Human Powered Hovercraft :: Steam Boat Willy

human powered hovercraft


This is a revised version of a lecture delivered at the Hovercraft Museum by Chris Roper.

Steam Boat Willy started as a project of the Human Powered Flight Club of the University College London Union, ( UCLU HPFC ).


Why did we start to design and build a human powered hovercraft ?

For myself, because I rather liked the idea, and had just returned from the Festival of Human Power at Interlaken, Switzerland. Here there were human powered boats, bikes and an aeroplane, the Velair. Whatever could be next ? One type of craft I hadn't seen was a hovercraft. So I felt that what would be next would be a human powered hovercraft and I would design and help build it.

For the Club, because we considered that it would be within our capability to do, would be within our brief as a human-powered-flight-club, and because of our location in a city centre, far from any airfield, but close to a river, the Thames. Hence we would be able to operate near our base, which we could not do with an aeroplane.

For the International Human Powered Vehicle Association, our craft could be amongst the first of its type, leading maybe to races etc. Since then we have learnt that we would be classified as a "watercraft" by that association. We will in fact be attending the World Champs in August 2008, but expect to be the only hovercraft there.

For the planet, because we would be able to demonstrate that yet another activity can be achieved without fossil fuels, ie the activity of hovering.

That was why we started. Now, years later, we sometimes get asked "Why do you do it ?", Individual's motivations differ. of course, just as motivations change over time, but we all agree that we have a craft which is good, but far from perfect, so we continue to improve it. And to enjoy operating it.


The UCLU HPFC was formed in 1994 by (now Dr) James Moult. The Club built a spool-drive bicycle and the cockpit of a human-powered-aircraft flight- simulator, which was shown at the Royal Aeronautical Society in Jan 2001. This cockpit is built just like a real one except it is supported on a display-stand instead of wheels. When you pedal, the propeller goes round. We actually did put it on wheels once, and by pedalling you could propel yourself forward. It was planned to serve both as a mock-up of the cockpit of a compete human-powered-aircraft and also as a display item. The construction techniques used for this were adopted for the hovercraft. Outside contacts. The electronics for this simulator were assembled by John and Gerald Wimpenny. John Wimpenny was the first to fly half a mile by his own leg power in 1962. We are in contact with, and get some help from John and Mark McIntyre, designers and builders of the Airglow, the only currently flying pedal plane in Britain. Also from hovercraft experts based at the Hovercraft Museum, Lee-on-Solent.


What were our stated aims at the start? Now, there have been a few, only a few, human powered watercraft which when static are displacement craft but, at sufficient speed, ride on hydrofoils. It is not easy to do this, and hence I appreciated from the start that for a human powered hovercraft to do the approximate equivalent, ie what`s known as "getting over the hump" would not be easy. Therefore, as I considered at the time, this is the stage that must be concentrated on, this is what the design must be optimised for. There are other approaches that could have been taken, this is the approach that I took. So, one of the definite goals that I saw as important at the outset was to be able to go over the hump, and the design was optimised for this. Another aim was to have the machine operating within 12 months. The intitial group were Joel Corcoran, Aleksi Halkola, Chris Roper, Simon Ward. All of us had been involved in the club`s previous project, the cockpit of a human powered flight simulator. It was the intention at this stage that we should all crew the craft and we now all have. In the words of Chris Cockerell "Hovercraft should go faster than boats and be fun", and these also were two specific aims. Since then, we have certainly had fun, and one of the videos on our website shows SBW going above hump-speed and leaving a canoeist well behind. What the video doen't show is that we had the wind behind us, and it has all taken much more time that we planned.


What precise type of machine did we consider would achieve the stated aims ?
The Club agreed that we should meet this specification :-
/ To be testable without thrust, ie static hover.
( This has been done. Many people of all ages, both genders and emanating
( from several parts of the globe have become airborne under their own leg
( power in the craft.
/ to be transportable without engined vehicle, ie handles
( Yes, but it takes four to safely carry it because of its size.
/ to be drawn before manufacture, and any drawings updated after modification.
( We have tried to keep to this procedure.
/ to float
( Yes, and it has hovered up from floating on the sea.
/ to go faster then a boat and be fun
( It is very subject to the wind. Any upwind progress is marginal.
/ to be optimised for minimum power at hump speed
( This is explained below. It is close to the calculated optimum
( for this, but we had to consider other criteria and practicalities.


The Club had run a series of tests at the UCL Stanmore Physiology Laboratory to determine our own output power from pedalling. Comparing the results of these with my estimate of the required power at hump speed showed that getting over the hump wasn`t something that we were all going to be able to do.
Of course, it would be great if I am proved wrong on this, but in order that the craft should ever get over the hump, we decided to tailor it for our fittest member. A simple running-up-stairs race showed us who had the greatest power compared to their weight. We appreciated that it is this single parameter which determines the ease with which you can lift yourself off the ground. Not weight, not power, but the power per weight. So, ft lb/sec per lb is ft/sec which is the vertical speed up the stairs. What it takes a dynamometer, a weight-balance and a calculator to do, we did straight away in one go with a flight of stairs. Aleksi Halkola was first up the stairs. The seat, the position of seat to bottom bracket, the assumed pilot-weight and the design cadence were all fixed to suit him. We considered that adjustment for leg-length would be too much of a complication. The result has been that those of similar leg-length find it very comfortable and ergonomic. Those who are heavier find that even a static hover is hard going. I suspect that this is because in those conditions the fan is operating outside its efficient range. Those who are shorter and lighter either have to use a makeshift seat-extension, or virtually lie down so as to reach the pedals. But everyone who has tried to hover has done so.

