Maintenance knowledge search for utilities: describe a fault in plain words, get a cited fix recipe from the manuals, procedures and past jobs
Decades of fixes, searchable before the expert retires.
See one case, screen by screen ↓An alarm raises a corrective work order, and the engineer on call gets an asset number and an alarm text. The answer is out there, but in pieces: the manufacturer's manual is a 300-page PDF, the service bulletin that amends it is filed somewhere else, the safety procedure sits in the document system, and the fix that actually works is a three-line closing note in a work order written by a fitter who has since retired. Failure codes are optional and inconsistent, so no search groups tens of thousands of free-text notes like “seal gone” or “pump died”.
So the engineer follows the manual's fault table, or swaps the most expensive part to be sure. The fault comes back within weeks, a crew goes out again and the same transformer is off automatic voltage control twice. Meanwhile the people who know are leaving: new engineers can take up to three years to work on their own, mainly because rare faults are learned by meeting them beside someone who has seen them.
Estimated split of one maintenance engineer's working days a year spent finding information, by hand and with the solution — about 50 of 220 days by hand, inside the published 20–40 % of time technicians spend consulting maintenance information.
Six specialist agents read the manuals, the procedures and every past job; the engineer decides what to fix, a reliability engineer what becomes standard.
Tuesday 13 October 2026, 06:42. SCADA at Lumina Utilities' Ashgrove Cross 132/33 kV primary substation raises “T2 OLTC — tap change incomplete”, then “Motor protection tripped”. The on-load tap changer on transformer T2 is a Corvane RS-III at 184,212 operations, resting between taps 9 and 10. Omar Haddad, two years at Lumina, is on call and has never seen this fault. The fitter who had, Graham Pell, retired in March 2025. Here is what happens next, screen by screen, in the working solution.
Omar types what he sees: “OLTC stuck between taps, Ashgrove Cross T2, motor breaker tripped”. The open job WO-2026-171902 and the 06:42 alarm are already on the question bar. The run log shows the agents at work: the RS-III manual rev 7 read at §8.4 fault finding, §8.4.3 hand operation and §9.2 drive shaft; 186,400 work orders and 1,240 root-cause reports searched; the safety steps locked to SOP-SUB-031.
Manual and procedure searcher: “found a newer service bulletin — SB-RS3-11 (2019): coupling wear past 150,000 operations, not in the manual's table”
Five agents, 8 steps, 6 documents, 14 past jobs. The answer opens with the symptoms it matched — tap change incomplete alarm, motor protection Q1 tripped, indicator between taps 9 and 10, past 150,000 operations — then six numbered steps, each with its source. Beside it, Manual says · Field says: the manual's table 8-3 points to the motor; the field record says check the coupling first, test the motor, and leave it if it passes.
What was done, and did it hold? Coupling replaced: 0 of 9 came back. Motor replaced: 2 of 3 came back.
Step 4 says inspect the vertical drive-shaft coupling, gearbox end first, for a sheared nylon insert or a missing shear pin — on RS-III drives past 150,000 operations this is the usual cause, which is why it comes before any motor change. Graham Pell's job note from WO-2022-118734, on this same transformer, is pinned to the step. The safety steps, 1 and 6, are quoted from SOP-SUB-031 and the Safety Rules, not rewritten.
Job note · Graham Pell · 18 Feb 2022: “Replace the drive-shaft coupling, not the motor. Motor tests fine every time, it's the coupling insert shearing. Check the gearbox end first.”
The work history miner read every closing note on this tap changer and symptom since 2019 and grouped them by the fix actually done: coupling replaced 9 times, 0 repeats in 90 days; motor replaced 3 times, 2 repeats in 30 days. At Oxley Road T1 the new motor was back in 9 days, at Kelbridge T1 in 11 — both times the next job found the coupling insert sheared. At Brayford T1 the motor winding really was open circuit, so step 3 still tests the motor.
WO-2025-166017, Hessle Gate T1, after Graham retired: “Took two calls to find someone who knew. Coupling insert. Replaced the kit.”
The parts list shows coupling insert kit 40-118-207 at $110, 4 in stock at Northgate depot, bin D-14-03, and shear pin 40-118-209, 12 in stock. The drive motor 40-118-150 at $6,200 is marked not needed. Omar reserves the kit against WO-2026-171902; three past jobs where this fix worked sit underneath — Ashgrove Cross T2, Fenmoor T1 and Asterdale T3, each done in about two hours.
Parts and tools checker: reserves only when the engineer asks, and never orders.
