Reliability
Reliability engineering for plant teams: where to start without a reliability engineer
Reliability engineering is the work of making machines fail less often and recover faster, using your own failure history to decide what to fix. You do not need a reliability engineer to start. Log every breakdown properly, rank the few machines that cause most of your downtime, find and remove the cause on one of them, and check MTBF a few months later. Repeat that every week and you are doing reliability engineering.
What reliability engineering means for a plant team
In a large company, reliability engineering can mean statistics, life data models and failure mode studies run by a specialist. In a typical Indian plant with one maintenance manager, a few supervisors and a team of technicians, it means something simpler: stop fixing the same failure twice.
Maintenance keeps machines running today. Reliability asks why the machine stopped and what would stop it from stopping again. The tools are the same people and the same records. The difference is that someone looks back at the history every week and acts on it.
Two numbers tell you whether it is working. Mean time between failures (MTBF) is how long a machine runs, on average, before it breaks down. Mean time to repair (MTTR) is how long each breakdown keeps it down. Reliability work pushes MTBF up. Better spares, tools and skills push MTTR down. If you want the formulas and a worked example, see MTBF vs MTTR.
You do not need special software, a statistics course or a new hire to begin. You need clean failure records and one hour a week of honest attention.
Step 1: Get failure data you can trust
Everything later depends on this. If breakdowns are told on the phone and fixed without a record, you have nothing to analyse. Agree on a minimum record for every stop, and make it quick enough that technicians will actually fill it in.
| Field | Why it matters | Example |
|---|---|---|
| Machine (from a fixed list) | Free text like "press 2" and "HP2" splits one machine into many | HP-02 Hydraulic press 250 T |
| Stop time and restart time | Gives downtime and MTTR without guessing | 10:40 to 13:15 |
| Symptom | What the operator saw | Ram slow on return stroke |
| Failure code | Lets you count failures by type | HYD: oil overheating |
| Cause found | What the technician found, not what they did | Return filter choked |
| Action and parts | Shows repeat parts and cost | Filter changed, 1 element |
| Closed by | Someone owns the record | Technician name |
Keep the failure code list short. Ten to fifteen codes per machine type is enough to start. A list of eighty codes gets ignored, and everything ends up as "other".
Step 2: Find your bad actors
Once you have two or three months of clean records, add up downtime hours by machine and sort from highest to lowest. This is a Pareto chart. In most plants a handful of machines sit at the top and account for a large share of the lost hours. These are your bad actors.
Rank by downtime hours, not by the number of stops. A machine that stops for ten minutes thirty times hurts less than one that stops twice for a whole shift. Then look at criticality. A bad actor on a bottleneck line matters more than one with a standby machine next to it.
Pick one machine. Not five. The point is to finish one fix and prove it worked, so the team believes the loop is worth the effort.
A worked example from a typical machining plant
Worked example (illustrative)
Take an auto component machining plant near Pune with 40 machines on two shifts. The maintenance manager has six months of breakdown records with machine, times and failure codes. All numbers below are illustrative.
Sorting downtime by machine gives this list. The plant logged 212 stops and 400 hours of downtime in six months.
| Machine | Stops in 6 months | Downtime hours | Share of downtime |
|---|---|---|---|
| HP-02 Hydraulic press | 31 | 96 | 24% |
| VMC-07 CNC machining centre | 22 | 58 | 15% |
| IH-01 Induction hardening | 9 | 41 | 10% |
| C-02 Air compressor | 6 | 18 | 5% |
| Other 36 machines | 144 | 187 | 47% |
What the list shows. Three machines, under a tenth of the fleet, cause almost half the downtime. The team picks HP-02, the top bad actor.
Worked example, continued (illustrative)
HP-02 had about 2,496 scheduled hours in six months (16 hours a day, 26 days a month). Take off 96 hours of downtime and it ran 2,400 hours. MTBF is 2,400 ÷ 31, about 77 hours. MTTR is 96 ÷ 31, about 3.1 hours.
The failure codes show 19 of the 31 stops were hydraulic: slow ram, oil overheating, pressure loss. The technicians had been topping up oil and cleaning the cooler each time. A short root cause session found the return line filter choking well before its 2,000 hour change, and a fouled oil cooler. Nobody was checking the filter's pressure drop or the oil temperature between changes.
