<p>Despite extensive research on the assembly of simple components, studies addressing geometrically complex workpieces assembly tasks in steelmaking remain limited. These tasks are characterized by high complexity, harsh environments, and significant physical demands, necessitating advanced automation solutions. To address these challenges, this paper proposes a four-phase assembly strategy for the robotic assembly of ladle sliding gate replacement in the steelmaking industry, particularly focusing on the plate-in-groove (PiG) task. PiG tasks involve peg-in-hole assembly, requiring robots to integrate precise vision and force sensing. The proposed method first re-estimates the workpiece pose using an alternating direction method of multipliers-based contact-aware pose estimation, reducing pose errors on five degrees of freedom. Then, during the rotation process, force/torque feedback is provided to the adaptive screw motion controller to adjust the rotation center position and angular velocity of the sliding plate, thereby reducing the contact force. Finally, the insertion process is completed through an admittance controller. Throughout the four-phase assembly framework, six-degree-of-freedom errors are iteratively minimized, improving success rate. Experimental results validate the stability and robustness of the proposed approach in real industrial field environments.</p>