EXAMPLE I
Air Lift Here is how the system was used to develop a method for handling spent reactor fuel
Because most of the operations in recovery of spent reactor fuel are of necessity conducted by remote control, ordinary mechanical devices are exceedingly difficult to maintain. As a substitute for remote-head pumps which occasionally suffer ruptured diaphragms at awkward times, air lifts are used where possible. Here also, of course, it is vital that there be a minimum of operational difficulty, because correction is usually impossible without an extensive and costly decontamination period. Further, intercycle flows are comparatively minuscule, and ordinary "engineering" assumptions cannot be applied to design. The system works like this: In conference with the experimental group it was concluded that not all five of the listed variables would be investigated. Thus, either "run slope" or "elbows" would be eliminated from this scheme as being nonrepresentative of plant applications, or at least scheduled for scrutiny at a later date after the major variables have been evaluated. Further, to effect symmetrical arrangement of runs, water rate would be at two levels, and at the same levels for both '/Vincli a n d 3 /4-inch pipe in order to obtain comparable data. The symmetrical arrangement is necessary in order to devise an "orthogonal" sequence of runs. In such schemes each factor is scrutinized the same number of times, and will appear in combination with each of the other factor-level combinations an equal number of times. Thus each effect can be estimated at the same level of precision. Construction of tests of statistical significance yields quantitative estimates of effects and also those areas where increased measurement effort would be profitable. Thus, if we scrutinize four variables once each at two levels, this would be 24 or 16 individual runs, without repeating any set of operating conditions. However, these 16 runs can be arranged in 16! or 2.1 X 1013 permutations. Accordingly, it would be fortuitous indeed if any intuitively selected arrange-
ment should correspond to the most efficient possible. In some cases this would not be a serious detriment, where the conditions of each run would be reproduced exactly. However, the difficulty of reproducibility is considerably more common than not, and is usually attributed to faulty analytical results or the like. Actually, however, there are run-torun and day-to-day variations that usually introduce errors of larger magnitude than mere analytical error. Fortunately, these runs can be scheduled in such manner as to confuse only unimportant effects with uncontrollable variables that adversely affect over-all precision. It is likely that such a program would be run in sets of four, on different days. It is characteristic of a factorial experiment properly designed for this case, that the data can be analyzed after each set of four runs, since each day's work has the same property of orthogonality as the entire set. Thus, if necessary, the course of the program can be altered as it proceeds. The reader is referred to relevant references (3, 4) for detailed descriptions of factorial experimental
designs, and subsequent statistical analysis. Following completion of these runs, principal effects and most secondary and tertiary effects can be expressed in quantitative terms. It will be possible to predict the performance of a particular design, based upon selection of the optimum combinations of operating conditions. Further, it will be known whether or not an operable system for the specific plant requirements can be designed. In this case the role of CONCEPT U A L was passed over because no unique technology was involved. Air lifts, as such, are sufficiently common that only sufficient data relevant to the specific unit operation were required to enable execution of a reliable design. The role of M E A S U R E M E N T was limited to physical observations, the reliability of which is affected only by calibration of flow devices. The systematization of these observations by STAT I S T I C A L greatly expedited the collection of data, and ensured the reliability of the ultimate proposal by DESIGN.
CONDITIONS Fluid Net lift, ft. Submergence,
%
W a f e r , ambient temperature 40 30
INDEPENDENT VARIABLES
Pipe size, in. IPS, Sched. 80 Horizontal run length,ft. W a t e r rate, liters/hour '/2 inch pipe: % inch pipe: Elbows, flat Run slope, degree
OBSERVATIONS 1. Air rate, c f . m . 2. Slugging a. frequency b. Volume 3. line holdup 4. Miscellaneous
DEPENDENT VARIABLE
VOL. 5 1 , NO. 12
•
DECEMBER 1959
55 A
