Data is the first Challenge!

We are all chasing the improvements that new technology provides of reducing downtime, improving maintainence practices and costs along with benefits of worker safety and overall sustainability of our operations in industrial settings.

One of the key challenges is the data, and yes we have a lot of it collected by control systems, within proprietry vendor systems along within maintainence systems as such as SAP. Additionally when attempting to solve an outage or optimise a process, acces to the design data is paramount. Some of these challenges include:

• Data is siloed or inacessible within Legacy or Vendor systems
• Lack of standardisation across different vendors and communication protocols
• Poorly calibrated sensors or manually entered data can raise questions about the Data Quality

The reality of some initiatives to implement Machine Learning (ML), Artifical Intelligence (AI) or even Agentic AI is scuttled due to the lack of data to be able to train and configure these models and systems. At the oputset it is important to understand the data landscape and how you may address any of these shortcomings. Some strategies in this area include:

• Instrument assets with smart sensors and ensure proper calibration
• Integrate data streams via unified platforms (e.g., OPC UA, MQTT)
• Tag and contextualise data using asset hierarchies and metadata
• Clean and validate datasets using automated tools and human oversight
• Building feedback loops to continuously improve model accuracy

AI & ML approaches can be utilised to support identifying gaps and defficiencies. Unstructured data, as such as pdf reports and manuals, can also be processed to extract data in particular from legacy systems. Extracting key parts of free-text fields from maintainence records to determine common issues can also be achieved through natural language processing approaches.

I would always recommend as part of the definition of a new initiative and project, that in conjunction with setting out the key functional requirements, is an assessment of the data required to achieve the goal. There have been a lot of instances that spending time on sorting out your data is more valuable than applying the latest trend.