Digitalization has become an increasingly urgent priority for industrial plants, with remote operations and maintenance offering previously underexplored possibilities for growth and reliability within new social distancing restrictions and supply chain disruptions.
Though COVID-19 has had an unequal impact on different industries and regions, the appetite for digitalization cuts across all industries and countries. A global survey of industrial CFOs published by the Boston Consulting group in July 2020 found that 66% believe that digital transformation will be a significant strategic opportunity in the near term, and that 87% are planning a digital transformation process.
As plants are recognizing the untapped potential of digitalization, many are tasked with making these significant organizational and infrastructure changes on a reduced budget. The stark reality is that in spite of some production increases in recent months, the total reduction in industrial production in the EU stood at 7% in September 2020. Simply put, capital-intensive investment in new initiatives are difficult to finance given current financial constraints and the short- to medium-term economic outlook.
Deploying an IIoT pilot can be challenging at the best of times, with over 70% of companies just starting their digitalization journey, or worse, failing to scale beyond the pilot phase. Best practices for conducting an IIoT proof of concept, such as SKF Enlight AI’s 4-step implementation process, help companies avoid pilot purgatory and its drain on valuable maintenance resources. However, before beginning to evaluate IIoT vendors for compatibility and reliability, plants can take steps to define and limit the scope of a Maintenance 4.0 pilot project by conducting asset criticality assessments.
Criticality should determine which asset data gets monitored and analyzed
Asset criticality refers to an asset’s importance to a business in relation to other assets’ importance to the business. This criticality is not determined based on OEM reliability guidelines but rather within the context of how the asset is used within the plant. The same make and model of plant equipment may have different functions within plants throughout the organization or within the same plant itself. Each of these variations in asset function, along with any changes made to its function or process over time, contribute to the asset’s dynamic criticality score.
SKF asset criticality rankings are calculated as follows:
[Asset critical ranking = likelihood (of failure) X consequence (of failing)]
The Consequence is based on an assessment of what will happen as a result of this asset failing including:
- Loss of production
- Impact on product quality
- Employee health and safety impact
- Environmental impact
- Equipment damage and repair cost
The Likelihood is based on how likely it is that this event will happen using the following as a guideline:
- Very unlikely — Once in 100 years (not known in history, but conceivable)
- Unlikely — Once in every 20 years (history of it happening elsewhere, rarely)
- Possible — Once in every 5 years (history of it happening elsewhere, occasionally)
- Likely — Once every year (has happened before)
- Highly likely — Once a month (happens regularly)
The potential impact of asset criticality assessments on maintenance spending can be substantial. Prioritizing asset work orders according to the frequency and extent to which they can cause damage can help maintenance catch issues earlier on and reduce unplanned downtime in the assets that matter most.
Beyond internal maintenance scheduling and triage, this ranking is similarly helpful in curbing unnecessary IIoT pilot expenses and justifying investments. The reason for this simple: A machine learning predictive maintenance solution tested on a non-critical asset is less likely to succeed, and not just because the asset may not have enough data for the machine learning models to train on.
Rather, a pilot that focuses on non-critical assets cannot generate returns that warrant the implementation of the solution in the long term. If an asset does not have a large number of failure incidents and/or does not trigger any serious consequences when it suddenly goes offline, early failure alerts provided by a machine learning predictive maintenance will do little to improve savings and revenue. In contrast, IIoT pilots that target critical assets can make demonstrable improvements to production objectives and operational efficiency.
The following example clarifies the correlation between the impact of a machine learning predictive maintenance solution and asset criticality based on a single consequence parameter, equipment damage and repair cost.
- Negligible: Minor damage to equipment, minor damage to other equipment (Cost guide: less than $1,000)
- Marginal: Moderate damage to equipment, some damage to other equipment (Cost guide: more than $1,000, less than $10,000)
- Significant: Major damage to equipment, some date to other equipment (Cost guide: more than $10,000, less than $100,000)
- Catastrophic: Destruction of equipment, major damage to other equipment (Cost guide: more than $100,000)
Based on this one criterion, a machine learning predictive maintenance solution deployed on an asset that can have a catastrophic impact on equipment damage and repair cost can provide greater returns than if deployed on an asset with negligible potential for damage. The overall potential for returns is ultimately dependent upon an asset’s overall criticality ranking, which measures its capacity to impact not only equipment damage and repair but production loss, damage to quality and OHS.
Summary and conclusion
Implementing a digitalization process during economic downturn and uncertainty is complex, but there are ways for plants to lower financial investment risks before they even contact IIoT vendors. Determining asset criticality is an initial step plants can take to evaluate which assets pose the greatest risks when they fail unexpectedly and to assign IIoT pilots to these assets. When investments must be made sparingly, determining which assets are most likely to derive value from digitalization processes is essential.
Theresa Farrell is Artificial Intelligence Business Development Manager at SKF.
