Today, sensors are everywhere. They do everything from counting parts on an assembly line to measuring the quality of products. But some of the biggest challenges occur after measurements have been made. At that point, you have to decide: Where do I collect the data, and how can I use it to improve my operations by decreasing variability and improving quality?
In the manufacturing arena, real-time operations require fast data collection for optimal analysis. Generally, manufacturing companies approach data collection in one of two ways: with a traditional relational database or with a plant data historian.
Each offers distinct advantages. A relational database is built to manage relationships, but a plant data historian is optimized for time-series data. For example, relational databases are great at answering a question such as: "Which customer ordered the largest shipment?" A plant data historian, however, excels at answering questions such as: "What was today's hourly unit production standard deviation?"
This type of database is an ideal option for storing contextual or genealogical information about your manufacturing processes. The relational nature of the database provides a flexible architecture and the ability to integrate well with other business systems. When extending the functionality of a relational database for manufacturing applications, companies leverage its openness by creating and managing custom tables to store data that comes from multiple sources, such as other databases, manually entered values via forms, and XML files.
As relational databases mature, I see vendors improving their system's performance in transactional manufacturing applications, such as capturing data from an RFID reader. When capturing contextual information or time-series data from a small number of sensors, a relational database may work best.
Plant Data Historians
On the other hand, plant data historians are a perfect choice when you must capture data from sensors and other real-time systems because this type of repository uses manufacturing standards, such as OPC, that facilitate communications. With plant data historians, you can streamline implementation by using standard interfaces.
With most of these systems, there is little or no management or creation of data schema, triggers, stored procedures, or views. You can usually install and configure a plant data historian quickly without specialized services, such as custom coding or scripting for the installation.
Plant data historians are also designed to survive the harshness of the production floor and feature the ability to continue capturing and storing data even if the main data store is unavailable. Another feature typically found in a plant data historian is the ability to compress data, reducing the amount of drive space required. When capturing time-series data rapidly (with a re-read rate of less than 5 s) for several thousand sensors, a plant data historian may work best.
The Best of Both Worlds
When relational databases and plant data historians are deployed in concert, companies can collect and analyze the tremendous volumes of information generated in their plants, improve performance, integrate the plant floor with business systems, and reduce the cost of meeting industry regulations. As stated by many Six Sigma quality experts, "You can't improve what you don't measure." Plantwide data collection can make this possible.
By using analysis tools, such as Microsoft Excel or other off-the-shelf reporting solutions, you can increase the quality and consistency of your products by comparing past production runs, analyzing the data prior to a downtime event, and plotting ideal production runs against in-process runs. Today's analysis tools make it easy to aggregate data, prepare reports, and share information using standard Web browsers.
Plantwide in-process data collection also serves as the vital link between plant processes and business operations, providing business systems with the data they need to gain a clear, accurate picture of current production status or historical trends.
Ultimately the decision should be to use a relational database and a plant data historian. The combined power of both provides the detailed information that yields numerous benefits, internally for the company and externally for customers.