1 — The real phenomenon
FDIR means Fault Detection, Isolation and Recovery. Detecting an anomaly is not enough: a bad sensor, degraded equipment, common cause and secondary effect must be distinguished before selecting a safe recovery. Observability depends on sensor quality and consistency models.
The guiding question is: How do we know what actually failed when several faults produce similar symptoms? Reasoning starts with the physical or operational function before introducing the mathematical relationship. The goal is not to accumulate terminology, but to know which quantity changes, why it changes and what becomes hazardous when it leaves its domain. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the first task here is therefore to identify the mechanism specific to this subject before searching for an equation or reference value.
2 — Vocabulary and problem boundary
In “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, distinguish the phenomenon, available measurement, any command, the margin and the success criterion. The calculation boundary states what is included and excluded; without that boundary, a percentage, mass or time may be mathematically correct but wrong as an engineering conclusion. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the chosen boundary also states what would otherwise be double-counted or omitted from a mission budget.
- Primary observable
- memory errors, resets, dose, current, temperature, sensor availability, data consistency and event logs
- Characteristic failure
- bit upset, misconfigured watchdog, bad time reference or redundant channels sharing the same supply
- Expected evidence
- error injection, irradiation where relevant, software tests, data verification and safe-mode simulation
3 — Course-specific system view
This lesson does not reuse one generic picture for every subject. The system view follows cause → measured quantity → decision or physical response → limit for “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”. The English text remains fully equivalent while large translated illustrations are intentionally deferred until their dedicated artwork is supplied. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the system view must expose inputs, outputs, measured quantity and the consequence of drift without relying on a generic module diagram.
4 — Mathematical relationship and reading the symbols
Read aloud : a residual is the difference between a measurement and the model prediction.
Before substituting numbers, write the unit of every term, state whether the relationship is a physical law, approximation or project indicator, and check dimensional consistency. This is especially important here because “Diagnostics and FDIR: detecting, isolating and recovering from anomalies” combines quantities that do not all have the same evidence status. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, this relationship is chosen because of the phenomenon under study; a different dominant quantity would require a different equation or model.
5 — Worked calculations and interpretation
1. 1. Residual
102.4 − 100.0 = +2.4 units
2. 2. Threshold
|2.4| > 2.0 → alarm in this example
3. 3. Sensor vote
Two sensors at 50 and 51; one at 88 → third is suspect, to be confirmed
6 — What the formula does not contain
The relationship “Résidu = mesure − prédiction” does not by itself contain all of “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”. It does not automatically tell us whether a sensor is valid, a structure is aging, a resource is accessible, a command arrives in time or a secondary failure removes margin. The example 102.4 − 100.0 = +2.4 units therefore remains a local calculation rather than a complete architecture.
To make the model useful, explicitly add the quantities that dominate this subject: memory errors, resets, dose, current, temperature, sensor availability, data consistency and event logs. We can then ask which variation truly changes the result, which is negligible and which forces an architectural change. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, this model limitation states exactly what a correct calculation still cannot establish about the real system.
7 — Instrumentation, observability and data quality
For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, observability relies on memory errors, resets, dose, current, temperature, sensor availability, data consistency and event logs. Each datum has a unit, acquisition rate, uncertainty, timestamp and validity domain. A value arriving without context can be more dangerous than no measurement because it creates unjustified confidence.
Consistency is checked with at least one independent piece of information when the function is critical. A trend, physical balance or second measurement principle helps distinguish a real system change from a drifting sensor. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the selected instrumentation must distinguish a real physical change from sensor drift or a bad state estimate.
8 — Phenomenon-specific failures and recovery
The reference failure is not a vague “broken component.” For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, test in particular bit upset, misconfigured watchdog, bad time reference or redundant channels sharing the same supply. Diagnosis asks which symptoms appear first, which are only consequences and which action preserves the most options.
The degraded mode must be defined before failure: minimum function, allowable duration, consumed stock, crew action, abort condition and return-to-nominal criterion. That sequence is topic-specific and cannot be replaced by one universal paragraph about redundancy. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the degraded mode is defined around the minimum function specific to this subject, with an abort threshold and a return-to-nominal condition.
9 — NASA / reference case
NASA material is used as an evidence dossier: requirements, reliability, maintainability, testing and configuration. The lesson never turns a generic failure rate into a universal truth; it shows how evidence is bounded to defined hardware, environment and duration.
The case is used only within what it actually demonstrates. Flight measurement, human-system standard, component test and architecture study are different kinds of evidence; the text therefore states what is observed, calculated, simulated or still prospective. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the cited NASA case is used as targeted evidence for this phenomenon and is never turned into one universal Mars architecture.
10 — Architecture trade
A good solution for “Diagnostics and FDIR: detecting, isolating and recovering from anomalies” does not maximize one metric. Compare nominal performance, mass, energy, simplicity, maintenance, crew time, common dependencies and recoverability. An option that improves |2.4| > 2.0 → alarm in this example can still be rejected if it makes failure detection or repair much harder.
The trade is recorded together with its assumptions. If environment data, mass or mission cadence changes, we know which conclusions must be recomputed instead of silently preserving an obsolete choice. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the trade is evaluated against the interfaces actually touched by this subject rather than a generic list of desirable qualities.
11 — Demonstration, testing and success criteria
The evidence strategy for “Diagnostics and FDIR: detecting, isolating and recovering from anomalies” combines error injection, irradiation where relevant, software tests, data verification and safe-mode simulation. Every test records exact hardware, software, configuration, environment, tolerances and success criterion. A successful demonstration outside the mission domain does not replace qualification inside it.
Evidence grows by levels: analytical relationship, simulation, component, subsystem, integrated system, duration and failure. This hierarchy prevents one spectacular test from being presented as validation of the whole mission. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, demonstration must reproduce the constraints that make this phenomenon difficult; a spectacular test outside the mission domain is insufficient.
12 — Decision exercise
Situation: revisit “Diagnostics and FDIR: detecting, isolating and recovering from anomalies” with a 20% increase in the most penalizing quantity from the first worked example while one measurement or backup path is unavailable.
13 — What to retain without over-generalizing
- Diagnostics and FDIR: detecting, isolating and recovering from anomalies has its own observables and failure modes.
- The relationship Résidu = mesure − prédiction remains attached to its units and boundary.
- NASA evidence is cited at the phenomenon level instead of reusing one reference bundle for an entire module.
14 — Topic-specific primary sources
These references directly document the phenomenon, technology or human constraint addressed in this lesson. They do not by themselves define an official Mars architecture. For “Diagnostics and FDIR: detecting, isolating and recovering from anomalies”, the bibliography is deliberately targeted to this page so that readers can trace each claim back to the relevant primary document.