The farmer-built Savonius rotor: A low-tech approach to renewable power for farms

Our project initially proposed to design, build, and evaluate a single "Mark I" prototype. Design and building phases were finished in 2010, and evaluation was wrapped up by late summer 2011. But rather than conclude the project at that point and begin outreach we elected to repeat the sequence of design, building, and evaluating with a second "Mark II" prototype that built on our experience and eliminated some of the shortcomings of the Mark I design. We feel that on the whole this effort resulted in a superior research product of greater utility and benefit to farmers. This section will describe in detail the sequential phases of design, building and evaluating the two separate designs.

Our first design is the product of project collaborator Victor Gardy's 20 year history of working with the Savonius VAWT concept. It is a modular unit based on two rotors and a drive disc. At nearly 6 feet in diameter, the rotor is as large as the nature of the construction material (1/2” thick plywood in 4' x 8' sheets) will easily allow. Other aspects of the design and construction follow from the sizing of the rotor. We chose to use standard exterior plywood and regular spruce lumber for our prototype with the understanding that a future permanent device could be built out of more durable materials. I would imagine that a rotor built from pressure treated components could last 10 years with a coat of paint or stain from time to time. Some sort of a roof over the rotor would not affect performance and would extend its working life.

Savonius rotors are also made out of steel, and I have seen photos and footage of such wind devices installed on flat urban rooftops in Australia and Scandinavia. We strongly considered testing the unit on a platform on top of one of our steel farm silos. Although we believe that the silos could have handled the load without problems, we couldn't devise a way to do the necessary pre-installation construction at that height (65 feet). This is too bad since there is superior wind available at that elevation. The Savonius windmill we built consists of a housing tower, two rotors, a drive disc, and a vertical shaft. The housing tower supports the shaft in bearings at the base and at the top, and also serves as a framework for the stator panels. Victor Gardy found the use of stator panels, which shield the back side of the rotor cups from turbulence as they turn into the wind, to increase efficiency. The rotors are mounted onto a 1” (inside diameter) pipe shaft. The pipe shaft is stabilized inside a sealed bearing at the top and is bolted to the plywood drive disc on the base. An automotive wheel bearing beneath the drive disc connects it to the frame. In early 2010, Victor Gardy built a 1/6th scale working model of the rotor. This helped us visualize the design and refine the construction methods as we began building the prototype on the farm.

The key feature of this windmill, or any other, is the “swept area.” This is the figure on which all wind-generation math is based. Our windmill features two rotors, the upper one set at 90 degrees to the lower one. At any given time, one rotor is catching the wind at the point of maximum mechanical advantage, and the other collecting rotor is at a point of reduced mechanical advantage. All together, we estimate the swept area to be 1.33 times that of an individual rotor sail. Each sail is four feet high and 32 inches across. This works out to a total of 14.9 square feet or 2.64 square meters of sail area. This seemed to be a good size for a basic portable experimental unit.

With the design outlined by Victor Gardy, we began building in the summer of 2010. The construction of the tower is all basic framing lumber, primarily 2' x 4's. We chose to use 1/4” bolts for strength at the joints. The draft design featured a great deal of diagonal bracing. We have eliminated most of this bracing from our recommended design in favor of a structural stator panel that will perform a bracing function. Few difficulties were encountered with the construction of the tower, other than the difficulty of moving it around once assembled. Each side is 6' x 12', and assembled the entire tower is 6' x 6' x 12' and rather heavy. We were still able to move it around with the tractor.

Assembly of the tower takes a few days at most. We recommend leaving the center upright and the stator panels off of one side for later insertion of the rotors. The rotors are assembled of pieces cut from 1/2” plywood. There is very little waste. We used elastomeric membrane (such as Grace brand “Ice and Water Shield”) at the assembled joints and reinforced them with blocking. We used weather-resistant decking screws, and reinforced the joints between the sail panels with light sheet metal, bent to conform with the angle by hand and attached to the plywood with sheet metal screws. At the center of the top and bottom panels, a hole is bored for the pipe shaft and 1 1/2” pipe flanges sandwich either side of each hole, attached to each other with stove bolts and nuts.

