Intelligensys neural computing modeling and optimization for product formulation

If you have good experimentally-determined cause-and-effect data for your formulation, then our software finds a usable model for you - one that can put you on the fast track to an optimum formulation.

Improving one property frequently means that another performance variable is degraded, or that cost goes up - historically, this has meant that balancing trade-offs to meet the needs of production and marketing often involves a lengthy and expensive series of lab trials, with no guarantee of success. Now, though, you can get models automatically. So, you can try out a number of ideas quickly, predicting what will happen if you make changes in ingredient amounts and/or process conditions.

But modelling alone won't solve all your problems - you need to bring in other techniques like visualization and optimization to get the full picture.

The figure below shows how this works - you can use your model for 'what if' investigations to see how changing specific inputs affects the properties. And you use the same model to find the formulation that best meets your current objectives. All you need to do specify the desired values for each property and rank the properties in order of their importance. Sophisticated optimization techniques then find the best solution for you.

see if you can meet specific needs of individual customers, with relatively small adjustments in your formulation

save research effort and development time

balance the trade-offs associated with producing your product at lower cost or to other specific requirements.

communicate the results clearly to colleagues and customers

Historically, statistics have been used to develop models, but using stats often requires a high level of mathematical expertise. More recently, neural networks have gained in popularity, especially for complex formulation problems. The combination of neural networks to develop models, with genetic algorithms for optimization, has proved exceptionally powerful in fields ranging from pharmaceuticals to coatings to personal and household care products. And neurofuzzy logic, a powerful combo of AI approaches, gives you insights into cause-and-effects that are simply unobtainable by conventional means. Now, these techniques are readily accessible to product formulators.

Intelligensys Solutions

1. "Do-it-yourself"
INForm combines neural networks and genetic algorithms, together with statistics and visualization capabilities, to give you all the tools in one package, specifically tailored to formulation optimization.

FormRules uses neurofuzzy logic to develop models that involve only the most important variables, so that you can see what variables actually affect each of the properties. It enables you to decide your future experimental strategy and, where possible, to week out extraneous ingredients that contribute to cost but not to performance.

Both INForm and FormRules are designed to be used by product formulators themselves, enabling them to make better decisions on the direction in which to take their projects. And if you want help in 'designing' the experiments so that all variables are explored fully, FormData offers an easy-to-use solution.

2. Get us to do it for you
We can also carry out the work for you, on a completely confidential basis. Just provide us with your data, and our contract research team can develop the models and suggest the optimum formulation to meet your requirements. To discuss this in more detail and find out how it could work for you, just contact us.

An Overview: Generating Models from Data