I am getting error for unit compatibility while solving equations using solve_block.
Please guide me to make correction.
thanks.
ACCEPTED SOLUTION
Accepted Solutions
You are NOT working with the same values when you omit the units!!
Without units you use values like 288.7, but with units you use 288,7 MPa which is internally represented by Prime in its base unit Pa.
So the numeric algorithm used by the solve block now uses the number 288700000.
On the other hand you use for example the number 50 for d but in the other sheet you use 50 mm, which is represented by the number 0.05 internally (base unit meter). Even worse when it comes to the areas. Unit-less value 942.5 versus 0.0009425 (m^2).
So the numbers in use when you use units cover a rather large range from very small to very large which may make a big difference concerning convergence behaviour within a limited tolerance.
So I think that's the reason why your solve block is so sensible concerning the guess values.
In the attached (unit-less) file I changed all values to the same numeric values which are used in the sheets with units.
As you can see, the calculations fail now!
You would have to set the system variable CTOL (convergence tolerance) to a rather large value like 10^-1 (default is 10^-3) to make the solve block work again and this at the cost of precision.
OR you let CTOL at its default value and provide other/better guess values.
I found that a guess value of 60*10^-3 makes for both sets the calculation work OK
Using other guesses was already the 'solution' in the sheet with the units.
In the new sheet here I found that 60*10^-3 and also 80*10^-3 are good guesses, but the second calculation (strain_500) fails with a guess of 70*10^-3. So there is a high sensibility concerning the guesses...
BTW, the "root" function, as shown by Luc, is usually much less sensible concerning the guess value and could also be use in a second way without having to provide a guess.
@MSRAZA wrote:
I am getting error for unit compatibility while solving equations using solve_block.
The error message I get is not about a unit mismatch! Its about Prime not being able to find a solution! ???
For some reason you are providing the guess value for x.u via the first function argument. If you want to do so, you must delete the hard coded assignment of the guess value (x.u := 50 mm).
Now the call with strain_500 returns a result, the first one with strain_415 still doesn't. Reason seems to be that your solve block calculation is very sensible with respect to the guess value. Changing the guess to 180 mm makes the first call work as well.
I appreciate all that you have done for me.
But when working without units, it is working perfectly without any error.
we can verify by root function, which was shared by you.
Kindly find the attached file.
Thank you.
You are NOT working with the same values when you omit the units!!
Without units you use values like 288.7, but with units you use 288,7 MPa which is internally represented by Prime in its base unit Pa.
So the numeric algorithm used by the solve block now uses the number 288700000.
On the other hand you use for example the number 50 for d but in the other sheet you use 50 mm, which is represented by the number 0.05 internally (base unit meter). Even worse when it comes to the areas. Unit-less value 942.5 versus 0.0009425 (m^2).
So the numbers in use when you use units cover a rather large range from very small to very large which may make a big difference concerning convergence behaviour within a limited tolerance.
So I think that's the reason why your solve block is so sensible concerning the guess values.
In the attached (unit-less) file I changed all values to the same numeric values which are used in the sheets with units.
As you can see, the calculations fail now!
You would have to set the system variable CTOL (convergence tolerance) to a rather large value like 10^-1 (default is 10^-3) to make the solve block work again and this at the cost of precision.
OR you let CTOL at its default value and provide other/better guess values.
I found that a guess value of 60*10^-3 makes for both sets the calculation work OK
Using other guesses was already the 'solution' in the sheet with the units.
In the new sheet here I found that 60*10^-3 and also 80*10^-3 are good guesses, but the second calculation (strain_500) fails with a guess of 70*10^-3. So there is a high sensibility concerning the guesses...
BTW, the "root" function, as shown by Luc, is usually much less sensible concerning the guess value and could also be use in a second way without having to provide a guess.
@MSRAZA wrote:
Can we set the default working units of MPa and mm?
Does it make any sense?
Kindly reply ASAP.
Thank you.
Short answer to your first question - no.
It may make sense but unfortunately Prime provides no means for changing the default units used internally nor the default units used for displaying results.
The desire for configurable default units often arises here, especially when users find it too inconvenient to have to manually change the default unit used to the desired one for each individual result displayed. And as your example also shows, this is not even possible with a matrix with mixed units.
But even if there were to be a solution for this at some point in a future version of Prime, it is questionable whether this would also apply to internal numerical storage and so avoiding the experienced effects.
So it may probably be the best way to set up the problem in a different way. Luc mentioned that it would be overdetermined which also may be the reason why we sometimes get different results using different guesses - for example
This happens much less often in your sheet without units, but still occasionally
Yes, the "root" function seems the best solution here.
Here is a way to set it up self contained in one single function and using a fix guess so you don't have to provide it.
The solutions look unique - at least for x>0
@MSRAZA wrote:
If any of you know or are in contact with the developer of PTC Mathcad,I requested to kindly suggest they provide an option for a default unit setting so that the user can choose accordingly.
Oh, hey, I know the developer of PTC Mathcad...!
Custom Unit Systems will be implemented in Mathcad Prime 11. Team PTC Mathcad just had a presentation last week (replay up now) about what will be (if everything happens as planned, anyway) in the next release. There were demos of the in-development build. Custom Unit Systems was one of the demos.
They'll resemble this:
I see no immediate units problem. Your case seems overdetermined.
When xu is known, then both eps values are known, and also the two f values fsc and fst.
This means that basically your problem is one-dimensional for each of the two scenarios, and can be solved with the root function:
First define:
Then setup the root function:
Then solve for the first scenario:
Check results:
Now for the other scenario:
Success!
Luc
Thinking of a different setup as Luc has shown sure is preferable, but in cases where this is not possible and you experience similar problems, you still can use units in your worksheet but strip the units for the solve block and add them again afterwards.
Its quite cumbersome and ugly looking and not recommended, but can be done if really necessary 😉
Here is one way to do it using the example of your solve block.
Prime 10 sheet attached
