Gong Dun-wei

Adaptive niche hierarchy genetic algorithm

The niche technology and hierarchy genetic algorithm are combined in this paper. Adaptive hierarchy genetic algorithm based on niche sharing and crowding strategy is presented. The algorithm improves genetic algorithm from not only encoding method but also operators. Thus it has strong search ability and it is easy to find many global optimums. Simulation results validate its efficiency.

Research on application of recurrent neural network in modelling jigging system of coal preparation plant

Recurrent neural network can describe complex nonlinear time-varying dynamic system. In this paper, model and unified mathematical description of recurrent neural network are presented. The BP algorithm to train this network is also presented. Finally, this network is applied to model jigging system of coal preparation plant. An example shows that this network has strong capability in dynamic approach.

Local robot path planning with predicting band trajectories of obstacles

We study the problem of robot path planning in an environment with dynamic obstacles and present a method of local robot path planning with the prediction of the trajectory of an obstacle being a band. First, the time to sample information of an obstacle is determined according to its position within the vision of the robot and the trajectory of the obstacle is predicted based on information above. Different from traditional methods, the predicted trajectory is not a curve, but a band; and then the risk of collision between the robot and the obstacle is evaluated based on the predicted trajectory; finally, for the case of their collision, a mathematical model of local robot path planning is formulated and particle swarm optimization is employed to solve it. We analyze the method above theoretically and conduct simulations. The results confirm that our method decreases the probability of collision between the robot and the obstacle greatly.

Representative selection for cooperative co-evolutionary genetic algorithms

The performance of cooperative co-evolutionary genetic algorithms is highly affected by the representative selection strategy. But rational method is absent now. Oriented to the shortage, the representative selection strategy is studied based on the parallel implementation of cooperative co-evolutionary genetic algorithms in LAN. Firstly, the active cooperation ideology for representative selection and the dynamical determinate method on cooperation pool size are put forward. The methods for determining cooperation pool size, selecting cooperators and permuting cooperations are presented based on the evolutionary ability of sub-population and distributive performance of the individuals. Thirdly, the implementation steps are given. Lastly, the results of benchmark functions optimization show the validation of the method.The performance of cooperative co-evolutionary genetic algorithms is highly affected by the representative selection strategy. But rational method is absent now. Oriented to the shortage, the representative selection strategy is studied based on the parallel implementation of cooperative co-evolutionary genetic algorithms in LAN. Firstly, the active cooperation ideology for representative selection and the dynamical determinate method on cooperation pool size are put forward. The methods for determining cooperation pool size, selecting cooperators and permuting cooperations are presented based on the evolutionary ability of sub-population and distributive performance of the individuals. Thirdly, the implementation steps are given. Lastly, the results of benchmark functions optimization show the validation of the method.

Research on applications of three layers neural network in dynamic modeling of jig system

The recurrent neural network has strong dynamic approximating capability. It can be used to describe a kind of complex nonlinear time-varying dynamic system. In the paper a model of a three layer recurrent neural network is proposed. The dynamic backpropagation method is put forth to train the weights of the network. Simulation shows the model has good performance in approximating dynamic systems.

Vapour temperature predictive control based on neuron prediction model

In accordance with the control characters of vapour temperature adjusting process, such as large time lag, large hysteresis and time varying parameters, etc., a kind of CAMIAX model of the process is suggested and the treatment of undetermined pure time lag is discussed. Based on the neuron learning mechanism, a kind of neuron identification algorithm of model parameters and the prediction model of the process are proposed. The predictive model combined with neuron controller constitutes vapour temperature loop control algorithm. The simulation results show the effectiveness of the approach.

Process out-of-control cause diagnosis of coal preparation plant based on fuzzy pattern recognition

A new method of fuzzy pattern recognition is proposed considering the importance of eigenvector in recognizing fuzzy patterns. Aiming at the coal preparation process of the coal preparation plant, the eigenvector of process out-of-control cause is attained by using of the integrated environment and other application systems of CIMS. The method to diagnose process out-of-control cause is given based on fuzzy pattern recognition. The data flow diagram of carrying out the system is also given.