Adaptive Input Shaping for Flexible Systems Using an Extreme Learning Machine Algorithm Identification

Abstract

In this paper, a promoted adaptive input-shaping (AIS) with extreme learning machine (ELM) is presented to get zero residual vibration (ZRV) of severely time-varying flexible systems. Firstly, the ZRV condition and the tradi- tional adaptive input-shaper is reviewed, together with its disadvantages of insuf- ficient adaptability caused by giant amount of data and low-accuracy calculation caused by noise. After that, online sequential-ELM (OS-ELM) algorithm is intro- duced to identify the impulse response sequences of the flexible system, its fitting impulse response sequences are gotten to update the shaper parameters with fixed length and less noise; therefore, the above-mentioned problems of traditional AIS could be significantly avoided; that is to say, AIS’s adaptability and identification-accuracy could be improved apparently, which means better perfor- mance to suppress the residual vibration of the flexible system. Finally, the veri- fication experiments of presented AIS are implemented on a two-links flexible manipulator, which is a classical flexible system with severely time-varying dynamics; the results proves the effectiveness of the presented AIS method for the vibration control of severely time-varying flexible systems