外文翻译---基于小波神经网络的Al（OH）3流化床焙烧温度的预测

《外文翻译---基于小波神经网络的Al（OH）3流化床焙烧温度的预测》由会员分享，可在线阅读，更多相关《外文翻译---基于小波神经网络的Al（OH）3流化床焙烧温度的预测（19页珍藏版）》请在毕设资料网上搜索。

1、 1 附录 A 外文翻译 原文部分 Prediction of Al(OH)3 fluidized roasting temperature based on wavelet neural network LI Jie(李劼 )1, LIU Dai-fei(刘代飞 )1, DAI Xue-ru(戴学儒 )2, ZOU Zhong(邹忠 )1, DING Feng-qi(丁凤其 )1 1. School of Metallurgical Science and Engineering, Central South University, Changsha 410083, China； 2. Ch

2、angsha Engineering and Research Institute of Nonferrous Metallurgy, Changsha 410011, China Received 24 October 2006； accepted 18 December 2006 2 Abstract（ cnki） The recycle fluidization roasting in alumina production was studied and a temperature forecast model was establishedbased on wavelet neural

3、 network that had a momentum item and an adjustable learning rate. By analyzing the roasting process, coalgas flux, aluminium hydroxide feeding and oxygen content were ascertained as the main parameters for the forecast model. The orderand delay time of each parameter in the model were deduced by F

4、test method. With 400 groups of sample data (sampled with theperiod of 1.5 min) for its training, a wavelet neural network model was acquired that had a structure of 7211, i.e., seven nodes inthe input layer, twenty-one nodes in the hidden layer and one node in the output layer. Testing on the predi

5、ction accuracy of themodel shows that as the absolute error 5.0 is adopted, the single-step prediction accuracy can achieve 90% and within 6 steps themulti-step forecast result of model for temperature is receivable. Key words: wavelet neural networks； aluminum hydroxide； fluidized roasting； roastin

6、g temperature； modeling； prediction 3 1 Introduction In alumina production, roasting is the last process,in which the attached water is dried, crystal water isremoved, and -Al2O3 is partly transformed into -Al2O3.The energy consumption in the roasting process occupiesabout 10% of the whole energy us

7、ed up in the aluminaproduction1 and the productivity of the roastingprocess directly influences the yield of alumina. As theroasting temperature is the primary factor affecting yield,quality and energy consumption, its control is veryimportant to alumina production. If some suitableforecast model is

8、 obtained, temperature can be forecastedprecisely and then measures for operation optimizationcan be adopted. At present, the following three kinds of fluidizedroasting technology are widely used in the industry:American flash calcinations, German recyclecalcinations and Danish gas suspension calcin

9、ations. Forall these roasting technologies, most existing roastingtemperature models are static models, such as simplematerial and energy computation models based onreaction mechanism2； relational equations betweenprocess parameters and the yield and the energyconsumption based on regression analysi

10、s3； staticmodels based on mass and energy balance and used forcalculation and analysis of the process variables and thestructure of every unit in the whole flow and system4.However, all the static models have shortages inapplication because they cannot fully describe thecharacteristics of the multi-

11、variable, non-linear andcomplex coupling system caused by the solid-gasroasting reactions. In the system, the flow field, the heatfield, and the density field are interdependent andinter-restricted. Therefore, a temperature forecast modelmust have very strong dynamic construction, self-studyfunction

12、 and adaptive ability. In this study, a roasting temperature forecast modelwas established based on artificial neural networks andwavelet analysis. With characteristics of strong faulttolerance, self-study ability, and non-linear mappingability, neural network models have advantages to solvecomplex

13、problems concerning inference, recognition, classification and so on. But the forecast accuracy of aneural network relies on the validity of model parametersand the reasonable choice of network architecture. Atpresent, artificial neural networks are widely applied inmetallurgy field56. Wavelet analy

14、sis, a timefrequencyanalysis method for signal, is named asmathematical microscope. It has multi-resolutionanalysis ability, especially has the ability to analyze localcharacteristics of a signal in both time and frequencyterritories. As a time and frequency localization analysismethod, wavelet anal

15、ysis can fix the size of analysiswindow, but allow the change of the shape of theanalysis window. By integrating small wavelet analysispacket, the neural network structure becomeshierarchical and multiresolutional. And with the timefrequency localization of wavelet analysis, the networkmodel forecas

16、t accuracy can be improved710. 2 Wavelet neural network algorithms In 1980s, GROSSMANN and MORLET1113proposed the definition of wavelet of any functionf(x) L2(R) in aix+bi affine group as Eqn.(1). In Eqn.(1)and Eqn.(2), the function (x), which has the volatilitycharacteristic14, is named as Mother-wavelet. Theparameters a and b mean the scaling coefficient and theshift coefficient respectively. Wavelet function can beobtained from the affine transformation ofMother-wavelet by scaling a and translating b.