4.4. Comparison between predictions and experiments
The stress measured at different speeds in the above experiments was substituted into the matlab fitting program, and the fitting curve was expressed as Eq.(17) based on Eq.(1). The fatigue life of fan blades under different working conditions was calculated (Fig. 9) as listed in Table 5. Substitute the number of cycles measured in the above test into Eq.(14).The cumulative fatigue life was shown in Table 6.The predicted fatigue life was compared with the experimental data, the relative errors between the predicted results and the experimental data were shown in Table 5 and Table 6. It could be seen that the relative deviations between theoretical prediction and experimental results with revolving speeds of 1600, 1800, 2000 and 2200r/min were about 10.0%, 2.5%, 3.9% and 11.1% respectively, and the error between cumulative fatigue life test and Miner’s theoretical calculation was 10.8% with acceptable dispersion. Therefore, the improved method based on the concept of nominal stress could effectively predict of the fatigue lifetime of complex fan blades.
In addition, it is well known that fatigue life outcomes tend to be much dispersed [45]. Therefore, in general, the general rule and analysis results of high reliability can be obtained through repeated large sample experiments. Under the circumstances, resources often limit the number of experiments. When more specimens and more stress levels are noticed for fatigue tests, more accurate fatigue properties would be determined which makes results more accurate.
It was clear that through modified method (6) and (7), fatigue lifetime of fan blades subjected to different rotating speed conditions could conveniently and easily be predicted by the minimum data set of a small number of experimental fans without any additional experimental studies. In short, the fatigue life of fan blades under different working conditions could be predicted only by obtaining the maximum stress of fan blades’ model and S-N curve of materials. The correlation between prediction results and actual experimental results was relatively good.
Conclusions
A method of predicting fatigue life based on the S-N curve and mean stress was proposed in this study, and a special application dedicated to the ventilation cooling system of high-speed-train was investigated. The effectiveness of the prediction model was confirmed through a comparison with fatigue bench test system. The primary conclusions were summarized as follows:
  1. The maximum stress of the fan was calculated by using the finite element method, and the stress concentration was located at the root of the blade.
  2. The fatigue prediction model was based on mean stress and also took several specificities of the fan structure (SCF, surface quality coefficient…) into consideration. By means of the S-Ncurve of the material, the fan’s life was predicted by using this new prediction model according to equivalent life curve and cumulative damage criterion.
  3. The fan fatigue test system with adjustable frequency was built. Experimental results showed that the method was successful, in which reasonable correlation was achieved between predictions and actual experiments. To reduce stress concentration and improve service life, making an arc or a stress groove at the blade root was recommended.