Flexible packaging plant in Nový Bydžov, part of Amcor plc (NYSE: AMCR). Measured over one year.
Cut down your diagnostics time
by 80%.
AI debugging platform for packaging lines. Edmund reads the PLC programs, drawings and OEM manuals of every machine on the line, plus your maintenance logs, and tells your technicians why the line stopped, with the source for every step.
- 01You send the fault
- 02A 20-minute call
- 03The root cause, within 48 hours
Searching „servo drive enable“ in programs
Found the servo control block FB 1343 Faulhaber MCBL. The drive is enabled when it is online, not cancelled, and a motion command (Home / Jog / Positioning) requests it.
#GoHallZerostarts it only while neither#Cancelnor#Stopis set, line 143.- When
#status.enabledis off, it first sets#state.enable.start, lines 144 to 145.
142//Go Hall Zero143IF #GoHallZero AND NOT #Cancel AND NOT #Stop THEN144 IF NOT #status.enabled THEN145 #state.enable.start := true;146 END_IF;147 #state.goHallZero.start := true;148 #status.done := false;149END_IF;
Searching „R1 waits at the pick“ in robot programs
R1 is waiting in plcWaitVacuum for its vacuum switches. It stays at the pick until diVacuumReachedModules12 is on, and diVacuumReachedModules34 too when it picks with both circuits, line 66.
- While it waits,
doRobAtPickis on anddoRobMovingoff, lines 64 to 65. - Once the switches close,
doRobAtPickdrops anddoRobMovingcomes back, lines 67 to 68.
62PROC plcWaitVacuum(bool bBoth)63 ! the switches of the circuits it switched on64 SetDO doRobMoving,0;65 SetDO doRobAtPick,1;66 WaitUntil diVacuumReachedModules12=1 AND (NOT bBoth OR diVacuumReachedModules34=1);67 SetDO doRobAtPick,0;68 SetDO doRobMoving,1;69ENDPROC
Searching „-94B1“ in documents
-94B1 is the product-detect photocell (SICK WL12-2P430) on conveyor 3.
It’s wired to input %I4.3 on DI module -A4, through terminal X1:11–12 in cabinet +SK3.
Querying „energy_by_state“ for IMM01, Jan 12–18, 2026
Pulled IMM01 power from the MES historian for Jan 12–18, 2026 and bucketed it by machine state:
The machine drew power while idle for 74 h, more than double its 31 h of actual production. Trimming idle time is the fastest saving here.
1SELECT state, ROUND(COUNT(*) / 4.0) AS hours2FROM energy_by_state3WHERE machine = 'IMM01'4 AND ts >= '2026-01-12' AND ts < '2026-01-19'5GROUP BY state6ORDER BY hours DESC;
| State | Hours |
|---|---|
| ON_NO_PROD | 74 |
| OFF | 63 |
| PROD | 31 |
Searching „photocell cleaning interval“ in documents
Clean the photocell lenses every 240 operating hours. That is the interval in the conveyor 3 manual, rev 04/2024.
Your maintenance log shows the lens was last cleaned 312 h ago, past the 240 h interval. That lines up with the repeated photocell dropouts.
“Clean the lenses of all photocells in the infeed section every 240 operating hours. Dust on a lens causes dropouts on the product-detect signal.”
- Status
- Indexed
- Machine
- Conveyor 3
- Device
- -94B1 · SICK WL12-2P430
- Event date
- 2026-04-12
Description
Photocell -94B1 dropped out twice in one shift. Cleaned the lens and the reflector; the signal held for the rest of the shift.
Searching „Station 52 photocell“ in maintenance journals
Yes. The Station 52 photocell replacement (Tomáš Marek, Apr 23) sits at the top of the Line 4 journal, indexed alongside every recent fix:
Open any entry for the full symptom, cause and fix. The next technician inherits it instead of rediscovering it.
- Status
- Indexed
- Machine
- Station 52
- Creator
- Tomas Marek
- Event date
- 2026-04-23
Description
The photocell at the Station 52 transfer stopped seeing pallets after a pallet knocked its bracket. Parts piled up at the transfer.
Solution
Replaced the photocell with the same type, aligned it to the reflector and ran 50 test cycles without a dropout.
Reading the dashboard of Assembly line 4
Assembly line 4 is fully indexed and every source is active:
- Spare parts catalog211
- Mechanical drawings128
- Machine & component docs112
- Schematic diagrams96
A plant lead can confirm coverage at a glance and catch a source that’s gone stale before someone hits the gap mid-fault.
- MES databaseSQL · synced 2 h agoActive
- Document storage547 filesActive
- OPC UA serverLine 4 PLCs · liveActive
Trusted by leading global manufacturers