An assembled rotor can be lifted by two people easily enough. Lastly we constructed the drive disc, and assembled it into the tower with the pipe shaft and accompanying bearings. The drive disc is 70” in diameter, and is assembled of two sheets of 3/4” BC plywood, glued and screwed together in two layers and cut to a precise circle with a circular saw and radius jig. It required some innovation to get a good circle with basic woodworking tools but we were able to get a good result. Initially we put a 1/4” rubber tire onto the outside of the circle with contact cement but this we later removed as it seemed unnecessary. The smooth, 1 1/2” wide edge of the drive disc is, all by itself, an adequate surface for a friction transmission to a generator. We installed a Chevy Lumina wheel bearing in the exact center of the drive disc with bolts. We chose that model of bearing because it is cheap and has a simple bolt pattern. The pieces left over from the building of the drive disc were sufficient to build a sturdy support box, five pieces assembled with screws. We mounted the stationary plate of the wheel bearing onto this support box. The giant wheel balanced perfectly and spun with very little effort on the bearing. We placed the disc on its support box into the tower. We then put the rotors into the tower and inserted the shaft through the two rotors from above, and threaded it into a pipe flange on the drive disc.

The unit is designed to spin clockwise so that it will not unthread this joint. The finished unit can be balanced by adjusting the placement of the support box below and the top bearing above. Finished clearances are fairly tight, within a few inches. Once assembled we found the array to be quite sturdy, and it could be lifted from any corner without distortion. The rotors spun in a light breeze. Initially one of the two towers we built had a slight bearing noise, a slight groan that happened once per revolution at low speeds only. But this disappeared after a month of operation and otherwise the unit is very quiet. We proceeded to the evaluation phase in early 2011. Overall the construction presented no problem. Modest competence with a circular saw and drill are all that is required to build a similar rotor. We had just under 80 man-hours in construction time.

We began evaluating the rotor's performance in February 2011, and concluded in September 2011. Initially we constructed a side-by-side installation of two rotor towers, each with their own drive disc. We believed that by placing the drive discs so that both would simultaneously bear on the drive wheel of a generator, that combined net generation would exceed generation of two independent generators, as pictured below: However we abandoned this approach in favor of independent generation. We also decided that the side-by-side installation also compromised our ability to evaluate the basic rotor array since each tower could only receive winds from three sides, so we separated them and continued to trial just one single tower. The generator can be an automotive alternator (the higher the amperage rating the better) or an exercise treadmill motor. Exercise treadmill motors make good generators since they are designed to function at a range of speeds. An inline skate wheel is mounted on the shaft of the motor, and the motor is fastened into the tower so that the wheel coasts against the moving drive disc, with enough pressure to turn, but not so much as to produce unnecessary friction. Sometimes we found the use of a nylon come-along to be a useful “spring” to help achieve the right amount of contact pressure. We had persistent problems with moisture getting into our generator and diminishing generation potential. It is important to house motors in a way that sheds water, yet allows modest ventilation. One of our mistakes was to mount the motor below the drive disc with the shaft pointing up. If mounted this way, any water accumulating on the surface of the drive disc will spill off the edge and into the motor. If the motor is mounted above the drive disc instead, can be protected by a simple bucket placed over it, and water on the drive disc will not affect performance. Only once in the winter did significant amounts of snow fall without enough wind for the rotor to keep itself clear. In other snowfalls there was enough winds that the rotor brushed all its working parts clear as snow fell. Our first installation site was right next to our house.

The rotor remained in this location through late May 2011. We then moved the rotor to a pasture location where we have observed winds to be stronger and steadier than at the farm homestead. The rotor was fairly easy to move; we lifted it onto a haywagon with a tractor bucket and a farm jack, and rolled it to the new location. It remained on the haywagon through the end of our trials in July 2011. We have gotten a good mix of strong winter/spring wind conditions and milder summer winds. We installed a data logger in order to record wind speed, direction, and generator voltage and current. The output of the generator ranged from about 10 watts at the minimum tip speed of about 5 knots to 1800 watts in sustained high wind conditions in the 30 knot range. We did not store the resulting electricity, but instead applied it to a load (12 volt lightbulbs, and at one point an electric stove burner). Our main focus was measurement. The rotor presented no mechanical issues during the testing period. We never stabilized it with guywires, relying instead on the broad 6' base for stability. It never fell over or sustained damage from high winds despite gusts of up to 45 mph during the test period, and despite the additional 3 feet of height when it was in the pasture location on the bed of the wagon. We found that the rotor would not spin much faster than 60 revolutions of the rotor shaft per minute, and beyond about 32 mph, additional wind seemed to decrease rpm slightly. Here is an approximation of shaft RPM under light load at various wind speeds: Although the rotor could not not exceed 60 rpm very often or by very much, under high-wind conditions it turns at that speed with greater force.

