Take a step back and contemplate the sampling of nuclear wreckage that has been laid out before you in the ten chapters of this too-brief narrative. You can see patterns developed in this matrix of events. There are hot spots, imprints, and repetitions. The markings are all over the developed world, left there by one very large experimental program that was trying to improve the lot of mankind, and not to destroy or degrade it. The boldness of this long-term program can raise an eyebrow or two, but from an engineering standpoint it was new, exciting, unexplored territory. In all, it killed fewer people than the coal industry, it caused less unhealthy pollution than the asbestos industry, and it cannot be blamed for global warming.[280] In the last chapter, we will discuss what it all means and reach for what conclusions there are.
Chapter 11:
Caught in the Rickover Trap
“If a man fires at the past from a pistol, the future will fire at him from a cannon.”
It was summer, 1985. We had successfully conceived, designed, birthed, programmed, built, worried over, installed, documented, and exhaustively tested the Safety Parameter Display System for the two General Electric BWR/4 reactor units at Georgia Power’s E. I. Hatch Nuclear Power Plant down in Appling County, Georgia. It was time to celebrate! Instead, the three of us, I, Jeff Hopper, and Mark Pellegrini, retired to our rat-hole motel in Hazlehurst, over in the adjacent Jeff Davis County, to catch up on our sleep.
There is seldom a great deal of civilization near a nuclear plant, and it was a half-hour to Hazlehurst, which was the closest place with a motel. I found my room, collapsed in bed, and extinguished the light. As soon as the springs stopped groaning, I noticed a periodic chirping noise. Eek… eek… eek… eek. Was it a cricket, or was it a dry bearing in the air conditioner blower? My attention was riveted to the sound, and I could not help but try to analyze it. Having just come down off an extended computer system verification exercise, I found that I had to submit it to a test. I turned on the light. The chirp stopped instantly. Yep. It was a cricket. Problem solved. Lights out. I drifted off to sleep and suffered through an incubus of an enormous squirrel-cage fan with a dry bearing, moving closer and closer.
These computer systems were built to satisfy new control-room instrumentation requirements set forth by the Nuclear Regulatory Commission in NUREG-0696, “Functional Criteria for Emergency Response Facilities,” written in response to inadequacies discovered in the TMI-2 disaster of 1979. Our sparkling new systems, bolted strongly to the control-room floor, each collected 128 analog and 512 digital data in real time from each reactor system and displayed them on demand using color video monitors. The equipment had been shake-tested on an earthquake simulator in California, run continuously for years in our lab back in Atlanta, and endured handling by inexpensive student labor from Georgia Tech. It was as solid and glitch-free as a dedicated team of engineers could possibly make it. I was overflowing with confidence that it would perform perfectly, as I would remind anyone standing near it.
The next day began an adventure that went down in the history of nuclear power as the “Rat Cable Problem.”
We wanted to give the systems ample time to run before we returned to examine the overnight performance logs, so we lingered over breakfast and retired to our rooms to relax. I’m not sure what the other guys did, but I watched cartoons on the television. That afternoon, we drove to the plant. Pellegrini fretted over the party that the control room operators must be planning for us.
The parking lot at the plant was blazing hot, with the power of the sun reflected back at us, making the air seem to sizzle and boil as we walked the distance. There was some race of enormous beetles living on the plant site — they were huge black things with horns. Pellegrini feared that they were mutants caused by irradiated bug-DNA. I scoffed, but they were big, and they were marching eight deep in a line across the parking lot, toward the river, moving perpendicular to our trajectory as we made a straight line for the guard shack.
They were a determined bunch of insects, keeping a disciplined column at a constant speed on the egg-frying concrete. We were equally determined. Our tracks were about to cross, and I was not about to hesitate as we intersected the beetle path. My eyes locked on an individual. He was not going to hesitate either. Estimating his speed and mine, it looked as if he would be walking either over the top of my right shoe or under it. Crunch. The big insect’s life ended suddenly, right under my foot. It was a bad omen. I wondered silently as we completed the hot trek. Why do silly omens seem to gain importance as we draw nearer to an active fission process?