The fix was small: a filter pressure drop and oil temperature reading on the daily operator checklist, a monthly cooler cleaning PM, and filter change on condition instead of only on hours.
What changed (illustrative). In the next three months HP-02 had about 1,248 scheduled hours, 6 stops and 14 hours of downtime. It ran 1,234 hours, so MTBF rose to about 206 hours (1,234 ÷ 6) from 77. MTTR fell to about 2.3 hours. The team then moved on to VMC-07.
Step 3: Fix the cause, not the symptom
A replaced part is an action, not a cause. Ask why the part failed, and keep asking until you reach something you can change: a missing check, a wrong lubricant, a loose mounting, a setting operators change between shifts. The five whys is enough for most shop floor failures.
The fix usually lands in one of four places: a new or changed PM task, a check added to the operator's daily list, a design or spare part change, or training. For a critical machine with many failure modes, a simple failure modes and effects analysis (FMEA, described in IEC 60812) helps you go through each part in turn. Start with the five whys and keep FMEA for the machines that matter most.
Then wait. Give the fix two or three months and compare MTBF before and after. If it did not move, the cause was wrong. Go back to the records.
A weekly reliability routine you can copy
Reliability needs a habit more than a department. One hour a week, the same day and time, with the maintenance manager, a production supervisor and the planner.
| Step | What you do | Time |
|---|---|---|
| Check records | Look for breakdowns with no failure code or no cause, and complete them | 10 min |
| Update the bad-actor list | Downtime hours by machine for the last 90 days | 10 min |
| Review the open fix | Is the action on the current bad actor done? Who is blocked? | 15 min |
| Check the last fix | MTBF and stops on the machine you fixed earlier | 10 min |
| Pick the next action | One machine, one cause, one owner, one date | 10 min |
| Write it down | Actions as work orders or PM changes, not meeting notes | 5 min |
Common mistakes
- Starting with tools before data. FMEA and life data analysis on bad records give confident wrong answers.
- Ranking by the loudest failure. The breakdown that stopped dispatch yesterday is not always the biggest loss. Use downtime hours.
- Fixing five machines at once. Effort spreads thin and nothing finishes. One at a time builds proof.
- Recording the action as the cause. "Bearing replaced" tells you nothing about why the bearing failed.
- More PM as the answer to everything. Some failures are caused by intrusive maintenance. Change the task, not just the frequency.
- Not checking whether the fix worked. Without a before and after MTBF, you are guessing.
- Leaving production out. Operators see the first signs of most failures. Their daily checks are your cheapest sensor.
Sources
- Reliability engineering (opens in a new tab), Wikipedia
- Assessing product reliability (Engineering Statistics Handbook, chapter 8) (opens in a new tab), NIST/SEMATECH
- IEC 60812:2018 Failure modes and effects analysis (FMEA and FMECA) (opens in a new tab), IEC
- Pareto principle (opens in a new tab), Wikipedia
Frequently asked questions
What is reliability engineering in maintenance?
It is the work of making machines fail less often and recover faster, using failure history to find and remove the causes of repeat breakdowns. It focuses on why a machine stopped, not only on getting it running again.
Do I need a reliability engineer to start?
No. A maintenance manager with clean breakdown records and one hour a week can rank bad actors, run a root cause on one machine and check whether MTBF improves. A specialist helps later with deeper analysis.
What is a bad actor in maintenance?
A bad actor is a machine that causes a large share of downtime or cost. You find them by adding up downtime hours by machine over a period and sorting from highest to lowest.
Which metrics show reliability is improving?
MTBF going up on the machines you worked on is the main signal. Watch MTTR, total downtime hours and the share of reactive work alongside it.
How long before reliability work shows results?
Give each fix two to three months of running before you compare MTBF before and after. Shorter periods often have too few failures to tell.
Is reliability engineering the same as predictive maintenance?
No. Predictive maintenance is one tool that watches condition to catch failures early. Reliability engineering decides which failures matter and which tool, check or design change removes them.