The importance of having a strategy to manage production data

The various types of data

First, production data is collected over time from industrial procedures and their associated processes. It can come from sensors, laboratory analyses, geolocation systems or any other method used in a decision-making process. Production data can vary in quality and nature.

Today, the Internet of Things (IoT) opens a door to a vast universe of data acquisition solutions. Many platforms are now offered on the market and can come with their own sensors, send data to the cloud and develop easily accessible dashboards and graphs.

Storing data

There are also so-called conventional solutions for storing production data. The data historian is a perfect fit for real-time archiving. It can offer features like compression, archiving by exception and buffering in case of communication issues during acquisition. All of these options ensure optimal use of the storage infrastructure while preserving the integrity over time of the collected information. The historian manages the storage space more efficiently compared with a relational database.

For laboratory and transactional data, a relational data model is generally the best solution. In this model, the data is organized as linked tables. The links provide integrity and consistency and make it easier to handle data. On the other hand, temporal data represents a state in time, without any relation, so it is therefore unnecessary to use this mechanism to ensure consistency.

The data warehouse: a valuable ally

Depending on the type of information collected and the available acquisition techniques, businesses use one of these solutions to manage their data, or sometimes both. To benefit from the data, it is a good idea to group it together, document it and ensure its validity. The tool for this is called the operational data warehouse, which has four main modules:

• The ETL module (extraction, transformation, load) extracts the information from the systems, transforms it (clean, validate, standardize) and loads it into the database. For example, the ETL module can be used to make scale changes or to interpolate missing data.
• The database stores data in a so-called dimensional model to optimize performance in terms of reading and consultation. This module makes it possible to relate facts (tonnage produced, quality obtained) with dimensions (work shift, sector, product type).
• The data marts make it possible to consult information and cross-check it by subject. For example, you can have a data mart that groups all the information dealing with an area of a factory or reagents used in a process.
• The visualization and analytics toolbox allows you to draw conclusions and infer information, which leads users to take action. For example, periodical report, graphic visualization tool, alerts triggered by certain programmed conditions.

Finally, we must emphasize that it is important to select the solutions that best suit the needs of the business, its business environment and the assets for which the data is being collected. After the solution is selected, users must plan to integrate it into the operational data warehouse, an essential tool in implementing information valuation initiatives. The warehouse will allow anyone to make decisions based on a common and accepted version of what constitutes operational data “truth” and, ultimately, to perform advanced analyses.