Using an artificial immune system for classifying aerodynamic instabilities of centrifugal compressors

We consider the problem of detecting aerodynamic instabilities of centrifugal compressors using the compressor's sound signal. Therefore, an immunity-based classification algorithm is applied. For detector generation a modification of the v-detector design is proposed that allows both optimal detector positioning and variable detector size and shape. Numerical results demonstrate the design and show the validity of the approach.