Engineering & Technology · Aerospace Engineering: Flight & Space Systems
Ground and flight testing, reliability and mission trade-offs
A successful ground test can reveal a component’s behaviour under those conditions. Establishing mission capability requires connecting many such observations to requirements, interfaces and the environment that will actually be encountered.
Review dynamics and vibrations, fluids, electrical power and machines, control and automation, and the physics of forces, gravitation and thermodynamics. Define symbols and units before substitution; 1 Wh=3600 J and 1 rad=180/π degrees. No lesson provides a flight, launch, engine, test or maintenance procedure.
Back to the chapterAfter reading, you can
Plan an exact mission requirement-to-evidence thread, interpret uncertainty under a declared decision rule, and compare reliability with physical resource constraints.
Different evidence answers different questions
Verification checks conformity to specified requirements. Validation asks whether the realised system satisfies the intended mission and users in its operational context. Qualification addresses a design’s suitability over an anticipated environment; acceptance checks the particular manufactured article. Certification is a separate decision by an appropriate authority, with its own required evidence. These terms are not interchangeable stamps.
Analyses, inspections, demonstrations and tests can each contribute. Material coupons and components isolate properties; structural or subsystem tests reveal interfaces and dynamics; integrated ground tests connect hardware, software, people and support systems. Instrumentation, calibration, configuration and uncertainty must be recorded so the observation remains interpretable.
Similarity has to match the relevant physics
A scaled wind-tunnel model may reproduce some flow conditions without reproducing structural mass and stiffness. Matching Mach number alone does not establish Reynolds similarity or full-scale aeroelastic response. A structural vibration test, a thermal-vacuum test and an end-to-end command/data test challenge different parts of a spacecraft. None individually covers all mission functions.
Flight tests add real environmental interactions but carry limits in the conditions and failures actually observed. Engineers plan controlled expansion and explicit criteria through competent programmes; this lesson supplies no flight-test manoeuvres. A software update, wiring change or altered payload can invalidate part of earlier evidence. Preserve what changed, the affected requirements and the closure of discrepancies.
Carry one requirement through an evidence decision
Use an original mission thread requiring the accepted 600 Mbit dataset to pass from a payload record through flight software, a represented communications channel and the actual ground decoder within 240 s of a defined start marker. “Accepted” includes complete identifiers, integrity checks, calibration and timestamps. The time limit is a classroom allocation, not a regulatory requirement. Separate component throughput results support analysis, but do not by themselves exercise this whole chain.
For declared configuration R1, an integrated ground test supplies complete data and an elapsed-time estimate 238 s with expanded uncertainty U=4 s under the specified channel/processing load. Suppose the project has chosen an acceptance rule t+U≤240 s for this exercise. The upper comparison is 242 s, so this record does not establish acceptance by that rule, even though its central estimate is below the limit. This is not proof the physical system always exceeds 240 s, nor a guarantee attached to an uncertainty interval.
A bounded processing revision R2 reduces elapsed time in a fresh represented-chain test to 232 s with the same stated U, giving 236 s and satisfying the selected comparison for those conditions. Record the revision, start/stop definition, input dataset, useful output, environmental/load representation and uncertainty evaluation. A later packet-format change touches the flight/ground interface and requires revisiting decoder compatibility and timing; an unchanged mounting inspection may remain relevant if its mechanical configuration truly is unchanged.
A quarter-scale wind-tunnel model at the same gas temperature, speed and properties can match freestream Mach number while having one-quarter the Reynolds number. A pressure or lift observation can still inform a narrowly stated aerodynamic comparison, with sensitivity evidence for the mismatch. It cannot silently prove full-scale viscous drag, a flexible wing’s modes or the end-to-end dataset requirement. Different tests close different missing links.
| Question | Method/evidence | Scope that remains to check |
|---|---|---|
| Does the mount match the drawing? | Inspection of identified article/configuration | Dynamic load and interface response not established by appearance |
| Can flight and ground exchange the complete dataset? | End-to-end payload–software–channel–decoder test | Represented channel and load must cover intended conditions |
| Does elapsed time meet the selected rule? | Defined timing result plus uncertainty, compared with t+U≤240 s | Only the stated rule/configuration/conditions |
| Does a packet-format update preserve the claim? | Change-impact analysis plus affected interface/timing verification | Prior R1/R2 records are not automatically new-format evidence |
Reliability needs a mission and dependencies
Reliability is probability of adequate service over stated conditions and duration. A redundant path helps only if a surviving path has enough capacity, selection works and the failure assumptions apply. A shared power bus, command error or environment can defeat multiple nominally separate paths. Extra hardware also adds mass, power, heat and interfaces.
