.png){kind=icon}
Return
An adaptive constrained multi-objective evolutionary algorithm based on problem attribute discrimination
.png){kind=icon}
PRE
AI
DOI：10.1007/s10586-026-06443-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Balancing feasibility, diversity and convergence is a major challenge for solving constrained multi-objective optimization problems (CMOPs). For solving CMOPs with different attributes, the emphasis on these three aspects should vary. This paper proposes an adaptive constrained multi-objective evolutionary algorithm based on problem attribute discrimination (PRAC-CMOEA). Firstly, three kinds of CMOP attributes (feasibility hardness, diversity hardness and convergence hardness) are defined, where feasibility and diversity hardness belong to global attributes while convergence hardness belongs to local attribute. Then, PRAC-CMOEA discriminates the types of global attributes exhibited by the problem after early stages of evolution and employs appropriate strategies to address different combinations of attributes. Local attributes are monitored in the later stages of evolution and escape strategy is activated if the convergence hardness is detected. Benchmark problems with different attributes are selected for experimental studies. Experimental results suggest that PRAC-CMOEA is highly competitive against the peer algorithms on benchmark problems. Furthermore, PRAC-CMOEA is used to optimize the control system of multi-wire cutting winding tension system. The results show that the optimized system has short rise time, small maximum overshoot, short setting time and better robust stability.
Keywords:
Constrained multi-objective optimization
Dual population
Evolutionary algorithm
Winding tension control system optimization
Problem attribute classification
Journal
IF:
2.9
Papers:
221
Citations:
1.1K