THE MECHANICAL SEAL by Kerr Roberge SUBMITTED IN PARTIAL FULFILIMENT OF THE REQUIREMENTS FOR THE DEGREE OF BACHELOR OF SCIENCE at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY May, 1960 ay pp Rl : sis Superviser Accepted Bye. ss. esse cess eee cc scene Chairmen, Departmental denior Thesis Committe:ACKNOWLEDGMENTS: I wish te express ay aos sincere gratitude te ay thesis adviser, Professer Leonard A. Gould. Prefesser Geuld was respensible fer the eriginal suggestien which. started me thinking about a breem balancer as wel) as guidamge throughout the course ef the werk, Witheut his excellent technical advice and enceuragement the device could net ave been ceapleted, I would alse like to extend my thanks te Mr. Walter Beyntea ef the instrument reom in the Electronic Systems Laberatory, In werk ef this type a large quantity ef equipment is necessary, Mr. Beynten always seemed able te lecate any required compenent in a minimum amount ef time, Om numerous ecassions he allowed me the use of his own persenal equipment. JKR, iLIl. III. Iv. CONTENTS Ackmowledgments sesseeeeceeecicceeeseeeeseee id AdStTACt eeee eee e ee eee eee 1 Intreduction ...... . 1 2 Preliminary Investigation of Stability so... Hare Descriptien of Components ........ eeeteinte ++ 10 Final Stability Considerations ........ reves 26 Results ...sseeee eee eee Bibliegraphy Appendix Operating Instructions ...... eens 38 aadABSTRACT A servi chanism has been designed and built which dalaaces a breem, This thesis presents the analytic tech- tion needed te main— niques used te determine the ceape: tain a breem in an inverted pesition, The construction of the components used to realize the indicated compen- satien is alse described.I. DNTRODUCTION The problem of stabilization of « set of fixed eles Renta which are dynamically unstable is an inherentiy ioteresting ene to the servo designer, This condition ts not commonly met with in practice, Usually any insta- Dilities in the feedback control system are a direct con- sequence ef feedback. This type of instability can of- ten be eliminated by reducing loop gain if the resulting decrease An performance can be tolerated. an excellent example of an inherently unstable de- vice ia the inverted pendulum er broom, This system has * transfer function with « pair of poles en the positive and negative real axis, If a servomechanism is used to Baintain a broom in an inverted position, any closed loop Anatabilities cannot be corrected simply by reducing loop Gain since, if the gain is decreased below a certain lin- it, the breem falls ever, A device haa been designed and constructed which ac- Compliskes the task of maintaining a breem in a vertical Position when constrained te fall in one plene only. This contre] mechani. 4s currently in operation and has been Gemenstrated at the recent M.I.T, Parents! Weekend. It is capable of sustained operation for long periods of tine. It generally functions correctly until meone (usually the designer) attempts te demonstrate the ability of the mechanism to correct for large disturbances and pushes the breom toe far off its null position,In its present version, there seems to be little practical application for a device of this type. Hewever, the equations apply almost directly to a rocket at take. eff. The recket at low speeds behaves exactly like an inverted pendulua, The basic method of stabilization outlined here should have medifications which apply to a wide variety of contre! systems in which fixed elements have one or more poles in the right-half plane. II, PRELIMINARY INVESTIGATION OF STABILITY Figure 2.1 shows the idealized physical situation from which the equations of motion of the breom can be determined. The broom is assumed to be pivoted at the base on frictionless bearings, such that it can fall en- ly in the plame ef the paper. The supporting rod is as- sumed te be m less and inflexible. x, is the variable which can be contrelled to maintain stability. The enly information available is the magnitude of the angle 0. Summing ferces at the breom head, Force = WX, = Mg sin @ - Bi, (2.1) sino +o@-+ 1, - x, (2.2) From this point on in the preliminary analysis, the damping, B, will be ai jumed to be zero. Later.in the de- Velepment it will be included primarily as a method of limiting the maximum required values of k, and X,. 