Japan Earthquake: Nuclear Plants at Fukushima Daiichi

[MadderDoc]

I think the discussion about terminology does belong in the political thread. I do NOT believe that the stuff about "what were they thinking" belongs in the management and government performance thread. What the crisis managers thought had real consequences on the engineering front and moreover interpreting the data is kinda what we do here.

I would also like to extend, again, an appeal to all of you to please use the Title: box.

Just to add a snippet from an IAEA OSART mission to a Japanese NPP, indicating the technical aspects relating to the plant which were examined by the IAEA. Note item 4.

"MAIN CONCLUSIONS
The OSART team concluded that the managers and staff at [name omitted] NPP are
very enthusiastic in their commitment to improve the operational safety of the plant. The
team found good areas of performance, including the following:
- Open, professional and productive approach to the OSART mission and willingness to
learn and improve
- Excellent material condition and housekeeping of the plant
- Cooperation with contractors and long term partnership
- Respect for the public and region and comprehensive programme to improve public
confidence
- Training facilities including simulator exercise reviews
- Very good comprehensive emergency exercise with involvement of both on-site and off-
site organizations"

 

[MadderDoc]

Unit 3 water injection early days (was Re: Japan ..)

I can't seem to recall exactly what line they were using to inject water at that time. Do you remember?

That's a very good question. The line, the injection point, the feed source of water. I remember being frustrated by some lack of data last time I looked into it. I'd better check, rather than taking this from memory.

Edit: Hum. This is going to take some time, zz.

 

[zapperzero]

Edit: Hum. This is going to take some time, zz.

Sorry. I'd have checked myself otherwise. I am swamped with work.

 

[MadderDoc]

Sorry. I'd have checked myself otherwise. I am swamped with work.

No worries, I'm not hung up and actually enjoy digging into it.

 

[tsutsuji]

http://www.tepco.co.jp/cc/press/betu12_j/images/120512j0101.pdf Establishment of an Implementation Plan on Reliability Improvement Measures at Fukushima Daiichi Nuclear Power Station (275 pages, Japanese)

Just a glimpse:

page 89/275

page 90/275

 

[MadderDoc]

I can't seem to recall exactly what line they were using to inject water at that time. Do you remember?

Water injection to the Unit 3 RPV would have been to the Fire Extinguisher System Line (FESL). As regards _unit 3_ until March 28th pumping would still have been by fire trucks, configured similarly to the setup by the evening of March 14th. (Edit: until March 26th, when fire trucks started injecting fresh water) Here's a schematic of that configuration cropped from http://icanps.go.jp/eng/120224Siryo04Eng.pdf :

 

[tsutsuji]

http://www.tepco.co.jp/nu/fukushima-np/images/handouts_120516_07-j.pdf Soundness of unit 4 reactor building (Government-Tepco mid-long term meeting, May 2012) (22 pages, Japanese)

 

[jim hardy]

would there be possible failure modes, with perhaps some unintended electrical interconnection within such a junction box, that could concertedly send the readings of some sensors downscale and others upscale?

Yes, moisture can do that.
The thermocouple effect produces tens of microvolts per degree so you translate signals of a few DC millivolts into temperature.
Should the wires get wetted by water that's conductive it makes a battery via galvanic cell action - different metals in presence of an electrolyte.
One of the metals would be one of the theromocouple wires and the other metal might be the other thermocouple wire or it might be nearby metal like the inside of conduit or junction box.
Galvanic cells make tens if not hundreds of millivolts , hundreds of times more than the meager millivolts you get from thermocouple effect.
So the temperature signal gets drowned out. :(
and can give you either polarity depending on the chemistry at the wet spot.

So wet thermocouples are not reliable.

what voltage would correspond to readings at about minus 130 deg C give or take some. Conspicuously many of the obviously faulty readings are in that range.

negative four to six millivolts DC.
I don't reemember which type thermocouples those were, for some reason i thought iron-constantan. Tsutsuji posted it once and i remember thinking "I'll never forget that" but i have.

