Operability and Controllability Issues

The operability of reactive distillation processes has received little direct attention in the literature but the recent discovery of multiple steady states in a reactive MTBE column (Jacobs and Krishna, 1993; Nijhuis et al., 1993) has significant implications for both the operation and control of these columns. Jacobs and Krishna (1993) and Nijhuis et al. (1993) appear to have collaborated in their research as they studied the same column configuration and produced similar results. Neither group was able to adequately determine the cause of the observed behaviour but their investigations made an important contribution in refuting several hypotheses, including CSTR multiplicity, the exothermic nature of the reaction, and crossing non-reactive distillation boundaries via the reaction.

Although the citations indicated above were the first reports of multiple steady states in reactive distillation, a similar phenomenon had already been discovered for ideal binary distillation (Jacobsen and Skogestad, 1991) and extensively studied with respect to azeotropic distillation (e.g. Kienle and Marquardt, 1991; Laroche et al., 1992; Bekiaris et al., 1993). The results of Jacobsen and Skogestad (1991) were of seminal importance: they provided examples of multiplicity, a thorough theoretical analysis of those results, derived recommendations from their analysis and presented two fundamentally sound explanations for the observed behaviour. Later papers from the same authors (e.g. Jacobsen and Skogestad, 1994; 1995) extended the original results and addressed operability and controllability issues more directly.

Bekiaris et al. (1993) considered only azeotropic distillation and proposed a different approach to the analysis of columns which exhibit multiple steady states. They developed a geometrical tool which could be applied to temai7 composition diagrams to predict the product compositions and flow rates for any given feed composition provided that the column operated with perfect fractionation (infinité number of stages and infinite reflux ratio; the oo/co analysis). Bekiaris and Morari (1996) reviewed these results to explain the implications for design and provide limited predictions for multiplicity in ternary systems using only VLE data (i.e. the composition of azeotropes). This work was important but disregarded other possible causes of multiple steady states and addressed a restricted range of column control configurations.

Recently, Guttinger and Morari (1997) extended the original results of Bekiaris et al. (1993) so that the technique could be applied to homogenous reactive distillation. The tool was applied to the hybrid reactive distillation of MTBE and produced some promising results but the extrapolation of oo/oo predictions to finite columns is uncertain and the tool cannot be used if a constant internal column flow (e.g. the reflux or boilup rate) or a duty specificalion is required.

Hauan et al. (1995; 1997) also attempted to address the problem of multiple steady states in reactive distillation, and focussed on the column that Jacobs and Krishna (1993) and Nijhuis et al. (1993) had analysed. They presented a mechanistic explanation of the behaviour but provided only a single example (i.e. the recreation of previously published results) to support their hypothesis and were unable to demonstrate its validity. Their explanation is flawed in some crucial areas (e.g. the discussion of azeotropic behaviour and relative volatilities) and includes references to irrelevant data (e.g. activity coefficient profiles) which further confuse their postulations. Their contentions are tenuous (at best) and directly contradict those of Jacobs and Krishna (1993) who tested their conclusions with more rigour and provided examples to support their claims. (Note: Hauan et al. (1995) gave their paper a catchy title which ultimately proved to be more important than the weaknesses identified above since their paper was eventually published three times in essentially the same form!)

Schrans et al. (1996) also analysed a similar MTBE column to that which was originally described by Jacobs and Krishna (1993). Their contribution was to extend the modelling to transient responses and they were able to show a transition between parallel steady states, indicating that both the high and low conversion steady states are simultaneously accessible. Unfortunately, other results reported by these authors serve to confuse rather than clarify the issue of reactive distillation multiplicity since the important distinction between input multiplicity (i.e. multiple sets of inputs producing the same set of ouptuts) and output multiplicity (i.e. multiple sets of outputs from a unique set of inputs) was overlooked.

Ciric and Miao (1994) considered a different problem: multiple steady states in the reactive distillation of ethylene glycol. They used a homotopy continuation method to construct bifurcation curves which showed up to nine multiple steady states. They tested a range of hypotheses for the observed multiplicity but were unable to find a consistent and comprehensive mechanism.

Subsequent to the simulation studies of multiple steady states in ideal binary distillation, azeotropic distillation and reactive distillation, the natural question of experimental evidence for the phenomenon was addressed. Sundmacher and Hoffman (1995) completed experiments on a bench-top reactive distillation column for MTBE synthesis and found oscillatory behaviour but could not confirm multiple stable steady states. Kienle et al.

(1995) and Koggcrsbol et al. (1996) experimented with the ideal binary separation of methanol and propanol, and both groups were able to produce results which supported previous simulation studies and confirmed the presence of multiple steady states in that system. Gtittinger et al. (1997) performed two sets of distillation experiments on an azeotropic system and were able to produce results which were consistent with simulations and with oo/oo predictions. Thus, the presence of multiple steady states in distillation has become irrefutable.

Conclusive experimental evidence for multiple steady states in reactive distillation does not yet exist but there seems little reason to doubt that convincing results are forthcoming since the multiplicity phenomenon has been demonstrated for other systems. However, the significance of the simulation results is somewhat equivocal since many of the reports of multiplicity (e.g. Jacobs and Krishna, 1993; Nijhuis et al., 1993; Hauan et al., 1995; 1997; Schrans et al., 1996; Mohl et al. 1997; etc.) rely on an imposed requirement for a physically unrealisable condition (e.g. a constant molar flow). This crucial weakness suggests that simulation efforts should be redirected and creates a new motivation for expenmentation.

The design of control schemes for distillation processes is an established area of research but there is little evidence of direct attempts to design controllers specifically for reactive distillation. Jacobsen and Skogestad (1995) addressed some issues concerning the control of columns which exhibit multiple steady states but did not consider possible reactive distillation.

Multiple steady states in the reactive distillation of ETBE and MTBE are analysed thoroughly in Chapter 8, considering predominantly physically realisable conditions. Mechanisms for this phenomenon are also developed and unite the postulations of several research groups. The implications for this unusual behaviour are addressed with respect to design and operation (Chapter 8) and control (Chapter 10). The substantial issue of specific control applications for reactive distillation is also discussed extensively in Chapters 9-10. Once again, the experimental work (Chapter 11 and 12) provides the basis for further study and an ideal platform to develop experimental evidence for multiple steady states in reactive distillation and to test control systems practically.

Was this article helpful?

0 0

Post a comment