(191a) Accelerating Viscoelastic Foam Development Via Descriptor-Based Experimental Design and Modeling | AIChE

(191a) Accelerating Viscoelastic Foam Development Via Descriptor-Based Experimental Design and Modeling


Tong, X. - Presenter, Louisiana State University
D'Ottaviano, F., The Dow Chemical Company
Nunez, A., Dow Chemical Company
Thiede, C., Dow Chemical Company
Springs, M., Dow Chemical Company

The progress in viscoelastic foam technology garners specific attention within the realm of flexible slabstock foam, especially when considering its applications within the bedding industry. The slow responsiveness, low resiliency, and desired weight distribution properties of viscoelastic foam make it essential in the development of sleep products. The desired formulation is usually defined by targeted ranges of performance specifications and the preference and availability of raw materials.

Various formulation approaches exist for viscoelastic technology development, and studies have shown how independent variables of raw materials affect viscoelastic foam properties. However, the relationships among the codependent variables and their synergistic effect on foam properties remain unknown. The lack of systematic knowledge of the descriptor-property relationship of viscoelastic foam leads to inefficiencies in conventional approaches where multiple formulation cycles are usually needed to achieve the desired properties.

Over the past few years, Dow Polyurethanes has invested substantial resources in advancing digitalization and predictive capabilities via its Predictive Intelligence program. The use of experimental design and statistical models within the chemical domain has empowered researchers and manufacturers to develop new products, optimize production processes, and improve sustainability in a more efficient way.

This study outlines a systematic approach for learning descriptor-property relationships of viscoelastic flexible foam. A descriptor-based experimental design was carried out and empirical models were built to capture the variability in foam properties. The model was validated based on the candidate formulations with predicted foam properties in a specific area of interest. With the implementation of the model, formulations with targeted properties were obtained in a significantly reduced formulation cycle compared to conventional formulation approaches. This work shows the contribution of statistical tools in directing and accelerating the development of intricate systems in the Polyurethane industry.