Optimization of pick-and-place in die attach process using a genetic algorithm

The entire semiconductor manufacturing process can be largely divided into two sequential sub-processes that are commonly referred to as front-end and back-end production. Both production processes involve many complicated and elaborate steps for wafer fabrication, probe testing and sorting, assembly, final test, etc. The front-end process refers to wafer fabrication process, whereas the back-end process refers to the assembly, packaging, and testing of individual semiconductor devices. Once the front-end process is complete, the wafers are transferred to the back-end process in order to facilitate its integration into electronic devices and to take final performance test. In back-end production process, based on quality and location information of individual chip (die) derived from the EDS process, only good semiconductor chips (dies) are individually picked up by an automatic robot arm and then attached to the support structure (e.g., the lead frame) on a strip. This process is called die attach process. In order to improve the production efficiency of die attach process, it is necessary to evaluate various robot arm operation methods (rules) and then to provide the most efficient operation method since the total transfer distance (or total transfer time) considerably depends on the robot arm operation methods as well as the distribution of good chips on wafers. Thus, in this research, a pick-and-place problem of the die attach process is mathematically formulated and several transfer methods are compared