Search

Additional file 1: Fig. S1. a FACS analysis of mouse plasma using RIP140 antibody. b FACS analysis of mouse CSF exosomes using RIP140 antibody.

In the semiconductor wafer fabrication facility (fab), the development of models to determine product mix for a multi-stage, multi-site, and multi-generation company is very important. In this paper, we present to fill the gap by using neural networks (NNs) to model the impact of product mix on cycle time factors, such as “arrilval time”, “process time”, “usable tool”, “Q-time constrain”, “CV of process time”, “number of recipes”, “sampling rate”, “hot lot ratio” and etc.. Through the interaction of various KPIs of genetic algorithms (GA), we seek a product mix that satisfies the optimization of the required cycle time. The results showed that seen the fab move could be improved 3.91% by the GA approach, and proposed approaches can help practitioners in a fab to determine the optimal product mix efficiently.

This study has demonstrated practical viability of the proposed approach employing datamining technics, Neural Networks (NNs), to estimate the productivity of individual process tool sets in a semiconductor factory, and to assess the efficiency loss by 15 related individual input factors, which included “process time”, “number of recipes”, “usable tool”, “Q-time constrain”, “standard deviation of lot size”, “batch size”, “sampling rate”, “hot lot ratio” and etc.. An empirical study was conducted by using the equipment data of a real fab. The results showed that the proposed approaches can define performance efficiency of Overall equipment efficiency (OEE) more reasonable, which discover underlying factors for efficiency loss, and help to improve performace efficiency from 91.23% to 94.03%.