MFMP: Automated experiment with Ni-LiAlH

The system just went through its first TEST mode where it turns off the power supply until the temperature drops 50C or for 300 seconds whichever comes first. Afterwards it goes back into PID regulation at its previous setpoint (1150C in this case). After having gone through this first TEST, the controlled power settled back to the same value, indicating that no LENR was apparently stimulated. This will happen again in another 4 hours.

21:00 UTC graph showing one of those tests: round2-pdf-1491083432.pdf

EDIT: 22:00 UTC update: round2-pdf-1491087971.pdf

EDIT: 23:00 UTC update: round2-pdf-1491089375.pdf

Well, so far I have not seen any evidence of XH in the data. The experiment is scheduled to run for several more days; as has been seen in other experiments, XH can appear after a long soak at >1100C ... or not. The experiment is pretty much autonomous now.

After 2 nights of only a little sleep, I am going to sleep tonight while the data is being taken. The data is being saved, but will not be uploaded to Google over the course of the next 8 hours. After I get out of bed in the morning there will be a big drop of all of the accumulated files.

While this experiment is running I am preparing for the next run with a different processing method for the fuel. With the automation in place, I am hoping for only a small gap between experiments.

Updated pdf-graph containing the latest data (not much difference compared to last time): round2-pdf-1491118453.pdf

Here follows some Python code which assembles experiment data from .csv files, calculates additional columns and creates averaged resampled data (i.e. reduced data) that will be easier to use with spreadsheets. This is a sample file containing such reduced data from the experiment at a 15 seconds sample rate: round2-1491118176-15s.zip (only 1.29 megabytes)

The code as it is works on my computer (with Python 3.6 64bit). It requires the third party Python libraries "numpy" and "pandas":

# -*- coding: utf-8 -*-
"""
This code reads data from a directory containing input .csv files produced during
Bob Higgins' (ClamShell) experiment and builds them into a single file containing:
    
    1) Human-readable full timestamps;
    2) Computed pressure and input power values;
    3) Calculated power output data.

An additional file containing reduced (and averaged) data is also created.
"""

import glob
import time
import os

import numpy
import pandas

###############################################################################

# Wildcard path to the directory containing all the source .csv files
filemask     = "d:/MFMP-data/clamshell/round2/*.csv"

# Output directory for the output .csv files
outputdir    = "d:/MFMP-data/clamshell/round2-output/" 

# Basename (without extension) for the newly saved .csv files
filebasename = "round2"

header_row    = 7       # Row from which data starts in the source .csv files
resample_time = '15s'   # Resampling time used for reducing .csv data
data_tz       = 'UTC'   # The data is in 'UTC' timezone format
drop_columns  = True    # Delete unused columns in the output .csv files

# Calibration coefficients from Bob Higgins' data
# Reference:
#    https://www.lenr-forum.com/forum/thread/5176-mfmp-automated-experiment-with-ni-lialh/?postID=54170#post54170
calib_coeff = [7.5046e-5, 8.5984e-2, -7.5065e-1]

###############################################################################

unixtime = int(time.time())

# Function that will generate data according to the calibration coefficients
f = numpy.poly1d(calib_coeff)

# The filenames that will be used for the output .csv files
csvdest_raw = os.path.join(outputdir, filebasename + "-" + str(unixtime) + "-raw" + ".csv")
csvdest_res = os.path.join(outputdir, filebasename + "-" + str(unixtime) + "-" + resample_time + ".csv")

# Create a collection of pandas DataFrames from .csv files in a source directory
data = []
for file in glob.glob(filemask):
    data.append(pandas.read_csv(file, header=header_row))

# Concatenate the Dataframes
data = pandas.concat(data, ignore_index=True)

# Convert mac timestamps to unix timestamps into datetimes usable by Pandas
# Reference: https://www.epochconverter.com/mac
data['Time Stamp'] = pandas.to_datetime(data['Time Stamp']-2082844800, unit='s')

# Make the dataframe indexed and timezone-localized
data = data.set_index(['Time Stamp'])
data.index = data.index.tz_localize(data_tz)

