IAC-08-C4.5.2 HEAT EXCHANGER DEVELOPMENT AT REACTION ENGINES LTD

IAC-08-C4.5.2 HEAT EXCHANGER DEVELOPMENT AT REACTION ENGINES LTD Richard Varvill Reaction Engines Ltd, United Kingdom richard.varvill@reactionengines.co.uk ABSTRACT The SABRE engine for SKYLON has a sophisticated thermodynamic cycle with heat transfer between the fluid streams. The intake airflow is cooled in an efficient counterflow precooler, consisting of many thousand small bore thin wall tubes. Precooler manufacturing technology has been under investigation at REL for a number of years with the result that flightweight matrix modules can now be produced. A major difficulty with cooling the airflow to sub-zero temperatures at low altitude is the problem of frost formation. Frost control technology has been developed which enables steady state operation. The helium loop requires a top cycle heat exchanger (HX3) to deliver a constant inlet temperature to the main turbine. This is constructed in silicon carbide and the feasibility of manufacturing various matrix geometries have been investigated along with suitable joining techniques. A demonstration precooler will be run in front of a Viper jet engine at REL s B9 test facility in 2011. This precooler will incorporate full frost control and be built from full size SABRE engine modules. The facility will incorporate a high pressure helium loop that rejects the absorbed heat to a bath of liquid nitrogen. INTRODUCTION The SABRE engine designed for the SKYLON SSTO spaceplane [1] and the SCIMITAR Mach 5 cruise engine both employ liquid hydrogen fuel and similar thermodynamic cycles. These engines are able to achieve high speed airbreathing flight by using the hydrogen fuel as a heat sink to lower the temperature of the decelerated inlet air so that it can be compressed and managed through combustion by relatively conventional turbo-machinery. Rather than employ the hydrogen to cool the airflow directly, the engines use the hydrogen as a heat sink for a closed helium power loop inserted between the hot airflow and the cold hydrogen. While the majority of the engine components are relatively conventional, lightweight high power heat exchangers are a new feature unique to this type of engine and pose a challenging manufacturing problem. To address concerns over the novel nature of the technology the recent experimental programme at Reaction Engines Ltd. has been directed at resolving those areas of difficulty relating to light weight compact heat exchangers operating in a frosting environment The helium temperature emerging from the precooler varies reflecting the changing air enthalpy with flight Mach number. Accordingly a top-up heat exchanger (HX3) employing heat drawn from a pre-burner maintains constant helium temperature for expansion in the main drive turbines. This high temperature heat exchanger is designed in silicon carbide and is subject to ongoing manufacturing research. PRECOOLER RESEARCH Precooler design The precooler is an efficient counterflow design, consisting of many thousand small bore thin wall tubes [2]. The cold helium flows inside the tubes whilst the hot air is arranged in external crossflow. Due to the high pressure difference between the two fluid streams plain tubes give the lightest matrix design. To maintain acceptable airside pressure drop the matrix has a 1

large frontal area but is relatively thin. Therefore to prevent excessive engine nacelle wave drag the matrix is wrapped into a cylindrical drum with radial inward airflow. To ease manufacturing and assembly difficulties the precooler is split into a large number of identical modules. A cross section through a fully assembled precooler is shown in Figure 1. The tube diameter and wall thickness is similar for the SABRE and SCIMITAR engines (ie: 0.88mm bore with 0.040mm wall thickness). Helium outlet headers Helium flow along tube spiral Helium inlet headers Radial inward airflow the LAPCAT project partly supported by EU funding. REL has been developing the technology for building precoolers in Inconel 718 due to its combination of high mechanical strength at temperature and oxidation resistance. Although this material is considered optimum for the SABRE where the time spent at high temperature is short, its elevated temperature creep strength is rather low for the hot end of the Scimitar precooler. Various alternative alloys have been explored as part of LAPCAT however none have been found that combine better creep strength with the necessary ductility for tube drawing whilst retaining high temperature oxidation resistance. Tube drawing technology has been successfully developed to produce 1mm diameter tubes with a wall thickness down to 30 microns in Inconel 718 (Figure 2). These tubes have been repeatedly pressure tested to 600 bar before rupture. Matrix spiral Figure 1 Precooler Cross section This type of precooler construction is extremely lightweight whilst achieving high thermodynamic efficiency. For example the SCIMITAR precooler has a predicted mass of 940 kg. At Mach 5 cruise conditions the precooler handles 172 kg/s of air at a recovered temperature of 1250 K that it cools to 665 K whilst incurring an airside pressure drop of 0.4 bar. This gives an installed power/weight ratio of about 110 kw/kg. Precooler manufacture Construction of flightweight precoolers requires technical advances in many areas such as tube drawing, cutting and forming, brazing, hole drilling, plating and assembly tooling. Reaction Engines has been developing this technology for the SABRE precooler for 7 years with private funding. Recently this work has been applied to Figure 2 Micrograph of a 30 micron Inconel 718 tube wall Following drawing the tubes are brazed into headers to distribute the coolant through the matrix (Figure 3). Conventional brazing techniques for thick wall sections are inappropriate due to excessive braze erosion. Consequently new braze methods are being developed to reliably achieve leak tight strong joints without impairing the tube material properties. Following this initial program of work a test heat exchanger was successfully manufactured employing the techniques 2

