Friction modifiers (FMs) are used to reduce rail-wheel contact noise, wear, derailment ratio, rolling contact fatigue, etc. without compromising the traction and braking requirements. In this work, a novel method for application of friction modifier coatings is presented. The method is environment friendly, energy efficient, and can be used to deposit thin coatings of a wide range of materials over large uneven surfaces. Here, zinc oxide, zinc peroxide, graphite, bentonite, molybdenum disulphide, and talc coatings are obtained using the new deposition method. Frictional performance of coatings is determined using ball on disc tests. Temperature of contact zone in ball on disc tests is monitored throughout the experiments in all tests. A surrogate model employing artificial neural network (ANN) is used to predict the coefficient of friction as a function of contact temperature and normal load for all friction modifier coatings at different sliding speeds. The model demonstrated robust performance with an average mean absolute error of 0.0358, and a root mean square error of 0.0462 across all testing cases, indicating reliable predictive capability despite some localized anomalies.