鈥淧ut your finger between the blades,鈥� says Merhdad Zangeneh, pointing proudly at the small, strange-looking turbine on the desk in front of him. Between each of the weirdly convoluted blades there is just enough room to slip a finger. 鈥淐an you feel the way they curve?鈥� he asks. The cold steel blades are warped like no others. This is a 鈥渄esigner鈥� turbine and it bears about as much resemblance to the conventional models found in aero engines, water pumps and gas compressors as a Gucci gown does to a ready-to-wear jump suit.
Designer turbines are not expensive accessories for fashion-conscious engineers but they are set to improve the performance of almost every type of turbomachine. They are the product of a revolutionary approach for designing and manufacturing turbines that Zangeneh has perfected. A mechanical engineer at University College London, he says his technique has made it possible to design turbomachines that work more efficiently over a wider range of speeds with less vibration than traditional designs. They can be made smaller and more powerful and, to top it all, they will be cheaper to build and operate.
His work has already been snapped up by the world鈥檚 largest manufacturer of pumps, a Japanese engineering company called Ebara. At its manufacturing plant in Fujisawa, a small town some 80 kilometres southwest of Tokyo, Ebara has invested heavily in developing and testing prototypes based on this approach. If the tests are successful, Ebara hopes to use the technique to produce full-scale designer turbines.
Advertisement
Sucking or spitting
The principle behind turbines is simple. When the blades rotate they are converting the energy of the liquid or gas flow into mechanical energy. This mechanical energy can then be used to do work, such as generating electricity in a power station. The same principle works in reverse. A turbine鈥檚 closest relatives are pumps and compressors in which rotating blades create a flow of liquid or compress a gas. In these machines, the set of blades is known as the impeller. A common example of an impeller is an office fan which sucks in air along its axis of rotation and spits it out in the same direction. Other impellers can change the direction of flow. These 鈥渟uck鈥� in the same way as an office fan but 鈥渟pit鈥� the gas or liquid away from the hub in a radial direction, changing the direction of flow by 90 degrees. Zangeneh has applied his novel design techniques to these radial impellers.
Although the ideas behind pumps and turbines are straightforward, the way a gas or liquid flows through them is immensely complicated. Ideally, the flow should be smooth but, near to the blade surfaces, friction can create havoc. As each blade moves, it drags a thin layer of the fluid along with it. This is known as a boundary layer and its behaviour has a profound effect on the performance of turbines and compressors.
Boundary layers occur on all aerodynamic surfaces and a common example of the problems they can cause occurs on aircraft wings. During normal flight the air in contact with the wing is stationary while the air several centimetres above is moving quickly. In between, in the boundary layer, the speed of the flow gets greater with increasing distance from the wing. But in certain conditions 鈥� for example, when the aircraft flies very slowly or at a steep angle 鈥� the boundary layer can become detached from the wing and the smooth flow becomes a turbulent maelstrom of eddies and vortexes. The result is the rapid loss of lift known as a stall.
In a pump or compressor the problems are more complicated. 鈥淒uring some types of stall the flow over the blade is like a tornado,鈥� explains Zangeneh. When a stall occurs over one blade, the flow passes to the other blades. But as a smoother flow re-establishes itself, the turbulence can pass to the adjacent blade which then stalls. If this process continues, the stall gradually rotates around the pump, creating dangerous vibrations as it jumps from blade to blade. In certain conditions all the blades can stall at the same time. When this happens the fluid plunges backwards and forwards through the blades as the flow stalls, re-establishes itself and then stalls again. This is called surge and the vibrations it creates can tear a pump apart.
The boundary layer also falls victim to other problems. Because each blade is rotating very quickly, the boundary layer is forced along its surface in a radial direction by Coriolis forces 鈥� the same effect that moulds weather patterns as the Earth rotates. 鈥淚t scrapes the low momentum fluid off the blade surface and dumps it at the trailing edge where the flow is fastest,鈥� he explains. This accumulation of low momentum flow can also cause stalls and until Zangeneh perfected his design techniques nobody had succeeded in controlling it.
Stalls and surge can be avoided by ensuring that the pumps or compressors never operate at certain critical speeds and flow rates. However, passing through these critical zones is often unavoidable when the pump is first switched on and limits the safe operating range of compressors. The vibrations this causes drastically reduces the operating life of an impeller or turbine. Zangeneh gives the example of the pumps used to supply cooling water to British nuclear power stations which must be replaced every few months because of the vibrations they experience every time they are switched on.
But designing turbomachines that are not bedevilled by these problems has proved well-nigh impossible until now. The problem is that tiny modifications to the shape of a blade can lead to enormous changes in its performance. Before the advent of computers, deciding what modifications to make was almost a matter of guesswork. This meant that the design of turbomachinery proceeded by trial and error, with each new shape being painstakingly built and tested. Several engineers could work for months to perfect a single design.
