A framework for brain learning-based control of smart structures

Abstract

In this approach, a deep neural network learns how to improve structural responses using feedback control. The effectiveness of the framework is demonstrated in a case study for a moment frame subjected to earthquake excitations. The performance of the learning method was improved by proposing a state-selector function that prevented the neural network from forgetting key states. Results show that the controller significantly improves structural responses not only to earthquake records on which it was trained but also to earthquake records new to the controller. The controller also has stable performance under environmental uncertainties. This capability distinguishes the proposed approach and makes it more appropriate for the situations in which it is likely that the controller will be exposed to unpredictable external excitations and high degrees of uncertainties.

Real-time measurement-driven reinforcement learning control approach for uncertain nonlinear systems

“…The Hierarchical Temporal Memory (HTM) and LSTM network approaches have been adopted to predict short-term arterial traffic flow [26]. A deep neural network that employs a feedback control concept is adopted to solve an intelligent structural control problem [27]. It makes use of a state-selector function to avoid forgetting key states by the neural-networks and hence improve the overall performance.…”

Novel Active Structural Vibration Control Strategy Based on Deep Reinforcement Learning

“…Recently developed algorithms, including deep deterministic policy gradient (DDPG) [7], trust region policy optimization (TRPO) [8], and proximal policy optimization (PPO) [9], have exhibited exceptional performance in robotics control tasks within the action-state domain. A modified DQN is used for structural control by optimizing a straightforward reward function [10]. Nonetheless, the control signal continues to be discrete, preventing accurate determination.…”

Tuned liquid wall damper for multi-hazard mitigation

“…Deep learning networks have also been proposed to build control rules for systems with complex dynamics. For example, Li et al (Li et al, 2010) employed a dynamic neural network to track nonlinearities in a building, and Rahmani et al (Rahmani et al, 2019) used a deep neural network for seismic control. The main advantage of deep learning algorithms is their capability to implicitly learn complex features (e.g., nonstationary dynamics) and adapt to uncertainties (e.g., system properties and external excitations).…”