NASA/TM—2007-214664

Investigation of Current Methods to Identify

Helicopter Gear Health
Paula J. Dempsey
Glenn Research Center, Cleveland, Ohio

David G. Lewicki
U.S. Army Research Laboratory, Cleveland, Ohio

Dy D. Le
Federal Aviation Administration, William J. Hughes Technical Center,
Atlantic City International Airport, New Jersey

July 2007
 NASA STI Program . . . in Profile

Since its founding, NASA has been dedicated to the • CONFERENCE PUBLICATION. Collected
advancement of aeronautics and space science. The papers from scientific and technical
NASA Scientific and Technical Information (STI) conferences, symposia, seminars, or other
program plays a key part in helping NASA maintain meetings sponsored or cosponsored by NASA.
this important role.
• SPECIAL PUBLICATION. Scientific,
The NASA STI Program operates under the auspices technical, or historical information from
of the Agency Chief Information Officer. It collects, NASA programs, projects, and missions, often
organizes, provides for archiving, and disseminates concerned with subjects having substantial
NASA’s STI. The NASA STI program provides access public interest.
to the NASA Aeronautics and Space Database and its
public interface, the NASA Technical Reports Server, • TECHNICAL TRANSLATION. English-
thus providing one of the largest collections of language translations of foreign scientific and
aeronautical and space science STI in the world. technical material pertinent to NASA’s mission.
Results are published in both non-NASA channels and
by NASA in the NASA STI Report Series, which Specialized services also include creating custom
includes the following report types: thesauri, building customized databases, organizing
and publishing research results.
• TECHNICAL PUBLICATION. Reports of
completed research or a major significant phase For more information about the NASA STI
of research that present the results of NASA program, see the following:
programs and include extensive data or theoretical
analysis. Includes compilations of significant • Access the NASA STI program home page at
scientific and technical data and information http://www.sti.nasa.gov
deemed to be of continuing reference value.
NASA counterpart of peer-reviewed formal • E-mail your question via the Internet to
professional papers but has less stringent help@sti.nasa.gov
limitations on manuscript length and extent of
graphic presentations. • Fax your question to the NASA STI Help Desk
at 301–621–0134
• TECHNICAL MEMORANDUM. Scientific
and technical findings that are preliminary or • Telephone the NASA STI Help Desk at
of specialized interest, e.g., quick release 301–621–0390
reports, working papers, and bibliographies that
contain minimal annotation. Does not contain • Write to:
extensive analysis. NASA Center for AeroSpace Information (CASI)
7115 Standard Drive
• CONTRACTOR REPORT. Scientific and Hanover, MD 21076–1320
technical findings by NASA-sponsored
contractors and grantees.
NASA/TM—2007-214664

Investigation of Current Methods to Identify

Helicopter Gear Health
Paula J. Dempsey
Glenn Research Center, Cleveland, Ohio

David G. Lewicki
U.S. Army Research Laboratory, Cleveland, Ohio

Dy D. Le
Federal Aviation Administration, William J. Hughes Technical Center,
Atlantic City International Airport, New Jersey

Prepared for the

2007 Aerospace Conference
cosponsored by the Institute of Electrical and Electronics Engineers and the American Institute
of Aeronautics and Astronautics
Big Sky, Montana, March 3–10, 2007

National Aeronautics and

Space Administration

Glenn Research Center

Cleveland, Ohio 44135

July 2007
 Acknowledgments

The authors would like to thank several experts in the Health and Usage Monitoring System (HUMS) field for providing
valuable information via interviews and survey responses for this paper: Alan Duke and Eric Bechhoefer of Goodrich, Paul
Grabill of Intelligent Automation Corporation, Andy Heather of Smiths Aerospace, Bill Hardman and Eric Carney of the
Naval Air Warfare Center.

Trade names and trademarks are used in this report for identification
only. Their usage does not constitute an official endorsement,
either expressed or implied, by the National Aeronautics and
Space Administration.

Level of Review: This material has been technically reviewed by technical management.

Available from
NASA Center for Aerospace Information National Technical Information Service
7115 Standard Drive 5285 Port Royal Road
Hanover, MD 21076–1320 Springfield, VA 22161

Available electronically at http://gltrs.grc.nasa.gov

 Investigation of Current Methods to Identify
Helicopter Gear Health
Paula J. Dempsey
National Aeronautics and Space Administration
Glenn Research Center
Cleveland, Ohio 44135
David G. Lewicki
U.S. Army Research Laboratory
Glenn Research Center
Cleveland, Ohio 44135
Dy D. Le
Federal Aviation Administration
William J. Hughes Technical Center
Atlantic City International Airport, New Jersey 08405

