Factorial Design of Experiments and Interpretation of Results: An Application to Ice Loads on a Conical Structure

The data for this study is taken from a publication by D. S. Sodhi et al. [1]. The experimental setup in their publication was made in a way which, after some manipulation, allowed it to be treated as a full two-level factorial design with four factors. The effects of the factors; sheet ice thickness, flexural strength, ice-structure friction and relative ice structure velocity are studied. For the mean horizontal response force the conclusion is that there are clear effects for the changes in friction, ice thickness and the interaction of these two factors, the other effects are not readily distinguishable from the noise. For the vertical force the only really clear effect on the force is the ice thickness, and possibly there is an effect from the flexural strength. The only effect on the frequency of the response is the velocity. Studying the results from the analysis it should be kept in mind that the experiment was not originally set up as a factorial design and that the results can be misleading, and the manipulation of the data has lead to the introduction of more noise. There is for example no randomization of the trials to eliminate unmeasured effects on the outcome and this study should be seen as a demonstration of a powerful methodology rather than a strict experimental design. Keywords: Experiment, Ice Loads, Factorial Design