Data Mining Technologies
Solutions, strategies and software for experiment design in research, high speed experimentation and process engineering
Our approach
Statistical design of experiments is based on, among other things, the application of regression models. It is particularly well suited for the design of experiments for the identification of key variables at the beginning of a study (screening) when there is little or no prior knowledge. If the parameter space is high dimensional, neural networks rather than regression models should be used in the modeling phase to reduce the number of additional experiments.
Regression models are much too expensive for optimization in high dimensional parameter spaces with discrete input variables. In such cases, evolutionary experiment design based on genetic algorithms can be used. The efficiency of these processes can be further improved if data mining methods are applied to data analysis. The significance of results, which are often difficult to interpret, can be improved and data mining can identify new correlations across several process levels with no prior knowledge or reveal hidden information.
So far, this form of integrated design of experiments has been applied successfully to catalyst screening (high throughput screening) and product quality optimization. In cooperation with our partners, we continuously improve and update our methods.
Our broad spectrum of methods and closeness to practice are our strength and ensure the success of our customer’s projects.
We provide our customers with consultancy services and take charge of complex DOE and data analysis projects. Typical examples are:
- improvement of recipes and products;
- process optimization;
- DOE and data analysis for high speed experimentation;
- consultancy for the development of DOE and data analysis workflows;
- optimization of reaction conditions;
- analysis for the economic feasibility of chemical reactions;
- development of customized tools (ECCO, VP-Tool).
Go to Top 
Print 
Recommend Page 
