Dissertation > Excellent graduate degree dissertation topics show

Integrated Intelligence Based Bi-Directional Prediction Approach for the Performances of Differential Fibers

Author: WangYi
Tutor: DingYongSheng
School: Donghua University
Course: Detection Technology and Automation
Keywords: bi-directional prediction neural networks multi-objective evolutionary algorithm performance prediction differential fibers
CLC: TQ340.64
Type: Master's thesis
Year: 2014
Downloads: 7
Quote: 0
Read: Download Dissertation

Abstract


This paper is based on the production industry of differential fibers and the optimization of this production process. By the studying of different production process of different fibers, the selecting and the collecting of the key production parameters and fiber performances, the relationship between them is aim to be found and simulated by intelligent algorithms. This bi-directional prediction approach, which is based on the production parameters and fiber performances of differential fibers, can optimize the production parameters, simplify the model building, reduce the production cost and improve the fiber performances. In addition, this method proposes a new idea in connection with the differential fiber production optimization.The bi-directional prediction approach of differential fibers is composed of two parts, one is the backward reasoning, which is using the fiber performances to predict the production parameters, another is the forward prediction, which is using the production parameters to predict the fiber performances.Firstly, because the final fiber performances is based on the production parameters, and the determination of production parameters is follow the required fiber performances, so the prediction of production parameters is very important to the fiber production optimization. Traditional way is following the experiences to adjust the production parameters until the whole production process can achieve the satisfactory fiber performances. Because all the production parameters experience interactions with each other and affect the final fiber performances, so there must be one more sets of production parameters to one fiber performance, which means it is a multi-answers problem, so it is very difficult to determine the production parameters. In order to solve this multi-answer problem, a multi-objective evolutionary algorithm which is based on particle swarm optimization is adopt to do the backward reasoning by the real experience data and its neural networks. All the particles will be lead by both its local optimum to fit all the objections and the global optimum to fit the worst objection. This algorithm can maintain the diversity, accuracy, efficiency and of the final solutions. Finally, use the real production data from orthogonal productions of polypropylene precursor to test the backward reasoning, and it shows that the backward reasoning has a good prediction results.Then, because the real production data of a fiber can’t be the results of orthogonal productions, so it must be a irregular big data set. In order to use the backward reason in to the fiber production optimization, the neural networks of the forward prediction and the clustering process of the input centres and output centres are improved in this paper. Firstly, in the clustering process of experience data, the K-means algorithms, particles swarm optimization and genetic operators are used to optimize the clustering results and the further prediction process. Then, the radial basis function are improved to optimize the neural network and the do the forward prediction process with the input and output centers. These works can increase the accuracy of the forward prediction results and lay the foundation of the backward reasoning. In order to test this process, IRIS data are used to test the clustering process, the polyester production data are used the forward prediction and the whole bi-directional prediction, and the results shows that the clustering process, the forward prediction and the whole bi-directional prediction all achieve good results.Finally, use the C#language and the visual studio based on.net to develop the bi-directional prediction approach of differential fibers as a software platform and to complete the data training, the production prediction and the performances prediction simply, and to be used in to real fiber production.

Related Dissertations

  1. The Brake Performance of Hydraulic Retarder and Simualtion Research on Its Application for Aircraft Arrestment,TH137.331
  2. High Speed Frequency Measurment and Non-Linearity Correction of Frequency Modulated Capacitive Displacement Sensor,TH822
  3. Research on Anti-periodic Solutions of Delayed Cellular Neural Networks without Assuming Global Lipschitz Conditions,TP183
  4. Spectrum Variation of Vegetation in Yanzhou Coal Mine Area and Heavy Metal Stress Characteristic,X173
  5. Stage Fault Diagnosis Based on Data Fusion,TP18
  6. Study on the Function Approximation Based on Neural Network,O174.41
  7. The Forecast Methods of Mining Risk of the Coal Resources Affected by Dynamic Disaster,TD713
  8. Rules Extraction from Artificial Neural Networks for Classification Based Improvedant Colony Algorithm,TP183
  9. Stock Prediction Based on the Principal Components-BP Neural Network,F830.91
  10. The Study of Coal Calorific Capacity Based on Texture Feature,TP391.41
  11. Multi- license plate location method based on CNN's Intelligent Transportation Systems Research,TP391.41
  12. Research of MSC Dynamic Reactive Power Compensation System in Mine Electric Power Network,TM761.1
  13. Research on the Models of GPS Height Fitting Based on BP Neural Network and Their Applications,P228.4
  14. Sensitivity Study and Application of Complex Variabal Weight Function Neural Network,TP183
  15. On-line Monitoring of Tool Wear for Mass Production,TG71
  16. Modeling of Photovoltaic Cells and Its Application in Energy Prediction,TM914.4
  17. Research on Evaluating the Feature Fatigue’s Effect on Customer Equip,F224
  18. Delay Neutral System Stability Analysis of Neural Networks,O175.13
  19. The Design of Identification System for Power Marketing Services,TP391.41
  20. The Study of Intelligent Intrusion Detection System Based on Neural Network in Linux,TP393.08
  21. Research on the Operating Performance Evaluation of the Central Enterprise Executives,F276.1

CLC: > Industrial Technology > Chemical Industry > The chemical fiber industry > General issues > Production process > Spinning
© 2012 www.DissertationTopic.Net  Mobile