Anyhow Omega teaches most engineers about thermocouples via their excellent books.

Here's their thermocouple intro page
http://www.omega.com/techref/thermcolorcodes.html

and millivolt vs temperature tables are accessible here
http://www.omega.com/thermocouples.html

clicking iron constantan opens a pdf table
type J(iron-constantan) is -5.8++ millivolts at -130 degreesC
similarly type T (copper constantan) is -4.2++.

as you see it's low millivolts irrespective of thermocouple type.

The TEPCO technicians went to great lengths to correct readings for damaged leadwires
but only they know how much faith to put in their results.
I had some luck with damaged thermocouples that were wetted right at their sensing end
but never with ones wetted midway.
Probably those guys were better than me. Certainly i never worked under the pressure they did..

hope this helps.

old jim

PS possible reason for lots of ~-130 degrees readings

it is common practice to interpose a device called transducer that translates a peculiar signal into a standard range. That allows use of a cheaper computer front-end(ADC) and kmakes things easier for the programmers.
A common range has max to min ratio of five, like is 1 to 5 volts or 4 to 20 milliamps or 10 to 50 .
So a bunch of thermocouples might be handed to the plant computer with 1 to 5 volts representing zero to 500 degrees ( just random pick, i have no idea how GE computer is scaled) .
When power is lost to the transducers they all go to zero volts which (with linear conversion) would represent -125 degrees . Computer probably adjusts for non-linearity of thermocouple so -130 degrees is a quite possible reading for thermocouples with expected range of ~500 degrees that are reporting maximum downscale temperature.
The five to one turndown is intentional so that a failed transducer or one that's lost power drives the indication 25% below bottom of scale.
You wouldn't want a loss of power to your instruments to put all your meters midscale - you'd want them downscale so it's obvious what happened. That's why zero center meters are rare in control boards.

Sorry for digression. But it's important to know what an instrument is really saying . They seldom lie we just misunderstand them.

 

[MadderDoc]

<..>Sorry for digression. But it's important to know what an instrument is really saying . They seldom lie we just misunderstand them.

Thank you jim, that was very helpful, and not least the last bit, which rings true, looking at data. In unit 3 it appears to be the case, that somewhere around March 20th something started interfering with readings, and under circumstances such that wetting of leads or junction seems likely.. This resulted in a period in gross instability of readings, and indeed readings which would seem physically impossible under the circumstances After that, there was a period of apparent gradual revival of sensors, i.e. readings wandered within the range of the possible to stay there albeit still with huge variations up and down.

From what you have told me this would seem to me to be true statements:

a) we have reason to believe the readings until the period of interference were still reasonably well calibrated.
b) during the period of interference, we can have no confidence that readings are accurate
c) during the period of gradual revival, we can have no confidence that readings are accurate
d) during the period after revival, we can have no confidence that readings are accurate
e) during the period of gradual revival we can have no confidence that the trends up or down in readings reflect reality
f) during the period after revival, we can have some confidence that trends up or down in readings reflect reality.
g) there would have been no way of recalibrating the sensors during the accident.
h) there would be no way -- later -- to assess the magnitude and direction of inaccuracy of the readings recorded during the accident.
i) the magnitude of inaccuracy we would be looking at might be in the XXXXXL league.

 

[zapperzero]

Water injection to the Unit 3 RPV would have been to the Fire Extinguisher System Line (FESL).

Thank you!

 

[MadderDoc]

<..>
PS possible reason for lots of ~-130 degrees readings

it is common practice to interpose a device called transducer that translates a peculiar signal into a standard range. That allows use of a cheaper computer front-end(ADC) and kmakes things easier for the programmers.
A common range has max to min ratio of five, like is 1 to 5 volts or 4 to 20 milliamps or 10 to 50 .
So a bunch of thermocouples might be handed to the plant computer with 1 to 5 volts representing zero to 500 degrees ( just random pick, i have no idea how GE computer is scaled) .
When power is lost to the transducers they all go to zero volts which (with linear conversion) would represent -125 degrees . Computer probably adjusts for non-linearity of thermocouple so -130 degrees is a quite possible reading for thermocouples with expected range of ~500 degrees that are reporting maximum downscale temperature.
The five to one turndown is intentional so that a failed transducer or one that's lost power drives the indication 25% below bottom of scale.
You wouldn't want a loss of power to your instruments to put all your meters midscale - you'd want them downscale so it's obvious what happened. That's why zero center meters are rare in control boards. <..>

