Design basis events and collusion
1. A design basis accident is the maximum credible accident a nuclear facility is designed to withstand. It is the worst accident that could credibly be anticipated. Nuclear facilities have to be designed to withstand such things as the worst earthquake, flood, or storm that could be expected in the area in which they are sited. Any accident that cannot be credibly anticipated is termed “beyond design basis.”
It should be pointed out that the design basis only considers credible events that can be planned for. Often cited examples of events that cannot be planned for include those that do not pose credible challenges, such as the sudden emergence of a new ice age. But the design basis does not include acts of war, terrorism, and human failings of operators, regulators, or design engineers.
2. The core damage frequency is a probabilistic assessment of frequency of core damage events, accidents that damage the fuel in a reactor. It is common for the terms “core damage event” and “meltdown” to be used interchangeably. Most core damage events are not catastrophic, however, and about two-thirds of the historic core damage events at commercial reactors went by without much notice. In commercial reactors, there have been four core damage events that were catastrophic, and, as far as we know, eight others not that bad.
3. The evaluation of the core damage frequency is in terms of reactor-years. A reactor-year is a measure of time of operating reactors, equivalent to one reactor running for one year. A core damage frequency of 1/10,000, means that there is a 1/10,000 chance of a meltdown in any given year at any given reactor. With 500 such reactors in the world, there would be a 1/20 chance of any year having a core damage event, with a much lower chance for catastrophic meltdowns. Only old reactors have an estimated 1/10,000 core damage frequency, however, and the newer ones are considered safer.
Industry safety predictions:
The GE boiling water reactors are among those estimated to have a core damage frequency of 1/10,000, at worst. Later reactors, which constitute more than 90% of those running today, were estimated to have core damage frequencies of, at worst, 1/20,000 or 1/50,000. For the current number of nuclear plants, this translates into one core damage event in 40 to 100 years, with catastrophic meltdown much less often than that. Please note that these are the most conservative industry numbers. Some estimates indicate core damage events could be expected once in a millennium, or even less often.
Empirical safety evaluation:
In the cold light of reality, the core damage frequency estimates have turned out to be grossly inaccurate. The Max Planck Institute has issued a report, “Probability of contamination from severe nuclear reactor accidents is higher than expected,” indicating that the actual frequency of catastrophic meltdown of nuclear reactors is about 200 times as high as the estimates indicated.
The report is not based on estimates, but on empirical data. Its calculation considers the number of reactor years accumulated and the number of core damage events that have happened. Worldwide, nuclear reactors have accumulated about 14,500 reactor years, and produced four catastrophic meltdowns. This means the actual likelihood of a catastrophic core damage event is about 1/3625. For all core damage events, including those that are not catastrophic, the frequency is about 1/1210 for all reactor types, and about 1/635 for older reactors.
The Max Planck Institute report says we can expect a catastrophic meltdown of current technology nuclear reactors every ten to twenty years, worldwide, with non-catastrophic core damage events somewhat more often.
The causes of failure:
The report on the Fukushima Disaster, issued by the Japanese Parliament, reveals the most important cause of that failure. According to that report, the Fukushima Disaster was not caused by an earthquake or tsunami, but by collusion between regulators, owners, and government. And in fact, we can observe that all of the core damage so far are arguably the result of human failure, which is not accounted for in considerations of the design basis or the estimates of core damage frequency.
An example of the implications:
To use the Vermont Yankee nuclear plant as an example, since it is an old-style reactor the empirical data we have says there is a 1/635 chance of a core damage event in any given year. Given a licence period of twenty years, there is a 1/32 chance it will have a core damage event, and the chance of a catastrophic meltdown is roughly 1/100. Since there are 23 such reactors in the US, we can expect an approximately 20% to 25% chance of a catastrophic meltdown, if they are all allowed to run for another twenty years. Of course, this does not take the effects of aging into account, so the actual number is probably worse than that. It also says nothing about the other 81 reactors in the US. Furthermore, it only accounts for reactors, and not for spent fuel pools.
We have two very different sets of numbers here representing safety probabilities. One is the industry-predicted rate of failure in an ideal world. The other is the actual failure rate in the real world. The industry-predicted failure rate is only about 0.5% of the actual failure rate.
One implication is that the engineering that went into safety of nuclear power plants is probably insufficient to prevent 99.5% of the failures.
Another implication is that the underlying assumptions on nuclear safety used by industry and regulators are nearly completely wrong.
Clearly, current standards for safety engineering of nuclear power plants have failed utterly, and need to be replaced completely. To fulfill its role properly, nuclear safety engineering has to account for human errors resulting from such things as inexperience, greed, misplaced loyalty, bad hair days, and hubris. If that is not possible, then nuclear safety is not possible.