A STUDY ON BIFURCATION CONTROL USING PATTERN-RECOGNITION OF THERMAL-CONVECTION

The Benard thermal convection field is known to have multiple solution s for a certain Rayleigh number condition. This means that several con vection flow patterns can be stably sustained without an external cont rol force. This study demonstrates that bimodal switching control is p ossible by means of flow state estimation. A neural network was used t o recognize the transition state of the convection flow pattern. In or der to change the flow pattern, an unstable feature of Benard convecti on was utilized. That is, the chaotic nature of wobbling around multip le solutions contributes to fast switching between stable modes. Numer ical simulations were carried out to show the ability of the current c ontrol method.