PT. Universfty of New Hampshire Lae ) hs Gat A MICS i SORA ASUDEVAN, feajra lle Durkan}, NH 03524 “yp epeess Control and Simulation is a 4-credit fc in the department of chemical engi Classe supple: tory se: tionall year. T namie describpd by differential equations, feed’ concept vaneed| Studi three, poring at the University of New Hampshire. m lectures (three hours per week) are ented with a two-and-a-half hour labo sion held once a week. The course is tradi- taught in the spring semester of the senior he topies include, but are not limited to, dy- havior of chemical enszineering processes control and techniques, stability analysis, and ad- rontrol techniques, its are usually divided into groups of two or ind each group is required to do six different experinfateover the course of the semester. These experi ‘ents are designed to expose the students to the prijetical aspects of almost all the theoretical topics equipay studen| the exp to colle and su| design various order, 13s overed in class. The basic materials and jent are supplied for all the experimonts. The js have to assist in designing and building priments, decide a priori what data they want t, perform the experiments, analyze the data, ymit a report. Some of the experiments are dd to pormit flexibitity in torms of simulating, process configurations (first order, second fhird order) or to demonstrate various pro- trol principles discussed in class, ‘The im- zr, vasudevan's chorea! enginserng tac Gy membor atthe Uneasy of ow Har [hve Ha osainod his PRD rs Ciaran Unk trsiy Us reseweh intros orem ta rea ot Nas ln b Cob ee [oa|q0 z eontblte portant features of each experiment will bet lighted in the following paragraphs. ‘The six experiments ave divided into two phases: ‘The first three experiments (phase one) are pe: formed by all the students prior to the spring bres while the remaining three experiments (phase two) are performed after the spring recess. EXPERIMENTS ‘Two of the three experiments in the first phase deal with determination of time constants of processés such as liquid level in a tank or pressure in an air eylinder. ‘The liquid-lovel process consists of three Plexiglass tanks with intorconneeting valves. Water can be pumped to any of the three tanks, and the feed water pressure is maintained co; avoid any Muctuations in the flow rate. The control valve is located on the food line. The students are at Kberty to select one, two, or all three tanks and sot the system up as cither an interacting or & noninteracting process. The level is monitored in the third tank by means of a pressure transducer mounted at a height of six inches above the tank bottom, and the signal from the transducer sent toa PC equipped with data-acquisition bilities (in this case, a Metrabyte DAS-8 card). Labtech Notebook is used to set up the various in- put and output channels. The PC is equipped with additional Metrabyte boards for process control. The students are thus exposed to various features of data acquisition and control, and instrumentation hardware very early in the semester. In order to determine the time constant of the process, the students have to use both a pulse- and a step-forcing function, These forcing functions are set up in an external file (in ASCID and ean be accessed by Labtech Notebook when necded. ‘Thus, data pertaining to the magnitude of the step or pulse and the type of pulse are stored in this external file. Chemical Bngiacering Education a.‘the eostroller (output channel) in open loop reads the infrmation from this file and changes the out- putto the control valve accordingly. The duration of the puke or the exact mament at which the step ehangefis to be introduced is controlled by adjusting the sanppling rate in the output channel. 'The entire operatipn is therefore carried out in a precise fash- fon, with yery little human intervention. The data are recorded ina file and are also con- tinuonfly monitored on the VGA monitor, so the studen}s can compare the experimentally obtained value fpr time constant with the theoretical vabu (knowihig the valve resistance and area of the tank), A Jinedr valve is used in the experiment, and the valve desistance can be easily determined experi- Jnentadiy, The deternination of time constant for a step-fo} cing fimetion is straightforward. For the ease of a pdlse-foreing function, the following method is, used fer determining the time constant for a first order flrocess (single tank). Welt = linear valve resistance ii = magnitude of the pulse = duration of the pulse hy(q) = height in the tank in terms of deviation variables: ‘Then, }n the Laplace domain a In the|time domain, for t > T, byl0= RH(e Pe") @ We fow define a function, f, equal to the product of h(t) and time, t. A plot of f versus Lime t will go through a maximum. By differentiating Bq. (2) with respedt to time and equating it to zero, we can show that the maximum occurs when t = 1. Thus, this methdd gives a