This work aims to find an efficient and fast solution for the problem of plant-wide optimization of a large-scale industrial process. The process is composed of a series of unit processes with coupling relationships between them, which leads to the problem of optimizing a large number of design variables and constraints. Therefore, the solution to the optimization problem is time-consuming and the global optimum is obtained with a low probability. To efficiently solve the problem, a new plant-wide optimization method is developed that decomposes the complicated optimization problem into several easier solvable sub-problems based on the coupling relationships between unit processes. Two metrics, constraints coupling and indices coupling, are defined in order to measure the extent of a coupling relationship between pairs of unit processes. Then, the global optimum is found by successively solving the sub-problems. Finally, the hydrometallurgy process of a gold smelter is demonstrated to evaluate the performance of the proposed method.