How Can I Trust My Data?

During my digital transformation journey with our customers, I receive many questions. Currently, I have encountered a new challenge: “How can I trust my data?”

The question from our customers is simple and extremely significant. With the introduction of new technology like Machine Learning (ML) and in the long-term Artificial Intelligence (AI), we become more and more dependent on our process data.

How do we know if the collected process data is correct? What if there is some measuring error? Some inaccuracies in the measurement? Did we collect all the data? Once it is in the “system”, the data is often considered as correct and is not validated anymore.

Data Integrity

An alternative wording of trust is integrity, to be more specific data integrity. Data integrity refers to the completeness, accuracy, and consistency of data. It is already crucial in the pharmaceutical industry to ensure that the end products meet the required quality standards. Data integrity is rather new for the process industry but as mentioned before increasingly more critical.

The Validation of Process Data

The validation of process data starts at the measurement device. The reliability and integrity of the measurement data can be increased by

• Option Number 1: Using multiple sensors for the same measurement or

• Option Number 2: Using sensor diagnostics proactively

Multiple Sensors

Traditionally, multiple sensors are often not an option due to financial constraints unless it concerns safety; rather than for measurement alone. The reason for this decision is simple: a transmitter costing 500 Euro, could end up costing the company over 5000 Euro once the installation and configuring into the control system considered.

But there is a game-changer around the corner. Increasing the reliability of measurement data will change with the introduction of the Industrial Internet of Things (IIoT) devices that offer easy installation and connection. How will the future situation look? Just imagine that all manometers are replaced by IIoT devices and connected via LoRaWAN to the Cloud and are directly available for users. This way, data can easily be cost-effectively verified through multiple measurements.

Sensor diagnostics

The existing alternative is proactive maintenance. The technology to support proactive maintenance has been available for a long time, especially after the introduction of Device Type Manager (DTM) / Field Device Technology (FDT). However, the change from preventive maintenance to proactive maintenance is a cultural change and not a technical one. I have seen many companies purchasing asset maintenance systems to have access to the DTM of field devices. Still, I have seen only a few companies who have been able to make the real cultural change and using the asset maintenance systems proactively. It is my firm belief that proactive maintenance is the future because of reliability management, safety management and of course data integrity.

Once a field device shows an error, the process control system will still collect the data and forward the data to the data historian system. Once the data is in the data historian system, it becomes more difficult to validate data integrity, though a quality flag can partially solve this. In the control system, the data conditions could be defined as bad, suspicious or good. Data out of range can be considered as bad, data with specific alarms marked as suspicious and the remaining data is considered as good. This quality flag can be added to every data transfer to validate the data in the data historian system.

Production Accounting

A third option to validate the data integrity at a sensor level is production accounting; also known asd yield accounting. At the heart of these systems, there is a mass balance model comprised of nodes and connections that represent the entire site including process units, storage tanks and all the streams that are part of the product movements. The inputs to the production accounting system come automatically from several sources, such as process control systems, data historian systems, laboratory information management system and automated tank gauging. Once all data is in the system, using a model and sophisticated mathematical routines, imbalances of mass can be detected. In most cases, these imbalances refer to inaccurate measurements. If the production accounting system is in use daily, maintenance will receive timely information concerning field devices that are possibly causing the incorrect measurements. Calibration or repair of these field devices results in more accurate measurements which, of course, increase the data integrity.

Data Integrity needs a Mind Shift

Although data integrity has been around for a long time, it has gained more importance due to the use of data for multiple purposes. The technology to increase data integrity is available. Now only the culture must change to make it happen. Do you feel ready for the change?