Estimation of Bearing Forces in Rotating Machinery: A Problem Revisited

Development of a new ship machinery installation may require combinations of prime movers, electric motors, generators, gearboxes and other items that have not been used previously. Large items of this type are expensive to develop, so that only existing production items are likely to be affordable. Modern quiet ship design requires careful attention to the source characteristics of the individual machinery items as well as the dynamic characteristics of any new combination, so that fluctuating forces transmitted to the hull via mounting systems and flexible connectors in the final installation are within acceptable limits. Almost periodic components, which arise at multiples of machine rotational frequency, are of particular concern. Machine source properties are often known only in terms of the vibration characteristics of previous installations, not in terms of the disturbing forces that cause that vibration. The aim of the techniques described in this paper is to allow deduction of those disturbing forces from a matrix of transfer functions at each frequency of interest, measured with an existing machine in operation. The number of forces to be estimated must be less than the number of structural degrees of freedom. Those degrees of freedom arise from rigid body motions and machine flexural properties, which may change significantly when shafts are rotating. Also, the matrices must be redundant in order to allow estimation of the accuracy of derived force estimates. The larger the machine, the greater the number of degrees of freedom that are likely to arise at a given frequency. A first use of the methodology was to establish the bearing forces in a marine turbo generator (TG) set with plain journal bearings. Measurements of transfer functions were made with the machine stationary and then with the machine in normal operation. Direct and reciprocal measurements in different directions were made for a large number of locations on the machine structure and bearings, covering the frequency range up to more than twice shaft rotational frequency. There were large differences in some frequency ranges between the static and operational conditions. Vibration due to machine operation was then measured to allow deduction of bearing forces using the transfer function matrices. Repeat measurements were made to establish whether machine source properties changed significantly with time, while statistical techniques were also used to identify and eliminate any suspect measurements. Those early experiments are described in this paper with a view to future application of similar techniques.