ADAPTIVE PHYSIOLOGICAL STATES CLASSIFICATION IN FED-BATCH FERMENTATION PROCESS

摘要

We present in this paper a model of fed-batch bioreactor states classification which it makes the differences between physiological states and the command action to maintain the nominal parameters. This method is based on the adaptive detection of noncorrelated bioreactor signals. A segmentation based on maxima of modulus of wavelets transform and Holder exponent is used to before the principal component analysis PCA. The segmentation allows the detection of operator's action during the fed-batch fermentation and the principal component analysis allows to define the influence of those operator's actions on the physiological states.