← Back to Architecture Edge · Sensing · Doors

Pod B
Doors · Audio

IoT sensing unit installed on the frame and actuator mechanism of each door. It monitors the open/close cycle, the actuator force and the acoustic signature to detect wear, misalignment and obstructions before in-service failure.

What it is

Pod B is IN-SIGHT's sensing unit specialised in door mechanisms, the main source of operational incidents in metro and commuter rail. It integrates three measurement modes: door position via Hall sensor, current of the pneumatic or electromechanical actuator, and acoustic signal via a digital MEMS microphone.

Combining these three signals makes it possible to build a complete profile of each door-open / door-close cycle, which is compared against the Golden Run baseline to detect any drift preceding an in-service failure: friction, degraded sealing, a leaking pneumatic valve or an actuator with abnormal consumption.

Cause #1 of operational incidents: Doors are the leading reason for cancellations and irregularity in metro and commuter operation. Pod B turns each door cycle into a structured data point, transforming reactive management into planned intervention.

Role in IN-SIGHT

Pod B captures three independent channels and delivers them to the CM4 Edge CPU where local analysis runs. Each channel provides a different dimension of the mechanism's condition:

  • Hall sensor — position: Detects the open and close end-of-travel with ±0.5 mm resolution. It measures cycle time, travel speed and intermediate position (door ajar, partial obstruction).
  • Current sensor — actuator: Current profile of the motor or proportional valve. A sustained increase in current at the same cycle speed indicates growing mechanical friction or sealing degradation.
  • MEMS microphone — acoustic analysis: The audio signature of each cycle is unique for every mechanism in good condition. Variations in spectral content in the 200–4,000 Hz bands identify seal friction, worn ratchets or unlubricated guides.

Acoustic analysis pipeline

Acoustic analysis is the channel with the greatest early-detection capability. The MEMS microphone captures PDM audio that the CM4 converts to PCM and processes in real time:

MEMS PDM microphone  (I²S @ 32 kHz)
    │  16 bits / sample · mono
    ▼
PDM → PCM decimation  (factor 64)
    │  Effective frequency: 500 Hz
    ▼
Door-event segmentation
    │  Trigger: Hall edge (cycle start)
    ▼
512-point FFT per segment  (Hamming window)
    │  Resolution: 1 Hz/bin up to 250 Hz
    ▼
Acoustic feature extraction
    │  Spectral centroid · Spectral flux
    │  MFCC (13 coefficients) · Energy per band
    ▼
Comparison vs. Golden Run (EKF)
    │
    ├─ Normal    → cycle log (30 s batch)
    └─ Anomaly   → alert + defect type

Detected failure modes

Pod B is optimised for the door failures with the highest statistical incidence in short/medium-distance fleets:

  1. Seal gasket wear: Change in the closing-force peak and variation of the acoustic spectral centroid in the final part of the cycle.
  2. Mechanism misalignment: Asymmetry between the opening and closing profiles, with a current increase at mid-travel.
  3. Partial obstruction: Cycle reversal (open → close → reopen) with a characteristic current signature.
  4. Actuator with pneumatic leak: Pressure drop reflected in a prolonged cycle time and irregular current in the solenoid valve.
  5. Unlubricated guide: Progressive increase in high-frequency acoustic friction (800–3,000 Hz) between overhauls.