be the first to admit that the machine that operated the Gresham Street bucket chain in 2nd century Londinium could have looked different, but this is inevitable given that so much of the original machine had not survived. The reconstruction is in keeping with the true spirit of experimental archaeology and is intended to be thought provoking and generate discussion. In doing so, it is hoped that the academic debate about the technology of Roman water-lifting machines will be moved forward and the significant and pivotal role that bucket chains played in large-scale water production across the Roman empire will be highlighted.

Acknowledgements

43MoLAS would like to thank Land Securities plc for their financial support of the excavations at 20-30 Gresham Street, and Shieldpoint Ltd for funding the work at 12 Arthur Street. The reconstruction project was made possible through the sponsorship of Swiss Re. Theoretical analysis and design of the water-lifting machine was developed by engineers Dr Robert Spain and Tony Taylor (Foggo Associates). The prototype and finished machine were built by Peter McCurdy and his team at McCurdy & Co.

44Thanks are due to Agnès Shepherd for compiling the French summary for the study day and for her help in translating the more technical engineering terminology in Bordeaux. Special thanks are due to our colleagues at Bordeaux University, especially Alain Bouet, Sophie Coadic and Nicolas Saulière for the hospitality they showed us throughout our short trip.

45The photographs to accompany this report are by Maggie Cox and Andy Chopping of MoLAS and the drawings by Sophie Lamb, Robert Spain and Tony Taylor. The archaeological projects were managed by Nick Bateman of MoLAS, with post-excavation management by Peter Rowsome and in-house editing by Susan M Wright.

Bibliographie

Bibliographie

Blair, I. et J. Hall (2003) : Working Water : Roman Technology in Action, Londres.

Jacobi, H. (1912) : “Römische Getreidemuhlen”, SJ, 3, 75-95.

Marsden. P. (1976) : "Two Roman Public Baths in London”, Transactions of the London and Middlesex Archaeological Society 27, 1-70.

Oleson, J. P. (1987) : The Spring House Complex, in : A. M. McCann, The Roman Port and Fishery of Cosa, Princeton, 98-128.

Wilkes, R. (2002) : examination of the Ironwork from a Roman Water-Lifting Device from the Eastern Blossoms Inn Well, London, English Heritage Centre for Archaeology Report 22/2002 (rapport non publié), Portsmouth.

Annexes

Supplementary technical information

Site

Dimensions of well

Depth

Bucket chain

Date of structure

Site 1

2.60 m x 2.60 m

4.50 in

type A

c. AD 63

Site 1

3.60 m x 3.60 m

5.00 m

type B

Early 2nd century

Site 1

2.60 m x 2.80 m

1.80 m (truncated)

(none found)

Early 2nd century

Site 2

5.85 m x 2.50 m

2.00 m
(truncated)

type A

Late 1st century

Site 3

stepped profile :
top 2.80 m x 3.70 m
base 2.80 m x 2.8 0m

2.00 m
(truncated)

type A

early 2nd century ?

The well structures.

Site

Bucket chain

Water boxes

Straight connecting links

Site 1

type A

c. 390 mm x 152 mm x 105 mm-125 mm

(not found)

Site 2

type A

c. 385 mm x 180 mm x 102 mm-110 mm

c. 170 mm x 20 mm x 10 mm

type A (small box)

c. 355 mm x 145 mm x 100 mm

(not found)

Site 3

type A

c. 390 mm x 170 mm x 105 mm-120 mm

(not found)

Dimensions of type A bucket chain elements.

Site

Bucketchain

Water boxes

Straight links on box sides

Cranked connecting links

Linkagepins

Site 1

type B

c. 400 mm x
250 mm x 170 mm (reconstructed)

c. 315 mm x
30 mm x 5 mm-9mm

c. 275 mm x 23 mm-28 mm x
13 mm-16 mm

c. 265 mm x 23-24 mm

Dimensions of type B bucket chain elements.