simple procedure for estimating the time donstant for a first-order process. Alternately, ‘a plot of cn h,(t) vs. time is a straight line with a slope fqual to the reciprocal of t. However, Raq. (2) does Hot take into consideration the response of the proces for values of t < ‘P. If the deration of the pulse js sufficiontly long, it is necessary to consider the cdmplete solution. ‘Thif problem is easily solved in the following man- ner. If is possible to delay the storage of information by spheifying a time delay cquat to the duration of ¥y setting up a “calculated channel” it js thekefore possible to monitor and store time in an exterhal file as (L-'f), (referred to hereafter as ted time"), In the next channel, the data are the pplse, er 1908 stored or displayed as the product of height (in de- viatipn variables, also easily sct up in. notebook through the use of “ealeuluted channels" once the initial steady-state height is known), and adjusted tie. The produet of height and adjusted time (Fan tion f) versus the adjusted tine is continuously di played on the screen (and alsa stored in an ext file) s0 that the information can he plotted later on. Such a plot is shown in Figure 1, (Since the sam- pling rate is 1 He, the data points are not shown.) From this plot the'time constant can be determined as the value on the abseissn corresponding to the value on the ordinate where the function f gues through a maximum. Or, to obtain an accurate esti- inate, a differential analysis of the data (function f swith respect to time) can be performed. ‘The value of thé time constant from the plot is abont 210 seconds. ‘Phis compares very well (within 5%) with the value of time constant obtained using: f@ step change. Por this particular experiment, the duration of the pulse was 100 seconds and the may nitude of the pulse (change in low rate) was 0.2 fishnin, It is interesting to note that the valve resis- tance can easily be determined once the time con- stant for the process is known, Setting t= tis F (2), we get na()= £ 0.968 281(67"*—1) ‘The only unknown in this equation is R, and it can be determined, ‘The same experimental setup is used to introduce concepts such as transmitter gain and dead time. For instance, since a pressure transducer is used to, measure the height in the tank or the pressure in the eylinder (in the air-pressure process exper- iment), the students are required to calculate the fon of tin constant from Tie se 1. Dotermin «a pilse test. Sasnpling ratetransfitter gain, This information is then entered jnio ihfe input channel. The students ¢ nin an uaderptunding of how a transducer warks and the rangelof the output signal for electrics! and pnew ‘Thelfinst fificen to twenty minutes of each Ta! tory yeriod is spent in demonstrating the proce contrdl principles discussed in la week For example, the liquid-le used [a demonstrate types of coatrotier action and how th set up the right action (rever the ofitput channel. The PO is equipped with a Metrdoyte DDA-06 controller card. The phenomenon of rept windup and the concept of stability a demofstraied soon after the theoretical material is prosehted in lass. Since Labtech Notebook uses the positibn form of the controller equation, reset windup is derhonstrated very effectively. s the previo or direet) in ‘Thd students also develop a good understanding of tht dynamics of PID control and the effect each elem}nt (P, I, D) has on the overall control process Impeftant concepts such as offset, oz how a simple first-peder process with PI control can behave in an oscillatory manner, or how a second-order overdamped process with sin sportional con~ trol fan hocome underdamped, are demonstrated witha ‘Thh pressure experiment is similar to the liquid evel) experiment, Students are required to deter- mind both the experimental and the theoretical time cons{ants and to compare the two. They must deter ming the transmitter gain (scale factor) and offset and fot up various channels in the Notebooks. ‘Tif: third experiment deals with control-valve eali- bratlon for both liquid and gas service. Here the students gain a practical understanding of eoneepts suet] as inherent and installed characteristies, valve cocfficient, and valve flow characteristics. Once again, Notebook is used to set up various channels. ‘A Metrabyte DAC-02 card is used to change the signhl to the transducers located on the control valves for foth liquid and gas, in inerements of 1V (range is -4 to +5V). The students take data of flow rale, vale stem position, current signal to transducer (4 20 ha), upstream pressure, and downstream pres suré. They are required to calculate and report the vale coefficients of the two valves as well as the typ} of valve (linear, equal percentage, quick open- ing] from suitable plots of the valve characteristics. In {heir report, the students are required to com- moft on the phenomenon of hysterisis obsoryed in a plof of valve coefficient versus