To develop a coarse-grained model that is both computationally efficient and accurate enough to capture the molecular interaction of long polymer chains, we form the segmented block copolymers of TPU with a lower degree of polymerization. The model is constructed with the amorphous module of Materials Studio with a density of 0.85 g/cm^{3}, which is close to the experimentally measured one $(0.97\u22121.24g/cm3)$ [20]. An initial length of a cubic simulation box is 7.7 nm. We define beads in the full-atomistic model for coarse-graining. In general, a bead in coarse-graining represents one monomer to describe the interactions among the repeated units, but in this study, we decompose TPU into multiple beads. It is known that the elasticity of TPU is mostly dictated by hard and soft segments, where the hard segments come from diphenylmethanediisocyanate (MDI) and the soft segments come butanediol [18]. Therefore, we consider the segments coming from MDI and butanediol as separate beads, which will enable us to obtain interactions between them. The segments from MDI consist of two phenyl rings connected by a $\u2212CH2$ bond. Due to the long chain of this segment, it may be difficult to capture the internal interactions. Therefore, we divide the hard segment into three beads for an improved coarse-graining of TPU. As illustrated in Fig. 1, the coarse-grained model of TPU has three types of beads: A, B, and C, whereas A + B + C symbolizes the hard segments, B + C represents diisocyanate and only A denotes the soft segment of TPU. Therefore, we construct a full-atomistic chain with a segmented block copolymer structure as $[(ABC)n1\u2212(A)n2\u2212(ABC)n1\u2212(A)n2\u2212(ABC)n1]$, where $n1=n2=4$. This makes 44 beads in a chain, equivalent to 644 individual atoms. Therefore, a total number of atoms of TPU with 50 chains in the full-atomic model is 32,200. To avoid extensive simulation time, we are restricted to the number average molecular weight of the chains as 4684.77 g/mol, which may be too small compared with the physical value $(25,000\u2212100,000g/mol)$ to capture the phase separation [19]. This will be discussed in the coarse-grained model section for more detail. However, the energy of the original model is minimized using the discover module of Materials Studio for 5000 fs with a 1 fs time step. The model is equilibrated using an NPT ensemble (constant number of atoms, pressure, and temperature) at atmospheric condition (298 K and 1 atm) for 500,000 fs with a 0.1 fs time step. To get an equilibrium model at the molecular level, the full-atomic model is heated to 600 K with a 25 K increment (run for 100,000 fs with a 0.1 fs time step for each 25 K), followed by cooling down to room temperature. At this stage, the density obtained through the simulation is 0.9 g/cm^{3}. Subsequently, an NVT ensemble (constant number of atoms, volume, and temperature) is run for 100,000 fs with a 0.1 fs time step. Twenty trajectories are saved to get the distribution of bond, angle, and radial functions among beads associated with the effective potential functions of the coarse-grained model.