Re: Transient stresses

From: <Christopher>
Date: Sun Nov 11 2007 - 23:51:00 EST

On Nov 11, 2007, at 6:50 PM, SEMTEC wrote:

> Specifically I must evaluate the
> transient thermal stresses that arise in a pump discharge piping
> system when
> the pumping product is changed from propane at -41oC to butane at
> -5 oC.
> This is an 18" OD header of 1 mile long. As butane displaces propane I
> realize that there will be local stresses in the "butane front" and
> there
> will also be expansion stresses as the system is changing its
> temperature.

I've done a certain amount of this sort of thing, although not in long pipelines. Probably the way you want to attack it is to solve the local problem at the 'butane front' separately and superimpose those results (which are secondary stresses) on the primary bending response from the overall temperature change. I don't think I'd try this with anything but a pretty good FEA program. This isn't a simple problem--but you probably know that already.

The heat transfer problem is going to be the real hooker. Strictly speaking the wall temperature lags the bulk temperature of the contents because of the convective resistance between the wall and the contents, but maybe you can ignore the lag and be conservative. I think what I'd do is try to get a feel for how the wall heats up when subject to a moving temperature boundary using a relatively short length of pipe modeled with plate elements. Figure the thermal time constant for the pipe section and set the length of the pipe section so that the travel time of the 'butane front' is much larger than the time for the pipe to heat up. Include convective resistance from the contents to the pipe and through the external insulation. You can see real quick how closely the pipe wall temperature follows the bulk temperature.

If you get lucky the pipe wall temperature follows the bulk temperature closely enough so you can figure the pipe element nodal temperatures change simultaneously with the arrival of the 'butane front.' Then you'd generate the model for the entire header with one dimensional pipe elements or beam elements) with the distance between nodes equal to the flow velocity times the thermal time constant. You'd start with the whole model at -41C and change the temperature of each node at times corresponding to the arrival of the 'butane front.'

The next step is a transient thermal stress model using the temperature time distributions you found in the heat transfer run. Most finite element software can apply these temperatures automatically in a time history run.

This sounds complicated, but I don't think there's another way. In effect you're assuming that a given section of pipe heats from -41C to -5C instantaneously which is conservative.

It occurs to me that you can probably estimate the effect if the pipe is a simple header without branches or elbows by considering the thermal loading in a one mile length of pipe at a temperature of -5C over an arbitrary fraction of the total length from an initial temperature of -41C. This would be a relatively simple manual calculation with a spreadsheet and you could generate a transient stress history pretty quick. But you'd still need to see how a given point in the pipe heats up as the 'butane front' passes. Christopher Wright P.E. |"They couldn't hit an elephant at <a href="/group/PipingDesign/post?postID=I54hkZ534oO9XrYPA6kl0wxj1lwQgYWI9K_K3CFBWgnSbAWLVu-Q5srQqPr3VITrCqCkFzo1VVn1Jw">chrisw@skypoint.com</a> | this distance" (last words of Gen.

.......................................| John Sedgwick, Spotsylvania

1864)
<a href="http://www.skypoint.com/~chrisw/">http://www.skypoint.com/~chrisw/</a> Received on Sun Nov 11 23:51:00 2007