NEi Advanced Optimization

Parameter Optimization Methods and Strategies

• SHERPA: Automated hybrid adaptive search that uses multiple search tools concurrently
• Multi-objective SHERPA (Pareto/tradeoff optimization)
• Global search methods
  • Genetic algorithm (hybrid, hierarchical, heterogeneous, mixed-variable)
  • Advanced proprietary evolutionary search algorithm
  • Simulated annealing
• Multi-start local search methods
• Surrogate-based methods (linear and quadratic response surfaces)
  
• User-defined methods via application programming interface (API)

Process Integration and Automation

• Direct portals to common CAE tools for data extraction
  
• Automated execution of multiple simulation and analysis tools within a design evaluation process
  
• Integration and sharing of data among separate simulations
  
• Support for parallel processing on networks, clusters and multiprocessors.

Design of Experiments

• Full factorial designs (2-level and 3-level)
  
• Fractional factorial designs (2-level and 3-level)
  
• Taguchi orthogonal arrays
  
• Plackett-Burman designs
  
• Latin hypercube designs
  
• Central composite designs
  
• D-optimal designs
  
• Taguchi robust parameter design (RPD)
  
• User-defined arrays
  
• User-defined response data

Quality Design Tools

• Taguchi robust parameter design (RPD)
  
• Structured sampling
  
• Random (Monte Carlo) sampling

Response Surfaces

• Linear and quadratic
  
• Multivariate adaptive regression splines (MARS

Solution Monitoring

• Process control and run-time adjustment capability
  
• Real-time solution monitoring with user-controlled graphs and tables
  
• User-specified termination criteria

Figure 1. System Level Model (pickup truck)

Unique Features and Capabilities

• SHERPA: Simultaneous Hybrid Exploration that is Robust, Progressive, and Adaptive
  
• Finds better solutions the first time, identifying the best method or tuning parameters for each problem.
• Enables non-experts to successfully apply automated optimization.
• Performs direct optimization based on actual model evaluations, rather than using approximate response surface models.
• Uses multiple strategies concurrently to effectively and efficiently search even the most complex design spaces.
• 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
• 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, so non-experts can achieve success every time.

User Interface

• Intuitive graphical interface: Pre-processing, run-time monitoring, and post-processing.
  
• Simplified coupling of simulation and analysis tools
  
• Guided problem set-up procedure
  
• Detailed and global views of problem statement
  
• Platform independent

Platforms

• Windows x86-32 - Windows 2000, XP, and Vista
  
• Windows x86-64 - Windows XP x64 and Vista x64
  
• Linux x86-32 or x86-64- Red Hat Enterprise Linux v4 and 5, SuSE Linux Enterprise v9 and 10
  

Figure 2. Reduced Subsystem Level Model