How diagnostics work in Wertek AI

How it works (concept)

Wertek AI diagnostics turn a field measurement into an actionable interpretation on an identified asset. It covers three modalities: vibration (bearings, unbalance, misalignment), ultrasound (compressed-air leaks, steam traps, partial discharge, and bearings), and thermography (thermal images with capture parameters). There are two ways to capture: with the phone as the instrument (a guided flow in the mobile app), or by uploading signal files from the web platform, coming from your own instruments.

Each measurement stays linked to the asset and returns an interpreted result: health index, severity, probable findings, and a recommendation for action. The approach is standards-based, trend-driven engineering — ISO-type severity bands and comparison against the same asset's history — not a "black box" that promises predictions without a basis. For bearing analysis, the platform includes a catalog of more than 30 common bearings and computes the fault frequencies (BPFO, BPFI, BSF, FTF) that it marks on the spectrum.

What the full solution is and what it solves: Condition diagnostics and monitoring. This guide teaches you how to use it.

First 15 minutes

Your first capture with the phone, start to finish:

1. Open the mobile app and start the capture flow ("Capture Measurement").
2. Choose the signal type: ultrasound, vibration, or thermography. The screen prompts you to choose the type of measurement you want to capture.
3. Choose the application: for example, in ultrasound — bearing, steam trap, leak, or partial discharge.
4. Select the asset: scan its QR/NFC or search it by name. If you scanned the asset first, the app takes you straight to choosing the application.
5. Capture: hold the phone (or the external microphone/probe) firmly against the measurement point and keep it steady during the live capture.
6. Review the result: when it finishes, the app shows the analysis result. You'll also see it on the web under Measurements (Diagnostics group), with spectrum and findings detail.

Tip: fill in the capture parameters you know (motor RPM, measurement point, mounting type). If you don't know a value, leave it blank — that's better than guessing, because an incorrect value degrades the analysis.

Step-by-step flow

Path A — capture with the phone: the first-15-minutes flow (signal type → application → asset → capture → result). Repeat the measurement at the same point and load conditions to build a trend.

Path B — upload signal files (web): the Upload menu opens "Upload Signal": (1) select the signal type (vibration or ultrasound), (2) select the asset, (3) choose the application type, (4) drag in the file — WAV, CSV, TDMS, or UFF up to 50 MB are accepted — and (5) press Analyze Signal. The result is associated with the asset just like a mobile capture.

How to read a result: each measurement shows a health index, a severity band, and the detected findings with their recommendation. Low severity = document and follow up on trend; high or critical severity = schedule intervention — from the finding you can create a work order (see CMMS guide).

Ultrasonic leak survey: in the app, open the Leak Estimator and start a Leak Survey: walk the compressed-air network, and at each suspect point use Document Leak — capture the audio, assign a tag (e.g., F-001), and press Save Leak. The app asks "Do you want to document another leak?" so you can continue the walk-down. The consolidated list stays in Leaks, on web and mobile, with each leak's estimate.

Screening or formal measurement? The phone is a screening tool: fast, comparative, and enough to prioritize and build a trend. Use a formal measurement (accelerometer or dedicated instrument, uploaded via Path B) when screening shows high severity, when you need evidence for a costly decision, or on critical assets. The app also includes a quick motor check using the phone's magnetometer (experimental feature, indicative only).

Common issues and fixes

• "The file upload fails" — the limit is 50 MB per file and supported formats are CSV, WAV, UFF, and BIN (on the Upload page: WAV, CSV, TDMS, UFF). Also verify the file isn't corrupted.
• "The analysis takes too long" — normal processing takes 30 to 60 seconds; during peak-demand hours there may be a queue. If it exceeds 5 minutes, retry the analysis.
• "The result looks incorrect" — check that the captured RPM, bearing model, and measurement-point orientation are the real ones. A wrong RPM shifts all the computed fault frequencies.
• "QR scanning doesn't work" — verify the app has camera permission, clean the code surface, and if it persists, use manual entry of the asset code.
• "Noisy or inconsistent capture" — press the sensor firmly on a clean metal surface, avoid capturing while the equipment is in transient (start/stop), and always repeat at the same point so the trend is comparable.

If the problem persists, write to [email protected] with the measurement ID.

What's next

• From spot checks to continuous: when an asset warrants permanent watch, add wireless vibration and temperature sensors — they self-discover on connection and are assigned to measurement points. Detail in Diagnostics and monitoring.
• Inspection routes: structure recurring captures as routes with checklists executable from the app — see the CMMS guide.
• From finding to action: turn high severities into work orders with evidence (CMMS guide) and complement with the asset's electrical dimension (energy guide).