Pole-Mounted Sonar Vibration Prediction using CMAC Neural Networks

C. Zhang, L.G. Kraft, and B.R. Calder (USA)


Neural networks, pole vibration, and oceanic engineering


The efficiency and accuracy of the pole-mounted sonar systems are severely affected by the pole vibration. Traditional signal processing techniques are not appropriate for the pole vibration problem due to the nonlinearity of the pole vibration, the lack of a priori knowledge about the statistics of the data to be processed, and the caustic environment. A novel approach of predicting the pole mounted sonar vibration using CMAC neural networks is presented. The feasibility of this approach is studied in theory, evaluated by simulation and verified with a real-time laboratory prototype. Analytical bounds of the learning rate of CMAC neural network are derived which guarantee convergence of the weight vector in the mean. Both simulation and experimental results indicate CMAC is an effective tool for this vibration prediction problem.

Important Links:

Go Back