ITRI's Prognostic and Health Management Software in Semiconductors (PHM), a 2017 R&D 100 Award winner, uses multi-level ensemble learning techniques which integrate a variety of advanced machine learning algorithms with domain knowledge. Along with model adaptation techniques to cope with individual equipment characteristics and drift in environmental conditions, PHM provides high-accuracy fault diagnosis and prognostic technologies. It supplies the means for engineers to early detect problems, effectively monitor the health of the production equipment, determine causes of downtime, and shorten maintenance time.