{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp04_retrieval_first_paired20_20260802","cell_id":"layer_decay_and_expiration_management__aviation_aeronautics","hypothesis_id":"H3","search_queries":["flight data monitoring selective retention sensor channel flight phase segment event triggered archive","patent flight data recorder selective data retention flight phase anomaly","aircraft flight data adaptive sampling anomaly compression archival safety monitoring","flight operational quality assurance raw data retention event data reduced dataset","commercial flight data monitoring software selective raw data retention event segments archive","EUROCAE flight data monitoring data retention raw decoded flight files standard event analysis"],"sources":[{"source_id":"C1","title":"Easy Access Rules for Air Operations — Revision 24, March 2026","publisher":"European Union Aviation Safety Agency","url":"https://www.easa.europa.eu/en/document-library/easy-access-rules/online-publications/easy-access-rules-air-operations?erules-id=ERULES-1963177438-12216","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["EASA already requires an explicit FDM retention strategy aimed at the greatest practicable safety benefit.","Raw or decoded data must remain until significant events are analysed, at least 80% of applicable raw or decoded flight files should remain processable for two years, and de-identified significant-event analyses are retained longer.","The guidance connects event status, analysis completion, security, accountability, parameter provenance, and algorithm dependencies to retention decisions."]},{"source_id":"C2","title":"US6628995B1 — Method and system for variable flight data collection","publisher":"Google Patents; patent assigned to General Electric Company","url":"https://patents.google.com/patent/US6628995B1/en","source_class":"PATENT","claims_supported":["The patent identifies recorder-capacity pressure and redundant or low-value measurements as flight-data collection problems.","It varies sampling rates and parameter lists by flight phase, records transient phases more densely than steady cruise, and escalates sampling when anomalies occur.","It therefore anticipates phase-by-channel differentiation and anomaly-sensitive preservation, although at acquisition rather than post-flight disposition."]},{"source_id":"C3","title":"US20160318622A1 — Aircraft operational anomaly detection","publisher":"Google Patents; application assigned to Rosemount Aerospace Inc.","url":"https://patents.google.com/patent/US20160318622A1/en","source_class":"PATENT","claims_supported":["The application stores selected operational data sparsely during normal conditions and more frequently during detected anomalies.","Storage or transmission frequency can depend on ranked anomaly severity, supporting later maintenance and root-cause analysis while limiting storage and bandwidth.","It overlaps anomaly scoring and selective preservation but does not disclose age-based post-flight tiering or restore testing."]},{"source_id":"C4","title":"Flight Data Monitoring Software — FlightVue FDM","publisher":"Fliant","url":"https://www.fliant.com/flight-vue-fdm","source_class":"COMMERCIAL_FIRST_PARTY","claims_supported":["A commercial FDM product ingests every flight, detects exceedances and precursors, splits data by flight phase, and applies customizable event logic.","This establishes deployed-style phase segmentation and event processing, but the opened product documentation does not claim phase-by-channel expiration or selective archival."]},{"source_id":"C5","title":"Explaining Aviation Safety Incidents Using Deep Temporal Multiple Instance Learning","publisher":"arXiv","url":"https://arxiv.org/abs/1710.04749","source_class":"PRIMARY_RESEARCH","claims_supported":["The research mines precursors in multidimensional flight time series using real commercial-airline data.","It identifies weak supervision, temporal behavior, and absent time-step labels as challenges, supporting the concern that low-frequency precursors may not coincide with predefined exceedances.","Its dependence on temporal multivariate context supports testing reconstruction links before deleting individual segments."]},{"source_id":"C6","title":"Commission Regulation (EU) No 965/2012 — Air Operations, consolidated version","publisher":"EUR-Lex","url":"https://eur-lex.europa.eu/eli/reg/2012/965/2025-02-13/eng","source_class":"GOVERNMENT_OR_REGULATOR","claims_supported":["Following an accident, serious incident, or authority-identified occurrence, operators must preserve original flight-recorder data for 60 days or as otherwise directed.","The rule demonstrates an investigation override that can suspend ordinary rolling retention and constrains any automated disposition system."]}],"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"EASA event-sensitive FDM retention strategy","similarity":"Already differentiates retention according to significant-event status, completion of analysis, long-term trend value, security, and accountability while permitting less than universal long-term raw-file availability.","remaining_difference":"It operates mainly at whole-flight-file and event-analysis-dataset levels rather than independently expiring phase-by-sensor-channel segments through an age, access, anomaly, and reconstruction score.","source_ids":["C1","C6"]},{"name":"GE variable flight-data collection patent","similarity":"Uses flight phase, parameter identity, redundancy, diagnostic value, and anomaly triggers to preserve high-value data at greater resolution while reducing steady-cruise storage demand.","remaining_difference":"It changes what is captured onboard prospectively; H3 proposes lifecycle scoring and tiering after full-flight ingestion, including holds and later restoration.","source_ids":["C2"]},{"name":"Rosemount anomaly-responsive storage patent","similarity":"Ranks anomaly conditions and varies storage or transmission density to retain more diagnostic evidence without burdening capacity during normal operation.","remaining_difference":"It lacks age and access decay, archival tiers, investigation-linked dependency holds, and sampled archive-restore verification.","source_ids":["C3"]},{"name":"FlightVue phase segmentation plus precursor detection","similarity":"Commercial documentation combines whole-flight ingestion, automated phase splitting, event rules, and precursor detection in one FDM workflow.","remaining_difference":"The public documentation does not disclose selective expiration or cold-tier migration at phase-by-channel granularity.","source_ids":["C4","C5"]}],"overlapping_components":["flight-phase segmentation","sensor or parameter-channel selection","anomaly and exceedance evidence","ranked anomaly severity","reduced treatment of steady-cruise data","significant-event preservation exception","investigation hold override","raw-versus-derived-data distinction","retention strategy","storage-capacity optimization","precursor mining","algorithm-to-parameter dependency information"],"remaining_contrastive_claim":"After complete post-flight ingestion, an FDM system can independently decay and tier phase-by-sensor-channel segments using age, access, anomaly evidence, and an explicit cross-segment reconstruction graph—while honoring investigation holds and verifying restores—and thereby reduce retained bytes without degrading predefined-event or weak-precursor detection.","claim_falsifier":"A pre-2026 patent, product manual, standard, or documented airline deployment showing post-ingestion phase-by-channel expiration or tier migration driven by anomaly/value scoring, with linked-context holds and restore validation, would erase the remaining distinction; alternatively, a controlled replay showing that such pruning reduces event or precursor recall or misses the stated 95% restore target would falsify its performance claim.","problem_support":"STRONG","recommendation":"RESEARCH","world_novelty_boundary":"The broad idea is not novel: aviation already uses event-sensitive retention, phase segmentation, parameter selection, anomaly-triggered recording, reduced normal-condition data, and investigation overrides. This bounded six-query search did not find a direct source combining those elements as post-ingestion, age-and-access-aware phase-by-channel lifecycle management with reconstruction dependencies and restore testing; novelty, if any, is confined to that narrow orchestration and its measured safety-preserving byte reduction."}