FAA Airworthiness Directives: The Federal Record of Mandatory Aircraft Safety Fixes

When investigators trace a crash to a cracked fan disk or a misrouted wire, the finding does not stay with the one airplane that fell. The FAA writes a rule—an Airworthiness Directive—and that rule reaches every aircraft of the same design, anywhere in the world that recognizes it. An AD is not advice and not a recommendation; it is a federal regulation, and once one applies to an aircraft, that aircraft may not legally be flown until the directive is complied with. This dataset captures the FAA's rulemaking actions, the Airworthiness Directives chief among them: roughly 22,900 rules and directives, each the legal hinge that turns a single safety finding into a mandatory, fleet-wide fix.

This article covers what the dataset is and how 14 CFR Part 39 makes an Airworthiness Directive a binding regulation rather than a suggestion; the anatomy of an AD—the unsafe condition, the affected fleet, the required action, and the compliance time; the safety chain that produces a directive, from an NTSB accident finding or a service-difficulty trend or a manufacturer's discovery through to a fleet-wide order; the emergency AD and the grounding-level actions, with the Boeing 737 MAX as the defining recent case; the recurring directives that govern aging-aircraft structures and engine and rotorcraft components; how the actions table joins to the aircraft registry by make and model and complements the accident-investigation record; the analytical uses, from ranking the models and components that draw the most directives to tracing the regulatory trail from finding to fix; a Python workflow that pulls FAA rules from the Federal Register, separates Airworthiness Directives from other rules, and tallies directives by manufacturer; and the caveats that every analyst must internalize before drawing conclusions.

What the dataset is

The FAA does not only certify aircraft and license pilots; it issues a continuous stream of rulemaking actions that govern how aircraft are built, maintained, and flown. This dataset captures those FAA actions—new and amended regulations, special conditions, airspace rules, and, dominating the record by count, the Airworthiness Directives that mandate fixes to unsafe conditions found in the existing fleet. An Airworthiness Directive is the instrument the FAA uses to discharge its statutory duty over continued airworthiness: the obligation to ensure that an aircraft type which was safe enough to certify stays safe across the decades it remains in service, as fatigue accumulates, as defects surface, and as accidents reveal flaws that certification testing missed.

In our database this record is stored as the table faa_actions, comprising roughly 22,900 FAA rules and directives, keyed by document number and publication date. The grain is one row per rulemaking action: a single Airworthiness Directive—say, a 2019 order on a particular Boeing model's wing structure—is one row, carrying the document that defines the unsafe condition, the affected aircraft, the required action, and the date. Because an AD names the aircraft it applies to, the table is joinable to the aircraft registry by make and model. The columns capture what kind of action it is, what it targets, and when it took effect:

document_number       -- Federal Register / FAA document identifier
action_type           -- airworthiness directive, final rule, special condition, etc.
title                 -- the rule title (names the affected make/model on ADs)
ad_number             -- the AD docket/number, when the action is an AD
publication_date      -- date published in the Federal Register
effective_date        -- date the rule/directive takes legal effect
is_emergency          -- flag for an emergency AD (effective immediately)
cfr_part              -- the CFR part(s) cited (14 CFR Part 39 for ADs)
make                  -- aircraft / engine manufacturer named by the action
model                 -- affected aircraft model(s) or product line
abstract              -- summary of the unsafe condition and required action
html_url              -- link to the published rule text

The load-bearing columns are action_type, make, and model. The first separates the Airworthiness Directives—the mandatory safety fixes—from the FAA's many other rulemaking actions, which include airspace redesignations, operating-rule amendments, and airworthiness-standard changes that are not ADs at all. The make and model fields are what tie a directive to a physical population of aircraft: an AD on a turbine engine names the engine manufacturer and model, an AD on a transport airplane names the airframer and the model series, and those names are the bridge to the civil aviation registry, which records which actual, tail-numbered aircraft are of that make and model. The is_emergency flag marks the rare directives that bypass the normal notice-and-comment timeline because the danger is too immediate to wait, and the effective_date distinguishes when the legal obligation attaches from when the rule was merely published. Together these turn a stream of published rules into a structured record of which designs the FAA judged unsafe, what it ordered done about it, and how urgently.

