However, a famous 1978 reliability-centered maintenance report from Nolan and Heap revealed that 89 percent of failures are random. The incline in failure probability over time is an oversimplification. In fact, performing invasive preventive maintenance actually introduces a new burn-in phase prematurely and can increase the probability of failure. So if not time, what factor can be used to determine the best schedule for maintenance activities?
Predictive Maintenance: Driven by Data
Fortunately, a benefit of rotating equipment like pumps is that they provide a signature of health. There are many opportunities to capture trendable and repeatable data, such as pressure, vibration and temperature. Performing predictive maintenance means using this data to end users’ advantage, keeping detailed records to capture baselines over time and then looking for overall increases that indicate a problem.
The integration of low-cost sensors directly on equipment is now becoming standard with OEM machines. There is also the wireless option, which is becoming more affordable and therefore practical to implement beyond only critical assets. The integration of mobile technology and the internet are making data accessible and actionable from anywhere in the world. As these technologies mature, the effort to shift from reactive to predictive is becoming much simpler and more affordable.
When a small North American oil refinery experienced a fire at the bottom of a vacuum tower in 2008, the resulting three-day shutdown caused $1.5 million in damages and lost production.
The root cause of the problem was a failed mechanical seal on a relatively small, overhung, inline API-style pump. This incident severely impacted the bottom line and created a dangerous situation for the refinery personnel.
In late 2009, the oil refinery made the decision to implement a wireless continuous monitoring system. With this technology, a data monitor examines the process and mechanical side of the pump, taking readings on bearing vibration, bearing temperature, suction pressure, discharge pressure and motor amps.
The data monitor wirelessly connects to a communications module that feeds this information to two sources:
- Pump key performance indicators—such as pump discharge pressure, pump suction pressure and motor amps—are fed to the refinery’s control room for operators to track where the pump is running on the pump performance curve and whether it has adequate net positive suction head to prevent cavitation.
- Machine health data—such as tri-axial bearing vibration and bearing temperature—are sent to a Web-based condition monitoring platform, enabling the head reliability engineer to monitor the pump at all times, from any location, through the Internet or a mobile device.
The reliability engineer was able to track rapid upward trends with greater convenience, which helped catch a step change in the pump’s outboard bearing vibration.
The pump was shut down, and the bearings were replaced to prevent another catastrophic failure.
Since implementing a predictive continuous monitoring system, the pump has increased mean time between failures (MTBF) and avoided catastrophic failure. In addition, the facility reported no unplanned maintenance activities on this pump for more than three years.
While predictive maintenance uses data to help foresee equipment issues, proactive maintenance takes this philosophy one step further. Also referred to as “precision,” it requires an understanding of why a failure occurred and a commitment to implementing change to prevent future failures and increase the MTBF.
This depends heavily on participation from cross-functional teams because improving reliability is everyone’s job, from maintenance to management.
Practicing proactive pump maintenance does not always require exhaustive root-cause analysis. Start with basic questions, such as:
- Where is the pump operating on the curve with respect to best efficiency point (BEP)?
- Is the pump the right size for the application?
- Does evidence of process upset conditions exist?
- Does the pump experience dry running?
- Is the pump run against a throttled or closed discharge?
- Is the mechanical seal and flush plan properly selected for
By answering these questions, top targets for change will be identified. It starts with data integrity. Standardizing on a thorough set of failure codes is important. More important is setting and enforcing the policy to incorporate these codes into a computerized maintenance management system (CMMS).
When everyone is adamant about putting in fault codes that accurately describe the failure modes, bad actors are quickly identified. By proactively working through the bad actors, repeat maintenance activities caused by these machines can be eliminated, reducing costs and increasing productivity.
Driving Procedural Change
A 700-pump North American chemical plant was under extreme pressure to reduce pump maintenance costs and improve MTBF. Its plan was to embed a pump specialist as a full-time resource to transform the facility’s maintenance practices.
The first task was to reconcile the pump work order history because vague or generic failure codes made the data less useful. The specialist made procedural changes to ensure that the fault codes and job materials were correctly assigned in the CMMS for future trending.