Test & Evaluation Adequacy Kernel

The January 2024 Nunn-McCurdy breach[1] exposed gaps beyond cost. When the first Sentinel flight test slipped from late 2023 to 2027,[7] no framework existed to quantify which certification claims were invalidated or which evidence chains broke.

Three crises converge: the model validation gap (digital twins without flight data), the test facility void (no schedule for security testing infrastructure), and the warhead integration paradox (W87-1 needs Sentinel data; Sentinel needs W87-1). TEA-K addresses all three with explicit, testable artifacts.

Minuteman III reliability was built on more than 300 Glory Trip test launches since 1970,[12] with the most recent (GT 255) on March 3, 2026. Sentinel has 25 EMD/test missiles from a total buy of 659.[13] That is roughly 1/12th the statistical base.

Demonstrating high reliability through flight testing alone requires large sample sizes. The binomial relationship — n ≥ ln(1−C) / ln(R) — means that for typical targets, tens of successful flights are needed with zero failures. With 25 test units split between developmental and operational testing, pure frequentist demonstration is insufficient.

But Not All Legacy Data Transfers

TEA-K assesses which MMIII data can inform Sentinel reliability based on design commonality. This is a TEA-K analytical output — no published DoD transferability analysis exists in unclassified literature.

Propulsion, guidance, and ground system reliability must be established almost entirely from new Sentinel data.

DOT&E's FY2024[4] and FY2025 annual reports[5] flag the central risk: digital twins "aid, but do not obviate" live testing. Without physical flight test calibration, models remain unvalidated constructs that may diverge from reality.

HWIL can simulate 10,000 flights in a week. HWIL cannot simulate stage separation, pyrotechnic shock, vibration, or reentry plasma. The first Sentinel flight test is targeted for 2027 by the Air Force,[7] though GAO projects no earlier than March 2028.[3]

What Physical Anchors Exist Today

Propulsion testing has advanced substantially, providing partial model calibration:

These anchor propulsion and separation models. Trajectory, reentry, and full-system integration models remain USED_UNVALIDATED until first flight.

GAO-25-108466[2] (September 2025) found that the Air Force "has not developed a schedule for construction of a Sentinel test facility." GAO-26-108755[3] (February 2026) reinforced this, recommending that the Air Force complete "Sentinel launch facility test and evaluation activities early in the transition."

The Physical Security Systems Test Facility is essential for:

Without this facility, Operational Certification cannot complete. TEA-K encodes this as an explicit blocking condition: INFRA_OK = FALSE for affected rows until PSSTF functions are validated.

Sentinel faces a circular dependency between missile and warhead certification:

W87-1 Timeline

The Interim Solution

Sentinel will initially deploy with W87-0 warheads in Mk21 RVs.[13] This requires delta certification (proving W87-0 survives Sentinel's flight environment) and a critical dependency on the ICBM Fuze Modernization program[24] — the replacement Arming and Fuzing Assembly managed by USAF/NNSA with Sandia as technical design agent.

Additionally, Sentinel's confirmed MIRV upload capability requirement[7] means the Post Boost Vehicle's multi-RV interface logic must be tested across single and multiple warhead configurations.

DoDI 3150.02[27] (reissued December 17, 2024) defines four nuclear surety standards. The reissued instruction explicitly integrates cybersecurity, supply chain risk management, and emerging technology threat assessment into the nuclear surety program framework — elevating cyber to peer status with physical surety.[28] DAFMAN 91-118[10] specifies the probability thresholds:

These probabilities are impossible to demonstrate through testing. Demonstrating 10⁻⁹ empirically would require billions of test-hours. This creates the fundamental bifurcation that TEA-K enforces:

Abnormal Environment Testing

Sandia National Laboratories[31] maintains unique facilities for qualification:

Nuclear certification follows a formal gate structure per AFI 63-125[25] and DAFMAN 63-119.[26] TEA-K binds adequacy predicates to each gate:

Critical sequencing: Design Cert must be substantially complete before CDOT. CDOT must be issued before dedicated OT can begin. OT results feed Operational Cert. NSCCA completion is required for Design Cert on all safety-critical software.

Completion of Gate 4 results in the item being listed on the Master Nuclear Certification List (MNCL) — the sole authority[25] for nuclear-certified items, maintained by AFNWC.

The Triangle of Verification

Three distinct pillars produce different evidence types. Certification requires all three:

TEA-K synthesizes outputs from established frameworks into a unified adequacy contract. No new T&E methodology is invented — the novelty is in the integration.

The Master Predicate

Gate passage is a computed state, not a committee opinion. If any condition evaluates FALSE, the gate cannot pass.

All TEA-K components use mature technologies (TRL 8–9): structured databases, Bayesian statistics per MIL-HDBK-189C, graph databases for MVDG, and operational systems (MNCL, JDRS). No component requires research-grade development.

Critical dependency: MVDG transitions from largely UNVALIDATED to partially VALIDATED only after first flight test. If the 2027 target slips, Phase 2 extends accordingly. TEA-K's binding to IDX_FIRST_FLIGHT_TEST rather than a calendar date ensures this propagation is automatic.

Risk Assessment

Path Forward: TEA-K must be established during the current restructuring period. Milestone B re-certification is expected end of 2026.[6] Once a new baseline is approved without TEA-K conformance, the opportunity closes until test inadequacy forces another review.