control of mean load, Pm, and of constant pressure,

P},  are  very small  as  compared with  that  of  the

servo-valve which is used for the usual servo-con-

trolled electro-hydraulic machine  and continuously

corrects the difference of the load or elongati:on signal

from the command signal of a sinusoidal waveform

or a random waveform.

  As mentioned above, this machine has three controlloops:  for constant pressure,  amplitude  value  andmean value.  Each control circuit has a proper inte-grating circuit to avoid the interference between each

other.

  So far, the control system was explained as for theload control. An elongation control can be conductedby interchanging a load cell with an extensometer.When a program test is required, the amplitude and

mean-value-setting units could be replaced by suit-able programmed signal generators.

Peak-detecting Circuit

  The peak-detecting circuit is shown in Fig. 3.  Aninput signal branches off in two:  one is used for thedetection of the upper peak, and the other, the lowerpeak.  The upper-peak value is detected as follows.Diode Dl allows current to flow only in one directionto charge a holding capacitor Cl·D2 provides feed-back for A1 after a peak is detected.  And C} holdsthe upper-peak value of the signal.4 In this situation,when the upper peak of the signal becomes small, thed-c voltage does not follow it.  So, to reset this value,an FET switch SW} is closed by a synchronized pulse

e,.1 from the mechanical hydraulic pulsator. Then euyvalue is reset to一V, },x at each cycle.  Before eunt isreset, this voltage has been transmitted to the holdingcapacitor C2 by a synchronized pulse erZ.  Thus, theupper-peak value responses with a delay of a quartercycle.  Now, the lower-peak value is detected in the same manner as in the upper-peak detector by re-versal of the direction of the diodes and by exchange of the synychronized pulse erl for ere.

  The amplitude value is obtained by subtraction ofelp from eup and pision of the remainder by two.The mean value is obtained by addition of eu to Pand pision of the sum by two.

  In this way, the amplitude and the mean values areobtained in d-c voltages from an input signal.  Asmentioned above, these values have high responses to the change in peaks through the agency of FET

Switches.

Test Results and Discussion

  Figure  4  shows  some  typical  load-time  records.These tests were run in constant load control at 10 Hz.Dimensions of specimens were 130 mm wide, 12.5 mmthick, 574 mm long, and 275-mm gage length with a5-mm-diam central hole.  The material used was asteel of 800-MPa tensile strength in the as-receivedcondition.

  Two test records were overlapped in Fig. 4.  Eachtest has its amplitude and mean-load records: one isthe record for the present control system; and theother, of an open-loop test approximately similar to

those of the mechanical hydraulic pulsator.l-3  SignCM indicates the  mean load of the present controlsystem; CA, the amplitude; OM the mean of the open-loop test; and OA, its amplitude.  The record OA in-dicates gradual changes from a stabilized value atthe beginning and at the end of testing, probably dueto the change in temperature of oil and that in com-pliance with crack formation in the specimen. On thecontrary, the records CM and CA indicate good sta-bility throughout the testing period.

  It is generally considered that the total accuracy offatigue-testing machines is from 3 to 5 percent, andthat the stability of the control formsthe greater partof it.  The stability of the present  control system,however, is concluded to be about 0.5 percent of therange.  Moreover, the stability of the control systemof the present machine is 0.5 percent of the range ina whole period of fatigue testing or in a week. This is superior to that of the servo-controlled electro-hydraulic machine, much more the usual mechanical hydraulic-pulsator  machine.   The  consumption  of electric power and cooling water is about one third