Folks,
The discussion today was very productive. Thanks to Kevin for sifting through the strain gage data and pointing out the most interesting features. (See attached file: Strain Gage Explorations...) Thanks also to Geoff for reviewing the pre-C1 strain gage qualification test data for us. The purpose of this note is to review what we learned yesterday and document the next steps we discussed.
Geoff started the meeting by going over the pre-C1 strain gage qualification test data. Recall that we put strain gages on a steel bar of known cross-section and pushed (or pulled) on it with a known force. We then cooled it down with LN2 and repeated the exercise. The bottom line is that (after accounting for the offset) the strain gages worked fine at LN2 temperature. The strain was linear with load and had the expected slope (Young’s modulus). This test appears to rule out temperature alone as the culprit.
Kevin walked us through the strain gage data and pointed out the most interesting features:
• Data does not match ANSYS in direction or magnitude except gage 15
• Strain readings in different directions at the same hole number give roughly the same delta strain.
• Strain readings from the two strain gages mounted in the same location with the same orientation (9 and 19) gave very similar results, even after cooldown.
• RT data is very noisy whereas cryogenic data is not nearly as noisy.
• Cryogenic data appears to change linearly with the current (I), not in a quadratic (I2) like the EM load.
• The change in strain during a pulse is much different at RT than at cryogenic temperature even with the same current.
• Turning the current off (i.e., a power supply trip) seems to create a different strain profile for both cryogenic and room – data gets much smoother. Step changes in strain readings can be seen (although there is no step change in current).
• Gage 15 away from the coils near the leads looks somewhat plausible even at room temperature.
There was considerable discussion over what factors contributed to these unexpected results. Two candidates emerged – power supply ripple and magnetoresistance. Power supply ripple is a likely candidate because of the noisy data when the coil is being powered and the different, quiescent data when the coil is not being powered but still has substantial current. Geoff agreed to provide current and voltage waveforms for cases of interest. We would like to see if the voltage ripple in the RT case was a lot higher than in the cryogenic cases which might account for the much noisier RT data.
A tangential concern arose during these discussions of power supply ripple. Each coil consists of 4 conductor paths. In the absence of any power supply ripple, when we ramp up the coil current, the outermost conductors carry more current than the others. When we ramp down the coil current, the innermost conductors carry more current than the others. The magnitude of this circulating current has been calculated in the past and found to be a small (I think order 10%) effect. However, substantial voltage ripple from the power supply could in principle excite this circulating current. Art agreed to provide a calculation of the circuit inductances to see if the driven circulating currents would be significant.
The other candidate is magnetoresistance. The dummy gages were located about an inch off the surface of the casting where the active gages were mounted. For the strain gages mounted on the winding packs, the dummy gages were on the opposing winding pack surface. If the active gage and the dummy gage are in different magnetic fields, there will be an apparent strain due to the difference in the magnetoresistance between the two gages. This apparent strain would track with current (which is what we see) and could swamp the true strain during a pulse. Art will determine the differences in magnetic field between the active and dummy gages. We will use that data to see if there is a correlation between the calculated difference in magnetoresistance and the change in apparent strain during a pulse.
We spent a considerable amount of time and money preparing for and testing the C1 coil. The lack of useful strain gage data means that [1] we have not verified our ability to predict the performance of the coil in ANSYS at temperature and under load (apart from the very good results matching the displacement measurements) and [2] we do not have a means to determine if the stellarator is performing as expected during commissioning and initial operation. The above investigations are interesting and may shed light on why we did not get useful strain gage data BUT they are [1] unlikely to make the existing data useful and [2] unlikely to provide a solution for future testing or use during stellarator commissioning and initial operation.
We had some discussion about if there were any success stories using strain gages on existing machines and if there were any solutions we could use in the future. Brad was going to contact Bob Walsh at the NHFML because he believes that they have used similar strain gages successfully. I will contact Irv Zatz and George Sheffield to better understand our history on TFTR with strain gages.
The most promising solution appears to be fiber optic strain gages. They might be immune to the effects that are plaguing us with conventional strain gages. We actually used these when doing the TF tests at ORNL. They were difficult to apply to the rough winding pack surface. One failed during testing. I believe this was attributed to the fiber optic rather than the strain gage itself. But they worked at cryogenic temperature! Brad reported that they were used on SNS with good results.
Following the telecon, I contacted Charlie Neumeyer to see if there was any NSTX experience with these fiber optic strain gages (after having no success with conventional strain gages). He reported that they are used extensively on the TF joints (2 per flag) and other places. They work very well. They are very fragile, requiring care during installation. However, once installed, they seem to be very reliable and give very accurate results. Charlie reported that the signal conditioning system that is used with these FISO fiber optic strain gages is particularly nice.
Regardless of whether we ever get to the bottom of why we did not get useful data from initial testing of the C1 coil, fiber optic strain gages are clearly the most promising option for future installations and we should lay the groundwork to qualify them for use. This work falls in ORNL’s bailiwick for coil I&C. It appears that a modest test program to develop/qualify the system components in our operating environment should be undertaken. ORNL has MTS test machines that could be used for cryogenic testing. (See last slide in Freudenberg’s PPT file.) ORNL (Nelson)should develop a test plan to support the design of a fiber optic strain gage system for use on NCSX. PPPL (Gettelfinger) should develop a test plan for re-testing C1 using this new system. Once we have these plans documented, we can fine tune them based on cost-benefit-risk assessments.
Let’s get together next week to review the input that was requested and further discuss the plan forward.
Regards,
Wayne