Manufacturing intelligence delivers new analyses
Business intelligence (BI) is the term given to the set of tools and processes used to make business decisions based on facts—instead of on guesses, intuition, and often irrelevant experience. Few companies can afford to guess any more, so decisions based on sound data and provable relationships are critical for business survival.
BI, coined in the late 1990s by the IT community, has come to describe the tools used for the collection of business data, integration of the data across multiple dimensions, analysis of the data, and presentation of the results. BI tools are the preferred method for extracting information and correlations from massive databases of customer and sales data. BI analysis typically focuses on data about sales, customers, suppliers, shipments, purchase orders, invoices, and deliveries.
The manufacturing equivalent of business intelligence is manufacturing intelligence (MI). Most manufacturing companies have engineers with considerable knowledge and experience in extracting information and correlations from production data using data historians and associated analysis tools. MI expands the concepts of traditional studies by looking across multiple dimensions of production information to discover correlations not seen in normal engineering analysis.
Multi-dimensional data, query
MI focuses on products, inventory, work in process (WIP), production schedules, production performance, yields, waste, and costs. MI usually does not need the high resolution data collected in data historians. The typical MI timeframe may be a production step, duration of a production run, or length of a batch operation. Like BI, MI is based on multi-dimensional data and advanced query and reporting systems.
Multi-dimensional data is maintained in “facts” and “dimension” tables. Facts are measurable pieces of information, such as a production yield or product quality measure. Each fact is associated with multiple dimensions, such as time, batch number, production line, shift, product supplier, batch step, and laboratory ID. An MI database can be huge, with billions of facts and dozens of dimensions per fact.
The analysis capability required for MI is the same as the capability required for BI. But in today’s economic environment, BI tools can be expensive and hard to justify. Fortunately, there are low-priced alternatives. Inexpensive servers built on recycled MS-Windows machines running Ubuntu can provide an inexpensive platform for open-source BI tools. Any Internet search will turn up multiple open-source BI tool sets, but there are a few tools that stand out. The Business Intelligence and Reporting Tools (BIRT) project is an Eclipse-based set of tools that provides Eclipse plug-ins and Java/J2EE code examples for BI analysis. Jaspersoft and Pentaho also offer free open-source BI tools. Any of these tools provide a good starting point for an inexpensive MI system.
The typical MI system would have server side software that collects facts and dimension data from data historians, manufacturing execution systems, and laboratory information management systems (LIMS) databases daily. MI is not designed for real-time decision making, so collecting data in a batch mode can be used to simplify data collection and storage. PERL, PHP, or Python programs would be used to quickly pull data using SQL queries and place it into an MI fact-and-dimension database. Open-source reporting tools would then be used to run analysis and develop correlations for intelligent manufacturing decisions.
Implementing a manufacturing intelligence program to augment Six Sigma and other improvement projects does not have to be a high-cost project. Medium-performance Linux based servers running open-source software are the tools to help turn your manufacturing data into manufacturing knowledge.
|Dennis Brandl is president of BR&L Consulting in Cary, NC, email@example.com .|