Design and Control of Heat-Integrated Reactors

Industrial & Engineering Chemistry Research is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036

Article

Design and Control of Heat-Integrated Reactors

Yih-Hang Chen, and Cheng-Ching Yu

Ind. Eng. Chem. Res., 2003, 42 (12), 2791-2808 • DOI: 10.1021/ie020324r

Downloaded from http://pubs.acs.org on November 28, 2008

More About This Article

Additional resources and features associated with this article are available within the HTML version: • Supporting Information

• Access to high resolution figures

• Links to articles and content related to this article

PROCESS DESIGN AND CONTROL

Design and Control of Heat-Integrated Reactors

Department of Chemical Engineering, National Taiwan University of Science and Technology, Taipei 106-07, Taiwan, and Department of Chemical Engineering, National Taiwan University, Taipei 106-17, Taiwan

The objective of this work is to quantify the controllability of a complex heat-integrated reactor. Similarly to the concept of ultimate gain in the control literature, the ultimate effectiveness is defined for a complex feed-effluent heat exchanger (FEHE) scheme. This parameter indicates the amount of heat that can be recovered (via FEHE) before the overall open-loop system becomes unstable. First, a systematic approach is proposed to model the complex heat-integrated reactors. A simple measure, the overall effectiveness, can be derived directly from the flowsheet. Given the reactor model, the controllability of a particular flowsheet can be evaluated on the basis of the stability margin of design. An interpretation based on the heat-generation and heat-removal curves is also given. With the controllability measure, implications for design are also explored. Because the loss of controllability comes from the positive feedback loop, several design parameters are studied, and design heuristics are proposed to improve the controllability of heat-integration schemes. Two examples, a simple two-FEHE example and an HDA example, are used to assess the controllability of different designs. The results show that, contrary to one’s intuition, some of the complex heat-integrated reactor design alternatives (e.g., alternatives 6 and 7 of the HDA example) are indeed more controllable than the simpler heat-integration schemes (e.g., alternative 1). The increased number of FEHEs allows for a greater number of candidate manipulated inputs and thus provides opportunities for multivariable control. Contrary to one’s intuition, the multivariable controlled FEHE/reactor system gives a steeper slope and a lower peak for the closed-loop load-transfer function. This results in an improved disturbance rejection capability.

1. Introduction

According to a second-law analysis, the efficiency of the majority of chemical processes ranges from 20 to 30%.13_{Thus, there exists an incentive for improvement} by heat integration. This is accomplished by controlling energy flows between process streams via heat exchang-ers. Approaches include heat exchanger networks11_and heat-integrated reactors and columns.9,13_{An important} industrial example is the feed-effluent heat exchanger (FEHE), where the heat of the exothermic reactions is recovered by preheating the reactor feed.1-3,7,10,14,18 Heat-integrated schemes can become very complex when plantwide energy management is taken into consideration [e.g., alternative 6 of the hydrodealkyla-tion (HDA) process9,15,16_{]. Literature examples show that} positive feedback makes the plants more difficult to control.9,18 _{This is an important example of the} ever-present tradeoffs and conflicts between steady-state economics and dynamic controllability. The objective of this work is to provide a systematic approach to model and evaluate the dynamic controllability of heat-integrated reactors on the basis of steady-state flow-sheets.

Luyben et al.6 _{probably presented the first study of} the effect of design on the control of FEHEs. This heat-integrated process is nicely analyzed using the heat generation-removal curve. In this work, a more control-oriented approach is taken to look at the interaction between design and control. First, relationships between the degree of heat recovery and open-loop poles are derived. Then, a stability limit for the FEHE, the ultimate heat recovery, is derived analytically. The derivation is then extended to complex FEHE schemes (more than one FEHE), and for different degrees of complexity, a single measure, the overall effectiveness, can be derived. Tradeoffs between heat recovery and system stability can be interpreted with the familiar heat generation-removal plots. Because the stability margin is a good measure of controllability, achievable closed-loop performance can also be evaluated at the design stage. The overall effectiveness provides insight into the controllability analysis, and heuristics for the design of controllable heat-integrated systems are gen-erated accordingly. Two FEHE examples are used to illustrate the closed-loop performance for different designs. The results presented here offer a simple way to synthesize controllable heat-integrated reactors.

Pioneering papers on the control of FEHE systems include those of Douglas et al.3 _{and Silverstein and} Shinnar.14 _{Recently, Luyben and co-workers}6,8,18

ex-* To whom correspondence should be addressed. Tel.: +886-2-3365-1759. Fax: +886-2-2362-3040. E-mail: [email protected]. †_{National Taiwan University of Science and Technology.} ‡_{National Taiwan University.}

10.1021/ie020324r CCC: $25.00 © 2003 American Chemical Society Published on Web 05/17/2003

plored the stability and controller tuning for this posi-tive-feedback system. Chen and Yu2_{studied the} interac-tion between thermal efficiency and control of the FEHE system. Seider et al.13 _{and Yan et al.}19 _{explored the} control of a heat exchanger network in a plantwide framework. All of these literature studies have consid-ered reactors with a single FEHE. The purpose of this paper is to explore the control of more complex heat-integrated schemes, i.e., multiple FEHEs heat-integrated with an adiabatic exothermic reactor.

2. Feed-Effluent Heat Exchangers (FEHEs) In an adiabatic reactor, the heat of the exothermic reactions can be recovered by preheating the feed. A typical example is the feed effluent heat exchanger (FEHE), where the reactor outlet stream (Tout) is used to heat the feed stream (Tf) to the reaction temperature (Tin), as shown in Figure 1A. As pointed out by several researchers,1,2,5,10,16,19_{the positive feedback nature} re-sulting from the heat recovery makes the simple FEHE scheme (Figure 1A) difficult to control. The process configurations can become even more complex as plant-wide energy management is taken into consideration. In the HDA example, Terrill and Douglas15_{show that,} from a heat exchanger network (HEN) analysis, an additional 20-40% of the energy can be saved, and this results in quite complex schemes in the FEHE.9 _For example, two or three feed effluent heat exchangers (Figure 1B and 1C) are required for alternatives 3-6.15 A systematic procedure is necessary for the modeling and analysis of such complex schemes.

2.1. Modeling. The thermodynamic effectiveness () of a heat exchanger was proposed as an important parameter by Kays and London in the 1940s.4_{For a heat}

exchanger (Figure 2), the effectiveness is defined as

where (m˘ Cp)H, (m˘ Cp)C, and (m˘ Cp)min are the capacity flow rates for the hot stream, the cold stream, and the smaller of the two, respectively. Depending on the heat-transfer area,  ranges from 0 (no heat exchangers) to 1 (infinite heat-transfer area). Assuming that the heat capacities of the hot and cold stream are the same, i.e., Cp,H) Cp,C, and because the mass flow rates of the hot and cold streams are identical in an FEHE, eq 1 can be rearranged to give

Equations 2 and 3 describe the propagation of the temperature disturbance as the inlet temperature changes, (Figure 2) and they serve as the basic formula for the construction of more complex schemes. Because there is no parallel definition of an effectiveness factor for a steam generator, a modified effectiveness (ex) is proposed to describe the energy balance for a steam generator (or reboiler). The hot-stream outlet temper-ature (TH,2) can be expressed in terms of the inlet temperature (TH,1) and the steam temperature (TH) as

Appendix A gives the derivation of this expression. Consider a simple heat-integrated reactor with an FEHE and a steam generator (Figure 1A) with ef-fectiveness factors 1 and ex,1, respectively. A block diagram can be constructed according to the flowsheet (Figure 3A)

where Tf is the feed temperature, Tin stands for the reactor inlet temperature, Toutdenotes the reactor outlet temperature, and Tout,1represents the exit temperature from the steam generator. GRis the transfer function that describes the reactor outlet temperature as the inlet temperature changes. Rearranging eq 4, we have

This is a positive feedback system, and the loop gain depends on the effectiveness factors (1 and ex,1) and the reactor gain [KR ) GR(0)]. For a heat exchanger network (HEN), the loop gain cannot exceed 1 (iand Figure 1. Heat-integrated systems with (A) a single FEHE, (B)

dual FEHEs, (C) triple FEHEs.

