SHERPA — Simultaneous Hybrid Exploration that is Robust, Progressive, and Adaptive
HEEDS uses a hybrid and adaptive algorithm called SHERPA as its default search method. SHERPA employs multiple search strategies at once and adapts to the problem as it “learns” about the design space. SHERPA requires significantly fewer model evaluations than other leading methods do to identify optimized designs, and often finds a solution the first time. This efficiency can save days or even weeks of CPU time during common engineering optimization studies.
During a single parametric optimization study, SHERPA uses the elements of multiple search methods simultaneously (not sequentially) in a unique blended manner. This approach attempts to take advantage of the best attributes of each method. Attributes from a combination of global and local search methods are used, and each participating approach contains internal tuning parameters that are modified automatically during the search according to knowledge gained about the nature of the design space.
This evolving knowledge about the design space also determines when and to what extent each approach contributes to the search. In other words, SHERPA efficiently learns about the design space and adapts itself so as to effectively search all sorts of design spaces, even very complicated ones. SHERPA is a direct optimization algorithm in which all function evaluations are performed using the actual model as opposed to an approximate response surface model.
All of the parameters within SHERPA are tuned internally, so there are no attributes to define for this method, except the number of evaluations you wish to perform.
Advantages of SHERPA
Finds better solutions the first time, without iterating to identify the best method or the best tuning parameters for your problem.
Enables non-experts to successfully apply automated optimization the first time.
Performs direct optimization based on actual model evaluations, rather than using approximate response surface models.
Identifies better-quality solutions for broad classes of problems, and performs global and local optimization at the same time.
Uses multiple strategies concurrently to more effectively and efficiently search even complex design spaces.
Enables non-experts to successfully apply automated optimization the first time.
Adapts itself to each problem, eliminating the need for user-specified tuning parameters.
Achieves both global and local search simultaneously.
MO-SHERPA (Multi-Objective SHERPA)
MO‐SHERPA (Multi‐Objective SHERPA) is a modified version of the algorithm SHERPA for multi‐objective Pareto search. It uses a non‐dominated sorting scheme to rank designs, but is quite different from NSGA‐II and NCGA in other aspects.
MO-SHERPA is designed to be used with projects with multiple objectives when those objectives are in conflict with one another. It works fundamentally like SHERPA but has the advantage of handling multiple objectives independently of each other to provide a set of solutions, each of which is optimal in some sense for one of the objectives.
Conventional parameter optimization methods take all objectives into consideration and provide solutions based on the weighted sum of all objectives. If all objectives get better or worse together, the conventional method can find the optimal solution. However, if the objectives conflict (as, for example, weight and load-carrying ability typically do), the tradeoff can be explored using MO-SHERPA Pareto optimization.
Advantages of MO-SHERPA
Performs multi-objective Pareto search using a modified version of the SHERPA algorithm.
Handles multiple objectives independently to provide a set of optimized solutions that represent trade-offs among the objectives.
Uses multiple search strategies simultaneously to more effectively explore the Pareto front.
Contains no tuning parameters, making it simple for non-experts to achieve success every time.
