===Article abstract===

This article will introduce the efficient way to use Abaqus python script to access Abaqus odb field outputs. The considerations for Abaqus python scripting are given. [[BR]]
 A comparison of several unpacking styles to access odb outputs is made. The demonstration code is provided in Github: https://github.com/tao364744553/Abaqus-DNN.git. 

----
=== Considerations for Abaqus python performance ===

 # Make use of appropriate object sequence unpacking. The style of sequence unpacking can significantly influence the efficiency of your Python scripts.
 # Make use of python built-in functions.  They are usually optimized for performance.
 # Make use of 3rd-party modules.  They are usually optimized for performance. 
 # In general, working with high-level objects is more efficient than working with low-level objects.
 # Avoid accessing attributes that require the use of computationally intensive methods.  For example: minimize the number of calls to getSubset() and addData() methods.
----

===Comparison of unpacking sytels=== 
The sequence unpacking style with the Abaqus object model can have a significant influence on performance. Consider the below three methods of access the same stress information:

[[Image(democode1.PNG)]]

For the first piece of code, it has to reopen the subdirectories in the odb dictionary every time it accesses stress. Thus, the cost will be significantly expensive for a large model.  [[BR]]
The second piece of code still needs to open the Value subdirectory. But it is more efficient than the first case. [[BR]]
The third piece of code shows the most efficient style. It is recommended to access odb outputs in this style. 

----

===Example to access multiple fields outputs===

A plane stress model is fixed at one end and under a displacement <math>u_x = 0.1</math>  inch. The FEM model is described below [[BR]]
[[Image(model.png,  600px)]] [[BR]]
[[Image(material.PNG,  300px)]]

[[Image(democode2.PNG)]]
[[Image(democode3.PNG)]]

The cost to access displacement for this model is shown below

[[Image(result.png)]]
----