Figure 2. Heat exchanger block diagram representation.

 ) (m˘ Cp)H(TH,1- TH,2) (m˘ C_p)_min(T_H,1- T_C,1)) (m˘ C_p)_C(T_C,2- T_C,1) (m˘ C_p)_min(T_H,1- T_C,1) (1) TC,2) (1 - )TC,1+ TH,2 (2) TH,2) TC,1+ (1 - )TH,1 (3) T_H,2) (1 - _ex)T_H,1+ _exT_C Tin) (1 - 1)Tf+ 1Tout,1 ) (1 - 1)Tf+ 1(1 - ex,1)Tout ) (1 - ₁)Tf+ 1(1 - ex,1)GRTin (4) T_in T_f ) 1 - ₁ 1 - ₁(1 - _ex,1)G_R (5)

products of the i’s must always be less than 1). However, for a heat-integrated reactor, depending on the reactor gain, the loop gain can exceed unity, and this results in an open-loop unstable system.2,18 _The stability of this FEHE is determined by the poles of the characteristic equation

with

Here, we use GR(1), instead of GR, to indicate that there is only one FEHE in this system. We can also derive the exit temperature (Te) in terms of the feed temper-ature (Tf) (Figure 3A) as

Please refer to Appendix A for the derivation of this equation (a simple FEHE/reactor system).

The case of a two-FEHE system (Figure 1B) can be viewed, again, as one complete FEHE (Figures 1A and 3A) embedded in another FEHE (2). This provides a systematic procedure for modeling more complex sys-tems, because the same modeling approach (i.e., eqs 4 and 5) can be applied. The relationship between the feed and effluent of a simple FEHE is expressed in eq 8; it can be treated as a single unit in the positive feedback loop, as shown in Figure 3B. We denote this as an extended reactor model (actually a combination of a reactor and an FEHE) at level 2. This is equivalent to

deriving the relationship between Tf,1and Tout,2and the result, certainly, is the same as eq 8

The superscript (2) denotes the extended reactor model for the case of two FEHEs. Again, the characteristic equation is simply

The same principle applies to the case of three FEHEs (Figures 1A and 3A). Here, two FEHEs are embedded in the positive feedback loop, resulting in an extended reactor model at level 3. That is

The characteristic equation becomes

We can generalize the derivation to the case of n FEHEs. First, the extended reactor model at level n is

and the associated characteristic equation is

With this general formulation (i.e., eqs 13 and 14), we are able to assess the stability for complex FEHE systems.

Before leaving this section, it should be noted that the complex network in Figure 1 (e.g., 1C) actually is quite similar to the heat-integration scheme of indus-trial processes. For example, in alternative 6 of Terrill and Douglas (shown below in Figure 20B), the reboilers in Figure 1C provide heat input to the recycle column, to the stabilizer, and then to the product column. In other words, the network presented in Figure 1C is generic, and it describes the heat-integration between FEHE and process units. More importantly, a complex network can be model as a simple FEHE scheme in a repeated manner.

2.2. Controllability. 2.2.1. Overall Effectiveness. The poles and zeros play an important role in the inherent controllability of FEHE systems. The charac-teristic equation, eq 14, describes the pole locations for systems with n FEHEs, and GR(n) can be expressed explicitly in terms of n-1and GR(n-1)(eq 13). Therefore, we can go back to the previous layer and derive the stability criterion explicitly in terms of all of the i’s. Substituting GR(n-1)for GR(n), eq 13 becomes the expres-sion shown in eq 15. Going back a step further, one obtains eq 16.

Repeating this procedure, we can express GR(n) ex-plicitly in terms of GR(1) (i.e., GR and the i’s). For Figure 3. Block diagram of multiple-FEHE heat-integrated

system. 1 - ₁(1 - _ex,1)G_R(1)) 0 (6) G_R(1)) G_R (7) Te T_f) 1+ (1 - 1) 2 (1 - ex,1)GR (1) 1 - ₁(1 - _ex,1)G_R(1) (8) GR (2)₎Tout,2 T_f,1 ) 1+ (1 - ₁)2(1 - _ex,1)G_R(1) 1 - 1(1 - ex,1)GR (1) (9) 1 - ₂(1 - _ex,2)G_R(2)) 0 (10) GR (3)_{) } 2+ (1 - ₂)2(1 - _ex,2)G_R(2) 1 - 2(1 - ex,2)GR (2) (11) 1 - 3(1 - ex,3)GR (3)_{) 0} (12) G_R(n)) _n-1+(1 - n-1) 2 (1 - _ex,n-1)G_R(n-1) 1 - _n-1(1 - _ex,n-1)G_R(n-1) (13) 1 - _n(1 - _ex,n)G_R(n)) 0 (14)

example, for the case of a dual FEHE, the extended reactor model becomes

Substitution of eq 17 into the characteristic equation (eq 10) gives

Rearranging eq 18, the characteristic equation becomes

Equation 19 reveals that, as for a simple FEHE, the pole locations are determined by two separate factors: one contains all of the i’s, and the other is GR. We now define the overall effectiveness (at level 2) as

Thus, the characteristic equation for the case of a dual FEHE is simply

Table 1 gives the overall effectiveness factors for the cases of one, two, and three FEHEs (o(i), i ) 1, 2, 3). Therefore, we can assess the stability of more complex schemes by evaluating the overall effectiveness.

2.2.2. Stability Margin. In other words, we can locate the poles of complex FEHE schemes by treating them as a simple positive feedback loop provided with the corresponding overall effectiveness (Table 1).

Con-sider a reactor model that relates Tinto Tout(e.g., Figure 3A)

Here, KR is the steady-state gain, and τ is the time constant. Generally, as we try to recover more heat, the overall effectiveness increases, and the positive feedback system tends to become less stable [i.e., the poles move toward the right half-plane (RHP)2_{]. The limit of} stabil-ity can be found by solving the characteristic equation (i.e., eq 21) provided with the reactor model (eq 22)

From eq 23, it is clear that the ultimate effectiveness (u) is simply the inverse of the reactor gain [KR) GR(0) ]-Therefore, we can evaluate the stability of FEHEs at

different levels of complexity using the overall effective-ness

Note that the applicability of eq 25 is not limited to the first-order reactor model only; it is valid for any stable reactor model with nonincreasing magnitude in GR(iw). This can be proved from the Nyquist stability criterion, following Tyreus and Luyben.18 _{From the} characteristic equation (1 - o(i)GR) 0), a stable FEHE requires that the Nyquist plot of GR(iw) does not encircle GR (n)_{) } n-1+ (1 - _n-1)2(1 - _ex,n-1)

[

]

[

]

{

[

]

[

]

}

{

[

]

[

]

}

[

]

1 -1(1 - ex,1) + 2(1 - ex,2)(1 - 21)(1 - ex,1) 1 - ₁₂(1 - _ex,2) GR) 0

(19)

o

(2)₎1(1 - ex,1) + 2(1 - ex,2)(1 - 21)(1 - ex,1)

1 - ₁2(1 - ex,2) (20) 1 - 0 (2) GR) 0 (21) GR(s) ) T_out T_in ) K_R τs + 1 (22) 1 - G_R) 1 -  KR τs + 1 )τs + 1 - KR τs + 1 ) 0 (23) _u) 1 KR (24) o (i)

{

1/o(i). Because GRis stable and the magnitude of|GR -(iw)| is a nonincreasing function, the limit of stability is

If we increase o(i)further, e.g., o(i)) 1/KRin Figure 4A, GR(iw) encircles 1/o(i), and the FEHE system becomes unstable.

