This work was carried out as part of a joint project with Allevard Springs Limited and the School of Metallurgy and Materials at the University of Birmingham.
In the UK, automotive suspension coil springs, which are highly stressed in operation, are made from silicon - manganese spring steel (BS 970 251A58), of which the main constituents are: 0.55 - 0.6% C, 1.8 - 2.1% Si and 0.8 - 1.0 Mn. Typical processing gives a nominal hardness of 49.5 Rc, equivalent to a UTS of 1600 to 1700 MPa. These springs are designed for a maximum stress under full bump conditions of 1100 MPa.
Useful weight savings could be made if the springs could be designed to operate under a full bump stress of 1300 MPa - but this would only be acceptable to the motor industry if there was no price increase. The cost of SiMn spring steel was approximately £320/tonne. Manufacturers in most other countries have moved away from SiMn steels and use a low alloy steel - the Japanese use a 0.45 % C, 1.5 % Ni steel with a UTS of about 1800 MPa. Alloying of this type would involve an increase in costs, probably to about £400/tonne, which was thought to be unacceptable. As most alternatives to SiMn steels would involve significant increases in costs, it was thought important to fully explore the capabilities of SiMn steels and minor variants.
Because fatigue testing complete springs in a hydraulic rig is time consuming owing to the long stroke required, the testing of single 'turns' cut from complete springs was assessed. This was found to be satisfactory when the stress distribution in such a configuration is considered even though it is different to that in a concentrically loaded coil spring.
Tests were carried out on single turns cut from coils and compared with the results obtained from tests on complete springs carried out at SRAMA (Spring Research and Manufacturers Association, Sheffield, UK). Initially the test programme contained two materials: a conventional SiMn spring steel and a special SiMn+V steel, each tested at 3 draw ratios. The results of the tests on the single turns indicated that the fatigue lives of the single turns from the SiMn+V steel were longer than those in the conventional SiMn steel. This trend in the results was also apparent from the tests on complete springs carried out at SRAMA.
Because the draw ratio did not seem to have a systematic or significant effect, the fatigue life results were brought together into four groups: SiMn and SiMn+V tested as complete springs and as single turns and plotted on Weibull graphs. The parameters of the distributions are summarised below:
| SiMn 1 turn | SiMn+V 1 turn | SiMn Spring | SiMn+V Spring | |
| eta | 117,155 | 140,309 | 144,409 | 166,337 |
| beta | 7.48 | 7.31 | 2.68 | 3.25 |
| R2 | 0.96 | 0.89 | 0.96 | 0.98 |
Although the load conditions of the single turn tests had been set to provide the same stress range (700 MPa) and average stress (750 MPa) as the complete coils (corresponding to 725 MPa PSWT), the lives are shorter. More significantly, the different beta values indicate significant differences in the failure mechanisms with the two different test methodologies. The consistency of the beta values for the single turn results does indicate that results from such tests can be used to estimate the performance of actual springs.
In order to carry out an assessment of defects in samples an automated method of counting and classifying defects using a vision system with appropriate software was developed. In high strength steels fatigue properties are related to the size of the largest inclusions present on a plane perpendicular to the direction of maximum principal stress. In concentrically loaded coil springs the maximum principal stress is on an axis inclined at 45o to the axis of the wire. It was observed that stringers of soft inclusions are normally significantly smaller than hard globular inclusions in this plane and can therefore be ignored when considering fatigue strength.
Before the automated analysis technique was adopted, checks were made to ensure that the parameters set for automated counting gave results which were very similar to those obtained by manual methods on the same samples. While agreement between the results of manual and automated counting of globular inclusions was reasonably good, the agreement between the two types of measurements for stringer inclusions was not so consistent. The reasons for this discrepancy are thought to be a combination of manual measurements overestimating the size of stringer inclusions and the fact that the software was deliberately set to ignore small inclusions and the stringers tended to have much smaller cross sections. It is believed that the hard globular inclusions are the critical ones, this discrepancy is not a matter of concern in this project.
One observation from this work was that considerable variation existed in the numbers of globular inclusions present, even in fields that were close together. Even after totalling up the areas from 40 fields taken while traversing 10 mm over a cross section, the largest total area for 40 fields was still 2 to 3 times greater than the smallest total area for 40 fields from the same material. For the 3 draw ratios of SiMn steel (incoming rod sizes 12.5 mm, 13.5 mm and 14 mm) the averaged % area fractions of globular inclusions were 0.035, 0.0107 and 0.68 % respectively with mean sizes of 39, 105 and 67 x 10-12 m2 respectively. The equivalent figures for the SiMn+V steel were 0.039, 0.08 ad 0.085% with mean sizes of 52, 85 and 97 x 10-12 m2 .
SiMn steels from two suppliers, which normally contained different inclusion levels, were investigated. The clean material was produced from a route involving a high scrap recycling content, melted in an electric arc furnace. The less clean material was obtained from a conventional steel supplier.
One way of increasing the strength of springs is to temper at a lower temperature, resulting in a higher tensile strength. It was not known if this would also result in better fatigue lives or, if the Charpy impact values would be reduced, so a series of experiments were carried out on single turns cut from complete springs that had been processed to give nominal hardnesses of 50, 52 and 54 Rc. These tests included fatigue testing and Charpy impact tests. The samples tested, which were cut from complete springs, had Rc values about 1 point below the nominal values.
| Hardness (Rc) | 49 | 51 | 53 |
| Batch 1 (J) | 103 | 113 | 110 |
| Batch 2 (J) | 102 | 112 | 120 |
The Charpy impact test results showed that steel hardened to 51 Rc gave higher Charpy energies than the 49 Rc material. The results for the samples cut from the 53 Rc springs were similar to those from the 51 Rc steel.
Because of the variation in fatigue life results of single coils and complete springs, and the desirability of having a clearly defined volume of material subject to the specified stress levels, it was decided that it would be preferable to test Wohler type testpieces in rotating bending rather than single turns from actual springs. All but one of the specimens in the less clean steel when tested failed at a visible non - metallic inclusion which was either surface breaking or very close to the surface. The size of the inclusions was between 12 and 30 x 10-6 m2. In the majority of cases the direction specimen fracture was within + or - 45o of the angular position where the stress was a maximum.
Of the clean steel specimens, it was only possible to see a defect in one of the fracture surfaces, all of the other failures were 'self initiated'.
Even in the clean steel there were a few significant inclusions, one 25 x 10-6 m2 inclusion being present on one fracture surface, this particular specimen having a life of only 10000 cycles, whereas other clean specimens tested at the same stress level typically had lives around 38000 cycles.
The investigators are grateful to the EPSRC for funding and supporting this project.