Results in Engineering (Dec 2024)
Data-driven continuous-time Hammerstein modeling with missing data using improved Archimedes optimization algorithm
Abstract
This research introduces the improved Archimedes optimization algorithm (IAOA) for data-driven modeling of continuous-time Hammerstein models with missing data. It addresses the limitations of the original Archimedes optimization algorithm (AOA) through two key modifications: rebalancing the exploration and exploitation phases and mitigating local optima trapping issues. The primary focus is on developing a novel data-driven approach for modeling continuous-time Hammerstein models, particularly in the presence of missing output data. Four levels of missing measurement data (5 %, 15 %, 35 %, and 50 %) were considered, with data points randomly replaced with zeros. Models were tested with both complete and missing output data to evaluate the robustness of the IAOA-based method. The proposed based method identified linear and nonlinear subsystem variables in a continuous-time Hammerstein model leveraging input and output data, validated through two practical experiments: a Twin Rotor System and an Electromechanical Positioning System. The performance was assessed by examining various factors, including the convergence curve of the fitness function and its statistical analysis, responses in the frequency and time domains, Wilcoxon's rank-sum test, and computational time. Across all experiments, the IAOA-based method demonstrated superior performance compared to AOA and other methods, including a hybrid approach combining the average multi-verse optimizer and sine cosine algorithm, particle swarm optimizer, the sine cosine algorithm, multi-verse optimizer and grey wolf optimizer. The findings showed that the proposed IAOA-based method delivered highly accurate and consistent solutions, proving it to be the most effective and reliable method compared to the others assessed.
