Experimental design for chemical mixtures: A case study on a molten salt system conductivity

Abstract

Experimental studies in many fields such as chemistry, biochemistry, food and environmental sciences require very well designed experimental strategies to achieve the best results at minimum cost. At this point experimental design means application of statistical methods to the selection of experimental process factors and to the modelling of system responses to them. Models in which experimental results are expressed depending on experimental factors are also called response surface functions. Mixture designs as a special type of response surface function are also a very effective method for determining the proportions of components in mixtures and are also of great importance in chemical industries like paint, food, glass, ceramic and polymer. Physicochemical properties such as thermal-physical properties, fusion enthalpy, heat capacity, density, thermal stability and conductivity for pure materials or their mixture can be determined experimentally by thermal analysis methods depends on their ingredients. In this study to design molten salt mixtures for thermal energy storage, computational thermodynamics approach has been applied on an experimental design sample. For this purpose conductivities of a molten salt system with three components were measured experimentally depending on the relative ratio of each component, and the effect of these ratios on the conductivity of the mixture was investigated. Here, mixture design model chosen for optimization of process parameters is Simplex Lattice, and Design Expert 700 software were used for statistical analysis. After the model selection for experimental design and transformation of the all factors, the correlation coefficients of the response surface function with linear, quadratic and cubic terms based on all factors for experimental measurements and pseudo and real factors were determined separately. Comparing with the others the best fit of the response surface function sufficient for optimization of process parameters to the experimental measurements was found out for that of with quadratic terms.