A hard-arcing precursor that causes charging damage to semiconductor devices was found. The micro-arcing that occurs during the plasma etch process was statistically classified based on the optical intensity variation by using high-speed optical sensing. According to the level of effect on semiconductor device failure, the micro-arcing was categorized into soft-arcing I (8–11σ), soft-arcing II (11–15σ), and hard-arcing (over 15σ). Experimental results suggest that a sufficiently large increase in soft-arcing is directly associated with the occurrence of hard-arcing. By using this characteristic and based on a Gaussian distribution, hard-arcing was statistically predicted within 12 wafers. A mathematical model is also proposed to explain the same. This is the first control system capable of predicting and effectively restraining micro-arcing during the plasma process.