If we had the means to engage multiple generators for these conditions, we would probably have been able to greatly increase produced wattage. We believe that a second generator, would slow the rotor less in 30 mph winds than in 20 mph winds, even though the observed rpm with a single generator is much the same. We did not have a torque sensor as these are very expensive, but we observed that stopping the rotor under very high wind conditions was much more physically demanding than stopping it under moderately high wind contions, even though the observed rpm was not much higher. The downside of having multiple generators permanently attached is that the additional load on the drive disc would increase the “tip speed,” and render the windmill unproductive in a greater percentage of light-wind conditions. Because the 70 inch drive disc has a circumference of 220 inches, each rpm of the shaft turns the 3 inch inline skate wheel on the generator 23.35 times. This is the gearing mechanism by which the slow-turning rotor shaft turns the generator at operating speeds with a minimum of friction. The value of this first prototype was assessed in various ways. On a strictly economic analysis of its electrical generation capacity, it proved to be of slight benefit. In our conditions, at which there is an average of 11 knots of wind at or near the surface Based on our collected data for the evaluation period, a single two-rotor tower would produce about 650 kWh of power per year.

At this rate, depending on whether labor costs were included and if so at what rates, the unit could take up to 16 years to recoup the costs of building it. In order to last this length of time, the rotor would need to be constructed out of pressure-treated lumber and plywood, and the exposed edges of the plywood should be encased in metal U-channel to prevent delamination. These upgrades might increase the material cost to about $870 per two-rotor tower. In order to enhance durability we began to consider building a second prototype from steel. 55 gallon drums are often used for Savonius rotors, but the resulting sail area is considerably less than on our plywood prototype. Steel 275-gallon fuel oil tanks on the other hand, present an ample sail area and seemed to have the potential to make a great rotor cup, and as a result of our experience with the Mark I we began to consider steel fuel oil tanks as an alternative low-cost material as part of a revised design. Another observation was that the tower could be much simplified. There may be advantage to the use of stator panels but we were not able to establish this for sure. At any rate most contemporary Savonius designs do not feature them. A simpler tower without stators can consist of just two uprights with a cross-brace at the top of the shaft and another at the bottom. These members can be made out of 4' x 4's or 4' x 6's, and joined at the corners with steel reinforcing plates. Such a tower would be quicker to build and more durable but would need guywires to keep it upright.

The Mark II design aimed to build on the baseline research established by the wooden rotor and enhance durability and tower design without substantial additional cost. The rotor array is all steel for which a different skill set is required than for wood construction, but metal cutting and welding skills are common and the design is easy to execute. Overall we aimed to retain the advantages of the Mark I including: 1. Simplicity of design and construction methods 2. Use of low cost, common materials 3. High ratio of power output to material costs While also: 1. Reducing the complexity of the tower 2. Eliminating the drive disc in facor of a PTO-style power train. 3. Enhancing durability Babbit bearings were an obvious improvement over the Mark I's automotive-style wheel bearings in terms of ease of installation and their ability to support the shaft at an intermediate position. This makes it easier to utilize power from the lower end of the turning shaft, which is accessible in the open area below the rotor. A variety of means of transmission are possible, including v-belts, bicycle or equipment chain, or flat belts. In the end we decided to go for PTO connections and universal joints as this seemed ideal for use with a planned archimedes screw pump. We decided to build a three-rotor tower using three scrap 275 gallon fuel oil tanks. This size would be small enough to still be easy to install without a crane yet powerful and durable enough to perform long-term service powering an irrigation pump to serve 6.5 acres. While others may find any number of ways to best utilize the power of wind, it became apparent over the course of our work with the Savonius that using our windmill's power to move water would result in a clear savings of both money and labor. As a result we planned to permanently install the second device adjacent to a reservoir used for our wet rice fields. With our revised design, we began construction in fall 2011.

