Abstract
With the rapid development of technologies, electronic devices are asked to be thinner, lighter, and more powerful to meet market demands. Apart from developing the IC design technologies, advanced packaging technologies are also crucial elements for accomplishing the goal described previously. The reliability of the packaging structure is the most effective factor that ensures the functionality of packaged ICs. Thermal cyclic test (TCT) is one of the important experimental approaches to obtain the reliability of the electronic packages. The drawback of the experimental approach is taking a great amount of time and cost to obtain the result. Thus, the finite element analysis (FEA) is introduced to the industry to reduce the number of experiments, that is, we can save the time and cost from experiments. Due to the great enhancement of the computer infrastructures that provide high-performance computation and the great amount of storage, machine learning techniques become realized, applying in many different research topics including the assessment of the electronic package reliability.The purpose of this study is to apply support vector regression (SVR) techniques to predict the reliability of wafer level chip scale package (WLCSP), then to provide an efficient and effective way for the front-end designers to check the feasibility of their design. After we construct the SVR model, we can further save the time and cost from FEA simulations.In order to accomplish the research goal, this study will be accomplished according to the following three steps: first, the WLCSP reliability obtained by using FEA will be validated by the reference experimental result; second, the validated FEA result will be served as training data and testing data, and adopt SVR techniques to train the predictive model; third, the predictive performance of the predictive model obtained by using SVR techniques will be evaluated.The predictive models obtained by using SVR techniques show good agreement with the FEA results. The predictive performance can be improved by increasing the number of training data. As the number of training data increases, the difference between SVR model and FEA result would decrease, but the time used to train SVR model would increase. In this study, we would discuss the relation between the number of training data, SVR model performance, and training time of SVR model.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.