Shell Design Help

[MikeArmstrong]

Hi
I have a unique problem when calculating the tank shell weight in the attached sheet.

The shell weight which is calculated is then required above to determine if the liquid level should be increased which might require the chosen thickness to be increased, which if true would change the tank shell weight.

Hope that makes sense.

The table in the spreadsheet in an inserted excel component.

Any suggestions on how to modify the calculation of the shell weight automatically instead of manually inserting as indicated by the arrows would be welcomed.

Thanks in advance.

Regards

Mike

[alnstevens]

On 7/7/2009 1:25:12 PM, Armo wrote:

Would you be able to
>have a look at the comments i
>have made on the
>calculation.

Cheers Mike

See my comments on the attached.

stv

[MikeArmstrong]

Stv cheers for looking at the calc, but if you look at the sheet i have attached you can see that when units are added the equations for thickness don't work.

Regards

Mike

[alnstevens]

On 7/8/2009 5:35:16 AM, Armo wrote:
>Stv cheers for looking at the
>calc, but if you look at the
>sheet i have attached you can
>see that when units are added
>the equations for thickness
>don't work.

Regards

Mike

Mathcad doesn't like parameters with different units in the same vector or matrix. Therefore you need to remove them during the solve and add them back after (see attached).

In fact, there is really no need to include the thicknesses within the solve block - but I didn't take the time to do this.




stv

[alnstevens]

On 7/8/2009 6:17:12 AM, stv wrote:
>In fact, there is really no need to
>include the thicknesses within the solve
>block - but I didn't take the time to do
>this.

Actually, it's a trivial task to remove them. There is then no need to remove units from the solve block - see attached.

stv

[MikeArmstrong]

But that doesn't calculate the td and tt thickness.

Regards

Mike

[alnstevens]

On 7/8/2009 6:39:33 AM, Armo wrote:
>But that doesn't calculate the
>td and tt thickness.
>
>Regards
>
>Mike

True, but you can just calculate them separately after finding the values of W.com, calling the appropriate function: something like: t.d(Level(Wcom[2))

stv

[MikeArmstrong]

Thanks alot i think it is sorted now..

God a need a bit of work on this programming.

Regards

Mike

[TomGutman]

Here is the sheet with units fully implemented. Some constants have been back-calculated to make the results the same as before.

Note that when you use Mathcad units the results are not in any particular units. They are physical quantities. When you display them you use the units placeholder to decide what units you want to show. Thus a length is just a length, and can be displayed in meters or millimeters or feet, indifferently.

I don't understand why you reverse the selected thicknesses. Here your heights are essentially the same, so it make no difference in the calculations, but it seems that if that were not the case the calculations would be wrong.
__________________
� � � � Tom Gutman

[MikeArmstrong]

Tom

I have just had a look and it work its fantastic.

Excellent effert.

Cheers Mike

[MikeArmstrong]

Hi

Thanks for the help so far with this tank design especially Tom & Stv.

I have tried to incorporate the units spreadsheet of Tom with the programming spreadsheet from Stv, the majority of the spreadsheet seems fine unit we get to the simultaneous equations, it doesn�t seem to like the units at this stage.

Any ideas anyone????

Mike

[TomGutman]

What simultaneous equations? I don't rememberm anything in your sheet that required, or implied, simultaneous equations.
__________________
� � � � Tom Gutman

[MikeArmstrong]

Sorry that is the part that Stv helped with.

Regards

Mike

[TomGutman]

Helped doing what? I don't remember anything in the sheet that called for simultaneous equations.
__________________
� � � � Tom Gutman

[alnstevens]

On 7/8/2009 5:09:29 AM, Tom_Gutman wrote:
>Helped doing what? I don't
>remember anything in the sheet
>that called for simultaneous
>equations.
>__________________
>� � � � Tom Gutman

I re-arranged the logic of the original sheet so that the manual iteration of the original was replaced by a solve block containing some simultaneous equations, which is what I thought Mike was asking for. Have a look at any of the files I've attached to posts in this thread

stv

[TomGutman]

I've looked at your sheet and could not make any sense out of it. There is no real iteration in the original sheet. The thicknesses are chosen manually and fixed at the chosen values. There are some constraints on the values, with the constraints very weakly dependent on the chosen values.