We chose a swivelling, variable pitch propeller for control since it seemed the only method of getting directional control at low forward speeds. This has proved satisfactory, although we have found that the technique of pedalling harder but at the same speed as you apply pitch to the propeller takes some learning and experience and good strong legs.


The bottom line in our optimisation calculations was required power. Ie, how hard you have to pedal. Reducing weight or drag or transmission losses are only means to this end. What if one proposal will be heavier but will produce less drag ? Then you need to know that for this craft :-

Saving 5 lb weight has the same effect as saving 1/6 lb drag or increasing efficiency, ie reducing energy losses, by 1 percent.

or how the main dimensions got chosen What length ? , what beam ? , what area ? I made a set of assumptions of how the weight would vary with size,
eg ( using dimensions of pounds, feet, seconds, radians )
weight of peripheral frame or "floats" = .25 * peripheral length
( We don`t yet know what our hovercraft will measure around,
( but it will help us to decide this if we bear in mind that we do know
( that each foot of float will weigh a quarter of a pound.
( We know this because we made a test-piece of float and weighed it.
weight of frame without floats = 6 + 0.1 * length
A set of assumption of how the drag would vary came mostly from Elsey,
(ref 1) eg wetting drag = .04 * weight

So for a series of possible sets of length and beam I could estimate the power required for lift and the power required for thrust. Then allowing for the power losses in the fan and the propeller, I could estimate the power required from the pilot for this shape at hump speed for that length, which is the value that we used for selection. It was found that within quite a range, the length was not very critical. So length was decided on being able to fit the fan and the crew into the frame, and being able to fit the frame into our workshop. Both of these criteria are only just met. I now feel that the extra pitch stability of a longer craft might be significant.

What hover height ?
We analysed the best height for each speed
More height implies more revs, more fluence, more fan-power required
More height implies less wetting-drag, less prop-power required
Hence there is an optimum which was found to vary with speed.
The calculations indicate that the optimum height for minimum total power
H/L/10000 varies from 3 at low speed to 12 at high speed,
but that this is only marginally better than H/L/10000 kept constant at 10
( H is hoverheight, L is craft length, other symbols defined below. )
In practice, during our early tests, we could see that the gap between
the bag and the ground was far from constant around the perimeter.
Since then, we have adjusted it locally using internal cords.

These are the assumptions made in optimisation of length L and beam B:-

All dimensions in feet, pounds, seconds, radians

PilotWeight =154, H/L=.001, Discharge Coefficient Dc=1, bag/cushion pressure ratio=1.2,

efficiencies:- FanEffy 0.61, transmission TransEffy .90, PropEffy 0.67,
with reciprocals rfe , rte , rpe
Having chosen an arbitrary Length and Beam B for investigation
we derive, using theory from reference 1, Elsey & Devereux :-
HumpSpeed V = 3.2 * square root of Length,
AirDynamicPressure q = V^2 /840
Bow Rad=B/2, Stern Rad=1.25, height = length * H/L
Area S =L*B - .41*BowRad^2 -.43 * SternRad^2,
Perimeter length Oce = 2*(L+B) - .82 * BowRad - .86 * SternRad
("floats" is the perimeter frame which provides bouyancy in boating mode)
("booms" are the cross-beams)
Empty weight without floats = 23.3 + 0.1* Length + .01 * Area + .1 * Oce
Weight of floats = greater of (.25 * Oce) or (.05*weight without booms )
Total weight W = PilotWeight + Weight without booms + Weight of booms
Pressure pc = Weight/Area, EscapeVelocity=29 * square root of Pressure
Fluence vf = EscapeVelocity*height*Oce*Dc cubic feet per second
FanExitStaticPressure= pc * bag/cushion pressure ratio
NetLiftPowerReqd = FanExitStaticPressure * Fluence
GrossLiftPowerReqd = NetLiftPowerReqd * rfe * rte

WavemakingDrag = pc^2 * B * (1.6 * 1.5*B/L)/62.5 ,
valid for hump-speed when 0.3 < B/L < 1.0
ProfileDrag = (2.89 + .75*B) * q
this is air resistance, it assumes a hull height of 1.5 ft and
drag coefficient of 0.5 based on frontal area
plus 2.89 * q which is the air-drag of a cyclist
( This implies that SBW has approximately three times the air-drag
( of a cyclist. I now conside that this is an optimistic
( assumption, bearing in mind that, unlike with an aeroplane, the
( relative wind is not always from directly ahead.
MomentumDrag = vf*V/840
fluence * airspeed * Air MassDensity
Trim Drag = W * sin( H/B )
WettingDrag = .042 * sqr(.001/(H/L)) * W , valid for hump-speed
Drag =WavemakingDrag+ProfileDrag+MomentumDrag+Trim Drag + WettingDrag
NetPowerReqd=Drag*V ,GrossDragPowerReqd = NetDragPowerReqd * rpe * rte
Total Power Required = GrossLiftPowerReqd + GrossDragPowerReqd

