.tA.-5-3._ta_b_o_r._a_t_o_r.:.y________ ) AN AUTOMATED DISTILLATION COLUMN For the Unit Operations Laboratory DOUGLAS M. PERKINS, DAVID A. BRUCE, CHARLES H. GOODING, JUSTIN T. BUTLER Clemson University • Clemson SC 29634 istillation is one of the most common separation pro­ variable using software such as Control Station.DJ Thus, the cesses; hence, it is important for undergraduates to senior-level UO Lab course incorporates process control con­ D have some hands-on exposure to this unit operation cepts into several of the classical unit operations so that stu­ during the course of their studies. Additionally, working with dents gain hands-on experience working with and tuning con­ batch distillation towers provides students with an opportu­ trollers in automated chemical processes. Ultimately, this nity to experience and learn about a dynamic process that is better prepares them for work with complex, real-world pro­ widely used in the pharmaceutical and specialty chemical in­ cesses than would conducting experiments with idealized pro­ dustries. It is equally important for undergraduate students to cesses, such as simple tanks in series. work with automated processes, since similar control features are commonly implemented in industry to reduce labor costs, EQUIPMENT provide greater processing flexibility, and improve process Though the basic design for the batch distillation appara­ safety and product purity. tus was developed by Clemson faculty, the detailed design For these reasons, an automated batch distillation column and most of the construction was accomplished by under­ was designed and constructed on-site for the Clemson Unit graduates involved in the projectYl This afforded the stu­ Operations Laboratory (UO Lab). The pilot-scale tower uses dents an opportunity to gain hands-on knowledge about metal sieve plates to separate a mixture of 2-propanol (IPA) and 1- machining and process engineering. The apparatus required propanol (NPA). Some background information about the approximately six months and $25,000 to build. The key com­ Clemson curriculum should be mentioned before going into ponents of the apparatus (shown in Figure 1) include more depth about this particular experiment. First, the se­ • A 180-liter jacketed vessel (Owens Mechanical & Fabrica- nior-level UO Lab course consists of groups of students (typi­ cally three to four students per group) conducting four ex­ Douglas M. Perkins is a recent chemical engineering graduate from periments during the course of the semester. Lab groups de­ Clemson University During his undergraduate career he worked closely velop their own experimental procedures to accomplish the with Dr. Bruce on designing and building multiple experiments for the Unit Operations Laboratory as well as being the lead student designer/ assigned objectives. They are given three three-hour lab pe­ builder of the batch distillation column. riods to conduct each experiment, and the results of these David A. Bruce is Associate Professor in the Department of Chemical Engineering at Clemson University He earned his BS degrees in Chem­ experiments are presented either in writing or orally in front istry and Chemical Engineering and his PhD in Inorganic Chemistry of a panel of their peers and professors. from Georgia Institute of Technology His research interests include heterogeneous catalysts with controlled pore geometries. advanced Having three lab periods to perform each experiment pro­ oxidation processes. and quantum and molecular mechanics model­ vides students with the ability to explore different aspects of ing. Charles H. Gooding is Professor of Chemical Engineering at Clemson a particular process and to collect enough data to explore sta­ University He earned his BS and MS degrees at Clemson and his PhD tistical variations. Additionally, the process control course, at North Carolina State University, all in chemical engineering. His teach­ ing and research interests are in chemical process design. analysis. while providing students with both the practical aspects and and control. fundamental mathematical principles of control, has no lab Justin T. Butler is a recent chemical engineering graduate from associated with it. Therefore, students can only simulate how Clemson University He aided in building/designing the batch column. a change in a manipulated variable affects a process control © Copyright ChE Division of ASEE 2005 104 Chemical Engineering Education tion), which serves as the reboiler The IPA/NPA system also exhibits relatively • A 4-in. diameter glass distillation tower (Lab glass) with six aluminum sieve ideal behavior as indicated by the vapor-liquid plates vertically spaced 6 in. apart equilibrium (VLE) data shown in Figure 2. This • A single-pass shell-and-tube heat exchanger using water coolant in the tube figure includes both experimentally measured side data[3l and compositions predicted by a Margules 4 • Coriolis flow meters (Micromotion CMF025) and