Abstract: The capabilities of a system developed on the basis of ANN are preliminarily evaluated, in suppressing the vibration field related to a general structural element; to this introductory aim, a numerical FE-model of a flat plate, part of an acoustic box for technology demonstration of architectures for sound radiation reduction, is referred to. The aluminum panel is excited by a 2-tonal point force, with the main peak at 50 Hz. The force may vary in frequency and amplitude ($POM@2% and $POM@10%, respectively). The time delay, between the two components of the signal is considered time-varying. A control system referring to 4 piezoelectric ceramic actuators and 8 accelerometers is used to minimize the vibration field; the placement of the different devices is defined by means of a proper tool based on Genetic Algorithms, and able to consider both the investigated frequencies; time delay (or phase shift), is not take into account in this process. An arbitrary signal, simulating the corresponding reference for the vibration suppression system, is transmitted to an ANN-based observer, aimed at identifying all the characteristic values, namely the amplitudes, the frequencies and the time shift. A classic feed-forward type minimization algorithm, ANN-based, is implemented. The identified solution is validated by means of a FE simulation, developed in the time domain, where the disturbance force is acting with the true values of the component sinusoidal forces (amplitude and frequencies), shifted by the correct value of the imposed phase. To have a first estimate of the capability of the proposed architecture in following the signal variations, a new signal is given to the ANN identifier, and the whole process is repeated.!22
展开▼