You have a CMMS. You have an MES. You have Copilot.
So why do faults still take hours to fix?
An hour of unplanned downtime at a consumer goods plant, on average (Siemens, The True Cost of Downtime 2024).
We bet your last downtime took more hours than you would have liked.
It takes ten minutes to repair the fault,
but an hour to diagnose it.
On a packaging line, clearing the jam takes minutes. Finding why it happened takes the rest of the hour: the sensor behind the fault code, and the page in the OEM manual that explains it. The answer is in your PLC programs, drawings, alarm history and maintenance logs, and in the heads of your senior people.
Send us your hardest fault.
The one your maintenance team is still trying to solve. We look into it for free: send the fault, your maintenance history or both, then pick a time for a 20-minute call.
Your data stays confidential. NDA available on request.
How we handle personal data
What happens next
- 01You send the fault
The fault in a sentence or two, and the files if you have them at hand.
- 02A 20-minute call
We go through it together and agree on any files we still need. The 48 hours start when we have them.
- 03The root cause, within 48 hours
Why the line stopped and where the cause sits in the machine’s PLC program and OEM manual, with the source for each step. A maintenance history comes back as the faults that cost you the most hours, and where Edmund would help.
Results from manufacturing groups using Edmund
Corrugated and solid board packaging, four Czech locations, part of the Swiss Model Group.
“AI is the only tool that can systematically work with data from multiple sources. Using Edmund has led to a lower frequency of a specific failure, or even its complete elimination.”
Every machine on the line, in one place.
Edmund reads each machine’s PLC program, electrical drawings and OEM manual, plus your maintenance log, and answers from all of them at once.
Reads the program of every machine on the line.
Siemens, Omron and Allen-Bradley projects from each OEM in one searchable model. Edmund answers with the block and the line, so a technician can follow code an OEM wrote.
142//Go Hall Zero 143IF #GoHallZero AND NOT #Cancel AND NOT #Stop THEN 144 IF NOT #status.enabled THEN 145 #state.enable.start := true; 146 END_IF; 147 #state.goHallZero.start := true; 148 #status.done := false; 149END_IF; 150//Position abnullen 151IF #SetPositionNull AND #status.online THEN 152 #state.setPositionNull.start := true; 153END_IF;
Reads the robot program and its I/O, down to the channel.
ABB RAPID modules and the robot controller’s I/O, next to the PLC programs of the same line, from the palletizer to the pick-and-place cell. Each signal comes with its address on the I/O module and the RAPID code that uses it. We are building it with an automotive supplier, on an end-of-line test cell.
Finds the photo-eye behind the fault code.
A tag resolves to the cabinet and the terminal on the filler or the case packer. Finding where a signal is wired takes seconds instead of a search through the OEM’s drawing set.


%I4.3 Checks the guess against stops and rejects.
Stop reasons and reject counts from SQL and MES. Edmund confirms or rules out a cause across shifts and SKUs, and comes back with a next step.
Finds the page in the OEM manual.
Manuals for fillers, labelers, case packers and palletizers, change-part instructions and datasheets, answered with the exact page and paragraph.


“Clean the lenses of all photocells in the infeed section every 240 operating hours. Dust on a lens causes dropouts on the product-detect signal.”
Keeps what one shift learned for the next.
Work orders and handwritten shift logs become searchable. The last fix for the same fault comes back with the answer.
| Name | Status | Creator | Last updated | ||
|---|---|---|---|---|---|
| Photocell stop, conveyor 3 | Processed | M. Novák | 04/25/2026 | ||
| Light barrier lenses cleaned | Processed | T. Marek | 04/12/2026 | ||
| Photocell stop, conveyor 3 | Processed | P. Doležal | 04/08/2026 | ||
| Photocell stop, conveyor 3 | Processed | Night shift | 04/03/2026 |
Shows which machines are covered, and what is out of date.
A live count of what is indexed for each machine, with the time of the last sync. The maintenance lead sees the gaps before someone hits one mid-fault.
What maintenance managers at packaging plants ask first.
Our lines mix machines from several OEMs. Does that matter?
No. Edmund reads each machine’s own program, drawings and manual, so a line built by five OEMs is five sets of sources in one search. Siemens, Omron and Allen-Bradley projects are read the same way.
What can we send?
Either of two things, or both. A fault with the data of that machine: the PLC project or an export of it, the drawings, the manual, the alarm log. And your maintenance history: work orders or downtime records from your CMMS or a spreadsheet. Attach the files to the form, up to 20 MB in total, or send them after the call. Missing pieces are fine: the answer says which sources it used.
Where does our data go?
Into Edmund on AWS, hosted in the EU (Ireland) or in the US, whichever you choose. It is encrypted and kept inside a single-tenant boundary for your company, and the language model runs under zero data retention, so nothing you send is kept for training.
Do we need to connect Edmund to the line?
Not for the first answer. It comes from the files you export. Live data from SQL or MES is connected later, if you want it, and Edmund only reads: nothing in your production systems gets changed.
Who is behind Edmund?
Edmund s.r.o., from Ostrava in the Czech Republic. The company came out of an industrial consulting practice, and Edmund now runs on 130+ machines across four countries.