valve stem-position, ‘The second phase of the lnboratory deals with eon- 186 le pe troller tuntrg based on Zieyl seitings, Cohen: Coot dota frou a pulse test, The liquid-tev tal sotup or the air-pressare protes Tn thn case of the Tiquid-k it, th vation they tik stude From the p reaction eueve generated (hers aytin, Lnbtoch Notebooks is used to set-up the char ls and store the infoewation}, the to determine th cia controller (P, PA, or PED) and ose to both serve and toad from the tings. determine the t changos. ‘The controller set reaction curve serve as preliminary est mates, and the students the optimum settings using a dynamic criterion such us IAB, ISH, or ITAR. ‘This is easily done Hhrough the use of varions “calculated channels” of Notebook, and IAB, ISK, and ITAB are set up in different channels. re required to obtain ‘The display window for the monitor is divided into four sections, and the students can observe the ai tual height in the tank, the error, IAB, ISH, or TTAR They are required to select one of the integral crits ria and try to obtain the optimum controller set tings. This is done by keeping the reset time eo stanly for instance, and changing the proportion gain and determining the response to a unit step change in the set point (always from the same value’. ‘The students then change the integral time (keep: ing the gain constant) and observe the response. In ‘each case the integral value is reported. ‘The second experiment also deals with controller tuning, This is done using the Ziegler Nichols elosed- loop tuning method. The second half of this experi- ment consists of using a pulse test to genorate a Bode plot. The objective of the experiment is to de- termine the open-loop transfer function and caleu- late the overall gain, time constant, and dead time, if any. The students have to decide on a proper pulse duration and magnitude. ‘The pulse is introduced by changing the position of the control valve, and honce the flow rate to the system, for a known duration. This is achieved by setting the output channct in “open loop,” which i turn accessos an external file to obtain values of the controllor output. Care is taken to ensure that the system returns to its original steady state, ‘The in- put and output data are then Fourier-transformed and divided to give the system transfer function in the frequency domain, Gliw). From the amplitude ratio and phaso angle, Bode plots are constructed and the various parameters determined. The calen- Chemical Rngincering EducationMagnitude Ratio Frequeney, Fad/sec Figure 2) Bode plot generated from a pulse Test: nade ratio versus frequen Bice Frequency, Figure 3} Rode plot generated from a pulse test:|phase angle versus frequency. Figure 4, Bode plot generated froma pulse test: second-order system, Sumner IBS lation of Gaw) from the pulse data is achieved by representing the transfer funetion ae Fu tipeostwt)ae-ify(U)sin (wtyat 0h Be Jali joos(oet} de Ff (sin (wre where x(t) and y(t) are the input and output functions, ‘Then cw) {ABD HAD—BE) CD Fotos ns Fre eincooat c= Pueeospna p= facaeingwtat ‘The duration of the inplut pulse and the time it takes the response to return to the original steady state, are ‘f, and Ty, yespectively. ‘The integrals are evaluated numerically by pick- ing different values for the frequency, sinframe computer nfier up-londing the data from the PC mainframe. The experiment yields reasonably accurate frequency response curves, Numerical integration beconies a problem beeause of the oscillatory behavior of the sine and ‘cosine terms at high values of frequency. Since there is practically no human input necessary white performing this experiment, and because of the resolution and sampling rate used, data notse is not a problem, The Bode plots generated for a first-order liquid-level process are shown in Figures 2 and 3. ‘The magnitude ratio and phase angle tt higher values of frequency are not shown because of the problems associated with integration. Figure 3 indicates that the phase angle reaches an asymptotic value around -90°, which is indicative of a first-order system without dead time. Tt js also evident from Figure 2 that the transfer funetion of the system is exactly first order. ‘These observations are not surprising considering the fact that the process is first order and there is no measurement lag. ‘The time constant for the process can be easily determined from Figure 2 once the corner frequency is known. ‘The time constant is found to be about 200 seconds and is within 3% of the value previonsly reported, "Phe magnitude ratio for ng tanks) is shown in high-frequency asymptot sgond-ordler process (two interact ure 4, Fis clear from the slope of the that the system is exaetly second order, Laltech Notebook also has a Fast Fourier Transferin (FFT) capability that ean be vsed to generate a power spectrum, In the above experiments, the students are elso required to study the effect of