← Back to Architecture Cloud · Dashboard · Maintenance

Portal IN-SIGHT
Alerts · Dashboard · Golden Run

The maintenance engineer's interface. It centralises alerts classified by severity and subsystem, visualises each vehicle's health drift against the Golden Run and plans interventions before in-service failure.

What it is

The Portal IN-SIGHT is the system's presentation layer. A responsive web application that aggregates, filters and visualises the telemetry processed by the cloud analytics layer (ADX + EKF) and turns it into actionable information for the maintenance team.

The design follows a single principle: the maintenance engineer should not have to interpret raw data. The Portal presents only classified events, health trends with operational meaning and direct comparisons against the approved Golden Run baseline. Everything else stays out of view.

Production access: The portal deployed in the IN³ programme is available at portal.in3-insight.cloud with Azure Active Directory authentication. Access is restricted to authorised operator users.

Portal modules

Fleet view — Fleet Overview

Visual map of all monitored vehicles with a global health indicator per vehicle. Colour code: green (nominal), yellow (active non-critical alert), red (critical alert pending attention). The view refreshes every 30 seconds with the latest data from ADX.

Alert centre — Alert Center

Chronological timeline of all active and historical alerts, classified into three levels:

  • CRITICAL: Detected degradation exceeding 3.5σ of the baseline. Requires inspection before the next service. Notification SLA: 2 minutes from detection.
  • WARNING: Sustained negative trend (2–3σ) without the threshold yet exceeded. Schedule a check at the next planned overhaul.
  • INFO: State change or system event (new installation, firmware update, approved Golden Run).

Golden Run comparator

Overlaid view of the subsystem's current signal against the approved Golden Run baseline. The X axis is the time within a normalised cycle (one wheel revolution, one door cycle), the Y axis is the metric in question (RMS vibration, temperature, current). The deviation is computed in number of sigmas and shown as a confidence band.

Subsystem dashboard

Detailed view per vehicle and subsystem (front bogie, rear bogie, door 1-L, door 1-R…) with a time series of the key metrics, the accumulated frequency spectrum of the last 24 hours, and a projected health trajectory via regression over the history.

Device management

Inventory of installed Pods, real-time connectivity status, active and pending firmware version, and access to the per-device EKF configuration parameters.

Technology stack

Frontend
    React 18 + TypeScript
    Recharts (time series and spectra)
    Tailwind CSS (dark theme by default)
    Azure Static Web Apps (global CDN)

Data backend
    Azure Data Explorer  →  KQL queries
    WebSocket (real-time alerts)
    Azure API Management (rate limiting)

Authentication
    Azure Active Directory B2C
    RBAC: Admin / Engineer / Viewer

Deployment
    Azure Static Web Apps  (CI/CD via GitHub Actions)
    Region: West Europe (latency < 40 ms Spain)

Engineer's workflow

The Portal is designed for a 10-minute daily review workflow and alert response in under 2 hours:

  1. Daily review: Login → Fleet Overview → identify vehicles with non-green indicators.
  2. Alert assessment: Open alert → see affected subsystem, severity, evolution time and projected trend.
  3. Comparison with Golden Run: Open comparator → check the magnitude of the σ deviation and confirm it is not a false alarm.
  4. Intervention decision: CRITICAL → immediate work order. WARNING → note in the upcoming maintenance plan.
  5. Closure: Log the action taken in the portal → the system restarts monitoring and updates the baseline if a repair has been carried out.