Search Results

We apply artificial neural networks to helicopter hydraulic pump condition monitoring. Several neural net models are used to perform pattern classification on the vibration measurements. Various pump conditions are examined using data from accelerometers in different places on the pump. The fundamental pump frequencies and its harmonics are used as input features to two neural net models: (1) a multi-layer neural net using back-propagation and (2) a Kohonen's feature map. Both neural net models have the ability to distinguish between pumps with different flow rates and mechanical conditions. A fundamental result is that the vibration signature can be used to classify pump condition.

A numerical computation scheme has been developed to determine the sound generated by propellers. A comparison of these calculations to the noise data taken in the the flight test of a propeller driven aircraft shows good agreement. The method is then applied in a parametric study of fixed pitch propellers designed to reduce noise. All these techniques reduce noise while maintaining shaft speed so that the method presented here may be used in a retrofit option for the general aviation fleet.