14 CFR Part 39 and why an AD is a regulation

The single fact that gives this dataset its weight is that an Airworthiness Directive is a federal regulation, not guidance. ADs are issued under 14 CFR Part 39, the part of the Code of Federal Regulations devoted entirely to Airworthiness Directives. Part 39 does two things in a few spare sections: it establishes that the FAA issues an AD when it finds that an unsafe condition exists in a product—an aircraft, aircraft engine, propeller, or appliance—and that the condition is likely to exist or develop in other products of the same type design; and it makes compliance mandatory by providing that no person may operate a product to which an AD applies except in accordance with that AD. That second clause is the teeth. It means an AD does not merely advise an operator to do something—it makes the aircraft un-flyable, as a matter of law, until the directive's required action is taken.

The two-part test in Part 39 is what distinguishes an AD from the rest of the safety system. A manufacturer's service bulletin tells operators that a part should be inspected or replaced, but a service bulletin is the manufacturer's recommendation, and complying with it is, for most operators, voluntary. An AD is what happens when the FAA decides the condition is serious enough and widespread enough that the fix must be mandatory across the fleet. The FAA frequently issues an AD precisely to make a manufacturer's service bulletin compulsory—the AD adopts the bulletin's inspection or modification by reference and converts it from advice into law. This is the mechanism by which a discovery made by one engineer at one company becomes a legal obligation on every operator of the affected type: the service bulletin identifies the fix, and the AD makes it mandatory.

Because an AD is a rule, it normally moves through the rulemaking process: the FAA publishes a notice of proposed rulemaking in the Federal Register, takes public comment from operators and manufacturers, and then issues the final rule. That is why the Federal Register is the authoritative publication channel for ADs and why this dataset can be assembled from it. Part 39 also builds in a relief valve: an operator who cannot comply with an AD as written, or who has an equally effective alternative, can apply for an alternative method of compliance (AMOC), which the FAA approves on a case-by-case basis. The AMOC mechanism is what keeps a rigid, fleet-wide mandate workable for the genuine variety of configurations and operations across the affected population—but it does not soften the core rule, which remains that the aircraft does not fly until the unsafe condition is addressed by the AD or an approved alternative.

Anatomy of an Airworthiness Directive

Every AD answers four questions, and the structure of the data follows them: what is wrong, which aircraft are affected, what must be done, and by when. Reading those four elements is how an analyst—or a maintenance organization—extracts the operational meaning of a directive from its regulatory text.

The first element is the unsafe condition. Every AD opens by describing the specific defect or hazard that prompted it: a fatigue crack found at a particular wing-to-fuselage attachment, corrosion in a fuel-tank boundary, a flight-control actuator prone to jamming, an engine fan blade that can fracture, a battery that can overheat. The unsafe-condition statement is the regulatory finding that justifies the whole directive, and it is what links the AD back to the safety event—the accident, the incident, the in-service failure—that surfaced the problem. The second element is the applicability: the precise definition of which products the AD reaches, by make, model, and often serial-number range or configuration. Applicability is the embodiment of the Part 39 likely-to-exist test—the AD applies to exactly the population in which the FAA believes the condition is present or could develop, which is why the make and model fields are the join key to the registry.

The third element is the required action, which is what the AD actually orders: a repetitive inspection of a part at stated intervals, a one-time modification, the replacement of a component, an operating limitation, or, at the extreme, a prohibition on flight. The required action is where the AD ranges from a minor recurring check to a major and expensive structural rework, and it is what determines the real-world burden the directive imposes on operators. The fourth element is the compliance time: the deadline by which the action must be taken, expressed in calendar time, in flight hours, or in flight cycles (one takeoff and landing). The compliance time is the FAA's judgment about urgency—a condition that could cause an immediate catastrophic failure carries a short, hours-based deadline or immediate grounding, while a slowly developing fatigue issue may allow hundreds or thousands of flight cycles before action is required. Together these four elements convert a safety finding into an enforceable instruction: a named hazard, a named fleet, a named fix, and a named deadline.

The safety chain: from finding to fleet-wide fix

An Airworthiness Directive is rarely the start of anything—it is the codifying step at the end of a longer safety process, and understanding that chain is what lets an analyst read the AD record as the output of the entire aviation-safety system rather than as an isolated stream of rules. Directives flow from several distinct upstream sources, and most ADs can be traced to one of them.

