AI Fault Predictor · Roof and basement

Hear about the chiller failure ten days before the tenants feel it.

The Fault Predictor studies each asset's readings, run hours and repair history. When the pattern starts to look like the run-up to a past failure, it flags the asset, estimates the window and opens an inspection job for the right team.

Manage one building free Connect with sales
0605040302GB1L2CriballtonRoof and basement
AI Fault Predictor · Watchlist3 assets flagged
Chiller 2 · RoofRisk high
Likely failure in ~10 days
Bearing vibration, mm/sAlarm 4.5
  • Vibration up 86% in 14 days
  • Discharge temperature 4 °C above its normal
  • Similar pattern before the 2023 bearing failure
WO-2481 inspection assigned to CoolAir HVAC · 48 h
Pump P3 · B1Watch
AHU-4 · Floor 4Watch

How it works

  1. 01Read

    Sensor data, PPM readings and job history per asset.

  2. 02Compare

    Against the asset's own normal and past failure patterns.

  3. 03Warn

    Risk score, likely window and the readings behind it.

  4. 04Inspect

    Inspection job opened, engineer confirms or dismisses.

What it does

Built for the way ai fault predictor really run.

Learns each asset

Vibration, temperature, current draw, pressure and run hours compared against that asset's own normal.

Uses your history

Past breakdowns and repairs on similar assets teach the model what a run-up to failure looks like.

A window, not a guess

Each warning shows the likely window, the readings behind it and how confident the model is.

Inspection first

Warnings open an inspection job. Engineers decide whether repair is needed, never the model.

Works with few sensors

Even manual PPM readings and repeat-fault history are enough to start flagging risky assets.

Gets better with feedback

Engineers mark each warning as confirmed or not. The model adjusts for that building.

Specification sheet

The details your engineers will ask about.

Everything listed is included in the plans that carry AI Fault Predictor. See Plans for what each plan covers.

Asset classes
Chillers, AHUs, pumps, DG sets, lifts, transformers, UPS
Inputs
Sensor readings, PPM readings, run hours, jobs, parts used
Output
Risk score, likely window, contributing readings
Action
Inspection work order to the assigned team or vendor
Feedback
Confirmed, not confirmed, already fixed
Minimum data
About 60 days of readings or 12 months of job history

Trust notes

How this data is used.

  • Fault predictions are estimates that prompt inspections. Engineers decide on repairs.
  • Facility data belongs to the client, is encrypted, and is never sold.
  • Models learn from your buildings for your buildings, not for other clients.

Questions

AI Fault Predictor: common questions

Is the prediction always right?

No. Predictions are estimates that prompt an inspection. Engineers decide on repairs, and their feedback improves later warnings.

How much data does it need?

Around 60 days of readings for an asset, or a year of job history across similar assets, to begin useful warnings.

Is our data used to train for other clients?

No. Models learn from your buildings for your buildings. Your data is never sold.

Next system · Upper floorsWork OrdersEvery complaint from a tenant, staff member or sensor becomes a tracked job with an owner and a deadline.

Key register · Handover

Take the keys to one building.

Bring your asset list and your floor plan. Within a day your QR tags are printed, your PPM calendar is built and your team is raising jobs. Run it free for 30 days.

Issued to: your buildingManage one building freeRunning a portfolio? Connect with sales
Trial
30 days, one building
Setup
Usually one working day
Card
Not needed to start