Francoise Couenne

Time series clustering is applied to inflammation and neutrophil cell development markers, CD16 and CD10, in sepsis, trauma and surgery patients and a dynamical model with an inflammation function, F , is used to represent their evolution over a two month period. Five patient clusters are identified, characterised and evaluated against medical assessment scores and the literature. A dynamical model for neutrophil and monocyte cell counts and maturity has been constructed based on mass balances and cell kinetics in both blood and bone marrow. Cell proliferation and flow rates, as well as expression of monocyte HLA-DR, depend on concentrations of pro- and anti- inflammatory cytokines, IL6 and IL10, via F . A good fit with the data is obtained for each cluster and the estimated parameters correlate to illness severity. The model is a potential tool for simulation of immunomodulatory therapies.

In this paper we propose a port Hamiltonian formulation of the two-phase Stefan problem. This problem describes the evolution of a pure material which is at two different states of matter: liquid and solid. This is a parabolic distributed parameter system which possesses a sharp moving interface. The two-phase Stefan problem results from the interconnection of two heat conduction equations at the interface position. The interface dynamics is governed by an ordinary differential equation. The port Hamiltonian formulation relies on the introduction of color functions which identify the liquid and the solid states. The contribution of this paper concerns the formulation of the two-phase Stefan problem as a sharp interface port Hamiltonian system. A thermodynamic-based modeling approach is considered to provide a physical insight on the proposed model.

The tortuosity factor of γ-alumina supports and catalysts used in hydrotreating applications was evaluated in the liquid phase with three different techniques: pulse field gradient-NMR (PFG-NMR), inverse liquid chromatography (ILC) and catalytic experiments in a batch reactor. In order to satisfy the specific experimental constraints associated with each technique, tortuosity factors were evaluated in a wide range of operating conditions: temperature was varied from 295 K to 613 K, catalyst particles were used as-synthesized (trilobe extrudates) or crushed, and the liquid composition was either pure toluene, n-heptane/squalane mixtures, or squalane/2,5-bis-(octadecyl)thiophene mixtures. It is demonstrated that not taking into account both the shape factor and the size distribution of the particles can lead to significant errors in the tortuosity factor. For both parameters, the optimal mean spherical radius depends on the contribution of internal diffusion to the overall performance. When internal diffusion is the limiting step, a new expression for the mean radius of the distributed particle is proposed and validated by comparison between ILC and PFG-NMR results. This expression is transposable to any catalytic system, making it possible to measure the tortuosity factors using either microscopic or macroscopic methodologies. Moreover, the tortuosity factors obtained with ILC and NMR are in good agreement with the one estimated from the catalytic experiments, showing that mass transfer parameters can be extrapolated from non-reactive to reactive conditions.

In this work, a model-based control of the melt temperature of a distributed parameter bi-zone model with moving interface of an extrusion process is studied. An energy-based controller is derived using internal energy as a storage function and controlled variable with the screw speed as control input. The stability of the controller is proved through the use of a Lyapunov-like candidate function. Finally, the practical usefulness of the method is illustrated by some closed-loop simulations of the experimentally verified model of the extrusion process. The interconnection of the moving interface is performed under the assumption of variable viscosity along the extruder and a finite-volume scheme is used for the discretisation of the dynamic model.