Rotating machinery is a critical piece of equipment in many industrial plants, and its operation is regularly accompanied by balancing problems that result in potentially dangerous vibrations for operators and equipment. The centrifugal decanter, for example, is sometimes subject to vibrations that force the operator to slow down the production rate. The nuclear environment in which this equipment operates makes it impossible to carry out the measurements and observations required for a purely experimental study. The aim is therefore to carry out modelling with limited data in order to gain a detailed understanding of the phenomena involved. The aim of this work is to combine Euler-Euler type CFD simulations of the mass distribution in the rotating bowl with mass-spring modelling of the mechanical connections in order to get closer to the vibration signals measured industrially. Such a numerical tool would be a valuable aid in investigating the various potential sources of mass imbalance without the need for experimental replication. Combined with deep learning methods, this type of model would also make it possible to build an unbalance predictor from short vibration signals, opening the door to active control of the decanter