This study presents an artificial neural network (ANN)-based surrogate modeling framework for predicting key performance metrics of automotive brake discs using data derived from finite element analyses. A dataset comprising 200 geometrically parameterized design points was generated via Finite Element Method (FEM) simulations, including static structural, transient thermal and modal analyses in ANSYS/Workbench. The target outputs, maximum von Mises stress, peak temperature, first natural frequency and total volume were used to train a multi-output ANN implemented in MATLAB. The trained model demonstrates strong predictive capability with high correlation coefficients (R>0.98) and low Mean Absolute Percentage Error (MAPE <2.5%) across all metrics. Validation on six independent test cases confirmed the ANN’s generalization ability and accuracy in replicating FEM results within acceptable error margins. The proposed approach significantly reduces the computational burden associated with conventional FEM simulations while maintaining reliable accuracy, making it an efficient tool for early-stage design evaluation and multi-objective optimization of brake disc components.