Visual walkthrough — Verification methods — analysis, test, inspection, demonstration

3.6.28 · D2 · Physics › Spacecraft Structures & Systems Engineering › Verification methods — analysis, test, inspection, demonstra

Yeh page EK sawal ka jawab deta hai, bilkul ground up se:

Jab hum shake karte hain ek spacecraft ko yeh prove karne ke liye ki woh launch survive karta hai, hum use 25% zyada kyun shake karte hain jitna launch actually shake karta hai? 1.25 number kahan se aata hai?

Parent note ne rule bataya tha. Yahan hum use earn karte hain — har symbol, har bell curve, har square root, pehle picture ki tarah draw karke, phir use karte hain. Ek smart 12-saal-ka baccha jisne kabhi "sigma" nahi dekha, woh line one se follow kar sakta hai.

Hum is ek final statement tak pahunchenge: $\text{Qualification Level}=\text{Flight Limit}\times 1.25$

lekin tabhi jab hum samjhein ki har piece kya hai.

Step 1 — "Load" actually kya hota hai: ek ruler par ek single number

KYA HAI. Jo bhi spacecraft fly karta hai use uske rocket se jhakzore lagte hain. Hum "kitna hard jhakzora lagta hai" isko ek single number mein compress karte hain. Vibration ke liye hum ise shaking strength kehte hain, g mein measure kiya jaata hai (Earth ki gravity ke multiples). Ise $x$ kahte hain.

• $x$ = shaking strength (g mein). Bada $x$ = zyada rough ride.

"RMS" ke baare mein ek baat. Parent note kehta hai "8 g RMS random vibration." Random shaking ek steady push nahi hai — yeh ek second mein hazaaron baar upar-neeche jhanjthanati hai, toh iska plain average zero hota hai (utna upar push karta hai jitna neeche). Kisi cheez ka ek honest "size" pane ke liye jo average karke zero aaye, hum har wiggle ko square karte hain (squaring se upar aur neeche dono positive ho jaate hain), un squares ka average nikalte hain, phir g mein wapas aane ke liye square root lete hain. Woh teen-word recipe — Root of the Mean of the Squares — RMS hai. Toh "8 g RMS" jhanjthanaahat ki typical effective strength hai. Hamara $x$ yahi RMS number hai.

YAHAN SE KYUN SHURU KAREIN. Pehle hum margins aur safety factors ki baat karte hain, hume kuch measure karne ki zaroorat hai. Baaki sab kuch bas "kahan $x$ land karta hai, aur hum kahan test karte hain?" hai. Toh hum ek ruler draw karte hain.

PICTURE. Orange tick dekho. Rocket engineers hume Flight Limit batate hain — sabse rough ride jo koi bhi launch dene ki expect ki jaati hai. Yeh ruler par ek fixed point hai, $x$ ki ek particular value.

Step 2 — Reality ek number nahi hai: yeh ek cloud hai

KYA HAI. Yeh woh twist hai jo poori derivation ko force karta hai. Koi bhi do flights identical nahi hoti, aur koi bhi do spacecraft identical nahi hote. Ek hi satellite do baar banao aur ek doosre se thoda stiff hoga (ek weld ek baal mota, ek bolt thoda tight). Ek hi rocket do baar fly karo aur ek launch thoda harder shake karega doosre se (mausam, fuel, wind).

Toh $x$ ki real value jo ek given unit feel karta hai woh single tick $\mu$ nahi hai. Yeh $\mu$ ke aas-paas scattered possible $x$-values ka pura spread hai.

YAHAN KYU MATTER KARTA HAI. Agar reality ek exact number hoti, hum exactly usi number par test karte aur ho jaata. Hum fun ke liye harder test nahi karte — hum harder test karte hain kyunki hum nahi dekh sakte ki koi particular flight ya unit $\mu$ ke kis taraf land karega. Margin exist karta hai purely is scatter ko cover karne ke liye.

PICTURE. Single orange tick ek possible $x$-values ke cloud mein khil jaata hai — chhote dots ka ek pile, beech mein sabse ghana, door jaate jaate patla hota jaata hai. Zyaatar flights $\mu$ ke paas land karti hain; kuch unlucky waale kaafi upar land karte hain.

Step 3 — Cloud ko naam dena: bell curve aur uski width $\sigma$

KYA HAI. $x$-dots ke us pile ko smooth karo aur tumhe famous bell curve milti hai (normal distribution). Do numbers ise completely describe karte hain:

• $\mu$ (Greek "mu") — center: woh $x$-value jahan peak baithti hai. Yeh hamara Flight Limit tick hai.
• $\sigma$ (Greek "sigma") — width: cloud $x$ mein side-ways kitna spread karta hai. Chhota $\sigma$ = tight, confident pile. Bada $\sigma$ = fat, uncertain smear.

