DenseNet Structured Neural Network-based Chaos Prediction for Stepped-chamber Tapered Air Bearing System

Stepped-chamber tapered air bearing (STAB) systems are advanced bearing arrangements designed for applications requiring superior stability at high rotational speeds, along with precise control and stiffness. These systems generate multidirectional supporting forces and offer higher stiffness, making them ideal for high-performance mechanical designs. Despite their advantages, STAB systems can exhibit chaotic motion due to factors like nonlinear pressure distributions, gas imbalances, and design flaws. To understand and control non-periodic motions, a comprehensive analysis is essential. Analytical methods, including dynamic trajectories, spectral responses, bifurcation diagrams, Poincaré maps, and Lyapunov indices, are employed. The study reveals that chaotic phenomena are sensitive to eccentricity and bearing number variations. The interplay of these factors is thoroughly examined. To address chaotic behavior, a DenseNet-structured neural network (DSNN) is developed. This computational framework predicts chaos by evaluating the maximum Lyapunov index of the STAB system. Comparative assessments demonstrate the superior predictive efficacy of the DSNN model, making it a robust tool for anticipating and managing chaotic behavior in STAB systems. These findings enhance our understanding of dynamic intricacies in STAB systems and provide practical design guidelines for industrial applications prioritizing precision, rotational speed, and stiffness.