Accruent - Resources - Blog Entries - How to Automate Asset Tagging and Referencing with Machine Learning  - Hero

Published: Oct 07 2021

How to Automate Asset Tagging and Referencing with Machine Learning

Have you ever struggled to find the manual to put a piece of furniture together or repair a coffee machine? How about searching for the right documents during tax season? It can be frustrating – and not finding the information you need can lead to broken items, snowballing costs and plummeting efficiency.

If you work in manufacturing, you know that this frustration is multiplied by about a million on the plant floor. And the potential costs of not knowing what documents go with which assets – or which documents accurately reflect the as-built environment – can be much higher, leading to lower efficiency, siloed teams, slow operations and more.

Automating your asset tagging and referencing can help resolve these concerns.

The Importance of Asset Tagging in your Asset Information Management System

If you want to digitally transform or automate your operations, you need to be able to tell your Asset Information Management system which assets are referenced by which documents -- and vice versa. This will not only allow your team to easily find the documents they need to do their job, but also ensure that your documentation is always up-to-date and that you have the information you need at hand to overcome pressing manufacturing challenges.

Until now, this hasn’t happened – and process engineers, document controllers and maintenance technicians have been the human interface to connect the assets (facilities, functional locations, equipment, pipes & lines) to documents (layout drawings, process and instrumentation diagrams, operating procedures, lockout tagout procedures). This has required extensive manual work – and it’s been more inefficient than it should be.

With the right tools, though, you can automate this process for easy referencing and increased efficiency.

You Can Now Automate Your Asset Tagging

With the availability of digital information— and accurate representations of facilities, utilities and other industrial assets— it is now possible to automate asset tagging and referencing multiple assets to multiple documents and vice versa. To enable this capability, you simply need to teach your Asset Information Management (AIM) system how to recognize asset tags in technical documents and match those tags to functional locations or physical equipment.

An obvious location to start looking for asset tags is on Process and Instrumentation Diagrams or P&ID's. P&ID's provide a schematic overview of the production process with an overview of pipes, lines, equipment and instruments. The items on the P&ID are usually tagged according to International P&ID Standards i.e., ISA-5.1, DIN 19227, PIP PIC001, ISO 14617 or BS 1646. These standards allow the machine or system to learn and recognize symbols that are used on a P&ID like:

  • T for Tank
  • P for Pump
  • V of Valve
  • I for Instrument.

Based on this classification information – in combination with a naming convention and sequence number— the system can learn how to recognize asset tags.

This Can Be Achieved Using Machine Learning

How exactly is this accomplished? By utilizing machine learning capabilities within your Asset Information Management system.

Step 1: Teach the System How to Recognize Tags

The first step in the process is to teach the system how to analyze the text describing asset tags based on the naming conventions. Keep in mind that there can be multiple naming conventions for a single tag type. For example, you can tag for usage and/or location of a particular asset. To recognize all tags, the system uses an algorithm with multiple patterns to recognize them. Also '*' operators and concatenation of multiple attributes can be applied in the algorithm to improve the recognition of asset tags.

Step 2: Create Relationships Between Tags and Their Originating Documents

The second step is to identify the asset tag in the multiclass classification and to create a relationship between the asset tag and originating document. When the asset tag is found, the link can be made. When the asset tag is not available or not found, the asset tag can be created or marked as unavailable. It is also possible that the algorithm will need to be optimized to further improve the matching criteria.

Step 3: Apply “Hotspots” to Identify Asset Tags

The third step is to apply 'hotspots' to the document where the asset tags are found. 'Hotspots' are image classifications in a document to identify asset tags. When a hotspot is selected, the user can navigate to the asset tag to find more information on the asset.

By following these steps, users will be able to find asset tags in documents, highlight the hotspots in P&ID's, view the asset tag details and navigate to the asset tag to find other related documents.

asset tracking meridian edms

Conclusion: Using Machine Learning for Asset Tagging Drives Manufacturing Efficiency

There are multiple benefits of machine learning for asset tagging, including:

  1. Increased efficiency of engineering, maintenance and operations
  2. Improved standardization and accuracy of asset tagging and naming
  3. Reduced time and cost finding asset tags in related documents

For more information -- or if you want to learn more about machine learning, asset tagging and hotspots – you can contact me at coen.vromans@accruent.com.

You may also enjoy these articles