We allowed our computer to quantify this for a variety of Lengths and Beams. The optimum length transpired to be 9.25 feet, but with a craft this length there was not space with our configuration for the crew and the fan, so we chose a length of 10.5 where the Power is only 1% greater. This analysis indicated that the power continues to decrease as Beam increases upto a Beam of 10 feet. However, such a width was considered impractical, and we chose 6.5 feet where the power is 1.12 times this optimum. The analysis told us that making it any narrower would have increased the power considerably. So, we did not just slavishly follow the computer's recommendations, but, in making our final decison on these major dimensions, influenced by practicalities, we knew we were close enough to the theoretical predictions that the required power would be close to the optimum. A more compact craft would certainly be more practical, but, by needing a higher pressure would be a lot harder to pedal.

The result of these decisions was a set of dimensions :-
Predicted Weight Empty 44 lbs, weight with Crew 198 lbs
Length Structure 11.5 ft, Cushion 10.5 ft, Overall 13.7 ft
Beam Structure 6.5 ft, Cushion 6.5 ft, Overall 8.2 ft
Cushion Area 64 ft^2 , Perimeter 31.34 ft, Pressure 3.09 lb/ft^2


The next stage chronologically was the design of components. In some cases we made a test-piece. These are described below along with the construction of each component.

A more detailed weight estimate now included the location of each item to give centre of gravity and moment of inertia, using methods as used on the club`s previous projects.

Not having any known theory for this, we decided to cater for the craft being able to turn ninety degrees while travelling one length forward, ie miss a wall or jetty one length in front of position of nose at start. We assumed that resistance was all at perimeter of cushion and that resistance was proportional to velocity. Calculations on these assumptions implied that we needed a swivel of Six degrees. Initially I decided to have a margin above this and hence decided on 10 degree swivel. However, other club members preferred a much larger amount. We have built it with 45 degrees of swivel. Pilots find that they use all of this at low forward speeds. This increase in pylon swivel had the unfortunate effect of causing the propeller to swing close to the bag and under certain semi-inflated-bag states to actually strike the bag causing damage. We have now added a guard-rack to keep the bag out of the propeller swing. There is no other guard around the propeller and ground-crew and bystanders need to keep well clear of it.


With our chosen major dimensions of the craft, the net power required for thrust was estimated at 130 ft lbs/second. With a 3 ft propeller, the efficiency would be 36%, leading to a gross power of 312. This means that two thirds of your leg-power is creating a propeller wake and only one third is pushing the hovercraft forward. To get an efficiency of 74%, we would need a 12 foot propeller. Human powered aircraft have had propellers of this size, but the bigger the propeller, the higher the shaft would need to be. leading to a nose down pitching moment which varies with thrust. Control in this sense is by the crew leaning, We conducted some tests on how far a person`s centre of gravity moves when leaning forward and backward. Knowing what our thrust would be, we estimated that the crew would only be able to cope with a 4 ft high shaft. This implied an 8 ft propeller with a 67% efficiency. In practice, this continues to be one of the major problems. As you pedal harder, the propeller thrust rocks you forward, so you have to lean back. This is not only awkward to do, but difficult to judge how much to do.


We asked ourselves, " How strong to make this craft ?."

What loads would it be subjected to, which we must design it to resist ?

Wave Impact

We assumed 6ft long by 1 ft high waves bufetting against any point of the craft and followed the theory of ref. 1 in calculating the effects. The calculations indicated: max impact 570 lbs, max shear force of 267lbs, max bending moment 8000 lb in, acceleration 10g at bow, 4g at seat and 6g at stern, local pressure 10 lb/in^2 at bow and 5 elsewhere. The frame has withstood quite a lot of battering and only suffered local damage which has been repairable.

Landing on Ground

We catered for the possibility of a sudden loss of cushion pressure and hence the craft falling onto the central pads. This has never happened, but we retain the deep central pads to keep the frame off the ground when landing on uneven terrain.

Pedalling, Power Transmission and Thrust

We calculated the loads, which would be on the pedals, and on each chain, sprocket and shaft at hump speed. And the centrifugal and gyroscopic loads on fan and propeller. All these transmision loads were factored by 6 to allow for cyclic variations in torque, repeated applications and wear. The transmission has withstood much usage. We have replaced some of the smaller of the plastic sprockets with stainless steel sprockets.

Crew Ingress

The rear transverse beam is used as a step for getting in, and this was made to carry the weight of the crew. Other parts of the craft are far too light to do so. Every now and them someone puts their weight on the wrong place when getting in or out. It breaks, and we repair it.