Instead of choosing the stability of any given design, we would like to have a quantitative measure for FEHE designs. The familiar gain margin (GM) is used for the following reasons: Most design modification is reflected in the overall effectiveness, which affects only the gain

in the characteristic equation. Given the characteristic equation and specifications on GR, the gain margin for a general FEHE is

Similarly to the familiar rules for the GM, an FEHE design is stable if the GM is greater than 1, and the FEHE becomes unstable if the GM is less than 1. This is useful in evaluating controllability for FEHEs.

Because an inverse response (wrong-way behavior) is sometimes observed in packed-bed reactors, this gener-ally results in an RHP zero in the reactor model. For example, consider the following reactor model

The model structure in eq 28 gives a nonmonotonic behavior in the magnitude of|GR(iw)|, and this might further limit the stability range (amount of heat that can be recovered) for FEHEs.10,18_{The reason for this is} that the frequency response of GR(iw) at -360° can cross the positive real axis further to the right than the steady-state gain (0°) (e.g., Figure 4C). For example, for a system with τ ) 1 and D ) 1 (eq 28), if we increase τ1 (move the RHP zero toward the origin), the limit of stability becomes 1/|GR(360)| (instead of the inverse of the magnitude when the phase is -360°), which is less than 1/KR, as shown in Figure 4. This complicates the stability analysis because we need dynamic information on GR(s) instead of GR(0) only. Figure 4 also reveals that inverse-response systems will deteriorate in the stability range only if|GR(360)| > KR.

The gain margin for the design then becomes

That is, instead of looking at steady-state value of GR, we have to find the largest magnitude of GR(iw) when it crosses the positive real axis (i.e., all phase angles are integer multiple of 360°, including n ) 0). It should be noticed that, throughout the derivation, we have Table 1. Structure Effect on Overall Effectiveness

a_{Number of FEHEs.}b_{Steam generator used.}

Figure 4. Nyquist pole location plot for a PFR reactor

wrong-way behavior. 1 o (i)) GR (0)_{) K} R (26) GM ) u o (i)) 1/K_R o (i) (27) G_R(s) )KR(-τ1s + 1)e -Ds (τ s + 1)2 (28) GM ) min n 1/|G_R(360n)| o (i) (29)

assumed that the heat exchanger dynamics is much faster than that of the reactor. This is a reasonable assumption for most FEHE/reactor systems. However, in some occasions, the heat exchanger dynamics might be significant, and all dynamics associated with FEHEs (’s) should be accounted for in the stability analysis (e.g., eq 21). Nonetheless, with the definition of the GM, we are able to quantify the degree of controllability for FEHE systems with different levels of complexity.

Up to this point, we have provided a procedure to find the pole location (or stability margin) for FEHE systems. The positive feedback system (e.g., Figure 3) shows a unique mechanism for pole-zero exchange, and this has important implications for layered FEHE structures (Figure 3). Consider the typical positive feedback system depicted in Figure 5 with GR(i)) n(i)/d(i), where n(i)is the numerator polynomial and d(i)_{is the denominator} polynomial for the generalized reactor model. The return difference transfer function becomes

Equation 30 clearly indicates that the poles of GR(i)(i.e., d(i)_{) 0) become the zeros (i.e., d}

i(i)) 0) of the positive feedback system. For the layered structure shown in Figure 3, this means that the poles of GR(1)becomes the zeros of GR(2)and the poles of GR(2)become the zeros of GR(3)_{. Thus, the sequence in arranging }

i’s and ex,i’s becomes important. For dual-FEHE system (Figure 3B), if we design a relatively large 1(i.e., 1> u) that leads to an unstable GR(2), then the system in Figure 3B will have positive poles as well as positive (RHP) zeros. This is, of course, undesirable, but it can provide useful guidelines for avoiding uncontrollable designs. 3. Implications for Design

Provided with the overall effectiveness, we are able to evaluate the stability margin for any given FEHE system. Using the GM of the design, we can assess the appropriateness of different FEHE systems. Before getting into the details, the following definitions are useful to the subsequent development:

QFEHE is the heat recovered by feed effluent heat exchangers (iin Figure 1).

QSTMis the heat recovered by steam generator (ex,i in Figure 1).

QRECis the sum of the heat recovered by the FEHEs and the reboilers (QREC) QFEHE+ QSTM).

QFUR is the heat supplied to the furnace (Figure 6). The following assumptions are also made. First, the heat-capacity flow rates are constants [(m˘ Cp)H) (m˘Cp)C]. Second, the feed temperature (Tf) and the inlet

temper-ature of the reactor (Tin) are kept constant for different designs, which implies that QREC) QFEHE+ QSTMis also constant.

3.1. Size of FEHE. For a system with fixed Tfand Tin(Figure 8), a large FEHE (i.e., QFEHE) implies a small furnace (i.e., QFUR) and vice versa. This case was first studied by Luyben et al.9 _{and analyzed by Chen and} Yu.2 _{As one tries to heat the inlet stream using} primarily the FEHE, the result is a larger QFEHEand, consequently, a larger overall effectiveness (o). There-fore, the stability margin (eq 27) decreases as one tries to recover more heat from the FEHE (QFEHE). A similar result can be obtained by inspecting the open-loop pole of the positive feedback system. For the simple first-order reactor model in eq 22, the open-loop pole is given by

Therefore, an increase in omoves the pole toward the RHP (Figure 6). This is always true irrespective of the complexity of the FEHEs, because, here, only the magnitude of the overall effectiveness matters.

Consider the following reactor model

If the FEHE accounts for 25% of the energy required to heat the reactant, one obtains

The pole of the positive feedback system is located at p ) -0.90, as shown in Figure 6. If we increase the size of the FEHE to provide 95% of the energy for the preheating, the effectiveness becomes  ) 0.85, and the system becomes open-loop unstable (i.e., p ) 1.49, Figure 6).

Figure 7A illustrates the effect of the size of the FEHE on the temperature-enthalpy (T-H) diagram for the system. This is a simplified version of the HDA example, and the numerical values for stream temperatures are given in Table 2 (cases A1-A3). As one tries to recover more heat via the FEHE, the minimum approach temperature (∆Tmin) becomes smaller, and a larger proportion of the heat is recovered, as shown in T-H diagram. Moving from left to right in Figure 7A, o becomes larger, and the pole moves toward the RHP. Figure 5. Pole and zero exchange in positive feedback system.

1 1 - iGR (i)) 1 1 - _ini (i) d_i(i) ) di (i) di (i)_{- } ini (i) (30)

Figure 6. Root location for different energy recoveries.

p )-1 + oKR τ (31) GR(s) ) 2.5 0.47s + 1 (32) _o) ₁) 200- 57 685 - 57) 0.23 (33)

3.2. Size of Steam Generator (QSTM). Consider the case in which we would like to recover additional heat from the reactor effluent, but unlike in the previous case, we integrate the heat away from the positive feedback loop, e.g., by providing steam to the reboiler of a distillation column. The flowsheets in Figure 8 show cases in which two FEHEs provide the same amount of energy for preheating (84%). For the first FEHE, the effectiveness can be computed according to eq 1 to give

Given the boilup temperature, the effectiveness of the reboiler can also be calculated as

From Table 1, the overall effectiveness becomes (for case B3 of Table 2)

Similarly, we can also compute the overall effectiveness for the case with no reboiler (case B1 of Table 2). The result is o(2)) 0.75 (Table 2), which is larger than the overall effectiveness for the case with a reboiler. Given a reactor gain of KR ) 1.36, we can compute the gain margin for both designs, and the results are

The results indicate that the system becomes stable (e.g., GM ) 1.13), if we recover heat from the reactor effluent via a steam generator. This is exactly the opposite of what we learned in the previous section, but the reason is quite obvious: we integrate the heat away from the positive feedback loop. This can also be seen from the equations in Table 1, which indicate that a larger value in ex,1or ex,2(i.e., recovery of more heat via the steam generator) leads to a smaller overall effectiveness. This also implies that, contrary to intu-ition, a complex heat-integrated plant, such as alterna-tive 6 of the HDA process, can be more controllable, at least from the stability point of view.