pneumatic control valves activity coefficient model. [ l Binary parameters 4 (Fisher Rosemount 5100 valves with 3661 positioners) on both the reflux calculated by Gmehling[ l for other activity co­ and distillate lines efficient models are shown in Table 1 along with • Data acquisition and control hardware and software (National Instru- the mean deviations associated with each ments). model. As can be seen in the table, all of the activity coefficient models do an excellent job All vapor lines are 2-in NPT pipe and all liquid lines are 1/2-in NPT steel at describing the VLE data for this non-azeo­ pipe. Additional information, such as vendor addresses, a wiring diagram, tropic system. and a more detailed equipment list, are available upon request from Profes­ sor Bruce. Another key design feature was the choice to use gravity, rather than a pump, to return reflux DESIGN METHODOLOGY flow to the column. This necessitated that the pressure drop across all flow measuring/control The chemical system as well as several key components of the batch col­ devices be kept to a minimum. For this reason, umn were chosen to be similar to those previously used in another pilot­ Coriolis flow meters were chosen both for their scale distillation apparatus in the UO Lab. This was done because the previ­ accuracy and their low-pressure drop charac­ ously built column had proven to be very safe to operate and there were ter istics. An added bonus in using the Co- well-established performance characteristics, such as optimal flow rates, heat duty for the condenser, and stage efficiencies. Specifically, the IPA/NPA bi­ nary system was chosen because: 1) the chemicals are relatively inexpen­ sive; 2) they have low toxicity; 3) they have high short-term exposure limits (> 250 ppm); 4) fires can be extinguished by water, dry chemical powder, or ~ 0.7 ; i 0.6 CO2 5) they have moderate vapor pressures (< 45 torr) at STP conditions; and 6) mixture compositions are easily analyzed by gas chromatography. ~ 0.5 +--_____,,_ _ _,,_ _____-------, "'§ 0.4 +----~~----,,~-------------, jj 0.3 City T o Compressed i Water Co~uter Air ·~ 0.2 +----+~~----------------, Heat Exchanger Steam <- - ToTT }-T'"---t==========J~..L-(,-----"""-£;Ye From Compute r 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Computer Sight weight fraction 2-propanol in liquid, Xi PA Glass T o Computer Figure 2. Vapor-liquid equilibrium data for ,,-. From mixtures of 2-propanol and 1-propanol at 1 From Floor Drain Compute r Com,pute r atm. Curve corresponds to compositions pre­ dicted using the Margules parameters listed in Table 1 and (*) represent experimental data collected by Ballard and Van Winkle, 1952. TT '¥ '¥ '¥ TABLEl T o Co mpute r Column :- ---- -• VLE Binary Parameters for Activity ]I 1' Q) Coefficient Models of the E '¥ E 2-propanol/1-propanol System[4l 0 T o a: Com puter Mean TT Activity Binarr Parameters Deviation from Coefficient AIPA-NPA ANPA-IPA Experimental Model Vapor Fractions Disti llate Margules 0.1321 -0.0621 0.0046 Vau Laar 0.0100 25765.3800 0.0090 Floor Drain Sample Wilson 818.3291 -481.0590 0.0048 UNIQUAC -399.2278 575.0956 0.0050 Figure 1. Schematic for batch distillation apparatus. Spring 2005 105 riolis flow meters is that they can not only measure mass Initially, the column is started in total reflux. This is the flow rate, but also volumetric flow rate, density, and tem­ simplest mode of column operation and involves control of perature of the fluid in the pipes. only the reboiler steam pressure and liquid level in the sight The reflux and distillate control valves were selected for glass. This mode allows the rising vapor and falling liquid to their ability to control low flows with a very small pressure heat the tower. Once the column has reached steady state, the drop across the valve. At first, an actuated ball valve was overall tray efficiency can be determined from composition considered to allow for the low-pressure drop requirements, analysis of samples collected from the top and bottom of the but after talking with several vendors it became apparent that column. By varying the setpoint of the reboiler steam pres­ a ball valve would not be a viable option, due to sizing diffi­ sure, it is possible to determine how tray efficiency varies culties. The selected needle valves were chosen based on with vapor flow rate. their ability to meet the above criteria, specifically, a low­ The column can be operated in three other modes. Con­ pressure drop across the valve and integral positioners for stant reflux ratio is used least often because it requires that adjusting the needle. the distillate and reflux valves be adjusted to maintain a con­ The Labview data acquisition and control software (Na­ stant fluid level in the sight glass as well as a fixed reflux tional Instruments) was chosen because it has several key ratio. Noise and interaction make this mode difficult to tune features that make it attractive for use in the UO Lab.