described above for testing in the frost control wind tunnel. Figure 4 Tube bending specimens before and after pressure testing Figure 3 Close up of brazing fillets A major technical issue is the manipulation of tubing with very small bore and walls only a few tens of microns thick, since commercial tube benders capable of handling this tube are unavailable. However great progress has been made with equipment developed within Reaction Engines so that the tubing can now be bent without buckling the walls or causing excessive thinning of the wall material. Pressure testing to destruction of many trial bends has shown that failures do not occur in the bends. The problem of cutting the tubes accurately with minimal burring has also been solved and several means are now available to do this in production. Tube cutters which trim the tubes to the correct length following bending and pressure testing are incorporated into the prototype tube manipulation machine. Figure 4 shows trial bends before and after pressure testing. To further investigate the assembly and manufacturing issues entailed in the construction of a real precooler, a Viper engine demonstration module was constructed in 2005. The module contains 460 tubes 2.2m long arranged in 4 columns and 115 rows. This entailed the drawing of 1180 m of tubing about 0.98mm outside diameter and 40 microns wall thickness. During this production run various difficulties arose whose solution has led to greater understanding of the issues involved in producing these exceptionally fine tubes in a nickel base superalloy. The module was brazed according to a time/temperature cycle devised by Reaction Engines. Figure 5 shows the completed module after disassembly of the braze jig. Figure 5 Full scale VIPER precooler module 3

Further practical work within the LAPCAT project resulted in the construction of two full scale SCIMITAR engine modules (Figure 6). These modules were successfully pressure tested to prove their mechanical integrity but suffered from a small number of leaks in the braze joints. Since the brazing technology is already proven at laboratory scale these defects were due to transferring to industrial process conditions. Subsequent investigation has identified the causes and current experimental trials are expected to eliminate the problem. work is now at an advanced stage with several hundred cryogenic test runs. Extended precooler operation is now possible at tunnel air temperatures down to 193 K ( 80 o C). Tunnel temperature is a well controlled variable, and steady state operation at higher temperatures is, of course, possible. Figure 7 Frost control wind tunnel Figure 6 Full scale SCIMITAR modules Precooler frost control Reaction Engines Ltd has initiated an experimental program to address those areas of difficulty relating to light weight compact heat exchangers operating in a frosting environment. The water vapour in the atmosphere up to an altitude of around 12 km is a problem for precooled engines, causing them to block with frost in a matter of seconds. A great deal of this water is precipitated in the liquid phase during the cooling of the air and has to be rejected from the engine before it can freeze. Provision then has to be made to stop the build up of ice within the matrix as it precipitates directly from the vapour. A major part of the experimental program has been to demonstrate that this can be achieved. Reaction Engines operates a frost control wind tunnel at its site in Oxfordshire (Figure 7) to develop efficient frost control systems. This Figure 8 shows steady state operation in wind tunnel tests, with air inlet temperatures of 80 o C, 100% relative humidity and constant pressure drop across the heat exchanger matrix, for a test duration of 8 minutes, limited by wind tunnel consumables. This compares with the requirement for the system to operate for some 4 minutes during an actual ascent to orbit. Figure 8 Wind tunnel operation tunnel temperature and heat exchanger pressure drop. Figure 9 shows tunnel operation again at about 80 o C, but now the heat exchanger temperature is being varied. The heat exchanger coolant temperature changes from 50 to 120 o C over a period of 8 minutes, with the tunnel air inlet temperature being maintained constant for this time. 4

