Description

During the last decades, the digital transition, along with the environmental one, has become one of the two research topics of main interest of the entire scientific domain. The coupling between the huge computational power increase for data treatment and the support of artificial intelligence to generate new knowledge from the available one makes the overcoming of the current technologies a viable opportunity to tackle challenging problems in multiple engineering fields. In particular, the variety of process engineering related topics, ranging from thermodynamics to control and logistics, offers a wide spectrum of possible applications to test the AI potential on different types of problem formulation.
However, the direct application on this tools to the chemical engineering field has not resulted yet in the identification of a methodology of general validity; on the contrary, the majority of studies propose solutions that are strictly case specific. Moreover, as already discussed in several literature research works, when exploiting data-driven models for unit operations involving thermodynamic equilibrium or chemical reactions, the outcome does not necessarily fulfill fundamental conservation laws.
Therefore, the ambitious purpose of this PhD thesis project is to thoroughly define the methodological steps to be included whenever applying a procedure that couples AI data-driven tools and process systems dealing with chemical transformations in order to have meaningful results from a physical point of view. Furthermore, a procedure to define the applicability boundaries of the transition from phenomenological to data-driven models should be implemented in order to be aware of the reliability related limitations of the digital tools and to have a proper interpretation of their outcome.
The development of this research topic will require the use of process simulation software, surrogate modeling software and coding languages to develop and analyze different machine learning algorithms including statistical approaches for data treatment all over the three years PhD program.

Skills required

Process System ENG.

Bibliography

ras

Keywords

Process Systems Engineering, AI