The figure separates an original conditional independent-failure model from a common event. Numerical probabilities are supplied teaching inputs, not estimates from test counts. A finite series of successful tests does not prove zero failure probability. Decisions must weigh uncertainty, consequences and the evidence available for the actual architecture and mission.
State what two paths actually provide
Supply two full-capacity paths, each with mission survival .96 conditional on no common event. Their failures are independent in that condition and either path suffices; switching is perfect in this model. Conditional system failure is .04²=.0016, so survival .9984. Now supply common-event probability .015 which disables both. Total survival is(1−.015)×.9984=.983424; failure is .016576.
A separate single-path comparison with supplied survival .96 has failure .04. In the two-path model, failure is .0016+.9984 p for common-event probability p, reaching .04 at p≈.0384615. This crossing does not establish an actual design choice: the single path may face its own dependencies, and assumed path survival, capacity and switching need evidence. The exercise makes the independence claim visible rather than proving redundancy adequate.
Now connect the retained probabilities to resources. Supply one full-capacity path with mass 2 kg and 15 W demand, versus a two-path arrangement of 3 kg and 30 W, including its assumed perfect selection. Let mission allocations be at most 3 kg, at most 25 W and failure probability at most 0.020 under the stated teaching model. The single path fits resources but its supplied 0.040 failure exceeds the probability allocation. At common-event p=0.015, two paths have failure 0.016576 and fit the mass cap, but exceed power by 5 W. Neither option satisfies all three allocations. Changing to standby, increasing power or adding a different architecture requires new power and failure evidence, rather than reusing the active-pair calculation unchanged.
Try the reasoning
Does a qualified design establish that every produced article is fault free?
Show the answer and reasoning
No. Article workmanship, configuration and acceptance evidence still matter.
Why is an end-to-end test different from separate component passes?
Show the answer and reasoning
It examines the connected interfaces and complete mission thread.
Can two devices sharing one vulnerable power source be assumed independent?
Show the answer and reasoning
No. The shared dependency must enter the failure model and verification.
A new represented-chain result is 235 s with U=6 s under the same chosen t+U≤240 s rule. Classify it, then state what a revised 230 s with U=6 s result would establish.
Show the answer and reasoning
235+6=241 s, so the first result does not establish acceptance by the rule. 230+6=236 s satisfies that rule for its recorded configuration and conditions. Neither comparison proves every future mission run meets 240 s or makes the uncertainty interval a hard guarantee.
Raise the illustrative power allocation to 35 W, but change the two-path common-event probability to p=0.05. Does the two-path arrangement now meet all allocations, and how does its model failure compare with 0.04?
Show the answer and reasoning
Its 30 W and 3 kg now fit the resource caps, but failure is 0.0016+0.9984×0.05=0.05152, exceeding both the 0.020 allocation and the single-path 0.040 comparison. Added nominal paths do not defeat the supplied shared dependency. These are assumed probabilities, not an inference from a handful of passed tests.
A packet-format software update leaves the mount and electrical connector unchanged. Which old evidence can be retained conditionally, and which fresh check is most directly missing?
Show the answer and reasoning
Identified mechanical fit/mount inspection can remain relevant if configuration review confirms no affected mechanical requirement. The payload-to-ground decoder data-integrity and timing thread is directly affected and needs change-impact analysis plus appropriate fresh verification. An unchanged connector shape does not establish software/interface semantics or mission delivery time.
What to carry forward
Bind every test and reliability conclusion to the requirement, configuration, environment and failure assumptions it actually supports.
References
Teaching resources and research supporting this lesson’s explanations and examples.