The T-H diagrams in Figure 7B reveal that, when the plateau becomes larger (moving from left to right in Figure 7B), the overall effectiveness becomes smaller, and the heat-integrated system becomes more control-lable (case B in Table 2).

3.3. Temperature of Steam Generation (TSTM). The previous section indicates that, if a large amount of heat is recovered via a steam generator, the system tends to become stable. Now consider the case when the same amount of heat, QSTM, is recovered, but at different temperature levels. Figure 9 shows the flowsheets where steam is generated at 110 and 338 °C. Note that Figure 7. T-H diagram for changes in (A) FEHE size, (B) STM

size, (C) size of two FEHEs.

Table 2. Stream Data for Figure 7

stream no. type supply temp (°C) target temp (°C) ∆Tmin (°C) 1 2 ex,2 overall effectiveness (o) case A A1 1 hot 685 550 493 0.23 0.23 2 cold 57 200 A2 1 hot 685 213 156 0.75 0.75 2 cold 57 529 A3 1 hot 685 152 95 0.85 0.85 2 cold 57 590 case B B1 1 hot 685 213 156 0.55 0.64 0 0.75 2 cold 57 529 3 STMa ₃₃₈b B2 1 hot 685 173 116 0.55 0.71 0.26 0.70 2 cold 57 529 3 STMa ₃₃₈c B3 1 hot 685 133 76 0.55 0.78 0.51 0.65 2 cold 57 529 3 STMa ₃₃₈d case C C1 1 hot 685 133 76 0.73 0.41 0.51 0.75 2 cold 57 529 3 STMe ₁₁₀d C2 1 hot 685 133 76 0.70 0.58 0.51 0.73 2 cold 57 529 3 STMe ₁₆₄d C3 1 hot 685 133 76 0.55 0.79 0.51 0.65 2 cold 57 529 3 STMe ₃₃₈d

a_{Steam generator used.}b_Q

STM) 0.cQSTM) 40.dQSTM) 80.

e_{Steam generator used.}

Figure 8. Heat-integrated system flowsheet (A) without steam

generator, (B) with steam generator.

_ex,2) 494- 414 494 - 338) 0.51 (35) _o(2))0.55 + 0.78[1 - (2)(0.55)](1 - 0.51) 1 - (0.55)(0.78)(1 - 0.51) ) 0.65 (36) (GM)_w/o)1/KR _o(2) ) 1/1.36 0.75 ) 0.98 (GM)w) 1/K_R o (2) ) 1/1.36_0.65 ) 1.13 (37) 1) 529 - 338 685 - 338) 0.55 (34)

the change in temperature changes the relative size of the two FEHEs. If steam is generated at a higher temperature (e.g., TSTM) 338 °C in Figure 9B), the first FEHE (1) is smaller and is followed by a larger second FEHE (2), as shown in Table 3. Considering the other case where the heat is removed at a much lower temperature, i.e., TSTM) 110 °C, the configuration is reversed to a large FEHE followed by a smaller one, as shown in Figure 9A. The results show that, if the steam is generated at a higher temperature, ois smaller, and consequently, the system is more stable (Table 3). The reason for this is that, in the flowsheet in Figure 9A, a larger portion of the heat is immediately fed back to the reactor because of the large 1value and, moreover, the energy recycled back to the reactor via the second FEHE (2) is discounted by the steam generator. On the other hand, the configuration in Figure 9B recycles a smaller portion of energy back to the reactor first (via 1) and then uses a larger FEHE after the steam generator, and this leads to a smaller o. This can also be seen from eq 20 (Table 1), where the effect of 2 on the overall effectiveness is discounted by a factor of (1 - ex,2), which is always less than 1, whereas 1 still has its original direct effect on o.

Moreover, a larger 1(e.g., Figure 9A) is also unfavor-able because of pole-zero exchange. As pointed out earlier, the pole of the inner layer becomes the zero of the outer layer, and therefore, a larger 1implies the possibility of an RHP zero. For example, consider the following reactor model

The pole (p) and zero (z) for the case of Figure 9A (case C1 in Table 2) become

and

For configuration A, we have p ) 0.63 and z ) 0.56, and for configuration B, the pole and zero are p ) 0.26 and z ) -0.10, respectively. It then becomes clear that recovering the steam at a higher temperature is always favorable and the advantage comes from considerations of stability as well as possible RHP zeroes. In other

words, any configuration that gives delayed positive feedback is always favorable.

Again, the T-H diagram can identity favorable configurations. As shown in Figure 7C, the T-H dia-gram in the far right of Figure 7C is dynamically favorable.

3.4. Heuristics. Several observations can be made from the preceding analyses:

(1) Integrate heat back to the positive feedback loop. If a larger amount of heat is recovered via the FEHE, ois larger, which implies that the system is dynami-cally less favorable (case A3 in Table 2 and Figure 7A). (2) Integrate heat away from the positive feedback loop. If heat is recovered via a steam generator (i.e., nonzero ex,i), then ois smaller, which means that the configuration is dynamically favorable (cases B2 and B3 in Table 2 and Figure 7B).

(3) Integrate the heat immediately away from the positive feedback loop. If the temperature level of the steam generator is closed to Tout, then 1is small (and 2 is relative large), which implies that the system is dynamically favorable (from the considerations of stabil-ity and possible RHP zeroes).

Because T-H diagrams are often used in the design of heat-integrated systems, the dynamically favorable configurations can also be identified, as shown in Figure 7.

These observations facilitate the assessment of the inherent controllability of any given design on the basis of the T-H diagrams (Figure 7) or process flowsheets (Figures 6, 8, and 9)

3.5. Simple Example. In the design phase, it is always helpful to reduce the number of heat exchangers, and the same principle applies to feed-effluent heat exchangers.15_{Here, we are interested in how this affects} the overall effectiveness. The original flowsheet is a system with two FEHEs, as shown in the center flowsheet of Figure 10A; Figure 10B gives the corre-sponding T-H diagram, where the horizontal line represents the reboiler of a distillation column. Two alternatives exist for reducing the number of FEHEs. One is to move the reboiler to the effluent of the reactor, as shown in the right flowsheet of Figure 10A, which results in the horizontal line being shifted to the right (Figure 10B). Figure 10C shows that the overall ef-fectiveness decreases (Table 3 gives the numerical values), because the heat was integrated away im-mediately. The second alternative moves the reboiler outside the positive feedback loop, as shown in the left flowsheet of Figure 10A. This leads to a larger value of o(case A1 of Table 3).

Figure 9. Heat-integrated system flowsheet.