the plates (Figure 10). This approach enables a much higher airside massflux since flow separations are largely eliminated. Figure 9 Wind tunnel operation tunnel and heat exchanger temperature variation. The use of ground based wind tunnels, the control of tunnel and heat exchanger conditions and the wide range over which parameters can be changed lead to a rapid and cost effective means of carrying out technology development and characterisation. HX3 can have large driving temperature differences since the choice of preburner temperature has no effect on the overall fuel flow. Since the preburner exhaust is oxygen rich it is not possible to manufacture HX3 in the refractory alloys and the matrix temperatures are too high for the Nimonic alloys. Consequently the baseline material for HX3 is silicon carbide since it has the highest thermal conductivity of the engineering ceramics. Since the helium pressure (up to 200bar) is much higher than the preburner pressure (around 5bar) the matrix plates have to withstand internal tensile stresses. Consequently it is planned to meet the life and strength requirement by a combination of pressure testing and low stress levels. Helium inlet manifolds Helium flow PRE-BURNER HEAT EXCHANGER HX3 HX3 design Both the SABRE and SCIMITAR engines require a heat exchanger to heat the helium emerging from the precooler to a constant temperature independent of vehicle Mach number before expansion in the main turbines. To achieve this HX3 is situated in the exhaust of a pre-burner. The SCIMITAR HX3 design point is currently at Mach 3.0 at an altitude of 18856m. The optimum matrix design depends on pressure drop constraints and stream pressures. For the SABRE engine where both the preburner exhaust and the helium are at high pressure a compact tubular matrix with crossflow gives acceptable pressure drops. However the SCIMITAR engine has a low pressure preburner and a design with external crossflow over a tube bank would necessitate a very large frontal area to minimise preburner-side pressure drop. Consequently the matrix design features a number of parallel plates arranged in an involute spiral pattern with axial preburner flow between Figure 10 Helium outlet manifolds Airflow SCIMITAR engine HX3 layout showing involute matrix spirals. HX3 manufacture research The production and joining technology of silicon carbide heat exchangers is not as well developed as the metallic pre-cooler technology. Consequently this activity has focussed on matrix (tube or plate) manufacture and joining technology. Four different potential production routes for silicon carbide components have been investigated. 5

Pressure-less sintering A sintering route is being explored with Saint- Gobain Advanced Ceramics for silicon carbide tubes. Hexoloy silicon carbide is produced by pressure-less sintering ultra-pure sub-micron powder, which is mixed with non-oxide sintering aids and is then formed into complex shapes and consolidated by sintering at temperatures of more than 2,000 C. The sintering process results in a pure alpha phase, fine-grain silicon carbide product with virtually no porosity. The material retains useful strength up to 1650 degc. During manufacture the material is first extruded in a green state in which condition it is readily machined. Saint-Gobain successfully manufactured tubes with an OD of 2.0mm and wall thicknesses of 0.50, 0.45 and 0.35mm. The tubes were 600mm long and generally of high quality. A number of the 0.35mm tubes were pressure tested to 700 bar. 2 tests failed when the end fittings blew off, 2 tubes failed at the end fittings (probably due to a stress raiser) and 2 tubes survived without failing. From these limited tests it is concluded that the material is of high quality with a hoop failure stress of at least 172 MPa. CVD onto a removable mandrel In this chemical vapour deposition process a gas mixture based on Ethyl Silane is flowed through a mandrel at high temperature. The silicon carbide is deposited on the surface in an autocatalytic reaction. This process is used in the electronics industry due to the ultra-pure silicon carbide which is deposited. In theory the material is 100% dense, non porous and can achieve mirror surface finishes. Due to the slow rate of deposition it is possible to manufacture very thin wall silicon carbide components by this method. In order to produce a structure of this type out of pure SiC, a graphite component acted as a male mould onto which >1mm of CVD SiC could be deposited. The graphite could then be removed by machining or oxidation. Graphite conversion by silicon This method entails conversion of a graphite foil structure into SiC by liquid or vapour reaction with elemental silicon. Various types of graphite foil were tested. Some foils were single layer whilst others were multilayer structures, with one including a metallic mesh between the foil layers. In order to convert the graphite foil to SiC, two routes to siliconisation were tested a liquid and a vapour route. The liquid route is well known for production of ceramic materials by siliconisation of C/C composites, with the resultant material being used for applications such as ceramic brake discs for high performance cars. In the liquid route the silicon melts and capillary forces draw the liquid into the pores of the graphite requiring conversion. At the same time as being drawn into the material, the silicon reacts with the carbon to form SiC. The vapour route was attempted however in this case very little weight gain or conversion was observed and hence it was concluded that this route was not useful for producing the required SiC components. Reaction bonding of a green extrusion A reaction bonded process (REFEL) has been explored with Tenmat Ltd. In this process a SiC/graphite body is exposed to liquid silicon and the graphite converted to bonding SiC. The green material may be formed by a number of routes including extrusion, isostatic compaction and warm moulding. The final component has a fine homogeneous grain structure with grains of Alpha SiC bonded together by Beta SiC with some residual free silicon. For simplicity an extrusion die was manufactured with only 4 cooling channels. Following extrusion the green extrudate was air dried to give adequate handling strength before loading into the furnace for siliconising. Some of the finished parts are shown in Figure 11. For a first attempt the process has been remarkably successful with a relatively uniform cross section and small variation in wall thickness. Since the parts are produced by extrusion it is easy to vary the length and strips up to 600mm have been formed. Also curved strips were successfully manufactured by simply bending the flexible extrudate into a curve before drying. 6

