(The Key Laboratory of Contemporary Design and Integrated Manufacturing Technology, Ministry of Education,Northwestern Polytechnical University, Xi’an 710072, China)
In the design process of die cavity of investment casting, the inaccurate enlarged die cavity that based on shrinkage rate can lead the die need mold-repair for many times. As an initial study, a shrinkage rate prediction method of typical structure casting in the solidification process is proposed in this article. The method can provide a way of thinking for shrinkage rate prediction of casting. As BP neural network has strong fault tolerance and robustness function. Thus, the mapping model between geometric parameters that attach to the structure and shrinkage rate is built based on BP neural network. As there is no determination criterion for the number of the hidden layer neurons of the BP neural network in different cases, thus, the influence of the number of neurons in the hidden layer on the accuracy of modeling is researched. The result is that for the typical structure casting, when the number of neurons in the hidden layer is three, the mapping model has the least prediction error. In this case, the shrinkage rate average deviation of the predicted and measured values is 0.09%. The mapping model can better realize shrinkage rate prediction of the casting in solidification process.