At present the packaging industry
currently relies heavily on test data to determine the bonding
parameters settings needed to meet the design requirements.
This trial-and-error process is time consuming and cost
ineffective [6]. Therefore, it is important to fully understand the
effects of the bonding process parameters and their interactions
to obtain the optimum parameter combination quickly.
Considering the difficulty in determining and representing the
nonlinear and interactive relationships between the bonding
parameters and bonding quality, this paper proposes a process
modeling approach using an Adaptive Neuro-Fuzzy Inference
System (ANFIS). For the wire bonding modeling process, sets
of experimental data are measured and used for the training the
ANFIS prediction model. The error of the prediction model is
then analyzed with separate data and the effects and
characteristics of varying the process control parameters on the
bonding quality can be obtained and used to optimize the
combination of bonding process parameter settings.