0 IteMASS 0} BROOM HEAD EL GENE & - ees os vISenue Fievre 2.1 Broom DYNAMIeSinclusion does not seriously alter the stability consid- erations and thus for the present it can be ignored with- out drastically changing the form of compensation necessary for stabikity, The equations of motion can be used to fora the block diagram shown in Figure 2.2. G(s) is the compensation required to yield a stabilizing X, as # function of @, Te eliminate the necessity of handling a aultiloop system, the compensation loop is’ opened at the indicated point and the loop containing the broom dynamics is col- lapsed, ‘The broom loop can be replaced by a block having the transfer function VE . e/g + an(t/gy® (242) g/e-L ” (ase1)(as-1) As an attempt at stabilization, let G(s) + K/s%, This preduces ac open-loop transfer function K 2 (2.4) Tass TyCaa=T) The corresponding closed-loop transfer function 1s ed 7 E 7 x SD ee=1 + —=3 1l+E 1 Bhas ie gas’ ~g@+K & TasrTytae=T) (4.9) From Routh's stability criterion we notice that this form of compensation has removed the pole from the right-balf plane previding K>g, but has left a pair ef poles on the imaginary axis, Notice that the term 1/s° must be « true double integration rather than of the form W/(ba+1)".|v ioe | Ficure 2.2 PRELIMINARY BLock DIAGRAMIf we assume 1/s° to be approximated by t/(bs+1)2, the resulting open-loop transfer function ia ae Ek & (asvl)(as-1)(bse1)= haa The corresponding characteristic equation of the closed- loop system is g(as+1)(ae-1)(b8+1)° + KBs? « (602d? 84 202s (a2ab2)82-2ba-1)) + KBr. 0 (2.7) We notice that there is no way to eliminate the negative 7 O in this equation. ‘thus at least coefficient of s” and s ene pole remains in the right-half plane. A similar re- sult is obtained if an approximation to 1/s” is made by d/a(bs+l). It is cencluded that, if damping in tae broom is ne- glected, at least a true double integration is necessary. The enly practical way to ebtain this seems to be the use of electro-meckanical integrators such as motor-tach units. It is assumed that such integrators will have some lag associated with them. Therefore, a reasonable expres— sion for the double integration would be 1/a@(ca+1)°. To offset the lag associated with the integrators, and to eliminate the oscillation which would be present if G(s) were K/s?, « lead network is employed. The re- sultant G,(s) is of the form K 1 dsel yard (2.8) (cael)? ~ (dav)When the dynamics of the broom are included, the overall open-loop transfer expression becomes: K 1 a mMds+l] psc ina Stderl 2. & (as+1)(as-1) (cael)? aavl sete The term 1/(as+1)(as-1) is a non-mininum phase expression. It has magnituce characteristica that are the same as the magnitude of 1/(as+1)° but has a phase shift of -150° at all frequencies, A Nyquist plot ef Equation (2.9) is shown in Figure 2.3, %It is assumed in this plot that the integrator time Constants, c, can be ade shorter than a by proper choice of integrators and the length of the broom. It is also assumed that oxand d can be found such that over some Tange of frequencies the net phase of the expression is greater than -180", We see from the Nyquist plot that stability can be attained. From the plot, there is -1 net encirclement of the -g/K point, Since there is one open-loop pole in the right-half plane the system is stable. ‘The diagran further shows that it is possible to use a Bode plot for the determination of gain and time constants of the lead network providing the D.C, gain of the expression is made greater than 0 dg,o (OD NOltweADA 40 Latd ASINDAN) STE aunai 4° ONT S¥aYONT MmIII, DESCRIPTION OF COMPONENTS Once the basic type of compensation necessary to stabilize « broom was determined, the individual cempon- ents of the loop were assembled. the data obtained from measurements on these elements enabled selection of the final cempensation. The platform used for movement of the handle of the breom was originally # 2(1,Y) generator constructed by the M.I.T, Dynamic Analysis And Control Laboratory. it perzits movement of a small output fixture along two per- pendicular axea, The drive is from two-phase servomotors via gearpasses, pulleys, and steel