Machine Vibration Monitoring

Analysis Server (vibration-analysis-mcp)

• load_signal — Load data from file or DATALOG2 acquisition folder
• compute_fft_spectrum — FFT magnitude spectrum
• find_spectral_peaks — Detect dominant frequency peaks
• assess_vibration_severity — ISO 10816 severity classification
• diagnose_vibration — Full automated diagnosis pipeline

Workflow: First-Time Setup

When a user wants to monitor a new machine, follow these steps:

1. Connect to the board

   • datalog2_connect — auto-discovers the STWIN.box over USB-HID
   • datalog2_get_device_info — confirm firmware version and board identity

2. Configure sensors for the application

   • datalog2_list_sensors — see current sensor configuration (ODR, FS, enabled)
   • Ask the user about machine type and expected shaft speed (RPM)
   • datalog2_configure_sensor — enable/disable sensors, adjust ODR and full-scale
   • Typical configurations:
     • Wideband vibration: IIS3DWB at 26667 Hz ODR — best for bearing analysis
     • General purpose: ISM330DHCX at 6667 Hz — good for unbalance/misalignment

3. Acquire baseline data

   • datalog2_start_acquisition(name="baseline", description="...", duration_s=5)
     • The duration_s parameter auto-stops the acquisition server-side (no round-trip delay)
     • Returns the acquisition folder path
   • load_signal(file_path="<acquisition_folder>", sensor_name="iis3dwb_acc")
     • Sample rate is auto-detected from device config
   • compute_fft_spectrum(data_id="...", channel="X") — compute the FFT
   • assess_vibration_severity(data_id="...", machine_group="group1") — ISO 10816
   • Present this as the baseline vibration signature

Workflow: Routine Monitoring

1. datalog2_connect → datalog2_start_acquisition(duration_s=5) → load_signal
2. compute_fft_spectrum and find_spectral_peaks — compare against baseline
3. Flag any new peaks or amplitude increases >6 dB (factor of 2)
4. assess_vibration_severity — report ISO 10816 severity zone

Key Guidance

• Always specify the axis when discussing vibration (X = horizontal, Y = vertical, Z = axial on typical installations)
• Vibration amplitude in g (acceleration) or mm/s (velocity) — clarify units
• Frequency resolution = sample_rate / N_samples. For 1 Hz resolution at 26667 Hz sample rate you need 26667 samples (~1 second)
• The IIS3DWB is the preferred sensor for bearing fault detection due to its high bandwidth
• The ISM330DHCX is suitable for general unbalance/misalignment detection
• Temperature from STTS22H can help correlate vibration changes with thermal effects

Units of Measurement

Sensor output (IIS3DWB)

• Raw data: int16 (signed 16-bit)
• Sensitivity: 0.000122 g/LSB at ±16 g full-scale
• After SDK decoding: acceleration in g (gravity units, 1 g = 9.80665 m/s²)

ISO 10816 assessment

• Requires RMS velocity in mm/s in the 10–1000 Hz band
• The diagnose_vibration tool converts acceleration (g) → velocity (mm/s) automatically via frequency-domain integration
• If using assess_vibration_severity directly, you must provide velocity in mm/s — do NOT pass raw acceleration in g

Conversion chain

1. Raw int16 × sensitivity → acceleration in g
2. g × 9806.65 → acceleration in mm/s²
3. Frequency-domain integration (÷ j·2π·f) + band-pass 10–1000 Hz → velocity in mm/s
4. RMS of velocity → ISO 10816 input

⛔ Distinguish MCP Tool Output from General Knowledge

When providing information based on your general engineering knowledge (e.g., typical speed ranges, recommended monitoring intervals, sensor placement best practices) rather than MCP tool output, MUST label it:

⚠️ General engineering knowledge — not from sensor data analysis.

Never present textbook knowledge as if it came from the MCP analysis pipeline.

ODR: Nominal vs Measured

When reporting sample rate after load_signal, you may see a slightly different value than the configured ODR (e.g., 26,584 Hz instead of 26,667 Hz). This is normal hardware behavior:

• Nominal ODR = configured value (26,667 Hz for IIS3DWB at ODR index 0)
• Measured ODR = actual rate from the crystal oscillator (~0.3% tolerance)

The measured ODR is used for analysis (more accurate FFT frequency bins). When presenting results to the user, say:

"Sample rate: 26,584 Hz (measured; nominal 26,667 Hz)"

Similarly, the acquisition duration may be slightly shorter in samples than the wall-clock duration_s because the actual ODR is slightly lower. For example, 10 seconds at 26,584 Hz = 265,840 samples (~9.97 s at nominal). This is expected.

Sensor Configuration Reference

See sensor-specs.md for detailed sensor specifications and recommended ODR (Output Data Rate) settings for different use cases.

Evidence & Assumptions Protocol

When presenting results, always separate:

1. Measured evidence (tool outputs, frequencies, amplitudes, ISO zone, statistics)
2. Inference (diagnostic interpretation with confidence)
3. Assumptions / prior knowledge (catalog values, typical fault heuristics, missing machine metadata)

If assumptions are used because required inputs are missing, declare this explicitly and ask for the missing data when practical.

Example Interaction

User: "I want to start monitoring the vibration on our pump"

Response approach:

1. Ask about the pump type, RPM, and bearing information if known
2. datalog2_connect to the STWIN.box
3. datalog2_list_sensors to check current config; adjust if needed
4. datalog2_start_acquisition(name="pump_baseline", duration_s=5)
5. load_signal(file_path="<folder>", sensor_name="iis3dwb_acc")
6. compute_fft_spectrum + assess_vibration_severity
7. Present the FFT spectrum and ISO 10816 assessment
8. Suggest a monitoring interval (e.g., weekly for Zone A/B, daily for Zone C)