Since there is only one true level, one or the other sensor - or neither - is correct.

Calibrate the easiest / most accurate sensor, then adjust the 2nd level sensor to match values across the overlap range.

There are probably actual spec's and procedures available online for bringing level sensors into agreement.  That is the best approach.

Do the 2 sensors overlap?

Are both sensors providing the same accuracy?

Is sensor 1 say 0-100% of the range of available level and sensor 2 is say 40-60% of available range and is more accurate over the smaller range?

And of course, what PLC platform are you using?

smsmith0977

I have a system that has two separate level detectors. The detectors read different values for the same level. When switching between the active detectors, the difference is too great to have an instantaneous switchover. What is the best way to softly make the transition? I need to start by reading 100% of sensor A and 0% of sensor B, wait half a second then read 90% of sensor A and 10% of sensor B and add the values together to give a new level reading to the PLC...each half second will read 10% more or less of the respective sensors until sensor A is at 0% and sensor B is at 100%. We solely use LAD.

If they are both providing analog inputs or inputs from remote IO or inputs from an external device via a communications method, once the data has arrived in the PLC, you are in control of the values from that point.

If you know the actual level using a huge measuring stick made of something with a much higher melting point - an exaggerated example, you can make the 2 level values INSIDE the PLC agree by using PLC logic instructions.

If for example, the radioactive source sees some variable amount of impurities, you can bracket the best and worst case accuracy, and over time you can probably improve that accuracy estimate.

For the other magnetic based level sensor that is blind to the impurities, you can artificially add the estimated amount of impurities based on the radioactive based level sensor reading and it's estimated accuracy - which may depend upon what is currently being measured - or from an experienced metallurgist who can provide good educated estimates - or you can contact the sensor manufacturers for methods of determining accuracy in liquid steel of semi-unknown impurity amounts.

I suspect the reason for the 2nd sensor is to account for the impurities.  True?

If so, then you should be able to artificially adjust the magnetic level sensor input according to the differences between the scaled level signals used internally within the PLC as "preliminary" scaling.

Once both MagLvl_scaled_cm and RadLvl_scaled_cm have been obtained using calibration data obtained from their respective manufacturer's calibration or device data sheets, you can then artificially increase the level of the MagLvl_scaled_cm by the difference between the level sensors once a sufficient level of liquid steel is present.

Presumably this will need to be mapped, probably based on an experienced Metallurgist's input, as the amount of impurities as well as the shape of the vessel containing the liquid steel will vary with actual level for a known mixture of liquid steel and impurities.

Perhaps you can obtain several known impurity level variations and imply the source metal content mixture currently in liquid form based on the impurity difference.

If you simply want to make them match, you can adjust the magnet level sensor scaled value UP by the difference between their scaled input values.

For s7-1200/1500, use NORM_X and SCALE_X to get input signals scaled into common measurement units to enable automatically adjusting the magnet level sensor.

For example, let's assume both sensors provide a 0-27648 analog input into an S7-1200/1500 PLC.

Let's also assume both level sensors read 0-100cm, and for over-simplicity, let's assume the vessel is a perfect cylinder at least 100cm tall.

So, initial scaling would be something like this:

-------|========|-------------------|=========|---|
     0-|        |-#tReal        0.0-|         |-#Lvl_scaled_cm
in_Lvl-| NORM_X |            #tReal-| SCALE_X |
 27648-|========|             100.0-|=========|

Now you can do a simple subtraction between the larger #RadLvl_scaled_cm and smaller #MagLvl_scaled_cm to get an average impurity delta.

SCL:
#ImpurityDelta := #RadLvl_scaled_cm - #MagLvl_scaled_cm;

With that impurity amount, you can offset the magnet level sensor signal to include the missing amount of level due to impurities on top of the liquid steel. 

SCL:
#adjMagLvl_cm := #MagLvl_scaled_cm + #ImpurityDelta;

#adjMagLvl_cm (adjusted, accounts for impurities) will be available for use for managing vessel level as well as #MagLvl_scaled_cm (unadjusted, does not account for impurities). 

Hopefully this makes sense, especially if you share it with a Metallurgist.  They should be better able to assist with real-world impurity calculations, or perhaps the radioactive level sensor manufacturer has a white paper, maybe written to help deal with impurities using their sensor in combination with a magnetic level sensor.

smsmith0977

The sensors cannot be calibrated to read the same because of their construction. One is a radioactive source, the other is a magnet. The radioactive source "sees" impurities that float on top as if they are part of the liquid steel. The magnet doesn't "see" these impurities. We are using TIA Portal V17.

There are 10 distinct time intervals, with 1 calculation to make at each of the 10 time interval points.

It is easy enough to simply count the system clock memory bit for 2Hz (0.5ms) leading edges using an INC instruction.

Math:

Let's say the INC tag is #TransitionStep, that ranges from 0..10, and when it is 0, no math takes place.

 #TransitionStep   #transitionStep
-------[>=]---------------[<=]---------[CALL FC to do math]
        1                  10

   #Op_Swap_LVLs  #transitionStep #Do_LVL_Transition
-------| |----------[>=]-----------------( )---
                     10

#Do_LVL_Transition      Clock_2Hz
-------| |-----------------|p|----------[INC]
                          ons_1     #transitionStep

The above sets up the transition based on the operator input #Op_Swap_LVLs, and calling an FC with the current transition step # from 1..10

That FC would use the transition step # [1..10] to do parameterized math, like algebra, but in a PLC.

FC: LevelTransitionMath
IN:
  X : REAL or LREAL
  Y : REAL or LREAL
  Step : UINT
OUT: PLC_LEVEL

// X * (100.0 - Step * 10.0) + Y * (Step * 10.0) = PLC Level //
//              -----------                -----             //
//                  100.0                  100.0

// use the CALCULATE instruction either one really big time, //
// or use several CALCULATE instructions to calculate the    //
// results.  Store the results to #PLC_LEVEL and pass back   //
// to calling program - or do the best solution              //
// ** BEST ** For S7-1200/1500 PLCs                          //
--- right-click on network 1 of FC LevelTransitionMath and
--- insert new SCL network:

SCL:
#PLC_LEVEL := #X * (100.0 - #Step * 10.0) / 100.0 + Y * (#Step * 10.0) / 100.0;
// 

Or, you can use a lot of MUL, DIV, SUB blocks to do the above math.  However, you already understand the required formula, so use it.

smsmith0977

This method is not feasible as the impurities vary widely with the different grades of material, additives etc. To try and simplify: Let's say sensor 1 is reading X and sensor 2 is reading Y. Sensor 2 is the active sensor in the PLC, but the operator wishes to switch to sensor 1. This needs to be done 10% at a time, every 500 ms, giving the PLC a "fake" level reading to use during the switchover. In this example it looks like this on paper:

X * 100% + Y * 0% = PLC Level

X * 90% + Y * 10% = PLC Level

X * 80% + Y * 20% = PLC Level

....

X * 0% + Y * 100% = PLC Level

Then the PLC allows sensor 2 to take over.

I'm looking for some ladder logic to facilitate this easier.