Automated, computational approaches to kinetic model and parameter determination

A major bottleneck in the transition from chemistry research at lab scale to process development is a lack of quantitative chemical synthesis information. Critical aspects of this information include knowing the correct reaction model and precise kinetic parameters. If this information is available, classical reaction engineering principles may be utilised to shorten process development times and lower costs.

Identifying the correct reaction model for a particular process, however, can be challenging and time-consuming, particularly for physical-organic chemists and kinetics experts that may be busy with other aspects of process development. The work presented herein describes computational approaches that automatically determine the most likely kinetic model and associated parameters based on the experimental data supplied, without expert chemical intuition.

The concept for these methodologies involves a comprehensive model evaluation tool. The experimental data and the species involved in the process are inputted. Based on mass balance, all mass-balance-allowed transformations between these species are identified. All possible models are then compiled from this list of transformations, featuring unique combinations of these model terms. Every model is then evaluated using ordinary differential equation (ODE) solvers and optimisation algorithms to maximise the convergence of simulated reaction progression with the experimental data, thereby identifying the kinetic parameters. Each model is then statistically evaluated to determine which model is the most likely to be correct.

Using these methodologies allows any chemist to automatically determine a reaction model and kinetic constants for a particular system, by performing all kinetic analysis autonomously. Their most expensive resource, time, can then be focussed on other tasks that cannot be automated.