The Research Of The Quality Control Model For Wild Aluminum Plate And Strip Based On Artifical Neural Network

Aluminum plate and strip is wildly used because of its light weight, high strength, corrosion resistance well, processability easily, surface appearance, etc. The demand of high-precision wild aluminum plates is growing lareger in national defense, aerospace and transtation areas. For now, the overall technology of mill control is still relatively backward in our country, the quality of the products has not reach the needs of the market, so there is still need to import large quantities of high precision wild aluminum plate and strip, especially 5××× series wild plate which is restricted the development of the national economy and national defense construction. Therefore, scientific and technological innovations should be emphasized to develop wild aluminum plate and strip, it is very significant to produce the quality control technology of super-size, high-performance wide aluminum plate and strip that own independent intellctual property rights.The acquistion of various high-precision control model is the primary task of the study in the quality control technology area, while the prediction mode is the basis of construction of the control model. The process of wild aluminum plate and strip has multi- variable, strong coupling, nonlinear and time-varying characteristics, It is difficult to achieve the desired results by using traditional methods which was simplified. Artifical neural networks and other artifical intelligence technology are good at ling with complex multidimensional nonlinear problems, but also have good adaptability and self- learning ability, so it has been widely used in the data-rich area. A lot of data was accumulated in the production process of wild aluminum plate and strip, so it is necessary to introduce artificial neural networks, expert systems, and other artificial intelligence technology to carry out research.The thickness precision and shape precision of the wild aluminum plate and strip are the main indicators of the quality control. The impact factors of the two indicators are analysed in this paper, high-precision prediction models were developed by using artificial neural network, which was laid foundation of the construction for the high precision control model.Aiming at the “1+4” hot tandem rolling line which was domestic leading and most wide, According to the processed measured data of as-rolled 5083 wild aluminum medium plates which was wildly application, a thickness prediction model was developed by artificial neural network based on the analysis of key factors. The best structure of the ANN is 4-10-1. The relative error of the model is within 0.5%, which has high precision and better generalization ability. The developed model was successfully to predict the thickness of 5052 wild aluminum medium plate and exhibited good generation ability.The flow stress is the main impact facter for the shape of the wild aluminum plate and strip. In this paper, isothermal, constant strain rate compression of as-cast 5083 aluminum alloys was carried out on a Gleeble-1500 thermo- mechanical simulator, in a wild range of temperatures 300-450℃, strain rate 0.001-1s-1 up to a ture stain of 0.7. The ture stress-strain data from isothermal hot compression tests were employed to develop the strain-compensated Arrhenius-type constitutive model and ANN constitutive model. The best structure of the ANN is 3-9-1. The result indicate that the ANN constitutive model can accurately predict the high-temperature deformation behavior over a wilder range of temperatures and strain rates, which was confirmed the advantages of ANN approach.In order to apply the prediction medels more convenience and accelereat the speed of the development, The expert system of perfoemance prediction of wild aluminum plate and strip was established by using the method of C#.NET and Matlab mixed programming in this paper. The expert systems contain two application modules currently, the high temperature flow stress prediction and the thickness prediction of wild aluminum plate and strip. The couple was realized through Matlab engine, which is between the human-computer interfaceo of the expert system. The human-computer interface was developed by using C# language combine with ASP.NET technology. The expert system was used to make a qualitative comparison of the influence of the fators on thickness vary of wild aluminum plate and strip. It is provided more meaningful information for the establishment of high-precision quality control model of wide aluminum plate and strip.