Green light for aluminium gas tanks
TWI Bulletin, November/December 1994
Graham Slater gained a degree in Engineering from Cambridge University in 1979, during which time he was employed as an Engineering Student Apprentice on a 'thick sandwich' course with the Atomic Weapons Research Establishment at Aldermaston.
Following graduation, he joined the Fatigue Section of The Welding Institute as a Research Engineer. In 1981 he moved to the Design Advisory Services Section and became involved in major research projects including welded repair of steel bridges and fracture toughness of offshore pipeline girth welds.
Graham is currently a Principal Engineer in the Structural Integrity Department. Areas of consultancy work with which he is concerned include: design and experimental stress analysis of welded structures; fatigue and fracture resistance; failure investigation and the application of codes and standards.
'Flat' tanks made from multi-cell aluminium extrusions provide a novel alternative to conventional cylinders for fuel storage in natural gas powered vehicles. Graham Slater describes a recent project for British Gas.
Natural gas is finding increasing popularity worldwide as an alternative fuel for motor vehicles. For many countries, a natural gas vehicles (NGV) programme is an important part of their national energy strategy, aimed at reducing dependence on imported oil supplies and overcoming environmental pollution problems. Conventional NGVs run on compressed natural gas (CNG) stored in cylinders at a pressure of around 200bar, and one drawback of such systems is the large amount of useful space occupied by the cylinders, especially in smaller vehicles.
An alternative storage mechanism for natural gas is by absorption within a microporous carbon matrix. Adsorbed natural gas (ANG) offers several key advantages over CNG:
- ANG provides two thirds the storage capacity of CNG at only a sixth the storage pressure ( Fig.1);
- the much lower storage pressure of ANG reduces the capital costs of compression;
- as a result of lower energy usage for compression, CO 2 emissions are reduced, lessening global warming;
- the reduced storage pressure improves perceived safety, and provides the opportunity to design and build specially shaped non-cylindrical storage tanks giving greater packaging flexibility to on-board storage systems.
Fig. 1. Comparison of pressure/storage characteristics of ANG and CNG
British Gas approached TWI for assistance in design and manufacture of experimental tanks that were rectangular rather than cylindrical in section - a 'briefcase' size and shape was considered the ideal objective.
Development history
The most important requirements specified at the start of the programme were as follows:
| Maximum working pressure: | 41 bar |
| Test pressure: | 69bar |
| Specific weight: | <1 kg/dm 3
(without carbon) |
| Temperature range: | -30 to +130°C |
| Operating cycles |
|
| Pressure range: | 0-41 bar |
| Design life: | 2500 cycles |
| Test life: | 40 000 cycles |
Initially, a conceptual design of tank as shown in Fig.2 was considered. It consisted of four sheet metal pressings, two forming the outer skins and two forming the corrugated cores. The pressings were to be welded together using laser stake welding to provide load-carrying internal joints.
Fig. 2. Conceptual laser welded tank
Finite element analysis of the design showed that it required 8mm thick outer walls, see Table, to prevent bulging, resulting in an unacceptably high specific weight of 2.5 kg/dm 3 (2.5kg of steel per dm 3 of internal volume). Substitution of initially curved outer skins, see Fig.3, produced a more efficient geometry, this time requiring an outer skin thickness of only 1.6mm to withstand the test pressure, see Table. This resulted in a specific weight of 0.87 kg/dm 3. However, fatigue tests on simulated longitudinal laser stake welded joints revealed an inadequate fatigue life.
Stress analysis results for three tank types, all subject to 69bar internal test pressure (results are transverse stresses remote from the ends)
| Tank type | Thickness, mm | Skin stress, N/mm 2 | Core stress, N/mm 2 | Specific weight, excluding ends, kg/dm 3 |
| Skin | Core | Bending | Membrane |
| Steel, flat skin | 8.0 | 1.6 | 180 | 21 | 170 | 2.47 |
| Steel, curved skin | 1.6 | 1.6 | -15 | 170 | 150 | 0.87 |
| Aluminium extrusion | 2.0 | 2.8 | 0 | 100 | 100 | 0.42 |
Fig. 3. Conceptual tank body with curved outer skin
The problem of the longitudinal joints led to the idea of forming the complete body of the tank from a single extrusion in aluminium alloy. The first step was to optimise the shape of the tank cross-section to make it as simple and efficient as possible, whilst still retaining the multi-cell features.