Strong points with loops for attaching lines or lifting are at each corner of the craft. These were designed in right at the start, and have proved absolutely necessary. On one occasion, the empty craft was tethered by only one of these in a strong wind and it took off and flew like a kite. ( That wasn't amongst our original aims either ! )

Structural Modifications

Since building, we found it necessary to add side bracing to the fan support.


Frame Perimeter or "Float"

One of our first test-pieces was a three foot long test-piece of the proposed float. It had to be of 4 inches square section to provide enough bouyancy. It had to have a certain bending strength, as calculated, and to withstand local wave-pressure. It had to be as light as possible. It had to be such that we could make it without too much time and effort. We made a hollow box of hot-wire-cut styrofoam with 1/4 inch square spruce booms at each corner. These spruce booms are then reinforced with carbon-fibre strands. The strength test proved satisfactory and the weight was within our estimates. This combination of spruce and carbon-fibre is used throughout the craft. It exploits the properties of both materials and produces a strong light structure at reasonable cost. As for manufacture of the float, the club-member who helped me make it referred to the task as "child's play". However, we found that making the entire float was very different from the three foot test-piece. It transpired to be an arduous task : we could have chosen a simpler method of construction for the float section, perhaps a solid core of a lighter weight foam.

Step Beams, or "booms"

Two of these, along with the perimeter frame and the central keel-like member make up the frame. The after of these needs to be able to take the weight of the crew when climbing in. The blue Styrofoam core has a full-width spruce cap. Then the whole is wrapped in carbon-fibre-mat. Climbing into the craft is one of the trickier operations for the crew. This is an experimental racing craft, intended only for those who know what they are doing. Most of the structure is not designed to bear the weight of a person. With each new crew member, we need to patiently spend a while explaining the technique. One person after prolonged pedalling of the craft developed cramp. and it was with difficulty that two of us helped them out without mishap. It has also been used for the ingress of a passenger. Carrying a passenger is one of our accomplishments that was not amongst the original set of aims.


This carries the weight of the pilot and the transmission loads. It was made similarly to the float and booms.


The many comparisons that have been done between a human's output capability upright v seated show little difference between the two. An upright cyclist position would have entailed a lot more structure, a lot more frontal area and would have raised the centre of gravity. Hence we opted for a seated position. The exact position relative to pedals was fixed to suit Aleksi. Our seat is a very special shape. On no other seat does the occupant need to be able to resist the effect of pedalling without sliding back and also be able to lean forward or backward. You need to lean back or forward to trim the craft particularly when there is thrust from the propeller. We had anticipated needing a restraint strap, but by careful contouring of the seat this has proved unnecessary. We made a plaster mould to suit Aleksi, and laid up layers of carbon and glass fibre in this mould.

Landing Pads

The centre pads made from semi-flexible packing foam, (cost zero) are shaped to be able to absorb the impact of a heavy landing. We made a test-piece of these.


This was designed using the XROTOR program on an Acorn computer. You tell the program revs, diameter, power and it outputs a shape. This shape is impractical near the root, so you amend and thicken the chord in this area. Then you re-enter your amended shape and it tells you what the efficiency etc is now. Then you can ask it to tell you how it will perform in various off-design conditions, ie different forward speeds and thrusts. It will tell you how much to adjust the pitch for these conditions. The program can`t predict performance at zero forward speed. Our propeller was designed for hump-speed and then I checked that it performed reasonably in other conditions. The variation of pitch is designed to cater for the range from a measure of reverse thrust upto beyond hump speed.

Propeller Shaft and Spars

The propeller spars and the shaft are tapered carbon-fibre tubes. These are customised for their application. The grain direction and width of each laminate is chosen and specified in order that the tube at each point of its length has the required torsional and bending strength. Even when the taper and the angle of the laminate is constant, the shape is far from simple. Initially, I tried drawing these out and cutting the cloth dry from a pattern. But we found it more practical to cut the piece oversize and to trim on assembly. The carbon laminates are laid wet onto a Styrofoam core. In general, we have only basic equipment, but we have made a tapered-cylidrical-foam-core maker, and a very basic sort of "lathe" which enables us to turn the tubes between centres by hand when laying-up. It is also used as a lathe to finally turn the cold-cast bronze bearings which are mounted onto the shafts. These bearings allow the pitch-change.

Propeller Contour

Construction of this is as for the fan, (see below), blue Styrofoam with
glass-fibre coating.


We opted for regular bicycle pedals for simplicity. Chains and sprockets were chosen because we had used them before on the simulator cockpit. The sprockets are Delrin, bought in. The chain is Renold 6mm. The bearings are skate-board bearings. We fix the sprockets to the bearings by locating them in a mould and then casting in Kevlar reinforced epoxy resin. The Kevlar having been previously stitched through small holes in the sprockets. This system has enabled us to build up the block of three sprockets used to distribute the power from the pedals to the fan chain and the propeller chain. Similarly, Kevlar lashing fixes the fan-sprocket to the fan, and in other places. Both chains twist along their length. The fan chain system needs a set of three idler sprockets to guide the chain on its slack side. The twist of the propeller chain varies as the pylon swivels.