'''
Calculate computed columns and then join them to the main DataFrame

References:
  https://www.lenr-forum.com/forum/thread/5176-mfmp-automated-experiment-with-ni-lialh/?postID=52744#post52744
  https://www.lenr-forum.com/forum/thread/5176-mfmp-automated-experiment-with-ni-lialh/?postID=52959#post52959

-Pressure (PsiA): [Pr (V),PrS (V)] = (100/(0.8*C13))*(C12-0.1*C13) 
-Line voltage (ACrms): [c11: Line (V)] = 16.08 * C11 + 6.304
-Heater DC Voltage (V): [c9: Vvolt (V), c8: Vcur (V)] = 5.0302 * C09 - C08
-Heater DC Current (A): [c8:  Vcur (V)] = C08 * 40.1606
-Heater Power (W) = (Heater DC voltage) * (Heater DC current)
'''

# Basic columns
data = data.join(pandas.Series((100 / (0.8 * data['PrS (V)'])) * (data['Pr (V)'] - 0.1 * data['PrS (V)']), name="Pressure (psiA)"))
data = data.join(pandas.Series((16.08 * data['Line (V)'] + 6.304), name="Line Voltage (ACrms)"))
data = data.join(pandas.Series((5.0302 * data['Vvolt (V)'] - data['Vcur (V)']), name="Heater DC Voltage (V)"))
data = data.join(pandas.Series((data['Vcur (V)'] * 40.1606), name="Heater DC Current (A)"))

# Derived columns
data = data.join(pandas.Series(f(data['Tube (C-k)']), index=data['Tube (C-k)'].index, name="Calculated Output Power (W)"))
data = data.join(pandas.Series(data['Heater DC Voltage (V)'] * data['Heater DC Current (A)'], name="Heater Power (W)"))
data = data.join(pandas.Series(data['Calculated Output Power (W)']/data['Heater Power (W)'], name="Apparent power gain (Out/In)"))

# Remove columns and data that will no longer be used
if drop_columns:
    data = data.drop(['PrS (V)','Pr (V)', 'Vcur (V)', 'Vvolt (V)', 'Line (V)', 'N/C'], 1)

# Finally, save .csv files
data.to_csv(csvdest_raw)
data.resample(resample_time).mean().to_csv(csvdest_res)

This looks really well run experiment thanks for that Bob. It's really good to see this.

Great collaboration with Can too. The graphs are perfect.

If anything shows up with here it will be really clear and I would have good confidence in it.

When nothing shows up it's a great benchmark for the future too.

Good luck with this and your future iterations of the experiment.

Good morning from New Mexico. I feel better after a night's sleep. I have just uploaded a block of data files from overnight. In the current state, it is operating at the same power for 1150C, so there has been no effect that produced sustained XH, because it does not appear to be there now. It is not yet time to consider giving up on this run. I will be working on the next fuel batch during the day anyway. There will be a change to the fuel and to the protocol for the next experiments.

I am open to hearing suggestions for changes to experiment protocol - so please feel to post them. Changes to some things will be easy to incorporate and others will require more work to change hardware.

Many thanks to can for sharing his expertise in data analysis and even his code! BobG has asked if I can get the data into Plotly that has a Python interface (he has a license I think). I would be interested in hearing comments about that. I think can's posting of the graphs through LENR-Forum has worked out really well.

BobHiggins

I don't have much experience with the Plotly API or dealing with streaming data in general. Creating plots from existing files is very straightforward: there's not much that can fail in the process.

Some time back I was about to propose that instead of posting manually the plots in the form of comment attachments I could have arranged a shared web folder (for example on mega.co.nz) where these would get posted automatically as new .csv files arrived, which is similar to what barty is suggesting in the comment above. However it would not be much discoverable and it would require my computer to be always available while your experiment is running, which I cannot guarantee. There's also a certain amount of small manual tweaking with my graphs - in addition to scale changes I have changed and improved the code a bit over time (although added flexibility came at the cost of complexity).

me356 however does have practical experience with plotly as far as I recall.

can,

Thanks for posting your code. I just noticed that there is a name inside the code that you may wish to redact (or not).

