Dissertation > Excellent graduate degree dissertation topics show

The Study of Prediction and Optimization of Ejector Performance Based on Artificial Neural Networks

Author: HuangLiangLiang
Tutor: CaoJiaZuo
School: Donghua University
Course: Heating,Gas Supply, Ventilation and Air Conditioning Engineering
Keywords: Ejector Neural Networks Wavelet basis function Ant Colony Algorithm Optimization
CLC: TU831.4
Type: Master's thesis
Year: 2010
Downloads: 114
Quote: 1
Read: Download Dissertation

Abstract

The energy crisis has become one of the most concern in the world today. With economic development, in particular the increasing popularity of building energy consumption of air conditioning. In our country, the refrigeration and air-conditioning energy consumption has accounted for about one fifth of the total energy consumption. Every summer, around the grid are faced with the test of the peak. If you can reduce the energy consumption of refrigeration and air conditioning, then this will greatly mitigate the energy crisis in our country, which will play an important role in promoting the work carried out on our energy reduction. The steam-jet refrigeration system wish to solve the above problems. Steam jet refrigeration system has no moving parts, and basically do not need electricity. If take refrigerant vapor jet refrigeration system using solar energy, waste heat, and geothermal energy system will that is energy-efficient and environmentally friendly. However, traditional jet refrigeration system COP is very low, the development of the steam jet refrigeration system has been a lot of restrictions. The injector is the core component of the steam jet refrigeration system. If you can improve the performance of the ejector, which will greatly improve the steam jet refrigeration system COP. Therefore, the injector to be the focus of research is particularly important. So far, the theoretical analysis of various injector performance, particularly in the design and operating practices, and ultimately still need experiment for inspection and correction. The injection experiments quite time-consuming and laborious, and the price is not small. Injector on the basis of experimental data, the prediction model can be used for similar or the same series of ejector performance prediction injector research and development, will be of great help in the design and operation of the same time saving time and costs expand the scope of the applicable parameters, has great practical significance. Ejector performance with its own structure parameters and operating conditions of the relationship is highly nonlinear, neural network prediction model is suitable for simulation of nonlinear systems. Neural network forecasting performance of the ejector is a new method to study the performance of the ejector. The transfer function of the neural network and learning methods have a major impact on the speed and the prediction accuracy of the neural network learning. In this article, refers to three types of wavelet neural network, of three wavelet neural network hidden layer transfer function were the Morlet, Mexi-hat and Gauss1 wavelet basis function. Numerical experiments show that these three wavelet neural network learning speed and predictive ability than traditional neural networks. Hidden layer neuron number have a significant impact on the performance of the neural network, through trial and error, to find the optimal hidden layer neuron number of for ejector performance prediction neural network. Neural network to learn the nature of the problem is an optimization problem. Traditional neural network learning methods exist learning is slow and easy to fall into local minimum value of the shortcomings. The ant colony optimization method is a new kind of global optimization algorithms. In this article, the first attempt the two continuous ant colony optimization algorithm CACS and ACO_R of as a neural network learning methods to improve the learning speed and predictive ability of the neural network. Regularization method can effectively improve the generalization ability of the neural network. The numerical results show that the regularization method and ant colony algorithm combined neural network can improve the prediction accuracy. In a large number of experimental data on the basis of the ejector, through the training of the neural network to obtain a neural network that can predict the performance of the ejector. The numerical results show that the predictive ability of the neural network injector performance is better than the theoretical calculation. Neural network optimized structure parameters and operating parameters of the ejector, the optimal combination of parameters, jet ejector coefficient. Particularly noteworthy is that the neural network can easily solve the two-parameter and multi-parameter optimization problem, traditional experimental methods injector multi-parameter optimization is not practical. Unlike traditional single parameter optimization, optimal parameters of the injector structure of the multi-parameter optimization results more accurate results. This neural network and the experimental data combining method simplifies the analysis of the process of the performance of the injector, the optimum design and operation of the injector provides an economic and simple, rapid comprehensive, sufficiently accurate new method.

Related Dissertations

CLC: > Industrial Technology > Building Science > Housing construction equipment > Air-conditioning, heating, ventilation,and its equipment > Air-conditioning > Air-conditioning machinery and equipment
© 2012 www.DissertationTopic.Net Mobile