How predictive maintenance transforms semiconductor fab maintenance - Customer story
European wafer manufacturer reduced unplanned pump failure significantly
Semiconductor fabs operate under pressure to maintain high safety, quality, productivity and reliability. Vacuum pumps in the subfab play a critical role in sustaining continuous production. If these pumps fail unexpectedly, the impact can be severe: costly process tool downtime, production delays, wafer losses, and increased maintenance expenses. While traditional maintenance strategies like reactive and preventive approaches are essential, they can still leave gaps when it comes to preventing sudden failures.
STMicroelectonics Fab in Catania, Italy
Preventing vacuum pump downtime in wafer production
One of our European customers, STMicroelectronics’ Catania main site wafer fab, is known for its expertise in analog and power semiconductor manufacturing. With a strong focus on innovation and operational excellence, our customer was looking for a smarter way to manage its extensive pump fleet and reduce the risk of unplanned downtime.
They faced a critical issue: process tool downtime caused by unexpected pump failures. These failures disrupted production, increased costs, and strained resources.
Traditional maintenance strategies, whether reactive fixes or scheduled preventive replacements, did not reduce unplanned downtime nor optimise asset utilisation.
Detecting pump issues before they occur
To address these challenges, ST’s main Catania site partnered with Edwards to implement a predictive maintenance platform that collects and analyses real-time data from vacuum pumps. This enabled the fab to:
- Apply predictive algorithms to anticipate vacuum pump issues before they occur.
- Take targeted actions that minimize failure risks and reduce unnecessary replacements.
By moving to predictive maintenance, ST gained greater control over pump uptime, process tool availability and operational efficiency.
Implementation
The change didn’t happen overnight. ST and Edwards started the journey together with a collaboration between ST and Edwards regional team, supported by a group of technical specialists and data scientists.
As part of this journey, Edwards team applied fleet monitoring to accurately identify symptoms for common failure modes and implement predictive actions. Expert-led analytics enabled objective, data driven decisions translating into reduced corrective maintenances. These results led ST to then explore how Edwards predictive maintenance could also be used to improve Mean Time Between Preventive Maintenance (MTBPM) by assessing pump condition before scheduling preventative maintenance.
This strong collaboration resulted in fewer unplanned vacuum down events, with fewer corrective interventions, extended MTBPM intervals and clear improvement in tool availability.
Results
Extended MTBPM intervals and corrective maintenance events at ST over time
- Unplanned events dropped significantly, from being the majority to just 10% of total replacements in the last two years. In years 1 and 2, the fab averaged 10 Corrective Maintenances (CMs) per year; by years 4 and 5, this was reduced to just 1.5 CMs per year, with one of those caused by external factors.
- 50% reduction in preventative pump replacements. Asset utilisation improved from around 10,000 hours MTBPM to around 20,000 hours from year 4 onwards, effectively reducing preventive replacements by half.
- Both unplanned and preventive replacements decreased simultaneously—a rare achievement in maintenance strategies, proving the effectiveness of predictive maintenance.
- Cost savings achieved by reducing unplanned events, improving predictability, and empowering teams to make proactive, data-driven maintenance decisions.
This achievement demonstrates that predictive maintenance not only delivers substantial improvements in process tool availability, improves subfab operational efficiencies and optimizes asset utilization but is also scalable across even the most complex manufacturing environments.
Testimonial
I was among the first to believe in EdCentra, sponsoring and organizing every initiative. By pushing equipment to its limits, we achieved unprecedented uptime. I fostered collaboration, embedding a datadriven culture that proves vision delivers extraordinary results.
Continuous improvements in predictive maintenance
Vision and courage in adopting innovative solutions have delivered measurable, lasting results. We have an agreed roadmap to deploy algorithms, ensuring continuous improvement. With this new predictive maintenance platform now in place, ST is ready to maximize equipment availability.
