LION-AG Experiment

  • The first LION-AG active run started at 23:50 UTC on 7 April. The streaming data is available at

    https://plot.ly/~QuantumHea...

    Plotly update is pretty slow, so be patient and it will eventually show the moving graph of temps and power.

    The fuel load and core construction are as close to the LION-provided description as possible, but with added thermocouples for measurement of the core temperature. Five calibrations were performed, showing good stability and measurement accuracy of around ±5°C and ±3 watts at 800°C. The initial program is a 48-hour linear ramp from 100 C to 800 C, followed by a soak period or temperature cycling, depending on conditions at that time. Full details of the experimental setup will be made available shortly.


    Images and data will be made available as the experiment progresses.


    AlanG

  • can

    I'm having my doubts about the average power trace. I recall seeing that it is supposed to be a 300 sec moving average but it doesn't look like that to me.


    Also, not temperature vs average power (although that would be a pretty interesting plot!). You are showing temperature and average power vs time

  • Bruce__H

    Sorry, I was incorrect. I meant temperature and power.


    The average power looks ok to my eyes, but it's hard to tell without the raw data available:



    I found I am not able to export data from the live graphs with my personal Plotly account. A possible solution could be digitizing it from the previously posted screen grabs using WebPlotdigitizer and then resampling it with other techniques to create a proper temperature vs power graph but I'd rather wait for the actual raw data first, or for the testers to post their own.

  • can

    As you suggested, Plotly no longer supports persistent data in the cloud. I am collecting the data locally from Labjack and will post the files as soon as the experiment ends, in a couple of hours. Thanks again for offering your perception and skill for this important analysis of the experiment.


    @Bruce_H

    The Average Power is a 300 second trailing average, so it will not be perfectly centered in the messy Watts graph. It is calculated from the Watt-Hours integration of the Tektronix PA1000, read once per second into a Python array and the delta then divided by elapsed seconds. Because the sampling is asynchronous with the PA1000 integration, there will be measurement jitter of one part in 300 using this technique, so the average power plot is not perfectly smooth, but it's close enough.


    Experiment details and links to data files (when available) can be seen in the Live Doc at:

    https://docs.google.com/docume…wa9T_mpyErhsCppMXxizazc5s

  • The Average Power is a 300 second trailing average, so it will not be perfectly centered in the messy Watts graph. It is calculated from the Watt-Hours integration of the Tektronix PA1000, read once per second into a Python array and the delta then divided by elapsed seconds. Because the sampling is asynchronous with the PA1000 integration, there will be measurement jitter of one part in 300 using this technique, so the average power plot is not perfectly smooth, but it's close enough.


    OK. It is not the lag produced by the trailing average that I am wondering about. But if what I am seeing is a discretation error of some sort then I think I understand.


    The Live Doc summary is fabulous!

  • can

    Interesting choice to show not temperature versus power but temperature differential versus power!


    If there is no anomalous power generation in the system, I would expect this plot to be a straight horizontal line ... and that is basically what we see for the primary thermocouples.


    What would you expect to see if there is anomalous power generated in the core? I would expect a straight horizontal line that droops at the right hand side and then possibly turns back on itself.

  • Bruce__H

    The underlying assumption with that graph is that there is a direct relationship with input power, which might not necessarily be the case over long periods. So for long term graphs it would probably be best to revert to temperatures and power against time.


    It looks like PID control is based on Tactive, so with anomalous power production Tactive would remain constant while Tnull decreases. With a sudden excess power release Tactive might increase above the setpoint before this occurred, though.

  • The underlying assumption with that graph is that there is a direct relationship with input power, which might not necessarily be the case over long periods. So for long term graphs it would probably be best to revert to temperatures and power against time.


    I partially disagree. If the relationship between input power and the rest of the system changes over a long time period then one would hope that the same change would occur in the control chamber too and these change would cancel out when you plot the temperature difference




    It looks like PID control is based on Tactive, so with anomalous power production Tactive would remain constant while Tnull decreases. With a sudden excess power release Tactive might increase above the setpoint before this occurred, though.


    I asked and had it confirmed that the PID control is based on T_active. You are correct that with anomalous power generation T_active should remain relatively constant while T_null decreases, but don't forget that input power will also decrease. That is why I think a plot of T_diff vs Power should look like a lower-case letter r that has been rotated 90 degrees clockwise.



    I've used data as it came from the xls file and haven't tried yet to compute average power values on my own.


    I tried out a 300 sec trailing average on the xls data and found that my calculations generally agree with the results in the "Average Power" column of the spreadsheet. I think that plotly has some bugs that was producing some odd appearances for the average power trace in the streaming data.

  • I partially disagree. If the relationship between input power and the rest of the system changes over a long time period then one would hope that the same change would occur in the control chamber too and these change would cancel out when you plot the temperature difference


    If there's the chance over time of a drift only affecting one side of the apparatus due to real excess power or thermocouple issue / artifact, this might not necessarily be the case, which is why I added that over longer periods a temperature vs time graph (as shown by default on the live Plotly pages) might be better. The next portion of my comment was under this context, although this was probably not sufficiently clear.

  • I note that the "Average Power" trace on the streaming coverage of the replication is currently (Tuesday Apr 10, showing temperature cycling between 500C and 800C) much smoother and more in line with my expectations of a 5 minute moving average than it was previously during the ramp. Not sure why.

  • I note that the "Average Power" trace on the streaming coverage of the replication is currently (Tuesday Apr 10, showing temperature cycling between 500C and 800C) much smoother and more in line with my expectations of a 5 minute moving average than it was previously during the ramp. Not sure why.

    Under the cycling regime, the temperatures really never settle, so the PID is always trying to catch up with the thermal settling time, and the sign of the differential (the "D" in PID control) only flips as the cycle reverses. The result is relative absence of servo hunting and a smoother curve.


    We will soon change from 1 hour to 15-minute cycling interval in response to advice received by BobG from LION.