The use of a multisensor array for measuring the emission from a production-scale baker's yeast manufacturing process is reported. The sensor array, containing 14 different gas-sensitive semiconductor devices and an infrared gas sensor, was used to monitor the gas emission from a yeast culture bioreactor during fed-batch operation. The signal pattern from the sensors was evaluated in relation to two key process variables, the cell mass and the ethanol concentrations. Fusion with the on-line sensor signals for reactor weight and aeration rate made it possible to estimate cell mass and ethanol concentration using computation with backpropagating artificial neural nets. Identification of process states with the same fusion of sensor signals was realized using principal component analysis. (c) 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 55: 427-438, 1997.