Skip to main content

Aerospace — fleet predictive maintenance

The problem. An airline operates a fleet of aircraft, each a deep hierarchy of systems and components, each component streaming sensor data every second of every flight. When one part starts trending toward failure, two questions follow immediately: is this part about to fail, and which other aircraft carry the same part and are therefore at the same risk? The second question is a graph question — and answering it fast is the difference between scheduled maintenance and a grounding.

This scenario leans on the knowledge graph: model the fleet down to the component, and when one component looks suspect, walk the graph to its whole blast radius.

Set up demo data​

New workspace? Run this once (Python) to create three aircraft whose pumps are all the same part type — so the blast-radius walk in step 2 returns the whole affected fleet. Safe to re-run.

import datahub_sdk

client = datahub_sdk.DataHubClient.from_env()

nodes = [datahub_sdk.Resource(external_id="part_type_hp_47", name="Hydraulic pump type HP-47", labels=["part_type_hp_47"])]
edges = []
for ac in ["ln_312", "ln_318", "ln_401"]:
    nodes += [datahub_sdk.Resource(external_id=f"aircraft_{ac}", name=f"Aircraft {ac}", labels=[f"aircraft_{ac}"]),
              datahub_sdk.Resource(external_id=f"hyd_pump_{ac}_1", name=f"Pump {ac}", labels=[f"hyd_pump_{ac}_1"])]
    edges += [datahub_sdk.RelForm.by_external_ids(f"aircraft_{ac}", f"hyd_pump_{ac}_1", "contains"),
              datahub_sdk.RelForm.by_external_ids(f"hyd_pump_{ac}_1", "part_type_hp_47", "is_part_type")]
client.resources.create(nodes, edges)

1. Model the fleet to the component​

A fleet contains aircraft, an aircraft contains systems, a system contains components. Components of the same part type also link to a shared part_type node — that shared node is what makes fleet-wide correlation possible.

ResourceForm aircraft = new ResourceForm();
aircraft.setExternalId("aircraft_ln_312");
aircraft.setName("Aircraft LN-312");
aircraft.setLabels(List.of("aircraft_ln_312"));

ResourceForm pump = new ResourceForm();
pump.setExternalId("hyd_pump_ln_312_1");
pump.setName("Hydraulic pump #1");
pump.setLabels(List.of("hyd_pump_ln_312_1"));

// the component is an instance of a shared part type
RelForm installed = new RelForm();
installed.setName("contains");
installed.setFromExternalId("aircraft_ln_312");
installed.setToExternalId("hyd_pump_ln_312_1");

RelForm isType = new RelForm();
isType.setName("is_part_type");
isType.setFromExternalId("hyd_pump_ln_312_1");
isType.setToExternalId("part_type_hp_47");

client.resources().create(List.of(aircraft, pump), List.of(installed, isType));

Each component streams series — hyd_pump_ln_312_1_vibration_mm_s, hyd_pump_ln_312_1_temp_c — ingested at fleet scale.

2. A part looks suspect — find the blast radius​

Rising vibration on one pump suggests a part-type defect. Walk the graph out from the shared part_type_hp_47 node: every component is_part_type of it, and every aircraft that contains one, is the blast radius — the list maintenance needs to inspect.

ResourceNetwork affected = client.resources().fetchRelated("part_type_hp_47", 3);

// every aircraft in the returned sub-graph carries this part type
affected.nodes().stream()
    .filter(n -> n.getExternalId().startsWith("aircraft_"))
    .forEach(a -> System.out.println("inspect: " + a.getExternalId()));

One traversal turns a single suspect reading into a precise, fleet-wide inspection list. See Correlate alarms with the graph for the pattern.

3. Record the finding​

Raise a component_anomaly event per affected component so the maintenance system schedules the work and keeps an auditable trail.

See the result​

Walking out from the suspect part type returns every aircraft to inspect:

inspect: aircraft_ln_312
inspect: aircraft_ln_318
inspect: aircraft_ln_401

See also​