Kofi Mensah issued permit PTW-ACX-2026-0412 at 07:36: T2 motor drive cabinet, Q1 open, locked, notice fitted. Omar works through the recipe as a checklist. His motor readings come in from insulation tester IT-0447 — 210 MΩ and 4.3 Ω — and each is checked against the manual's range as it lands. Motor good, so the motor stays.
“✓ within the manual's range” · insulation resistance > 20 MΩ, winding 4.1–4.6 Ω
Coupling insert sheared at the bevel gearbox end; insert and shear pin replaced, grub screws at 12 N·m, step time 5.1 s. T2 is back on REMOTE and automatic voltage control at 09:52, 2 h 12 after the job started. When Omar closes the job, the knowledge steward drafts his closing note from his readings and finding and links it to WO-2022-118734 on this unit. The note is his to edit and sign.
Drafted note: “Second coupling failure on this unit in 4 years and 8 months: suggest a coupling check at each tap changer maintenance on RS-III units past 150,000 operations.”
Omar signs and the note goes to Field fixes, waiting for Dr. Ines Varga. Beside it, the knowledge steward suggests adding this job to the RS-III field fix as its 10th coupling job. Ines can publish the note or ask Omar a question first. She publishes.
“Nothing changes in an answer until a reliability engineer approves it.”
The proposal puts the answer today, which follows the manual's table, beside the proposed recipe, with the evidence: at least 3 jobs with the same fix — 9; none back within 90 days — 0 of 9; consistent with bulletin SB-RS3-11; safety steps unchanged, quoted from SOP-SUB-031. It changes the answer for 12 assets. Ines approves it as field-proven; she could instead ask the manufacturer to confirm or reject it with a reason.
“Approved by Dr. Ines Varga at 14:20. Answers now lead with this fix.”
Doreen retires in 66 days after 34 years on dosing, filters and pumps; 41 % of her 1,412 jobs have a usable note. The coverage map shows where her knowledge is written down and where it is not — filters at 29 %. The knowledge steward drafted questions from the gaps for her next session on Thu 22 Oct, such as what “as before” means on 9 high-lift pump 2 vibration jobs.
Session 1 · 8 Oct · 47 minutes: 5 notes drafted from the recording; Doreen has signed 3.
The same tap changer, seen by the five people around it — what their day looked like, and what it looks like now.
Six specialist agents read the manuals, the procedures and every past job; the engineer decides what to fix, a reliability engineer what becomes standard.
Reads the engineer's question, the open work order and the latest alarms; finds the asset, the site, the equipment model and serial, and lists the symptoms in plain words.
Searches manufacturer manuals, service bulletins, the utility's procedures, safety rules and standards for this model and fault, and returns each passage with document, revision, section and page.
Finds corrective jobs on the same model with the same symptoms, reads every closing note, groups them by the fix actually done in the technician's words, and links repeats on the same asset to count which fixes held.
Writes the numbered fix recipe — symptoms, steps, tools and parts — and puts the field fix beside the manufacturer's procedure with the outcomes of every past fix.
Maps the parts to stock codes, depot stock and bins from a stores feed refreshed every 15 minutes, and lists the tools and test equipment for the job.
Drafts the closing note from the engineer's readings and finding, proposes a field-proven fix to a reliability engineer when the evidence meets the rule, and drafts questions for experts before they retire.
Approves what becomes standard. The agents propose; a named person decides.
Ask a follow-up in plain words, or ask for something to be done. Every answer names its sources.
Why does the recipe say not to replace the motor?
Because on the RS-III the motor is rarely the fault. Of 14 jobs with this symptom since 2019, the motor was replaced 3 times and the fault came back within 30 days on 2 of them. The coupling was replaced 9 times and none came back within 90 days. Bulletin SB-RS3-11 says coupling inserts wear past 150,000 operations; T2 is at 184,212. The recipe still tests the motor: at Brayford T1 in 2022 the winding really was open circuit.
Which other RS-III units are past 150,000 operations?
Twelve of Lumina's 38 RS-III units, including T2: Ashgrove Cross T2 at 184,212, Fenmoor T2 at 176,940, Hessle Gate T3 at 171,300, Oxley Road T1 at 168,450, and 8 more from 151,200 upwards. Three have had the coupling replaced since 2022. The field fix waiting for Dr. Varga would put the coupling check first for all twelve.
Reserve the coupling kit for this job
Reserved against WO-2026-171902 at Northgate depot: 1 × coupling insert kit 40-118-207, bin D-14-03, 3 left; 1 × shear pin 40-118-209, bin D-14-04, 11 left. Nothing has been ordered: stores re-orders the kit when stock reaches 2.
