Run ANSYS Mechanical APDL from Python in an efficient manner

Question

I have the following code, which writes an Input file and executes ANSYS Mechanical APDL using a Windows command. My issue is the execution time is much longer (15 minutes inside the software, upwards of 1 hour when calling from Python). I need it to be faster because I am varying as many as 'all' the input parameters.

def RunAPDL(E,t,w,p,aa,bb,lz,alpha,delta):

    ansyspath = r'C:\Program Files\ANSYS.Inc\v181\ansys\bin\winx64\MAPDL.exe'
    directory = r'C:\Users\Erik\Documents\ANSYS'
    jobname = 'file'
    memory = '4096'
    reserve = '1024'
    inputfile = r'C:\Users\Erik\Documents\ANSYS\ShellBucklingInput.inp'
    outputfile = r'C:\Users\Erik\Documents\ANSYS\OutputFile.txt'
    resultsfile = r'C:\Users\Erik\Documents\ANSYS\ShellBuckling.csv'

    # Start time count
    start = time.clock()

    # Write input file
    input_parameters = ('/NERR,200,10000,,OFF,0 \n'
                        'pi = acos(-1) \n'
                        'E = {:6.0f}     ! N/mm2 Young\'s modulus\n'
                        't = {:4.2f}       ! mm thickness\n'
                        'w = {:3.2f}       ! Poisson\'s ratio\n'
                        'p = {:7.2f}    ! N/mm2 external pressure\n'
                        'aa = {:6.2f}   ! mm horizontal radius at u = 0\n'
                        'bb = {:6.2f}   ! mm vertical radius\n'
                        'lz = {:6.2f}   ! mm model height\n'
                        'lu = 2*asin(lz/2/bb) !mm \n'
                        'lv = 2*pi      ! model perimeter at u = 0 \n'
                        'nu = 2*NINT(ABS(aa)/SQRT(ABS(aa)*t)) \n'
                        'nv = 3*nu      ! number of elements along v-axis \n'
                        'alpha = {:4.2f}   ! ratio of lu that is not loaded by p \n'
                        'delta = {:4.2f}*t ! prescribed imperfection magnitude \n'
                        '*ULIB,ShellBucklingLibrary,mac \n'
                        '*USE,ShellBuckling,pi,E,t,w,p,aa,bb,lz,lu,nu,nv,alpha,delta \n'
                        '/CLEAR'
                        ).format(E,t,w,p,aa,bb,lz,alpha,delta)
    with open(inputfile,'w') as f:
        f.write(input_parameters)

    # Call ANSYS
    callstring = ('\"{}\" -p aa_t_a -dir \"{}\" -j \"{}\" -s read'
                  ' -m {} -db {} -t -d win32 -b -i \"{}\" -o \"{}\"'
                  ).format(ansyspath,directory,jobname,memory,reserve,inputfile,outputfile)
    print('Invoking ANSYS with', callstring)
    proc = subprocess.Popen(callstring).wait()

    # Update pressure field for next analysis
    with open(resultsfile,'r') as f:
        lambdaS = float(list(csv.reader(f))[-1][16])
    p = 1.2*lambdaS*p
    print('Updated pressure is',p,' N/mm2.')

    # Stop time count
    stop = time.clock()
    print('Elapsed time is ',stop-start,' seconds.')

    return(p)

I execute it by pressing F5 on the Python shell, then messages appear in the shell:

Invoking ANSYS with "C:\Program Files\ANSYS 
Inc\v181\ansys\bin\winx64\MAPDL.exe" -p aa_t_a -dir 
"C:\Users\Erik\Documents\ANSYS" -j "file" -s read -m 4096 -db 1024 -t -d 
win32 -b -i "C:\Users\Erik\Documents\ANSYS\ShellBucklingInput.inp" -o 
"C:\Users\Erik\Documents\ANSYS\OutputFile.txt"
Updated pressure is -0.0046478399999999994  N/mm2.
Elapsed time is  4016.59094467131  seconds.

For anyone still wondering. The function file above works fine (in Windows). The input parameter was creating files which were computationally very expensive for ANSYS (very thin shells with too many nodes). — Erik, Apr 05 '18 at 12:52

score 2 · Answer 1 · answered May 30 '20 at 15:11

2

I don't know why your code is slower in Python but I suggest you to use PyAnsys package from PyPI. It offers a robust infrastructure to combine Python and APDL.

answered May 30 '20 at 15:11

Christos Sevastiadis

131
3

Here's the link to the GitHub repo. https://github.com/akaszynski/pyansys – Alex Kaszynski Jun 05 '20 at 23:37

Run ANSYS Mechanical APDL from Python in an efficient manner

1 Answers1