Results comparison

Average increases in temperature indicate that the models temperatures are increasing at a higher rate than the actual weld trials. This is possibly due to the convection coefficients specified in the model as well the difference in rotation speeds specified in the model. The reason for the comparison of the different rotation speeds for the FEA with that to the weld trials is that the FEA model for those rotation speeds had more results to work with and compare then the previous models set out. One can notice though however, that the convection coefficient for the bottom of the work piece to the backing plate has an effect in the model. The higher this convection coefficient the faster the rate of temperature increase rate is. Also, therefore in the processing of the weld trials, the percentage of convection coefficients can be changed that hQ2 has a contributing factor than originally thought for the hq2 value. The convection coefficients for the clamped part of the tool and shoulder of the tool surface can be investigated further. After processing the results again to obtain better convection coefficients, one is able to make a better estimate of these values

Comparison of the heat flux transfer depicts higher values for the model than for the weld trials. It was mentioned that the convection coefficient for the bottom of the work piece should be changed and since the heat flux for the weld trials are much lower and hence their convection coefficients, it is suggested that a lower convection coefficient could be used for the bottom surface. Based on the bar graph for the coefficients, a relationship in terms of a percentage from view should be noted.

116  For the models, it is noted that the heat flux for q2 is much lower than that of Q2 (which these relate to the top and bottom surface respectively). It can then be seen that for the weld trials this is in the reverse (but this is also the result of the percentage contributions chosen) Since in literature the bottom surface has a higher heat transfer rate and it is a result observed closely above, it can be suggested that for the weld trials, these values be altered and that the model does indeed depict the correct relationship between these two values.

 For the clamped and shoulder part of the tool, the models depict that indeed the clamped part of the tool has a higher value than that of the shoulder which can be seen more clearly in comparison for the heat transfers for these two values for the weld trials. Therefore, it can be concluded that the clamped tool part should indeed have a higher value than the shoulder, which also corresponds to the values obtained for the weld trials.

 However, as mentioned, the FEA model rotation value is at a higher degree and so although the values do not compare exact in number, in terms of correlation in percentage wise to each other it is comparable. It should also then be noted that for higher rotations speeds the convection coefficients and hence heat flux are much a higher and for a future recommendation a relationship of the rotational speed and convection coefficients can be further investigated.

Temperature comparisons for the weld trials and model show some varying degree between the values obtained. Although the model temperatures are far higher than the weld trial temperatures, they do follow the same trend of a curve and then increasing. The models do exhibit a higher increase in temperature, but this is due to the parameters inserted into the model. It is difficult to assume from these temperature values that the rotation inserted has a great affect (which under normal FSW circumstances should exhibit). The only conclusion that can be drawn as to why there may be the difference in the trends and that is again the convection coefficient specified for the bottom surface of the work piece, since a higher one was specified for model 10 and this is clearly shown a higher temperature increase and temperature result.

To see whether any relationship exists between the weld trials and temperatures, although the temperatures do not coincide, a polynomial line is fitted to each of the data points. Each of the lines were fitted with a polynomial line to the power of 5. This fitted very well for the trends with the weld trials having data points lying almost equally in and outside the line, indicating some error in the experimental results as is expected, but only slightly. Only one data trend does not consist with this polynomial fitting as this gave a linear trend of temperature

117 increasing for model 8. This could be an inaccurate result from the model, since the other trends follow quite well with one another and a possible reason could be the iterative workings for that particular model run. It is apparent however that for the weld trials temperatures obtained, higher rotations, results in higher temperature results.

After changing the convection coefficients percentage for the weld trials, one is able to depict a more comparable relationship between the bottom and top surface coefficients when compared between the experimental weld trials and model results. However, as noted in the Results section, this value for the backing plate is highly unstable since it has been assumed that it takes on a convection coefficient for modelling purposes when in fact it is conduction. So although, this comparison can be made, it is an assumption for both to compare.

In conclusion to the model developed, it is apparent that there are similarities between the model and weld trials in terms of the way the model behaves and one can conclude that the model does indeed depict the heat generation, but care and more extensive data is required for the model to operate within the correct conditions in order to achieve exact results.