G_R(s) ) 1.73 0.47s + 1 (38) p )-1 + o (2) K_R τR )-1 + (0.75)(1.73) 0.47 ) 0.63 (39) z )-1 + 1KR τR )-1 + (0.73)(1.73) 0.47 ) 0.56 (40)

Table 3. Process Stream Data for Figure 10 (Case A)

stream no. type supply temp (°C) target temp (°C) ∆T(°C)min 1 2 ex,2 overall effectiveness (o) A1 1 hot 685 133 10 0.75 0 0.77 0.752 2 cold 57 529 3 STMa ₁₁₀b A2 1 hot 685 133 64 0.73 0.42 0.51 0.746 2 cold 57 529 3 STMa ₁₁₀a A3 1 hot 685 133 64 0 0.86 0.14 0.742 2 cold 57 529 3 STMa ₁₁₀a

a_{Steam generator used.}b_Q

The results indicate that, if the reboilers are moved closer to the reactor effluent, the positive feedback system becomes more controllable. However, Figure 10D reveals that, if such a change is made, then a much larger total heat-transfer area is required, and the tradeoff between steady-state economics and dynamic controllability becomes evident. In this example, the marginal improvement in the overall effectiveness does not justify the shift of the reboiler position toward the reactor.

4. Steady-State Design Illustration

4.1. HDA Example. In a series of papers, Terrill and Douglas15-17 _{used temperature-enthalpy diagrams to} synthesize heat exchanger networks for the HDA pro-cess (Figure 11). The total annual costs (TACs) for six design alternatives are compared in ref 16, and the steady-state operability is evaluated in terms of the ranges of manipulated variables.17_{Terrill and Douglas}16

also show that, compared with the base case, the energy savings range from 29% (alternative 1) to 43% (alterna-tive 6), with a saving in TAC of 5% for alterna(alterna-tive 6. Luyben et al.9_{analyzed the heat pathway for the HDA} plant and proposed control schemes for one of the more complex flowsheets, namely, alternative 6 (Figure 11E). As pointed out earlier, the inherent dynamics of these heat-integration alternatives can be evaluated on the basis of the steady-state design data, and the open-loop pole of the positive feedback loop can be identified once the reactor dynamics is available.

Consider the T-H diagram for an HDA process alternative such as that presented for alternative 7 in the flowsheet in Figure 12A.9,15_{The hot stream starts} from the reactor effluent (685 °C) and ends at the outlet of the last FEHE (130 °C). The cold stream starts from the inlet to the FEHE (65 °C) and ends at the reactor inlet (529 °C). The horizontal lines correspond to the reboilers for the recycle column, stabilizer, and product Figure 11. HDA alternative structures: (A) alternative 1, (B) alternative 3, (C) alternative 4, (D) alternative 5, (E) alternative 6.

column. In the T-H diagram (Figure 12B), we associate a heat exchanger with each of the intervals between the break lines except for the one on the far right, which corresponds to a furnace. The circled numbers in Figure 12B stand for the feed-effluent heat exchangers num-bered from the outlet of the reactor toward the inlet stream. This base case is the completely heat-integrated system and is hereafter denoted as alternative 7. This is a very complex heat-integration scheme with four FEHEs, and more importantly, the positive energy feedback to the highly exothermic reactor can result in an open-loop unstable system.

One way to reduce the complexity is to decrease the number of exchangers. We can shift the stabilizer reboiler so that the outlet of the second FEHE drives the stabilizer and, consequently, the product column. (In this way, we can eliminate the third FEHE, as shown in the T-H diagram denoted as alternative 6 (Figure 13E). If we use the utility to drive the recycle column reboiler, another FEHE can be eliminated, as shown in the T-H diagram of Figure 13D (alternative 5). Also note that, in alternative 5, there are still two FEHEs and the first FEHE is much larger than the second one. Further modifications are to drive the stabilizer reboiler with the utility (alternative 4) or to drive the product column reboiler with the utility (alternative 3) as shown in Figure 13C and B, respec-tively. As for alternative 5, these two alternatives also have two FEHEs with a relatively large exchanger right

after the reactor effluent. Yet another alternative is to heat all three reboilers with the utility, which leaves only one FEHE, corresponding to alternative 1 (Figures 11A and 13A). Actually, alternative 1 is the original HDA process (without heat integration to distillation columns) of Terrill and Douglas16_{and Luyben et al.}9

The next question then becomes: Which of these alternatives are dynamically favorable? Certainly, the overall effectiveness provides a quantitative measure. Before getting into computation details, note that the following assumption is made: the capacity flow rates of the hot stream (Cp,H) and the cold stream (Cp,C) are the same, i.e., Cp,H ) Cp,C) 0.0378 MW °C-1.

The effectiveness of each exchanger and reboiler can be calculated according to eqs 1 and A8, respectively. We consider alternative 6 here to illustrate the proce-dure. From the inlet and outlet temperatures, the effectiveness factor of the first FEHE (1) can be obtained as

Given the boilup temperature, the effectiveness of the recycle column reboiler (ex,2) can also be calculated as

We can repeat the same procedure for all three FEHEs and both of the steam generators. Table 4 lists the numerical values for all these effectiveness factors. The Figure 12. HDA alternative structure 7: (A) flowsheet, (B) T-H

diagram.

Figure 13. T-H diagram for HDA alternative structures: (A)

alternative 1, (B) alternative 3, (C) alternative 4, (D) alternative 5, (E) alternative 6. 1) 533 - 338 685 - 338) 0.56 _ex,2) 490- 477 490 - 338) 0.09

overall effectiveness can then be computed using Table 1

The o values for all six alternatives can be obtained similarly, giving the results reported in Table 4. 5. Control System Analysis

Figure 14 shows two FEHEs coupled with an adia-batic exothermic reactor. The heat of the reaction produces a reactor effluent temperature (Tout) that is higher than the temperature of the feed stream (Tf). Therefore, heat can be recovered from the hot stream. The heat is utilized to heat the feed stream and to provide the heat input to a steam generator at a temperature of Tb1. Typically, a furnace is also required for the startup of these systems, as shown in Figure 14. The control objective is to maintain the reactor inlet temperature (Tin).

5.1. Manipulated Control Variables. There are three candidate manipulated variables for this dual-FEHE/reactor system. Because the startup furnace is available, the furnace temperature could be used as a manipulated variable. The other two manipulated in-puts come from the bypass flows of the two FEHEs. Note that mixing of the heated and bypass streams provides an immediate change in the outlet temperature (Tin,F or Tf,1), and fast dynamics can be expected. These two bypasses (b1and b2) provide two additional manipulated variables. Also note that, to provide an additional

manipulated variable, the area of the bypass heat exchanger is larger than that of the heat exchanger without the bypass. In terms of effectiveness, it can be shown that

where  is the effectiveness; the subscripts wo and w denote without and with, respectively; and b represents the bypass ratio.

Given three manipulated inputs, the controlled vari-ables are the exit temperatures of the heat exchangers and furnace (Tf,1, Tin,F, and Tin). If all three temperatures are controlled, intuitively, the pairing should be Tin -QF, Tin,F-b1, and Tf,1-b2 because of the preference of fast dynamics between manipulated and controlled variables. Figure 14 shows the cases in which one, two, and three temperatures are controlled.

