A Study On Quality Control Of Molding Sand

Sand casting is most popular casting process and moulding sand is the most important casting material because of its low cost and high efficiency. Above 80% castings are produced by sand casting. Green-sand, using clay as bonding material, has a long history and is most popular and convenient. 70~80% molding sand is clay green-sand. The quality of moulding sand directly affects the appearance quality and defects like sand hole, sand inclusion or gas porosity. More than 50% casting defects are caused by sand quality or improper use.Strict control of moulding sand properties is the best way to deduce casting defects. Meanwhile, quality prediction of sand will be benefit to the evaluation of sand processing and then improve the quality.Properties of component materials determine the final quality of the green-sand. Key component materials (silica sand, coal dust and bontonite) are studied in this dissertation. In foundry workshop of Shanghai Huizhong Automobile Company performance parameters of green-sand material are measured. According to results of comparison with the domestic and foreign foundry workshop, a new standard for moulding sand is proposed. Key factors in sand process like moisture, temperature, clay content and mixing process are mainly analyzed and discussed. Based on the present equipment and products, the proposed sand process should improve the casting defects and deduce the rejection rates.There are so many indicators with the characteristics of fuzziness and gray character which determine the quality of sand molding that the forecast of sand mold quality is extremely complicated and difficult. Therefore, this paper presents Support Vector Machine (SVM)-based molding methods of sand mold's quality prediction. The SVM quality prediction methods, based on SRM (Structural Risk Minimization), not only setup the relation between parameters of sand properties and sand quality, but predict the quantitative defects. Also a comparison between the gray relation method and SVM method is conducted. Instance analysis results showed that this method was more objective, scientific and rational.