Abstract: A multi-agent based adaptive optimization scheduling method is proposed to address the problems of delayed response, high cost, and insufficient equipment utilization caused by dynamic disturbances such as order changes and equipment failures in mechanical manufacturing workshops. Firstly, a multi-agent system interaction behavior mechanism framework is constructed to achieve information sharing and task collaborative allocation through standardized interaction, generate executable scheduling plans, and quickly trigger rescheduling when dynamic disturbances occur. After completing the interaction design, an adaptive scheduling model is constructed and constraints are set. The differential evolution algorithm is used to solve the model. Firstly, the scheduling parameter set is defined and merged to generate the initial population. During evolution, the objective function value is dynamically evaluated to screen parameters. Later, a non dominated solution set management strategy is used to maintain diversity. After reaching the preset iteration times, the non dominated solution set is output as the optimal parameter combination to achieve adaptive optimization scheduling. The experimental results show that compared with traditional scheduling methods, this method reduces the cost of mechanical production operations by 28.1% and 33.3%, respectively. While ensuring maximum economic benefits, the comprehensive equipment full load rate reaches over 95%, effectively reducing resource waste.