Abstract
This paper provides an overview of current vibration methods used to identify the health of helicopter
transmission gears. The gears are critical to the transmission system that provides propulsion, lift and
maneuvering of the helicopter. This paper reviews techniques used to process vibration data to calculate
conditions indicators (CI’s), guidelines used by the government aviation authorities in developing and
certifying the Health and Usage Monitoring System (HUMS), condition and health indicators used in
commercial HUMS, and different methods used to set thresholds to detect damage. Initial assessment of a
method to set thresholds for vibration based condition indicators applied to flight and test rig data by
evaluating differences in distributions between comparable transmissions are also discussed. Gear
condition indicator FM4 values are compared on an OH58 helicopter during 14 maneuvers and an OH58
transmission test stand during crack propagation tests. Preliminary results show the distributions between
healthy helicopter and rig data are comparable and distributions between healthy and damaged gears show
significant differences.

I. Introduction
Helicopter transmission integrity is important to helicopter safety because helicopters depend on the
power train for propulsion, lift, and flight maneuvering. A study of 1168 helicopter accidents from 1990
to 1996 found that after human-factors related causes of accidents, the next most frequent causes of
accidents were due to various system and structural failures (ref. 1). In 1999, of the world total of 192
turbine helicopter accidents, 28 were directly due to mechanical failures with the most common in the
drive train of the propulsion system (ref. 2). Liu and Pines (ref. 3) continued the study performed by
NASA from 1998 to 2004. Their study showed the number of accidents caused by vehicle factors
compared to the total number of accidents have been reduced by more than one half from earlier accident
data. Pilot error continues to be the major cause of all rotorcraft accidents and failure or malfunction of
the propulsion system remains the primary reason for vehicle factor related accidents. In order to reduce
helicopter accidents, the original study (ref. 1) recommended the design of HUMS capable of predicting
impending equipment failure for on-condition maintenance, and more advanced systems capable of
warning pilots of imminent equipment failure. In order to make these predictions, the system must
provide health monitoring of the transmission components and must also demonstrate a high level of
reliability to minimize false alarms.

NASA/TM—2007-214664 1
 Figure 1 shows the potential economic and safety benefits of diagnostics and predictive maintenance
of critical mechanical systems. If usage is more severe than design life of a component, health monitoring
will provide a safety benefit. If usage is less severe, the service life of critical components may be
extended. Although commercially available HUMS provide safety benefits when installed on rotorcraft,
the damage detection rate of today’s helicopter health monitoring systems through vibration analysis can
be improved. HUMS experience documented by the United Kingdom Civil Aviation Authority (CAA) in
1997, and informally updated in 2002, shows a success rate of 70 percent in detecting defects (ref. 4).
The following challenges still need to be addressed to achieve the safety and economic potential of an
integrated health monitoring system (ref. 5):

• Limited helicopter vibration fault data.

• Diagnostic tool operational constraints are undefined.
• Systems complex—Experts required for data analysis.
• Decision-making on system health difficult.
• Standard thresholds for types/levels of damage are not clearly defined.
• Performance assessment methods have not been developed.
• Tradeoffs between fault detection and false alarms.
These challenges must be addressed at the subsystem level prior to development of an overall
integrated helicopter health monitoring system.
This paper will provide an overview of current vibration based methods used to identify gear
component damage in commercial HUMS. The standard techniques used to process vibration data and
calculate vibration algorithms to detect gear damage will be discussed first. Then a brief overview of the
guidelines provided by the FAA and CAA in developing and certifying HUMS will be provided. Next,
the vibration condition and health indicators used in commercial HUMS will be discussed including
results of a survey sent to several manufacturers. Then, different methods used to set thresholds to detect
damage will be outlined. Initial assessment of a method to set thresholds applied to flight and test rig data
will be discussed.

Figure 1.—Economic and safety benefits of diagnostics and prognostics (ref. 40).

NASA/TM—2007-214664 2
 II. Gear Diagnostic Algorithms
Numerous diagnostic techniques have been developed from vibration data to detect damage and
abnormal conditions of the dynamic mechanical components in rotorcraft propulsion systems. A majority
of the technology developed focused on the gear health. Using vibration data collected from gearbox
accelerometers, algorithms are developed to detect when gear damage has occurred. Other areas
addressed include bearings and driveshafts of the main transmission system. This work will focus on
vibration algorithms available in commercial HUMS used to detect gear damage.
Gear damage produces changes in the vibration signatures measured by accelerometers installed on
the gearbox. Over the past 25 years, numerous vibration-based algorithms for mechanical component
damage detection in transmissions have been developed. The traditional methods of vibration based gear
feature detection and extraction methods in rotating equipment, discussed in detail by (ref. 6), are
typically based on some statistical measurement of vibration energy. The primary differences are based
on which of the characteristic frequencies are included, excluded, or used as a reference. Figure 2 is a
block diagram of the steps required to calculate FM4 (Figure of Merit 4), a common vibration algorithm
used in commercial HUMS.

