Swarm Intelligence Optimization Algorithm For Complex Optimization Problems And Applied Research

Swarm intelligence algorithms cover a class of heuristic search algorithms whose optimization process does not rely on structural information of the algorithm.Therefore,swarm intelligent optimization algorithms can well solve optimization problems with nondifferentiable,nonlinear,or non-convex objective functions.Particle Swarm Optimization(PSO)algorithm is one of the swarm intelligence optimization algorithms,which has simple implementation and fast convergence speed.Due to these characteristics,it can solve many practical application problems well and has been widely used in various scientific research and industrial production applications related to artificial intelligence.However,the PSO algorithm has some disadvantages,such as premature convergence and insufficient solution accuracy.Random Drift Particle Swarm Optimization(RDPSO)algorithm is one of the variants of the PSO algorithm,which has good search performance,fewer parameters,and is easy to experiment.Therefore,in this paper,the RDPSO algorithm is chosen as the focus of research to improve it for different complex optimization problems.The research of this paper is as follows:In order to solve the situation that Random Drift particle swarm optimization algorithm is difficult to solve problems with complex constraint spaces,this paper proposes Random Drift particle swarm optimization algorithm(RMRDPSO)based on Riemannian manifolds.Based on the property that the Riemannian manifold is locally homogeneous in the Euclidean space,the complex constrained problem is converted into an unconstrained problem on the Riemannian manifold.In each optimization process,the particles on the manifold are mapped onto its tangent space,the tangent vector represents the velocity of the particles,the velocity is updated on the tangent space,and then the velocity is retracted back onto the manifold to update the position of the particles.This paper also proposes to use inverse retraction mapping instead of logarithmic mapping to achieve the mapping from the tangent space to the manifold.Experiments have shown that the convergence performance of the RMRDPSO algorithm proposed in this paper is superior to other optimization algorithms on manifolds,and the inverse contraction mapping can reduce the running time of the algorithm.For the multi-objective problem,this paper proposes a multi-objective Random Drift particle swarm optimization algorithm(MORDPSO-DA)based on a dual-archive mechanism.First,Random Drift particle swarm optimization algorithm is applied to the multi-objective optimization by adopting a multi-scale chaotic variational operation on the particles to enhance the ability of the particles to jump out of the local optimal solution.A dual-archive mechanism is also proposed to establish an auxiliary archive with a capacity threshold in addition to the main archive external to the non-dominated solution set.The particles deleted from the main archive are censored according to the congestion distance,and the particles deleted from the main archive are saved in the auxiliary archive.When the capacity of the auxiliary archive reaches the threshold,the particles in the auxiliary archive are compared with those in the main archive,and the main archive is updated according to the congestion distance,and the particles that are beneficial to the diversity of the solution set can be retained.Finally,this article conducted experimental tests on the ZDT and DTLZ benchmark test functions,and the results showed that the MORDPSO-DA algorithm achieved better results in IGD,GD,and SP,demonstrating good convergence and diversity.A data-driven intelligent optimization method is proposed and applied to the propeller optimization design.First,in order to improve the accuracy and interpretability of the performance fast prediction proxy model,the binary random drift particle swarm optimization algorithm and BP neural network are combined to feature select the propeller design parameters,and then the BP neural network is used to establish the propeller performance fast prediction proxy model.Then the open water efficiency,the minimum pressure coefficient of the back of the blade and the pulsation pressure performance surrogate model are simultaneously used as the objective function,the expert empirical knowledge is the constraint of the design parameters,and the MORDPSO-DA algorithm is the optimization algorithm to realize the propeller design optimization.Finally,the visualization platform is built to realize the surrogate model training function and model optimization function.And the propeller surrogate model is optimized under different operating conditions.The experiments show that the proposed method can provide a better propeller design solution with better performance and meet the actual production requirements.In summary,this paper uses Random Drift particle swarm optimization algorithm and proposes two improved Random Drift particle swarm optimization algorithms for different complex optimization problems,and the effectiveness of the algorithms is confirmed by extensive experiments.And a data-driven intelligent optimization method is proposed and applied to the propeller design,which can design a propeller with better performance.Swarm intelligence optimization algorithm for complex optimization problems and applied research...