5.2. Process Transfer Functions. Similarly to the overall effectiveness (o(i)), the multivariable process transfer functions can also be derived systematically, provided that the reactor, furnace, and heat exchanger dynamics are known. Because a furnace and a bypass are added to the system to provide candidate manipu-lated variables, the block diagram differs from that of the design phase (i.e., Figure 3). A simple FEHE/reactor system, shown in Figure 15 is used to illustrate the derivation. The effect of the bypass ratio b on the inlet temperature to the furnace (Tin,F) and the exit temper-ature of the hot stream Tecan be derived as

Linearization of eq 42 gives the steady-state gain between b and Tin,Fand Te(Figure 15). Figure 15 also shows that two transfer functions aare ssociated with the furnace. One is the load-transfer function describing the relationship between Tin,F and Tin (GFL), which, generally, has a gain close to unity. The second is the transfer function (GF) relates the fuel flow (QF) to the inlet temperature of the reactor (Tin). With the slightly Table 4. Effectiveness Factors of Heat Exchangers and Reboilers for HDA Alternatives

alternative 1 ex,1 2 ex,2,1 ex,2,2 3 ex,3,1 ex,3,2 4 ex4 o

QFEHE (MW) QREB (MW) QF (MW) QRECa (MW) QHUb (MW) 1 0.85 0 0 0 0 0 0 0 0 0 0.850 19.9 0.0 1.2 19.9 4.6 3 0.79 0 0.58 0.24 0 0 0 0 0 0 0.820 19.4 1.0 1.7 20.3 4.1 4 0.80 0 0.34 0.23 0 0 0 0 0 0 0.816 19.2 1.9 1.9 21.1 3.3 5 0.77 0 0.46 0.42 0.49 0 0 0 0 0 0.778 18.5 2.9 2.5 21.4 3.0 6 0.56 0 0.62 0.09 0 0.42 0.37 0.45 0 0 0.729 17.7 3.4 3.3 21.1 3.3 7 0.55 0 0.53 0.08 0 0.37 0.18 0 0.41 0.44 0.725 17.6 3.4 3.5 21.0 3.5

a_{Energy required for FEHE and distillation reboiler (Q}

FEHE) QREB+ QREC).bHot utility usage for process (QHU) QF +QREB,max

-QREB).

Figure 14. FEHE/reactor system with two feed-effluent heat

exchangers and three temperature control loops.

_o(3)) [₁(1 - _ex,1) + ₂(1 - _ex,2)(1 - 2₁)(1

-_ex,1) + ₃(1 - _ex,3)(1 - _ex,2)(1 - _ex,1)(1 - 2₂)(1 -2₁) - ₁₂₃(1 - _ex,3)(1 - _ex,1)]/[1 - ₁₂(1 - _ex,2)

-13(1 - ex,3)(1 - ex,2)(1 - 22) - 23(1 - ex,3)]

)_{{0.56(1 0) + 0.62(1 0.09)[1 2(0.56)](1} 0) + 0.42(1 0.37)(1 0.45)(1 0.09)[1 2(0.62)][1 2(0.56)] (0.56)(0.62)(0.42)(1 -0.37)(1 - 0.45)(1 - 0)}/{1 (0.56)(0.62)(1 0.09) (0.56)(0.42)(1 0.37)(1 0.45)(1 0.09)[1 2(0.62)] (0.62)(0.42)(1 -0.37)(1 - 0.45)} ) 0.73

Figure 15. Block diagram for the simple FEHE/reactor system

with a heat exchanger bypass (b).

wo) w(1 - b) (41)

modified block diagram in Figure 15 (compared to Figure 3), we can derive the process transfer function. For the simple process, we have two possible con-trolled variables, Tin and Tin,F, and two manipulated inputs, QFand b. The relationship between Tinand QF can be derived directly from the block diagram

where

This approach can be extended directly to a more complex scheme. Once the transfer functions become available, one can analyze the fault tolerance and disturbance rejection aspects of complex FEHE/reactor schemes.

5.3. Analysis. For comparison, PI controllers were used for three different control structures (Figure 14). A sequential tuning approach was used here,12_along with a +3-db maximum in the logarithm of the closed-loop modulus (Lc,max ) 3 db). Similarly to the BLT tuning method,5_{the detuning factors, f}

i, were adjusted such that the value of Lc,maxin each loop reached +3 db (when the other loops were closed). Figure 16 describes the sequential tuning procedure.

5.3.1. Disturbance Rejection. Three different con-trol structures are compared here. These three struc-tures correspond to 1× 1, 2 × 2, and 3 × 3 systems (i.e., control of one, two, and three temperatures, respectively, in Figure 14). The process transfer func-tions are simply the principal submatrices of eq 43. Here, we are interested in the closed-loop load-transfer function (GL,CL) with the corresponding controllers present.

When PI controllers are used, each Sihas a slope of +20 db/decade in the loop frequency asymptote. In the case when three temperatures are controlled, this results in a slope of +60 db/decade in the low-frequency range, compared to +20 db/decade when only Tin is controlled (see Appendix C for the derivation). This means that, in the low-frequency range, good control of Tincan be achieved if more heat exchanger exit tem-peratures (Tin,Fand/or Tf,1) are controlled. In terms of time-domain responses, Figure 17 illustrates that Tin will return to the set point more quickly if additional temperatures, e.g., Tin,Fand/or Tf,1, are controlled.

Next, the initial responses are analyzed. In the case of a simple FEHE/reactor system, Figure 15 shows that, if the furnace is used to control the reactor inlet temperature (Tin), the high-frequency asymptote of the closed-loop load-transfer function becomes

If both the furnace (QF) and the bypass (b) are used, then GL,CL(∞) can be derived from eq 43, which gives, after some algebraic manipulation

Equation 45 clearly indicates that the additional tem-perature loop lowers the high-frequency asymptote and, thus, a better initial response is obtained. Similar results can be derived for more complex FEHE/reactor schemes. Given the closed-loop load-transfer functions for all three cases, the high-frequency asymptotes for Figure 16. Sequential tuning procedure with a +3-db criterion.

[

]

[

][

]

[

]

Figure 17. Closed-loop load-transfer function between Tinand

Tfwith different control structures: control structure 1 (CS1; one

temperature loop, QF-Tin), control structure 2 (CS2; two

temper-ature loops, QF-Tinand b1-Tin,F), control structure 3 (CS3; three

temperature loops, QF-Tin, b1-Tin,F, and b2-Tf1). (Note that

temperature measurement lags are not included in computing frequency responses.)

G_L,CS1(∞) ) 1 - ₁ (44)

G_L,CS2(∞) ) (1 - 1) 1 + K2KC2τI,2

one to three temperature loops are

respectively. These results show that better initial responses can be obtained if more temperature control loops are installed (Figure 17).

Frequency-domain analyses reveal that complex heat-integration schemes provide the opportunity for more manipulated inputs and, consequently, more tempera-ture control loops. The additional temperatempera-ture control loops improve the frequency responses in both the low-and high-frequency ranges. This implies that the reactor inlet temperature (Tin) will return to the set point more quickly and also with a smaller initial peak if more temperature loops are added.

5.4. Dual FEHEs. Two examples are studied here. One is a system with two FEHEs integrated with a reactor, as shown in Figure 14. The other is the HDA example.

Two reactor dynamics are considered here. The reac-tor dynamics are taken from the examples of Tyrens and Luyben.18

5.4.1. Reactor with First-Order Dynamics. The reactor inlet and outlet temperatures are described by the following transfer function

The furnace dynamics is given by

The system has an overall effectiveness of 0.778, which gives an open-loop unstable system (GM ) 0.51). A third-order lag with a time constant of 0.1 min is applied to all temperature measurements.

Three control structures were studied. CS1 controls only one temperature (Tin); CS2 controls both Tinand Tin,F; and CS3 controls all three temperatures, Tin, Tin,F, and Tf,1(Figure 14). As expected, when more tempera-tures are controlled, better disturbance rejection can be achieved, as shown in Figure 18A. The initial peak in Tinwas reduced from 1.87 for CS1 to 0.59 with CS2 and further down to 0.46 using CS3. The load responses confirm the results from frequency-domain analyses. Smaller settling times are also observed for CS2 and CS3 compared to CS1. Figure 18B shows that better set-point responses can be obtained if more than one temperature is controlled (e.g., CS2 and CS3).