The most consequential source is an accident or serious-incident investigation. When the National Transportation Safety Board investigates a crash, its findings frequently identify a design or maintenance deficiency—and the NTSB then issues a safety recommendation, often addressed to the FAA, calling for corrective action. The FAA, holding the regulatory authority the NTSB lacks, responds by issuing an AD that makes the corrective action mandatory across the affected fleet. This is the canonical path: an accident reveals a flaw, the investigator recommends a fix, and the regulator turns the recommendation into a binding directive. A second, higher-volume source is the service-difficulty trend: operators and repair stations report problems they encounter in service through service difficulty reports and similar channels, and when the FAA and the manufacturer see a defect recurring across many aircraft—cracks turning up at the same location, the same component failing repeatedly—that trend can prompt an AD before any accident occurs at all. This is the preventive face of the directive system: catching a fleet-wide problem from the accumulation of small in-service failures rather than from a catastrophe.

A third source is the manufacturer's own discovery. A type-certificate holder, through its continued-airworthiness obligations, its own testing, or analysis of in-service data, may identify an unsafe condition in a product it designed; it issues a service bulletin, and the FAA, agreeing that the condition meets the Part 39 test, issues an AD mandating the bulletin's remedy. A fourth path runs through harmonization with foreign authorities: because aircraft and engines are designed and built across borders, an unsafe condition identified by the state of design—the European Union Aviation Safety Agency for an Airbus, for example, or Transport Canada for a Bombardier—is mirrored by the FAA, which issues a corresponding AD so that the same fleet operating under FAA jurisdiction is covered. Whichever path produces it, the AD is the point where a finding becomes a fleet-wide legal obligation, and the dataset is the record of every time that conversion happened.

Emergency ADs and the 737 MAX

Most Airworthiness Directives move through the normal rulemaking timeline, with a proposal, public comment, and an effective date set weeks or months out. But some unsafe conditions are too dangerous to wait, and for those the FAA issues an emergency airworthiness directive—a directive that takes effect immediately, often issued directly to the affected operators before it is even published in the Federal Register, because the agency judges that the time the normal process would consume is itself an unacceptable risk. An emergency AD can require action before the next flight, ground a fleet outright, or impose an immediate operating limitation. The is_emergency flag in the dataset marks these directives, and their frequency and distribution are themselves a signal—a cluster of emergency ADs against a particular model or component is a sign of a problem the FAA regarded as acute rather than chronic.

The defining recent example is the Boeing 737 MAX. After two fatal crashes of the type within five months—Lion Air Flight 610 in 2018 and Ethiopian Airlines Flight 302 in 2019—investigators implicated the Maneuvering Characteristics Augmentation System (MCAS), a flight-control function that could repeatedly push the aircraft's nose down based on a single faulty sensor input. In March 2019 the FAA issued an emergency order grounding the entire US 737 MAX fleet, joining aviation authorities around the world in halting the type. The grounding lasted roughly twenty months; the aircraft did not return to service until late 2020, after the FAA mandated a package of changes—to the MCAS software, to the flight-control architecture, to crew procedures and training—through the rulemaking record. The MAX episode is the textbook illustration of the AD system operating at its most forceful: an accident finding, an emergency action that grounded an entire fleet effective immediately, and a return-to-service governed by a mandatory directive that the aircraft could not legally fly without satisfying. It also illustrated the system's limits—the questions it raised about how the type was certified in the first place fall outside the AD record and into the certification and oversight reforms that followed.

The MAX was an extreme case, but grounding-level and immediate-action directives recur across aviation history whenever a condition threatens imminent, catastrophic failure: lithium-ion battery overheating events that grounded a widebody fleet, engine fan-blade failures that triggered urgent fleet-wide inspections of a popular turbofan, and flight-control or structural findings serious enough to demand action before the next flight. For an analyst, the emergency ADs are the most consequential rows in the dataset—the rare moments when the continuing-airworthiness machinery shifted from routine maintenance scheduling to halting airplanes on the ground.

Aging aircraft, engines, and recurring directives

Not every AD is a dramatic grounding. The large majority are the quieter work of keeping an aging fleet safe, and two clusters dominate that work: structural directives on aging aircraft and component directives on engines and rotorcraft. Understanding these clusters is what explains the bulk of the directive count and its concentration in particular models.