belts. At the start of the work, it was intended to give the broom freedom to fall in any direction, This would have made use of both axes of the Z(X,Y) generator, However, construction of the second axis was not carried out due to lack of time, Accordingly, only the x axis of the generator is being used. The only work done on the generator was mech- anical maintenance and modification of the input adapter on the x axis to accept a new servomotor, The generator had been out of use for a long period of time prior to the start of this project and some of the components were badly rusted, Various adjustments such as gearpass back- lask and belt tension were incorrect, When these condi- tiens were corrected, the table functioned excellently. The x axis was originally driven by a Bendix servonotor. A Diehl unit of higher output power was substituted topermit Righer acceleration, The waximum travel in the X direction en the table is .55 meters, The gear ratio is such as to limit the maximum slewing speed to adeut +45 meters/second, ‘The "broom" was constructed as follews. A piece of steel conduit 1.55 meters long was cut. A mass was con- nected to one end, A hole was drilled through the other end to accept a shaft. The shaft is held in place with gears pinned te the shaft and bolted to the conduit. The ends ef the shaft are supported on ball bearings. This arrangement insures that the broom falls in one plane enly, Since the conduit is not massless, the effective length and mass of the broom were determined by measure- ment, A natural frequency of 2.05 rad¢:ans/second was ob- served. This corresponds to an effective length of 1.4 meters, The mass was measured as 1.6 kilograms. The an- gle between the broom and vertical is measured by a Clif- ten Precision Preducts type CT-11-B-2 synchro. ‘This is driven frem the shaft at the base ef the broom through & seven:one stepup gearpass, An attempt to reduce back- lash in this drive was asde by employing # rubber band te keep the small gear always wound up against the large gear in the same direction, Figure 3.1 shows the con- struction of the support and synchro assembly. The synchro is intended for eperation at 400 cps with 26 volts on the rotor, For system considerations it fe necessary to operate the synchro at 60 cps. To ac-oe! OCAXHINLS ONY WAV 4 ne / eae ee Wood g r€ ayyno4 “a feu es Cty eee . L. y/ Lansans 7 Y BIGNyH Wooyer Hycomplish this, the rotor voltage was decreased to 6.5 velts which is obtained from the line through a Variac driving an 8 volt transformer. ‘The maximum output of the synchro is 2.5 volts rms. Thus, with the gearpass, the sensitivity is 16 volts/radian for regions where the small ° at the broom). angle approximation is valid (2°-3 D.o, power for the electronics in the system is ob- tained from two regulated power supplies, one delivering 300 volts at 250 ma. , the other 750 volts at 250 ma. . 90 volts D.C, is obtained from batteries for fixed bias on the output amplifier and as a supply for a transistor amplifier serving as an active compensation network, The first integrator (Figure 3.2) consists of a 5 tt Diehl servomotor driving » 20K pot through a 10:1 gear reduction. Tachometric feedback is applied around the moter to impreve the frequency response of the inte- grater, The fixed phase of the motor is connected to the line: phase shift for the control signal is obtained by the network shown as part of Figure 3.1. This network provides mere than 90" of lag at 60 cps. However, the voltage at the centrol terminals of the motor is only about 80° lagging with reapect to the line, indicating some positive phase shift through the amplifier. The aup- lifter is a 15 watt serve amplifier ebteiced from the instrument reom in the Electronic Systems Laboratory, The gain from the tachometer output to the contro) wind- ing of the motor is set at 15.wy jae WoivYOaINT asuiy Te aynsi4 ADAM ee wort Batsinges : ' | PY) sa209 2A SH u + u t Iw alndinoFrom measurements. made on the first integrator, the gain with 90 volts across the pot is 140 volts/second D.C. output per volt rma in. The log magnitude plot falls at 10 dg/decade until 50 radians/second where the slope be- comes -20 dg/decade. ‘The transfer function for the first integrator