What cloud analysis is
Cloud analysis is the layer where data from all vehicles is gathered, stored and correlated. While the edge decides what is urgent on board, the cloud provides what a single train cannot see: deep history, comparison across vehicles and models that learn from the whole fleet.
IN-SIGHT relies on Azure: IoT Hub manages the secure connection of each device and Azure Data Explorer (ADX) stores and queries time series at large scale with low latency.
Incipient drift detection: the system does not wait for the alarm threshold; it identifies the anomalous trend weeks before the component reaches a critical state.
How IN-SIGHT does it
The cloud chain turns aggregated telemetry into actionable diagnosis:
- Secure ingestion: Azure IoT Hub receives the telemetry via MQTT over TLS, with per-device identity (device twin).
- Analytical storage: ADX stores the time series and allows KQL queries over millions of records in seconds.
- EKF against Golden Run: each signal is compared with the vehicle's health baseline to quantify the real deviation.
- Drift classification: the algorithms separate normal variation from progressive degradation, distinguishing root cause by subsystem.
In railway practice
For the fleet engineer, the cloud turns scattered data into a health picture comparable across vehicles. If a bogie on one unit starts to deviate from its Golden Run, the system detects it as a trend —not as an isolated spike— and places it in the context of the rest of the fleet.
This makes it possible to plan maintenance weeks in advance: moving from reacting to failure to scheduling the intervention in the optimal window, minimising downtime and cost.