Chain Tensioners.

Our first hovers were made with fixed tensioners, adjustable on the ground. Various systems of tensioning were made and tried out. With all of them, the chain would occasionally jump a sprocket. Then we switched to spring-tensioners working on the same sort of principle as on a bicycle with derailleur gears. These have been completely satisfactory.

Fan Design

Based on Elsey (ref 1) page 30 et sec, and having an estimate of our output requirements, we considered a series of diameters and speeds. Following a few such calculations, we decided to abandon the radial fan and to have an axial fan. This decision was influenced by the fact that the club had built a propeller for a human-powered-aircraft but we had no experience of radial fans. I was not confident that we could make one with predictable performance. Having said that, we didn`t actually have a design-method for an axial fan, only for a propeller, the difference is that a fan is inside a duct. Aleksi Halkola and I produced a fan-design-method by adapting Glauert`s 1926 propeller theory. We looked through his theory and changed the equations to account for the fact that the airscrew is inside a duct. This was then written as a computer program. ( Glauert`s theory, unlike Larrabee`s, ignores the effect of the number of blades. ) With this, we could predict the performance of a fan of any proposed dimensions. So, we had a fan-design method, and an estimate of the pressure and fluence that we were expecting to need to produce. A large number of sets of fan parameters were analysed. We found that we would need a rotational speed considerably faster than a person can pedal. We worked on 1200 rpm on the basis that this would be practical to achieve with the crew`s pedal cadence of 75 rpm and two sets of 4:1 step-up gearing. We had decided to use a chain and sprocket transmission, and we felt that 4:1 was the biggest practical ratio. ( 75 * 4 * 4 = 1200 ) The fan was optimised for a fluence of 17 cubic feet/second with a pressure of 4.2 lbs/ft^2 at fan exit. This is what would be needed to raise our estimated all-up-weight to a height to length ratio H/L of .001. This was the lowest value shown on Elsey`s graphs, I didn`t think we would have the power to get higher. The choice of chord was constrained by a requirement for practicality of constuction. This was to have a straight taper, a minimum of four inch chord and a maximum chord to not overlap the adjacent blade. The constraint of having decided a maximum rotational speed was a tricky criterion to meet. The performance of many notional fans were analysed on the computer and the design chosen was :-

30 inch diameter, 15 inch hub diameter, 4 blades
Rad Chord Angle

inch inch degrees
7.5 10.6 11.20
9.375 7.96 8.36
11.25 6.54 7.02
13.125 5.35 6.56
15 4.24 6.85
Section E193

I also checked, by the theory, that the fan performance would be reasonable for when fluence=2* nominal and/or pressure=5/4 * nominal. However, as mentioned above with regard to the difficulty that heavier pilots experience, I suspect that this is because the propeller is then being asked to operate outside its efficient range. These sections are, of course all on the surfaces of a cylinder. For manufacture, I derived the resulting plane sections 2 inches apart by interpolating from points on the cylidrical sections. I also extrapolated, so as to be able to make the blades overlength, and then trim to the cylindrical end shape. The fan has performed satisfactorily, enabling hovering at the expected rotational speed without excessive input power. However, heavy pilots find hovering difficult and lightweight pilots find it relaitively easy. I suspect that this is because with the heavy pilots, the fan is having to operate outside its efficient range. A group at Leon High, Talahassee, Florida, copied our fan shape for their craft,and have hovered successfully. Though the shape was the same, they used different constructional methods. See the construction and operation of a copy of SBW's fan on an excellent video on U-tube at
You Tube

Fan Construction

The fan blades for SBW were made using a technique widely used for human powered aircraft propellers. A set of aerofoil-section shape templates are cut. These conform to the required cross-section at various distances along the length of the blade. An adjacent pair of these is used to hot-wire-cut each section of the Styrofoam core. (For cost of hot-wire-cutter, see Finance, below) These are threaded onto a carbon fibre tube. The outside profile is then covered with glass-cloth impregnated with epoxy resin. Then any imperfections are filled. On our Progress Schedule we had written an item "Balance Fan". This was our quickest done item, because we found that it was already in balance. The fan proved satisfactory in operation.


As first built, there was no freewheel in the fan-drive. So, when the crew stopped pedalling, the fan`s inertia used to cause a tension in the normally slack side of the chain, which could disrupt the mechanism. You had to remember to slow down pedalling slowly. The addition of a freewheel was one step towards making SBW into a practical vehicle. The story of the freewheel is an example of the occasions when we don't get it right first time. I designed a freewheel mechanism. It worked - for about a minute till it wore out. And it made a deafening racket while it was doing so. The off-the-shelf freewheels that we looked at either would not fit into our existing fan or were excessively heavy, or both. Then we found the freewheel off a very cheap bike. It is not completely friction-free when free-wheeling. But, for this application, this is acceptable because the only time that it operates is at the end of a hover, unlike on a bicycle.


