Building a PM Schedule From Failure Modes and Run Hours

Copying the maintenance chart from a machine manual gives you a PM list. It does not give you a good one. A mid-level technician builds the schedule from how the machine really fails, sets each interval from data, and writes tasks that any qualified person can do the same way. Done well, the schedule catches problems with less labor. Done badly, it burns hours on checks that find nothing while the real failures keep coming. Reminder: every task that opens a guard or reaches into a machine starts with lockout, by an authorized employee.

What You Will Be Able to Do

• Break a machine into failure modes and pick the right kind of task for each one.
• Use the P-F interval and failure history to set an interval.
• Choose between calendar and run-hour triggers.
• Write a task card with real limits.
• Adjust lubrication intervals for real operating conditions.
• Balance PM labor against the hours you really have.

Start From How It Fails

A failure mode is one specific way a part stops doing its job. "Conveyor fails" is not a failure mode. "Drive belt cracks and slips" is. Each mode gets its own decision.

Ask these questions for each one:

1. What is the part supposed to do?
2. How does it fail?
3. What happens when it fails? Think of safety, production, quality and repair cost.
4. Is there a warning before it fails?
5. Does the risk go up with age or use?

Your answers point to one of five kinds of task.

If none of these works and the failure is serious, the answer is a redesign. No PM will save a part that cannot do the job.

Worked example. Here is part of a schedule for a conveyor drive.

Note the last line. A hidden failure gives no sign until the day you need the device. Only a test can find it. Safety devices, backup pumps and alarms all belong here.

Not Everything Wears Out on a Schedule

It is tempting to replace parts at a fixed age. That only works when the risk of failure really rises with age. A well-known study of airline components found that most of them did not follow a simple wear-out curve. That study led to reliability-centered maintenance. Many parts fail at random, or fail early because of installation errors.

This has a practical result. If a bearing fails at random, replacing it every six months does not lower the risk. It adds labor, and it adds the risk of a bad installation. A condition check is the better choice. Fixed replacement suits parts with a clear wear life, such as filters, belts, brake pads and some seals.

Setting an Interval

From the P-F Interval

The P-F interval is the time between the first detectable sign of a problem (P) and failure (F). Your check has to be shorter than this window, or the failure can start and finish between visits. A common rule is to check at about half the P-F interval, so that you get a good chance of seeing it and still have time to plan.

Worked example. Vibration data on your fans shows about 12 weeks between the first bearing tones and failure. Check at no more than 6 weeks. Parts take 2 weeks to arrive. In the worst case, the problem starts right after a visit. You find it at the next visit with 6 weeks left, which still gives you time to order the part and plan the work.

If a failure gives you only a week of warning, a monthly check is useless. Either check more often, use a continuous sensor, or choose a method that sees the problem sooner.

From Failure History

For a part with a clear wear life, use your own records.

Worked example. Your CMMS shows five belts on one drive failed at 1,900, 2,400, 2,100, 2,250 and 1,750 run hours.

1. The average is 10,400 / 5 = 2,080 hours.
2. The shortest life is 1,750 hours.
3. Replacing at the average means about half of them fail first. So set the interval below the shortest life, not at the average.
4. A replacement at 1,500 hours gives a margin of about 15 percent below the shortest life.

Five failures is a small sample, so treat this as a starting point and keep recording.

Calendar or Run Hours

Worked example. A machine runs two shifts, 80 hours a week. A 1,500-hour replacement falls about every 19 weeks. If the plant drops to one shift, a calendar PM set at 19 weeks now fires at 760 hours, about half the life you paid for. A run-hour trigger in the CMMS stays right on its own. It does need someone to log the hours, or an automatic meter.

Writing a Task That Works

A task card is an instruction for a person who has never done this job. It should give the same result no matter who does it.

Weak: "Check gearbox."

Strong:

Every inspection step needs a limit. "Check tension" lets each person decide what is good enough. "Deflection 5/8 inch at 8 pounds of force, adjust if over 3/4 inch" gives you the same answer every time, and a number you can trend.

Add what to do when a limit fails. Small fixes can be made on the spot, within a time limit your site sets. Larger ones need a follow-up work order.

