A failure window, a confidence level and the evidence. Then you decide.
The AI Failure Predictor compares each machine's trends with its own history and with known failure patterns such as bearing wear, misalignment, unbalance and looseness. It estimates when a failure is likely and how sure it is.
Fault type
Bearing, alignment, unbalance
Window
Likely failure in days
AI Failure PredictorGB-3 output bearing 6312
Outer-race wear pattern
Likely failure window from today
Evidence
- Peak at outer-race defect frequency, 3.2x baseline
- Bearing temperature up 8 °C over 14 days
- Trend shape matches 2 past failures on this asset
Engineer decision
Raise work orderWatch 48 h moreDismiss with reasonConfidence
Low, medium, medium-high, high
Fault type
Bearing, alignment, unbalance
Window
Likely failure in days
Confidence
Low, medium, medium-high, high
What it does
AI Failure Predictor, part by part
Compares each machine's trends with its own history and known failure patterns, then estimates likely failures with a confidence level for engineers to judge.
Machine-specific baselines
Each asset is judged against its own normal behaviour at its own load and speed.
Fault pattern matching
Recognises common signatures: outer and inner race wear, misalignment, unbalance, looseness and cavitation.
Failure window
Gives a range of days, not a single date, because real machines do not fail on schedule.
Evidence shown
Lists the readings and frequencies that led to the prediction, so an engineer can check them.
Suggested action
Proposes the likely repair and the next planned shutdown that fits the window.
Learns from outcomes
When a job closes, the finding is fed back so the next prediction on that machine is sharper.
Sequence of operation
How it runs in your plant
- 1
Watch
Trends from Sensor Watch arrive every 10 minutes.
- 2
Compare
The model checks them against the machine's own history and known fault patterns.
- 3
Advise
A prediction appears with a window, confidence level and evidence.
- 4
Decide
An engineer raises a work order, chooses to watch longer or dismisses it with a reason.
Spec sheet
- Learning period
- About 14 days of readings per machine
- Typical warning
- Several days to several weeks before failure, depending on fault type
- Confidence levels
- Low, medium, medium-high, high
- Output
- Advisory only, never sent to machine controls
- Feedback
- Every closed job updates the model for that asset
Safety and responsible AI
- AI predictions are advice with confidence levels. Maintenance engineers decide what to do and when.
- Schafte never starts, stops or controls machines. It only reads data.
- Machine data stays private to your company, is encrypted and is never sold.
Connected parts
Hang a tag on your first 5 machines.
- Machines
- 5
- Days free
- 60
- Card needed
- No
Signed: maintenance head, ____________
Control station
- Start with sensors we rent you, or with PLC data you already have
- Asset register, work orders and PM live in the first week
- Your first AI baseline is ready after about 14 days of readings