Special Contributed Session: Statistics at the Indian Statistical Institute
Optimisation of Concrete Strength Using Mixture Designs: An Application
A Civil Construction Company was facing the problem of inadequate strength in a certain mix of concrete. Concrete is basically a mix of cement, sand and mortar with water. The company desired to determine the optimum concrete mix design. An experiment was planned and conducted using the principles of ``Mixture Designs'', where the proportions of cement, sand and mortar were varied within certain limits. The proportion of water was kept constant. Different empirical models were fitted representing the strength as a function of proportions of these ingredients. The best fit model was selected. Using this model, the optimum combination was selected employing direct search technique. The model and the search also indicated suitable sub-optimal feasible domains of the concrete mix. These solutions could be made use of when a specified strength is to be achieved, though it might not reflect a global optimal solution. The methodology adopted along with the problem faced is detailed in this paper.
[ Somnath Ray, Indian Statistical Institute, 8th Mile, Mysore Road, Bangalore 560 059 INDIA; email@example.com ]
Sensitivity of SN Ratios Analyses: A Case Study
Inferences from SN ratios analyses are entirely different from that of original data analysis. Validity of SN ratios analyses has not been well established. To explore similar remarks, a comparison of analyses of original data and Taguchi's signal to Noise (SN) ratios was made in relevance to an expereiment on paper-pulp composition. The experiment was aimed at exploring the scope for loading higher amount of filler, than the existing 15% level, in 60 GSM paper with a view to conserve scarce pulp materials - bamboo and wood. OA design with 27 trials was used to experiment with 8 factors at 2 or 3 levels in order to determine pulp-composition that balances 5 paper characteristics, viz., Porosity, Ash Retenition, Breaklength, Tear factor and Burst factor. ANOVA of data on the charactaristics, and transformed SN ratios was supplemented with probability plots. Analyses of SN ratios was observed to discriminate the factor effects (and interactions) better and provide more precise estimates in respect of all the 5 characteristics. The parameter combinations that permit 25% loading of filler, and even higher level could be identified.
[ J. Venkatappaiah, Indian Statistical Institute, Street No. 8, Habshiguda, Hyderabad, INDIA 500007; firstname.lastname@example.org ]
Date created: 6/5/2001