My understanding is that under normal operation they would have the signal leads from the temperature sensors connected to a set of multi-channel recorders. The output would come on paper, in the form of differently colored pen traces. The paper would be scaled, for each group of sensors with a convenient scale for that group of sensors, say 0-500^oC.
The recorder would be equipped to do the necessary signal conversion, perhaps including transduction, offset calibration, linearity correction. you can probably think of more.

The data set I am looking at though, is from after all that functionality was lost after the tsunami on March 14th, with the power to the recorders, and these data represent the first temperature measurements emerging from Unit 3 after all the explosions at the plant. The earliest data are from March 19th. The output appears as discrete measurements, and indicate much more of a human hand was involved in the production of it. From looking at the data set I get an impression of men with notebooks at irregular intervals reading sets of signals, with those signals then on the spot or with some delay converted by some method to a temperature value. There is a 36 hour gap in data from about noon on March 21st, as if the men with notebooks during that period were otherwise engaged, or were not able to be at the spot where the signals could be read. I would think the signals would have been most likely read from the disconnected leads from the sensors, at the recorders in the Unit3+4 Main control room.

I am trying to imagine what the conversion from signal to final output in this scenario would involve, and would appreciate your thoughts on it.

 

[jim hardy]

There is a 36 hour gap in data from about noon on March 21st, as if the men with notebooks during that period were otherwise engaged, or were not able to be at the spot where the signals could be read. I would think the signals would have been most likely read from the disconnected leads from the sensors, at the recorders in the Unit3+4 Main control room.

I am trying to imagine what the conversion from signal to final output in this scenario would involve, and would appreciate your thoughts on it.

does that gap align with the temporary evacuation?

If the wires they accessed are thermocouple wire they can be read directly with a hand held multimeter. Better ones will accept thermocouple wire and do the conversion to temperature for you.

If there is an interposing transducer they would have had to power it up (probably with car batteries?) and read its output with a DMM and convert to temperature. That's an easy sliderule , well nowadays pocket calculator, operation. But in early days working in respirators and maybe double gloves with flashlights - it's unimaginable.

I am tying to remember the details of those drawing Tsutsuji posted, specifically whether there was a remote "reference junction" where the wires transition from TC to copper. If so that's one more calculation but instrument technicians are well versed in it. Will go back and look for Tsutsuji's remarkable finds on that subject.

So the readings are not difficult to get, in concept.
Fortunately thermocouples are extremely rugged devices. Wet insulation is their achilles heel.

Electronic transducers are good until they've gotten probably a few tens of thousands of rads (hundreds of sieverts) . If there are any they must be outside the worst radiation areas.

old jim

 

[MadderDoc]

does that gap align with the temporary evacuation?

I wouldn't say align, but the gap does include an event with black smoke coming out of the reactor during the afternoon/evening of March 21st, when most employees were temporarily evacuated to a safe place, leaving only a minimal crew at the site. However the gap continues for most of the next day, until late evening on March 22nd.

I am tying to remember the details of those drawing Tsutsuji posted, specifically whether there was a remote "reference junction" where the wires transition from TC to copper. If so that's one more calculation but instrument technicians are well versed in it. Will go back and look for Tsutsuji's remarkable finds on that subject.

I think this is the post you are looking for:
https://www.physicsforums.com/showpost.php?p=3765763&postcount=12381

 

[MadderDoc]

Unit 3 steam outlets and hotspots

The thermographic image below was taken by SDF in the morning of March 20th 2011 during the SFP3 brouhaha. It is likely still after all those months the best evidence available to us, to show the location of steam outlets and hotspots of the top of the building of Unit 3. It was with evidence like this in hand Tepco maintained before the public their estimate that the steam plumes coming from the building was all coming from the pool, and not from the reactor.