Scheduling Lubrication

Lubrication is the largest set of PM tasks in most plants, so it needs its own care. The chart from the equipment maker is the starting point. Then adjust for how the machine really runs. Two common adjustments are temperature and contamination.

Bearing makers often use a rule of thumb that the regrease interval halves for every 15°C (27°F) of running temperature above 70°C (158°F). Some use a slightly different step, so check your bearing maker's data.

Worked example. A fan bearing has a chart interval of 12 weeks, based on a running temperature of 70°C. A contact reading shows 100°C. That is 30°C over, which is two steps of 15°C. Halve it twice: 12 to 6 to 3 weeks. Before you accept that, ask why the bearing runs so hot. Over-greasing, misalignment or a bad fit can cause it, and those need fixing, not more grease.

Each lube task carries the same details as any other task: the exact lubricant, the amount in strokes, the gun to use, whether the machine must be stopped, and the interval. Group points into a route, and match the stops with planned downtime.

For oil, sample on a schedule and change on the results when you can. A gearbox on a fixed yearly change may be throwing away good oil, or running on bad oil for months.

Balancing the Workload

A schedule that needs more hours than you have will fail quietly. Tasks get pencil-whipped, which means signed off and not done. Or they slide, and your compliance drops.

Worked example. You have two technicians, each with about 160 paid hours a month. Your site expects about 60 percent of that time to be available for planned work, after breakdowns, meetings and walking. That gives 2 x 160 x 0.60 = 192 hours a month.

The weekly row is 12 x 0.5 x 4.33 weeks = 26 hours. The total is 147 hours, which fits. But the averages can hide a trap. If all 45 quarterly tasks land in the same month, that month needs 135 hours for them alone, or 237 in total, and you are over capacity. Spread the quarterly tasks over three months, 15 a month, and the load stays even.

Group tasks by location and by machine stops, so one lockout covers as much work as possible. Ask operators to take on the simple daily checks, such as levels, leaks and cleaning. That saves technician hours and it gets people who know the machine looking at it.

Reviewing the Schedule

A PM program needs regular review.

• Findings rate. A task that has found nothing in 20 visits may be set too often. A failure between PMs means the task is too rare or is the wrong task.
• On-time window. Many sites count a PM as on time if it is done within 10 percent of its interval. A 30-day PM has a 3-day window on each side. Use your site's rule.
• Failures with a PM in place. If a machine fails with a current PM, ask whether the task could have seen this failure. If not, it is not protecting you.
• Changes in use. A new product, speed or shift pattern changes the wear.

Change intervals through your approval process, with the data attached. Do not extend anything that the law, a code or the machine's warranty sets. Inspections required by OSHA, insurers or state boiler rules are not yours to stretch. Safety function tests stay on their schedule, even if they never find a fault.

Common Mistakes

Check Your Understanding

1. A hydraulic pump gives about 4 weeks of warning through rising case drain flow. How often should you check it?

About every 2 weeks, using the rule of half the P-F interval. A monthly check could miss the whole window.

2. Six seals failed at 3,000, 3,400, 2,800, 3,600, 3,200 and 3,000 hours. A coworker wants to replace them at 3,167 hours, the average. What is wrong?

About half would fail before that. Set the interval below the shortest life of 2,800 hours, and keep recording.

3. A standby fire pump has never failed a test in five years. Your manager wants to drop the test. What do you say?

A backup device hides its failure until it is needed. Its test is the only way to find a fault. It also may be required by code or by your insurer, so it stays unless the right authority changes it.

4. A bearing's chart says 8 weeks at 70°C, and it runs at 85°C. What interval does the rule of thumb give, and what else do you do?

85°C is one step of 15°C over, so about 4 weeks. Also find out why it runs hot.

5. You have 180 available hours a month and 170 hours of average PM. Your quarterly tasks all fall in January, April, July and October. Do you have a problem?

Yes. The average fits, but those four months will run over. Spread the quarterly tasks evenly across each quarter.

Key Takeaways

• Build from failure modes. Each mode gets its own task and its own reason.
• Use a condition check when there is a warning, and scheduled replacement only for real wear-out.
• Check at about half the P-F interval, and set replacements below the shortest life, not at the average.
• Use run hours when running causes the wear. Use tests for hidden failures.
• Every inspection step needs a limit and a record.
• Balance the workload by month, and change intervals only with data and approval.