Roman Wooden Force Pumps
Use and Performance
p. 7-17
Texte intégral
1The purpose of this brief paper is to give an overview of the use and performance of Roman wooden force pumps. Much more information on them, and on metal pumps, can be found in the texts noted in the short bibliography1. In this paper «pump» means the wooden two-cylinder force pump, unless the context indicates otherwise. The French equivalents of some of the technical terms are given in the notes.
Concept
2As with many very original ideas, the basic concept of the force pump2 appears simple: a cylinder is filled with water; a piston is then forced into it, displacing the water to a higher level through a pipe, or as a jet. The cylinder requires two non-return valves; one at the inlet, and one at the outlet. The concept is attributed to Ctesibius of Alexandria (ca. 270 B. C.)
3Three ancient authors - Philo, Vitruvius, and Hero-describe the force pump. They say that it is made from separate pieces of metal joined together, and two say that it is made of bronze. They do not mention a design in wood, nor does any other ancient author.
4Thirteen pumps made of metal, or parts of them, are now known. The latest appears to date from the 3rd c. A. D. The bronze pump is made by soldering together many separate parts. For example, the pump from the Antiquario comunale in Rome has more than 30 parts (fig. 1). Using bronze has a number of disadvantages; bronze is expensive; a skilled bronze worker is needed; the parts have to be made and assembled with care to withstand high water pressure; it is complicated to take the pump apart to maintain or repair it; and it is difficult to fix the pump in position by attaching it to anything else.
5Much later - perhaps in the1st c. A. D. -a pump was designed that works on the same principle but is made in a completely different way (fig. 2). The spaces for the cylinders, valve chamber3, and connecting passages4, are made by cutting holes into a single large block of wood (usually oak). Internal spaces are then made safe against high water pressure by tight plugs and planks of wood (fig. 3). The flaps of the valves are made of leather (fig. 4). The plates of the valves5 (fig. 4) and the liners6 of the cylinders (fig. 5) are made of metal-usually lead.
6Wooden pumps use cheaper materials than metal pumps, and they are probably easier to manufacture, to assemble, to take apart for maintenance or repair, and to fix in position; and their valves are probably more flexible and effective. In many respects wooden pumps are an improvement over their bronze predecessors.
7Eighteen wooden pumps have been found7, and remains of thirteen survive. This paper is based on detailed examination of the remains8, and on the reports of all eighteen made when they were found. As shown in the map (fig. 6), ten come from the area of the Rhine and Moselle, centered around Trier, a major Roman city. Only one is known from Rome, but three have been found in Milan. Two come from southern England, and two from southern France-from Lyon and Périgueux. But none are known from Spain, Africa, or the East.
Use
8The pump has three great advantages:
- It can raise water from a shaft that is both deep, and narrow.
- It can evacuate water in a pipeline along an uneven, twisting, incline-for example, from a mine.
- It can produce a strong jet in any direction-for example, to fight fires.
9Pumps could therefore be used in fixed installations, positioned at the bottom of a well, cistern, or sump; or as portable pumps, placed in a portable tank which is filled with buckets from the nearest available source. Ancient texts contain a number of references to using pumps, and to their operators. Ten mention pumps used to fight fires, and three mention the vigiles who operated them. Many mention the high jet they produce. Discussions with Fire Officers confirm that these pumps would be very useful in fighting a fire, and much more useful than throwing buckets of water at it. Perfume was sprayed to sweeten the air in crowded places, and perhaps pumps were commonly used for this purpose. Only one ancient text mentions using a pump to raise water from a well, and none mentions using a pump to dewater a mine.
10Of the eighteen known wooden pumps, fifteen were found in wells, and two more may have come from wells. The wells are circular, and often narrow; their diameters range from 0.7 m to 1.7 m, and only two wells have diameters exceeding 1.2 m. In twelve cases the depths of the wells are known; half of them are over 7 m deep. The shallowest and the deepest are both in England; 2.7 m at Silchester, and 26.4 m at Tarrant Hinton. This means that the Tarrant Hinton pump could withstand a water pressure of at least 2.5 Bar.
