TwinEdge / Platform · Remaining lifeWorking out what is wrong
How long it has left
Plan around developing condition.
Use degradation trends, operating duty, and asset history to estimate remaining life and focus the maintenance plan.
What a prediction gives a planner
| What to do | Replace the bearing assembly |
| When | Within the next two to three weeks |
| How long | About 2.5 hours, including alignment |
| What is needed | Bearing, seal kit, 400 ml of lubricant |
| How sure | Stated as a range, not a single day |
§ 01
What predicting failure can and cannot do.
It works on wear: bearings, seals, impellers, insulation. These degrade in ways that show up in the readings for weeks, which is what makes them predictable at all.
It does not work on sudden events. A power surge, a foreign object, or a valve left shut will not be predicted by anything, and a vendor who says otherwise is selling you something.
It also needs history. A machine instrumented last month has nothing to learn from. On a machine with two years of readings and a couple of past failures, it does well.
“Wear can be predicted. Accidents cannot. Most of your unplanned downtime is wear.”
The honest scope
§ 02
What it produces.
Every prediction carries its uncertainty. A confident-sounding single date would be easier to read and worse to plan with.1
- 01Enough warning to be useful
Long enough to order the part, book the crew, and pick a window. A prediction that arrives the day before failure is only a more precise way of being surprised.
- 02A number, and how sure it is
Days remaining, with a range around it and a health score. The range matters as much as the number, and it is never hidden.
- 03Which part, not just which machine
Bearing inner race, seal face, winding insulation. A technician can bring the right part rather than a diagnosis of “something is wrong”.
- 04A window that suits the plant
Suggested timing balances what life is left against the production schedule and who is available, so the repair lands somewhere it costs least.
Point it at a failure you already had.
Take a machine that failed, with its history intact. We run the readings up to the day before and show you what it would have said, and how early. You already know the answer, which is exactly why it is a fair test.
- 1Prediction quality depends on your instrumentation, how much history exists, and whether past failures were recorded well enough to learn from. We measure it against your own failures during an evaluation rather than quoting a benchmark figure.