The aging-aircraft program is the FAA's response to a problem that certification cannot fully anticipate: metal fatigue and corrosion accumulate over the structural life of an airframe, and as fleets remained in service far longer than their designers originally envisioned, the risk of structural failure from fatigue cracking grew. A 1988 in-flight structural failure—in which a large section of an aging passenger jet's upper fuselage tore away in flight—crystallized the concern and drove a sustained program of supplemental structural inspections, fatigue-management requirements, and a steady stream of ADs targeting specific cracking sites on specific older models. Many of these are repetitive inspection ADs: rather than a one-time fix, they require operators to inspect a known fatigue-prone location at recurring intervals— every so many flight cycles—for the remaining life of the aircraft. Repetitive ADs are a distinctive feature of the data because a single directive imposes an obligation that recurs indefinitely, and they concentrate on the high-cycle, long-lived transport models that have flown the most.

The engine and rotorcraft cluster reflects the reality that the most safety-critical components fail in characteristic ways. Turbine engine ADs frequently target rotating parts—fan blades, disks, hubs—whose failure can be uncontained, sending fragments through the airframe; these directives mandate inspections, life limits, or replacement of specific part numbers, and a finding on one engine model propagates as an AD across every airframe that uses it. Rotorcraft (helicopter) ADs are their own dense category, because helicopters subject critical components—rotor blades, gearboxes, mast and hub assemblies, tail-rotor parts—to severe, repeated stress, and the failure of any of them is typically catastrophic, so the directive cadence on rotorcraft components is correspondingly intense. Across both clusters, the pattern is the same: the AD record concentrates on the parts and the models where the consequence of failure is gravest and the population at risk is largest, which is exactly why tallying directives by make, model, and component is so revealing.

Joining to the aircraft registry and the accident record

The faa_actions table is most valuable not in isolation but as one node in the integrated aviation-safety record, and its make-and-model fields are the keys that connect it to the other datasets. Two joins matter most, and each answers a question the AD record alone cannot.

The first join is to the aircraft registry. The FAA civil aviation registry records every US-registered aircraft by tail number, with its make, model, year, owner, and airworthiness status. An Airworthiness Directive, by contrast, names an abstract population—a make and model series, sometimes a serial-number range—but not the individual tail numbers it reaches. Joining the actions table to the registry by make and model is what converts a directive from a rule about a design into a rule about a countable, real-world fleet: it answers how many actual registered aircraft a given AD applies to, where they are, and who operates them. That join is the foundation for any assessment of a directive's reach and burden—an AD on a model with thousands of registered examples is an enormous compliance undertaking, while one on a rare type touches a handful of aircraft. It also enables the inverse lookup that a maintenance organization actually needs: given a specific aircraft's make and model, which ADs apply to it.

The second join is to the accident and incident record. Because so many ADs originate in an accident finding or a service-difficulty trend, relating the directive record to the NTSB accident database and the FAA's broader aviation-safety data closes the loop on the safety chain. By make and model—and, where the documents allow, by the specific component or system implicated—an analyst can trace the path from a recorded accident, through the investigative finding, to the AD that the finding produced, and then forward again to whether the directive succeeded in preventing recurrence. That end-to-end view is what turns a pile of separate datasets into a picture of how the safety system actually performs: did the accident lead to a directive, did the directive reach the fleet, and did the condition stop killing people. The AD record is the regulatory pivot in the middle of that chain, and it is only legible as such when it is joined to what came before and after it.

Analytical uses

A national, date-stamped, make-and-model-resolved record of mandatory safety fixes supports a distinctive set of analyses that the accident data alone cannot deliver.

Which models and components draw the most directives is the most immediate use. Tallying ADs by manufacturer, model, and—where the text allows—by the affected system or component reveals where the continued-airworthiness burden concentrates. A high directive count is not a simple verdict on a design's safety: it is confounded by fleet size (the most-flown models accumulate the most in-service experience and therefore the most findings) and by age (older types have had more time to generate directives). But normalized against the registered fleet and the years in service, the directive count becomes a meaningful measure of how much ongoing intervention a design requires to stay safe.

