Biologically Inspired Computation And Blade Optimal Design Technical Of Axial Compressor

Compressor blade optimization design is critical to the enhancement of the overall performance of the compressor.The traditional gradient-based optimization algorithm requires the objective function to be continuously derivable,and the derivation information of the objective function about the optimized variables needs to be obtained.The procedure is complex and easy to fall into a local optimal solution or repeated oscillation with poor generalizability.Bionic intelligent algorithm and swarm intelligence algorithm have been broadly applied as excellent global optimization algorithms.This dissertation is devoted to studying the performance improvement technology of the novel bionic intelligent algorithm and its application in the optimization design of the compressor blades with the following achievements:1.This study develops the evolutionary and swarm intelligence algorithm based on distributed and parallel computing architecture.Genetic algorithm,differential evolution and artificial bee colony algorithm are accelerated and optimized by virtue of the parallel architecture,which greatly eased the time-consuming problem of solving the objective functions.This study also optimized the running mechanism of the algorithm per se.The algorithm of variable ratio trisomy crossover genetic algorithm,Gaussian random scaling differential evolution algorithm and neighborhood selection bee colony algorithm are applied,which effectively improves the global optimization performance of the algorithm.2.This study addresses the issue of difficulties in choosing parameters for SVM model by integrating SVM with bee colony algorithm to achieve the global optimization of SVM parameters.Compared with the traditional neural network,SVM based on the theory of structure risk minimization not only fully considers the sample's influence on the system,but also considers structural features of the optimization problem to achieve the optimal generalization performance.Experiments show that SVM performs better than BP network in the test sample set.The proposed support vector machine based on bee colony algorithm(ABC-SVM)balances the model fitting on the training samples and generalizability to new samples,,which helps the model to reach the global optimum.3.This study designs and implements the parameterized system.The implemented algorithms include Hicks-Henne,CST and NURBS.By studying NURBS in depth with trials and errors in model adjustment,this study puts forward the empirical formulas for compressor blade fitting.The problem that NURBS is prone to unconventional kinking is also solved by using NURBS parameterization.4.Adsorption leaf blade increases the aerodynamic bending angle by the suction of the attached layer with strong pressure gradient on the surface of the airfoil,so that the compressors improve their work ability.In order to obtain suitable leaf shape and suction parameters,this paper designs and develops a set of intelligent optimization systems for the shape design of leaf blade using artificial bee colony algorithm and NURBS parameterization method.The system can intelligently optimize conventional leaves and adsorptive leaves.The effectiveness of the optimization results is verified by cascade experiments.5.In order to reduce the time consumption in the optimization procedure,a DE-ABC-SVM algorithm based on Latin hypercube sampling is implemented,which replaces the real flow solver.This algorithm acquires the simulated values of flow field parameters within a shorter period of time,and continuously improves the accuracy of the algorithm by constantly adding new sample points.The entire optimization process is a continuous model optimization process,from which the final outputs are the optimized results and the optimal algorithm model.The set of algorithms is verified by optimizing the existing adsorptive leaf model with improved results.