During multi-layer additive manufacturing, multiple thermal cycles and addition of dissimilar-metal powder, even functionally graded materials (FGMs), lead to complicated transport phenomena and solidification behavior in the molten pool, which significantly impact the microstructure evolution and mechanical properties of deposited part. In this study, a predictive three-dimensional numerical model is developed to understand the multi-physical processes such as thermal behavior, Marangoni effect, composition transport, solidification behavior, and dendrite growth in multi-layer additive manufacturing of Ni-based alloy on cast iron. Dimensional analysis is performed to simplify the force balance equation on the liquid-gas interface, which determines the dynamic profile of molten pool. The conservation equations of mass, momentum, enthalpy and concentration are solved in parallel. Transient temperature distribution and thermal cycles at different locations are obtained. The solidification parameters at the liquid-solid interface are evaluated to interpret the solidification microstructure. The distribution of alloy elements and composition profile (Ni and Cr) are also present and compared with the relevant experimental results. The results show that the cooling rate declines progressively as the subsequent layers deposit, which results in the coarser solidified grains in the upper of part. Even though the powder and substrate can be efficiently mixed to be a homogeneous molten pool, a non-uniform concentration distribution is observed at the bottom of the deposited part, which agrees well with the experimental composition profile using Energy Dispersive Spectrometer (EDS). (C) 2017 Elsevier Ltd. All rights reserved.