How often emergency ADs are issued exploits the emergency flag to measure acuteness rather than volume. The rate of emergency directives, their clustering by model or component, and their timing relative to accidents distinguish the conditions the FAA judged immediately dangerous from the routine maintenance directives that make up the bulk of the record—and a spike of emergency ADs against a single type is one of the clearest early signals of a serious, emerging problem. Tracing the regulatory trail from a safety finding to a fleet-wide fix is the deepest use: ordering accidents, recommendations, and directives in time for a given make and model reconstructs how long it took for a finding to become a mandate, how completely the mandate addressed the finding, and whether the fix held. Finally, tracking continued-airworthiness enforcement over time uses the publication and effective dates to study how the cadence of directives shifts with the aging of the fleet, the introduction of new types, and the major safety-program reforms that followed events like the aging-aircraft failures and the 737 MAX grounding.

Python workflow: pulling ADs from the Federal Register

The script below pulls FAA rules from the Federal Register's public API over a date range, separates the Airworthiness Directives from the FAA's other rulemaking actions on the basis of their titles and abstracts, and computes two of the core metrics: the share of FAA rules that are ADs and the count of directives per year, plus a tally of directives by aircraft manufacturer. No API key is required. The Federal Register is the authoritative publication channel for ADs as final rules; for richer fields—the full required action, the precise applicability, alternative methods of compliance—the FAA's Dynamic Regulatory System (DRS) and its AD listings hold the complete, structured AD documents, and a production pipeline should reconcile the Federal Register stream against them.

import requests, re
import pandas as pd
from collections import Counter

# Federal Register public API -- no key required.
# FAA Airworthiness Directives are published as final rules in the
# Federal Register under agency "federal-aviation-administration".
# Each AD is a rule citing 14 CFR Part 39; the document text and the
# affected make/model live in the rule body and the abstract.
FR_API = "https://www.federalregister.gov/api/v1/documents.json"


def faa_rules(year_from, year_to, per_page=1000):
    # Pull FAA rules over a date range, paging through the result set.
    docs, page = [], 1
    while True:
        params = {
            "conditions[agencies][]": "federal-aviation-administration",
            "conditions[type][]": "RULE",
            "conditions[publication_date][gte]": f"{year_from}-01-01",
            "conditions[publication_date][lte]": f"{year_to}-12-31",
            "per_page": per_page,
            "page": page,
            "fields[]": ["document_number", "title", "publication_date",
                         "abstract", "regulation_id_numbers", "html_url"],
        }
        r = requests.get(FR_API, params=params, timeout=120)
        r.raise_for_status()
        batch = r.json().get("results", [])
        if not batch:
            break
        docs.extend(batch)
        page += 1
    return pd.DataFrame(docs)


def is_ad(row):
    # An Airworthiness Directive almost always says so in its title, and
    # cites Part 39. Separate ADs from other FAA rules (airspace,
    # operating rules, certification standards) on those two signals.
    title = (row.get("title") or "").lower()
    abstract = (row.get("abstract") or "").lower()
    return ("airworthiness directive" in title
            or "airworthiness directive" in abstract)


def manufacturer(row):
    # Crude make extraction from the AD title, which conventionally names
    # the type-certificate holder, e.g. "Airworthiness Directives; The
    # Boeing Company Airplanes". Match a handful of major OEMs; a rigorous
    # pipeline joins to the aircraft registry by make/model instead.
    text = (row.get("title") or "") + " " + (row.get("abstract") or "")
    for maker in ["Boeing", "Airbus", "Bombardier", "Embraer", "Cessna",
                  "Textron", "Piper", "Gulfstream", "Sikorsky", "Bell",
                  "Robinson", "Pratt & Whitney", "General Electric",
                  "Rolls-Royce", "CFM", "Honeywell", "Leonardo", "ATR"]:
        if re.search(re.escape(maker), text, re.I):
            return maker
    return "(other / unmatched)"


df = faa_rules(2018, 2024)
print(f"FAA rules pulled: {len(df):,}")

# --- 1. Airworthiness Directives vs other FAA rules ------------------
df["is_ad"] = df.apply(is_ad, axis=1)
ad = df[df["is_ad"]].copy()
print(f"Airworthiness Directives: {len(ad):,} "
      f"({len(ad) / max(len(df), 1):.1%} of FAA rules)")