INPUT
Sensor Data (accelerometers, tachometer)

Time synchronous averaged (TSA) vibration data

Fast Fourier Transform FM4—Gear Meshing

Frequency Harmonics, 1st
Order Sidebands
Filter

Inverse Fast Fourier Transform

Difference Signal (Filtered TSA) = d

∑ (d )
N
1 4
n −d
N n =1
2
⎡1 N ⎤
⎢
⎣⎢ N
∑
n =1
(d n − d ) ⎥2

⎦⎥

OUTPUT
Condition Indicator FM4

Figure 2.—Flowchart for calculating standard vibration algorithm FM4.

NASA/TM—2007-214664 3
 Referring to figure 2, once the vibration time series data is collected from the accelerometers, the first
step to define the gear vibration algorithm is to calculate the time synchronous average (TSA) of the data.
Gears produce vibration signals synchronous with speed. Noise in the vibration signal is reduced using
time synchronous averaging. Synchronous averaging refers to techniques for extracting periodic
waveforms from additive noise by averaging vibration signals over several revolutions of the shaft
(ref. 7). The typical signal time synchronous average is obtained by taking the average of the signal in the
time domain. Using the once per revolution signal, the vibration signal is interpolated into a fixed number
of points per shaft revolution. The number of interpolated points are typically a power of two (512, 1024,
etc.). This makes implementing frequency domain signal processing techniques significantly easier. The
desired signal, which is synchronous with the shaft speed, will intensify relative to the non-periodic
signals. It should be noted that sample rates of the vibration time series data determines the number of
averages that can be obtained for the time synchronous averaging. The vibration data is also filtered with
a low band pass anti-aliasing filter prior to data acquisition to prevent aliasing caused when the sample
rate is less than twice the bandwidth of the input signal.
After the accelerometer data is sampled and time synchronous averaged to shaft speed, the time
domain data is converted to frequency domain data using a fast Fourier transform. The amount of filtering
is defined by the algorithm. FM4 (Figure of Merit 4) was developed to detect changes in the vibration
pattern due to damage (pitting, small, cracks) on a limited number of teeth (ref. 7). For FM4, gear
meshing frequency, harmonics, and 1st order sidebands are removed from the original signal. This signal
is converted back to the time domain. Then a statistical parameter, the fourth normalized statistical
moment (normalized kurtosis) of the difference signal, is calculated from this data, where d is the
difference signal, d is the mean value of the difference signal and N is the total number of interpolated
points per reading. The kurtosis of a normal distribution equals 3. In theory, for healthy gears, the
difference signal would be Gaussian noise, resulting in a value of 3. When one or two teeth develop a
defect (such as a crack or pitting) a peak or series of peaks appears in the difference signal and FM4
reacts by increasing above this nominal value of 3.

III. FAA and CAA HUMS Guidance

In 1999, the FAA published Advisory Circular (AC) 29–2C, Section MG–15, hereafter referred as the
HUMS AC (ref. 8). The HUMS AC provides guidance for achieving airworthiness approval for
installation, credit validation, and instructions for continued airworthiness (ICA) for a full range of
HUMS application. Installation includes all the equipment needed for the end-to-end application that is
associated with acquiring, storing, processing, and displaying the HUMS application data, including
airborne and ground-based equipment. Credit validation includes evidence of effectiveness for the
developed algorithms, acceptance limits, trend setting data, tests, etc., and the demonstration methods
employed. A plan is needed to ensure continued airworthiness of those parts that could change with time
or usage and includes the methods used to ensure continued airworthiness. The AC establishes an
acceptable means, but not the only means of certifying a rotorcraft HUMS. Certification of HUMS
addresses the complete system, including quality assurance of data produced from the airborne modules
of the HUMS, and the potential subsequent intervening actions resulting from the ground processing of
the airborne data.
Setting reliable thresholds is related to the credit validation phase of HUMS certification by providing
evidence of condition indicators effectiveness through reliable threshold limits. In order to obtain
maintenance credits, alert limits and intervention actions must be defined based on direct evidence of the
failure mechanism. If seeded fault tests are used to measure the performance of condition indicators, the
tests must be verified as representative of flight data.
In 1999, the CAA also came out with Civil Aviation Publication (CAP) 693 that provides operators
with guidelines for installation and use of health monitoring systems. The CAP described the level of
monitoring required and provided advice on ensuring that the monitoring was effective and reliable. The
section on threshold setting and adjustment states that thresholds must be defined in the form of absolute

