Efficient hybrid optimization method and its role in computer-aided design
An efficient, reliable, and accurate optimization algorithm is proposed and developed. The algorithm has a local superlinear rate of convergence. It uses a cost function bounding concept initially and a constrained variable metric (CVM) method in later stages of the iterative process. This type of a hybrid algorithm uses potential of the constrained variable metric methods to their fullest extent. The variable metric methods have superlinear rate of convergence and are fully analyzed. A higher order of convergence is often accompanied by a smaller domain of convergence. Thus, a cost function bounding algorithm is used to reach the domain of convergence and then a switch is made to the CVM method. An attempt is also made to establish a sharper and reliable lower bound on the optimum cost function value. This is done by incorporating second order information into the cost function bounding algorithm. Some efficient numerical schemes such as normalization of the QP subproblem are incorporated. An improved active set strategy is suggested. Also an improvement to Pshenichny's descent function is proposed and implemented. Besides developing a superlinear optimization algorithm, an efficient programming structure of computer-aided design of engineering systems is suggested and implemented.
- Research Organization:
- Iowa Univ., Iowa City (USA)
- OSTI ID:
- 5569078
- Resource Relation:
- Other Information: Thesis (Ph. D.)
- Country of Publication:
- United States
- Language:
- English
Similar Records
Large-scale sequential quadratic programming algorithms
Large-scale sequential quadratic programming algorithms
Related Subjects
42 ENGINEERING
ALGORITHMS
OPTIMIZATION
COMPUTER-AIDED DESIGN
ENGINEERING
COST ESTIMATION
HYBRID SYSTEMS
METRICS
PROGRAMMING
MATHEMATICAL LOGIC
990200* - Mathematics & Computers
420100 - Engineering- General Engineering- (-1987)