Executive Summary
- Preventive maintenance is defined by a fixed timeline and equipment wearing down on a consistent schedule – yet AI and HPC workloads may render this out of date for data centres
- Predictive maintenance, alongside condition-based maintenance (CBM), driven by AI, is therefore emerging as a strong use case – but don’t think it is in competition with preventive maintenance
- According to Schneider Electric research, early CBM implementation can lead to 40% reduction in on-site maintenance and 20% reduction in operational costs
Anyone who is involved in running data centre operations will appreciate the savings made from preventive maintenance and predictive maintenance versus reactive. Deloitte research, for example, found that predictive maintenance can increase productivity by 25% while reducing breakdowns by 70%.
Preventive maintenance is defined by maintenance on a fixed timeline. It aims to stop failures before they ever get a chance to happen, with equipment wearing down on a relatively consistent schedule. At least, that’s how the theory goes. Yet as regular readers of Data Centre Insight will know, AI and high-performance compute (HPC) workloads are pushing the boundaries in more ways than one.
The ‘extraordinary demand’ placed on the grid by AI and HPC means power quality is a ‘rapidly ascending’ concern, according to Christopher Butler, president, embedded and critical power at Flex. “While data centres running traditional workloads have largely solved for power quality, AI/HPC applications are raising new challenges as the nature of compute changes,” wrote Butler in September.
Predictive maintenance, therefore, comes in utilising technologies from IIoT sensors, to analytics, to AI, to remediate issues when they start. As ABB points out, preventive and predictive maintenance are often positioned in opposition to each other. Yet this is a misnomer, as predictive maintenance is part of the wider preventive maintenance family, alongside time-based and condition-based practices.
AI is making its impact here too. Condition-based maintenance (CBM), as a Schneider Electric (SE) blog post from October outlines, is ‘powered by AI and is fast becoming a necessity in ensuring both competitiveness and resilience.’ According to the company’s own research, early CBM implementation can result in up to 40% reduction in on-site maintenance, and a 20% decrease in operational costs.
“With CBM, data centres can evolve into efficient, resilient, and sustainable facilities, ready to harness AI to optimise operations and drive innovation in an increasingly demanding digital world,” Canninah Dladla, cluster president at SE noted.
Schneider Electric is cited by GlobalData as a provider of ‘sophisticated’ predictive maintenance solutions to the power industry, alongside GE Vernova and Siemens. While the July study, ‘Predictive Maintenance in Power: Strategic Intelligence’, looks more at wider facilities such as power plants, the numbers are similar. AI-enabled predictive maintenance, in the analyst’s opinion, has the potential to decrease maintenance expenses by as much as 30% and boost equipment availability by 20%.
“As the power market continues to evolve, predictive maintenance emerges as a pivotal driver of innovation and efficiency,” explained Rehaan Shiledar, power analyst at GlobalData. “It not only bolsters the industry’s shift toward digitisation but also aligns with the increasing focus on sustainability by aiding power companies in managing their assets in an environmentally responsible manner.
“The adoption of predictive maintenance is poised to rise as stakeholders in the power market acknowledge its capacity to foster operational excellence and propel business success,” added Shiledar.