We reached the electronics bay adjacent to the control room. Hopper turned on the maintenance terminal and started reviewing the system’s performance over the last 18 hours. It was bad. The error log was filled with digital data reception failures, and the size of the log was growing as we watched. It was not every reception. It was just every once in a while, a parity check would fail.[281] It could go a minute or an hour between errors, or a burst of errors could occur, as the digital data set was polled once per second, and, even more unnervingly, the failures appeared to be completely random and did not follow any particular pattern or appear to be connected to any one particular system failure.
The digital data were collected using a wired loop running all over the plant and passed on to the computer systems by a Cutler Hammer Directrol Multiplexer Communications Station, bolted to the back wall of the electronics bay. A fat, multi-conductor cable (no. 5769) ran between the Directrol and the ROLM 2150 I/O expansion box. Both the Directrol and the expansion box had been exercised continuously back at our shop in Atlanta for years without a single error. Had something broken at the plant? We decided to let it keep running to see if the system self-regulated, and also so we could collect more data and try and figure out what the problem was.
The errors would not stop coming. The next day, the error log was in crisis mode, and it was eating up all the reserve space on the disc. Another week of this, and the working space on the disc would be gone. The operating system would lock up. The log was now larger than the 30-minute sliding record of everything in the plant that was held on the disc as a “black box” recording.[282] Why? It had to be the only one difference between the development setup in our shop and the operational installation in the power plant.
Back at Georgia Tech, we had built a special room in the high-bay of the Electronics Research Building, complete with enhanced air conditioning and a raised computer-room floor. The room was a good replica of the Plant Hatch electronics bay, except for one detail. There was an open pipe through the wall, leading to the outside. It had been added to the building so that power lines from a gasoline-powered 400-hertz generator could be connected. The pipe was not in use, except as an access walkway for vermin. Somehow, a brown rat, rattus norvegicus, had found our lab, and he visited often. By the time we discovered him, he had chewed up the insulating sheath on our very expensive cable no. 5769, connecting the ROLM I/O box to the Directrol. Why he so enjoyed the taste of plastic insulation on that particular cable, I could not comprehend. It had zero food value, but he made a meal of it.[283]
280
Statistically, for every person who dies as a result of the generation of electricity by nuclear means, 4,000 people die as a result of fine-particle pollution caused by burning coal. Between 1972 and 2002, 730,000 lawsuits were filed because of asbestos dust inhalation, costing the asbestos industry $70 billion. The deaths caused by asbestos are in the half-million-people range. Global warming is believed to be caused by a buildup of carbon dioxide in the atmosphere, caused by our digging or drilling up and burning carbon that was sequestered underground. This carbon was geologically kept out of the total carbon inventory on the surface, but the need for power in the industrialized world has made it necessary to release energy any way we can. Most of the extra carbon dioxide comes from burning coal, gasoline, and natural gas. Nuclear power generation does not oxidize anything, but derives energy from nuclear fission. Some have pointed out that building a nuclear plant requires a lot of fuel to pour concrete and workers drive to the plant in cars, but these are ridiculously weak arguments against non-polluting nuclear power.
281
Digital data consisted of the open/closed status of valves and the on/off status of electrical motors and solenoids, collected at nodes all over the plant. Each datum was a 16-bit digital “word” with an extra parity bit. If the binary bits in the 16-bit word added to an even number, the parity bit was set to a one. If the addition was an odd number, the parity bit was zero. Upon reception back at the computer, the binary bits were added up, and the odd/even character of the resulting number was compared to the parity bit. If the results did not agree, then the datum had been corrupted somewhere, having either dropped or gained an odd number of spurious bits. This condition was logged as an error, and a repeat transmission was requested.
282
We started the project specifying a 32-megabyte hard disc. Remember, this was the early 1980s, and 32 megabytes was considered to be a large data-storage capacity. It was housed in a box that fit in a 19-inch rack. Casey Lang, head of the software development group, came to me requesting a 50-megabyte upgrade to the specification. That was huge! It was hard to see how we could ever fill a monstrous 50-megabyte bucket, but I changed the order. It is amazing how data capacities have drifted up in the decades since then.
283
Having had 30 years to think about it, I now believe that the rat’s interest in the plastic cable insulation was caused by salt left on the surface by people having handled the cable. Human sweat contains a lot of salt, and when the water evaporates, the salt is left on the surface. Both ends of the cable had been handled many times, both at the factory in Cupertino and at the application point in Atlanta.