I have outlined the approximate extent of the spent fuel pool to the left in the image. The measured apparent temperature of the pool was 62^oC (144^oF) at the time, according to the source.

The white arrows point to the two main steam outlets from around the reactor area, one outlet is close to the gate to the spent fuel pool, and another source at the opposite side of the reactor area, at the gate to the equipment pool. This latter source had an apparent temperature of 128^oC (262^oF). These two steam outlets from the reactors were in fact in the main responsible for the observed twin steam plumes from the building at the time. However the image shows also other interesting steam outlets and hotspots outside the area of the spent fuel pool, some of which probably contributed to the plumes.

Source: The Evaluation Status of Reactor Core Damage at Fukushima Daiichi Nuclear Power Station Units 1 to 3

 

[jim hardy]

I think this is the post you are looking for:
https://www.physicsforums.com/showpos...ostcount=12381

That's it ! THANK YOU

Signal arrives on Copper-Constantan.. type T. I have bookmarked the link.

they can read directly in degrees with a good DMM.

Here's the T millivolt table. http://www.omega.com/temperature/Z/pdf/z207.pdf

If you only have an ordinary DMM that doesn't handle thermocouples,
then you must compensate for "cold junction" and convert from millivolts to temperature:
Measure the incomiing millivolts,
look up millivolts for temperature of your connection point to the TC wire(room temperature),
add that to measured millivolts,
find in table temperature corresponding to that sum.
Sounds harder than it is, it quickly becomes second nature.

Thanks again

old jim (absent minded)

 

[MadderDoc]

<..>
If you only have an ordinary DMM that doesn't handle thermocouples,
then you must compensate for "cold junction" <..>
Sounds harder than it is, it quickly becomes second nature.

compensate for "cold junction" .. because the mere connection of the DMM to the sensor leads creates two additional metal-metal junctions? (say, Cu/Ni and constantan/Ni, if the grips of my DMM are Ni-plated)

 

[jim hardy]

compensate for "cold junction" .. because the mere connection of the DMM to the sensor leads creates two additional metal-metal junctions? (say, Cu/Ni and constantan/Ni, if the grips of my DMM are Ni-plated)

Yes, that's right.

When you get right down to it a thermocouple is a differential device.
It reports the difference between its hot and cold ends.

It is natural to assume the EMF is developed at the hot tip but that's not so.
The EMF is developed along the wire in regions where there's a temperature gradient.

Here's my oversimplification of it:
Because the different metals have different 'affinity' for electrons, a temperature gradient shakes charge toward the cold ends of the two wires with different EMF's. So equal ΔT produces unequal ΔV in the two wires. You read that difference.

If you place a thermocouple on the bench and connect a meter you get no millivolts.
When you change the temperature of either end a voltage appears.

Here's Omega's introduction:
the text around figure 4 describes what you said.

http://www.omega.com/temperature/z/pdf/z021-032.pdf

again sorry for the digression.

 

[MadderDoc]

<..>
Here's Omega's introduction:
the text around figure 4 describes what you said.

http://www.omega.com/temperature/z/pdf/z021-032.pdf

again sorry for the digression.

Nah, it is great! (I'll head straight for my messy-box, should I have a piece of constantan wire somewhere I can play with, heh :-)

 

[jim hardy]

Page 93 of those DAEC drawings has a sketch, albeit not detailed, of that deck.

The passage way to dryer pool is as you'd expect wider than the one to SFP.

I couldn't make a lot out of trying to overlay them mentally.

I am curious how main steam piping is routed out of vessel and where are safety valves relative to your steam plumes.. for they don't quite line up with those passageways to pools.
The DAEC drawing pages 4 and 5 is so crowded it's hard to read.

