Lowering manufacturing cost of material by formulating it through statistical modeling and design

At Hoechst Celanese, the methods of statistical modeling (experimental design) and multivariate analysis are used extensively to enable scientists and engineers to better understand their processes, improve the quality of existing products, develop new processes/products and often reduce the manufacturing cost. As an example we discuss how a commercially available material AA (polyester bend) was reformulated through a series of experiments, and a series of relevant properties were analyzed by multivariate analysis (PCA and PLS) techniques with two objectives: (a) reproduce and/or improve the performance properties of the material and (b) reduce the raw material cost by at least 25%. By using statistical design we were able to plan experiments more effectively; by using multivariate techniques we were able to analyze a series of highly correlated properties, derive mathematical models, and identify the formulation(s) that not only have acceptable performance but also reduce the manufacturing cost. All the methods and techniques used to help us achieve our objectives are discussed in this paper.