NASA/TM—2007-214664 4
signal values, number of standard deviations above the mean fleet value, or other means. Trends for
indicators that require trend monitoring must also be defined. In the threshold changes section it states
that changing the limits requires an understanding of the relationship between the algorithm values and
the increase in the severity of the fault detected based on the history of the component (ref. 9).
In 2006, the Civil Aviation Authority came out with CAP 753, to provide additional guidance for
Operators using vibration health monitoring (VHM) in helicopter rotor and drive systems. In this
document they define VHM as, “the monitoring of vibration data and characteristics that can provide
advance information relating to the development of incipient failures in the engine(s) rotor drive systems”
(ref. 10). Gear tooth indicators that can detect gear tooth damage and cracks are identified as required
indicators. The signal acquisition and processing section of this document discusses the importance of
sampling rate and signal averaging for VHM. Thresholds, discussed in the alert generation and management
section of this document, may be set at absolute vibration levels based on fleet experience or learned for an
individual helicopter. Thresholds can be set as alerts, to indicate inspections, or as alarm to indicate that
flights should be avoided until maintenance has been performed. The thresholds are dependent on different
flight regimes. The VHM system is required to detect 70 percent of the failure modes it is designed to
monitor. Data must be capable of being downloaded within a 10 hour time period. System performance
should be evaluated for alert/false alarm rates and success/failure rates to detect component damage.

IV. Commercial HUMS Vibration Algorithms

An article in Rotor & Wing magazine provided a good overview of commercial HUMS (ref. 11).
HUMS consist of a main computer for data processing and acquisition of in-flight data from
accelerometers and tachometers installed on critical aircraft components. If the aircraft is equipped with a
flight data recorder, the system may also collect data from the recorder. Components that are monitored
include the main and tail rotors, their gear boxes and drive trains. Vibration data is typically downloaded
by a mechanic post flight. The main function is to collect data to identify normal and abnormal changes in
vibration levels on an aircraft's critical components. Typically the boundaries of “normal” performance
are identified during flight testing. All commercial HUMS use vibration algorithms for gears calculated
using the steps discussed in the gear vibration algorithms section of this paper. Threshold configurations
can be different for each system. Techniques utilized by three HUMS manufacturers are reviewed in the
following paragraphs.
The Goodrich Integrated Mechanical Diagnostics Health and Usage Management Systems (IMD-
HUMS) mechanical diagnostics capabilities include drive train vibration information analyzed in-flight to
determine drive train health. A number of condition indicators are calculated from vibration data to
characterize component health. The IMD-HUMS has limit exceedance capabilities on drive train
vibration levels for condition indicators (CI) and health indicators (HI) (ref. 12). Condition Indicators (CI)
refer to gear vibration algorithm used to reflect the health of the gear, such as FM4. Goodrich calculates
several CI’s for a component. The U.S. Army and the Goodrich Corporation have undertaken a battalion-
level demonstration of the Goodrich Integrated Mechanical Diagnostics Health and Usage Monitoring
System (IMD HUMS). A paper by Wright (ref. 13) provides an overview of benefits utilizing the
automatic acquisition of drive train component CI’s. Significant detections have been made to date of
problems on the helicopter. Comparisons between aircraft are being used to verify and validate some
indicators. However, final health limits and thresholds are not defined for every component.
In 1993, the U.S. Navy provided the groundwork for the current Goodrich IMD-HUMS through work
on an H-60 Helicopter Integrated Diagnostic System (HIDS) (ref. 12). Flight and ground tests were
performed to develop and evaluate diagnostic algorithms for the HIDS. Additional work was performed
installing the IMD-HUMS on the H-53 platform. These programs demonstrate the capabilities of the IMD
HUMS as a tool for rotorcraft health management and diagnostics. Navy researchers have also performed
preliminary research applying multivariate statistical process control techniques to produce a single
indicator on the state of a system. This technique was applied to drive system seeded fault tests with
promising results (ref. 14).

NASA/TM—2007-214664 5
 Intelligent Automation Corporation developed a HUMS system with the U.S. Army under the Vibration
Management Enhancement Program (VMEP) to monitor the health of drive system dynamic components.
The system consists of an on-board computer to collect and processes vibration information in flight, ground
based software for displaying information to maintenance personnel, and web–based infrastructure tools for
data archiving and analysis (ref. 15). Data acquisition can be triggered at a specific time or during a specific
flight regime. FM4 is one of several algorithms used as a condition indicator to threshold or trend for fault
diagnostics. Keller (ref. 16) provides and overview of the drive train faults, 4 bearings and 1 shaft,
discovered to date in over 70 aircraft equipped with the VMEP system. A database of drive train condition
indicators has been developed from the data collected to date.
Smiths Aerospace developed the GenHUMS (Generic HUMS). GenHUMS has been in operational
service in the UK Chinook fleet since 2000. The GenHUMS also includes powertrain and gearbox
diagnostics in their health monitoring system. Smiths Aerospace and Bristow Helicopters are also
developing a HUMS anomaly detection system under support from the CAA because HUMS data has
shown that not all defect related trends or changes in HUMS data are detected with current threshold
setting methods. They provided an example of large crack in a bevel pinion of a main gearbox that did not
generate alerts from the vibration indicators. The objective of their work is to develop a method to
identify abnormal behavior by comparing it to a data set of “normal” behavior (ref. 17).
Preliminary work was done by Boeing to investigate a method to evaluate diagnostic algorithms and
define criteria for threshold settings and compiled into a report, “Monitor the Monitors, RITA Metrics
Document.” A conceptual design of a Graphical User Interface was developed for a tool, but
implementation procedures were not defined and the tool was never validated due to the limited amount
of transmission seeded fault data available for assessment of vibration algorithm performance (ref. 18).

