This paper presents an overview of recent experimental and numerical investigations on industrial railway brakes. The goal of the present study is to discuss the relevance of the mechanical modeling strategy for squeal prediction. Specific experimental set-ups based on transient and controlled braking tests are designed for this purpose. Measurements are performed on it to investigate the dynamic behavior of TGV squeal noise and its squeal characterization through experiments. It will be demonstrated that it is possible to build consistent and efficient finite element models to simulate squeal events in TGV brake systems. The numerical strategy will be presented, including not only the modeling of the TGV brake system and the stability analysis, but also the transient nonlinear dynamic and computational process based on efficient reduced basis. This complete numerical strategy allows us to perform relevance squeal prediction on industrial railway brakes. This study comes within the scope of a research program AcouFren that is supported by ADEME (Agence De l'Environnement et de la Maîtrise de l'Energie) concerning the reduction of the squeal noise generated by high power railway disc brakes

International audienceAs noise reduction tends to be part of environmental directives, predicting squeal noise generated by disc brakes is an important industrial issue. It involves both the transient and stationary nonlinear dynamics of self-excited systems with frictional contact. Time simulation of the phenomenon is an attractive option for reducing experiment costs. However, since such computations using full finite element models of industrial disc brake systems is time-consuming, model reduction has to be performed. In this paper, both the transient and stationary nonlinear behaviors of the friction destabilized system and the effect of dynamical reduction on the nonlinear response of a simple friction destabilized system are carried out. The first part provides a description of the general modeling retained for friction destabilized systems. Then, discretization and solving processes for the stability analysis and the temporal evolution are presented. The third part presents an analysis of a sliding elastic layer for different operating conditions, in order to better understand the nonlinear behavior of such systems. Finally, spatial model reduction is performed with different kinds of reduction bases in order to analyze the different effects of modal reductions. This clearly shows the necessity of including static modes in the reduction basis and that nonlinear interactions between unstable modes are very difficult to represent with reduced bases. Finally, the proposed model and the associated studies are intended to be the benchmark cases for future comparison

No abstract

The complex eigenvalue analysis is a widely used technique to investigate the stability of a dynamical system with frictional contact. In the case of brake systems, it is the most frequently employed method to study the propensity of the brake to generate squeal noise. When finite elements models are considered, iterative solvers are needed to calculate the complex modes and eigenvalues with good precision. In practice, reduced real bases are often used in order to reduce the computational times. However, great attention should be focused on the errors introduced by the reduction, which is rarely done. In this paper, the performances of some reduced bases are investigated in the case of a simple disc/pads system. Bases composed of real coupled modes and bases provided by Component Mode Synthesis (CMS) techniques are tested. An enrichment of these bases is proposed in order to improve the precision of the results. In particular, new rubbing attachments modes are proposed to adapt free-interface CMS techniques to frictional contact. When real coupled modes are used, it is suggested to complete the basis by the static response to a first-order approximation of the friction forces. Applied to the disc/pads model, the different enrichment options allow a reduction of the errors on frequencies, divergence rates and mode shapes by a factor comprised between 10 and 100 without significantly increasing the computational times.