Resumen de Efficient simulation of the pantograph-catenary dynamic interaction. Catenary optimisation and installation error analysis
Modelling and simulation of the dynamic interaction between pantograph and catenary has become a powerful tool to expedite the catenary design process since, among other advantages, it helps in reducing the number of the costly experimental in-line tests.
In order to tackle these numerical simulations, in this Thesis the catenary system is modelled by the Finite Element technique, while a simple lumped-mass model is used for the pantograph. The interaction between the two systems is accomplished with a penalty formulation. After solving the initial nonlinear configuration problem, the equation of motion is linearised with respect to the static equilibrium position and it is then solved by applying the Hilber-Hughes-Taylor (HHT) time integration method. However, dropper slackening and pantograph contact losses are two sources of nonlinearities which must be considered in the solution procedure at the expense of an increase in the computational cost.
The main objectives of this Thesis are both to find optimal catenaries in terms of current collection quality and to analyse the effect of installation errors in the dynamic behaviour of the system. To achieve these goals, it is mandatory to perform a large number of pantograph-catenary dynamic simulations for which the computational cost can become prohibitive.
In order to reduce this computational effort, the first proposal made in this Thesis is to precompute a parametric solution of the pantograph-catenary dynamic interaction for all values of the design variables, by means of the Proper Generalised Decomposition (PGD) technique. If dropper lengths are considered as design variables, this parametric approach is successful when applied to the static equilibrium problem. Nevertheless, in the dynamic case, when dropper slackening is considered, the solution exhibits a great sensitivity to small changes in the parameters and therefore, a huge number of PGD modes are required to obtain the parametric solution with enough accuracy.
The impossibility of having a parametric solution leads the author to propose a fast strategy to simulate the dynamic interaction problem, providing remarkable saves in computational cost. The method is divided into two stages which are based on moving the nonlinear terms to the right hand side of the dynamic equation. In the first stage, the response of the system under unitary forces is precomputed and stored. Then, in the second stage of the method, the treatment of the nonlinearities is condensed into a small system of equations, whose unknowns are now the forces associated with the nonlinearities instead of the nodal displacements of the whole system.
With this proposed algorithm, it is possible to carry out efficient optimisations of the catenary geometry. Specifically, contact wire height and dropper spacing are considered as design variables in order to obtain the most uniform interaction force that leads to the optimal current collection. The optimisation problem is solved by means of a classic Genetic Algorithm, applied to both simple and stitched catenaries. The results obtained show that an optimal catenary design can remarkably improve the current collection quality of the actual catenaries.
Finally, the influence of the installation errors on the dynamic behaviour of the system is analysed under a stochastic approach in which variability in dropper length, dropper spacing and support height are involved in the simulations. The use of a Monte Carlo method allows the propagation of the uncertainty to the magnitudes of interest of the dynamic solution and therefore, to obtain their probability density function. The results of Monte Carlo simulations demonstrate that dropper spacing errors are slightly influential, whilst dropper length and supsupport height installation errors have a strong influence on the dynamic behaviour of the system.