V. HUMS Survey Results

A survey was sent to the 3 HUMS manufacturers (Goodrich, Intelligent Automation Corporation,
Smiths Aerospace) to compare the methods used for identifying drive train component damage. The
survey contained the following questions related to their systems:

• Number/locations of sensors (accelerometers, 1X/rev)

• Time synchronous averaging techniques
• Vibration algorithms used for gears and bearings
• Data collection information (frequency, amount, type)
• Regime/maneuver recognition
• Operational constraints
• Decision making tools using condition indicators (CI’s)
• System performance and reliability metrics
All three manufactures provided valuable information on the capabilities of their HUMS (refs. 19 to
21). Results of the survey indicated commercial HUMS shared many similar traits. The number and
location of sensors and time synchronous averaging techniques were comparable. The CI’s for drive train
components were similar. Data collection rates and amount/type of data saved is user selectable. There
are some operational constraints regarding data collection during stable regimes with minimal transitions.
Data collection can be automatic based on specified conditions. Smiths has found that data collected in
normal cruise conditions at 60 to 70 percent torque provides the best flight regime for monitoring gear
CI’s. Some systems monitor torque for regime/maneuver recognition but the threshold adjustment based
on torque varies. Some systems access several parameters in addition to speed and torque such as
airspeed, altitude, rate of descent that can be correlated with CI’s. Decision making tools using vibration
algorithms or condition indicators (CI’s) integrated into health indicators (HI’s) are under development by
several manufacturers. HI’s provide decision making tools for the end user on the status of system health.
Overall, CI and HI performance and reliability is very difficult to evaluate due to limited failure data

NASA/TM—2007-214664 6
available to assess performance. Published CI and HI threshold setting methods currently under
development by the three manufacturers will be discussed in the next two paragraphs.
Intelligent Automation Corporation created a tool for developing a standard dataset of CI’s (ref. 22).
New CI data is compared to the standard dataset to determine if the new data fits within the standard data
set as normal or a specific fault class. Then data is accepted or rejected based on a rejection threshold. If
the data is accepted it can be added to the standard data set after review by qualified personnel.
Goodrich is developing proprietary statistical procedures to define sets of threshold for each aircraft
type and component by modeling the effect of variance across aircraft and torque on transmission
components (ref. 23). Thresholds use a “healthy” data set that spans a subset of the fleet and the operating
conditions. Variance for different helicopters and torque levels is estimated using a least square estimator,
then corrected for small sample by using Student’s t-distribution to test for statistical significance between
two sample means. Their system is also capable of integrating several CI’s into one health indicator (HI)
that can be used to identify when a component requires additional analysis or a component is faulted
(refs. 23 to 25). Fusing CI’s sensitive to a specific failure can minimize the amount of information used
by end user to make a decision on system health.
Smiths is currently developing and fielding health indicators (HI) for the end user to easily interpret
the system health (ref. 21). In the past, simple thresholds were identified for condition indicators. More
work was performed on statistical analysis of CI data distributions. Rule based systems have also been
used. Smiths has found that the mentioned threshold setting tools may identify classic faults, but cannot
identify novel faults that can occur 50 percent of the time. Using Artificial Intelligence (AI) reasoners,
Bayesian statistics, data mining, clustering and failure modes and effects analysis models, they have been
successful identifying novel, unanticipated faults such as sensor failures. They have rerun 6 years of
archived data using their anomaly detector HI with promising results. This anomaly detection technique
will not replace the current HUMS threshold setting methods. Rather, it will provide additional
capabilities to the current HUMS threshold setting methods.

VI. Environmental Effects on Vibration Methods

The usage monitoring piece of the HUMS consists of measuring torque, engine operating hours and
flight hours to track usage of aircraft. Flight regimes and environmental operating conditions can also be
included. The sensitivity of the CI’S to environmental effects in varying flight regimes must be taken into
consideration when setting thresholds. Understanding the significant effects of operating conditions on
CI’s is important to defining reliable thresholds. Preliminary work has been performed evaluating gear
CI’s under varying load, speed and flight conditions, but more work is required (refs. 26 and 27). Limited
work has been performed verifying that thresholds defined in test rigs can classify helicopter transmission
health. This is due to limited availability of flight fault data to verify damage detection sensitivity
demonstrated in test rigs can be maintained in flight in different flight regimes (refs. 28 and 29).

