I

t was early evening in the dimly lit cocktail lounge, and I was relaxing after a long day by nursing a parched martini, the pair of olives nestled in its depths blankly staring back at me. The hectic nature of the day had nearly immobilized my body, but had inspired my brain to frantic activity.

I had just attended the first day's sessions of an operating and support costs symposium. The presentations had been highly thought-provoking. Letting my mind wander back to the previous day, when Dr. Marco Fiorello and I had discussed various issues involved in modularity, I coalesced that conversation with the more stimulating points of the symposium and conceived three sharply different approaches to aircraft avionics modularization.

GALRU

In my first case, all of the avionics are in one Giant Avionics Line Replaceable Unit (as is the bureaucratic wont, we'll call it a GALRU). Naturally, a few physical difficulties might be encountered; for instance, a crane would probably be necessary standard ground equipment to install and remove the GALRU (in the interests of cost-effectiveness, the crane could do double duty on other jobs as well). There might also be a center of gravity problem, but we'll

by Rustle Em Jenay DoD Analyst?

presume that it could be overcome by our aircraft design people, who typically seem to be in search of a challenge.

With but a single LRU to malfunction, flightline diagnosticians wouldn't be bothered with pulling the wrong unit. This would presumably reduce false removals to a very low rate. And integrated avionics would now come into its

own.

Certain life cycle cost enthusiasts, hung up on introducing new LRUs into the inventory, would predictably anathematize the GALRU. But limiting the LRUS to one per new aircraft would substantially reduce such problems as LRU management, assignment of Federal Stock Numbers, and so on. The number of technical orders would also be trimmed to one, although it would be 27 feet thick (wheels and a motor could be attached to its cover to make it sufficiently mobile). Costly hot mock-ups in the field would be avoided since the GALRU would serve as its own hot mock-up. Finally, those taxing decisions on which LRUs to modify would be eliminated, there being only one LRU, and a significant amount of study money and modification board time would be saved.

Of course, there might be a minor drawback or two. Since any failure would require removal of all the avionics from the aircraft, there would inevitably be quite a few of these giant LRUS being shuffled between the flight line and the intermediate-level shop, which might prove to be a trifle expensive in terms of spares. How

ever, proper application of design to cost could keep this under control.

While 100 percent field-level repair of the LRU is theoretically possible using Shop Replaceable Units from the depot, in actuality, with complex avionics some of the LRUS are always NRTS'd back to the depot for repair for one reason or another. Again, stocking this pipeline could be expensive.

Those people promoting avionics standardization across various types of aircraft would no doubt tend to frown on the GALRU, but Reliability Improvement Warranty proponents would be delighted in that one RIW would cover all. Some analysis would be needed to verify the GALRU concept, but the fact that there is only one LRU should simplify this process enough to compensate for the less-consequential problem of having 5,000 SRUS.

FMB

As you have probably anticipated, in our second case every resistor, transistor, integrated circuit, etc., is an LRU. The biggest problem here is that there would have to be as many connectors as there are discrete components, not to mention the mass of interconnect

ing wires. One might initially conclude that 5,000 pounds of connectors is an insurmountable problem, especially since there must be easy flightline access to each LRU, i.e., component. Obviously, surface area would have to be rather extensive.

No problem for the imaginative government manager: we simply turn these difficulties into an asset by replacing the aircraft skin and structure with one huge mother board. The monocoque construction would be the absolute last word in avionics-structure integration. Plugged into the stressed-skin mother board would be the 127,532 LRUs. Should there prove to be insufficient surface on the inside of the aircraft, the LRUS could also cover its outside. Significantly, this would save money by eliminating any requirement for speed brakes or vortex generators.

The standardization folks would be extremely pleased with the FMB, since standardization would be completely captured at the LRU level without infringing on design freedom.

I will admit that it may take a little longer to find the right LRU to replace on the flight line since there are 127,532 from which to choose, but this would certainly be more than offset by

the complete elimination of intermediate- and depot-level maintenance. Just think how much could be saved by closing depots and excluding all intermediate and depot automated ground equipment! And SRUS would be eliminated forever!

Another enormous benefit would be the elimination of all the expense and bother of optimum repair-level analysis, as there would be only one level of repair. We would also have achieved the impossible dream of all true R&Ders: 100 percent throw-away LRUS (thank goodness you can't fix a resistor). While there would probably be corrosion problems when dealing with 127,532 connectors, space limitations in this article preclude further discussion of this topic at this time.

Since there isn't a computer simulation program that can handle 127,532 LRUS, analysis to verify the flying mother board concept would have to be done by hand.