These tests will take place at Reaction Engines new test facility at its site in Oxfordshire, shown in Figures 12 & 13 and known as the B9 Test Area. The facility is on two levels. The top level supports the fuel tank, a liquid nitrogen tank, pipework for these tanks and the control room. The lower level is where a Viper jet turbine is located, inside a protective concrete structure, together with a specially made silencer to muffle and cool the exhaust. Figure 11 Extruded and reaction bonded SiC heat exchanger strips A problem with the process is the formation of a number of silicon beads inside the channels. These are thought to form during furnace cooling since the free molten silicon in the matrix expands as it freezes forming beads on the surface. Beads also form on the outside of the strips but these can be relatively easily removed by abrasive blasting. For production a means of cleaning the channels will have to be found and future work should address this point. Nevertheless of the manufacturing routes investigated to date it appears that reaction bonding of silicon carbide/graphite extrudate has the most promise for a flat plate heat exchanger with a complex internal cross section. In particular this method is amenable to mass production and would be much cheaper than CVD. FUTURE EXPERIMENTAL PROGRAM Over the next three years Reaction Engines plans to build and test a 9% scale SABRE engine precooler. This precooler will be built of the same modules as SABRE but have only 9% as many. It will be operated at the same temperatures, pressures, mass fluxes and Reynolds numbers as full scale. Apart from verifying the heat transfer and frost control system the test will investigate mechanical integrity issues such as operation in close proximity to a high power axial flow compressor. Figure 12 Top level of the area, showing the butane and nitrogen tanks. The facility has been specifically designed to operate at cryogenic temperatures. In order for the engine to run under these conditions it is fuelled by liquid butane, rather than aviation fuels which become viscous at low temperatures. A large liquid nitrogen tank provides the heat sink for precooler operation. In Viper shakedown tests the liquid nitrogen is simply injected directly into the airstream to investigate compressor handling and potential low temperature embrittlement problems etc. For full precooler operation a high pressure helium loop will be installed into the facility. The cold helium will pass through the precooler cooling the airflow before rejecting the absorbed heat to a bath of liquid nitrogen. The engine is run with a constant fuel flow at a predetermined rate, the engine then moving to a new operating point corresponding to the reduced inlet temperature. 7

REFERENCES 1) R.Varvill and A.Bond The Skylon spaceplane, IAA 95-V3.07, 1995. 2) H.Webber and S.Feast, Heat Exchanger Design in Combined Cycle Engines, IAC-08- C4.5.1, 2008. Figure 13 Viper engine running on butane with water cooled exhaust exiting the silencer. CONCLUSIONS Reaction Engines has designed an efficient high power/weight ratio precooler that enables single stage to orbit with near term technology. The Company has been evolving the manufacturing and frost control techniques for this precooler for a number of years. This has now reached the stage where the Company is planning to build and test a 9% scale precooler which will be run in front of a Viper jet engine. The Company is also working on silicon carbide heat exchangers for high temperature applications. This technology is less advanced than the precooler however early findings suggest that it is possible to manufacture thin wall tubes by sintering and also to reaction bond flat plates with complex internal cross section. 8