Dynamic Intelligence of Self-Organized Map in the Frequency-Based Optimum Design of Structures

By leveraging the capabilities of machine intelligence, metaheuristics can be enhanced to achieve improved performance and convergence. This chapter presents an investigation into the dynamic scenarios for utilizing machine intelligence in metaheuristics, focusing on their application to optimization problems. The main contribution of this research lies in the proposal and evaluation of five distinct methods within the dynamic intelligence (DI) scenarios for optimization using machine learning (ML) and cross-entropy (CE) framework. These methods include ML cooperation in full or fixed number of iterations, adaptive usage, performance-based decision-making, and randomized intelligence usage. Through extensive experimentation, the DICE framework is applied to solve various benchmark truss optimization problems involving frequency constraints and size/shape variables. Prototypes such as the 10-bar planner truss, 72-bar space truss, 120-bar truss dome, 37-bar truss bridge, and 52-bar truss dome are considered. Comparative evaluations against competing techniques are conducted to assess the effectiveness and efficiency of the proposed DICE methods. The findings indicate the efficiency of the proposed methods in fast and optimum convergence.