Intelligensys Formulating Knowledge

Developing better products faster

Adhesives

Adhesives are used in most manufacturing industries, from aerospace and marine, to automotive, to building and construction. Adhesion is a complex process, and adhesives can function in a variety of ways (solvent evaporation, hardening upon cooling, chemical crosslinking and so on). Generally adhesives are complex formulations designed for a particular end-use application.

Here, we show how data mining, modeling, optimization and (in some cases) simulation can be used to understand this complex process.

In the first instance, we used neurofuzzy logic, the technology in the FormRules program, to find out the key relationships in three different data sets, taken from published data (1-3). Customers with a password can find all the data sets here, in an Excel 'book'. The full results of our data mining study can be found here.

Of course, the objective of many formulators is to optimize a coating to obtain specific values for properties like hardness and gloss. So, we have also used INForm's modeling and optimization capabilities to look at the best combinations of ingredients, and the best processing conditions, to find strong adhesives. These used the same data sets as the FormRules study. Our full results can be found here.

Simulation helps to understand how changing the relative amounts of materials can affect the crosslinking, and hence end-use properties. Here, we have considered a specific system - an epoxy resin. Epoxies are among the most popular thermoset adhesives. See the full results for the simulations here.

If you are interested in melamine-formaldehyde adhesives, you might want to look at the notes that we did for coatings, using DryAdd to simulate melamine-formaldehyde crosslinked systems. You can find information about MF crosslinkers here, and a simulation of a MF-polyester crosslinked system here.

All these approaches gave useful information. The data mining approach showed the rules that govern the relationships between the measured properties and the ingredients and processing conditions. In addition, the data mining highlighted those cases in which perhaps the data was not really adequate for good models to be developed.

Modeling and optimizaton allowed optimum formulations to be suggested 'in silico'. Finally, simulation showed how varying the relative amounts of the key ingredients (e.g. the resin and crosslinker) affected the crosslink density and, consequently, properties like hardness and Tg which relate directly to the mechanical performance of the adhesive.

(1) R H Myers and D C Montgomery, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, p 712, John Wiley & Sons, New York (2002)

(2) Patent Specification 1 515 727 Improvements in or relating to Copolyesters

(3) S Setz, M Semling and R Mülhaupt, Fuzzy set approach for fitting a continuous response surface in adhesion formulation, Journal of Chemometrics, 11 403-418 (1997))