An optimisation method in Python 2.X that runs on a set of abaqus models to minimise the difference (in a least square sense) between the FEA output and a corresponding set of data

The following Python modules are used: scipy.optimize, subprocess, numpy, and importlib if run with Python 2.7

The module opti4AbqScalar defines a residuals function which is minimised by the scipy.optimize.minimize_scalar method. That residuals function computes FE models defined in python files stored in a given directory (modelsDir) and compare the output of those models with experimental data stored in a directory (expDir) - expDir and modelsDir can be the same directory.

The experimental data has to be stored for each model in a .ascii file whose name is the same as the name of the python file defining the corresponding model: if modelsDir contain modelNumber1.py and modelNumber2.py then expDir should contain modelNumber1.ascii and modelNumber2.ascii

Currently (commit 05/03/15), only data consisting of one single scalar values are supported.

The python files defining the abaqus models must define a model, a job, and a job analysis (with the job.submit() and job.waitForCompletion() commands). The abaqus model must be function of the scalar parameter to optimise. That scalar is passed to the abaqus model as a system argument: i.e. the model be run by the optimisation with the command abaqus cae noGUI=path_to_model -- parameterToOptimse A typical python file will contain the following block:

import sys
optiParam = float(sys.argv[-1]) #reads the string argument as a float
job = parametrisedTests(optiParam) #defines the job through a function taking as argument the parameter to optimise
job.submit() #submit the abaqus job
job.waitForCompletion() # waits for the job to complete before allowing the system to do anything else

The post-processing of the FEA model must also be defined in this python file with a function called postPro and taking as arguments the name of the abaqus odb: def postPro(odbName). The values of interest (i.e. those similar to the experimental data) must be recorded in a file called file output.ascii in the same format the experimental data is A typical postPro function will open the odb, read some output values, do some mathematical operations on those values and write the outcome into output.ascii

The opti4AbqScalar module has to be called through its main function: main(modelDir,expdir) A typical example using the module is defined in runScalarOpti.py

The main function also takes as optional arguments pBounds, a tuple of two floats/integers defining the boundaries of research of the parameter to optimize, and options, a dictionary of optimisation options as described in the example file

The main function returns the value of the parameter after optimisation, the value of the residual function for that parameter, the number of evaluations of the residuals function, and an optimisation message.

At each iteration of the optimisation process, a verboseValues_X.ascii (X being the iteration number) file is written in the working directory. It contains the iteration number, the value of the parameter at that iteration, the value of the residuals function at that iteration as well as a list of the values of each model output.