We were amazed to observe an upward airflow above the fan. We attributed this to air flowing up through the gap between the tip of the fan and the duct. This gap varied upto about 1/8th inch, 3mm. So, we carefully added material as required locally to reduce this gap to a constant .02in, 1/2 mm all round. We still observed the upward airflow, just the same. All this was before the dome and the curved funnel inlet were fitted. Now that the dome and inlet funnel are added, there is no upflow.


Skirt Design Right at the start of the project, we made a model hovercraft which lifted the weight of the vacuum-cleaner that powered it. This had a segmented skirt. It hovered but it was unstable and biased us in favour of a bag-skirt which is rather like a tyre on its side. The lift comes from the area in the middle, the pressure in the bag being slightly greater. Amyot's book, ref 3, contains an account of how to design a bag-skirt for a small,( assumed engine-powered), craft. A bag skirt seemed to me the right thing for us, and I followed the book on having the suggested larger diameter at the nose, and the gradual change of diameter around the bow. The shape of this was designed first and accurately calculated to derive the various radii and angles for each segment of cloth. The shape was drawn at quarter scale. This shape, of the bag, dictated the shape of the front of the frame. We decided to incorporate an angled planing surface on each of the seven areas of bow of the frame. The rest of the frame, the float, is a simple square section.


We parted from the book in the method of the ducting between the bag and the space underneath. The book shows a series of holes. We reckoned that pedal power would not be sufficient to blow any water out of the bag, if it came in through the holes. So we arranged a system of ducting such that the transfer is above water level when floating. Hence the rigid duct behind the seat and the tent-like appearance of the deck behind the seat. Air from the rear part of the bag flowed up into the "tent" and then after passing through the vent holes, just behind the seat, flowed down through the rigid duct into the area beneath the craft. Since then we have shown that you can pedal and blow water out of the bag, but that was the way that we first built it, and it worked.

Skirt Material

Balloon fabric was a material we had already used on another project, and seemed ideal. It has served us well, being easy to stitch, lightweight and tough. The original proposal was to make all joints with tape. We tried a variety of tapes and none were fully satisfactory. Surprisingly, the least bad is masking tape. On our very first few hovers, we were airborne with the balloon fabric fixed to the frame with masking tape only. Since then, and before venturing onto the sea we stitched loops of cotton tape onto the bag fabric which fix to toggles on Kevlar cords bonded to the frame.

Skirt Manufacture

We were inspired by Fig 8 in ref 3 , the model hovercraft book, to make effectively a skeleton "dress dummy" from white polystyrene sheet. There was a temporary frame ready for each proposed joint in the bag. We made no templates for the cloth itself, we calculated no dimensions of the cloth of the bag. We just "wall-papered" the cloth onto the temporary frame and joined it to the next piece with double-sided tape. Then the entire thing was carefully removed and the seams were stitched. The seams were reinforced with cotton tape or strips of Melinex in places where the grain of the cloth of the bag is not in line with tensile forces. After a while these Melinex strips wore or cracked from contact with the ground, and were replaced with cotton tapes. Actually, cotton, is also not the ideal material for this application because, when wet, it shrinks more than the balloon fabric.


On this craft, most of the deck is balloon-fabric. This is far and away the lightest way to do the job, there is absolutely no need for a rigid board. This was made in a similar way to the bag, but in some places we measured and cut without using temporary frames. The side deck panels are arranged such that there is only tension at the sides, not back and front. Thus at the back and front there is tape only, no need for toggle fixing. Making the rear deck was a complete adventure into the unknown for us. It is satisfying that it inflates to such a pleasing shape. Someone once called it a "fairing", which, to a certain extent, of course, it is.

Attachment of Bag and Deck to Frame

In most places, the edge of the cloth wraps around the frame. Thus we can form a seal by having a strip of double-sided tape along the edge of the frame, and at the very edge of the cloth are the tapes and toggles. It all had to be made such that it could be removed, and that we would have some idea what the tension under pressure would be. Both the bag and deck have given good service. The occasional rip has been stitched. A couple of toggles have been torn off by pounding of waves. These were easily replaced. However, it takes several people several hours to either remove it or to replace it. In theory it is as easy as doing up a duffle-coat and sealing an envelope. But you have to remember to do things up in the right order before some of the toggles become inacessible.

Skirt Pressure Test

The pressure in the perimeter bag needs to be a specified amount above the main lifting pressure below the craft, as prescribed by ref 3. We planned to arrange this, by adjusting the size of the vent holes between these zones, until the pressure difference was correct The club had made a water tube manometer for a previous project and we reckoned on making a second one of these so we could measure two pressures at once. Then we realised that with such low pressures, the difference between the two that we were aiming for would hardly register. So we came up with another idea. We would directly measure the shape of the inside region of the bag. The difference in the air pressure underneath SBW to that outside is as small as the difference in atmospheric pressure that you experience in climbing five floors of a building. The difference we were trying to measure was equivalent to a couple of floors. We rigged up a dipstick as shown, a lightweight stick in a guide tube. So while it was being hovered we allowed the stick to be supported by the inner bag fabric. It worked well. Aleksi was doing all the crewing at this stage.