VII. Setting Thresholds

The goal for setting thresholds to indicate drive train health is to provide the minimum number of
false alarms while maintaining sensitivity to gear damage. As mentioned in the previous sections, HUMS
manufacturers use several methods to define thresholds. One of the biggest challenges when defining the
threshold or limit of a diagnostic tool is the tradeoff between the sensitivity of the limit to indicate
damage and the number of false alarms. If a limit is decreased, damage may be detected, but more false
alarms may result. If a limit is increased, false alarms may decrease, but the algorithms will be less
sensitive to damage. Figure 3 illustrates 2 distributions used to represent a no damage response and a
damage response of a condition indicator.

NASA/TM—2007-214664 7
 Figure 3.—Probability of no damage/false alarms and damage
detected/missed hits.

Based on the CI value and the health of the component, table 1 provides the results of the decisions
made on a given CI. Interpreting the overlapping region illustrated in figure 3 is the challenge to setting
reliable thresholds based on the CI.
TABLE 1.—RESULTS OF DAMAGE/NO DAMAGE DECISIONS
Health of component
No gear damage Gear damage
Decision Indicate no damage Correct decision Missed hit
Indicate damage False alarm Correct decision

A. Hypothesis Tests
Figure 3 and table 1 are related to hypothesis tests used in statistical analysis. A hypothesis test is
defined by testing the value of a population parameter, the null hypothesis, Ho. The alternative hypothesis,
H1, is the statement that must be true if the null hypothesis is false.
Table 2 represents the hypothesis parameters for the decisions listed in table 1. The hypothesis is
defined to test that the mean of the CI from the undamaged gear looks significantly different than the
mean of the CI of the damaged gear. A test statistic is used to make the decision to reject the null
hypothesis based on the probability of acceptable error. Referring to table 2, if the null hypothesis is
rejected when it is true, this results in Type I error (false positive) If you fail to reject the null hypothesis
when it is false, this results in Type II error (false negative). An earlier analysis of flight data found the
mean could vary 3 standard deviations within a healthy set of CI data (ref. 29). The variance within the
data must be considered when defining a test statistic to set thresholds to minimize the occurrence of false
alarms.
TABLE 2.—HYPOTHESIS OF DAMAGE/NO DAMAGE DECISIONS
True state of system
Ho is true Ho is false
Ho: μ < CI H1: μ > CI
Gear is healthy Gear is damaged
Decision Reject Ho False alarm Correct decision
(Type I error)
Fail to reject Ho Correct decision Missed hit
Healthy component (Type II error)

B. Bayesian Statistics
Bayesian inference is another statistical analysis technique that can be used for setting thresholds to
make a decision on component health. Bayesian inference updates the likelihood of a hypothesis
(probability a hypothesis is true) given a previous likelihood estimate and additional evidence
(observations). Bayesian inference can be used to determine the probability that a diagnosis of gear
damage is correct given a priori information. An example of an equation for Bayesian inference to set
damage detection thresholds is:

NASA/TM—2007-214664 8
 P(O / f i ) ⋅ P( f i )
P( f i / O ) = (1)
∑P(O / f i ) ⋅ P( f i )
i

where P(fi /O) equals the probability of fault (fi) given diagnostic output (O), P(O/fi) equals the probability
that a diagnostic output (O) is observed with fault (fi), and P(fi) is the a priori probability of (fi) occurring
(refs. 30 and 31).
Due to limited fault data, it is challenging to apply Bayesian inference to identifying component
health. Bayesian inference requires knowledge about the diagnostic system to generate the a priori
distributions and a priori probabilities of the hypotheses. The complexity of the data due to multiple
hypotheses (several vibration algorithms and damage levels) and multiple conditional dependent events
(maneuvers, time dependent data) also make it difficult to define levels of probability for each scenario.
Bayesian statistics may be useful when applied to fault classification, where a large amount of fault data
is acquired on different types of faults (ref. 32).

