When I speak with CIOs and their staff, the topic of digital transformation always leads to a discussion around how process (i.e., time-series, historian, sensor) data is the starting point, but it’s difficult to work with and organize data in a way that represents how equipment and assets exist in the physical world.
Industrial companies have begun to address the problem by adopting Asset Data Models, which represent the physical structures and relationships of industrial equipment and processes. Asset Data Models are crucial for equipment benchmarking, cross-site comparisons, and underpin every kind of analytics. These models are also the central point for both internal stakeholders (e.g., engineering, compliance and safety) as well as external data consumers, including OEM and services vendors who need organized process data delivered in a useable format.
IT teams are often tasked with creating one perfect master model that meets subject matter expert (SME) and business requirements, but it’s impossible to accommodate everyone’s needs (e.g., field engineers need one asset view, data scientists need another). As a result, sites and individuals in the organization often reject such a singular, rigid model and instead continue to use a multitude of incompatible, localized, and siloed models in separate databases. This creates significant data integrity problems within the organization.
To meet the needs of all the data consumers in an industrial organization while maintaining data integrity, IT needs to adopt a flexible, adaptable model that lets different data stakeholders use and add to it the way they want.
For the industrial enterprise, a sustainable Asset Data Model needs to be two seemingly opposing things at once:
Standardized across the enterprise.
Yet flexible for each person, function, and application.
File vs. Pile analogy: Instead of trying to organize everything into one rigid, perfect model, let data pile and focus on enriching it with metadata to enable powerful downstream capabilities such as rapid querying.
A graph-based approach to Asset Data Models is needed. Like in Facebook, LinkedIn, or Google, a graph structure containing all process data serves as a single source of truth, allowing all data consumers to get the data they need in the format that they need. It supports the 3 core tenets of Build, Maintain, and Sustain.
Process Engineers, Operations Managers, and SMEs can use Asset Data Models to augment process data with key contextual information. Then, using the complete set of information captured from experts and other systems, IT can enrich assets in the data model with metadata, building a network of relationships (e.g., process relationships, instrumentation relationships, asset-based relationships, hazard relationships). This relational network of metadata results in a graph-based, malleable Asset Data Model. It can be used to generate any structure needed by all stakeholders, business units and applications in the organization.
One of the biggest challenges today is that as processes and equipment changeover time, existing models become outdated and irrelevant, and therefore unusable. Often, model maintenance falls on the shoulders of a site engineer who is already tasked beyond capacity.
The Asset Data Model is a “living graph” that evolves, continuously and automatically updating with changes over time (e.g., new process data comes online in the form of tags, instrumentation degrades, equipment undergoes improvements). The model ensures that process data will always have an updated, accurate, and readily usable model for analytics to deliver greater value for the business.
An additional maintenance challenge is ensuring the data is high quality, free of issues, and trustworthy. An Asset Data Model must be able to surface both data model and data value issues for IT and engineering to address.
Often overlooked is the need to export an Asset Data Model in diverse forms for applications. Whereas countless databases increase infrastructure and support costs, a sustainable enterprise Asset Data Model is a single source of truth for process data that can be consumed by any business application or used to generate countless unique views for various stakeholders within the organization (e.g., Optimization, Reliability, Process Safety, Compliance).
An Asset Data Model should provide the flexibility and freedom that people need to see and use data the way they need to, in their own internal language and existing data structure — no retooling required. This unlocks data collaboration for the wider organization, allowing people to easily perform analytics that support business goals such as reducing operational risk and improving asset utilization.
About your Guest Blogger:
Andrew Soignier is the Head of Industry Solutions for Element Analytics, Inc. With over 20 years of experience, he has held leadership positions spanning sales, product management, strategy and marketing. His first 10 years in industry were dedicated to process control and automation with a focus on electrical, process safety, critical controls and rotating equipment. He has spent his last 10 years in the enterprise software space partnering with client leadership to solve tough problems in the areas of trading & marketing, supply chain, asset management, mechanical integrity, and process safety.
The Element GraphTM is a graph-based approach to enterprise Asset Data Models. The Element Graph takes shape as process data is prepared from a raw state into a workable state. It also enables IT to surface data issues to engineers, who can then address them in the field (e.g., check for missing sensors, fix configuration). In this way, the Element Graph enables greater collaboration between IT and engineering.