The proposed shape is shown in Fig.4. The geometry was crucial to the stress efficiency and was designed so that bending stresses were eliminated and membrane stresses were uniform throughout. This was subsequently confirmed by finite element analysis (see Table). This design was much lighter at 0.42 kg/dm 3.
Fig. 4. Proposed 20-cell extruded aluminium tank (dimensions in mm)
The novelty of the extruded tank design presented two major technical challenges:
- could a satisfactory multi-cell extrusion be produced?
- could a satisfactory method of closing the ends be found?
Extrusion manufacture
Although strength, ductility, weldability and corrosion resistance were all factors relevant to alloy selection, the prime requirement was reliable extrudability.
Maintenance of wall thickness tolerances and extrusion seam weld integrity were essential features, and in this respect the 6000 series alloys were most favoured, having a long history of successful use for complex shapes. Thus, the final choice was alloy 6063A T6 on the basis of best extrudability.
To reduce development costs, a smaller ten-cell extrusion was commissioned. This maintained all the major dimensions of the 20-cell version, apart from overall width. Despite the complexity of the design, extrusion was successfully achieved by Alcan Speciality Extrusions. A section of this ten-cell extrusion is shown in Fig.5.
Fig. 5. Section of ten-cell extrusion
Design of end closures
The simplest form of end closure for the multi-cell extrusion is a flat plate: this could be either a single 'set-on' plate covering the whole end of the extrusion, or individual 'set-in' plates inserted into each cell. Three methods of attaching the end closures to the body were considered, namely: welding, mechanical fastening, and adhesive bonding. However, the main thrust of the development programme was directed towards the welding options. To minimise the end cap thickness it was necessary to provide attachment to the extrusion body at both the external and internal walls. Four basic types of welded end closure were considered initially, and these are shown in Fig.6. Type (i) is a single set-on end cap covering all the cells, types (ii) to (iv) are variations of an individual set-in end cap design.
Fig. 6. Alternative end closure details
Welding procedure trials
Welding procedures were developed for four processes: MIG, TIG, electron beam (EB) and laser. Each process was evaluated for two of the four end closure designs as identified in
Fig.6. For each type of end closure, two distinct joint types required investigation: the end cap to outer wall joint and the end cap to internal wall joint. Trials were carried out using tube and plate assemblies in the relevant grade of alloy.
Arc welding processes were already fairly well established and needed relatively little development. Both MIG and TIG welding were generally successful, notably with a 2-3mm protrusion for the set-in (proud) type (iii) geometry. Power beam welding processes were initially less well established for the specific joint geometries under consideration, and required greater development: this was limited by time and resources available and was potentially capable of further refinement and improvement.
Electron beam welding was found to be capable of producing satisfactory joints, but was less tolerant of dimensional variations.
In particular, accurate joint fit-up (less than 0.25mm gap) and beam alignment were required. Laser welding produced similar results to EB welding, but was generally less successful, with problems in particular of weld metal cracking.
Preliminary tests using cylindrical specimens
A pilot test programme was undertaken to evaluate the potential performance of the end closure joint designs using simple cylinders of the same thickness and radius as the cellular tanks. Sixteen test cylinders were manufactured using extruded cylinder and bar in 6063A alloy and three welding procedures (laser welding was rejected in preference to EB welding). The bodies were 100mm long, 2mm wall thickness and 58.6mm outside diameter. The end caps were all 8mm thick. The types of cylinder are shown in Fig.7.
Fig. 7. Geometries of cylindrical test specimens (not to scale)
Each test cylinder was first subjected to a proof pressure hydrotest to 69bar. The cylinders were then either pressurised to failure or pressure fatigue tested. The pressure fatigue cycle applied was 1.4-42.8bar (41.4bar range) at a frequency of 1.5Hz.
The results showed that all combinations of process, joint type and material produced end closures capable of withstanding 69bar proof pressure.
Burst tests generated maximum pressures in the range 99-165bar, all failures occurring in the weld throat or cylinder wall. Fatigue lives were in the range 64 800-3 363 000 cycles, all well over the design requirement of 40 000 cycles. Fatigue failures occurred either in the weld throat or in the cylinder wall, depending on joint geometries. A selection of tested cylinders is shown in Fig.8.
