Download as pdf
Download as pdf
You are on page 1of 54
Moyo secon 3 PROCESS DYNAMICS 3-1 THE IMPORTANCE OF PROCESS DYNAMICS A omer of things most be considered inthe design and application of s control techniques. The overall operating strategy of the plant, the ‘stuation, and the market position othe produc mut be considered ‘orking fom the top down 40 to speak, one arrives atte lowest of of operation and contol where indindual lows, temperature, pr uid Teves, and so on must. be contolled. Invariably, as previously 1.00 40. © er () 4.0 =(0922){1.66F 6. pogo fgg = Hime for twansmiter output = 50 sée to reach 41.425 ‘age = time for transmitter output = 80 see to reach 48.28% ‘osq,_ = Hime for transmitter output = 125 sec to reach 47.126 Using these values inthe &-step procedure gives the following = 27.290 52.2 see 9.9 sec It must be admitted that there isno right answer to a problem such as this, Furthermore, some calculation would show thatthe esulls are very sensitive to the reading of the “characteristic” times from the time response. This points out the difficulty in finding two time constants from experimental data. “The computational methods pve fa feng ao Sal erSet convenient, and accurate enough for many purpose in proces contol, How. cver, for more accurate representations, the more complicated computer regres sion technique hold promise (Rel 4, ‘What follows is an introduction to a fundamentally different approach to. the problem of analyzing process dynamics in operating plants — frequency respons methods, preemie a * out 2 Output Zo 2 Tine ‘Figure 43. Snusoil Veriton of Process input and Output What would happen fasted of astep change ninputto te process, there ‘were an np variable hat changed periodically asasine wave (Figures-3?Such {change could be due to the temperature variations that occur froméay tonight in the air surrounding the process. Or the changes might be intentionally intro> aced by manual thanging a valve positon. Whatever the source of the {might be seen afterenoughtime has passed for asteady rexponse (o_ becstablshed) thatthe output — perhaps a potential cont Figure 43). eo ae Example 44 Amplitude Ratio of «FirsOrder Process Develop agraph of the amplitude ratio ofa istorder process as function orthe put nine wave frequency (a log-leg plot is usual). The amplitude ratio, saree see only on the Frequency but ase on the gain and time constant as Amplitude Ratio = A.R, = where: = input frequency (radians/ see) K= process gain rocess time constant (4) Solution All that is required is plot of the amplitude ratio for various values of frequency. However, because numerical values are not specified for the time constant and gain, the plot will necessarily be a rather general one (Figure 4-4). ‘The plot shows that ifthe input sine wave has a zero frequency (the input isnt really period) the size ofthe output is the gain multiplied by thesize ofthe input — just as with a sep test. As the frequency ofthe sine input increases, the smplitude ratio decreases. This indicates thatthe input is attenuated by the process and the output shows less variation. At very high input frequencies the utput variations would not even be detectable because ofthe extreme atenua tion ofthe process. This isthe effet of time constants on the input; the capacity ‘ofthe process tends to filter orsmooththe upsets. The greater the capacity (larger time constant), the greater the attenuation at lower frequencies, ‘The use of frequency response methods in developing process models is based on the way the amplitude ratio plotchangesasmore and more complicated ‘Procees Dynamics From Plant et “Amplitude aio pales + 10 700 Frequency, redians/see Figure 4-4, Fa.orde Proc Solution itis necessary to plot rather general situation because no numerical values as given forthe gain or time consants (Figures 4S and 4-6) A compat rennet thet plots shows the affect ofan additional time constant on the slope roti amplitude ratios the frequency irreases,Iisthis change nslope thas eee aeapuncloping process models experimentally. The dashed lines are called ater (simply extensions ofthe fail sraight portions of the curves); the saitotesintesce ata poimt where the produto requency and time constant SRunls one a very valuable piece of information in later work in this tet Again tnisinteresting to note that ifthe process had contained dead time, it would not have affected the amplitude ratio plot at all. How then does this important ‘gnamic clement show up in fequency sponse studies? Its through the other fart of frequency response — the phase ferences between the input and output sine waves. - ‘As mentioned earlier, Oy 7 10 Frequency, radians/see Figure 45, Second-Order Process Amplitude Fatlo (Equal Time Constants) o Process Bynamies From Pan este ‘change in signal, perbent Figure 47, Phase Diference between Inputand Output Signals Example 4-6 Phase Angle of First-Order Process Develop the graph of the phase angle plotted against the frequency ofthe input sine wave (a semi-log plot is normally used) fora first-order process. The ‘equation relating phase angle @ frequency, and the time constant is; of 10 Phase angie, degrees - ‘on + Frequency, radians/sec Figure 4-8, First-Order Process Pnave Angle Phase angle, degrees 2 mets Oya Pr lest 9 09 rr) aT ory 1 10 109, Frequency, degrees igre 410. Fra.Ordr pus Dead Time Process Phase Ang It is important to point out that the magnitude and phase plots can be ‘combined for individual elements to quite easily provide the plots for more ‘complicated processes with such elements in series. The overall amplitude ratios ‘Simply the product of the individual amplitude ratios, while the overall phase langle for the process isthe sum ofthe individual phase angles. The next discus- ‘Sion will show how these basic definitions of frequency response can be used asa tool for finding the plant dynamics from plant testing. 