GALRUFMBC

For those pessimists not sold on either of the above alternatives, there is a middle groundthe GALRUFMBC (the C standing for compromise). The trick is to take the best features of each extreme while avoiding the bad. This would necessarily involve some compromises and trade-offs requiring study, analysis, and even hard thinking-not to mention voluminous reports, lengthy meetings, and state-of-the-art seminars. But who is better suited to such a task than a professional bureaucrat? The crane required for the GALRU might be a difficult notion to sell, so some upper limit on LRU weight would be essential. With women now participating more fully in maintenance tasks and some of us old men having weak backs, this limit would be set at 30 pounds. The potential center of gravity problem on the GALRU thus disappears as the LRUS can be distributed around the aircraft a bit (but 500 pounds should be added to account for interconnecting cables).

Because there would now be more than one LRU to pull, we would need to be concerned with pulling the right one most of the time. As the number of LRUS increased, high diagnostic error rates would become less tolerable. Integration and software problems between LRUS would need careful attention. Hot mock-ups would probably have to be provided to the field

so that those personnel wouldn't have to sneak one together under the cover of darkness.

By adroitly arranging LRUS by functional characteristics and failure rates, diagnostic errors could be reduced, and pipelines would not be filled with loads of nonfailed and expensive material. The flying mother board advantage of standardization could be partially retained if the functional arrangement of LRUS was so ingenious that it even considered multiple aircraft-type applications. Lost in the compromise, however, would be throw-away LRUS. Finally, we would have to retain optimum repair-level analysis, not to mention depots.

But then being an analyst by profession myself and having a distinct inclination to stay one, I generally prefer the compromise position, which coincidentally would require the use of my skills. DMJ

production, and scheduling.1 In doing so, DTC policy attaches great importance to design by making cost a design parameter and stressing the need for tradeoffs between design and system capability.

Reflecting the central tenets of DTC policy are the cost-per-unit ceiling and the general absence of performance requirements in requests for proposals. By stipulating a cost-per-unit ceiling, DTC RFPs make cost overruns and insufficient acquisitions less likely. Because of this ceiling, costly improvements in one area of system capability must be balanced against possible, economically mandated deterioration in another. Since performance requirements may require such improvements, properly stated DTC RFPs minimize or eliminate performance requirements and give contractors considerable latitude in designing systems for certain missions in certain scenarios. As a result, they challenge contractors to design systems maximizing the military worth, or benefit, of their total performance within a cost constraint. Since system designs reflect contractors' different corporate, industrial, and technological capabilities and interests, and their different interpretations of the missions of the system, contractors are likely to propose highly diverse systems, each promising to satisfy the military service's needs.

Evaluating Design Proposals

The cost-per-unit ceiling and the general absence of performance requirements make evaluating diverse design proposals more difficult than it has been in the past, since all contractors are likely to propose designs meeting

the cost-per-unit ceiling. Paradoxically, for those designs proposing and then estimated to meet the cost-per-unit ceiling, cost itself is no longer a significant consideration in evaluating the proposals. Moreover, RFPs stating missions and scenarios rather than performance requirements alter the focus of the evaluation process from a direct consideration of system capabilities to an analysis of the military worth of the total performance of the system.

Accordingly,

the military service confronts the crucially important and extremely complex problem of evaluating diverse design proposals without employing the customary criteria of cost or common performance requirements. The logic of evaluation and the nature of competitive bidding require that any program for evaluating the military worth of proposed systems must be carefully constructed to meet four criteria; namely, it must discriminate effectively among the alternative design proposals, and it must be reliable, intelligible, and equitable. Any program meeting these criteria has but one operational purpose: to relate measures of performance of the controllable parameters of a design proposal to the total military worth of the proposed system. Only after such an evaluation does the service need to estimate the contractors' abilities to meet the cost-perunit ceiling.

The first two criteria, discrimination and reliability, are satisfied or not by the nature of the

1Department of Defense Directive 5000.1, "Major System Acquisitions," July 13, 1971 (latest version issued January 18, 1977); and DoD Directive 5000.28, "Design to Cost," May 23, 1975.

Aspects of the Model

Structuring

The many considerations addressed in developing an MAU model are finally organized hierarchically; more general elements are successively decomposed until a structure is developed that aggregates measures of performance characteristics up through TSU of each design proposal or that justifies any higher-level figure of utility or effectiveness by reference to lower-level figures. For these reasons, correct structuring of the problem is the most important step in the process. Moreover, correct structuring helps to reduce possible errors in judgment and to ensure that variations in the design proposals are reflected in different figures of TSU.

Utility and Effectiveness

The words "utility" and "effectiveness" are used in this article in senses consistent with, though not necessarily so specific as, their usage in related disciplines. "Utility" is mainly applied to evaluation (in the narrow sense), "effectiveness" to prediction, and each has its important part in evaluation (in the broad sense). The distinction, admittedly slight (see Figure 1 for the area in which simulation runs may or may not be used), is that between determinations involving military tradeoffs and those involving technical system capabilities.

Scaling

Since an MAU model involves many different kinds of elements and many different units of measurement, it is important to have a common unit of measurement that makes all measures commensurable to facilitate aggregation up through the model. The convention is to scale either utility or effectiveness from zero to 100. Accordingly, various performance characteristics may be measured as percentages which become amenable to aggregation at higher levels in the MAU model.