型砂质量在线检测与控制模型的研究

型砂质量的在线控制基于两大基本前提:1.实现型砂性能的自动检测;2.寻找并建立型砂成分与性能之间的相互关系.为此,常规试验中既能反映型砂成分变化又适于在线快速检测的紧实率、水分、砂温、透气性与壁裂强度五性能被选作为在线检测项目.文中详细讨论了型砂性能在线检测仪的研制过程、诸项目的检测原理以及仪器所能达到的检测精度.由于采用了80C552微处理器的控制,本仪器可在1min内完成全部五项目的检测与结果的打印.以此为基础,本文进一步探讨了利用人工神经网络(ANN)原理建立型砂制备这一非线性系统中成分与性能相互关系的原理与具体实现过程.研究表明,采用由非线性变换单元组成的前馈网络(B一P网络)方法建立的型砂成分与性能关系的网络模型(CPN),可用于型砂的预测,且预测精度高于由传统的回归方法建立的模型,从而为实现型砂质量的在线控制打下了基础.

On-line Testing and Control Model of Sand Qualities

The on-line control of molding sand qualities is based on two essential prerequisites. First, the automatic testing of molding sand properties should be carried out. Second, the relationship between compostitions and properties of molding sand should be established. Thus, compactability, percentage of moisture, sand temperature, split strength and permeability have been chosen as the testing items in this paper, as they not only reflect the change of molding sand compositions, but also are suitable for on-line testing. The automatic testing method has been discussed in detail. In order to predict the variation of molding sand compositions from the measured results rapidly and to adjust the preparing procedure promptly, the method of establishing model has been studied, which deals properly with the technology of artificial neural network. The testing results indicate that the predicting precision of molding sand compositions with artificial neural network molel is enough for production practice, and the molel may be used for on-line automatic control of molding sand preparing system.