Inductive sensor linearisation

[Pandiani]

Hi everyone,

This is my first post on this form and I hope you'll be able to give me useful advices and share opinions.
I have an inductive sensor and when it is new, it is linear i.e. output voltage depends lineary on distance from metal object, so this can be described with the following equation y = ax+b.
y is voltage, and x is distance. This voltage signal is transmitted to acquisation system and this system asks for slope and offset so it can calculate distance back and show it on operator's screen. System calculates distance by formula: x = m*y+n, "m" and "n" are entered to system like parameters. After two or three years, sensor's charactersitcs become more nonlinear (something like "s" shaped curve). This situation can be easily solved without replacing sensor by recordin its characterstics and then obtaining new "m" and "n" from excel worksheet using least square method. When sensor's nonlinearity is huge, only replacement is solution.
Now suppose the following situation:
system is showing distance 4 mm and voltage from sensor is say 10 V. After recording charactersitc, it shows that sensor has nonlinearity and using excel table new "m" slope and "n" offset are obtained. Now when new m and n parameters are entered in the system, new charcterstics is more linear (appeared to operator).

Now, where to return sensor? If I return on old position (by output voltage), system will now show that distance is not 4 mm but for example 3.8 mm?
There is other possible option: to try to return sensor to new position and fix it when system show 4 mm (now voltage from sensor will not be 10 V).
What would you do?
What is your opinion?

Thank you very much.

 

[reidh]

I would collect a large series of data points for your inductive sensor, say 20 data points (every half volt if it is a 0-10V output). Perform this test twice, once when you are increasing in distance, once when you are decreasing in distance. Plot both of these and compare the trendlines. This will tell you if hysteresis is affecting your sensor.

Assuming there is a slight difference between the m and n values, average them, plug in your new m and n values, and re-check your results by varying your sensor distance.

If the readings still are not correct, it sounds like your sensor is not repeatable (a very bad thing for a sensor). Replace it, and move on.

-Reidh

 

[Pandiani]

Thank you for the reply. You didn't quite understand what was the actual question. I get satisfactory results when linearizing, but problem is when returning sensor back into the field. If I return on the same place, now I have different distance shown on screen (because ma and n are changed in the system). Should I move sensor to a new distance (very small) in order system to show value that was before...

Not sure if I was able to explain, but it comes to this.

Before any work, sensor was in its position and its voltage output is 9.6V. That values is calculated to the distance in acquisition system and shown on operator screen d = m*9.6 +n.

Now after "m" and "n" are "corrected", when sensor is returned to its original physical position, its output is 9.6V, but now little different value is shown on the operater screen.
Question, should I leave it that way, or physicaly move sensor a little bit to new value (say 9.7V) in order system to show same distance that was on the screen before any work performed?

 

[danbarber]

TURCK & Balluff has some interesting inductive sensors that may fit your application.

http://www.clrwtr.com/Turck-Inductive-Capacitive-Proximity-Sensors.htm