 

[tsutsuji]

http://www.tepco.co.jp/nu/fukushima-np/images/handouts_120516_07-j.pdf Soundness of unit 4 reactor building (Government-Tepco mid-long term meeting, May 2012) (22 pages, Japanese)

An English pdf is provided on the same date on the same topic, but the contents, although overlapping for some part, are not the same:

http://www.tepco.co.jp/en/nu/fukushima-np/images/handouts_120516_05-e.pdf Soundness Verification of Unit 4 Reactor Building at Fukushima Daiichi Nuclear Power Station (9 pages only)

http://www.tepco.co.jp/nu/fukushima-np/images/handouts_120516_07-j.pdf page 19/22 links to http://www.nisa.meti.go.jp/shingikai/800/25/003/3-3-4.pdf dated 28 October 2011, which is a presentation about the JNES studies completed in May and in July to cross-check the Tepco studies on the earthquake resistance of units 3 and 4.

Edit: The English pdf mentioned above is a translation of yet another Japanese pdf, that one : http://www.tepco.co.jp/nu/fukushima-np/images/handouts_120516_06-j.pdf (9 pages)

 

[MadderDoc]

Page 93 of those DAEC drawings has a sketch, albeit not detailed, of that deck.

This sketch is more like the deck of the Fukushima reactors no.2-4:

I am curious how main steam piping is routed out of vessel and where are safety valves relative to your steam plumes.. for they don't quite line up with those passageways to pools.
The DAEC drawing pages 4 and 5 is so crowded it's hard to read.

I am not sure one can necessarily conclude much from the position of the hotspots. There is a leaking PCV down there under several meters of more or less cracked up concrete shield and the steam leaking out likely just take the easiest way out. There could also be weird chimney effects.

ps. jim, please check your method of quoting/replying (are you using the quote button?) The title of the sub-thread seems to be reset to the title of the main thread when you reply.
pss. I found the constantan wire, cheers :-)

 

[MadderDoc]

Building soundness testing (was Re: ..

An English pdf is provided on the same date on the same topic, but the contents, although overlapping for some part, are not the same:

http://www.tepco.co.jp/en/nu/fukushima-np/images/handouts_120516_05-e.pdf Soundness Verification of Unit 4 Reactor Building at Fukushima Daiichi Nuclear Power Station (9 pages only)

tsutsuji, the 9 pages English version mentions a method for tilt/skew testing applied to the walls of the buildings, the Japanese version does not seem to mention it. Have you encountered any results of the soundness testing using that method?

 

[tsutsuji]

tsutsuji, the 9 pages English version mentions a method for tilt/skew testing applied to the walls of the buildings, the Japanese version does not seem to mention it. Have you encountered any results of the soundness testing using that method?

I had not realized that there were actually two Japanese pdfs about unit 4's reactor building. So I edited my previous post adding:

Edit: The English pdf mentioned above is a translation of yet another Japanese pdf, that one : http://www.tepco.co.jp/nu/fukushima-np/images/handouts_120516_06-j.pdf (9 pages)

I think it is the first time that I hear about testing the walls that way.

I also found the following press release in the "Important report from Tepco" section of their website:

http://www.tepco.co.jp/en/nu/fukushima-np/info/images/120426_01-e.pdf "We affirm that the Reactor Building and Spent Fuel Pool of Unit 4 will not collapse in the event of an earthquake" (dated 26 April 2012) where they talk about their plan to use "non-destructive inspection" such as Schmidt hammer.

http://www.tepco.co.jp/en/nu/fukushima-np/images/handouts_120515_02-e.pdf "May 14: While installing the data logger (data collection equipment used to measure the direct current resistance of thermometer in Unit 1), we found that the wiring for safety valve 4B thermometer (TE-261-13B) and safety valve 4C thermometer (TE-261-13C) connected to the digital recorder in PCV was done in a opposite manner. At 7:12 PM on the same day, the connection was redone properly. The cause of this miswiring is currently being investigated. These temperatures are not included in the monitoring list of the technical specification (Article 138 and 143)."

 

[NUCENG]

Page 93 of those DAEC drawings has a sketch, albeit not detailed, of that deck.

The passage way to dryer pool is as you'd expect wider than the one to SFP.