sampling rate on data sequisition and on the vontral characteristics, Continued om page 198technicallbackground. Although the oral and writ- ten repo s are addressed to a technical audience, when wolking individually with the CI the students must ex] dience. I derstand(ng of the material communi Finally, cation tr: for as m separate time to r¢quire helpful reading mater ing and ‘ess technical idens to. non-technical au- is actually helps to develop a better un- nd is a challenging ative exercise in itself. we recognize that integrating communi- ining into existing courses does not allow ch instruction as could be offered in a not enough on speake iting, or to evaluate and discuss pub- ommunication course. There lished arfieles, or to offer workshops on writing and speaking, Many students would benefit from more intense i articularly on technical writ- ing, But] acknowledging that good communication skills arf never “learned” once and for all, we feel that by providing some limited instruction and sig- nificant yractice and evaluation, we are at least help- ing studi approact We also cation campus ts to improve their skills. As one student his writing improved partly “because (he ally writing for a change.” An integrative is certainly a step in the right direction. sill encourage students to take communi- jurses outside the department and to use ‘esourees such as the "Writer's Workshop,” tutorial conter sponsored by the Center for Writihg Studies. As we \work to provide our students with better communjcation skills, we must remember that, de- veloping|expertise in writing and speaking is a life- 23s, Integrating communication training ing chemical engineering courses may not he extenbive enough for some students, but it does provide speakiny significant smount of practice in both and writing, leaving students with some professifnal experience and, hopefully, with an awarend ACKNO} The co was dev) 8 of the value of communication. HLEDGMENT smnunication work reported in this article Joped by a number of individuals in addi- tion to te authors: Dr. Charles A. Eckert, who initi- ated ou tion wid Edward] Lauffen! K. Dan 1. Can} Suremer emphasis on communication in collabora- h Marsha Bryant and Wayne Howell; Drs. hv, Funk, Thomas J. Hanratty, Douglas A. jurger, Richard f. Masel, Mark A. Stadtherr, Wittrup, and Charles F. Zokoski, who all .¢ courses; and Dr. Ruth Yonts, forsner Com- jon Instructor, NCES fina, Claudia M. “More ChE De eations Shite Chem. Eng. Pe 299 ents Stress Come 87, 1000990 Gicsslmnn, Hebert D, and Ruth A. Yontz, in "Writing Across the Curriculoin The Case of Engineering.” ASEIE Spring Conf. Papers, Hinoisl nig, Janet, “Writing as a Mode of Learning,” The Web of Menuing: Besays f, Tracking, end Thinking, Baynton/Cook, Po Grifin, CW, “Using, Weiting to ‘Teach Many Disciplines” Inprovings Call. an Unie, Troehing, 84, 121 (2983) Deli, Airy Heth, "Collab nival Writings Theory and Practice, Bertie E, Fearing, and W. Keats Sparrow, cde, MLA, New York, 33 (1989) 6, Ferman, dain and Prive etsy The Grp Repurt A Prbte in Sl Croup or Writing Proceso Conny 23,28 (1888) 7. Some cf the idens for this checklist were adapted frm i, David G. and Denatd Zimmerman, The anton ‘aus! Guide to Technical and Sttentne Communication, Ranson Hause, New York, Chap. 11582) 1 Process Control Lab Course Continued from page 187. ‘The Inst experiment in the sccond phi “Hardware. the features of Meirabyte ears such 02, DDA-06, PIO-12, and to ois ealled Init, the students are required to study DAS-8, DAC ‘uequi- sition system for monitoring temperature in six poly- mer renetors i different initiators or different initiator concentrations. A multiplexer board (Metrabyte EXP-16) is used to conncet the different, thermocouples. The students thus learn about mul- tiplexers, thermocouples (how the cold junction is set up on the EXP-16), A/D converters, D/A convert- ers, electro-pneumatic transducers, and other im- portant features in data acquisition and digital con- trol. The reaction is then started, and the students monitor the temperature change in each reactor si- multaneously. The students study the effect of chang ing sampling rate on data acquisition since si ferent temperat dif. -08 are monitored simultaneously. CONCLUSIONS "These six Jaboratory experiments are an e! supplement to classroom lectures. St nts gain hands-on experience in controller tuning, data 2e- cepts are emph thorough under ‘the concepts. The laboratory all the topi vanced control strategies such trol or ca: puter simulation few of these adv ind control. Various process eontrol con- ed, and the students develop a ding of the practieat meaning of jons cover almiest cept certain ad- feedforward con- ome of the available com- ce nsed to illustrate a egies. Interested discussed in clacs ade control. S kages need control readers may obtain complete information on the equipment or writeups of the experiment by con- tacti the author. 0 198