C. Fuzzy Logic
Fuzzy logic has been successfully used to set thresholds to identify gear damage and to make a
decision on the health of the system (refs. 28 and 33). Earlier studies found fuzzy logic, as compared to
production rules and neural networks, the most robust when monitoring transitional failure data on a
gearbox (ref. 34). Another study comparing automated reasoning techniques for condition-based
maintenance found fuzzy logic more flexible than standard logic by making allowances for unanticipated
behavior (ref. 35). Fuzzy logic applies fuzzy set theory to data, where fuzzy set theory is a theory of
classes with unsharp boundaries. The data belongs in a fuzzy set based on its degree of membership
(ref. 36).
Fuzzy logic starts with a fuzzy set, extending boolean set theory to a continuous valued logic via the
concept of membership functions valued between 0 and 1. A membership function is a curve that defines
how each point in the input space is mapped to a membership value or degree of membership between 0
and 1. The only condition a membership function must satisfy is that it is a continuous function that varies
between 0 and 1. Once membership functions are defined for a fuzzy set, fuzzy rules must be defined.
Fuzzy rules are defined by experts in the field. Experts express their field knowledge in rules with an IF-
THEN format. An example of a fuzzy if-then rule is, “if x is A then y is B.” The final step is to convert
the fuzzy membership information and rules into a crisp output using defuzzification methods.
Figure 4 is an example of using fuzzy logic to define health indicators from gear condition indicators.
Gear condition indicators are calculated from sensor data. Membership functions are then developed for
the CI’s. Defining membership functions that accurately reflect the CI data is the challenge when using
fuzzy logic to set thresholds. In this example, trapezoidal shaped membership functions were used. A
certain amount of overlap is required or no rules will fire for values in the gap between the membership
functions. The x-axis of the membership function refers to the CI values and the y-axis refers to the
degree of membership. DL refers to Damage Low (Gear O.K), DM refers to Damage Medium (Inspect
Gear) and DH refers to Damage High (Shutdown System). The basic rules are defined based on the level
of damage indicated by the CI and experience with the CI to detect specific faults in certain
environmental conditions. Membership functions and rules can be added for additional CI’s. The degree
of membership for the output of the fuzzy model provides the status of gear health. For this example O.K
indicates no gear damage, Inspect indicates initial pitting and Shutdown (due to damage) indicates
destructive pitting. The output was defined to give the end user a simple function based on the state of the
gear. A value of 0 to 0.33 indicates the gear is O.K., 0.33 to 0.66 indicates the gear should be inspected,
and 0.66 to 1.0 indicates shutdown the system, the gear is damaged.

NASA/TM—2007-214664 9
 CI’s: FM4, NA4, Debris

Membership Functions

Gear Health
Shutdown
Inspect
O.K.

Figure 4.—Setting threshold from CI’s using fuzzy logic.

D. Relative Frequency Plots

Preliminary research was performed comparing probability distributions of CI’s from a GRC spur
gear component rig with and without gear damage and flight data from an OH58 helicopter under normal
conditions (ref. 29). Histograms of the CI values were created for CI’s measured in the test rig with and
without damage and the CI’s measured on a helicopter during several maneuvers. In order to compare the
histograms with probability distributions, relative frequency plots were created scaling the histograms so
that the area under the curve was equal to 1. The relative frequency plots of the flight data were similar to
the curves for NASA Glenn Research Center (GRC) component test rig data without gear damage.
Distributions of the data also changed when damage occurred. This earlier work compared the data from
helicopter flights to a simple test rig consisting of 2 meshing spur gears. In order to evaluate if differences
in distributions are observed between comparable helicopter and test rig transmissions work has begun to
evaluate the difference in distributions between flight data and a comparable transmission test stand. The
next section will provide preliminary data to demonstrate this analysis.

VIII. Setting CI Thresholds on OH58 Flight and Test Data

Figure 5 illustrates the OH58 helicopter used to collected healthy CI data from accelerometers located
on the transmission housing (refs. 37 and 38). For this analysis, only one accelerometer mounted
horizontally and radial to the transmission will be discussed. No mechanical component damage was
observed during flights. Data was collected from 14 different maneuvers each repeated 12 times. The
maneuvers and average torque level during the maneuver are shown in table 3. The pilot set-up the
maneuver, then collected data for 34 sec. The component monitored for health was the main transmission
19 tooth input spiral bevel gear pinion. Vibration data was sampled at a rate of 50 kHz with an anti-
aliasing filter set at 18 kHz. For each 34-second maneuver, 48 time synchronous averages were calculated
with 71 rotations in each average. In addition to vibration, torque and speed were also measured. The
pinion was run at an average speed of 6278 rpm.

NASA/TM—2007-214664 10
 Figure 5.—OH58 helicopter.

TABLE 3.—OH58 FLIGHT MANEUVERS

Man Torque, Maneuver description Man Torque, Maneuver description
percent percent
A 55 Level, forward ~55 percent torque H 74 Hover, ~10 ft
B 80 Level, forward ~80 percent torque I 74 Hover ~10 ft, turn left
C 58 Level, sideways left J 73 Hover ~10 ft, turn right
D 58 Level sideways right K 61 20° bank left turn
E 55 Climb, ~55 percent torque L 61 20° bank right turn
F 15 Descent, ~15 percent torque M 80 Climb, ~80 percent torque
G 26 Flat pitch on ground N 35 Descent, 35 percent torque

