The aim
After a breakdown, the first job is to get the equipment safely back into service. The second is to understand why it failed. Root cause analysis, often shortened to RCA, is a family of structured methods for the second job. The purpose is to find the underlying reasons a failure occurred, so that the fix addresses more than the visible symptom.
Without this step, a plant can end up replacing the same part again and again. A seized bearing is replaced, the machine runs, and a few months later the replacement seizes too. The visible failure was real, but the reason behind it was never touched.
Failure has layers
It helps to separate three levels of cause:
- Physical cause. What broke or wore out, such as a fatigue crack in a shaft.
- Human or process cause. The action or gap that allowed it, such as a bolt tightened without a procedure.
- Latent or system cause. The deeper reason in training, planning, design or management, such as no procedure existing, or no time allowed for a task.
Good investigations usually go beyond the physical level. Stopping at “operator error” is rarely helpful, because it points at a person and not at the conditions that made the error likely.
Step one: protect the evidence
Evidence disappears quickly once repair starts. Before parts are cleaned and discarded:
- make the equipment safe, using the plant’s lockout and tagout style procedures
- photograph the failed parts and the surrounding area
- keep failed components, bagged and labeled, until the investigation is done
- record readings, alarms, control system data and the operating conditions at the time
- speak to the people who were present, soon after, and separately
- note maintenance history and anything recently changed
The facts should be collected before conclusions are drawn. Early guesses tend to shape what people notice and what they ignore.
Simple tools
Several techniques are common, and all are only aids to thinking.
Five whys. Ask why repeatedly, each answer becoming the next question. Here is a hypothetical chain for a failed pump bearing:
- Why did the pump stop? The bearing seized.
- Why did the bearing seize? It ran hot and lost lubrication.
- Why did it lose lubrication? The grease had broken down.
- Why had the grease broken down? It was not replenished on schedule.
- Why was it not replenished? The lubrication task was missing from the routine after a recent equipment change.
Five is a rule of thumb, not a requirement. The danger is that a single chain follows one path and ignores others, and different people may give different answers to the same question. Each answer should be backed by evidence, not assumption.
Cause-and-effect (fishbone) diagrams. The failure sits at the head of a diagram, and branches group possible causes under headings such as people, methods, materials, machines and environment. This helps teams brainstorm broadly before narrowing down.
Fault tree analysis. The failure is placed at the top, and the combinations of events that could cause it are drawn beneath, using logical “and” and “or” connections. It suits complex failures with several contributing factors.
Timeline. Laying out events in order often shows that something changed shortly before the failure, such as a new operating mode or a different batch of material.
Barrier analysis. Asking which defenses should have prevented or caught the failure and why they did not. This is useful for finding that an alarm was ignored or a check had become a formality.
Looking at the part itself
Physical examination can reveal a lot. The pattern of wear on a bearing, the appearance of a broken surface, discolored metal or a leaking seal each give clues. Specialists called failure analysts or metallurgists are sometimes brought in for serious or repeated failures, and a laboratory examination is a reasonable choice when the cost of recurrence is high. For the common case of bearings, reading bearing wear patterns and gearbox failure causes can point towards lubrication, contamination, misalignment or overload.
Data supports the physical evidence. Vibration trends, temperature records and the history of earlier repairs can show whether the problem developed slowly or suddenly.
Keeping the process fair
Investigations work only when people feel safe describing what happened. If the aim is to find someone to blame, information dries up and the same failures recur. Focus on conditions and decisions, not individuals, and keep in mind that most people want to do their work properly. This does not mean ignoring genuine misconduct, but that is a separate matter from learning about failures.
Turning findings into actions
An investigation is only useful if it changes something. Strong corrective actions address the cause directly: fixing a design, adding a task to a routine, changing a procedure or improving training. Weak ones, such as “remind staff to be careful,” tend not to last. For each action, assign an owner and a date, and later check that the problem has actually stopped returning. Findings may also feed back into other reliability tools, including FMEA worksheets.
Deciding which failures to investigate
It is not realistic to investigate every stoppage in depth. Many plants set triggers such as:
- any failure involving injury or a safety near miss
- failures of high-criticality equipment
- repeat failures of the same item
- failures with large cost or production loss
Smaller events can get a quick check, perhaps just a few lines in the work order recording the apparent cause.
Common pitfalls
- Settling on the first plausible explanation.
- Stopping at a physical cause without asking what allowed it.
- Accepting statements without checking them against evidence.
- Writing actions that nobody owns.
- Letting the report sit unread.
Results are rarely perfect: some failures have more than one cause, and a few cannot be fully explained. Recording what is known and what is uncertain is better than inventing a neat story. This is general information, and safety procedures and local regulations should always be followed with trained personnel.