11Various features of a wooden pump indicate that it was used in a fixed installation, and was not portable. For example:
- Two pumps were found in their operating positions in wells.
- In one case guides for the driving linkage9 found in the well show that the pump was used there.
- Four pumps have an extra valve to guard against high static pressure. This only arises in fixed installations.
- In five cases a long wooden delivery pipe found in the well, or the attachment for one, indicates that the pump was used there.
- At least thirteen pumps were found near the bottom of a well with their metal liners intact. If a pump is found in a well and the metal has not been removed for recycling, we may reasonably conclude that it was used in that well, if there is no evidence to the contrary.
12The evidence suggests that the eighteen pumps were used as follows:
13It appears that sixteen pumps were used to raise water from wells; in ten cases the nature of the surrounding area suggests a particular use for the water, but for the other six there is no indication of the use to which the water was put.
14The use of one pump is not certain; no remains survive and the drawing is unreliable, but it was probably used in the well in which it was found.
15The pump found in 1908 in the cellar of the Amphitheatre at Trier is particularly interesting. Unfortunately there is very little information about it; no remains or drawings exist, and the only mention of it is very brief. It may perhaps have been used for a number of purposes: to dewater the cellar by pumping infiltrated water into the drain; to fight fires; and to spray perfume into the arena. If this is correct, then this pump is portable, and it is the only clear example of a portable wooden pump.
Dating
16All eighteen pumps come from a Roman context. Dates which are more or less reliable are available for eleven of them10 ; they lie in the range1st to 4th century A. D. Of these eleven pumps, two are dated by dendrochronology to the 3rd century A. D. ; the other nine are dated by context. There is nothing to indicate that the remaining seven pumps are not also of the 1st-4th century11.
Design
17All the pumps follow the same overall generic design (fig. 7), but there are nevertheless many differences between them. Some of the more important are as follows:
- In twelve pumps the liners are made of lead, but at Lyon they are of bronze12. If bronze liners can be made more circular, or smoother inside, this might reduce the wear in the piston seals, and the leakage past them.
- In most pumps the outlet valves are horizontal, but in two they are vertical. This may make it easier for them to shed grit.
- Four pumps have an extra valve to combat the effects of high static pressure.
- In most pumps the two outlet valves are located in a single valve chamber, but in two pumps they are in two separate chambers. There is no reason to think that this gives any particular advantage.
- Twenty-nine of the valve plates have a straight edge, to align with the hinge of the valve. Twenty-one of these are circular but have a straight edge on part of the circumference, and eight are square or hexagonal. The remaining thirteen are completely circular, and have no straight edge. Presumably their valves perform less well.
18None of these features shows that a particular pump is clearly more advanced than others, but the Périgueux pump (fig. 8) has many improvements not known in other pumps. In particular:
- In other pumps the plates of the valves are rivetted to the leather flaps, but at Périgueux two clenched nails13 are used14. This makes it easy to replace the leather flap of the valve, and it ensures that the straight edge of the plate remains parallel to the hinge line of the valve.
- The connection between the block and the delivery pipe is made by forcing the pipe into a mortise in the top of the block. This produces a very robust joint.
- The piston seals are attached to the pistons by a cylindrical iron rod. The rod is spring loaded by the seals15, and kept in place by a cotter pin16. This makes it easy to replace the seals when they become worn.
- In other pumps four connecting passages are needed; two vertical, and two horizontal. In the Périgueux layout only three are needed-all bored horizontally. This also enables the outlet valves to be hung vertically.
- The pump has a filter chamber17, constructed as part of the block itself.
- The seats for the inlet valves are tapered18. This enables the tightness of fit between the seats and the cylinders to be adjusted.
- In other pumps the core of the tree trunk lies near the back of the pump, but at Périgueux it lies near its front. It appears that a tree has been selected whose core lies far to one side. This enables the block to be rather large, and for the apertures to be placed in the strongest parts of the trunk.
