January 15, 2019

Parallelization in pSeven: Design Space Exploration Batch Mode

Design space exploration (DSX) is a key part of the decision-making process. It allows pruning unwanted designs and highlighting valuable ones based on given objectives and constraints.

Quality of design is estimated by one or several objectives and admissibility is defined by constraints. In engineering problems, it is assumed that designs can be easily constructed but testing them against constraints and evaluating objectives takes time, because such evaluations may be carried out by a sophisticated 'black-box'. 

 

DSX methods include many approaches, most prominent of them belong to design of experiment (DoE) and optimization areas.

Due to their nature most of DoE methods may produce a whole set of points. Optimization methods are usually considered as sequential processes, in which a new promising design to evaluate is constructed based on information collected on previous steps. However, there are still a lot of situations when optimization can construct a set of new points that need to be evaluated. For optimization techniques that are available in pSeven, sets of designs are generated by:

  • gradient-based methods when regular or irregular numerical derivatives are used;
  • in surrogate based optimization (SBO);
  • in robust design optimization to sample uncertainties.

In all these cases, DSX constructs a pool of points that can be passed to 'blackbox' for evaluation.

 

From the user’s point of view, to handle points one by one is usually simpler than to organize processing of a queue of points. But on the other side, knowlegde about a set of points that need to be evaluated may contribute to more efficient allocation of computational resources.

Accordingly, the Design space exploration block in pSeven can work in different modes: the normal mode and the batch mode, which can be easily switched in block's run options.

The batch mode is useful when parallelization of the 'blackbox' work is possible, since it can significantly reduce the overall computation time.

 

By Alexis Pospelov, Senior Researcher, DATADVANCE

LinkedIn
VK

Interested in the solution?

Click to request a free 30-day demo.

Request demo