Brake linings have complex microstructure and consist of different components. Fast growing automotive industry requires new brake lining materials to be developed at considerably shorter time periods. The purpose of this research was to generate the knowledge for optimizing of brake friction materials formula with mathematical methods which can result in minimizing the number of experiments/test, saving development time and costs with optimal friction performance of brakes. A combination of processing methods, raw materials and testing supported with the Artificial Neural Network (ANN) and Taguchi design of experiment (DOE) allowed achieving excellent results in a very short time period. Friction performance and wear data from a series of Friction Assessment and Screening Test (FAST) were used to train an artificial neural network, which was used to optimize the formulations. The averaged COF, COF variation and wear were used as the output parameters. Weight percentage of raw materials denoted as an input parameter and these data were used to train ANN. A two layer feedforward ANN with back propagation was used in this study. Back propagation learning algorithm can be divided into two phases: i) propagation and ii) weight update. The friction performance of the optimized friction materials was considerably better when compared to the baseline commercial brake lining materials. This method can be applied to development of any type of complex friction materials.