Control of smart non-linear base-isolated structures using optimal adaptive neural network-based PID controller

This paper aims to propose an adaptive control approach for the PID controller known as ANN-OPID controller with simultaneous use of the advantages of the classical PID controller and neural networks so that it can provide better seismic performance than the optimal PID controller tuned by the teaching–learning-based optimization (TLBO) algorithm. In this approach, the structural dynamic is estimated using a neural network block and hence the parameters of the PID controller are adjusted adaptively. The seismic performance evaluation of the proposed controller is compared with the conventional optimal PID controller for an 8-story smart base-isolated structure subjected to various near-field and far-field earthquakes. Overall, the results obtained for the studied structure subjected to six earthquakes show that the TLBO-PID controller leads to a reduction of 24% and 25% in the maximum base displacement and its root mean square (RMS), an increase of 24% and 11% in the maximum acceleration and its RMS of superstructure floors and an increase of 6% and 6% in the maximum drift and its RMS of superstructure floors. However, the proposed ANN-OPID controller approach causes a reduction of 35% and 33% in the maximum base displacement and its RMS, a reduction of 9% and 7% in the maximum superstructure acceleration and its RMS, and a reduction of 5% and 6% in the maximum drift and its RMS of superstructure floors. Consequently, the proposed ANN-OPID controller can give a simultaneous reduction of the base displacement, acceleration, and drift of superstructure floors, while the TLBO-PID controller reduces the base displacement at the cost of an increase in acceleration and drift of superstructure floors.