Minitab- Test Selection

#1
If in my project I have to manufacture roof shingles.
Following things of the roof shingles will be compared to the standards (being used in the roofing shingles)
1. Tensile Strength
2. Durability
3. Bond Strength
4. Water Absorption
Now I need to know what I should keep my design factor.
The samples will be
1. Fiber Glass+ Asphalt
2. Fiber Glass+ Asphalt+ Limestone
3. Fiber Glass+ Asphalt+ Rice Husk
4. Fiber Glass+ Asphalt+ Limestone+ Rice Husk
In these types of shingles I’m also varying the weight percent of materials.
Hence on weight percent basis we will have:
1. Fiber Glass+ Asphalt
2. Fiber Glass+ Asphalt+ Limestone
Fiber Glass+ Asphalt (50%) + Limestone (50%)
Fiber Glass+ Asphalt (70%) + Limestone (30%)
Fiber Glass+ Asphalt (90%) + Limestone (10%)
3. Fiber Glass+ Asphalt+ Rice Husk
Fiber Glass+ Asphalt (50%) + Rice Husk (50%)
Fiber Glass+ Asphalt (70%) + Rice Husk (30%)
Fiber Glass+ Asphalt (90%) + Rice Husk (10%)
4. Fiber Glass+ Asphalt+ Limestone+ Rice Husk
Fiber Glass+ Asphalt (50%) + Limestone (25%) + Rice Husk (25%)
Fiber Glass+ Asphalt (50%) + Limestone (10%) + Rice Husk (40%)
Fiber Glass+ Asphalt (50%) + Limestone (20%) + Rice Husk (30%)
Fiber Glass+ Asphalt (50%) + Limestone (30%) + Rice Husk (20%)
Fiber Glass+ Asphalt (50%) + Limestone (40%) + Rice Husk (10%)



Goal is to compare all the sample’s tensile, bond strength etc. to standard and then chose the best out of it.
What will be the design used for data analysis and How many replicates will be needed?
 

Miner

TS Contributor
#2
This is a specialized type of experimental design called a mixture design. Minitab has a sub-menu that addresses mixture designs. While I have 30 years experience running designed experiments, I have never run a mixture design, so my help will be somewhat limited there. Regarding the number of replicates, Minitab has a sample size calculator that can assist there. The big driver of replicate size will be the size of the effect vs. the standard deviation of the response variable. As long as the effect size you are looking for is larger than the standard deviation the number of replicates will be reasonable. However, if the effect size is smaller than the standard deviation, the number of replicates will rapidly increase.

What is the goal of this experiment? Are you trying to screen factors to find the important control factors? To develop a predictive model, or to optimize the output?

Plan on a series of sequential experiments to build knowledge. Don't try to learn everything in a single experiment.