19These improvements are straightforward in concept and simple to implement. However the Périgueux pump appears to be one of the earlier pumps, and these improvements are not seen in later pumps. Therefore, if the dating is correct, it appears that the improvements do not come from a long process of gradual change, and that although the overall generic design was transmitted widely, important improvements were not transmitted in the same way.
Operation
20Wooden pumps are submerged in water, and force water upwards by pressurising it. They do not lift water by suction. The reasons for saying this are as follows:
- At least two pumps were found in their operating position at the bottom of the well; if they could operate from a higher level, it would be attractive to install them there.
- For technical reasons, water can only be raised about 7 m by suction. Half the wells are deeper than this.
- To raise water by suction a pump must create a partial vacuum19. It is unlikely that a wooden force pump can create the vacuum required.
- All wooden pumps are made from materials which include leather, wood, and iron. If they alternate between wet and dry, leather cracks, wood decays, and iron rusts. The crucial leather valves and seals would suffer particularly badly, and the pump would fail. Keeping the block submerged avoids this.
21To operate smoothly, the motion of the pistons should be co-ordinated, so that one goes up as the other goes down. The easiest way to arrange this is for them to be driven by a rocker20, through connecting rods (or “con rods”)21. No complete driving mechanism has yet been found, but the iron fittings of a rocker were found at Bertrange. In addition, connecting rods were found at Sablon, and significant elements of the drive22 were found at Herrenbrünnchen. Analysis suggests that in pumps operating in wells the pistons are fixed rigidly to the con rods to form a rigid driving linkage23, as shown in figure 9; the pistons therefore follow the slope and movement of the con rods.
22It is vital to ensure that the pistons do not over-run, which would damage the pump, at either end of the stroke. As far as is known, no tests have ever been made, using a replica pump with long heavy con rods, to determine how fast a pump in a well can be driven. But in wells the driving linkage can be very heavy, and have high inertia24 and it must therefore be difficult to start the stroke, and to control the drive thereafter; as a result pumps in wells cannot be driven fast. On the other hand, portable pumps fighting fires are presumably driven as fast as possible, using many operators. The drive is not heavy, but the fast cycle must be controlled. It would be very difficult to use an animal to produce the controlled reciprocating motion needed; the evidence suggests that pumps in wells, and portable pumps, were driven by human operators.
23Many commentators assume that the con rods must lie vertically over the pistons. If they do, the rocker can have only a small radius, and this in turn produces (in the known pumps) a stroke which is very much shorter than the height of the liners. But it is very unlikely that pumps would be made with long liners if the stroke uses only a fraction of their height. It is much more likely that the pistons travel the full distance available in the liners, and that the con rods slope to allow this, as shown in figure 9. Calculations show that in deep installations the slope is too small to create a problem. In portable pumps the slope is probably eliminated by joining the pistons to the con rods with hinged joints.
24It must be very difficult to install a pump at the bottom of a deep well. The block is under water; the parts are heavy and unwieldy; and the wells are narrow and often deep. The block is subjected to strong oscillating forces. The block, the pipeline, and the driving linkage, must all be fixed very firmly. Portable pumps are presumably fixed firmly to their portable tank.
Factors affecting performance25
25The performance of a pump depends on a large number of factors. Some are specific to a particular installation-they include: the height through which water is to be raised (the lift26); the diameter of the liners; and the stroke27. Others apply more generally-they include: the power28 lost through friction; the volume lost through leakage; the force that the operators can exert; the power that they can produce; and the limitation on the cycle speed due to the inertia of the system.
26The power required to drive a pump depends on the rate of output29, and on the lift. Doubling the rate of output, or raising the water twice as far, requires twice as much power (if all other parameters remain the same).
27The volume expelled from the cylinders is determined by two things: the cross sectional area30 of the liners; and the length of the stroke.
28In addition, the operators must be able to produce sufficient force to start the stroke; their ability to do this depends on the strength of their arms, and on their weight. The force required depends mainly on the cross sectional area of the liners, and on the lift.