It's at most a pass/fail situation, not an iteration for a value. Values are chosen by the user, the limits for the values are calculated, and it is determined (manually, although some automation could be included) if the chosen values are compatible with the limits. If not, new choices need to be made.

Since AFAICT the whole purpose of the sheet is to have the user make the thickness choices in a table that also shows the lower limits, separating the chosen thicnesses into a separate table, earlier in the sheet, seems to contradict the basic purpose of the sheet.
__________________
� � � � Tom Gutman

[MikeArmstrong]

Tom

In my original calculation the thicknesses are chosen by the user based on the minimum allowed thicknesses for design and test conditions.

But if you look at the first sheet in the first few posts, once the thicknesses had been chosen the tank un-corroded weight was calculated which had to be manually inserted above (shown by the arrows) so the combined weight could be used in the design liquid height criterion equation.

This is why I asked Stv is he could suggest an alternative method.

Regards

Mike

[ptc-1368288]

Mike,

That is a very long thread now, not well targeted maybe. You have essentially missed the dimensioned diagram. Also some reason why API looks so complicated. Pressurized, non pressurized thanks, horizontal, spherical, boiler drums ... guess how many I have dealed with as a Process Control Consultant. Browsing through your work sheet only indicated you were checking some design from suppliers, specified for such or such applications as per some code. In fact you didn't abstract your work sheet justifying for it.

Just a comment, forget it.

jmG

[MikeArmstrong]

JMG

A not too sure what you�re getting at, I posted my work so some of the more advanced competent Mathcad users could help and suggest improvements?

API 650 seems a little trivial with some of their calculations and requires a lot of information to be extracted from them.

Regards

Mike

[TomGutman]

Yes, but there is still no actual feedback. The weight is based on the user's input, that input doesn't change based on the criteria calculations. Did you want to replace the user's input with the calculated minima? Then why ask for a user input at all?

Further, the calculated weight does not figure directly in the calculations of the criteria. It is used only for a logical decision as to whether or not L_cri is to be used. As such, find is inappropriate and would not work anyway. Find requires a continuous (and differentiable) relationship between the variables involved.

The ability to use the calculated weight for figures in the table is provided by the variable define component. The misuse of a solve block (there is no point in finding values that are simple functions of other values, degrees of freedom need to be matched between constraints and unkowns) does not help this.
__________________
� � � � Tom Gutman

[MikeArmstrong]

Hi Tom

This is just the way the original calculation was set up in my company I have tried to improve the calculation and would be happy to incorporate any comments.

Remember these tank designs are sent out to clients who have to check and approve them, so when they ask for a tank to be designed to a standard, they want direct correlation between them.

Lcir is a condition where a nominal figure is added to the design liquid height if the uplift pressure in the tank is greater than the combined tank weight, this can then increase the required thicknesses and in turn increase the shell weight.

Regards

Mike

[TomGutman]

I'm not sure I understand what your point is. I was not questioning the calculations in your sheet. But those calculations do not involve, nor imply, a solve block.

It may not make much logical sense that at some weight you don't use L_cri but if you have one Newton less you do. But that's the standard and I did not change that, nor suggest changing that. But regardless of the reasonableness of this usage, or of the reason for it, the result is still a discontinuous behaviour in terms of W_sh, and find simply doesn't do that kind of dependency.

Further, a change in weight may change the required thicknesses. But even if it does so, that does not change the actual selected thicknesses. Those will change, if at all, only if and when the user explicitly changes them. In terms of the calculations on the sheet there is no iteration or feedback. Nothing that suggests the use of a solve block. The only reason for having the out of order calculations is to show the results of the calculations (the required thicknesses) at the same point that the input to the calculations is placed (the Excel table). The calculations could be done just as well with the thicknesses input in a table near the top and the calculated required thicknesses shown at the bottom. It just would not be a convenient for the user to verify that the thicknesses are sufficient.

Ideally the Excel sheet itself would compare the input values to the calculated minima, and flag any values are are too small (perhap putting a red background for them). That would make it much harder for somebody to overlook a too-low value.
__________________
� � � � Tom Gutman

[MikeArmstrong]

Tom cheers for the advice.

Regards

Mike