Catalyst Deactivation - Spent Hydroprocessing Catalysts - page 90

ideal mixing in which radial and axial dispersions as well as back mixing could be neglected.
There was a reasonably good agreement between the predicted and experimental data for the
top layer of the catalyst. Theoretical part of this study includes several correlations in which
the values of Thiele modulus were arbitrarily chosen to identify the effect of the presence and
absence of diffusion on the V distribution profiles on the particle level. With such correlations
available, the meaning of Thiele modulus becomes much clearer.
The thermal monitoring for iso-performance desulfurization of oil residues (THERMIDOR)
model was developed by the French Institute of Petroleum to simulate operation of the
HYVAHL process comprising two guard reactors operating in the perturbating mode upstream
from the series of four fixed-bed reactors [290] . The model considered both the grain scale and
the bed scale. The shape of catalyst particles (sphere, cylinders, polylobe, etc.) were considered
at the grain level using the fractional radius of the catalyst particle as space variable, whereas
the fractional axial position was used to describe changes between the inlet and outlet of the
reactor. More than 40mathematical equations, which were considered represent one of the
most comprehensive approaches used for the model development. These equations were
beyond the scope of this review. The pseudo Langmuir-Hinshelwood law was used to express
the rate of hydroprocessing reactions as first-order in disappearing reactant (e.g., sulfur, metals
and asphaltenes) using the assumption that about 10% of asphaltenes present in the feed were
responsible for coke build up. Figure 4.27 [290] shows the experimental and simulated
deactivation curves for Boscan heavy feed using the macroporous HDM catalyst. Remarkable
agreement between the experimental data and those predicted by the model should be noted.
The THERMIDOR software incorporates Fick law to express the local molar fluxes in the
liquid phase. The Stokes-Einstein law and Andrade law were combined to describe the effect
of temperature on the feed viscosity and molecular diffusivities. The random spheres, random
needles and random coins models were used for the representation of the catalyst porous
media. Other equations derived as part of the THERMIDOR project enabled simulations of the
additional parameters, e.g., the change in the inlet and outlet temperature with time on stream,
longitudinal profiles of temperature, Tamm factor, surface area, porosity, etc. with time on
stream. Another model suitable for the simulation of a multistage system performance (e.g.,
ARDS) was developed by Kodama et al. [291] was used by Al-Adwani et al. [292] to simulate
performance of the four reactors in a series, such as in the ARDS process. Apparently, this
model was supposed to be less dependent on the catalyst and feed properties than the other
models. The model could predict the effect of LHSV on the set-point temperature for
achieving certain level of HDS. For a given degree of HDS, the set-point temperature
increased with the increasing LHSV, i.e., the decreasing contact time. In general agreement,
the model predicted that the degree of HDS parallels that of HDM.
Few attempts have been made to develop the models to simulate performance of the
ebullated-bed reactors. In this regard, it has been recognized that the artificial neural network
may be a suitable modeling tool [293,294] . The neural networks comprise computing systems