The automotive industry is one of the drivers of CAE-based virtual product development. Due to a highly competitive market, development of innovative, high quality products within a short time is necessary and it is only possible by using virtual prototyping. It is important to note that increased application of virtual prototyping itself increases the necessity to perform robustness studies. If the number of hardware tests has to be reduced, it is essential to implement the scatter, which is always present in these tests (such as loads, material, geometry), into the computational model. Consequently, probabilistic methods using CAE-based stochastic analysis have to be utilized in order to quantify robustness, safety and serviceability.
Brake noise is one of the most important problems in the automobile industry due to the high warranty costs. The generation of brake noise is due to the development of instabilities in the brake system. The analysis of brake squeal is highly complex and it is also very sensitive to the operation conditions. Therefore, the scope of this work is to carry out a robustness analysis for analyzing the behaviour of the system due to a change in the above-mentioned parameters. Robustness analysis is primarily carried out to determine the variation range of significant response variables and their evaluation by using definitions of system robustness due to the unavoidable scatter of design parameters. The imperfections of the design parameters are usually modelled by either random variables that are constant in space or random fields that vary in space.
In this paper, material and geometrical tolerances are considered. Material tolerances are modelled using random variables which have been in existence for long time, but the geometrical tolerances are modelled using the random field, which has been used in this field recently. From robustness analysis, the transfer behaviour of design parameter dispersion to important NVH performance criteria is investigated. As a result, the design parameters responsible for the main scatter of responses are identified, which in turn leads to the information regarding the improvement of design. The probabilistic and structural analyses are performed with optiSLang and Nastran software programs.