Inside Fukushima

More images at NGS and Gawker.

Years ago I worked in the defense software industry.  We worked indirectly on various things related to flight control, radar, space flight and so on.

During that time I had a lot of opportunity to talk to and work with people that were developing front line systems - flight avionics (things having to do with controlling an airplane in flight).  In those days (the early 1980's) there weren't flight simulators to test things on before the airplane, missile or helicopter first flew.  Things had to be designed and built right "out of the box".  The flap controller on a commercial airliner, for example, had three completely separate computer systems to interpret the motion of the flap controls.  At least two out of the three had to agree on what was to be done in order for the flaps to work and the system would work if only one computer was functioning.

Now these systems were well defined and "closed" in the sense that input, e.g., from controls operated by pilots during flight, was limited.  The flaps would only move so far up and down, the controls only so far back and forth, hence there were limits as to what the inputs and outputs might be.

People that worked for me in those areas eventually moved on and did some work for a local nuclear energy company working on nuclear reactor simulators.

The idea behind the simulators was to create the experience of operating a nuclear reactor without any of the danger of melting it down.  Now, as far as I now today, and certainly not in the 1970's when Japan's reactors were built, computer do not operate nuclear reactors.  People do.  The images of vast control rooms with thousands of dials, controls and gauges on the walls, were set up for operators to control the reactor directly.  So the simulators provided responses to the input of the operators.

Now on the flight control side, as I said, things were fairly limited.  The limits, for example of how much stress the wing would take, were calculated and tested outside the flight software.  (Somewhere there are videos of the first Boeing 747 wing being stressed tested to failure.  This test tells the engineers exactly how much stress the wing an endure during flight.  The wing fails in the test at just about the point the engineers predict.)  Things like the flap controllers are designed to prevent the airplane for operating in such a way as to not even come close to putting that much stress on the wing.

The aerodynamics of flight are a fairly well understood engineering discipline these days.  For example, its possible to calculate  the position of the plane in any positions (yaw, pitch, roll) and to calculate airflows over the surfaces of the wings and determine how the plane will (or will not) recover if placed in that situation.  Plus, in flight, a pilots experience (as well as gravity) help to right any wrong.

On the nuclear side, things are not so clear cut - particularly in training.   There are hundreds or thousands of valves, controls, and sensors in a nuclear reactor.  There are engineering modifications, particularly as in Fukushima, which have been made over four decades of operation.  There is staff turn-over.  There are numerous regulations and requirements placed on the design, function and operation of the plant.  There are engineering issues (errors, omissions, design changes) that impact the functioning of the plant and its systems.

Simulating this is not so easy.  Sure its easy to simulate things as they should be.  But what about an incorrectly connected or broken sensor.  (Three mile island, for example, began with a problem in a secondary (non-critical system) and was followed by a "stuck valve" which operators failed to recognize.)

How would this be represented in a simulator?  It can only be simulated if someone can A) figure out that it would be a realistic problem, B) determine what would happen should it occur, and C) accurately predict how to fix it.

Its very unlikely anyone would be able to create or guess all possible such scenarios and program them into a simulator.

The bottom line here is that something like a nuclear power system is very unpredictable and very prone to human error.  Not just operational error but systemic error in the sense that bureaucratic meddling over time in terms of design, function, safety, etc. create a compounding of unpredictability that no simulator can express. 

Leaving operators to guess at problems and fixes - just like Three Mile Island.

So in Fukushima you have the same issue, except compounded at a national level - regulators,  companies, people, protestors, all meddling in a forty year old design and creating an unsafe safe system.

(Simple explanation of the Fukushima problems and why the three safety systems failed here.)