Prediction Model for Wire Bonding Process through Adaptive Neuro-Fuzzy Inference System

In the wire bonding process, different combinations of parameter values will directly affect wire bonding quality. The optimal combination of these parameter values is very important to ensure the overall process quality response. Therefore, it is necessary to investigate the effects and interactive relationship of the bonding parameters on the bonding quality. This paper chooses the response factors of shear strength and Squashed Ball Diameter (SBD) for the bonding quality evaluation. Through the design of experiments (DOE) method, 60 sets of experimental samples of the ball bond, varying 9 process parameters are manufactured and tested. The effect of the variation of these parameters on the shear force and SBD are analyzed and 6 out of 9 parameters are determined to be controlling factors in the prediction model. Considering the difficulty in analyzing the nonlinear and interactive relationships between the parameters and bonding quality, this paper proposes a process modeling approach based on an adaptive Neuro-Fuzzy Inference System. For the construction of the prediction model 50 sets of samples, where the 6 process control parameters are varied are prepared. These are used for the training of the process prediction model and a further 15 sets of samples are used for model validation. An error analysis is then performed to evaluate the model created. Based on the process prediction model, the characteristics of the bonding parameters affecting bonding quality are obtained, that can then be used for the optimization of wire bonding process.