Abstract

For enterprise quality improvement in manufacturing, data for decision making can exist in a variety of sources. A data warehouse facilitates the delivery and analysis of such widespread data. Many manufacturers have a variety of quality improvement, resource planning, statistical process control, and other data systems, all having a great amount of detailed data and all disconnected from each other. Enterprise resource planning (ERP) systems are not the solution to this problem of disparate data sources, for ERP is more amenable to transactional needs than to data analysis needs. Data warehousing is the solution, by providing for consistency in data analysis throughout an organization. A data warehousing system does this by linking to the operational environment as well as to data marts and info marts. To implement a data warehouse, an organization might already have accomplished the initial steps of setting up automated data collection, data transfer, and data conversion systems. Next, a data model is needed; it should have components for transactional data and derived data. Modeling of the physical implementation of the data warehouse should be based on facts and dimensions. Facts are analysis variables such as attribute data, process status, process statistics, control limits, specification limits, and target values. Dimensions are categorical variables such as product and process identifiers, time intervals, and defect categories. The nature of warehouse design is set by constructs such as summary tables, multidimensional tables, star schema, and snowflake schema.