Takahiro Ishikawa1, Kiminao Kogiso1, Kenichi Hamamoto
15:10 - 15:30 | Wed 22 Aug | Schackenborg | WeB6.6
The study analyzes a fault of McKibben Pneumatic Artificial Muscles (PAMs) focusing on unknown parameters used in nonlinear models of the PAMs. In the analysis, a change in dynamics, such as the steady-state responses to increasing the number of contraction movements, was measured through experiments and was subsequently associated with a change in model parameters. Subsequently, the fault state, which involves making a hole in the rubber tube of the PAM, and the normal state were classified by using a support vector machine (SVM) based on the change in model parameters. Furthermore, a few model parameters that can characterize the state of the PAM were extracted in terms of generalization performance. The results can be applied to other systems, such as hydraulic systems of a construction machine, and used to improve control performance by designing control systems to alleviate the influence of changing dynamics.