Advanced LS-OPT®: Deterministic & Probabilistic Optimization
Instructor
- Anirban Basudhar, Ph.D.
Advanced LS-OPT®: Deterministic & Probabilistic Optimization
Prerequisites
- Required: Basic knowledge about metamodel-based optimization and result analysis using LS-OPT.
- Strongly recommended: Introduction to LS-OPT class since it provides a foundation for some of the advanced topics.
- Recommended but not required: An introductory class in LS-DYNA® for familiarity with a few keywords.
Syllabus
This course is intended to help engineers with a basic knowledge of LS-DYNA and LS-OPT to become proficient in advanced optimization and probabilistic design methods. With this course we hope for you to become more productive at design and parameter identification of complex systems, such as multidisciplinary systems with competing objectives, advanced material testing and models, and systems with discontinuous responses. We will also provide insight into reliability and robustness in order to facilitate higher quality product design. Additionally, we will introduce classification-based adaptive sampling constraints as a tool for enhancing the efficiency.
In this course, we will discuss both the theoretical and practical aspects of design. We will also cover advanced topics, such as multi-objective and collaborative optimization, digital image correlation, statistical classification, and probabilistic optimization. During workshop sessions, we will apply the discussed theoretical topics. We will use the LS-OPT Version 6.0 graphical user interface to teach input preparation and post-processing. We will also emphasize interfacing with LS-DYNA.
Content
Day 1
- Course Outline
- Introduction to Design Optimization & LS-OPT Basic Features Summary
-
Theory: Parameter Identification
- Noisy Data: Filtering Computed Curves
- Dynamic Time Warping (DTW)
- Digital Image Correlation (DIC)
-
Examples: Set up, run and post-process parameter identification examples
- GISSMO failure model example
- Defining multi-point histories for spatial data
- Full field calibration using DIC data
-
Theory: Collaborative Optimization
- Multidisciplinary Optimization (MDO)
- Multilevel Optimization
-
Examples: Set up, run and post-process collaborative optimization examples
- Mode tracking
- Variable screening
- MDO using a reduced set of variables
- Multilevel Optimization
-
Theory: Classification-Based Constraint Handling
- Discontinuous and binary responses
- Classification-based constraint boundary definition
- Support Vector Machine Classification (SVC)
-
Example: Optimization with Discontinuous Constraint Response
- Defining a classifier
- Optimization using a constraint defined by an SVC classifier
Day 2
-
Theory: Multiobjective Optimization (MOO)
- Pareto front definition and MOO algorithm
- Analyzing the Pareto front using the Viewer
-
Example: Setting up, running and post-processing MOO example
- Create Pareto Optimal front
- Trade-off Plot, Parallel Coordinate Plot (PCP), Self Organizing Maps (SOM), Hyper Radial Visualization (HRV)
-
Theory: Probabilistic Analysis
- Statistics fundamentals
- Probabilistic analysis methods
-
Example: Direct Monte Carlo Analysis
- Uncertainty quantification using noise variables and statistical distributions
- Latin Hypercube Sampling
- Failure probability calculation
- Statistical post-processing tools
- DYNAStats
-
Example: Metamodel-Based Monte Carlo Analysis
- Reliability calculation with noise variables and control variables
- Statistical post-processing tools
- Stochastic contribution
- DYNAStats
-
Theory: Probabilistic Optimization
- Reliability-based design optimization (RBDO)
- Robust design
Day 3
-
Example: Reliability-Based Design Optimization
- Optimization of Control Variables
- Target probability of failure
-
Example: Robust Design
- Noise and Control variables
- Standard deviation composite
- Minimize effect of noise variables
- Example: Sequential Metamodel-Based Monte Carlo Analysis
- Example: Sequential Monte Carlo Analysis with Classifier-Based Adaptive Sampling Constraint
- Stochastic Fields
- Outlier Analysis (optional)
- Metal Forming (optional)
- Tolerance Optimization (optional)