Machines with high accuracy on factory floors

Abstract

This thesis will propose a new method to predict the response of sensitive equipment, once it is installed on the factory floor. Currently, only the stiffness of the floor is taken into account when the dynamic response of a machine on a floor is predicted. To improve these methods, a short literature survey is done about the dynamics of factory floors. From this survey it follows that a floor is better characterised by an SDOF oscillator. To use this characterisation in practice, dynamic substructuring will be used. With dynamic substructuring it is possible to predict the coupled response of a machine, when the dynamic response of the floor is known. This could either follow from a detailed model or from a measurement. In this thesis it is shown how this should be done with measurements. This dynamic substructuring can be easily extended to improve the method to predict the response of the machine to the vibrations of the oor. It is found that an equivalent system can be defined which predicts the oor vibrations more accurately. This method is again based on a dynamic measurement of the floor, as well as the free vibration level of the floor. In the second part of this thesis, the theory is validated with an experiment. First it is shown how to obtain the dynamic response of the oor from impact measurements. It is found that a floor typically has a lot of damping, which damps the response quickly. This limits the achievable frequency resolution. Also problems are encountered with harmonic vibrations of a floor, which appear as a resonance, and the combination of the low response and an exponential window, which might cause an artificial anti-resonance. These measurements are than used to predict the coupled response of a test case. This prediction is also validated with a validation measurement. It was found that these floor measurements together with the dynamic substructuring predicts the coupling very well. The proposed method to predict the coupled vibration level of the oor could however not be validated with this experiment. When the results obtained with the dynamic subtructuring are compared with the results as obtained with the existing methods, it is found that dynamic substructuring improves the prediction very much.