It is probably faster to loop over all entries in your fieldOutput set, like in this code snippet, you can simply say: for stressValue in stressSet.

for step in odb.steps.values():

# print 'Processing Step:', step.name, step.number

endtime=step.timePeriod

if (step.number==3):

# for frame in step.frames:

# Access the final frame with index -1

frame = step.frames[-1]

if (frame.frameValue==endtime):

print "Reading at time", endtime

allFields = frame.fieldOutputs

if (allFields.has_key(Stress)):

isStressPresent = 1

stressSet = allFields[Stress]

if elemset:

stressSet = stressSet.getSubset(

region=elemset)

for stressValue in stressSet.values:

if (stressValue.maxPrincipal > maxPrincipal):

maxPrincipal = stressValue.maxPrincipal

maxElem = stressValue.elementLabel

maxStep = step.name

maxFrame = frame.incrementNumber

I think as you correctly stated the problem is that you need to access the whole stress field for everything at every iteration, then get its subset. Maybe if you know the node numbers (or pick them up from a set of nodes), you could iterate on them and use the in-built history output extraction, and save the stress values and do your calculation. That way you don't need to extract the rest of the model at least.

Best regards,

David

I struggled with a similar requirement for most of the duration of my PhD. In my extreme case I had an odb file of around 200K nodes with 4000 time frames each having U1, U2, and U3 displacement fields. I needed the data at each node, process it and then plot it back on the odb file mesh. My script was running into months trying to do this simple task. Try the method suggested by Martin Baeker first and see if that solves your problem.

For my case I found a very unconventional solution. What I did was that I would save the data for all the nodes from the odb file into three big .rpt file (for U1, U2, and U3) with Report>XY Data, select 'Separate table for each XY Data'. Sometimes for my case these text .rpt files might even run into 20GB+ and it takes some time for Abaqus Viewer to even save this file. I would then split this file into one file for each node (there is a linux terminal command to do this quickly). Remember you now have 3x200K files. When splitting remember to keep the instance name, and node number in the file name. I would then loop through these files for my calculations and write down the node number and processed result into another file. I would then create a new odb and write the results from the processed file into it. This process was really fast and saved me tons of time. I can send you a sample script which automates this process, if what I said doesn't make any sense to you (the rpt file still has to be generated manually).

If you really just want all stresses at all elements in all frames, I recommend ditching the subset and building a list as follows

stresses_in_all_frames = [

...frame.fieldOutputs['S'].values

...for frame in myOdb.steps[stepName].frames

…]

if you only wish to read the stresses at certain elements, generate a set containing the labels of theses elements before:

label_set = set([element.label for element in my_important_elements])

stresses_in_all_frames = [

...[

…… value for value in frame.fieldOutputs['S'].values

…… if value.elementLabel in label_set

…… ]

...for frame in myOdb.steps[stepName].frames

…]

this should be much faster, than generating a subset every time. If you dont know, which elements would be inside the subset, you could generate it once on an arbitrary frame and then read out the element numbers.

A different approach that would be much faster is to write data to the result file (.fil) instead of the odb file. Then use one of the subroutines in Abaqus that can read the data you want.

Best regards

Dennis