Monday, May 2, 2011

LPSC Notebook: A Special Note on Risk Mitigation

In my experience, no federal agency is better at marketing and public relations than is NASA. They have to be. Unlike most of the other divisions of the government, NASA survives on its ability to inspire the nation and to involve them in the exciting work that it does. And while everyone can see the economic sense of keeping air travel safe, building roads and protecting the border, NASA is something of a luxury that a rich country can afford as an investment in its future. If people stop believing in what NASA is doing and supporting it, the agency will disappear or at least contract down to its directly justifiable parts.

How does such an organization deal with risk? Unfortunately, not very well. This is particularly true when it comes to human space flight. This has always been a risky business. Eight astronauts died in the lead-up to the moon landing, all on training missions. It was a different time, with higher stakes and these losses were accepted. The astronauts themselves were test pilots and it was widely known that they understood the risks and took them in stride.

But by the 1980s, I think the agency made a fundamental mistake by changing course and attempting to appeal to taxpayers on a more mundane level. The pre-Challenger posters are still up in some places around the country, describing a shuttle flight as "Going to work in space." Moving the bar like that means that when a fatality happens, it isn't just a tragedy, it is an outrage*.

Knowledge of that kind of a response means that space systems designed today are much more risk-averse than were systems designed back in the 1960s. In many ways that is a good thing, but it also adds enormously to the cost of these systems. There's a reason that repeating Apollo with a return to the moon is economically beyond the ability of NASA to complete within a decade, despite the fact that we have much more knowledge and much better technology than was available in 1961.

In a similar way, this explains one of the reasons why NASA's Robotic Space Missions are also very costly. They too are engineered to accept very little risk of failure. There's two different contributors to this which I'll go over one by one.

First, and perhaps most importantly, there is nowhere to repair in space and therefore the reliability of the components needs to be exceptionally high. Reliability for space hardware is much much higher than 99%.  For comparison, when it comes to consumer products, it's not unusual to plan for a failure rate of 2%, and this is built into the price that we pay for these items. Why not make consumer products more reliable? The reason is that it would cost more to do so. But that's not immediately clear. You see, as reliability increases, price increases. But returns and repairs decrease with increasing reliability, saving the company money.  The key is that the increase in reliability isn't a linear increase in cost; as you try to squeeze more and more out of a metric, the price to do so rises exponentially.

To illustrate, consider this apocryphal example using motor vehicles: take a Kia Rio ($13,695 sticker price), a BMW M ($53,000 sticker price) and a Maclaren F1 (no longer sold, but something like a $1,000,000 sticker price) and consider their acceleration and top speed as the key performance metrics. The BMW and the Maclaren are both built for speed, but the Maclaren isn't 20 times as fast as the M, instead it's acceleration is something like 0-60 mph in 4.7 seconds (top speed near 300km/h) and the Maclaren is at 3.2 seconds (391km/h to speed). The Rio can apparently only do 0-60 mph in 9.8 seconds (top speed 202 km/h) or about 1/3rd the acceleration or about one half the top speed for 1/73rd of the price.

Because of this exponential increase, there is a sweet spot beyond which increased metric performance (acceleration or reliability, for instance) costs much more than accepting some failures. If you're a speed demon, it's probably worth it to you to spend the extra $40,000 for the BMW and half your 0-60 mph time from the Rio, but not worth paying the additional $947,000 to drop that time an additional third. In the same way, if you want a very high reliability, you're going to have to pay more and more for it, the closer you get to 100%. In the past, this meant that NASA would launch two copies of many missions, such as Pioneer, Voyager and MER. Because the missions were "single-fault tolerant," you could greatly reduce the cost and accept a higher risk.

I should point out one fine issue here - there is a difference between a systematic risk and a random failure. I bring this up because Bob stated in his talk that Lindberg was right to use a single engine design because two engines would just have failed at the same time. That is not correct. No matter how consistent items are manufactured and used, they will undergo slightly different manufacturing and life cycles and be made of slightly different stuff. Thus, the specific time to failure for each unit will be different. Of that 2% of returns mentioned above, we're mainly discussing random failure, as most systematic risks are eliminated prior to a product getting to market. In manufacturing there is an acceptable range for any parameter (say the tightening torque on a bolt, or the stiffness of structural steel). If a component ends up with values near the middle of that range, its lifetime will likely be greater than a component with marginal values (note that the reliability argument above mainly has to do with adding cost to reduce these ranges or "tolerances" as we call them).

For a starker example, consider the usage life. Say that one of the two engines Lindburg was considering came from a good mechanic who always kept the oil topped off and another who was less meticulous. Or, if they were new engines, what about encountering a flock of seagulls over the Atlantic and perhaps sustaining a bird strike? If you have two engines, you can survive a failure in any one. But if you go with a single engine you have to have the well maintained one, and hope that you are lucky and that all the birds miss you.

The second major factor relating to reliability that drives up costs is technical innovation. Every time you do something from scratch, you increase the cost. As Ross Beyer said in his recent interview with, each NASA spacecraft "is its own unique and delicate flower." That does offer advantages - you can tailor a spacecraft specifically for the environment that it will encounter and the instrument suite that it will carry. But the advantage of reusing an old design is that there is less time to be spent "working out the bugs" and you get an economy of scale, after a sort. For example, consider the MER rovers. Initially the project was to build one for a cost of about $600 million, but it was possible to add a second identical model for much less, around $250 million. If were were to build a whole fleet of  MER rovers, that unit cost could fall even further.

This begs the question - is there a use to creating a standardized platform for spacecraft? ESA certainly thinks so. With fewer funds available than NASA, ESA has done very well with reusing the same bus over and over again with slight changes. This is why Mars Express, Venus Express and TGO all look so similar to each other - they are essentially the same spacecraft with different payloads. NASA has also used this strategy with launch vehicles, having at one point bought the rights to multiple launchers to bring down the per-unit cost (the end of this program is a big part of the reason why launch services have exploded in cost lately and the planetary decadal reccomended excluding launch costs from future cost-capped mission AOs).

This argument dovetails into another tantalizing tidbit from Bob's talk - the possibility that accepting higher risk can lead to up to an order of magnitude decrease in the overall cost. What might planetary exploration look like at 1/10th the cost?

Well, let's consider Phoenix, which cost something like $475 million to produce. What changes with Phoenix at $47.5 million? Well, right off the bat you'll notice that the cost has come down to something like what a mid-range movie costs to make. Would a lander or a small rover of this range be of interest to a movie studio? You could make the launch an event - document a movie or make a series out of it (if you did a 24-episode season, it's again on par with a moderately expensive dramatic series) and involve the public directly.

That's the kind of marketing I can get behind.
* To their credit, following Challenger and Columbia, the trend is swinging back the other way. At a screening on "Hubble," the narrator pointed out repeatedly that each astronaut on every mission was putting their life on the line to go into space.

* As an aside, I offer my apologies for the length of time required to get this post to market. But soon I will make clear the reasons for which my blogging rate has been low these last couple of months

No comments:

Post a Comment