Prediction Coking

The best linear regression model to predict coke has the form for catalyst A ... 

Figure 2 Predicted coke-conversion selectivity as a function of catalyst activity (crackability) for REY catalysts.

A parallel fouling model has been developed to represent experimental observations for hydrotreating a coal oil in a trickle-bed reactor over a commercial NiMo/Al catalyst. This model accurately predicts coke profiles with time and reactor position, and hydrogenation and hydrodenitrogenation as functions of coke content. The following conclusions can be drawn from this study. 

By high API gravity reduced crudes it is meant reduced crudes with an API gravity of 20 or greater. Some of the variables in coke production were taking effect as can be seen in Table V, where predicted coke yield falls far short of the actual coke yield. 

Table 3 Predicted coke yields from various feedstocks...

Table 2 compares the predicted coke profile to the experimentally observed profile. Coke was measured by carrying out thermogravimetric analyses (TGA) on the three sections of catalyst bed after the experiment was completed. Therefore, the values shown in the following table are averages over the first, middle, and last thirds of the bed. 

Petrographic analyses are also used to predict coke properties (such as strength) which would be produced from the coal. Other uses include the estimation of the chemical properties of fresh coals from the reflectance of weathered specimens. 

Coal petrography (Chapter 4) has become widely used for predicting coke quality based on coal analysis and has led to a system for predicting coke stability based on petrographic entities and reflectance of coal (Schapiro and Ciray, 1960). Thus, an optimum blend of coals could be selected to produce desired coke quality. 

Seam correlations, measurements of rank and geologic history, interpretation of petroleum (qv) formation with coal deposits, prediction of coke properties, and detection of coal oxidation can be deterrnined from petrographic analysis. Constituents of seams can be observed over considerable distances, permitting the correlation of seam profiles in coal basins. Measurements of vitrinite reflectance within a seam permit mapping of variations in thermal and tectonic histories. Figure 2 indicates the relationship of vitrinite reflectance to maximum temperatures and effective heating time in the seam (11,15). 

Knowledge of the composition of coal ash is usehil for estimating and predicting coal performance in coke making and, to a hmited extent, the folding and corrosion of heat-exchange surfaces in pidverized-coal-fired furnaces. 

Feed residue coke is the small portion of the (non-residue) feed that is directly deposited on the catalyst. This coke comes from the very heavy fraction of the feed and its yield is predicted by the Conradson or Ramsbottom carbon tests. 

This interpretation of the experimental data is supported by the differences observed in the deactivation patterns and carbon contents after test, since one notorious effect of Hjp is the capacity to diminish the deactivation caused by coke deposition on the active sites [21,22]. This is supposed to be due to a reaction with the coke precursors, very likely a hydrogenolysis. In pure silica-aluminas, where no source of spillover is present, no special protection against deactivation should be observed. Indeed, the silica-aluminas lose most of their activity (about 80%) before reaching the steady-state and present the highest carbon contents after catalytic test. On the other hand, in the case of the mechanical mixtures, where spillover hydrogen is continuously produced by the CoMo/Si02 phase and can migrate to the silica-alumina surface, the predicted protection effect is noticed. The relative losses of activity are much lower... 

There are also several possibilities for the temporal distribution of releases. Although some releases, such as those stemming from accidents, are best described as instantaneous release of a total amount of material (kg per event), most releases are described as rates kg/sec (point source), kg/sec-m (line source), kg/sec-m (area source). (Note here that a little dimensional analysis will often indicate whether a factor or constant in a fate model has been inadvertently omitted.) The patterns of rates over time can be quite diverse (see Figure 3). Many releases are more or less continuous and more or less uniform, such as stack emissions from a base-load power plant. Others are intermittent but fairly regular, or at least predictable, as when a coke oven is opened or a chemical vat... 

The values in the first two columns of Table IV show the distribution of original carbon in products of distillation values in the next two columns show the distribution of available carbon in products of oxidation. The large difference between the value of cox(s) at 317°C predicted by Equation 21, 17.4%, and the observed value of 0% underscores the validity of our proposed change in mechanism near 285°C. Additional evidence for this change is provided by the carbon contents of the residual cokes ... 

There are several factors that may be invoked to explain the discrepancy between predicted and measured results, but the discrepancy highlights the necessity for good pilot plant scale data to properly design these types of reactors. Obviously, the reaction does not involve simple first-order kinetics or equimolal counterdiffusion. The fact that the catalyst activity varies significantly with time on-stream and some carbon deposition is observed indicates that perhaps the coke residues within the catalyst may have effects like those to be discussed in Section 12.3.3. Consult the original article for further discussion of the nonisothermal catalyst pellet problem. 

Parameters in the model are listed in Table I. The flow, structural, and boundary conditions are known quantities. The frequency factor and activation energy for coke burning were the values determined by Weisz and Goodwin (1966) from the experiments discussed earlier, and the catalyst diffusivity D was measured directly in the laboratory. The value of a was determined from direct observations of the CO/CO2 ratio in each zone of the operating kiln. The remaining parameters are known quantities. Thus, there are no adjustable parameters available to tune the fitting of predicted values to observed data, for the fraction of coke remaining and for the vertical temperature versus distance from the top of the kiln. 

The response of the kiln to a 20% increase in coke on the catalyst from the reactor is shown in Fig. 27 for no control and for schemes (a) and (b). The simpler scheme (b) is clearly superior to scheme (a) although steady-state considerations predicted that scheme (a) would be the better strategy. The fluctuations in the total air rate of scheme (a) that maintains a constant amount of oxygen to the kiln causes this difference and outweighs the effects of the fluctuations in oxygen amounts present in scheme (b). This comparison showed that control strategies designed... 

Leeder, W.R. Gransden, J.R. Price, J.T. Botham, J.C. Prediction of Coke Quality with Special Reference to Canadian Coals, Proc. Ironmaking Conference (AIME), 1979, 38, 385. 

---