5.4.2. Reactor with Wrong-Way Behavior. Be-cause wrong-way behavior (inverse response) is often

observed in the packed-bed reactor, the control perfor-mance of this type of reactor dynamics was also studied. Consider an example of Tyreus and Luyben18

with furnace dynamics again given by eq 50. The overall effectiveness is 0.782, which gives an open-loop unstable system (i.e., o(1)KR> 1).

The load responses in Figure 19A show that, again, if more temperatures are controlled, better disturbance rejection can be achieved. This can be seen in the peak temperature (which decreases from 0.79 to 0.5 and then to 0.29 °C for C1-C3, respectively), as well as the settling time. CS2 and CS3 also give faster set-point responses, as shown in Figure 19B.

The dynamic response results presented in this sec-tion confirm the finding from the frequency-domain analysis. The complexity of the FEHE/reactor scheme leads to more candidates for manipulated inputs, and G_L,CS1(∞) ) (1 - 1)(1 - 2) 1 - 12(1 - ex,2) (46) G_L,CS2(∞) ) (1 - ₁)(1 - ₂)(1 - K_b1K_C2) (1 + K_b1K_C2)[1 - K_b1K_C2- ₁₂(1 - _ex,2)] (47) G_L,CS3(∞) ) (1 - ₁)(1 - ₂)(1 - K_b1K_C2) (1 + K_b1K_C2)[(1 - K_b1K_C2)(1 + K_b2K_C3) - ₁₂(1 - _ex,2)] (48) GR) K_Re-Ds τRs + 1 (49) GF) K_F τFs + 1 (50)

Figure 18. Closed-loop performances of CS1, CS2, and CS3 for a

first-order plus dead time reactor dynamic under: (A) +10 °C feed temperature change, (B) +10 °C set-point change in Tin.

G_R)KR(-τ2s + 1) e

-Ds

this results in better disturbance rejection while main-taining good robustness.

5.5. HDA Example. Alternatives 1 and 6 of the HDA flowsheets are studied here.9,15_{The reactor geometry} is taken from Luyben et al.,9_{and corresponds to a length} of 57 ft and a diameter of 9.53 ft with a feed flow rate of 4382.5 lb‚mol/h. Two reactions take place in the reactor

with the following reaction kinetics

The effectiveness factors of the heat exchangers and

corresponding temperatures are reported in Table 5 (and shown in Figure 20). In nonlinear simulations, 10 lumps were used to model the distributed parameter system. The heat exchanger dynamics were assumed to be instantaneous. At nominal steady-state, the linear-ized reactor model becomes

Also, recall that overall effectiveness factors for alternatives 1 and 6 are 0.85 and 0.73, respectively. Both heat-integrated schemes are open-loop unstable, and the GM values are 0.47 and 0.55, respectively. Two temperature control loops are used in alternative 1, and four temperature control loops are used in alternative 6, as shown in Figure 20. Following the sequential tuning procedure (Figure 16), the tuning constants for alternatives 1 and 6 are listed in Table 6. Figure 21 shows the closed-loop load-transfer functions (GL,CL) for both schemes. As pointed out earlier, the slopes of the low-frequency asymptotes are +20 and +80 db/decade for alternatives 1 and 6, respectively. Alternative 6 also gives better high-frequency responses. (Because mea-surement lags are added to the temperature measure-ment, the magnitudes approach zero at high frequency.) The frequency responses for the HDA process are consistent with our earlier findings in which better disturbance rejection can be expected for alternative 6. Time-domain simulations (Figure 22) reveal that much better load rejection is obtained using the more Figure 19. Closed-loop performances of CS1, CS2, and CS3 for

wrong-way reactor dynamics under: (A) +10 °C feed temperature change, (B) +10 °C set-point change in Tin.

Figure 20. Control structures for HDA design alternatives.

GR(s) ) Tout(s) T_in(s) ) 2.5 1.3s + 1e -0.5s (56) toluene + H2f benzene + CH4 (52) 2benzene T diphenyl + H2 (53) r1) 3.6858 × 10 6 exp(-25 616/T)pTpH 1/2 (54) r2) 5.987 × 10 4 exp(-25 616/T)pH 2 - 2.553 × 105exp(-25 616/T)p_Dp_H (55)

complex heat-integration scheme, alternative 6. The peak in Tinis reduced by 79% from 0.96 to 0.2 °C) and the settling time (to reach (0.05 °C of the set point) is also reduced by 0.9% compared to the responses of alternative 1. However, little difference in the set-point response is observed (Figure 22B).

Again, time-domain and frequency-domain analyses confirm that better load responses can be achieved using the more complex heat-integration scheme.

At the design phase, alternative 6 gives a slightly better gain margin in design (GM ) 0.47), and at the control phase, it gives better load rejection as a result of the increased number of temperature control loops (giving a steeper slope at low frequency and lower peak at high frequency).

6. Conclusion

In this work, a systematic approach is proposed to model complex FEHE schemes. A simple measure, the overall effectiveness, can be derived directly from the flowsheet. Given the reactor model, the controllability of a particular flowsheet can be evaluated on the basis of the gain margin (GM) of design. An interpretation based on the heat-generation and heat-removal curves is also given. With the controllability measure, implica-tions for design are also explored. Because a loss of controllability comes from the positive feedback loop, several design parameters are studied, and design heuristics are proposed to give more controllable heat-integration schemes. Two examples, a simple two-FEHE example and an HDA example, are used to illustrate the assessment of controllability based on process

flowsheets. The results show that, contrary to expecta-tions, some complex heat-integrated reactor design alternatives (e.g., alternatives 6 and 7 of the HDA example) are indeed more controllable than some of the simpler heat-integration schemes (e.g., alternative 1). For the control part, because of the availability of more manipulated inputs, e.g., the bypass ratio for the FEHEs, the system naturally leads to a multivariable control system. Contrary to one’s intuition,

multivari-Table 5. Steady-State Process Data for FEHE/Reactor Systems with Dual FEHEs18_{and Dynamic Parameters}a

reactor dynamics 1 2 ex,2 KR

τ1 (min) τ (min) D (min) τF (min) τM (min) 1∆T1 2∆T2 b1 b2 first-order (A) 0.77 0.46 0.70 2.5 0 10 1.6 1 0.1 -442.8 -34.5 0.15 0.15 wrong-way behavior (B) 0.80 0.80 0.70 2 0.5 1 1 0.5 0.1 -8.964 -8.964 0.10 0.10 a_K F) 1.

Figure 21. Closed-loop load-transfer functions for HDA design

alternatives 1 and 6.

Table 6. Controller Parameters for HDA Design Alternatives

temperature

loop 1 temperatureloop 2 temperatureloop 3 temperatureloop 4 alternative KC1 τI1 KC2 τI2 KC3 τI3 KC4 τI4

1 6.41 2.04 -0.0029 0.72 - - - -6 6.41 2.53 -0.017 0.70 -0.037 0.51 -0.038 0.49

Figure 22. Closed-loop performances for alternatives 1 and 6 of

the HDA process under: (A) +10 °C feed temperature change, (B) +10 °C set-point change in Tin.

able control of FEHE/reactor systems leads to a steeper slope and lower peak value in the load-transfer function, and this results in improved disturbance rejection. Acknowledgment

Financial support provided by the National Science Council of Taiwan is gratefully acknowledged. We also thank Professor W. D. Seider for insightful comments. Appendix A

To exchange sensible and latent heat between two streams, the design equation for a typical shell-and-tube exchanger becomes

where Q is the heat transferred; T1and T2are the hot-stream inlet and outlet temperatures, respectively; T3 is the cold-stream (steam-generator) temperature; (m˘ Cp)H is the hot-stream heat-capacity flow rate; (m˘ Cp)Cis the cold-stream heat-capacity flow rate; UA is the overall heat-transfer coefficient into the heat-transfer area; and λ is the latent heat of the cold stream.