I couldn't make a lot out of trying to overlay them mentally.

I am curious how main steam piping is routed out of vessel and where are safety valves relative to your steam plumes.. for they don't quite line up with those passageways to pools.
The DAEC drawing pages 4 and 5 is so crowded it's hard to read.

Main steam piping exits the vessel below the RPV Head flange. The piping nozzlkes are spaced 90 degrees apart and are then routed down to the first floor and become parallel as they exit the drywell to the steam tunnel to the turbine building. One set of MSIVs (inboard) are located inside the drywell and the second outboard MSIVs are in the steam tunnel.

Safety and relief valves are on the main steam piping near the vesssel. Most of them have piped discharge to the torus. Some BWRs have unpiped safety valves that discharge to the drywell, but these are the valves with the highest relief pressure setpoints. From what I have read the Safety Relefs Valves that were opened were probably all piped to the torus.

 

[jim hardy]

main steam piping

Main steam piping exits the vessel below the RPV Head flange. The piping nozzlkes are spaced 90 degrees apart and are then routed down to the first floor and become parallel as they exit the drywell to the steam tunnel to the turbine building. One set of MSIVs (inboard) are located inside the drywell and the second outboard MSIVs are in the steam tunnel.

Safety and relief valves are on the main steam piping near the vesssel. Most of them have piped discharge to the torus. Some BWRs have unpiped safety valves that discharge to the drywell, but these are the valves with the highest relief pressure setpoints. From what I have read the Safety Relefs Valves that were opened were probably all piped to the torus.

Thanks Nuceng

so they'd all have to be inside drywell (or PCV whichever's right name...) i assume they're up high, very near the vessel penetrations?

Thaks MD, note title block :) :)

 

[MadderDoc]

Miswired sensors (was.. Building soundness..)

<..>
http://www.tepco.co.jp/en/nu/fukushima-np/images/handouts_120515_02-e.pdf "May 14: While installing the data logger (data collection equipment used to measure the direct current resistance of thermometer in Unit 1), we found that the wiring for safety valve 4B thermometer (TE-261-13B) and safety valve 4C thermometer (TE-261-13C) connected to the digital recorder in PCV was done in a opposite manner. At 7:12 PM on the same day, the connection was redone properly. The cause of this miswiring is currently being investigated. These temperatures are not included in the monitoring list of the technical specification (Article 138 and 143)."

Looking up in the latest reliability testing from May 1st, all that can be seen is that -- compared to the other two TE-261-13's -- TE-261-13A has returned clearly higher readings, and appears to be more sensitive to temperature change. This is not quite what I'd expected to see, but perhaps there's something I am missing.

I interpret Tepco's explanation that 'the wiring for safety valve 4B thermometer (TE-261-13B) and safety valve 4C thermometer (TE-261-13C) connected to the digital recorder in PCV was done in a opposite manner' to mean both of the sensors had been connected with reversed polarity, whereas TE-261-13A has been connected correctly.

This is not the first time we hear about miswired sensors.

 

[MadderDoc]

I think it is the first time that I hear about testing the walls that way.

Measuring the verticality of the walls would seem to be suitable to answer the valid question whether the building has been flexed to the side, something which the water level measurements in the reactor well and the SFP might not be able to detect.

 

[tsutsuji]

I interpret Tepco's explanation that 'the wiring for safety valve 4B thermometer (TE-261-13B) and safety valve 4C thermometer (TE-261-13C) connected to the digital recorder in PCV was done in a opposite manner' to mean both of the sensors had been connected with reversed polarity, whereas TE-261-13A has been connected correctly.

I think they mean that the TE-261-13C thermocouple was connected to the TE-261-13B digital recorder and vice versa.

 

[jim hardy]

pss. I found the constantan wire, cheers :-)

put some water in a coffee cup and stick the two wires in it. Observe voltage when they're touching vs separated. A few grains of salt will help your water conduct the current necessary to drive your meter. Then try constantan vs a galvanized nail, a strip of aluminum foil, an iron paperclip, and of course a piece of copper wire..