is then 140/s(.02s+1). Since the carrier fre- quency fer this system is 60 cps, a bandwidth of 50 radians/ second seemed reasonable and no attempt was made to im- prove it. The first integrator produces a D.C, output proport— ional to the integrah with respect to time of the envelope of a 60 cps input signal from the synchro. Because the output is D.C., tnis seemed a good point in the loop to insert compensation, From considerations which will be discuased in more detail later on compensation of the form K(.28+1)/(.0158+1) was desired. The D.C, attenuation inherent in a purely passive network could not be tolerated 50 a transisterized active network was designed. The drift problem is not teo severe in the design of this cir- cuit since it is preceded by an integrator, Any shift in the D.C. output can be compensated for by a shift of the output of the first integrator and will not be detri- mental to closed-leop operation, assuming the drift is not too rapid. Tae schematic for the network is shown in Figure 4.3. The operation is as follows, A signal from the first in- tegrater is attenuated by the 100K pot. This serves asWYOMLAN Baisoy FE Zana : Sho S+- ; wee $ Lh wee ; vid SCN IBIN Sata so 4ad1.na 1038 4nd 140 rocco BENT Ba! : S+- ' 1 > ' 2 SE wort g Yore $ you oot Sha gain control for the network and subsequently for the entire loop, Beceuse the 2N35 has «# of 75, the input impedance is greater than 250K for any frequency and thus the gain pet is not seriously loaued, The first stage behaves basically as an emitter follower to the emitter of the 2N35. Thus on an incremental basis the gein to the output of the first stage is - aaa eon, 73h an g73Hs VAAR vOoTs Noes Suonyge 4 beans 70 TOA 08, ‘SaTyaaas wos NOUINOT Timur “bo vice comes from this strip, At this point it is wise te check to be sure that: 4. The filaments in the output amplifier are lighted, B. Both motors (the output and the first integra- ter) kave 115 velts on the fixed windings. C, The Brown Converter (gray can op the output amplifier) is operating. The operation can be neted by lifting off the can. If the reed is vibrating, the converter is receiving pow- er. 4, Turm beth the filament switches en the power sup first cennected fer the transistors in the netwerk te stabilize, Cemnect a D.C, veltmeter between the eutput terminal (center wire ef the black shielded cable cennect- ed te the amall terminal strip en the netwerk) ef the ac- tive network and the point labeled Ba on the terminal beard, Adjust the screwdriver adjust balance pot en the back ef the network for a veltage mull en the neter, This adjustment skeuld be made with the gaia centrel (the kneb furthest left on the front panel ef the device) « If the network is turmed comphetely counter-cleckwi: fumotioniag preperly the balance contrel sheuld all: at least a + 10 velt swing at the eutput. 6, Adjust the centre] em the tep of the Variac te100, With this setting the output of the small filanent transformer connected directly te the Variac should be about 6.5 velts A.C, This is the synchre supply voltage. 2. Push the broom transport bar all the way te the left when viewed from the front. Leck the set screw se- curely. 8, Turn on the plate switch of the 750 volt power supply. Then turn on the plate switch of the 300 volt supply, The power must be applied in this order, or else the screens of the 807s will be ruined. 9. Check the gain setting on the first integrator amplifier, The 10 millivolt imput gain pot should be set to 8, If everything to this peint is operating cerrectly the first integrater should run into the stops on beth ends as the broom is tilted to either side of vertical. 