Why Robot Faults Get Fixed Twice: A Diagnosis and a Remedy

Symptom: a gripper on one line stops releasing parts cleanly, a technician spends a shift tuning it, and two weeks later an identical fault appears on a sister line, where another technician starts from nothing. The same defect gets diagnosed and fixed twice, at double the cost. This article diagnoses why robotics operations teams repeat their own repairs, and describes a remedy based on keeping work visible: shared fault notes, open configuration changes and drafts that others can read before they are final.

The presenting problem

In a plant or warehouse with several robots, work tends to fragment. Each shift, each cell and each technician accumulates knowledge that never leaves its origin. Typical observations include the following.

  • A parameter was adjusted on one machine and never mentioned to anyone else, so a later replacement unit behaves differently.
  • A workaround is passed on by word of mouth and slowly distorts with each retelling.
  • An engineer begins a software update on a controller without knowing that a colleague has already started one.
  • A recurring nuisance stop is treated as a one-off each time, because nobody can see how often it happens.

Diagnosis: the work is done where nobody can see it

Skill is seldom to blame. What goes wrong is that the working state of the operation lives in private places: personal notebooks, chat threads, memory, and the heads of the most experienced staff. Incidents get closed, but their reasoning does not get recorded anywhere shared. The organisation therefore cannot learn faster than its most talkative member can talk.

There is a second, subtler cause. Teams often delay writing things down until they feel finished, so the record only ever contains tidy conclusions. The messy middle, where most of the learning occurs, disappears.

Remedy 1: a shared fault log, written during the work

Create one log that every technician and engineer can read, and make a rule that an entry is opened when a fault investigation begins, not when it ends. Each entry needs only a few lines: the machine or cell, the symptom, what has been checked, and the current guess. Update it as the work proceeds. A colleague facing a similar symptom elsewhere can find it within minutes, and may add a note that closes the question.

For teams working with sensing hardware, the background in our article on how sensing devices are built into industrial robot systems gives a useful vocabulary for writing these entries consistently.

Remedy 2: configuration changes proposed in the open

Parameters, program versions and calibration values should change through a visible proposal, not a quiet edit. Even a simple shared page listing “proposed change, reason, who will do it, when” prevents collisions. Reviewers can notice that a value conflicts with the settings of the neighbouring cell, or that the same change is already planned. Where the organisation uses version control for robot programs, drafts should be opened as soon as work begins.

Remedy 3: fleet-level notes for mobile equipment

Automated guided vehicles and similar fleets are especially prone to hidden fixes, because their faults depend on routes, floors and battery habits. Keep a shared notes page for the fleet where operators record oddities as they appear. Our guide to planning and running warehouse automation with guided vehicles describes the operating context in which such notes become most valuable.

Remedy 4: record the dead ends

Add a section to the fault log headed “tried, did not help”. It prevents the next person from repeating an expensive test, and it normalises the idea that investigations include failures. Over time this section becomes one of the most heavily consulted parts of the record.

A comparison from outside the factory

Visible information changes behaviour in everyday life as well. A local-shopping website, BuyMLocal, lays out a straightforward idea: visitors search for something they need locally and read what nearby shops have put up as in stock. The shopper decides with something to read; the technician in a plant should have the same advantage when deciding how to handle a fault.

What a good entry looks like

Consider a short example. Entry opened: “Packing cell 3, intermittent stop at the pick position, roughly once per shift.” Checked: gripper pressure within range, vision camera lens clean, nothing logged by the controller at the moment of the stop. Current guess: reflections from a new overhead light affecting part detection. A technician on another line reads this, remembers the same light fitting being replaced there, and adds a note. Within the hour the investigation has two sources of evidence instead of one, and nobody has attended a meeting.

Notice that the entry is not polished, and that the guess may prove wrong. That is acceptable. The value lies in making the thinking available while it can still be corrected.

Safety and sensitive information

Working in the open does not override safety procedure. Anything that affects machine safety, such as guarding, interlocks and emergency stops, remains subject to your site’s formal change control and sign-off. The shared log is a communication aid, not an approval path. Likewise, keep out of shared notes any commercially sensitive details, customer information or security credentials for control systems. If a detail feels borderline, check with whoever owns plant security before it goes on the page.

Implementation in two weeks

  1. Days 1-2: choose a single location for the fault log and add the investigations currently under way.
  2. Days 3-5: agree the rule that entries open at the start of an investigation, and brief each shift.
  3. Days 6-8: add the “tried, did not help” section and seed it with known dead ends.
  4. Days 9-12: introduce the visible proposal page for configuration changes.
  5. Days 13-14: review the entries together and note any fault that appeared in more than one place.

Verifying that the remedy works

Select two measures that are easy to collect and hard to argue with: the number of faults investigated more than once, and the time between a symptom first appearing and a colleague finding a relevant earlier entry. Compare a month before and a month after, noting any other changes that could explain the difference. A modest improvement is a success; a sudden dramatic one deserves a second look. If the log is not being read, the fix is usually a change of habit, such as consulting it at shift handover, not a new tool.

Related Posts

Eight Things a Finished Robot Cell Teaches You About Operations

A technician reads a finished pick-and-place cell against its commissioning binder. Eight structural lessons for anyone running robotics operations.

Read More

Robot Cell Fault Messages: Vendor Style Against Plain Words

The same robot fault written twice: once as controller shorthand and once for the operator at the screen. A practical side-by-side.

Read More

Robotics Operations on a Short Parts List: A Beginner Primer

A precise beginner primer on why robot builders and operators benefit from limits on sensors, parts and behaviours, with a glossary of the terms you will meet.

Read More

Commissioning a Robot Fleet Slowly, Step by Step

A warehouse team learned that bringing robots up at walking speed, one check at a time, finished the job sooner than a fast start. Here are the steps.

Read More

Robotics Operations and the Done List: Practical Answers

What belongs on a done list for robots, how it differs from the maintenance schedule, and who writes it? Direct answers for teams running automated equipment.

Read More

Fallback Skills That Robotics Operations Employers Look For

Robotics operations rewards people who know what happens when a step fails. Seven fallback skills to build, and how to turn them into a career plan of your own.

Read More

Leave a Reply