Over many years, individual designers built up experience and developed a feel for how to solve certain problems. In the case of radial turbomachines they discovered that most could be tackled by changing the tip of each blade while leaving the overall shape unchanged. The result is that most blades for a given application are essentially the same shape. 鈥淚t was almost a black art and some designers were very secretive about their work,鈥� says Zangeneh.
Fiendish flows
Calculating the flow over the surface of a blade is fiendishly difficult. The flow is governed by up to six mathematical equations known as partial differential equations and these must all be solved to work out the flow at a given point at any instant. The result of this calculation can then be used to work out the flow at the same point at the next instant and so on, a process known as iteration. By repeating this series of calculations for many points on the surface, engineers can build up a picture of the flow and how it changes.
This task is ideally suited to a computer and about twenty years ago the science of computational fluid dynamics (CFD) began to have a big influence on the design of turbomachinery. 鈥淭oday the best workstations can simulate the flow at 40 000 points on the blade surface in around five hours,鈥� says Zangeneh. They test the effect of the smallest tweak in a huge range of flows without having to build a prototype. And they do it in a fraction of the time that a human could manage. 鈥淐FD has been very successful. Most of the very best designs are based on this process,鈥� he points out.
But although CFD is a powerful tool for testing design changes, it tells engineers nothing about what modifications to make in the first place. Even with CFD, turbomachinery design is almost as much of a black art as it was twenty years ago and most engineers still only vary the shape of the blade tip in their designs.
His approach is different. It relies on an idea first proposed around forty years ago by James Lighthill, a mathematician at UCL. Instead of modelling the flow around either side of a wing or blade, Lighthill suggested working backwards by specifying the velocity of the flow and calculating the aerofoil shape that will produce it.
But this type of 鈥渋nverse鈥� design is not as easy as it sounds. Lighthill鈥檚 approach works well for simple surfaces such as some aircraft wings but there are severe shortcomings when this approach is applied to the more complex flows in turbines and pumps. Many attempts produce impossible designs where the top and bottom surfaces of the blade didn鈥檛 meet up. Since then this 鈥渃losure鈥� problem has been the major difficulty.
The problem is that the blade shape and the flow are so closely linked that specifying one automatically limits the other. A computer which has been given the flow on either side of the blade can work out the blade shape but is powerless to change anything if this design is impractical. In the late 1970s, William Hawthorne, an engineer at the University of Cambridge, came up with a trick that could get around this. Instead of maintaining a specific flow on either side of the blade, Hawthorne鈥檚 idea was to specify only that the difference between the flow velocities remains the same. In this way, he could vary the flow until he came up with a design which is possible to create in the real world. Zangeneh, who worked with Hawthorne in the early 1980s, has developed these ideas into computer programs that design the blades. 鈥淚 never have a problem with blade closure. My blades always have leading and trailing edges,鈥� he says.
Unhindered by the hit-and-miss techniques that have governed turbomachine design in the past, Zangeneh鈥檚 blades are shaped like no others. They are also better. With the secondary flow under control, the blades can operate over a wider range of speeds and flow rates without fear of stalling or surge. As a result, they will last longer and cost less to maintain.
They are also vastly more efficient. Zangeneh says that preliminary studies show that his radial pumps may be up to 10 per cent more efficient than traditional models. 鈥淎 couple of per cent would be an incredible breakthrough with other methods,鈥� he adds. With the increased efficiency, it may even be possible to build smaller gas turbine engines that produce the same power.
Zangeneh believes his ideas will revolutionise manufacturing techniques. Because inverse design relies less heavily on iterations, it drastically reduces the development time of new designs. 鈥淲e鈥檙e talking about days rather than months for one engineer to optimise a design,鈥� he explains.
Ebara has glimpsed the potential. Zangeneh is reluctant to give exact figures but admits that Ebara has invested millions on developing prototypes at its factory in Fujisawa. Ebara hopes to change not only the way turbomachines are designed but also the way they are built. Another advantage of Zangeneh鈥檚 method is that his program produces a set of coordinates that exactly specifies the shape of the blade. This can be fed directly into a computer-controlled cutting machine that could carve out the design from a single block of metal 鈥� the ultimate purpose-built designer turbine. 鈥淔or big blades this may turn out to be cheaper than casting,鈥� he says.
And what of the future? For the moment, Zangeneh鈥檚 techniques work only for the radial impellers used in water pumps and small gas turbines. But he is already applying his ideas to the axial blades used in aero engines. He believes better blades and more efficient engines could translate into fuel savings of up to 5 per cent for airlines. 鈥淲e know how to do it but it鈥檒l be three years before we perfect the technique,鈥� he predicts.
