SystemManufacturing Intelligence

Equipment-Related Deviation Analysis

Equipment-related deviation analysis that places each deviation on its asset and part, lays its work orders, calibrations and alarms on one timeline, and carries the fix to every machine with the same part.

See one seal's failures traced screen by screen ↓
Watch with transcript →
The problem

Each equipment-related deviation closes as wear, and the same part fails again

Maintenance and quality keep two histories of the same machine. The maintenance system holds the work orders and calibrations; the quality system holds deviations whose category says equipment and whose equipment field often names the line, not the machine. The investigator emails maintenance for recent work orders, asks automation for trend screenshots and writes the timeline by hand.

So the investigation stops at the part. It closes as seal wear, seal replaced, and the preventive maintenance plan stays on the calendar it was written with, because changing it needs a change control nobody has the evidence for. The same part on the next machine is left until it fails too, and quarterly trending shows equipment as one bar.

Built forReliability engineersMaintenance managersQA equipment and validation leadsCalibration coordinators

The Problem Today

Maintenance and quality keep two histories of one machine, so each deviation closes as wear and the part fails again.

164
FDA FY2025 drug citations of 211.192 for weak investigations published

Two histories of one machine

  • QMS record
  • CMMS notes
  • Cal certs
  • Historian
  • A calendar plan for hourly wearThe seal is due at an overhaul it never reaches
  • Sister machines left uncheckedNobody looks for the same part on other machines

The Solution

  1. Each deviation on its partFrom the narrative and the equipment log down to the component
  2. One timeline, in run-hoursWork orders, calibrations, alarms and deviations lined up by wear
  3. Every machine with the partSame part number across plants, near-miss numbers struck
  4. Signed into change controlMaintenance approves, QA countersigns, the plan change joins the CAPA

Built-in controls

  • Agents read the QMS and CMMS and write to neither
  • Maker, approver and QA countersign are three people
  • A signature counts only once its evidence is opened

What You Get

Maintenance and quality work from one history of each machine, and a finding on one part reaches each machine that carries it.

  • One fix on every sister machineSister machines change the part on run-hours, below the wear seen
  • Work orders as inspection evidenceEach technician's note links to what happened next
  • History that outlasts its engineerJoined records and approved rules stay when people leave
Features

Equipment-related deviation analysis, followed through one granulator seal

Reliability engineers, maintenance managers, calibration and QA use these screens to find the part behind repeat deviations and change its plan on every machine that carries it. They follow one high-shear granulator's impeller shaft seal.

demoThe plants, machines, people, records and run-hours in this walk are sample data.

  1. 01Asset tree

    Bad actor analysis on the asset tree, weighted by joined quality losses

    • Branches as thick as their quality losses. Deviations, held batches and unplanned stops weight each branch, so the bad actors stand out before anyone draws a pie chart.
    • Placed down to the component. Each deviation lands on a machine, an assembly and a part, instead of the line someone typed as free text.
  2. 02One timeline

    Maintenance and quality history of one part, laid on one timeline

    • Five lanes from four systems. Work orders, calibrations, historian episodes, changes and deviations sit together, so the investigator sees which came first without emailing maintenance.
    • Agents show what they read. The plan, sources and confidence of each agent stay open beside the timeline, and each mark opens its record.
  3. 03Work order

    Equipment failure analysis from the technician's own note, as written

    • The note read as written. Words such as weeping and purge raised, which no category field captures, become evidence the reliability engineer can cite.
    • Run-hours on the record. Each work order carries the wear counter at the time, so the note can be placed against the part's life.
  4. 04Calibration

    GMP equipment calibration checked before an instrument is trusted or blamed

    • As found against tolerance, point by point. The calibration coordinator rules the instrument in or out from its own certificate before the investigation leans on its readings.
    • Failed calibrations start an impact check. An out-of-tolerance result lists the batches since the last good check, and the coordinator sends QA a structured impact request.
  5. 05Run-hours

