Abstract
A unique adaptive neural network control scheme is proposed for active suspension systems by employing bioinspired nonlinear dynamics, so as to address several critical engineering issues including energy efficiency, input delay, and unknown/uncertain dynamics simultaneously. A novel constructive predictor is firstly designed to solve the effect of input delay. Neural networks are then adopted to approximate the uncertain/unknown dynamics, and importantly, a unique finite-time adaptive control is established which can not only online update the input and output weights of the neural networks but also intentionally introduce beneficial nonlinear dynamics to vibration control. The significant difference from most existing controllers lies in that the designed controller can effectively utilize beneficial nonlinear stiffness and damping characteristics of a novel bioinspired reference model and, thus, purposely achieve superior vibration suppression and obvious energy-saving performance simultaneously. Theoretical analysis and experimental results vindicate that the proposed controller can effectively suppress vibration with much more improved control performance and considerably reduced control energy consumption more than 44%. This should be for the first time to reveal both in theory and experiments that a superior suspension performance is achieved simultaneously with an obvious control energy saving, by employing beneficial bioinspired nonlinear dynamics, compared to most traditional control methods.
Funding Information
  • Innovation and Technology Fund (ITF) Project of HK ITC (Ref. ITP/020/19AP)
  • Strategic Research Fund of the Research Institute of Urban Sustainable Development
  • HK Polytechnic University
  • Project of Strategic Importance of HK PolyU
  • General Research Fund of HK RGC (15206717)
  • Key Research and Development (Special Public-Funded Projects) of Shandong Province (2019GGX104058)
  • National Natural Science Foundation for Young Scientists of China (61903155)

This publication has 42 references indexed in Scilit: