Ammonia Syngas Plant: Process Improvement Using Multi-Input-Multi-Output (MIMO) Surrogate Models

Ammonia (NH₃) is among the most manufactured industrial chemicals in the world. In most cases, it is produced from natural gas that is taken through a series of complex reaction steps together with steam and air. The key to successful process optimization is a good understanding of how various process variables interact. The process is characterized by high energy demands, complex reaction paths, and relatively high operating temperatures and pressures.

Process interactions that are directly influencing production and energy efficiency can become increasingly complex to detect. The application of detailed process models and common sensitivity studies is often not enough for making correct optimization decisions when a process shows as complex and non-linear variable interactions as the Ammonia Syngas Plant does.

The application of an extensive Multivariate Sensitivity Study (MSS) allows the simultaneous examination of multiple input and output relationships and their effect on simulation results. MSS contributes significantly to the understanding how changes in the plant's operating conditions influence conversions and energy efficiencies and therefore increases process knowledge. The data generated by the MSS is transformed into surrogate models (SM), which are used to set optimal operating conditions to achieve economical benefits.

In this contribution, a multivariate sensitivity study is carried out on a cluster of DWSIM instances to calculate data and to train a multi-input-multi-output surrogate model. We present results from an optimization for energy efficiency by simultaneous adaptation of 9 process variables.