calculating mean time to failure of flexing ss tank top

huangyhg

calculating mean time to failure of flexing ss tank top
is it possible to calculate the mean time to failure for a ss tank top with a top mounted agitator. what we have is a tank with top mounted agitator that stirs a viscous solution (not sure of viscosity) at a speed of @ 60 rpm. this solution will "gel" when exposed to the atmosphere. i believe that we have build up on the agitator blades which is causing an imbalance scenario. the nozzle that this agitator is mounted to is not gusseted which inturn allows the agitator to rock in a circular motion flexing the top @ 1/16 - 1/8 in with every revolution, which will eventually, i believe, cause failure of the tank top (which would not be good). the shell is @ 3/16 - 1/4 in thick ss with a reinforcement plate of the same thickness welded to it. the failure is inevitably going to occur at the weld seam between tank top and this reinforcement plate (my opinion).i've told management about this,what will happen and that we need to gusset this nozzle, but this is a very integral part of our process and shutting down is a very rare option.  can i calculate stresses at the flex points and how long before we start to see signs of fatigue. formulas and or website info would be deeply appreciated.
roy gariepy
maintenance and reliability dept.
bayer corporation  dorlastan fibers div.
goose creek, south carolina  usa
check out our whitepaper library.
prediction of fatigue by analytical means is always approximate, especially for weldments. look at the fatigue design curves from the alloy at different numeber of cycles. these curves are often based on a 95% survival rate at a 95% confidence level from the underlying data.
first, the allowable stress range at the design load for each detail are divided by the structural stress range predicted by stress analysis (shell element fea) to obtain the global design margin. this margin is multiplied by the load multiplication factors for the "test life".  
from this you can predict the mean life from lognormal distributions.  standard statistical analysis techneques can be applied to find the probability of survival at test life and cumulative probability of failure.
calculation of stresses in the vessel head may be done with wrc107 bulletin; however unfortunately this bulletin has no formulae for deflections, whilst you are working with imposed deflections in this case.
wrc297 has such formulae, but only for cylindrical shells, not for spherical heads.
so you are unlucky, but by wisely merging the two methods you could get an estimate of stresses at both the nozzle attachment and at the outside of the reinforcement plate.
however again you won't obtain any information about the stresses in the weld of the reinforcement plate.
personally i would expect either a failure in the shell plate at the weld of the rp or in the weld of the nozzle neck to both the shell and the rp, depending on the outer diameter of the reinforcement plate.
also i confirm what boo1 said: methods for estimating fatigue failure are very approximate, and anyway they give you an absolute minimum life, never a mean life. it is quite possible that, after doing an analysis, you would conclude that the vessel should be already broken...
prex