AI-assisted optimization design of seismic performance parameters for timber structures.

Abstract

Timber multi-story buildings offer environmental benefits, lightweight construction, and seismic resilience, but Artificial Intelligence (AI) based integrated frameworks for optimizing seismic parameters, including inter-story drift and roof displacement, remain limited. The Gradient Boosted Random Forest Machine with Scalable Cheetah Optimizer (GBRF-SCO) is proposed for improving prediction accuracy while facilitating optimal design decisions. The dataset consists of 4,000 timber building samples obtained from a publicly available Kaggle repository (Timber Seismic Performance Dataset). Data pre-processing employs normalization and outlier detection using Robust Scaling and Isolation Forest, ensuring high-quality inputs. For exploratory analysis, t-Distributed Stochastic Neighbor Embedding (t-SNE) is applied to visualize high-dimensional feature relationships and identify structural parameter patterns relevant to seismic performance. The proposed framework uses GBRF to predict seismic response metrics, with the SCO tweaking hyperparameters to optimize model performance. It also enables the optimization of seismic performance characteristics, guiding engineers in selecting structural designs that minimize drift and enhance robustness. Multiple Linear Regression (MLR) was employed to examine the influence of key structural and seismic elements on roof displacement, providing insights into the overall seismic performance of wood buildings. Comparative evaluation shows superior performance over conventional regression and ensemble methods, demonstrating a higher accuracy of 0.949, which corresponds to the classification of roof displacement levels (low, medium, high) under seismic loading conditions and seismic intensities using Python 3.10. By providing a strong and clever method for designing sustainable and earthquake-resilient buildings, the suggested GBRF-SCO framework successfully improves the seismic performance optimization of timber structures.