BELL KYU AUR KUCH OONCHI KYU NAHI. Strictly, $x$ negative kabhi nahi ho sakta — tum kisi cheez ko zero se kam strength se shake nahi kar sakte. Ek perfect bell curve zero se neeche probability ka ek chhota sliver leak karta hai, toh yeh sirf ek approximation hi hai. Yahan yeh achi hai kyunki hamara scatter chhota hai: $\sigma$ $\mu$ ka lagbhag 10% hai, toh "zero" center se poore $10\sigma$ neeche baithta hai — tail par itni door ki leaked probability bilkul negligible hai (ek trillion trillion mein ek se bhi kam). Ek comfortably-positive mean ke aas-paas chhote scatter ke liye, neat symmetric bell honest, standard choice hai. (Agar $\sigma$ $\mu$ ke comparable hota toh hum ek one-sided model jaise log-normal pe switch karte — lekin hum us regime ke qareeb bhi nahi hain.)

$\sigma$ KYU CHAHIYE. Margin widths mein measure hoga. "Center se kitni door jaane par safe hoon?" ka jawab tab tak nahi hai jab tak pata na ho ki uncertainty ka ek step kitna wide hai. Woh step exactly $\sigma$ hai.

PICTURE. Bell, $\mu$ par centered. Teal double-arrow ek $\sigma$ hai — beech se ek standard step of uncertainty bahar. Shaded areas dhyaan do: saare outcomes ka 68% $\pm 1\sigma$ ke andar aata hai, aur 99.7% $\pm 3\sigma$ ke andar. Woh percentages is exact shape mein baked in hain.

Step 4 — Do clouds, ek nahi: manufacturing scatter AUR environment scatter

KYA HAI. Hamari uncertainty do independent sources se aati hai, aur parent note ne dono diye:

• Manufacturing scatter ${\sigma }_{\text{mfg}}=0.1$ — units ek doosre se $\pm 10\%$ alag nikalte hain.
• Environment scatter ${\sigma }_{\text{env}}=0.1$ — launches ek doosre se $\pm 10\%$ alag shake karte hain.

($0.1$ ka matlab hai "Flight Limit ka 10%," toh hum un units mein kaam kar rahe hain jahan $\mu =1$ aur $0.1$ ka ek $\sigma$ matlab 10%.)

DO ALAG NUMBERS KYU. Ek rough launch par ek stiff unit dono effects se akele se bura hai. Margin size karne ke liye hume combined scatter chahiye, har piece alag nahi. Toh Step 5 ka sawal hai: do clouds kaise add hote hain?

PICTURE. Do narrower bells side by side — plum wala manufacturing ke liye, teal wala environment ke liye — har ek ki apni width $0.1$. Hume inhe ek mein fuse karna hai.

Step 5 — SQUARES KYU ADD KARTE HAIN (variances add hote hain; statistics mein Pythagoras chhupa hai)

KYA HAI. Do independent scatters combine karne ke liye hum simply $\sigma$'s add nahi karte ($0.1+0.1=0.2$ galat hai). Balki hum unke variances add karte hain (${\sigma }^2$, Step 3 se squared widths) aur square root lete hain: ${\sigma }_{\text{total}}=\sqrt{{\sigma }_{\text{mfg}}^2+{\sigma }_{\text{env}}^2}$

Term by term:

• ${\sigma }_{\text{mfg}}^2$ — manufacturing variance (right triangle ki ek leg).
• ${\sigma }_{\text{env}}^2$ — environment variance (doosri leg).
• $\sqrt{\cdots }$ — squaring ko undo karta hai taaki plain width wapas mile (hypotenuse).

SQUARES KYU AUR PLAIN SUM KYU NAHI? Yeh Step 3 ka deep fact hai: independent sources ke liye, variances (${\sigma }^2$'s, squared distances) add hote hain, $\sigma$'s khud nahi. Geometrically iska matlab hai dono effects "different directions" mein point karte hain — jaise East phir North chalna. 0.1 East aur 0.1 North chalo aur tum ghar se 0.2 dur nahi ho; tum $\sqrt{0.1^2+0.1^2}$ door ho. Independent uncertainties exactly perpendicular distances ki tarah combine hote hain: Pythagoras se. $\sigma$'s seedha add karna yeh maan lena hoga ki dono ki worst cheez hamesha saath hoti hai, jo bahut zyada pessimistic hai (aur bahut expensive bhi).

PICTURE. Ek right triangle. Horizontal leg ${\sigma }_{\text{mfg}}$ hai, vertical leg ${\sigma }_{\text{env}}$ hai, aur slanted hypotenuse — real combined uncertainty — ${\sigma }_{\text{total}}$ hai.

Ab numbers plug karo: ${\sigma }_{\text{total}}=\sqrt{0.1^2+0.1^2}=\sqrt{0.02}\approx 0.14$

• $0.1^2=0.01$ each — do variances (squared legs).
• $0.01+0.01=0.02$ — total variance (squared hypotenuse).
• $\sqrt{0.02}\approx 0.14$ — combined width: 14%, 20% nahi.