29The two crucial dimensions of the pump are therefore the cross sectional area of the liners, and the stroke; and the designer of a pump must strike a balance between them. If the force required to drive the pump is too great, it can be reduced by making the liners narrower. The output of the pump can be increased by making the stroke longer. We might therefore expect the liners of pumps in deep installations to be long and narrow.
30In contrast, a portable pump probably has a short stroke, which reduces its output. The output can be increased by making the liners wider. The force required to move the pistons would increase, but if a portable pump is used to extinguish a fire, it would be driven by many operators, to produce greater force and power. We might therefore expect the liners of portable pumps to be short and wide.
31It can be seen that the relationship between the various parameters of a pump installation is complex. It is not suggested that the Roman maker understood either how these parameters arise, or how to calculate their effects, but presumably rules of thumb31 emerged from practical experience, and were used to determine the dimensions that should be used in any particular case.
Dimensions
32The two crucial dimensions of the known pumps
- the stroke, and the cross sectional area of the liners
- are shown in figure 10. As can be seen, the pumps vary widely. The strokes range from 16 cm at Lyon to 41 cm at Silchester; and the cross sectional areas from 20 cm2 at Wederath to 53 cm2 at Lyon. But the diagram shows that the pumps are not of two separate classes, corresponding to “deep” and “portable” use. Unfortunately the pump from the Trier Amphitheatre, which may be portable, cannot be shown, as its dimensions are not known.
33The Lyon pump has the shortest stroke, and the largest area. Although this might suggest that it is portable, it was found in a well with its liners and five valves, so it is probable that it was used there.
34The two pumps from southern England - Silchester and Tarrant Hinton - have the two longest strokes; perhaps this indicates a local tradition of manufacture. But apart from this the pumps from the four geographical groups (Rhine/Moselle; Italy; southern England; southern France) do not share any particular characteristics.
35The pumps found in the two deepest wells are those from Tarrant Hinton and Wederath; and these pumps have the smallest areas. This is consistent with a need to reduce the force required to drive a pump with a high lift. In compensation, the strokes of these two pumps are very long.
Performance
36Calculations have been made of the performance of a theoretical pump of typical dimensions, using the following assumptions32 :
- The diameter of the liners is 7 cm.
- The stroke is 25 cm.
- The water is raised 10m
- 10 % of the power input is lost through friction and turbulence
- 20 % of the water discharged from the cylinders33 is lost through leakage
- The rocker gives a mechanical advantage34, or leverage effect, of 3:1.
37The results are as follows:
- If the rocker is driven at 15 cycles per minute, the pump delivers 23 litres per minute, and requires a power input of 54 watts. Working a rocker, one man can probably produce about 40 to 50 watts, so it would be difficult for one man to drive this pump. But two could do so easily.
- If however, it is assumed that there are two operators, and that they can still control the machine if they drive it faster - at, say, 20 cycles per minute - then the output rises to 31 litres per minute; and the power requirement rises to 71 watts.
38Similar calculations have been made for all the pumps for which dimensions are available.
Summary
39The force pump was invented in the3rd c. B. C. The original design in metal was radically improved, perhaps in the1st century A. D., by creating spaces in a single wooden block and making them pressure proof. Eighteen wooden pumps have been found; the evidence suggests that only one was portable, and that seventeen were used to raise water from wells.
40Many texts refer to pumps (or pumpers) fighting fires, or to the jets that they produce, and it may be that portable wooden pumps were fairly common. But as they are used at ground level they are unlikely to survive; it is attractive to recycle their metal, and to use their wood as fuel. In contrast, wooden pumps in wells should leave some evidence. To recover the liners it is not necessary to recover the complete pump; one could remove the liners, and leave the block and the valve plates in situ - this was probably done at Périgueux, and may have been done at Wederath. Even if all the wooden parts perish, the metal valve plates should survive, though they might not be recognised.
