Model-in-the-Loop Engineering · Capstone brief

ENG-06 · Fault injection lab

When you inject named sensor faults into an approved loop, does the controller fail gracefully (detected, safe state) or silently (wrong output, no alarm)?

The question

When you inject named sensor faults into an approved loop, does the controller fail gracefully (detected, safe state) or silently (wrong output, no alarm)?

System / materials

Teacher-approved controller + sensor set. Fault list fixed by teacher (stuck-at, drift, dropout, scale error).

Expected failure modes

Only injecting faults the controller already ignores. No safe-state definition.

Done looks like

Fault matrix: fault type × detection time × safe state reached, with one silent-failure case honestly reported.

Five C's

CT: silent vs. loud failure. CR: choosing faults that matter. CO: peer injects one fault blind. CM: matrix a maintainer can scan. CZ: hazard if silent.

Mentor role

Controls engineer reviews fault list at Checkpoint 1 and matrix at Checkpoint 2. School-supervised.

Rubric calibration

R1: approved fault set. R2: injection procedure logged. R3: nominal baseline run. R4: silent failure addressed. R5: matrix complete. R6: no production system.

Two ways this goes wrong

(a) Only faults that trip obvious alarms. (b) Safe state undefined — “it stopped” is not a case.

Credit lane fit

Lane A immediately. No verified credit claim.