DRAFT SOO_AI for Mx Assistant.pdf

DRAFT

STATEMENT OF OBJECTIVES (SOO)

AI-Enabled Aircraft Maintenance Data Standardization & Troubleshooting – Pilot

1. Purpose. To evaluate the feasibility, performance, and operational integration of a commercial AI capability that cleanses and standardizes historical USAF aircraft maintenance data and provides prescriptive troubleshooting to maintainers, to inform an aircraft repair decision.

2. Background. The Department of the Air Force's reliance on the Integrated Maintenance Data System (IMDS) and the Reliability and Maintainability Information System (REMIS) is severely constrained by unstructured data. These databases house millions of records, critical qualitative details that are trapped in free-text "narrative" fields written by maintainers.

Because these entries lack standardization, suffer from rampant typos, and use non-standardized acronyms, advanced analytical tools cannot easily parse the data. This "dark data" prevents aircraft maintenance technicians, analysts, and engineers from conducting targeted automated data cleansing and analysis in a timely manner to aid in diagnosing an aircraft problem and comparing to historical effective and non-effective repairs.

Aircraft maintenance technicians and analysts require the ability to automatically ingest, clean, and analyze this unstructured text. Without this capability the technicians are left with a massive "blind spot" to quickly troubleshooting and repair aircraft.

Maintenance analysts are forced to manually review thousands of logs to identify "bad actor" parts or recurring failure modes—a process that is incredibly labor-intensive and slow.

Consequently, aircraft technicians and analyst teams miss historical repair data that increases aircraft downtime during troubleshooting and repair, delaying supply chain demands, and degraded mission-capable rates across the fleet.

Aircrew use of historical aircraft reported discrepancies and effective repairs will enable aircrew to precisely describe an identified problem. Providing aircraft maintenance technicians, a focused and detailed description of a problem with the aircraft enables efficient troubleshooting and diagnosis with a faster aircraft repair and return to service.

The combined AI capability to assist in aircrew debriefing actions leads to aircraft technicians completing troubleshooting and aircraft repairs faster and with greater precision. Culminating in greater aircraft availability.

3. Performance Objectives (Salient Characteristics).

# Objective Measurable Threshold

1 Automated data cleansing/standardization

Ingest unstructured records, parse the records and assign standardized Air

DRAFT

Transport Association codes to the discrepancy and corrective action narrative at ≥ 90% accuracy, without requiring pre-formatted source data

2 Discrepancy reporting guidance

Guide aircrew documentation of aircraft discrepancies using accurate fault reporting codes and descriptive narratives based on previously reported and corrected discrepancies.

3 Prescriptive troubleshooting

Provide repair recommendations from cleansed historical data, each with a stated confidence/probability based on the effectiveness of the historical repairs.

4 Natural-language query Allow technicians to query historical records in plain language

5 Chronic trend ID Identify top recurring codes driving chronic degradation

6 Rapid deployment Deploy and integrate with historical data within 30 days of award, enabling ≥ 5 months live testing

7 Security & hosting Operate in an IL4 compliant environment

8 Operational validation Demonstrate ability to deploy within the pilot timeline.