Hacker News Reader

We have accumulated a lot of stories that can be a bit hair-raising. But looking back, the actual rate of casualties is much higher for coal, oil, & gas. Which makes this feel like obsessing about airline fatalities (0.2 deaths per 10 billion passenger-miles) and ignoring automobile fatalities (150 per 10 billion vehicle-miles)

I mean, look at it from the other angle. At San Onofre, a vessel was installed backwards. It then proceeded to deliver power for 25 years & was decommissioned. Additionally, we're reading about all this mismanagement as a result of regulatory oversight. Not as a result of a post-catastrophe investigative report.

I never liked comparing air travel to car travel. It's just not a fair comparison.

Per mile danger is the most favorable towards airlines, but you could use per-journey and then airline travel is 3-4x more dangerous than cars.

The best you can do is either say "airline travel is plenty safe enough" or make direct comparisons like "traveling from NYC to LA via airline is safer than driving the same distance".

> you could use per-journey and then airline travel is 3-4x more dangerous than cars.

This wouldn't be a fair comparison either, unless you normalized for number of people on the journey. (As done in passenger-miles.)

My point was that the 3-4x stat is also not fair. My point was that there are very few fair comparisons, and gave some examples that I think are fair statements.

On investigation, the replacement steam generators from 2011 in both units were found to show premature wear on over 3,000 tubes, in 15,000 places.[47] Plant officials pledged not to restart until the causes of the tube leak and tube degradation were understood.

This particular case at least sounds like the system working as expected. The system was inspected, the tubes were in bad shape, so the plant shut down.

Yes. My point was that these systems are complex, we do not understand them, and even after accidents we learn little, except that they were caused by poorly understood interactions between system components. I also used to say that "nuclear power is safe and well understood by now", but I recently became much more restrained in my statements.

> we do not understand them

You don't. I bet you don't understand how a coal or hydro plant works either. People operating them do.

I am pretty sure he was describing the metaphorical "we", and is postulating that even the "experts" have a misplaced confidence in controlling these systems.

Furthermore, I feel this this could be an ideal application for AI. AI is outstanding at recognizing patterns that are often impossible for other systems to discover, so what would be better then to "train" a group of AI on nuclear plant safety and let them monitor all the systems in real time, looking for any group of subtle changes that, in the past, have lead to severe problems?

I'm sure I'm hardly the first person to consider this, so I was wondering if anyone has any info about using AI in this way?

There are indeed utilities utilizing ML with their plant computer archives (~OSISoft PI or similar) to find trending problems, e.g. identifying things like "This pump usually runs at this temperature for these conditions, but now it is behaving differently".

Per the point of the article, take "radioactive material" and just change it to "toxic" or "hazardous material".

Which avenues produce the most? Even PVs + batteries require mining and by-product disposal with environmentally devastating materials, and the fatalities cause by burning coal is almost immeasurable.

> I would respectfully suggest that you read the book I quoted.

I got the Kindle edition and have read up through most of Chapter 2 now. While I see the author's general point, at least regarding nuclear power, and can understand where he was coming from when he wrote it, I already see two obvious flaws looking at it today:

(1) In Chapter 2 he predicts that there will be more TMIs as more reactor years of operation accumulate. But that didn't happen. If the expectation is that we should have, say, one TMI for every 35 reactor years of operation (35 is his estimate of how many reactor years of operation there were of reactors similar to TMI-2 at the time of the incident there), we should have had 50 or more TMIs by now. So evidently there was some major factor that he overlooked in his analysis. (The obvious candidate for such a factor is that, as someone else mentioned elsewhere in this thread, the nuclear industry learned from TMI and incidents like it, and the level of understanding of how nuclear plants work and how to avoid problems became much better than he indicates in the book.)

(2) He focuses on systems in which what he calls "normal accidents" can take place--incidents that are outside the intended operation of the system, but which are unavoidable (and hence "normal") due to the tightly coupled interactions of the parts of the system and the system's complexity being too high to be fully understood. But he completely ignores systems in which the intended, normal operation of the system causes harm, which is simply ignored because it is too diffuse to attract attention. For example, in the introduction to the book he says that we could have chosen coal plants instead of nuclear plants; he seems completely oblivious to the fact that coal plants, by emitting particulates into the atmosphere, have killed orders of magnitude more people through respiratory disease than have been harmed by nuclear incidents. But coal plants emitting particulates isn't an "incident"--it's the normal operation of the plants--and people dying of respiratory disease isn't a single newsworthy incident, so it flies under his radar and he never considers it.

> I also thought there were "no incidents".

You mention the San Onofre plant, but that was not a US Navy reactor, it was a commercial generating plant. I was specifically talking about US Navy reactors on ships and submarines.

I have not read the book you referenced, but I have added it to my reading list. It seems like it is focused on commercial nuclear plants.

> tell me that we know what we're doing and that we "know how to safely operate nuclear power plants" after reading things like

These are not errors that happened because nobody knows how to do these things correctly. They are errors that happened because technically ignorant managers and business people were making decisions they should not have been making. The US Navy's program solves that problem by not allowing technically ignorant people to be in decision making positions. And even technically non-ignorant people who make minor errors of judgment--far short of the kind illustrated in the incidents you mention--are fired and no longer allowed to operate the reactors, so they never reach the point where they can make a wrong decision that would have significant adverse consequences.

The reasons why civilian nuclear power in the US was never held to the same standards, again, are political, not technical. And it's also worth noting that no other energy source in the US gets held to such standards, even though all of them have caused far more harm than nuclear ever has.

> which newer designs do you mean?

Pebble bed reactors and small modular reactors, for example. But even a standard light water reactor can be passively safe against a TMI or Fukushima type scenario if the cooling system is designed to utilize natural circulation, which US Navy submarine reactors are (they do it in order to be quieter, since they don't require coolant pumps unless the sub is at high speed, but it also gives the passive safety benefit).

The one that curls my toes is Davis-Besse.

https://en.wikipedia.org/wiki/Davis%E2%80%93Besse_Nuclear_Po...

Had that hole perforated, pressurized water/steam would have jetted straight up into the control rod drive mechanisms.