Cheers also !

old jim

 

[MadderDoc]

I think they mean that the TE-261-13C thermocouple was connected to the TE-261-13B digital recorder and vice versa.

Yes, thank you, that must be it. I can see the switch back to correct on May 14 has now left a discontinuity in the 6 hour data series for those two sensors. I suppose Tepco will have logs to determine when the incorrect connection has been done, and so retrospectively will be able to correct the designation of the data. There is no other obvious discontinuity to be seen in data as far as it goes back in time, so the faulty connection would have happened either before that, or at a time when the two sensors returned similar readings.

 

[MadderDoc]

Of most considerable interest is the area around the gate to the equipment pool. It was this area of the building, which had
- the public's eye intently kept focused away from it,
- the hottest hotspots by far,
- the massive north major twin plume from the building originating from it, and was
- the likely source point of the black 'this is not styrofoam' smoke events of March 21st and March 23rd.

To the March 20th thermographic overlay of the 5th floor above, I have inserted a photo-detail, which shows
how the equipment pool gate area appeared from a crane camera much later, by the end of August 2011.
Emission from the cracks at the sides of the equipment pool gate was still ongoing, but had by then been reduced to a trickle
compared to its heydays during March, April, and May.

 

[jim hardy]

your photos of upper deck & steam leaks

i think this is same area March 16. Cryptome photo 47, cropped a bit.
Looking West, and the column is #4 the centerline of building.

That crane picture you posted is closest to high resolution I've seen of that locale.

 

[MadderDoc]

There are no published photos of the top of unit 3 that are suitable to produce a complete thermographic overlay from.
This is probably as good as it gets at the current level of BO.
Green-yellow-orange to red represent areas above the building top which are at a significantly higher temperature than ambient.
The efficiency of IR blocking by the huge fallen overhead steel crane across the reactor areas is well demonstrated.
http://www.gyldengrisgaard.eu/fuku_docs/111008_05enhovl_thumb.jpg

http://www.gyldengrisgaard.eu/fuku_docs/111008_05enhovl.jpg and the http://www.gyldengrisgaard.eu/fuku_docs/111008_05enh.jpg on which the overlay was made,
enhanced from the original photo released by Tepco

 

[jim hardy]

There are no published photos of the top of unit 3 that are suitable to produce a complete thermographic overlay from.
This is probably as good as it gets at the current level of BO.
...

The efficiency of IR blocking by the huge fallen overhead steel crane across the reactor areas is well demonstrated.

I am carrying as an open question whether that crane was always in just that spot.
The dark circle in that Cryptome #47 shows in early helicopter videos pre- 24 March
but not in later photos or videos
my circles aren't perfectly aligned , sorry

... later ............... 3/16 ...

i don't know whether to dismiss it as an artifact of lighting.
In march the sun still comes up somewhat to North lighting from right side of photo.
By summer it's rising more southerly.
Fukushima at 37 deg North is about latitude of San Francisco and central Missouri
so midday sun never gets directly overhead but always lights from south (left).
Shadows in left photo show it was taken shortly before noon.

 

[MadderDoc]

I am carrying as an open question whether that crane was always in just that spot.<..>

That's pretty exotic. It'd be the last thing I'd imagine moved anywhere since it fell from its rails under the roof in connection with the explosion on March 14th.

The dark circle in that Cryptome #47 shows in early helicopter videos pre- 24 March
but not in later photos or videos

It's called apophenia. If you look carefully in the March 16th source video, you can see the outline of the dark shape is actually the inner edge of steam plumes moving about it. Consequently the dark shape is 'alive', changes shape,and does only in a brief moment appear as a circle. The video has been taken a few hours pre-noon, i.e. the sun is behind the camera to the left, and the wind and the plume is going in the direction of the camera. Hence the space over the reactor where no steam is condensing appears relatively dark, while there is no condensate there to scatter light, and the area is being shaded by the outgoing plume.

The photo you have to the left is from March 24th, when the building for a brief period paused steaming. It may have been exhausted after producing the grand black smoke event on the night before.