    Wear compared in run-hours, so two seals that failed apart line up

    • Calendar or run-hours at one click. Seals that failed after different months in service line up once the impeller counter is used, which is how the part wears.
    • The warning before each deviation. Each purge-raise work order sits weeks before a major deviation, so the stop-gap shows as the warning it was.
  6. 06Pattern

    Failure pattern analysis at part level, challenged as an inspector would

    • Each claim opens its evidence. Seal lives, weeping work orders and calibrations behind the pattern each open the record they came from.
    • A wrong draft line struck. The challenger strikes a draft line on seal life that the records do not support, before the pattern reaches a signer.
    • A second cause kept apart. An alarm limit widened by a software update is kept as a separate cause for the process engineer, so the seal is not blamed for everything.
    • Batch look-back asked of QA. The batches made while the seal wept go to QA, who decides whether they need an impact assessment.
  7. 07Interval rehearsal

    Preventive maintenance interval rehearsed against the wear already seen

    • Both plans played forward. The reliability engineer sees when each plan would change the part and whether it would pass the wear at which the last one failed.
    • Assumptions open to change. Forecast run-hours a month and the trigger are editable, and a trigger at or above the earliest failure is refused.
  8. 08Same part elsewhere

    Every machine carrying the same seal kit, found across plants

    • Exact part number only. A similar number on another machine is struck with its reason, so a seal change is not planned on the wrong machine.
    • Margin to the trigger for each. Run-hours now, the monthly rate and the date each reaches the trigger tell the planner which machine needs a work order first.
  9. 09Change control

    Maintenance plan change through GMP change control, tied to the CAPA

    • Maker, approver and countersign. The reliability engineer submits, the maintenance manager approves, and the QA equipment lead countersigns with a requalification decision and the client notices.
    • Evidence opened before any signature. Each signer opens the evidence pack in person; the signature then records its meaning and session.
    • One finding, four plans. The plan change reaches each granulator with the part, and it starts only after the CAPA review board approves the plan.

The calibration coordinator rules the gauge in before anyone blames the seal

Maintenance and quality each held half of the seal's story; these four read the same joined history before any of them signs.

Reliability engineers

Read failure history and propose changes to maintenance plans

  • Each part's joined history, with sources
  • Patterns QA can follow and challenge
  • Every machine with the same part
Maintenance managers

Own the preventive maintenance program and approve work orders and plan changes

  • Plans on run-hours where wear demands it
  • Evidence opened before approving
  • Work orders drafted, never filed alone
QA equipment and validation leads

Countersign GMP plan changes and decide requalification

  • One evidence pack for each change
  • Rules from patterns wait for QA
  • Client notice decided at countersign
Calibration coordinators

Run the calibration program and start impact checks on failed results

  • Certificates on the part's timeline
  • Batches since the last good check
  • Impact requests logged for QA

QA's countersignature carries the requalification call and the client notices

Reliability engineer
Sees
The pattern, the challenger points and the rehearsal
Signs
The proposal, as maker; never its approval
Maintenance manager
Sees
The evidence pack, opened in person
Signs
The work order, and the maintenance plan change
QA equipment and validation lead
Sees
The same evidence, the requalification question and the client notices
Signs
The countersignature, learned rules and any settings change
Calibration coordinator
Sees
Out-of-tolerance results and the batches since the last good check
Signs
The impact request to QA, who decides batch impact
More screens

The agents left 59 deviations unplaced rather than guess the machine

The screens around the seal: what waits for signature, what ranks highest in every plant, what the agents would not place, and the rules behind it all.

