Weijian Ren

An improved ant colony algorithm and its application in TSP

The default that the factor of pheromone remnant is constant can make the algorithm fall into the local optimal solution easily, a new factor of pheromone remnant is proposed which is varying with the iterative time. But this can not overcome the local one completely, the bit exchanging mode with the shifting windows is proposed by the analysis in the solution set. Above all an improved ant colony algorithm is got, then it applies in the TSP, the simulation shows that this algorithm can improve the ability of searching global optimization and overcome premature convergence. Comparison with the basic ant colony algorithm shows that the algorithm is effective.

Fault diagnosis of progressing cavity pump well based on wavelet package and Elman neural network

In this paper, the fault diagnosis problem is studied for progressing cavity pump well base on wavelet package and Elman neural network. The signals of active power can fully reflect the status of progressing cavity pump wells. A new fault diagnosis method for cavity pump wells is presented. This method uses wavelet time-frequency analysis technology for de-noising and filtering of active power signals, uses 3-layer db4 wavelet packet to decomposition fault signal of different frequencies, extracts fault feature based on changes in band power spectrum, then use Elman neural network to identify the fault. By use of Matlab simulation, the results show that this method can effectively improve the diagnostic accuracy of progressing cavity pump wells.

A kind of adaptive Immune Genetic Algorithm based on Chaos and its application

In order to overcome the shortcomings of Immune Genetic Algorithms relatively slow convergence rate during the process of solving large-scale optimization problem, given Chaos optimizations benefits of sensitive to the initial value, easy to jump out of local minimum point, the fast search speed, global asymptotic convergence and so on, basing on the both search advantages of Immune Evolutionary Algorithm and Chaos Optimization Algorithm in their own space, considering that chaotic sequence could be simulated the proliferation of immune cells approach, we combined their characteristics of Chaos Optimization Algorithm and Immune Genetic Algorithm, making use of “exchange” and “shift” operations to solution matrix and memory matrix during the optimization process of chaos, and doing adaptive adjustment to selection probability and mutation probability during the Genetic operation, a kind of adaptive Immune Genetic Algorithm based on Chaos is proposed. The validity of the new algorithm is verified by real application.

Fault diagnosis based on WNNs with parameters optimization by immune evolutionary Particle Swarm Algorithm

The immune evolutionary mechanism of artificial immune system is used into Particle Swarm Optimization(IEPSO). A new training algorithm in wavelet neural networks(WNNs) based on IEPSO is presented, it can avoid early ripe of PSO and traditional BP algorithm. In the course of optimizing the parameters of WNNs, new algorithm use the immune evolutionary principle to improve the process of PSO, it determines the probability of their choice based on the size of fitness and concentration in antibodies, and dynamically adjusted crossover probability and mutation probability by use of fitness function. With the parameters optimized by IEPSO, the convergence performance of the WNNs is improved. The fault diagnosis of progressing cavity pumps well shows that the WNNs optimized by IEPSO can give higher recognition accuracy than the normal WNNs.