Step 6 — Almost sab kuch cover karna: $3\sigma$ reach

KYA HAI. Ab hum decide karte hain ki cloud ka kitna hissa cover karna hai. Industry $3\sigma$ choose karti hai — center se teen widths upar pahunchne se 99.7% cases capture ho jaate hain (Step 3 ki shaded bell yaad karo). Test level hai: $\text{Test Level}=\underset{\text{center}=1.0}{\underbrace{\mu }}+\underset{\text{safety reach}}{\underbrace{3{\sigma }_{\text{total}}}}$

Plug in karo: $\text{Test Level}=1.0+3(0.14)=1.0+0.42\approx 1.4$

Term by term:

• $\mu =1.0$ — Flight Limit khud (un units mein kaam kar rahe hain jahan limit 1 hai).
• $3{\sigma }_{\text{total}}=0.42$ — combined scatter ke teen standard steps, hamari safety reach.
• Sum $\approx 1.4$ — flight limit se 40% upar test karo taaki saare real flights ka 99.7% swallow ho jaye.

$3\sigma$ KYU AUR $1\sigma$ YA $5\sigma$ KYU NAHI. $1\sigma$ sirf 68% cover karta hai — teen mein se ek flight hamare test se bahar jaati, unacceptable. $5\sigma$ 99.9999% cover karta hai lekin ek monstrous, hardware-destroying shake demand karta hai bahut chhoti extra safety ke liye. $3\sigma$ sweet spot hai: survivable cost par near-total coverage.

PICTURE. Full bell jisme orange test-level tick $\mu +3{\sigma }_{\text{total}}$ par planted hai, right tail par bahut door. Us tick ke left mein sab — reality ka 99.7% — hamare test se covered hai.

Step 7 — 1.4 se real-world 1.25 tak (degenerate aur practical cases)

KYA HAI. Clean statistics, hamare illustrative $\pm 10\%$ inputs feed karne par, $\approx 1.4$ dete hain. Published industry standard (jaise NASA-STD-7001 / GEVS aur MIL-STD-1540 heritage) 1.25 hai — historically vibration spectrum par +3 dB qualification margin ke roop mein express kiya jaata tha, jo RMS level ko $\sqrt{2}\approx 1.41$ se multiply karta hai... energy mein, lekin flat-level factor jo customarily apply aur quote kiya jaata hai woh $1.25$ hai. Hamare 1.4 se 1.25 tak ka gap koi fudge nahi hai: yeh reflect karta hai ki $0.1/0.1$ figures worst-case teaching numbers the, aur decades of flight data dikhate hain ki mature, well-characterised hardware aur launch environments isse kam scatter karte hain.

1.25 KYU CHOOSE KIYA JAATA HAI. Teen edge/degenerate cases ise pin down karte hain:

1. Assumed se kam scatter — $0.1/0.1$ pair deliberately conservative hai. Jab flight-heritage data tighter, empirically-measured scatter deta hai (roughly ${\sigma }_{\text{mfg}}\approx {\sigma }_{\text{env}}\approx 0.06$ characterised hardware ke liye), toh ${\sigma }_{\text{total}}=\sqrt{0.06^2+0.06^2}\approx 0.085$, aur $1+3(0.085)\approx 1.25$. Toh standard ka 1.25 factor wahi $3\sigma$ jawab hai jab tum assumed-worst ki jagah measured scatter plug in karte ho — exactly yahi wajah hai ki standards bodies wahaan settle huin.
2. Zero scatter (perfect world) — agar ${\sigma }_{\text{mfg}}={\sigma }_{\text{env}}=0$, toh ${\sigma }_{\text{total}}=0$ aur Test Level $=\mu =1.0$. Koi uncertainty nahi ⇒ koi margin nahi. Yeh sanity check hai: poora margin exist karta hai sirf scatter ki wajah se.
3. Acceptance vs qualification. Design ek baar $1.25\mu$ par ek qualification unit par prove hoti hai. Phir har flight unit ko acceptance test $1.0\mu$ par milta hai — manufacturing defect pakadne ke liye itna kaafi, bina kisi achhe unit ko overtress kiye jise actually fly karna hai.

PICTURE. $1.0$ se $1.4$ tak number line jisme teen landmarks hain: $1.0$ (acceptance / zero-scatter), $1.25$ (qualification, standard, measured scatter se), aur $1.4$ ($3\sigma$ result pessimistic $0.1/0.1$ teaching inputs ke liye). Chosen band 1.25 par baithta hai.

Ek-picture summary

Upar sab kuch, ek single frame mein: Flight Limit tick $x$-values ke cloud mein khil jaata hai, cloud ko width $\sigma$ milti hai, do variances Pythagoras se ${\sigma }_{\text{total}}\approx 0.14$ mein combine hote hain, teen widths 99.7% cover karne ke liye bahar nikalte hain, aur worst-case ki jagah measured scatter use karna standard factor ko 1.25 tak kheench laata hai.

Dekho bhi: Verification Methods — Analysis, Test, Inspection, Demonstration · Vibration Testing · Acceptance Testing · Margin Philosophy · Requirements Development · Traceability Matrix · Model Validation