Why organisations need to implement asset intelligence

For organisations in industries such as utilities, manufacturing and other asset heavy industries, the company’s physical assets form a significant part of their balance sheets. These assets could be dams and filtering plants for water utilities, network infrastructure including poles, transformers and conductors for electrical utilities, large network infrastructure for telecom organisations, road and vehicle Infrastructure for transportation organisations and plants and machinery for manufacturing and mining.

It is essential that these assets are maintained impeccably for the following key reasons.

1. Significant revenue loss, for example if a mining company’s drilling machinery breaks down during a critical time, delaying critical infrastructure projects.
2. Environmental impacts, for example when offshore Oil rig malfunction- causing oil spills into the sea and affecting marine life.
3. Health and safety impacts– like when an overhead conductor of an electrical distributor snaps at a busy location.
4. Loss of Company reputation, when a Telecom organisation has network issues due to malfunction of ageing network infrastructure.
5. Customer dissatisfaction- for example when an underground water pipe breaks, causing supply issues to hundreds of customers of a water utility.
6. National Security issues- for example when critical defence and other equipment malfunction.

Maintenance of Assets:

In order for the above eventualities not to occur, organisations spend billions of dollars in the upkeep, maintenance and replacement of their physical assets. The maintenance of assets is of two types:

• Planned maintenance: Planned maintenance is when an asset is replaced as per plan. It can be replaced as per plan for two reasons.
  • Age based replacement, which is when an asset has typically reached its end of life as prescribed by the manufacturer.
  • Condition based replacement, which is when an asset needs to be replaced due to the existing poor condition of the asset.

• Unplanned maintenance: Unplanned maintenance is the biggest pain point for an organisation. This happens because an asset has already failed, possibly in the middle of a critical process. This leads to unplanned downtime, leading to the consequences described earlier.

How can Asset Intelligence manage Asset failure?

Asset Intelligence is used to decrease unplanned maintenance and move all maintenance into planned /condition-based maintenance. This is because by the time a failure of an asset has occurred, it is often too late to avoid large impacts. Therefore, asset Intelligence is used to predict/monitor a failure event before it occurs. The larger the lead time that is available between a prediction/monitoring event to actual failure, the team gets a longer lead time to reduce any potential impacts. Better understanding of this at the organisation levels helps it to manage their supply chain and workforce allocation for the year.

Asset Intelligence techniques

1. Asset conditional monitoring: Asset conditional monitoring is specific to those assets that have sensors that monitor the operational parameters of a particular asset class. Obviously, this does not include low value, high volume asset types for example poles in an electrical network. Many organisations elect to just replace these low value assets on failure rather than have any sensors attached to them as long as there are no catastrophic effects of failure.
   
   Asset condition monitoring continuously evaluates real time data from the monitoring devices. It is set up to alert in the event of key KPI’s being breached. This could be for simpler systems when the temperature variable crosses a threshold. In more complex systems, this could be set up based on complex rules that mimic a failure signature of a particular asset class. For example, in PCP pumps, this could mean when the torque of the pump is over a particular threshold along with rapid temperature increases. Once this happens, this sets off an operational alert that is actioned upon. What is also important to understand is that there is a lot of human knowledge that required to define the conditional monitoring thresholds and business rules.
   
   
2. Predictive Asset maintenance: Predictive asset maintenance goes one step ahead of conditional maintenance. Using a combination of supervised, unsupervised and semi supervised machine learning techniques on the available data, the algorithm can
   • Use regression to calculate probability of failure of an asset.
   • Use classification techniques to look at images/Video data of an asset to classify those assets that are at high risk of failure.
   • Use semi supervised methods by combining unsupervised techniques like clustering to arrive at conditional age of asset and combine that with supervised techniques that use conventional techniques.
     
     
3. In a marked difference from conditional monitoring, historical data is used to arrive at the patterns that are used in this methodology.

Data for Asset Intelligence:

The data that is typically required for Asset Intelligence are of the following categories.

• Asset master data: This data contains the hierarchical asset classes data and all their attributes. These are typically maintained in SAP, Ellipse, Maximo and other large enterprise Asset tools. This will also contain Asset failure history data, material data etc.
• Field Inspection data: Inspection data brings data from all physical inspections of the asset. Inspections are usually scheduled periodically. While this data is entered in field entry systems and is structured, there could be some unstructured text data from comments.
• Operational data from systems: This could be network data, Electrical or water parameters etc.
• Environmental data: Environmental data could be several variables that potentially affects failure. This could include rainfall, distance from sea (corrosion), vegetation data, temperature data etc.
• Sensor data: Real time data that is attached to a particular asset class and shows variables in real time. This is typically time series data.
• Unstructured data: This includes data from photos, videos, any unstructured text data etc.

Critical Functionalities required in a modern Data ecosystem to enable Asset Intelligence:

A modern data ecosystem is central to implementing good asset Intelligence. There are certain key functionalities of a modern data ecosystem that is required to enable a strong baseline for Asset Intelligence namely:

• Single source of truth or Asset 360: Like a customer 360, an asset 360 refers to a single view of asset data. This is created by bringing in asset data from all the different sources mentioned in the previous section and creating a unified single view of an asset across all sources of data. This is often complex due to the complex nature of assets and their attributes.
• Real time streaming data ingestion and transformation: The critical sources of asset data often come from devices and sensors such as IoT devices. This requires capabilities in the modern data platform to be able to ingest, transform and consume real time data in scale.
• Strong data governance and Data quality frameworks: Like many other use cases, but even more so, asset Intelligence requires strong data governance and stewardship to identify the business glossary, manage the data catalogue, understand data lineage and apply the right data quality rules to be able to offer datasets of the highest quality.

Challenges of Asset Intelligence

There are several challenges that an organisation goes through to implement Asset Intelligence.

• Complexities with Assets: Assets have intrinsically complex mechanisms. An asset class in turn can have hundreds of component assets. Each of these components have different failure signatures and complexities. Very strong domain knowledge is required to understand failure signatures and complex thresholds to implement asset Intelligence.

• Volume of assets: For a medium size organisation, there could be potentially hundreds of thousands of asset types and classes. Therefore, there needs to be strong priority matrix that needs to be built based on the cost of the asset class/impact of failure vs volume of failures. This will help the organisation focus on high impact/high volume asset classes.

• Model inaccuracies: Unlike simpler use cases like customer analytics, AI/ML for assets can sometimes be an inexact science. It is likely that the model will require several rounds of iterating and fine tuning to be reasonably accurate.

• Change management: For a technically advanced field force that has been doing asset maintenance from decades, it is sometimes difficult to use insights of an AI system. It requires careful change management and creation of processes that allow a human to make a final decision based on the intelligence available to them.

For more information on how Altis Consulting has helped organisations implement Asset Analytics please connect with us today.