Abstract: In the chip wire bonding process, due to the coupling and nonlinear relationship of the multi-parameters including bonding temperature, power, pressure, speed and time, it’s very difficult to describe the relationship of these multi-parameters in a mathematics way, which affects the improvement of wire bonding quality. The result of how these parameters affecting the bonding quality in terms of shear force and squashed ball diameter is obtained through the orthogonal experiment, and six key parameters are fixed for the bonding process modeling. The prediction model of wire bonding process is proposed on the basis of adaptive neuro-fuzzy inference system(ANFIS). The model is trained through experimental data on the bonding machine. Through the comparison of the predicted data and the real measured data, it shows that the mean error of the shear force is 3.16%, and the mean error of the squashed ball diameter is 1.24%. Based on this prediction model, the influence of key process parameters on the bonding quality is determined, which can be further used in the parameter optimization of the wire bonding process.

Key words: Design of experiment, Neuro-fuzzy inference system, Prediction model, Wire bonding