MIL-HDBK-217F · RAC 217Plus · ANSI/VITA 51.1

Reliability predictions in minutes,
not weeks.

AI-driven BOM ingestion classifies every component, runs the handbook model, modernizes it with RAC 217Plus and ANSI/VITA 51.1, and lets your real life data override the estimate — a prediction that reflects the article you're actually building, not a 1990s table.

No spreadsheets. No transcription errors. Audit-ready by construction.

COMPONENT · RCR07G103JSMIL-HDBK-217F §9.1
familyresistor, composition
P_rated0.25 W
stress ratio S0.42
T_ambient55 °C
π_T 1.6π_S 0.71π_P 1.0π_Q 3.0π_E 6.0
λ_b · π_T · π_S · π_P · π_Q · π_E3.42 FITs

// λ_b 0.0017 — MIL-HDBK-217F Notice 2, p. 9-2 · triangulated ✓ handbook ✓ golden BOM ✓ oracle

Standards: MIL-HDBK-217F Notice 2RAC 217PlusANSI/VITA 51.1Telcordia & FIDES in work
The shift

From a three-week spreadsheet to a traceable prediction.

Reliability prediction today means manually classifying every line of a BOM, hunting π-factors across handbook tables, and praying the transcription was right. One wrong cell invalidates the report.

The manual way

  • Classify every component by hand
  • Look up π-factors table by table
  • One model — raw 217F, unmodernized
  • No record of where a number came from
~3 weeks

With ReliabilitAI

  • BOM classified and parameterized on ingest
  • Modernized with RAC 217Plus + ANSI/VITA 51.1
  • Four parallel scenarios, life-data corrected
  • Every value traces to a handbook page
minutes
Four predictions, one BOM

The 1990s handbook is your baseline — not your verdict.

Every part returns four parallel λ values. The raw handbook number is just one of them, kept for comparability. The number you ship reflects the article as built. Built for satellite reliability — the method generalizes to any domain.

S1Default

As-built

217F + RAC + VITA where the part has known pedigree, then life-data corrected.

870 FITs
MTBF ≈ 1.15M hr
S2

Theoretical cheapest

Every part forced to the floor grade for your domain. "What if we built it all from COTS?"

1,510 FITs
MTBF ≈ 662k hr
S3

Theoretical max

Every part upgraded to the top grade for your domain + RAC modernization. The best-case bracket.

690 FITs
MTBF ≈ 1.45M hr
S4Baseline

Raw 217F

Unmodernized MIL-HDBK-217F. The standard number another tool would give you.

1,240 FITs
MTBF ≈ 806k hr

S1 is the customer-facing number. S4 is what everyone else ships. Life data corrects S1–S3, never S4.
Theoretical brackets default to space/commercial grades — retarget them to your domain.

Methodology

Three layers between the handbook and the truth.

Constant-failure-rate tables from the 1990s are a starting point. We treat them like one.

01

BaselineMIL-HDBK-217F Notice 2

The handbook model, computed exactly and cited to the page. This is the comparability number every reliability tool can reproduce.

02

ModernizeRAC 217Plus · ANSI/VITA 51.1

Updated failure-rate models and modern derating, applied where the part's pedigree warrants it — so a mil-spec or space-qualified part isn't penalized by a 30-year-old table.

03

Supersedeyour life data

Where you have vendor or field life data, observed reliability overrides the handbook estimate. Empirical evidence beats a generic prediction.

// Base engine is a swappable axis — Telcordia and FIDES in work. The scenario shape holds across engines.

How it works

Four steps from BOM to audit-ready report.

01 / INGEST

Upload the BOM

Drop in a BOM in any common format. Components are parsed line by line.

02 / CLASSIFY

AI classification

Each part is identified by family and its parametric values extracted.

03 / COMPUTE

Run the scenarios

Four parallel models run — baseline, modernized, and life-data corrected — each citing its source.

04 / TRACE

Traceable output

Complete λ and MTBF per scenario, every number sourced, ready for audit.

Capabilities

An engineering tool, not a black box.

Built so the question "where did this number come from?" is always answerable.

Multi-standard engine

217F, modernized — not frozen in 1995

MIL-HDBK-217F Notice 2 as the cited baseline, with RAC 217Plus and ANSI/VITA 51.1 modifiers composed on top. Telcordia and FIDES on the roadmap.

Ground truth

Every calculator triangulated against the handbook, a hand-verified golden BOM, and a VBA oracle. No single source is auto-trusted.

Life data

Observed beats estimated

Vendor and field data override the handbook where you have it.

Traceable

Every π cites its page

Each factor carries a handbook citation in the output.

Defense-ready

Tenant isolation & full audit log

Row-level security per tenant; every mutation written to an immutable audit trail.

Ground truth

Three independent checks. Any disagreement blocks the result.

01
Handbook citation

Every π-factor, formula and range cites its MIL-HDBK-217F page number.

02
Hand-verified golden BOM

Owner-calculated reference values, derived independently from the source.

03
VBA oracle

A read-only spreadsheet oracle cross-checks every computed value.

// resistor §9.1 — composition
λ_b = 0.0017 // p. 9-2
π_T = 1.6 // T_amb 55°C, p. 9-3
π_S = 0.71 // stress 0.42, p. 9-3
π_P = 1.0 // 0.25W, p. 9-4
π_Q = 3.0 // non-ER, p. 9-4
π_E = 6.0 // G_F, p. 9-5
λ_p = 3.42 FITs // ✓ all three sources agree
One engine, four surfaces

Use it where your team already works.

The reliability math lives in one framework-agnostic service layer. Every surface is a thin adapter over it.

web

Platform UI

Programs, parts, and predictions in the browser.

/api/v1

REST API

Drive predictions from your own pipeline.

mcp

MCP server

Expose the engine to AI agents with engineering context.

$ cli

CLI

Scriptable predictions, human tables or --format json.

Early access

Reliability predictions you can defend.

Built by a reliability engineer for teams who have to show their work. Request access and bring your first BOM.