The rotors are comprised of halves of 275-gallon used fuel oil tanks, which were obtained from a fuel company dumpster for free. The legs and pipes were removed, and they were cut in half and welded onto a 2” outside diameter steel shaft. Once cut the tanks had a slight element of floppiness, so 3/8” steel rods were added as stabilization straps as shown. In addition, extra steel plating was added to reinforce the weld joining the tank to the steel shaft at the top and bottom. Where the tanks joined each other, it was sufficient to weld them to each other. The three rotors are set at 60 degrees intervals on the shaft, rather than the 90 degree intervals on the two-rotor shaft. We chose to weld the rotors permanently to the shaft, creating a complete structural unit, rather than to have any bolted connections. The shaft stems out above and below the rotors. We used babbit bearings that were easily bolted to the tower and allowed the shaft to stem out below the bottom brace for easy connection to a 90-degree gearbox in that space as shown. A short spacer pipe transfers the weight of the array from the weld on the lowest sail to a locking collar, which rides upon the lower babbit bearing. We welded a category 1 male PTO stem to the bottom of the shaft and were able to use this to couple a 90 degree belt drive gearbox using this connection. To the gearbox we welded an additional PTO stem to which a shaft can be coupled. The belt drive pulley can also be used for power transmission with a flat belt, and can also be used to friction-brake the rotor for servicing. The revised tower is comprised of pressure-treated 4 x 4s and stands about 19' tall. We installed the tower without a crane or gin pole. First the ground post 4 x 4 s were set in concrete footings and cut off at 5 feet of height and cut for a lap joint. Next the 16' tower sides were cut for the opposite face of the joint below and assembled with a 4 x 4 and a 2 x 8 at the top. 2 x 6 x 8' members carriage bolted to the outer face of the lower section of the tower add strength to the lap joint area. Once the top of the tower was complete, we bolted the lap joint connection with a single bolt. The top beam was now laying on the ground with the lower legs of the tower bolted onto the ground posts so as to serve as a “hinge” during raising. We laid out ground anchors for guywires as follows. The guywires were made out of 1/4” steel cable and were cut to approximate length. We attached the two cables on the same side of the ground posts as the upper part of the tower was lying before we lifted. This way the cables would come taut when the tower became vertical and would not allow it to fall over to the opposite side. The remaining two cables were attached to the upper corners but hung free for the time being. Using a tractor bucket and poles for initial lifting and a light winch from the opposite side once lifing was underway, the tower went up easily. Once bolted together with 3/8” galvanized carriage bolts the tower was easily made plumb by adjusting the guywires. With the guywires tight the tower was now strong enough to lean an extension ladder against the top beam and climb. In December 2011, the assembled rotors on the shaft were brought into position on an 8' x 16' haywagon. The base of the shaft was moved close to the lower beam, with the upper end lying on the ground. Some denting of the rotor cups occurred during this moving but we were able to pound them back out with sledgehammers. Then the bearing and bearing bracket were assembled ont o the shaft and strapped to the beam with nylon ratchet straps, allowing enough slack for the bracket to rotate during raising. We chained a pulley onto the top beam and threaded 1/4” steel cable through it, with one end looped around the upper stem of the rotor shaft and the other attached to a tractor drawbar. A farm jack was used to assist with the initial lifting of the upper end of the tower—the tractor pulling has little mechanical advantage until the shaft is 20 degrees or so above vertical. It would have been good to have several volunteers to help start the lift but we accomplished the job with just two of us. Once the shaft approached vertical, we climbed the tower and passed the top babbit bearing over the top of the shaft and bolted it to the top beam using 1/” x 8” carriagef bolts. Then we the farm jack and the tractor bucket to lift and position the rotor array to affix the lower bearing bracket to the lower beam with carriage bolts. The weight of the device being considerable, we found it necessary to install a 4 x 4 diagonal brace to support the steel bracket and to prevent it from rotating from the downward pressure of the shaft onto the babbit bearing. We checked the final plumbness of the bearing bracket with a spirit level. Once plumbed the rotor turned under light winds with no bearing noise.