After quite a lot of operation with the bag as described above, we made the decision to add fingers to the bag. To convert from one of the standard hovercraft skirt types to another. In designing the fingers we had to begin by learning the fundamental principles on which they worked, getting advice from experienced people at the Hovercraft Museum. The task was made more difficult by the fact that we had to ensure compatibility with an existing flexible bag. Direct analysis was not possible, so I wrote a computer program which had two Newton-method loops, and, after crunching the numbers, showed a picture of the bag and finger which resulted from each set of input parameters. After many runs of this we chose a design for a side-finger and a different one for the aft. We took the opportunity of extending the bag aftwards to better trim the craft. This was based on ballast tests, see videos. We made a few full-size test fingers and tested these. Our first test looks like a toy theatre with the fingers being the curtain. The next test-piece was a circular board with a ring of twelve, full-size, fingers. This was effectively a small hovercraft and it lifted the battery-powered vacuum cleaner that powered it. Unlike our first attempt at such a device, this one was stable: it stayed upright. Following our usual policy of making everything as lightweight as possible these fingers were made from half thou, (.0005inch, 12micron), Melinex. This is the standard material for the wing skin of human-powered-aircraft. We also learned, after we had made them that in fact it is the identical material used for the skirt of the 1966 Alikat, a sidewall craft which hovered on flat calm water but could not make forward progress or operate at all on land. SBW, fitted with these fingers performed better than without them. However, the slightest roughness on the ground would cause rips and they needed frequent repairs. Before long some of the fingers seemed to consist of more repair tape than Melinex. One advantage of the material was that they did look rather fine: their transparency showed that we were really hovering off the ground. Also the material has the advantage of being isotropic whereas balloon fabric stretches across the bias and sometimes needs diagonal strips sewn on. However, their lack of durability implied replacement. We have now made a new set of fingers, shaped as the first ones, but from the same material as the rest of the skirt. These were first tested on 14th June 2008 and so far seem to perform O.K. On this set, the aft fingers are half the previous width.


The Pilot`s feet are busy pedalling and the pilot's upper trunk is sometimes also leaning. The right hand is controlling direction and the left hand is controlling the propeller pitch. Propeller pitch is more or less the equivalent of "gear" on a bicycle. You choose pitch appropriate to your forward speed. A coarse pitch offers more resistance to the pedals, with the same sort of effect as a "lower gear". Whether or not it would also be beneficial to have a choice of "gears" is one of the ongoing debates amongst those involved with the project. Another is whether the current steering handle is appropriate. Some people say it works "backwards". It is necessary that leaning and pedalling will not interfere with hand movements or cause inadvertent hand movements. This is effected by the gripping effect of the lower part of the seat and the position of the control handles. A simple push and pull on the right causes the pylon to swivel via a push-rod. Operation of the handle on the left alters the pitch of the blades via a Kevlar cable which passes around pulleys and moves a rod which passes through the propeller shaft. This rod has a ball joint partway along its length such that the rear section is free to rotate with the propeller. The rod operates levers on the root of each of the two propeller blades. The pitch is adjustable from 30degrees finer than nominal upto 23degrees coarser than nominal. When set to finest pitch, the craft is propelled backwards.


The website shows videos of many of the people who have become airborne under their leg-power alone on Steam Boat Willy. Pilots have an age-range of 17 to 70 and are from many parts of the globe. Everyone who has tried to hover has been successful. Weight-range is 8stone to 16stone,(112 to 224 lb, 50 to 100 Kg). Passengers of the same weight as the craft have been carried.


Maybe you are wondering
" How much did all this cost ? "
and " Would I be able to afford to do it ? "
or " Where can I get a grant from ? "
or even " Can I make a profit out of it ? "


What we asked ourselves was
"How can we arrange to start/continue with this project under our current financial circumstances ?" Here I describe how we went about doing this with very litle in the way of grants and with none of us being millionaires. To give a proper "account" of historical costs would not be very helpful since your circumstances will be different. As probably will be the materials you choose to use, etc. We sometimes get asked "Are you going to produce and sell them ?" When we started, we never thought that we could make anything that even looked marketable, so it is cheering to be asked this. A production version would need to be a lot more practical, and take a lot less hours to make.

Logical choices

It would clearly be illogical to spend a lot of money to save a small amount of weight and later say you can`t afford a small amount of money to save a large amount of weight, when making a different decision. Whereas, if a standard for Pounds-Sterling per Pound-Weight-Saved had been agreed at the outset, both choices would have been made the other way round, thus ending up with a craft which was both lighter and cost less. We never actually did this. If we had then maybe SBW would be even lighter and have cost even less. As things are, weight is not our biggest problem, any more than finance is.



