In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing convention...In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing conventions problem is solved absolutely. Heuristic methods are used to constrain the topology mutation probability and the trend of mutation kind choice. Also, the niching method is used to protect the network topologies evolution. The experiment results show the efficiency and rapidity of NEGA forcefully.展开更多
The complexity of aeroengine external piping systems necessitates the implementation of automated design processes to reduce the duration of the design cycle.However,existing routing algorithms often fail to meet desi...The complexity of aeroengine external piping systems necessitates the implementation of automated design processes to reduce the duration of the design cycle.However,existing routing algorithms often fail to meet designer requirements because of the limitations in providing a single solution and the inadequate consideration for route constraints.In this study,we propose the multi-solution pipe-routing method for aeroengines.This method utilizes a hybrid encoding approach by incorporating fixed-length encoding to represent route constraints and variable-length encoding and indicate free-exploration points.This approach enables designers to specify route constraints and iterate over the appropriate number of control points by employing a modified genetic iteration mechanism for variable-length encoding.Furthermore,we employ a pipe-shaped clustering niche method to enhance result diversity.The practicability of the newly proposed method is confirmed through comparative experiments and simulations based on the“AeroPiping”system developed on Siemens NX.Typical solutions demonstrate significant differences in circumferential and axial orientations while still satisfying engineering constraints.展开更多
文摘In this paper, a new neuroevolution algorithm (NEGA) for simultaneous evolution of both architectures and weights of neural networks is described. A whole new network encoding method is shown. The competing conventions problem is solved absolutely. Heuristic methods are used to constrain the topology mutation probability and the trend of mutation kind choice. Also, the niching method is used to protect the network topologies evolution. The experiment results show the efficiency and rapidity of NEGA forcefully.
基金supported by the National Science and Technology Major Project,China(Grant No.J2019-I-0008-0008).
文摘The complexity of aeroengine external piping systems necessitates the implementation of automated design processes to reduce the duration of the design cycle.However,existing routing algorithms often fail to meet designer requirements because of the limitations in providing a single solution and the inadequate consideration for route constraints.In this study,we propose the multi-solution pipe-routing method for aeroengines.This method utilizes a hybrid encoding approach by incorporating fixed-length encoding to represent route constraints and variable-length encoding and indicate free-exploration points.This approach enables designers to specify route constraints and iterate over the appropriate number of control points by employing a modified genetic iteration mechanism for variable-length encoding.Furthermore,we employ a pipe-shaped clustering niche method to enhance result diversity.The practicability of the newly proposed method is confirmed through comparative experiments and simulations based on the“AeroPiping”system developed on Siemens NX.Typical solutions demonstrate significant differences in circumferential and axial orientations while still satisfying engineering constraints.