Rearranging eqs A1 and A2, we have

Using the exponential operation for eq A3 gives

Letting B ) exp[UA/(m˘ Cp)H], we obtain

Rearranging eq A5 then gives

Now, define the modified effectiveness of a reboiler as

Then, eq A6 becomes

Finally, we can find the relationship between the effectiveness of the reboiler and the steam temperature as

Appendix B

From the definition of the effectiveness of a heat exchanger, we have

and

From the definition of the effectiveness of a reboiler, we have

Substituting eq B3 into eq B1 gives

Here, the cold-stream temperature for the reboiler remains constant. Hence, we can cancel the term 1ex,1Ts,1from the RHS of eq B4. In addition, because we know that the reactor transfer function is GR (describing the dynamic response between Tinand Tout), eq B4 becomes

Rearranging eq B5, we obtain

From eq B2, we know the relationship between Teand Tout,1. Substituting eq B3 into eq B2 gives

Here, the relationship between Tinand Toutis given by GR. Again, we obtain

From eq B6, we obtain the relationship between Tinand Tf. Substituting eq B6 into eq B8 then gives

Then

Appendix C

It is assumed that integral action (e.g., a PI controller) is employed in all three temperature loops. From the definition, the sensitivity function of the first loop (QF -Tin) is

where GFis the process transfer function and K1is the controller transfer function. At low frequency, the integrator (1/s) in the denominator of the loop transfer function (GFK1) makes GFK1dominant. Taking the limit, we have Q ) (m˘ Cp)H(T1- T2) ) (m˘ Cp)Cλ (A1) ) UA(T1- T3) - (T2- T3) lnT1- T3 T₂- T₃ (A2) lnT1- T3 T2- T3 ) UA (m˘ Cp)H (A3) T1- T3 T2- T3 ) exp

[

]

(

)

T_out,1) (1 - _ex,1)T_out+ _ex,1T_s,1 (B3)

Tin) (1 - 1)Tf+ 1(1 - ex,1)Tout+ 1ex,1TS,1 (B4)

T_in) (1 - ₁)T_f+ ₁(1 - _ex,1)G_RT_in (B5) Tin) (1 - 1) 1 - ₁(1 - _ex,1)G_RTf (B6) Te) 1Tf+ (1 - 1)(1 - ex,1)Tout (B7) T_e) ₁T_f+ (1 - ₁)(1 - _ex,1)G_RT_in (B8) Te) 1Tf+ (1 - ₁)2(1 - _ex,1)G_R 1 - ₁(1 - _ex,1)G_R Tf (B9) Te Tf ) ₁+(1 - 1) 2 (1 - ex,1)GR 1 - 1(1 - ex,1)GR (B10) S₁) 1 1 + GFK1 ) 1 1 + KF τFs + 1 K_C1

(

)

From the block diagram, the closed-loop load-transfer function (from Tfto Tin) becomes (Table 1)

Substituting eq C2 into eq C3 and taking the limit, one obtains

Therefore, in the low-frequency range, the magnitude of the closed-loop load-transfer function can be ex-pressed as

Taking the logarithm and using decibel (db) as the unit, we have

This gives a straight line with a slope of +20 db/decade. Similarly, the closed-loop load-transfer function for control structure CS2 (from Tfto Tin) is also shown in Table 1. Taking the limit, we have

Again, taking the logarithm, we have

This is a straight line with a slope of +40 db/decade. A similar result can be derived for the CS3 structure

Equation C9 reveals that the magnitude plot at low frequency is a straight line with a slope of +60 db/ decade.

Literature Cited

(1) Bildea, C. S.; Dimian, A. C. Stability and Multiplicity Approach to the Design of Heat-Integrated PFR. AIChE J. 1998,

44, 2703.

(2) Chen, Y. H.; Yu, C. C. Interaction between Thermal Ef-ficiency and Dynamic Controllability for Heat-Integrated Reactors.

Comput. Chem. Eng. 2000, 24, 1077.

(3) Douglas, J. M.; Orcutt, J. C.; Berthiaume, P. W. Design and Control of Feed-Effluent Exchanger-Reactor Systems. Ind. Eng.

Chem. Fundam. 1962, 1, 253.

(4) Kays, W. M.; London, A. L. Compact Heat Exchangers; McGraw-Hill: New York, 1984.

(5) Luyben, W. L. Simple Method for Tuning SISO Controllers in Multivariable Systems. Ind. Eng. Chem. Process Des. Dev. 1986,

25, 654.

(6) Luyben, W. L. External versus Internal Open-Loop Unstable Processes. Ind. Eng. Chem. Res. 1998, 37, 2713.

(7) Luyben, W. L. Effect of Kinetic, Design, and Operating Parameters on Reactor Gain. Ind. Eng. Chem. Res. 2000, 39, 2384. (8) Luyben, W. L.; Luyben, M. L. Essentials of Process Control; McGraw-Hill: New York, 1997.

(9) Luyben, W. L.; Tyreus, B. D.; Luyben, W. L. Plantwide

Process Control; McGraw-Hill: New York, 1999.

(10) Morud, J. C.; Skogestad, S. Dynamic Behavior of Inte-grated Plants. J. Process Control 1996, 6, 145.

(11) Polley, G. T.; Heggs, P. J. Don’t Let the Pinch Pinch You.

Chem. Eng. Prog. 1999, 95 (12), 27.

(12) Shen, S. H.; Yu, C. C. Use of Relay-Feedback Test for Automatic Tuning of Multivariable Systems. AIChE J. 1994, 40, 627.

(13) Seider, W. D.; Seader, J. D.; Lewin, D. R. Process Design

Principles: Synthesis, Analysis, and Evaluation; John Wiley: New

York, 1999.

(14) Silverstein, J. L.; Shinnar, R. Effect of Design on the Stability and Control of Fixed Bed Catalytic Reactors with Heat Feedback. 1. Concept. Ind. Eng. Chem. Process Des. Dev. 1982,

21, 241.

(15) Terrill, D. L.; Douglas, J. M. A T-H Method for Heat-Exchanger Network Synthesis. Ind. Eng. Chem. Res. 1987, 26, 175. (16) Terrill, D. L.; Douglas, J. M. Heat-Exchanger Network Analysis. 1. Optimization. Ind. Eng. Chem. Res. 1987, 26, 685.

(17) Terrill, D. L.; Douglas, J. M. Heat-Exchanger Network Analysis. 2. Steady-State Operability Evaluation. Ind. Eng. Chem.

Res. 1987, 26, 691.

(18) Tyreus, B. D.; Luyben, W. L. Unusual Dynamics of a Reactor/Preheater Process with Deadtime, Inverse Response and Open-Loop Instability. J. Process Control 1993, 3, 241.

(19) Yan, Q. Z.; Yang, Y. H.; Huang, Y. L. Cost-Effective Bypass Design of Highly Controllable Heat-Exchanger Networks. AIChE

J. 2001, 47, 2253.

Received for review April 30, 2002 Revised manuscript received December 3, 2002 Accepted March 19, 2003 IE020324R lim sf0S1) limsf0 1 1 + G_FK₁) limsf0 1 G_FK₁) limsf0 τI1s K_FK_C1) 0 (C2) GL,CS1(s) ) (1 - ₁)(1 - ₂) 1 - ₁₂(1 - _ex,2) S₁ 1 - o (2) S1GR (C3) lim sf0GL,CS1(s) ) limsf0

[

]

[

]

[

]

[

]

[

]