This second prototype was, as mentioned earlier, installed in a permanent location adjacent to an irrigation reservoir pond. This site is in a broad flat plain with no wind obstructions within 1000 feet in any direction. We becan evaluating this unit on January 7th, 2012. It was found to be very quiet except when turing at very slow speeds, when bearing noises could sometimes be heard at close range. In the following two weeks we were able to observe it during snowstorms, light winds, and in heavy winds up to 35 mph. It was observed to turn in very light winds (5 mph or less) and did not suffer damage in high windspeeds. In fact, regardless of windspeed the RPM of the unit was never observed to exceed 60, which mirrored our experience with the mark I. This Mark II unit does not incorporate a generator so we opted instead to evaluate its production by measuring the velocity and force ot the turning driveshaft in order to establish horsepower under typical wind conditions. The windmill drave shaft terminates in a PTO stem at its lower end. To this is coupled a 90-degree belt thresher gearbox designed to bolt onto a tractor. This bolts onto the lower bearing bracket which has holes drilled to receive it and to allow the output shaft to be pointed in multiple directions. Our gearbox increases the RPM of the primary driveshaft at a ration of 2.5 to 1. The thresher belt pulley has a circumference of 28.26 inches (2.36 feet). Under winds of 6 mph (just under the site average) the rotor turned at 30 RPM, which results in the thresher belt pulley turning 75 RPM. 30 RPM tower shaft * 2.5 pulley gear ratio = 75 pulley RPM Establishing torque is slightly more invloved, and involves the use of a “prony brake,” which we constructed using a short leather belt, a lever and a pair of spring scales. One spring scale is hooked to each end of the belt. One of the scales is hooked to a fixed point on the frame of the rotor, the other to a lever. When the belt is looped over the rotor's belt pulley as shown when the pulley is not moving, and force is applied to the lever, both scales register equal poundage of force applied. We then removed the belt and allowed the rotor to resume turning with the wind. When it had reached full 30 RPM in 6 mph winds, we reapplied the belt and applied tension with the lever. When the lever scale registered 35 lbs and the fixed scale read 5 lbs, the rotor began to slow somewhat from the braking force, so we took a reading at this point. The fixed scale reading is subtracted from the lever scale for a net load of 30 lbs. To determine horsepower, we used the following equation: Hp. = net load x circumference x RPM / 33,000 Inserting our numbers into the equation gave us a horsepower reading of 0.16 This exceeded our expectations given the light winds. Even under such light wind conditions the windmill has enough torque to turn our archimedes screw pump to pump 4500 gallons of water per hour, or to turn a grain mill. Using this rough measurement, engineer Sam Gorton and I approximated the efficiency of Mark II device (or, in other words, the percentage of the energy in the wind that it captures) at around 30 %, which is quite respectable for a VAWT. This is an estimate of course, as the exact efficiency could only be completely and accurately measured over time with a torque transducer that was beyond our budget for the project. Nevertheless these results suggest that this device is likely to be a cost effective energy-capturing tool. Of course any use of this captured energy must be adaptable to the lower output shaft speed and any efficiency losses through gearing or transmission. Using our approximate efficiency rating, we estimate that the Mark II device would produce 182 watts of actual power under typical site conditions of 11 mph winds, for an estimated total annual production of around 1600 kilowatt hours per year, which for our farm is equal to about 3 monthly bills of about $110, or about $330 per year. The unit only required $935 in materials and thus has a quite low payback period of only 3 years. If the fabricators are paid $25 per hour to build and install it (a total 104 hours in our case) for a total labor bill of $2600, then adding labor costs still results in a low payback period of 10.7 years. Most likely this Mark II device could function well beyond that payback horizon.

• the tanks halves are stabilized on a vertical shaft with steel straps and reinforcement plates
• rpm of the mark I unit plotted against windspeed
• The Mark I unit ready for testing
• Mark II performance curve
• Mark II Savonius dimensioned drawing