Calibration with insufficient calibration references

[af-norce]

Hi!
Great package idea, I'm just going to get started using it.

How does the code behave if there is not sufficient information available to fully solve for all calibration coefficients?
e.e. 3 independent reference measurements of temperature are not available

There are two key examples for my use cases:

• where it is not possible to use sufficient calibration baths (perhaps one internal to the interrogator, one outside the interrogator, and that is it - the far end of the fibre is inaccessible).

• where the fibre experiences transient attenuations at point locations (bad splices, damage etc.) but it is not possible to have additional calibration baths either side of the loss.

The temperature calculated on the device can show discrepancy in some of these scenarios. But would this package provide a different answer when solving with insufficient information? And why?

Thanks in advance.

[BSchilperoort]

Hi,

It's been a while since I used the package myself, so I am not sure what will happen if there isn't enough information to fully constrain the calibration. I think it will lead to high uncertainty, weird drifts in temperature, etc. I don't think the package will raise an error.

• where it is not possible to use sufficient calibration baths (perhaps one internal to the interrogator, one outside the interrogator, and that is it - the far end of the fibre is inaccessible).

If you have only a single bath it will be basically impossible to calibrate the data properly. To get a ballpark estimation you can calibrate gamma (temperature sensitivity) and alpha (differential attenuation) for the specific fiber you used in a lab setting. Only the C parameter then would need to be calibrated. However, while setting gamma to a fixed value generally is acceptable, you should really use a double-ended calibration to properly estimate the differential attenuation along the length of the fiber. Strain has a strong effect on the differential attenuation, and due to the nature of fiber optic deployments this strain will never be constant along the length of the fiber.

In a double-ended setup you can probably get away with having a single bath, but two would give you more certainty in the accuracy of your setup.

You could also make use of opportunistic "baths". I.e. a section of the fiber in a medium with a relatively constant and homogeneous temperature in which you place the fiber, along with a temperature sensor with a separate logger. It will depend on your deployment what this can be.

Internal reference sections of the interrogator are basically useless for calibration purposes. The only use for this as end-user of the interrogator is to determine the relationship between signal strength and measurement noise of your device.

• where the fibre experiences transient attenuations at point locations (bad splices, damage etc.) but it is not possible to have additional calibration baths either side of the loss.

If you have the two fibers co-located in some way, you can correct for this by assuming that these sections of fiber are at the same temperature, giving you extra constraints.

---

Could you tell us a bit more about who you are and what you're using DTS for? It's nice to have an idea who we're talking to!

Lastly, I would be up for providing commercial services such as consulting on the use of this software & calibrating DTS data and/or maintaining this software if you are interested (and I assume @bdestombe would be as well).

[af-norce]

Interesting, thanks for the quick answer!

I'm with the NORCE Norwegian Research Centre, where we are looking at applications of different types of fibre optic measurement technologies in many scenarios. We work with DTS, DAS and FBGs, and also develop novel fibre sensors for chemical sensing alongside other environmental parameters.

The applications I'm thinking of right now concern an industrial process, with a single ended measurement where the far end of the cable is steadily consumed in a harsh environment. Temperature measurements as far along the cable as possible (at any given moment, given that the end is being eaten away) are the aim.

We generally do set gamma to be a constant value in these cases and use a single bath for in-situ calibration. Sometimes these cables will be used with unusual and tight bending profiles, so we then calibrate what we can with access to both ends of the cable in the laboratory first. But the strain will change when the fibre is pushed, pulled and bent by the materials in the process, and new step losses due to damage in a tough environment are unavoidable, whatever prior work we try to perform.

I will have a play with applying the code to these situations and see how the results come out. Time permitting, I will dig deeper and see if I can see anything inside this software to tweak for these scenarios, maybe I'll be in touch 😄

[BSchilperoort]

The applications I'm thinking of right now concern an industrial process, with a single ended measurement where the far end of the cable is steadily consumed in a harsh environment. Temperature measurements as far along the cable as possible (at any given moment, given that the end is being eaten away) are the aim.

Yeah that's going to be challenging to calibrate accurately. I don't think there's a way to properly constrain the calibration in this case apart from extrapolating the differential attenuation from a known section of cable, and only calibrating the C parameter.

You will get artifacts along the cable from differential strain, and there might be an error that scales with the distance along the fiber (i.e. local variations in attenuation and a general under/over estimation of alpha).

The calibration can take into account any uncertainty in the fixed alpha and gamma values, if you have some information on that the uncertainty should hopefully be properly propagated (and show that the further along the fiber you get, the uncertainty gets bigger and bigger)