Pump and motor repairs decrease with predictive maintenance

June 27, 2011
In this installment of What Works, a fuels refinery uses vibration monitoring to increase mean time between failures.

The ConocoPhillips Ferndale Washington Refinery (www.conocophillips.com) was constructed in 1954. Ferndale is an integrated, single-train fuels refinery whose operating units include FCC, HF alkylation, reformer and the typical array of supporting units. The refinery’s capacity is 110,000 barrels/day (BPD) and it has about 1,000 pieces of equipment. The rotating equipment group has full responsibility for the machinery’s condition, effectiveness and mechanical integrity. The staff includes three rotating equipment reliability engineers, one part-time vibration monitoring technician and a maintenance superintendent.


Dating back to 1991, the ConocoPhilips Ferndale plant had no formal vibration monitoring program and used a reactive approach to maintenance. The team was so busy responding to problems that required immediate attention that there was no time to make long-term improvements. Operations had such little confidence in the plant’s pumps and motors that it was a common expectation that even spared equipment had to be repaired and returned to service the same day. This generated a lot of machinist overtime as the team routinely spent its weekends overseeing machine failures.

It became apparent that something had to change. The plant needed a way to get away from reactive maintenance (run to failure) and move to a more strategic, proactive approach to addressing machine reliability and uptime.

When the plant investigated machine condition assessment technologies, it evaluated oil analysis and thermography technologies and concluded that, for the machinery and the types of failures being experienced, vibration monitoring offered the highest potential return on invested money and time. The team also recognized the necessity of regular, repetitive data collection to establish a representative history and statistical baseline data from which deviations could be recognized accurately.

“In my opinion, vibration monitoring is the cornerstone to identifying machinery problems,” says Ryan R. Barnes, maintenance superintendent of rotating equipment services at the refinery. “How else can you stand next to a running piece of equipment and determine, without shutting it down, that it suffers from impeller unbalance or misaligned shafts, or is at the early stages of bearing damage.”


The team began the search for a vibration monitoring system, one that emphasized accurate and repeatable data collection and easy analysis. ConocoPhillips Ferndale considered several approaches to data collection. One was to use Ferndale operators or machinists to collect the vibration data and have reliability engineers or contractors perform the analysis. The other was to outsource the entire condition monitoring program. The team decided to hire a committed, on‐site data collector to be involved with the entire program, from data collection to working with an outside vendor that would handle primary analysis. This way the data collector wouldn’t be pulled away for pump repairs or to perform operator rounds. Also, because the plant has such a small staff, it was critical that the vibration monitoring system be fast, accurate and repeatable, and most importantly, diagnose potential problems easily.

The vibration condition monitoring program

After reviewing available vibration monitoring offerings, ConocoPhillips Ferndale selected the Azima DLI vibration data collector with ExpertAlert software. This incorporated a number of features that would help the team be successful.

Triaxial simultaneous data acquisition collects vibration data in three dimensions in less than 20 seconds per test location (for a typical 3,600 rpm machine). This includes both low (0‐10 orders) and high (0‐100 orders) frequency ranges.

A bar code scanner allows point identification without searching through a list of pumps or test point locations. Data collectors point and shoot the bar code label glued to the pump base or foundation and the data collector begins recording immediately. The labels are designed to withstand years of harsh conditions in outdoor refinery environments.

Test location attachment pads with a screw-mounted connection to the transducer is an accurate means of collecting repeatable vibration data, especially in the higher frequencies. The bronze pads have an alignment notch to orientate the triaxial transducer and a 10x32 threaded captive screw for connection.

Vibration monitoring programs use some sort of data screening technique, either enveloping or mask alarms. Azima DLI uses an analysis program called ExpertAlert. This software uses statistical analysis to compare the accumulated average baseline signature against the test data. The diagnosis is based on comparing the tested machine’s spectral peaks to this baseline using a complex set of rules for the machine. About 20% to 30% of the hundreds of pumps and motors collected in one day will generate a significant diagnostic report that requires follow-up manual analysis.

Lessons learned

For the Ferndale plant, it was imperative to have one person, with real interest, enthusiasm, accountability, and appropriate training and certification, who owned, and was responsible for, the predictive monitoring program. In addition to collecting and analyzing data, this person must have enough time to analyze the few machine problems that arise and track the statistics and metrics needed to refine the program and recommend plant improvements.

During the past 16 years, unplanned machine failures have diminished to near zero. The plant’s pump mean time between failure rate increased from 22 months to 52 months in a 10-year period after implementing the Azima DLI condition monitoring system. The annual maintenance costs for pump repairs have dropped from $1,300,000/yr in 1994 to $600,000/yr within 10 years. Virtually every maintenance task at the plant is planned. Equipment failures are few, and machinist overtime has been nearly eliminated.

Since 1991, the Conoco Phillips Ferndale plant has come a long way. What was once a run‐to‐failure plant in which machine breakdowns were common and machinist overtime was inevitable, is now a plant with a 91% on-stream availability with a staff of only five machinists.