10, Set the Kp adjust (middle knob on the front pa- nel) so that 0.5 velts is measured acress the slidowire on a high impedance A.C. veltmeter. ‘This corresponds to a netting abeut 1/3 of the way cleckwise en the pet. 11. Set the mull set (rigat-hand knob witk large dial en the centrel panel) te abeut the center ef its mechanical limits, 12, Leosen the set screw on the broom transport bar. Be sure to hold the breem handle, since drift in the n werk may cause the output integrator to drive quite quick- ly, If the speed of travel ef the transport bar is grest- er than 0.1 meters/second, repeat Step 5, To check theba output integrator leop, try to move the transport bar against the direction of travel, Heavy resiatance ekould be emcountered. Force the bar to the center ef the table and leck down the set screw, Be sure not to allow the broom to fall during these tests, Until familiarity is gained with the equipment it is prebably wise te use two People to make these tests, 13. The first integrator should still go to either limit as the breom is tipped te either side of vertical, Check the null position on the first integrator, Motion Should cease when the breom is perfectly vertical, If this is not the case, slip the large gear between the broom Randle and the synchro on its shaft to position tae syn- chro such that there is no first integrator motion when the breem is vertical, This adjustment is fairly critical. 4, Leesen the set screw and allew the breom trans- Port bar to drift to either side. furn gain (K,) pot about 5° cleckwise, At this point it should be pessible to tip the breom te either side and have the breom trans- port bar follew, 15. Try to get the broom as nearly vertical as poss- idle. Turn the K, pot about 1/4 turn clockwise (to the indicated line) and let the broom fall simultaneously. This pert. Tequires a little bit of practice, If every— thing ie working correctly the breem will oscillate for ® few seconds and then stabilize, Do not allew the breom to fall ever as « result of excessive initial offkang.If dirticulty is encountered repeat steps 14 and 15, 16, When the device is operating, if a tendency is noted for the broem te kunt about « point not in the cen- ter of the table, adjust the null set pot, Turning this Pet clockwise makes the breom move right. IF there is not enough latitude in thie control, repeat step 13, 12. Adjust Ky and K, slowly to minimize the magni- fude of hunting oscillations. Tap the broom handle slight- dys The respense should be quick and have 2-3 high-tr quency oversheets., If the overshoots are excessive, de- crease E,. If none ere present increase K.. To repeat * warning given previously, at no time should the broom ever be allowed to fall over for any reason whatsoever, Tf this eccurs, the aynchro shaft will be severely bent, and the required mechanical rebuilding on the synchro and platform takes about ene hour, 4 commen cause of iaproper operation is slippage of the gear driving the first integrator pot. This gear can be tightened by the Allen Head screw on the gear collar, Tf the above procedure fails to achieve operation, check the output integrator first. Remove the D.C, input deck from the output integrator and apply 1t:¥olte from a battery, This should result in a transport velocity ef about .05 meters/secend. Alse lock the transpert bar and a ure the voltage on the motor-contrel winding, This voltage should be greater than 125 volte. If tt is ever necesaary to replace a tube in thebh output amplifier the following adjustment procedure sheuld be ebserved. 1, Remeve the moter leads from the 84 ohm and 500 ohn taps on the output transformer, Also remove the tach- eneter output lead from the amplifier A.C. input, 2, Connect a 100 watt light bulb between the 84 and 250 ohm taps. 3. Remove the 807s from their sockets and apply the 300 volt supply to the amplifier. 4, apply a 60 cps % volt signal in phase with the line to the A.C. input jack. With the gain pet turned up about K of a turn clockwise adjust the A.C. balance control te drive both 807 grid pins an equal amount with respect te greund. 5. Replace the 807s and turn on the 750 volt supply. Put in both the A.C. signal and « 1% volt D.C. signal, Observe the output waveform. Adjust the gain pot to yield 2 100 velt peak-te-peak output. This waveform shoudd be the sum ef a sime and square wave. Adjust the phase con- trel so that the zere crossings of these two component waves ceincide. 6. Remove the light bulb and replace the output drive motor. Adjust the gain control of the output ampli- fier to the highest possible setting consistant with stable operation ef the leop, If all of the above adjustments still rail te yield stable broom operation, the best advice is simply te be-4s co familiar with the theory of operation described earlier and start troubleshooting,