What did Graham Pell write about the RS-III?
Graham closed 31 jobs on RS-III drives between 2003 and 2025, and his notes are short and exact. On this fault he wrote the same thing four times: the coupling, not the motor. After he retired, the next job took “two calls to find someone who knew”. His note now sits in step 4 of the recipe.
13 screens from the working solution, on its sample data. Pick one to see it large.
Which sources are trusted first, what counts as a field-proven fix and who decides are settings, chosen by reliability engineering.
| Setting | Default | Choose from |
|---|---|---|
| Sources, in the order they are trusted | Safety rules and procedures first — locked | then manufacturer manuals and bulletins, standards, work orders with a closing note, root-cause reports; bulletins outrank the manual they amend |
| Jobs with the same fix before a field fix can be proposed | 3 | Before a field fix can be proposed |
| Repeats allowed, and the window to hold | 0 within 90 days | Jobs that came back inside the window, on the same asset |
| A repeat is | 30 days | The same asset and failure within this window |
| Who approves field fixes | Reliability engineering | Reliability engineers · Dr. Ines Varga, Marcus Feld |
| Ask the manufacturer | On | On · Off — drafts a message when a field fix contradicts the manual |
| Confidence threshold | 0.70 | Below it the answer says so and names who to call |
| Stores feed | On | Stock and bins refreshed every 15 minutes · reserve only, never order |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the pilot on the customer's maintenance system · “estimated” = our arithmetic on the demo utility's inputs, replaced by the customer's: 31,000 corrective jobs a year × 4 points fewer repeats = 1,240 jobs × $950 a repeat job (a crew of two for four hours, a vehicle and parts); the days split is our estimate · “typical” = published figures: T&D World, utility workforce crisis (33–50 % of the workforce eligible to retire in 5–10 years); AWWA 2025 (21 % of water staff within 5 years); Avers et al. via arXiv 2608.18465 (20–40 % of technicians' time consulting maintenance information). IEC 60214-1 and ISO 14224 are named as context. Lumina Utilities and the people, sites, equipment, documents and records named on this page are fictional — characters and sample data in the working solution.
Maintenance knowledge search lets maintenance engineers ask about a fault in plain words and get an answer drawn from OEM manuals, maintenance procedures, standards and past work orders. Maintenance Knowledge Search answers with a fix recipe: matched symptoms, numbered steps each cited to its source, tools, and parts with stock codes — and the fix that stopped the fault coming back.
Six agents share the work. The fault reader finds the asset and symptoms; the manual and procedure searcher finds the manual sections, bulletins and procedures; the work history miner reads past closing notes and counts which fixes held; the fix recipe writer builds the cited recipe; the parts and tools checker finds stock; the knowledge steward drafts the closing note. In the sample case the answer took 9.4 seconds.
Yes. It reads the technicians' own closing notes, not only failure codes, groups them by the fix actually done and links repeats on the same asset. A job without a closing note is never counted as a success.
The answer shows both side by side, with the outcome of every past fix. When a fix meets the field-proven rule — set in the demo to at least 3 jobs, none back within 90 days — it goes to a reliability engineer, who approves it as field-proven, rejects it with a reason or asks the manufacturer to confirm. Nothing changes in an answer until a reliability engineer approves it.
Yes. Expert capture lists experts retiring in the next 18 months and shows how much of their work has a usable note. The knowledge steward drafts questions from the gaps in their record, and each session becomes notes the expert checks and signs.
Yes. The working solution also answers a hypochlorite dosing pump losing prime at a water treatment works, and an SF6 low-density alarm on a 33 kV circuit breaker. Questions and answers can be in English and Arabic.
Yes. Safety steps are quoted word for word from the procedure and safety rules, and only the procedure owner can change one. The engineer decides what to replace and signs the closing note; the authorised person issues the permit; a reliability engineer decides what becomes standard. The agents never order parts, change a work order's status or publish a fix.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your manuals, procedures, safety rules and work order history — and it goes live once every quality gate has passed. We will walk you through it on your own closed corrective jobs first.
Defects caught while they grow: every structure compared with last year's flight.
1-min demoAsset Health IndexThe same capital budget removes 2.5× the risk, every score traced to its evidence.
1-min demoTransformer DGA InterpretationEvery oil result trended the day it lands, with advice set by how fast the gas is rising.
1-min demoWe'll run Maintenance Knowledge Search on your manuals, procedures and a sample of your closed corrective jobs.