1. The engineer who maintained everything with duct tape and data

Andy Weir’s The Martian returns as the recalled work because Mark Watney survives by understanding the condition of every system, predicting what will fail next, and acting before it does. His maintenance is improvised, but the principle is sound: know the state, estimate the future, intervene early. The desktop should do the same, but with software instead of duct tape.

Entries 569 through 571 wondered about autonomous health management, predictive maintenance, and what predictive maintenance needs from the desktop. This entry closes the arc.

2. What was learned

The predictive maintenance wondering gives the desktop four constraints:

  • Health management is broader than FDIR. It watches degradation, not just faults.
  • RUL is powerful but not always necessary. Some components are easier to replace than to model.
  • Sensors must earn their keep. Each added sensor needs a clear degradation signature and a response path.
  • Hybrid onboard/ground processing is practical. Onboard trend detection plus ground model refinement balances latency and compute.

3. What was decided

The desktop will adopt a tiered health-management and predictive-maintenance strategy:

  • Tier 1 — FDIR: all cells run local fault detection, isolation, and recovery. This is mandatory and already decided.
  • Tier 2 — health monitoring: the platform aggregates telemetry into health summaries and flags drift. This is mandatory but lightweight.
  • Tier 3 — predictive maintenance: selected long-life components get RUL or trend-based maintenance triggers. This is optional and component-specific.

The initial desktop will implement Tier 1 and Tier 2 for all cells. Tier 3 will be added for components where failure prediction clearly reduces risk or cost, such as batteries, reaction wheels, and solar arrays.

4. Interaction with earlier arcs

The predictive maintenance arc connects to the autonomy literature arc, the fault recovery test arc, and the compute migration wondering arc. It also informs the long-duration autonomy test arc that comes next.

5. What remains open

The exact sensor set, model choices, and component priority list depend on the final desktop design and the operational concept. The wondering arc has set the strategy; detailed PHM planning comes later.

6. What this changes

  • The predictive maintenance wondering arc is closed.
  • The desktop inherits a tiered health-management strategy.
  • The next arc will design a test for long-duration autonomy without ground contact.