4-6 PROCESS MODELS FROM FREQU ee aes "Lanett xe mess ae eos wbeeos fide Brat ee, | ober ets sgnitude and Phase Characteristics ents (Ret) ‘Assume that the magnitude and phase plots fora particular process have been drawn using a pulse test ofa prt of the plant (Figure 4-12) The nextstepis « to approximate the magnitude ratio plot with a series of straight line segments (he ssymtotes) that have whole eumbers for their slopes (Figure 4-13). An ‘ckamination of the slopes of these asymtotes shows what dynamic elements Should be included inthe model ane gives an approximate value fo the parame {ers such as time constants, An initial slope of -1 indicates the presence o Integraing element (1/2 i Laplace transform notation). A positive change i Slope from Ito indicates a ead erm (rs+ 1) inthe process. A decrease inslope DfT unit to-1 means there is a firstorder lag (1/15 * 1) as does the next change froma slope of -| to-2, Thetime constants and lead time arefound by usingthe frequencies where the slopes changeand noting that at these pointsthe frequency times the ume constant equals one. The scady-state gain is found by extending the slope ofthe lowest frequency element to a frequency of one radian/ second. ‘The amplitude at this point is the steady-state gain It jp still necessary Be Paces Ovni Hom Pan Tot 2 g4 go 5 ae oor oo + 0 Frequency, rediansse Figure 412, Plan Test Results Log amplitude ratio es | output oteies! oo Kas He = Jus completes the very brief inroduction to frequency response tec niques, Frequeney response can be Very powerful method it contol sytem Gesign and application. In general, for any time response method used for plant {esting or control system design, there i an equivalent frequency response nique that could be used. Often the choice of method depends only on the personal preference a the person doing the work. On the other hand, there are ‘umes when one technique is preferred based on the nature of the process being studied 4-7 PRACTICAL CONSIDERATIONS IN PLANT TESTING. There are many practical problems with step testing, just as there are with frequency response and statistical methods of process testing. The process opera tion in an existing plant must be upset to get the information needed. The results {depend on the size of that input disturbanee, the amount of measurement noisein ‘he signal, and the accuracy of the test recorder. Some ofthese problems will be considered in more detail How large should the input change be in a step test? Some authorities recommend a plus or minus $ percen! of span imposed on a base level of $0 percent of span. However, this assumes that all components of the instrument system are properly sized and free from substantial external interference. As an ‘example, consider the test data shown in Figure 4-15. What can be determined about the gain and time constant ofthe process? Virwually nothing, The output signal i so corrupted with noise that ile can be discerned. Increasing the size of the input step change gives an output where the useful signal overshadows the noise and gives useful information (Figure 4-16), LRH eA Cae YY teprtte peso payee | Sau /auane Pi eee a) Pierre tol iaeue eter lia Erte 1 et iG Ti dB CETTE ad Process Dynamics From Pant Tee Gutput - Time" Figure 4-16. Step Test Data sof Output Time Figure 417, Step Test with Other Upsets Finally, her is some controversy about the entice subject of plat testing ining tne eapenteinenpingering time and lost produetion sme sy it ean he Jehute One author as pointed out seme rather simpletest utilizing the Love loop contol sytem coupled with knowledge ofthe process to farm the Sonamic characevstis ofthe plat (Re. 7). Undoubiedly such knowledge ‘SNoutd bc used, but there maybe many ies when isjust cheaper and faster 10 oahead and make the stp ts or pub ex rather hana in-depth study ofthe hemstry oF physies ofthe process, Certainly use should alvays be made of the Salucs of flow rate, the volmes of vss, andthe lengths pipe but this may not be enough. After one has learned the caractersis of a varety of processes, the job of finding the process dynamics becomes much case, A systematic procedure of process testing an help one gin ths experience. Its aso wise 0 fead about the experiences of others im studying the dynamics of industrial processes, For design work itis nat posible toase the operating plant fr making ets and, economically usable, computetsimlation of processes might be relied ‘nas asource of data, While modes based on pant data are very helpful for both onttol system design and controller tuning, simulation data would normally be ‘sel for contro system desig but not fr ti “This completes the rather extensive discussion of proces dynamics, the? topics or those interested in pursing fre study include mathematical model- ing. numerical methods of analysis, computer simulation techniques, and the Tundamentals of whatever proceses are ef interest (chemical, clerical, exile, ‘mechanical, petroleum, ee) QUESTIONS: | Usethe graphical technique to find afits-order plus dead time approxima: tion forthe sep response shown i the figure below. The stepchange inthe ‘controller manual output was from 10 psig to 8 psig. Find the gain ime. ‘constant, and dead time, 4 2. Use the calevlational method to find a first-order plus dead time approxi- ‘mation forthe step response shown in the figure below. The step change in the controller manual output wasfrom 13 mA to 14mA. Find the gain, time constant, and dead time, 55.0] see ee ; Process oynamcs From Pant te 4. Calculate a two time constants plus dead time approximation for the step fesponse shown in the figure below. The step change in the controler ‘manual output was 18 percent of span, Find the gain, two time constants, and dead time Step response 500 ars] 48.9] Output, % 40.9} 375} 35. e 0 ata) 4 Ooms eataTa, % Time (minutes) i “The phase angle plo ‘hase angle of 45 deg is observed ata frequency of I$ radins/ minute, What is the time constant of the process? ANSWER ‘

You might also like