1. The precogs of the machine room

Philip K. Dick’s Minority Report returns as the recalled work because its precogs see a crime before it is committed, allowing intervention in advance. Predictive maintenance tries to do the same for machinery: detect the conditions that precede failure and act before the failure occurs. The desktop could use such foresight to schedule swaps, reconfigure loads, or defer risky operations.

This entry wonders about predictive maintenance and remaining useful life (RUL) on the desktop.

2. From reactive to predictive

Maintenance strategies form a spectrum:

  • Reactive: fix it after it breaks.
  • Preventive: replace it after a fixed interval or cycle count.
  • Predictive: replace it when data indicates it is actually degrading.

A 2024 review of PHM for predictive maintenance identifies RUL prediction as the key enabler: estimating how much useful life remains before maintenance is required. This avoids both premature replacements and unexpected failures.

3. What RUL means for spacecraft

Remaining useful life is the expected operating time before a component can no longer perform its function. For the desktop, candidate RUL applications include:

  • Batteries: cycle count, depth of discharge, and temperature history predict capacity fade.
  • Reaction wheels: bearing wear, vibration, and current draw predict impending failures.
  • Mechanisms: cycle count, motor current, and positional repeatability predict lubrication degradation.
  • Solid-state storage: program/erase cycle count and bad-block growth predict media wear.
  • Solar arrays: power output degradation predicts end-of-life margin.

An aerospace ISHM survey notes that model-based, data-driven, and hybrid reasoning techniques can all be used for detection, diagnosis, and prognosis.

4. The prediction problem

Predictive maintenance is not free. It requires:

  • Sensors that capture degradation signatures.
  • Historical data linking those signatures to actual failures.
  • Models that can distinguish normal variation fromincipient failure.
  • Compute resources to run the models.
  • A maintenance concept that can act on the predictions.

The NASA small-satellite health monitoring guide asks a sharp question: do you really need RUL? Sometimes anomaly detection is enough; sometimes RUL is essential for mission planning.

5. What this changes

  • Predictive maintenance shifts the desktop from time-based replacement to condition-based replacement.
  • RUL is valuable for long-duration components but may be overkill for easily replaceable items.
  • The next entry will ask what predictive maintenance needs from the desktop architecture.