# --- 2. Directives per year ------------------------------------------
ad["year"] = pd.to_datetime(ad["publication_date"]).dt.year
print("\nADs by year:")
for yr, n in ad["year"].value_counts().sort_index().items():
    print(f"  {yr}  {n:>5,}")

# --- 3. Directives by aircraft manufacturer --------------------------
ad["maker"] = ad.apply(manufacturer, axis=1)
print("\nADs by manufacturer (title match):")
for maker, n in Counter(ad["maker"]).most_common(15):
    print(f"  {maker:<22} {n:>5,}")

Two notes about the script deserve emphasis. First, the manufacturer extraction is deliberately crude—it matches a handful of major OEM names against the AD title, which conventionally names the type-certificate holder. This is a first pass only. A rigorous attribution does not parse names out of free text at all; it joins the directive's make and model fields to the aircraft registry, which carries the authoritative, normalized make-and-model vocabulary and, crucially, the fleet counts needed to normalize directive volume against the population at risk. Second, separating ADs from other FAA rules on title and abstract keywords is reliable enough for a survey but should be confirmed against the Part 39 citation in the full rule text for any analysis that depends on a precise AD count, because the FAA publishes many rule types and the boundary cases—corrections, withdrawals, superseding ADs that replace an earlier directive—need careful handling. The script computes the survey-level metrics; the make-and-model join to the registry is the natural next step for anything more.

Limitations and analytical caveats

The FAA actions record is the authoritative public account of mandatory aircraft safety fixes, but it carries structural features that an analyst must internalize before drawing conclusions from it.

Directive count is not a safety ranking. The most tempting and most dangerous misreading is to treat the number of ADs against a model as a measure of how unsafe it is. A high count is driven as much by fleet size and age as by design quality: the most popular, longest-serving transport airplanes accumulate the most in-service hours, surface the most findings, and therefore draw the most directives—and those same directives are the reason the type stays safe. A rarely flown type with few ADs may simply have generated too little in-service experience to reveal its problems. Any model-to-model comparison must normalize against the registered fleet and the cumulative years in service, and even then the count measures the intensity of ongoing oversight, not a verdict on the underlying design.

Superseding ADs and the active-versus-historical distinction complicate counting. An AD is frequently revised or replaced: a later directive supersedes an earlier one on the same condition, tightening an inspection interval or adding a terminating modification. A raw count of every published directive therefore over-counts distinct safety issues, because several documents may address one evolving problem. Whether an analysis wants the count of distinct unsafe conditions, the count of currently-in-effect directives, or the full historical stream of every action ever published is a real modeling choice, and the superseding relationships have to be resolved—not just the documents counted—to answer it correctly.

The record is the rule, not the compliance. An AD records that the FAA mandated an action; it does not, by itself, record whether each affected aircraft actually complied. Compliance is tracked in individual aircraft maintenance records and, in aggregate, through other oversight channels—it is not a field in the directive itself. The dataset answers what was required of which fleet and when; it does not answer whether every operator did it on time. Treating the existence of an AD as evidence that the condition was fixed across the fleet over-reads the data; the directive is the obligation, not the proof of its discharge.

Free-text make and model need entity resolution, and there is publication lag. The make and model named in an AD title and abstract are free text, written in evolving conventions—a manufacturer that changed its corporate name, a model series referred to several ways, an engine identified by part number rather than model. Joining to the registry by these fields demands the same careful entity resolution that any name-based join requires, and the crude title-matching in the example script is no substitute for it. And while emergency ADs reach operators immediately, the published rule record is assembled through the Federal Register and the FAA's regulatory systems, so the most recent actions can lag in any snapshot, and the boundary between an emergency directive issued directly to operators and its later formal publication must be handled with care. The dataset is authoritative for the established record of mandated fixes; it is not a real-time feed of what was ordered an hour ago.

Held with these caveats, the faa_actions table is a uniquely valuable resource: roughly 22,900 FAA rules and Airworthiness Directives—the legally binding record, design by design and component by component, of every time the federal government found an unsafe condition in the flying fleet and ordered it fixed before the next accident could repeat the first one. It is the regulatory hinge between the accident-investigation record that precedes it and the registered aircraft that must obey it.