Dear Milind,

Can you explain your problem more elaborately?

Dear Sobhan, thanks for reply.

I am executing a 5-factor 54 run DOE based upon CCD, RSM and Desirability for RPD. I have a fundamental confusion.In my experimentation I am obtaining a reading of 30 minutes combustion process which is unsteady and data is not normally distributed (within the reading for 1800 reading points, sampled per second). Is averaging of these 1800 sample points fundamentally alright; over an individual reading to generate 54-single-redings for further analysis by ANOVA and fitting RSM and optimization by Desirability? Or any special method for such unsteady data analysis? Because each averaged reading for given factorial combination will have its own SD and Uncertainty. In all the books I read on DOE and RSM the SD and uncertainty is calculated for group of such 54 points (as if an individual reading was firmly read at that value only; without any uncertainty or SD for that particular reading).

Dear Milind P. Kshirsagar

For unsteady data, a better way is to carry out FFT analysis. However, it depends on your data or objective (which is unfortunately not clear to me from the discussion so far).

While averaging, make sure the averaged quantity is a true representative. You may consider averaging of data based on the physical observations rather than per minute. To elaborate this, let us consider any parameter "A" which repeats itself with a period "T" in an unsteady process. In this case your averaging time period should be

Dear Kuldeep, thanks for replying. Can you suggest some good books or research articles related to FTT?

Dear Milind!

You can go thru the link given below:

https://www.youtube.com/watch?v=3FAIXEkxyBs&t=686s

You have experimental data, you can rearrange in the correct format and can do FFT analysis in Excel/MAtlab/Tecplot etc.

My work might be useful for you. Although, it is not in your field. It can help you to present and interpret results after FFT analysis.