Our approach on materials is to use the grades of material that have become standard for human powered aircraft. These include carbon-fibre, aircraft-grade epoxy-resin which are expensive and Styrofoam and spruce which are not. Something to avoid is to submit to a feeling of "bad-value" when you are paying a high Pounds-Sterling per Pound-Weight rate. For a craft which must be lightweight, this can signify good value. Also, bear in mind that only part of a sheet material will end up on the craft. Whereas with carbon rovings for instance there is virtually no waste, you take only the length that you need from the roll. We made a balance which weighs to an accuracy of a tenth of a gram, thus enabling us to accurately measure out resin in small quantities, thereby saving money The spruce strips used for the first tests were bought from an aeromodelling shop. Luckily one of our members had access to a bandsaw and after carefully selecting lengths of spruce from an ordinary timber yard, was able to cut these lengths down to the section sizes that we mainly use. 1/4 by 1/4 inch and 1/4 by 1/8 inch ( 6 by 3 and 3 by 3 mm). These strips provide the basis for a construction method that is both strong and economic. After making the component with the spruce, it is reinforced with a small amount of carbon-fibre in the areas of high stress. Similarly with Styrofoam. The shape is carved, then reinforced with glass-fibre and/or carbon-fibre. This is quicker and cheaper than making a mould for the carbon-fibre parts. We only made moulds for the seat and for the mounting of the bearings in the sprockets. The skill to do all this was acquired by making a part, then looking at it, and making a better one, then making another better one until we made a good one. Usually these practice pieces could be made using scrap offcuts, but our first propeller stub was not satisfactory, so it became inadvertently a "practice piece" and we had to make another one. Then we had to make a third. The stub is a tapered carbon-fibre tube. Not something that is easy to make. Or, you could say, not something that you can be sure of getting right first time. The point is that although we were using this expensive material, each scrapped stub only cost about the same as eight pints of beer. Some of the materials were leftover from a previous project, some were given, some, such as the packing foam used for the landing pads literally came out of a rubbish-skip.
See materials-stockists on RAeS HPAG site


The hot-wire cutter that shapes most of the Styrofoam parts on SBW comprises the actual hot-wire, and a few feet of ordinary electrical wire and a couple of crocodile clips. Total cost less than a pret-a-manger sandwich. Other tools are makeshift from timber offcuts and cardboard. It is the expendable items like the various grades of abrasive paper and masking tape that are an ongoing expense. Probably amounting, in some weeks to as much as I pay for water through my home-water-rates. There is a standing joke that we spend more on masking-tape than we do on carbon-fibre. If you throw in the cost of other such mundane expendables, then its probably true. We do have a donated pillar-drill, and the use of a member's sewing machine which cost about the same as a low-price new bike.


We have no lathe, and so we either had to design parts to not need such machines or we had to get someone else to do the job. Alan Craig, a Bognor BirdMan prizewinner, made most of the machined parts and John McIntyre, designer of Airglow, one or two essential components. Most of our sprockets were off-the-shelf from HPC, but this firm also excellently made some special parts for us.


If you do budget, bear in mind that we have had to pay for the materials for our Transport-Box, and tarpaulins to cover this, and further tarpaulins for operating over grass. We have had to pay fares and transport costs, and location fees. Incidentals have been a substantial proportion of our expenditure.


We did have some income from grants from the College Union at the outset. But it was a lot of hard work applying for these, and then more work each time we wanted to actually spend any of the money that was supposedly "ours". As a rate-per-hour method of acquiring income, you're probably better off begging on the street. Our main source has simply been member's pockets. Now that we can put on a reasonable performance of hovering, we have been paid for doing displays. Event Organisers please note. Here is something unique. If you book us then we are yours for the day. When you say "Hover", we say "How high ? ". Indoors, outdoors, on land or water.


Deck drains. Previously spray onto the deck lingered there adding to the
Propeller-bag guard. We had some mishaps with the propeller striking the
bag when partly inflated on water.
Chain and sprocket spray-guards. Salt water had splashed onto these
causing corrosion problems.


The continued operation of the craft under various conditions will undoubtedly indicate where development will be beneficial. The effect of modifications will be reported on in


1 "Hovercraft Design and Construction", G H Elsey & A J Devereux, David and Charles, 1968
2 "Hovercraft Technology, Economics and Applications", J R Amyot(ed), Elsevire, 1989
3 "A Guide to Model Hovercraft", Paul Taylor(ed),HCGB,1998
4 "The Constructors Guide",Jeremy Kemp,HCGB, 2nd edition 1994
5 "HPV News",International Human Powered Vehicle Association
6 "Steam Boat Willy", University College London Union Human Powered Flight Club,April 2000
7 The project website


British Engineering Units are
feet, pounds, seconds, and accelerations are measured in g = 32 ft/sec/sec
1 pound mass produces a force of 1 pound weight per g.
1 ft = 3.28 metres
1 ft lb/sec = 1/550 HP = 1.356 watts
1 lb mass = 0.454 Kg
1 lb weight = 4.45 Newtons

:: join in.

To find out how you can join in with this exciting project please get in touch.

All content © 2023 Steam Boat Willy :: Sitemap :: Links :: Contact :: London Website Design by Coffee Bean Design