For those of you who are curious, I have added screen captures of my control panel on my computer in the Photos folder. The underlying code is Labview, the license for which National Instruments graciously provided 1 license for FREE! Thank you National Instruments and Dr. Truchard! I had to learn Labview from scratch to write this code (but I was already experienced in other languages). It is a strange language and is fundamentally multi-threaded code, not sequential. It seems most programmers have abandoned the "virtual instrument" coding in favor of more conventional straight function access to the peripherals. This can present problems in multi-threaded code and I have created, what I believe, are true virtual instruments for the code I wrote to handle the power supply and the back pressure regulator. You can see their virtual instances in the Power/Heater status tab (3).

Here is the progress of the programmed experiment. At any point, if I decide that I am no longer interested in continuing along the present track, I can jump forward to the cool-down sequence which is stabilized at each step. More than anything, this would provide some measure of system aging - likely thermocouple. Note that during regulation, if the heater coil ages, the power should be the same for the same temperature while the voltage may not.

barty,

Thanks for the hosting offer There are more experiments to come in the weeks ahead, so I will certainly keep in mind how we can work to better share what is going on.

Note that I opted for photos instead of video. I would be interested in hearing how other people feel about this. In the area of my reactor, not one single atom has moved (kidding) - but nearly so. Video seems such a waste of bandwidth for something that never moves. It may be possible with video to capture that rare explosion - but I am working hard to see that never happens (the lab is under my bedroom in my house!). So, post some comments about this issue. If video is really desired, I will see what I can do.

BobHiggins

About that specific name, I don't particularly care either way, but I removed it just in case, thanks for notifying.

I don't think a video of the reactor in this experiment would have added much value. Perhaps if it mostly showed streaming data, that could have made it more accessible to some people, but in general I agree that the information/bandwidth ratio would have not been very high in this case.

As for the reactor, at this point more tests could for example include carefully cycling temperatures through the temperature ranges where reversible hydride reactions occur, although this is not what Parkhomov has done. This should significantly increase heater coil aging.

Quote from BobHiggins

I am open to hearing suggestions for changes to experiment protocol - so please feel to post them. Changes to some things will be easy to incorporate and others will require more work to change hardware.

At the risk of being a boor, a question of mine is whether the presence of a current-carrying filament made of something as heavy as or heavier than tungsten, say, inside the reactor volume and exposed to the hydrogen, will make a difference. Platinum and tungsten are two examples, but much heavier would be even better. (You probably aren't prepared in the present experiment to look at this, but I hope a future experiment will.)

Speaking of heavy elements, it would not be suitable for electrodes, but United Nuclear sells depleted uranium at an affordable price (USA only):

http://unitednuclear.com/index…=16_17_69&products_id=211

It has interesting properties: it absorbs a lot of hydrogen, increases its volume significantly when it does and pulverizes into very fine particles after several loading/unloading cycles.

Of course, it would take lots of guts to actually use it in these experiments ...

@Adrian,

When we were examining Parkhomov's reports initially, we thought high pressure could be generated by the potential amount of H₂ that could be released inside his closed reactors - it could release enough for 400+ bar. Then we found out that some of his reactors exploded, apparently from high pressure and a better seal. The ones that didn't explode and worked, we were very curious about the internal pressure. So Bob G. got him to include a manometer for reading pressure and he reported his reaction with both temperature and pressure data. From that report, I extracted the following curve. From this, it is clear that his seals leaked, you can even see the leaks starting after the peak in pressure. In the region where XH was seen, the pressure has been partial vacuum. This doesn't mean that XH can't occur at high pressure. Piantelli reported that the pressure must be partial vacuum. Mizuno was in partial vacuum. Rossi's early reports were for 10 bar of H₂ on the other hand, but that was for a different fuel mix and low temperature. For the Lugano experiment, the described casual sealing of the reactor after the fuel was added suggested a high probability of leak as well. Once the temperature goes above about 800C, the fuel draws the pressure lower - in Parkhomov's case, it would not have gone below 1 bar absolute (0 bar gage as shown in the graph below) without the fuel drawing the H₂ pressure lower. My script for this experiment includes programming the back pressure regulator to mimic the pressure vs temperature shown in the graph below.

Quote from barty

Yes, live video streams are unnecessary.

Live data streams would be more interesting

I will try to set up the plotly, just for my own interest. Then we can use it any time you want

Thanks Barty, I am back in LENRForum