Skip to content
Service & spare partsPumps & ValvesWater & EnvironmentMachinery & Plant EngineeringConveying & Intralogistics+3

Classify damage from a photo

A photo of an impeller, belt or sealing face: AI recognises the type of damage, compares it with earlier cases and proposes the action.

Today

Whether an impeller shows cavitation, a conveyor belt longitudinal cracks or a sealing face scoring is often judged by a single expert, once they find the photo in their inbox. Until then the machine sits dismantled, or a new part is simply fitted and the damage returns because the cause was never found. Photos of earlier findings on the same machine are somewhere, never next to the new one.

An example

A service technician from Conveyra Fördertechnik inspects a B 800 belt conveyor at a gravel plant and photographs the belt edge, frayed on one side over some 30 metres. demi recognises edge wear from mistracking, not a longitudinal rip: the belt is not cut through, only the edge facing the conveyor frame is worn. Photos from the inspection six months ago show a clean edge, and four similar cases at other customers were caused by material build-up on the tail pulley and return idlers. demi proposes cleaning pulley and idlers, checking the belt scraper and watching the belt run under load before a new belt is ordered; the belt expert at head office confirms the classification in minutes instead of days.
Fictional company, typical workflow.

Your data, wherever it lives

Damage photos live today in inboxes, chats and on technicians' phones, earlier findings in the maintenance system, materials and bills of materials in the PLM, damage catalogues and maker notes in the DMS. demi reads, understands and links them per machine and component, with your rights. AI classifies a new photo in seconds and puts the most similar cases next to it; the expert decides which action applies.

Where the data lives today

  • Photos, video & audio
  • E-mails
  • Chats
  • Tickets
  • DMS & archive
  • PLM
Data LayerOne view of everything, with your rights

How demi solves it

Three layers in this order: connect the data first, then the rights, then the AI.

  1. 1 · ConnectData Layer

    Data sources in this case: maintenance system, field service app, DMS and PLM. demi replaces nothing; it links every photo to machine, component and material from the bill of materials and PLM, to earlier findings and photos of the same spot and to solved cases on similar machines. Damage catalogues, maker notes and your own rating rules sit alongside.

  2. 2 · PermitPermission Layer

    Photos and findings are visible to those who look after the machine, and customers see only the findings for their own machines. An expert authorised to do so confirms the classification before it goes into the finding. Every confirmation and correction is recorded in the case with name and date.

  3. 3 · ActIntelligence Layer

    The agent recognises component, location and damage pattern in the photo, such as a pitted surface on the suction side of the blades, scoring, cracks or deposits, and classifies the type of damage by your damage catalogue. It shows the most similar earlier cases with their confirmed cause and proposes checks and actions. If a photo is blurred or an angle is missing, it says so and asks for a new picture.

How it runs

  1. 1

    The technician photographs the part with a phone and selects the machine or scans the nameplate.

  2. 2

    demi recognises component and damage pattern and classifies the type of damage by your damage catalogue.

  3. 3

    Photos of earlier findings on the same machine and similar cases with their confirmed cause sit alongside.

  4. 4

    The agent proposes checks on site and the action, each with the case or document it comes from.

  5. 5

    An expert confirms or corrects the classification; the finding goes into the machine file and any spare part into planning as an order.

What changes

  • Damage classified in minutes instead of days in an inbox

  • Causes are fixed, not just parts replaced

  • Consistent findings across technicians and sites

What does it bring you?

Check it with your own numbers

Every starting value is an assumption you can change. The calculation sits right below, step by step.

Assumption, editable

Your numbers

Result

+120 hfreed per year

Assumptions and calculation

demi classifies the damage and puts earlier cases alongside; you confirm and decide.

25 min → 17.5 min per case

How does it run today?

What is already digital saves less: the factor lowers the share demi brings.

How we calculate

80 damage assessments × 25 min × 30% × 1.0 (manual) × 12 months = 7,200 min = 120 h

Value per year, before demi costs

Freed per year+120 h

Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.

= Value per year, before demi costs

+€7,800

gained per year

120 h ÷ 1,760 h per role = 0.1 full-time roles

An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.

Combine with more cases and get the package

Questions about this use case

Which damage can demi recognise in a photo?

Visible damage on surfaces: cavitation, abrasion, corrosion, cracks, scoring, chipping, deformation and deposits. Anything inside or anything that must be measured, such as wall thickness, hardness or imbalance, no photo shows; for that demi proposes the right test. The agent marks uncertain spots as unclear.

How does demi tell cavitation from wear?

By location and pattern: cavitation shows as a spongy, pitted surface where pressure is low, for example on the suction side of the blade inlet edges. Abrasion by solids leaves smooth, washed-out surfaces in the direction of flow. The agent also compares with confirmed cases from your own history.

Do we need our own damage catalogue for this?

No, but it helps. If you have one, for example as a PDF or in inspection instructions, demi uses your categories and terms. If not, demi proposes one from your past findings, which your experts check and release.

Does the classification improve over time?

Yes. Every classification an expert confirms or corrects is stored with photo and cause and is ready as a comparison for the next similar case. The base grows from your own checked cases.

See this case with your data

In the demo we show you the workflow on your own systems.

Your configurator

No use cases yet

Your configurator is still empty

Add a use case from any card or case page and see what it brings per year, with the calculation. Free, no sign-up, no commitment: try it out and change any number.

Browse use cases