Experimental Planning – in the context of Taguchi

I remember when I did my undergraduate research project. I often rushed to carry out the experiments without thorough considerations beforehand. The end result is I realise there is a flaw in my experimental design, and make my result difficult to interpret or impossible to draw a valid conclusion. The consequence is plenty of re-do's and re-work which were both time consuming and expensive.

Today (Wednesday) I continued working on Taguchi experiment. I find the initial phase of research and experimental planning to be very time consuming. I needed to consider many aspects of experiment (selecting arrarys, control factors, assigning factors) all before I even start folding the planes. This is to ensure I do not make costly mistakes later on in the conducting experiment. This is the same idea as DFSS, where we want to make all the changes as early as possible to prevent costly changes later on. One book I frequently refer to for this PMA is Taguchi - Hands on approach (Peace 1997) is quite useful in this respect because it takes me through a step by step process, emphasising pitfalls and common mistakes, ensure all necessary precautions are taken to avoid making mistakes.

But the truth is even if I gave my best shot at paying careful attention to the experimental design, there always seems to some slip-ups somewhere in your experiment which you realise much later. For example, after I determined my design variable, and started folding planes. I forget to consider the effect of combinantorial (?) layout of orthorgonal arrays, which means some of the control factors are mutually incompatible with each other. Specifically, I couldnt have  planes with a 6 inch wing span AND a 1.5 inch high wing tip both at the same time (due to paper size constraint)! This means I had two planes that turned out to be impossible to build according to my design specifications. It's too late to change things now because that would take too much effort. I will simply point that out in my pma reflection.

I guess the important lesson to take away from this, is the importance of spending time and effort on the initial experimental planning. Making sure you have considered everything that can go wrong in your experiment and take steps to mitigate the effects in case they happen. In most PMA's this is not so important because we can easily make changes to our essays on the computer. But for the project, especially if it involve surveys or experiment. It is handy to study books on experimental designs , many of which are mentioned in the REME module.

: 12 Mar 2009 05:11 |  Tags: Mbe Piuss |  Comments (1) |  Report a problem