Application of Rules Extracted from a Multilayer Percentron Network to Moulding Machine Efficiency Improvement

It has been found that artificial neural network (ANN) has been widely applied to manufacturing industry. Most of ANN-based applications focus on yield prediction and fault diagnosis. This paper describes a multilayer perceptron (MLP) network based method for tuning a moulding machine parameter. Once the MLP has been trained, we will draw ´IF- THEN´ rules from it and find out which input parameters may significantly affect the moulding machine process efficiency. In additon, we can learn from rules that how to tune those input parameters. Moulding process engineers can accordingly tune a moulding machine input parameters and increase its efficiency. Practical numerical data obtained from a moulding machine at an IC packaging company in Taiwan justifies the feasibility of this method. It appears that the proposed method can be applied to other manufacturing process.