Drafts waiting for the maintenance managerThe work order and change control as the agents drafted them, the sign-off chain with the maker's signature, and earlier changes this solution produced.
Bad actors ranked on one score in each plantParts ranked by severity points, held batches and stop hours under the same rules in all four plants, with where each one stands.
Deviations the agents would not place59 deviations whose equipment tag was missing or matched two machines, each with its candidates and a due date; bulk placement asks for sampled records first.
A learned rule waiting for QAA rule from the confirmed pattern flags any weeping or purge raise on the same seal kit and tells the QA equipment lead; it applies once QA approves it.
Signatures, reasons and sessions in the trailWho signed, which records they opened first and for how long, the reason and what each signature means, exported as CSV or a JSON evidence pack.

demoSample data, from the working solution.

Two seals failed 7 run-hours apart, after 17 and 15.3 months in service

What three investigations on one granulator had closed past, and a near-miss on the sister list.

Two seal lives that agree

The seals lasted 17 and 15.3 months, which looks like chance, but failed at 417 and 410 run-hours.

A purge raise before each deviation

Each weeping was met by raising the purge from 0.5 to 0.8 bar, and a major deviation followed 3 to 4 weeks later, 3 of 3.

An alarm limit widened outside validation

A software update widened the torque alarm to 380 N·m, so the last batch reached 361 N·m with no alarm.

A part number that only looked the same

The first list of sister machines held a homogenizer with a different seal kit; the challenger struck it before any work order.

demoSample data, from the working solution.

The history joiner ties each record to the asset, never to a date alone

Results pass by name from agent to agent, and the drafts end with the maintenance manager.

Hierarchy mapper

Places each equipment-related deviation on its asset and component from the equipment field, the narrative and the equipment log.

  • Guesses no asset when the tag is missing or matches two machines
History joiner

Lays work orders, calibrations, historian episodes, changes and deviations for one part on one timeline, with run-hours.

  • Edits no source record
  • Joins nothing by date alone without the asset
Failure pattern reader

Finds a repeat failure at component level: what came first, at what wear, with sources and confidence.

  • Calls no pattern below the minimum number of events
  • Names no root cause for an open investigation
Inspector challenger

Asks what an inspector would ask and answers each point from the records, hands it to a person, or corrects the agents' own draft.

  • Closes no point a person must answer
Interval rehearser

Plays the next two years forward under each maintenance plan, with editable assumptions.

  • Presents no forecast as a measured result
Sister-asset finder

Finds every asset with the same part number, with its run-hours and its margin to the trigger.

  • Matches no similar part number
  • Raises no work order itself
Change drafter

Drafts the work order and change control with citations and links both to the CAPA.

  • Submits, approves and edits no plan in the CMMS

What QA and maintenance ask about equipment-related deviation analysis

What is an equipment-related deviation?

An equipment-related deviation is one the quality system classes as caused by a machine, an instrument or its upkeep. Here each one is placed on its asset and part and joined to that part's work orders, calibrations, alarms and stops, so a repeat failure shows across deviations instead of inside one.

Is this a CMMS or an asset performance tool?

No. Those tools hold work orders, health scores and failure analysis for maintenance. This solution joins that history to the deviations, held batches and CAPAs in the quality system at component level, and sends any plan change through GMP change control with a QA countersignature.

Can the agents change a maintenance plan?

No. They draft both the work order and change control and link them to the CAPA. The reliability engineer submits, the maintenance manager approves and the QA equipment lead countersigns; no agent writes to the CMMS or the QMS.

Does it name the root cause of an open investigation?

No. The failure pattern is evidence for the investigator. The root cause of an open investigation is approved by people in the quality system, and a pattern needs the minimum number of events QA has set.

What happens to batches made while a part was failing?

The reliability engineer can ask QA for a look-back on the batches made between the warning and the deviation. QA decides whether they need an impact assessment; the plant's head of quality dispositions, and the client's own person releases to market.

What is bad actor analysis in maintenance?

Bad actor analysis ranks the assets that cause the most trouble. Here each branch of the asset tree is weighted by the deviations, held batches and unplanned stops joined to it, down to the component, so the parts behind repeat quality losses stand out first.

Bring one part that
keeps failing.

Send the deviations, work orders and calibration certificates for one machine, and we will place them on its asset tree and line them up in run-hours.