High product quality is the main target of the petroleum refining industry. It is widely admitted that there are some limitations of traditional product-quality-monitoring methods. Data-mining (DM) is a method to get useful information, which other regular methods cannot find, from enormous data. Data warehouse (DW) is the best way to store and manage massive enterprise data and provide a strong support to data analysis methods. This paper presents a new framework to deal with quality analysis, which combines the soft sensor, DM and DW. It promises to overcome the limitations of soft sensor and apply soft sensor in quality analysis in the petroleum refining industry.