We study the early work scheduling problem on identical parallel machines in order to maximize the total early work, i.e., the parts of non-preemptive jobs executed before a common due date. By preprocessing and constructing an auxiliary instance which has several good properties, we propose an efficient polynomial time approximation scheme with running time $O(n)$, which improves the result in [Gy{o}rgyi, P., Kis, T. (2020). A common approximation framework for early work, late work, and resource leveling problems. {it European Journal of Operational Research}, 286(1), 129-137], and a fully polynomial time approximation scheme with running time $O(n)$ when the number of machines is a fixed number, which improves the result in [Chen, X., Liang, Y., Sterna, M., Wang, W., B{l}a.{z}ewicz, J. (2020b). Fully polynomial time approximation scheme to maximize early work on parallel machines with common due date. {it European Journal of Operational Research}, 284(1), 67-74], where $n$ is the number of jobs, and the hidden constant depends on the desired accuracy.