Fuzzy Neuro-based Co-relation of Process and Performance Relationships for Semiconductor Encapsulation Machine Design
This project is to explore and capture the dynamic characteristics and optimal solutions of the auto-moulding production systems for a local semiconductor equipment manufacturer. SIMTech researchers have carried out a study to extract, learn and establish the co-relation between these characteristics and parameters, and to map out the complete process knowledge of the auto-moulding product system performance.
Contact PersonLi Xiang(xli@SIMTech.a-star.edu.sg)
SolutionA fuzzy neural network was developed as an optimisation controller which assists the machine designers in sensitivity analysis, performance prediction and optimal process setting, as well as for decision support.
Auto-moulding production system is used to encapsulate lead frame with thermoset mould compounds. The encapsulation machine is designed and constructed based on information with some degree of uncertainty as a result of inherent and spatial variability, limited historical data or imperfect modelling. The performance of such a system is uncertain and the final design involves a high level of risk. Firstly, it is unable to foresee long-term dynamic effects due to the interactions of the transfer mechanisms. Secondly, it does not favour experimentation, as a change in any one parameter requires the timing diagrams to be re-drafted and analysed. Since the auto-moulding machine has to cater for a range of curing times which is dependent on the plastic material used, the optimal process settings of the transfer elements tend to be varied accordingly. These optimal process settings are difficult as well as time-consuming to determine.