41The fact that so few wooden pumps are reported from the large number of excavated wells indicates that not many were used in wells, even though many may have been used on the surface. Two reasons for this can be suggested:
Technology
42A pump used to raise water from a well may be operated for long periods. If the pump is used for (say) 6 hours a day, 365 days a year, at 15 cycles per minute, the leather seals and valve flaps will move at a rate of about 4 million movements per year. They may fail fairly frequently under such heavy use. If they do, it is very burdensome to repair them, as they are submerged at the bottom of the well. In contrast, a portable pump on the surface may be used much less often; and if it fails, it is much easier to repair.
Economics
43It may be that the greater output obtained from a well by using a pump, as compared with a bucket, did not justify the effort of making, installing, and maintaining a pump at the bottom of a well. In contrast, a portable pump that can fight fires is very valuable.
44It appears probable that seventeen of the eighteen known wooden pumps represent the less common way in which Rome used the wooden pressure vessel - in wells; and that only one - that from the Amphitheatre at Trier - represents the more common use of this technological innovation - in portable pumps on the surface. Most unfortunately, there are no remains of this one example, and very little information about it. The wooden pumps used in wells are now well understood, and the portable pump probably shared many of its features; but confirmation of its detailed arrangement must await the discovery of a further example.
Bibliographie
Des DOI sont automatiquement ajoutés aux références bibliographiques par Bilbo, l’outil d’annotation bibliographique d’OpenEdition. Ces références bibliographiques peuvent être téléchargées dans les formats APA, Chicago et MLA.
Format
- APA
- Chicago
- MLA
Bibliography
Fischer 2002 : FISCHER R., Eine römische Doppelkolben-Druckpumpe aus der Villa von Bartringen, Bulletin d’information du Musée national d’histoire et d’art, Luxembourg 15, juin 2002, p. 38-40.
Krier 2000 : KRIER J., Grandes découvertes à Bertrange – “Bourmicht”, Bulletin d’information du Musée National d’Histoire et d’Art, Luxembourg 13, février 2000, p. 6-7.
Langouët 1996 : LANGOUËT L., La Cité d’Alet; de l’Agglomération Gauloise à l’Ile de Saint-Malo, Les Dossiers du Centre régional d’Archéologie d’Alet (St. Malo), supplément S, 1996, 50-57.
Lehmann 1922 : LEHMANN H., Eine römische Saug-und Druckpumpe aus Trier, Trierische Heimatblätter, 1, 1922, p. 24-27.
Neyses 1972 : NEYSES Α., Eine römische Doppelkolben-Druckpumpe aus dem Vicus Belginum, Trierer Zeitschrift 35, 1972, p. 109-21.
10.3138/9781487577926 :Oleson 1984 : OLESON J. P., Greek and Roman mechanical water lifting devices. Toronto, 1984.
10.2307/jj.5973115 :Oleson 2004 : OLESON J. P., Well-Pumps for Dummies: was there a Roman tradition of popular, sub-literary engineering manuals?, in F. Minonzio (ed.), Problemi di Macchinismo in Ambito Romano, Archeologia dell’Italia Settentrionale 8 (Como), 2004, p. 65-86.
Oleson 2005 : OLESON J. P., Design, Materials, and the Process of Innovation for Roman Force Pumps, in J. Pollini (ed.), Terra Marique-Studies in Art History and Marine Archaeology in Honor of Anna Marguerite McCann (Oxford). 2005, p. 211-231.
Ringeisen 1870 : RINGEISEN Kertzfeld, Bull. Soc. Conservation des Monuments historiques d’Alsace (Paris), II série-septième volume, séance de 9 dec. 1869, p. 63-65.
Roberti 1964 : ROBERTI M., Tracce di vita romana in via Speronari, in Ambrogio Palestra, Ritrovamenti di età romana presso S. Satiroe loro rapporti con la documentazione del sec. IX, Archivio Ambrosiano, 16, 1964, p. 39-44.
Sablayrolles 1988 : SABLAYROLLES R., with the collaboration of Lacour, M., La pompe Romaine de Périgueux, Aquitania, Tome 6, 1988, p. 141-56.
