How predictive maintenance reduces unplanned downtime in processing plants

Maintenance is essential for processing plants to improve uptime, enhance productivity and reduce the need for rework. Many facilities, however, are stuck in break-fix maintenance modes that create a reactive cycle: Assets fail unexpectedly; maintenance teams respond, but because speed is the priority, they are more likely to miss root causes and only solve symptoms.

Predictive maintenance offers a better approach. By leveraging analytics to track process and machine performance, teams are better prepared to spot and solve issues before they lead to unplanned downtime.

According to IBM research¹, predictive maintenance can reduce costs by 25% to 30% by identifying small issues before they become big problems. This approach also allows teams to schedule necessary maintenance during slower production periods, in turn minimizing operational disruptions.

The impacts of unplanned downtime

Unplanned downtime is naturally disruptive. Processing plants depend on reliability and repeatability to ensure production outputs meet cycle time expectations and quality standards. Common impacts of unplanned downtime include:

• Lost revenue: When production lines go down, so does revenue. Even short periods of downtime can result in significant losses and create bottlenecks that drive ongoing challenges. For example, if a production line depends on a specialized piece of equipment for quality control or packaging, and this equipment goes down, the result is both immediate and long-term. Even if teams get machinery up and running right away, it still takes time to get production timelines back on track.
• Asset damage: The reactive model of maintenance can also lead to asset damage. Because companies do not know when maintenance issues will occur, they also do not know how severe these issues will be. In some cases, problems may be fixed with a quick parts replacement or system upgrade. In more severe cases, unaddressed issues can accelerate wear or lead to failures that require extensive repair or full asset replacement.
• Increased maintenance workloads: Unplanned downtime also creates increased maintenance workloads. This is because maintenance cannot simply focus on solving the issue at hand — teams also need to try to track root causes while balancing the need for a quick restart. This can be further complicated by supply chain and spare parts issues. If failures are not expected, organizations may not have the correct type or number of spare parts on hand.
• Lowered customer satisfaction: Customers and clients depend on the reliable delivery of quality parts and components. If product plans cannot meet these expectations due to unexpected downtime, customer satisfaction suffers. In a best-case scenario, buyers come back but are less tolerant of mistakes. In a worst-case scenario, customers find new and more reliable suppliers.

Key components of predictive maintenance strategies

It is one thing to talk about a predictive maintenance strategy — and it is another to put it into practice. Predictive operations rely on data: information collected from assets, devices and technicians that help identify issues and create targeted responses.

The first component of an effective predictive maintenance strategy is data gathering. This requires a combination of connected devices, such as vibration, temperature, pressure and acoustic sensors, that collect data in real-time. 

Next is centralization. Data from these devices must be funneled into solutions such as manufacturing execution systems (MES), computerized maintenance management systems (CMMS) and enterprise resource planning (ERP) tools. This provides a single source of truth for all maintenance decision-making and processes, which helps target key issues and reduce the risk of redundant work. 

Collected data is then analyzed to identify patterns and trends, which inform maintenance operations. Teams create assessment and repair schedules based on this data to limit the risk of unplanned downtime and proactively address issues such as wear and tear.

Consider a processing plant struggling with the unexpected failure of a key assembly line asset. Despite best efforts, the equipment fails at least once a month, sometimes more. While maintenance teams are quick to respond, the facility has been unsuccessful in identifying the root cause.

To solve the issue, the business takes a predictive approach. Teams install IIoT sensors that collect operational data and funnel it to software solutions. Data analysis reveals a recurring issue preceded by a change in vibration and acoustics. This leads to a structural assessment of the equipment, which pinpoints a stability issue. Reinforcing the machine’s base and bracketing eliminates the vibration, stabilizing performance and preventing the recurring unplanned downtime associated with the issue.

Equipped with this information, teams can also create a schedule for regular review to ensure the equipment remains stable. 

Benefits of adopting predictive maintenance operations

Adopting predictive maintenance operations offers several benefits, such as:

• Improved visibility: Data analysis helps identify symptoms, root causes and possible fixes. It also improves process visibility. The more teams know about how machines act and interact, the better prepared they are to spot problems early and take action quickly.
• Increased accuracy: More accurate problem identification reduces the time required to complete repairs, test new fixes and bring machines back online. In addition, greater accuracy allows teams to leverage data for long-term maintenance planning. For example, if predictive analysis suggests that equipment will last four years with proper maintenance, companies can plan for capital spending rather than simply hoping for the best.
• Enhanced scheduling: Predictive maintenance operations allow teams to schedule maintenance during low-impact windows. This minimizes the disruption to operations while still enabling companies to stay ahead of possible failures. Data-driven scheduling also helps teams create repair, upgrade and replacement schedules that reduce the impact of wear and tear on machine performance and throughput.
• Extended asset lifecycles: Finally, predictive maintenance helps extend asset lifecycles. Consider a high-temperature, high-pressure piece of machinery that suddenly fails. If teams can control the damage, it may be repairable. If sudden pressure and temperature spikes render the machine unusable, significant spending may be required to remove and replace equipment.

Data analysis combined with scheduled maintenance helps ensure that machine lifecycles match operational expectations.

Making the move to predictive processes

Adopting predictive processes does not happen overnight. Even with a solid strategy in place, processing plants need a plan for effective implementation, measurement and evolution.

When it comes to implementation, facilities typically face two paths: Building predictive maintenance capabilities internally or partnering with a predictive maintenance services provider. Both have potential advantages. For example, staying local means more control over data and processes. The caveat? Most maintenance teams are stretched thin already, creating the potential for new gaps to emerge as focus shifts to predictive operations.

Partnering with a trusted service provider, meanwhile, delivers both visibility and value. Providers can take on as much or as little of the predictive process as companies prefer, allowing teams to balance change and consistency.

Measurement is essential to verify that predictive plans are working as intended. Achieving consistent measurements requires the use of key performance indicators (KPIs), which help quantify the effects of predictive maintenance.

Some of the most common KPIs for process plants include overall equipment effectiveness (OEE), mean time to detect (MTTD), mean time to repair (MTTR) and mean time between failures (MTBF).

Consider MTTD, which measures how long it takes for the team to detect a problem after it occurs. Under a reactive maintenance model, detection often occurs only after failure or noticeable performance degradation. As a result, the MTTD can vary from days to weeks to months, depending on the original cause.

Using a predictive maintenance strategy, meanwhile, enables the use of analytics to identify trends. These trends allow teams to take targeted action that identifies issues based on first causes rather than eventual impacts.

Finally, organizations need to make room for evolution. As processes change, assets are upgraded, and machinery is replaced, the nature and frequency of failures can shift. Operations that were once reliable and accurate may become more irregular, in turn requiring a shift in maintenance focus. 

Addressing evolution requires regular reevaluation of maintenance processes and their impact. For example, if OEE is steadily decreasing despite best efforts, companies may need to reconsider their maintenance strategies. If production throughput begins to fall, teams may discover that previously low-impact repair windows are no longer viable, necessitating a shift in predictive priorities.

Staying one step ahead

Reactive maintenance remains common — and costly — for many processing plants. Predictive strategies help companies stay one (or more) steps ahead of potential problems, in turn reducing maintenance complexity and cost.

Building a predictive maintenance strategy starts with data collection, followed by centralization and targeted action. Combined with low-impact, scheduled downtime for regular assessments, repairs and upgrades, companies can get the best of both worlds: Reduced unplanned downtime and improved productivity.

References

¹https://www.ibm.com/think/insights/ai-in-predictive-maintenance