Parallel Coevolution Of Intelligent Optimization Algorithms And Application In Fault Diagnosis

In the gradual development of the modern engineering field,many new application scenarios and problems have emerged,which are usually complex and variable,and require comprehensive research and solution across disciplines.Cooperative swarm intelligent optimization algorithm is a cooperative learning optimization algorithm,which is also based on intelligent algorithm.In tool condition monitoring,coevolution algorithm can be applied to tool fault diagnosis.Through the integration and analysis of different sensor data,the state model of the tool is established,and the coevolution swarm intelligence algorithm is used to predict the tool life and optimize the maintenance,so as to realize the efficient use of the tool and reduce the maintenance cost.Through the appropriate swarm structure design and parameter adjustment,the cooperative swarm intelligence optimization algorithm can realize the search and optimization of the global optimal solution.In view of the above content,the specific research content of this thesis is as follows:(1)This thesis proposes a parallel symbiotic lion swarm optimization algorithm based on Latin hypercube initialization distribution.The algorithm used Latin hypercube to generate the initial population to improve the diversity of the population.The mutualistic symbiosis mechanism was proposed to reduce the probability of boundary crossing when the lioness updated,and the directionality of the algorithm was strengthened.To improve the ability of jumping out of local optimum,the opposition-based learning strategy for dimension by d imension hole imaging was proposed.The parallel evolutionary computing is combined with the improved lion swarm optimization algorithm to improve the convergence speed and speed up the calculation efficiency.The algorithm was simulated on several benchmark functions.The experimental results show that the accuracy and stability of the improved algorithm are greatly improved,and more operation time can be saved after adding parallel evolutionary computation.(2)This thesis proposes a clustering competition and coevolution mechanism to further improve the efficiency and stability of the algorithm,increase the diversity of the optimization mechanism of the lion swarm optimization algorithm,enhance the scope of application of the algorithm and accelerate the calculation.Clustering competition and coevolution mechanism generates sub-populations through clustering algorithm.The sub-populations were combined and optimized by ring topology structure,and the optimization was performed in parallel.The information space was introduced,and the historical information of sub-populations was shared during parallel computing to assist competitive coevolution computing.GPU-assisted computing is used to improve the operation efficiency.Benchmark functions were used to conduct simulation experiments to evaluate the performance of different combination methods of coevolution algorithms.The experimental results show that the clustering competition coevolution algorithm is better in optimization accuracy and convergence speed,and the performance of the combination of lion swarm optimization algorithm,particle swarm optimization algorithm and equilibrium pool optimization algorithm is better.Some computation time can be saved by GPU acceleration.The clustering coevolution algorithm is applied to the feature selection classification problem,and the results show that the classification effect is good.(3)This thesis,a tool condition monitoring model integrating multiple methods is proposed to improve the monitoring efficiency and accuracy.Specifically,in this thesis,wavelet transform,swarm intelligence optimization algorithm and machine learning techniques are used to extract a variety of features of tool vibration signals,and the optimal feature subset is screened out by clustering competition cooperative optimization algorithm.Then,the support vector machine was used to train the integrated feature matrix,and the swarm intelligence optimization algorithm was used to optimize the model to reduce the feature dimension.Finally,combined with the training classification model of neural network,the tool state was output by sound and display.The experimental results show that the method proposed in this thesis can effectively diagnose whether the tool is damaged,and has the advantages of fast and accurate.