Prototype Testing, Part 1.

Prototypes should be tortured, not licensed.

The Gordian Knot Group is an unabashed advocate of market based regulation of nuclear power, which we call Underwriter Certification. The market is by far the most efficient way to balance nuclear power benefits and safety. But this will not happen automatically. The government has a critical role in making this work. A quick glance at our outline of Underwriter Certification will reveal a multi-page list of national government responsibilities, while the other players only merit a paragraph or two. Among those national government responsibilities is provision of a protopark. What does this mean? This is the first in a three-part series outlining how new nuclear designs should be tested.

Figure 1. Loading Breeder Reactor into Fermi Containment

In 1956, Detroit Edison proposed building a 100 MWe sodium cooled, fast breeder at the western end of Lake Erie between Detroit and Toledo, at a place called Lagoona Beach. At that time the only US fast breeder was the 1.7 MW EBR-1 which had just suffered an unexpected excursion and partial meltdown. The problem was traced to unanticipated bowing in the core elements. Yet the AEC was pushing ahead with Lagoona Beach.

The Advisory Committee on Reactor Safety(ACRS) wanted a much bigger prototype tested at a remote location. The AEC argued that was unnecessary. Any problems could be handled by modelling and sub-system tests. The head of the ACRS sub-committee looking into Lagoona Beach was a guy named Harvey Brooks. In a letter to the AEC, Brooks wrote:

In any technology as new and untested as that of the sodium cooled reactors, there are likely to be serious surprises which were not anticipated by the designers. Experience indicates that such surprises always occur in connection with any new development, even when the technology is much more thoroughly tested than in the case of the fast reactor. Many of these surprises can be matters of apparently trivial detail which may nevertheless seriously influence the safety or operation of the reactor. The purpose of a prototype is primarily to minimize the possibility of such surprises rather than to find the answer to specific technical questions which are anticipated now, and which presumably can be answered on a piecemeal basis by experiment and theory. [Emphasis mine.]

The AEC went ahead and approved the reactor anyway. The plant, called the Enrico Fermi Nuclear Generating Station, encountered a series of problems including a coolant blockage which shut down the plant for four years. The safety systems operated properly and there was no radiation release. But the accident created a great deal of publicity and a strongly anti-nuclear book with the catchy title of ``We Almost Lost Detroit". The plant never operated at full power and was a commercial disaster.

The GE Boiling Water Reactor Mark 1 containment was duly approved and licensed by the AEC. But when the Germans decided to do a full scale test at Wurgassen in April, 1972 by opening up all eight steam relief valves at the same time, as was supposed to happen in a real casualty, the suppression torus started oscillating, banging back and forth, and badly damaging the reactor. Something similar happened when the Swiss tried a similar test a few months later. The AEC deemed such testing was unnecessary.

The Super-Phenix was one of the most thoroughly analyzed nuclear systems ever constructed. But during start-up, it became evident that the control rod worth, the ability of the control rods to stop the chain reaction, had been greatly over-estimated.

A promising nuclear technology is the molten salt reactor (MSR). MSR's combine low pressure, high temperature, with a liquid fuel which can be moved around with a pump and passively drained in an upset. One of the uncertainties with this concept is the amount and location of plate out of the fission products, which could build up in heat exchangers. The computer cannot help us much here. The only way to find out is long term, full scale testing.

And by testing I mean not only tests that confirm that the design behaves as predicted under normal operation; but physical imposition of the upsets and failures that the designers claim the plant can handle. If and only if the design passes these tests, can underwriters/regulators have confidence in the design, and that design be insured/licensed.

Test, then license. This is simple common engineering sense. But the NRC policy is: don't test, but license anyway. This puts sensible developers of new nuclear technology in an impossible quandary. They know they need rigorous, fullscale prototype tests to prudently license their new technology. But they can't do the tests without a license. Unless this Catch 22 is eliminated, new nuclear is going nowhere.