Without context, the value of data is self-limiting. It’s not just the collection of data that matters, but the ability to transform raw data into decision-ready data. Attempts to do this at the analytics layer can often lead to significant challenges where disparate data sets must be combined in a meaningful way. Only through context does raw sensor data deliver an answer to complex questions, for example:
- Do environmental or geographic effects cause global operational performance variations?
- Is process performance related to the make or model of equipment types?
- What does maintenance history tell us about an asset’s reliability?
- Which operating behaviors lead to the greatest long-term asset integrity indices?
- Did a singular event in the past have residual effects on the current process?
In most organizations, such questions are asked not only by engineering, but also teams outside of operations, including cost accounting, capital procurement, customer service, project planning, outage response, contracts management, compliance reporting, exploration, and so on. For these teams outside of core operations and engineering, context becomes critical in order to derive value from sensor-based data; context becomes mandatory.
For example, at one of our customers who runs their operations in remote areas across the world. When large, critical equipment breaks down, they incur not only the cost of parts and repairs but also the opportunity cost associated with stalled production. In order to manage downtime more effectively, they have been analyzing historical data to identify patterns that may be leading indicators of part failure.
Pattern recognition alone is of significant value. However, in order to leverage this knowledge to make business-driven decisions on what action to take, they consider additional business context. They overlay information from their maintenance, planning, and finance systems along with market conditions. For instance: Are the parts readily available if they run in to failure? What is the current market price of the product? Strategic decisions can then be made to choose between running at 110% capacity where parts are available and market conditions are favorable vs. running at 60% capacity to extend the life of the equipment where parts are not easily available. These questions are best answered by understanding the physical behavior combined with relevant context to optimize business value. It may be natural to assume that a point solution must be deployed to solve each challenge, but in many cases the solution lies in a disciplined approach to applying context to existing systems.
There is an iterative process at each level starting with access to data in relevant context for the stakeholder, a better understanding of physical behaviors based on insights from the data, and ultimately, optimization based on those insights. With low-cost sensors and pervasive networking, access to data now extends far beyond the domains of automation and control.
Context and modeling play an ever-more critical role to get data into a meaningful format for analysis. Data models that are tied to the physical equipment and process models are a key success factor to provide context at a fleet level. With the right models, and data shaped in an appropriate manner, data from fleets of assets can quickly lead to novel insights. With insights, action can be taken to continuously improve operations, achieving a more predictable mode of operating. This cycle is repeated at all levels (asset > process > fleet > enterprise) and success is measured by using business performance as a key metric
Is your data decision-ready? If not, investigate how context can be collected along with your sensor data. Trying to apply context at the analytics level may come with high cost of time, money, and perhaps even the project.
Register to attend the Energy Forum.