Notes

1 Oleson 1987, 98-128.

2 Marsden 1976, 40 figs 16 and 26.

3 Wilkes 2002, 6.

4 Blair & Hall 2003, 43.

5 Blair & Hall 2003, 6.

6 Jacobi 1912, 75-95.

7 Spain & Taylor in preparation.

Table des illustrations

Légende Fig. 1 : Plan of the western half of Roman Londinium in the 2nd century AD showing the location of sites 1-3.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-1.jpg
Fichier image/jpeg, 106k
Légende Fig. 2 : Cruciform cross-bracing structure in the massive 1st-century well on site 1, dated by dendrochronology to AD 63.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-2.jpg
Fichier image/jpeg, 280k
Légende Fig. 3 : Softwood barrel at the bottom of the 1st-century well on site 1 ; one of the type A water boxes can be seen in situ in the left-hand side of the barrel.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-3.jpg
Fichier image/jpeg, 293k
Légende Fig. 4 : Near-complete water box found in the 1 st-century well on site 1 ; the lower face of the box is missing revealing the inner water chamber.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-4.jpg
Fichier image/jpeg, 171k
Légende Fig. 5 : The type B wrought-iron bucket chain during excavation.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-5.jpg
Fichier image/jpeg, 302k
Légende Fig. 6 : Articulated section of type B wrought-iron bucket chain from the 2nd-century well on site 1.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-6.jpg
Fichier image/jpeg, 55k
Légende Fig. 7 : The reconstructed box buckets of the 2nd-century type B bucket chain from site 1.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-7.jpg
Fichier image/jpeg, 253k
Légende Fig. 8 : Close-up showing domed head of the linkage pin of the type B bucket chain.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-8.jpg
Fichier image/jpeg, 130k
Légende Fig. 9 : Close-up showing the iron connecting ring through the linkage pin of the type B bucket chain.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-9.jpg
Fichier image/jpeg, 196k
Légende Fig. 10 : Complete type A water box with linkage from site 2 ; the narrow entry port can be seen in the upper right-hand side of the box.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-10.jpg
Fichier image/jpeg, 239k
Légende Fig. 11 : Schematic design for an octagonal faceted drive wheel and type A water boxes.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-11.jpg
Fichier image/jpeg, 86k
Légende Fig. 12 : Schematic design for an open cage drive wheel.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-12.jpg
Fichier image/jpeg, 83k
Légende Fig. 13 : Early schematic diagram of type B box buckets showing how the bucket chain may have engaged with registers on drive wheel.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-13.jpg
Fichier image/jpeg, 107k
Légende Fig. 14 : Preliminary reconstruction drawing of type B bucket chain powered by a treadwheel.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-14.jpg
Fichier image/jpeg, 435k
Légende Fig. 15 : Preliminary reconstruction drawing of type B bucket chain powered by a capstan and gears.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-15.jpg
Fichier image/jpeg, 430k
Légende Fig. 16 : Testing the water discharge sequence of the type B bucket chain using a 7-register prototype wheel.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-16.jpg
Fichier image/jpeg, 521k
Légende Fig. 17 : The drive wheel (left) under construction alongside the prototype drive wheel (right) in the McCurdy workshop.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-17.jpg
Fichier image/jpeg, 312k
Légende Fig. 18 : One of the gear wheels being assembled.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-18.jpg
Fichier image/jpeg, 266k
Légende Fig. 19 : The overhead shaft and gearing on the finished machine.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-19.jpg
Fichier image/jpeg, 276k
Légende Fig. 20 : Preliminary engineering drawing showing the elevation of the proposed water-lifting machine.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-20.jpg
Fichier image/jpeg, 90k
Légende Fig. 21 : The finished water-lifting machine assembled on site outside the Museum of London.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-21.jpg
Fichier image/jpeg, 223k
Légende Fig. 22 : The two-stage water transfer system of the reconstructed type B bucket chain in operation.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-22.jpg
Fichier image/jpeg, 243k
Légende Fig. 23 : The finished full-size reconstruction of the 2nd-century water-lifting machine being demonstrated outside the Museum of London in 2003.
URL http://books.openedition.org/ausonius/docannexe/image/10458/img-23.jpg
Fichier image/jpeg, 294k

Auteurs

Museum of London Archaeology Service.

Mechanical Engineer.

Foggo Associates, Architects and Engineers, London.

© Ausonius Éditions, 2005

Conditions d’utilisation : http://www.openedition.org/6540

Cette publication numérique est issue d’un traitement automatique par reconnaissance optique de caractères.
Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search