Schiøler 1980 : SCHIOLER T., Bronze Roman Piston Pumps, History of Technology, 5, 1980, p. 17-38.
Schiøler 1989 : SCHIOLER T., Rekonstruktion einer rômischen Feuerloschpumpe in Antiquarium comunale, Mitt. Leichtweiss Institut 103, 1989, p. 281-321.
Schiøler 1999 : SCHIOLER T., Die Mechanik und Technologie rômischer Bronzepumpen. Zu einem Fund aus Aventicum/Avenches VD, Helvetica Archaeologica 30, 1999-117, p. 10-30.
10.1017/S1047759400008230 :Stein 2004 : STEIN R. J. B., Roman wooden force pumps: a case study in innovation, Journal of Roman Archaeology 17, 2004, p. 221-50.
10.1163/9789004473829 :Wikander 2000 : WIKANDER O., Handbook of ancient water technology. Leiden / Boston, 2000.
Notes de bas de page
1 Wooden force pumps are treated in detail in Stein 2004, and in Roman Wooden Force Pumps, a PhD thesis, by R. J. B. Stein, for The University of Reading.
2 Pompe foulante à pistons.
3 Chambre des clapets.
4 Forages de liaison.
5 Plaques métalliques des clapets.
6 Chemisages des cylindres.
7 Stein 2004 mentions nineteen pumps but notes that the Unione pump and the San Giovanni pump may be the same device; this is now confirmed, reducing the number of known wooden pumps to eighteen.
8 Except those of the Martberg pump.
9 Tringle.
10 These eleven include Lyon; M. Savay-Guerraz explains its dating in his paper in this volume.
11 In addition, the two installations at St. Malo (one block with three cylinders, and one with eight, all “in line”) are now thought to date from perhaps the late 2nd - middle 3rd c. A. D. (personal communication L. Langouët).
12 In five cases the material of the liners is not known.
13 Clous rabattus.
14 For the inlet valves. The outlet valves are still in the closed valve chamber and are not accessible.
15 L’assemblage est rendu élastique par les joints.
16 Clavette.
17 Chambre à filtre.
18 Biseautés.
19 Dépression.
20 Balancier.
21 Bielles.
22 Tringle.
23 Tringle rigide.
24 Inertie.
25 The relationships between the various parameters and characteristics of a pump are complex. The brief description given in this paper is a simplified account of the main factors involved.
26 Hauteur de relevage.
27 Course du piston.
28 Puissance.
29 Débit
30 Section.
31 Règles empiriques.
32 Données de base.
33 Strictly; of the swept volume (SV) of the cylinders. The SV is the stroke multiplied by the cross sectional area of the liners.
34 Rapport de force.
Auteur
Le texte seul est utilisable sous licence Licence OpenEdition Books. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.
Les bois sacrés
Actes du Colloque International (Naples 1989)
Olivier de Cazanove et John Scheid (dir.)
1993
Énergie hydraulique et machines élévatrices d'eau dans l'Antiquité
Jean-Pierre Brun et Jean-Luc Fiches (dir.)
2007
Euboica
L'Eubea e la presenza euboica in Calcidica e in Occidente
Bruno D'Agostino et Michel Bats (dir.)
1998
La vannerie dans l'Antiquité romaine
Les ateliers de vanniers et les vanneries de Pompéi, Herculanum et Oplontis
Magali Cullin-Mingaud
2010
Le ravitaillement en blé de Rome et des centres urbains des début de la République jusqu'au Haut Empire
Centre Jean Bérard (dir.)
1994
Sanctuaires et sources
Les sources documentaires et leurs limites dans la description des lieux de culte
Olivier de Cazanove et John Scheid (dir.)
2003
Héra. Images, espaces, cultes
Actes du Colloque International du Centre de Recherches Archéologiques de l’Université de Lille III et de l’Association P.R.A.C. Lille, 29-30 novembre 1993
Juliette de La Genière (dir.)
1997
Colloque « Velia et les Phocéens en Occident ». La céramique exposée
Ginette Di Vita Évrard (dir.)
1971