Fig. 8. Cylindrical test specimens
Manufacture and testing of ten-cell tanks
Following the work on end closure design and welding, it was decided to focus on two types of end closure for the ten-cell tanks: type (ii), flush fitted, end caps welded by EB, and type (iii) set-in (proud) end caps welded by TIG.
The tank bodies were cut and faced off to 200mm lengths. Ports were machined in nine of the inner walls to provide gas passageways between all cells. Two shapes of end cap were made to fit the corner and central cell shapes: these were produced by machining extruded round bar to the appropriate dimensions. The end caps for one end of each tank were drilled for pressure fittings to be attached. The first end closures welded by the EB process were subsequently discovered to contain solidification cracks in the welds, and because of time limitations it was decided not to pursue this process further in this programme, and to concentrate on TIG welding instead.
Solid 'immobilised' carbon blocks machined to the cell profiles were inserted in one tank prior to closure.
Four ten-cell tanks were manufactured from 6063A alloy using the TIG process and set-in (proud) joint design: three empty and one containing carbon inserts. A selection of tanks, end caps and carbon inserts is shown in Fig.9. The carbon-filled tank and one empty tank were set aside for future work. The remaining two tanks were subjected to fatigue and burst testing.
Fig. 9. Tanks, end caps and carbon inserts
The tank subjected to the burst test failed at an internal pressure of 119.5bar. Failure occurred in the weld throat at one of the corner cells, blowing out the end cap from that cell ( Fig.10). This result compared favourably with the corresponding cylindrical specimen test result (99bar), demonstrating the effectiveness of the increased weld throat size achieved in the tank welds.
Fig. 10. Ten-cell tank after burst test
The second tank was pressure fatigue tested in a similar manner to the cylindrical test specimens, using the same pressure range (1.4-42.8bar) and frequency. This specimen also failed in the weld throat at the outer wall on a corner cell. The number of cycles sustained was 522 130; much greater than that achieved by the corresponding cylindrical specimens, again demonstrating the expected improvement because of the increase in weld throat thickness achieved in the ten-cell tank.
Summary
The results of this project have been remarkably encouraging, considering the technical uncertainties at the outset. The project has demonstrated that:
- a ten-cell extrusion in 6063A T6 alloy can be manufactured satisfactorily with sound seam welds and acceptable dimensional tolerances;
- a ten-cell tank with TIG welded end closures can be successfully manufactured and will resist a pressure of 1.73 times the specified test pressure without leakage before bursting;
- a ten-cell tank with TIG welded end closures can survive approximately 13 times the required number of service pressure cycles.
Future work
The project demonstrated emphatically that the concept of a cellular extruded aluminium tank with welded end closures can be realised in practice with the required levels of pressure containment and fatigue performance. The objectives set out at the start of the project were met. The successful outcome paves the way for further work to refine and improve the design and performance of the multi-cell tank.
Adequate burst strength was readily achievable. Fatigue performance was more than adequate to cope with pressure cycling alone, and probably sufficient to accommodate additional cyclic loading inputs such as thermal cycling and vibration, although this still needs to be demonstrated. Other aspects such as corrosion and impact resistance also need to be evaluated in future designs.
Choice of alloy is limited by extrudability. It has been demonstrated that 6063A alloy can be extruded to the required shape and dimensions in ten-cell form, but further trials will be needed to evolve a die design for wider extrusions.
Inspection and testing of completed tanks in production will need particular attention. Two important aspects are the integrity of the extrusion seam welds and the end cap welds. These are not readily amenable to non-destructive inspection, so quality control and assurance will have to rely heavily on destructive testing of extrusion material prior to assembly and leak and proof testing of completed tanks.
Detail design and manufacture of tanks to satisfy certifying authorities will be a major task, because there is little precedent from which to work. The major barrier to introducing such novel tanks in road vehicles is the lack of specific standards and specifications covering this type of design for ANG containers. Existing codes and regulations for gas storage and transportation are not applicable to multi-cell non-cylindrical designs, and this is an aspect currently under consideration.
Acknowledgements
The author is grateful to British Gas plc for permission to produce this article, and to Dr M H Ogle, Principal Design Consultant at TWI, who put forward the proposal to use aluminium extrusions for this application and the design of the prototype tank.
