DataModeler in Chemical Industry

DataModeler has earned its place at the leading edge of data-driven modeling and analysis tools for industrial applications. The chemical industry (DataModeler’s cradle) has benefitted particularly from its unique features: elegant handling of challenges of real-world data, ease of use, transparent models enabling insight, quantified uncertainty, and field-tested & success-proven modeling workflows.

In process analytics DataModeler is used for modeling, prediction, and subsequent optimization of production processes, e.g. for yield maximization, off-spec minimization, time, energy, cost reduction and improvement in control strategies.

Inferential Sensors built with DataModeler are widely used for real-time predictions of difficult-to- measure process variables through combinations of available and easy-to-measure inputs. Reasons for success are the capability of DataModeler to automatically handle superfluous factors, correlated inputs, nonlinear couplings, noise and measurement uncertainty.

Model transparency and potential for insight makes the technology the first choice for building the emulators of complex process models and simulators. High fidelity empirical emulators built with DataModeler are used to tackle larger problems by substituting CPU-intensive blocks, to dramatically cut evaluation time. These surrogates for first-principle models can be used to accelerate plant or product optimization as well as facilitate operational training.

The hypothesis generation capacity of DataModeler has made it an indispensable tool for research analytics and material design. The competitive advantages in these areas include identifying and predicting property-property relationships in very high-dimensional descriptor spaces, as well as in the active design of experiment workflows for understanding fundamentally new materials and focusing exploration on the right areas of design space.

DataModeler accelerates Materials Science:

  • The variable selection functionality of DataModeler helps to focus new materials research on only a handful of significant variables in the descriptor space which influence desired response behavior.
  • Predictive modeling enables robust prediction and optimization of material properties for efficient product design.
  • Active design of experiments guided by developed models offers effective and fast exploration of the design space while reducing of the lab costs.
  • Capability of creating plausible transparent models helps interpret and guide high throughput experiments and facilitate fundamental modeling.