This work reviews the major published studies, both theoretical and experimental, that address the impact of wave motion on packed tower performance. Current practice is to add excess packing to guarantee the required separation is obtained, though there is little data available to derive a safety factor for packing height. This work highlights deficiencies in the current knowledge base and analyzes general trends to address common misconceptions about tower design for floating production. Tilt and motions imposed on a fractionation column have a significant impact on product specifications due to reduced packing efficiency. Improved awareness of motion impacts will assist the analysis of tower design and allow feedback to the process design. Identifying gaps in available data shows limits in the current understanding and allows the development of appropriate simplifying assumptions. The current understanding of towers in motion allows only very basic design rules and the safety factor for packing height in literature varies from 1.1 to 2.0. This provides little confidence in the ability to predict packing efficiency for floating production. Static tilt is more detrimental to liquid distribution than motion at a given amplitude. Still, liquid maldistribution from motion approaches that of static tilt as the period increases. Liquid sloshing, often cited as a significant concern, is a relatively minor contribution to maldistribution except for short periods and tall towers. The relative bed size has a significant impact on liquid maldistribution, limiting the recommended maximum bed height:column diameter to 2 – 3 to maximize the efficiency. Approach to equilibrium is often overlooked as the major determinant of sensitivity of efficiency to maldistribution. Separations that operate near equilibrium are more sensitive to maldistribution than services with a large driving force. Thus, distillation towers are generally more sensitive to motion than absorbers and sensitivity may vary over column height. Recommendations are provided for a bed-by-bed analysis and feedback to the process design. Floating production units require additional complexity and conservatism in tower design. An improved awareness of motion effects on towers, and the factors involved, will lead to improved designs and a reduction in the cost of floating production facilities.

Dynamic analysis models are customarily employed in turbomachinery design to predict critical whirling speeds and estimate dynamic response due to loads imposed by unbalance, misalignment, maneuvers, etc., Traditionally these models have been assembled from beam elements and been analyzed by transfer matrix methods. Recently there has been an upsurge of interest in the development of improved dynamic models making use of finite element analysis and/or component mode synthesis. We are currently developing a procedure for modelling and analyzing multi-rotor systems [1] which employs component mode synthesis applied to rotor and stator substructures. A novel feature of our procedure is a program for the automatic generation of the component modes for substructures modelled as Timoshenko beam elements connected to other substructures by bearings, couplings, and localized structural joints. The component modes for such substructures consist of constraint modes and internal modes. The former are static deflection shapes resulting from removing the constraints one at a time and imposing unit deflections at the constraint locations. The latter have traditionally been taken to be a subset of the natural modes of free vibration of the substructure with all constraints imposed. It has however been pointed out [2] that any independent set of geometrically admissible modes may be used. We take advantage of this and employ static deflections under systematically selected loading patterns as internal modes. All component modes are thus obtained as static deflections of a simplified beam model which has the same span and same constraints as the actual substructure but which has piecewise uniform dynamic properties. With the loading patterns we employ, all modes are represented by fourth order polynomials with piecewise constant coefficients. We have developed an algorithm for the automatic calculation of these coefficients based on exact integration of the Timoshenko beam equation using singularity functions. The procedure is illustrated by applying it to a simplified system with a single rotor structure and a single stator structure. The accuracy of the procedure is examined by comparing its results with an exact analytical solution and with a component mode synthesis using true eigenfunctions as internal modes.

No abstract