Dissertation
Dissertation > Industrial Technology > Automation technology,computer technology > Automated basic theory > Artificial intelligence theory > Artificial Neural Networks and Computing

Process Early Warning Technology Based on Neural Network and Intelligent Optimization Algorithms and Its Application

Author LiuTao
Tutor LiuManDan
School East China University of Science and Technology
Course Control Science and Engineering
Keywords Neural Network Culture-MPSO Algorithm Chaos-MPSO Algorithm Process Early Warning Technology
CLC TP183
Type Master's thesis
Year 2011
Downloads 109
Quotes 0
Download Dissertation

Early warning technology is that forecast the abnormal situation and send a warning signal before it occurs. Using early warning technology can improve the response speed of abnormal situation and find it before it occurs, and can reduce the future possible loss effectively. Early warning technology plays an important role in safety production.Two hybrid algorithms (Culture-MPSO and Chaos-MPSO) were proposed to optimal the parameters of RBF-NN. Culture-MPSO algorithm fusions cultural algorithm and particle swarm optimization algorithm. On the one hand, Culture-MPSO uses PSO to optimal the population space of CA, and continuously improves the parameters of PSO, such as inertia weight. On the other hand, the accept rules of Culture-MPSO are more adaptive because of increasing a parameter to measure evolution direction. Chaos-MPSO algorithm fusions chaos theory and particle swarm optimization algorithm. On the one hand, Chaos-MPSO continuously improves the parameters of PSO, such as inertia weight. On the other hand, Chaos-MPSO uses chaos theory to find the better extreme value nearby the current global extreme value. This algorithm is more efficient and has good effect to optimal the early warning model.Two models were proposed, including warning level model and soft sensor model. Warning level model has three warming levels, which is a discrete output model. Soft sensor model takes the key parameter as the output of neural network, which is a continuous output model. Many experiments have done to simulation the two kinds of models, then the two models are combined to decisive the warning level.Under the same conditions, simulation results shows that:compared with BP network and Elman networks, RBF network shows better fitting and predictive result; Culture-MPSO algorithm and Chaos-MPSO algorithm have better search performance and faster search speed than the traditional optimization algorithms, and Chaos-MPSO algorithm is better than Culture-MPSO algorithm; the two models were combined to decisive the warning level, which has improved the correct rate of the early warning model.Using the above intelligent optimization algorithms and neural network modeling method, and setting up a comprehensive early alarming model, which can predict the abnormal situation of process timely and prevent it occurs.

Related Dissertations
More Dissertations