Figure 6 shows the relative frequency plots for condition indicator FM4 from an accelerometer
installed on the transmission housing during NASA Ames OH58 Flight Tests. For this data, histograms
were created from the FM4 CI data for each maneuver with a bin width of 0.05 and a range of 1 to 8.5.
The histogram data was then rescaled so the area under the curve (integral under the estimated
distribution) is equal to 1 by dividing the counts by the bin size and total number of readings
(counts/(readings*bin size). Although the minimum FM4 value for this accelerometer across all
maneuvers was 1.87 and the maximum was 4.92, the majority of values fall within 2 and 3. The
distributions varied for each maneuver and the tails of the distributions shift to the right for higher torque
levels, with the maximum FM4 value observed during maneuver B, 80 percent torque. The changes in the
distribution for each maneuver shows that FM4 is effected by operational conditions.
Figure 7 illustrates the NASA GRC OH58 Helicopter Transmission Test Stand used to collect healthy
and damaged CI data from accelerometers located on the transmission housing during a matrix of pinion
crack propagation tests (ref. 39). The main transmission 19 tooth spiral bevel pinion gear, like the
helicopter, was monitored for health. And, for this preliminary analysis, only one accelerometer mounted
horizontally and radial to the transmission will be discussed. For these tests, a pinion tooth was notched
and run several hours to facilitate a tooth fracture. In addition to vibration, torque and speed were also
measured. The pinion was run at design speed of 6060 rpm for all tests. Torque values ranged from 80 to
150 percent torques. Vibration data was sampled at a rate of 150 kHz with an anti-aliasing filter set at
56 kHz. Data was sampled every 15 sec for a 1.5 sec duration. Time synchronous averages were
calculated with 150 rotations in each average.

NASA/TM—2007-214664 11
 Figure 6.—OH58 Relative frequency plots of FM4 from NASA Ames OH58 flight tests.

Figure 7.—NASA GRC OH58 Transmission test stand.

Figure 8 shows the relative frequency plots for condition indicator FM4 from an accelerometer
installed on the transmission housing of the NASA GRC OH58 Test Stand. The pinion was run for
60.5 hr at 150 percent torque with an initial notch fabricated on one tooth of the pinion. The notch was
then deepened and the pinion ran an additional 5.5 hr at 120 percent torque, 1.34 hr at 140 percent torque,
then 0.2 hr at 150 percent torque and inspected several times within each torque cycle. A 60X microscope
was used for visually inspecting the gear. It was known that no crack initiation occurred prior to the
deepening of the notch. Crack initiation was also not observed after running at 140 percent torque. At
completion of the test at 150 percent torque a tooth was found to be fractured. A picture of the tooth is
shown in figure 9. For this data, histograms were created from the FM4 CI data at different torque levels
(120, 140, and 150 percent) with a bin width of 0.05 and a range of 1 to 8.5. The histogram data was also
rescaled so the area under the curve is equal to 1 by dividing the counts by the bin size and total number
of readings (counts/(readings*bin size). Note that the majority of values fall within 2.5 and 4.5 for the data
sets with the healthy pinion. The distribution of the data shifts to 4.5 to 6.0 for the data set with the damaged
pinion.

NASA/TM—2007-214664 12
 2.5
150 % torque,
2 original notch
120% torque,
Frequency deepened notch
1.5
140% torque,
deepened notch
1 150% torque,
deepened notch

0.5

0
1.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5
FM4
Figure 8.—Relative frequency plots of FM4 from NASA GRC OH58 Transmission test stand.

Figure 9.—Pinion cracked tooth at test completion.

Additional work is required applying threshold setting methods discussed in the previous section to
this data set and additional data from both the helicopter and the transmission test stand.

IX. Summary
This paper provided an overview of guidelines used by the FAA and CAA in developing and
certifying HUMS, the techniques used to process vibration data to calculate conditions indicators (CI’s),
condition and health indicators used in commercial HUMS, and different methods used to set thresholds
to detect damage. Initial assessment of a method to set thresholds applied to flight and test rig data was
presented. Gear condition indicator FM4 distribution values on a healthy OH58 helicopter and a health
OH58 transmission test rig were comparable. On the helicopter, distributions varied for each maneuver
and the tails of the distributions shifted to the right (FM4 values increased) for higher torque levels.
Distributions between healthy and damaged gears on the test rig showed significant differences, shifting
to the right (FM4 values increased) with increased damage.

References
1. Aviation Safety and Security Program, the Helicopter Accident Analysis Team, “Final Report of the
Helicopter Accident Analysis Team,” 1998.
2. Learmount, D., Rotary Woes, Flight International, no. 4725, vol. 157, pp. 34–35, 2000.

NASA/TM—2007-214664 13
3. Liu, L. and Pines, D.J., “Analysis of U.S. Civil Rotorcraft Accidents Caused by Vehicle
Failure/Malfunctions, 1998 to Present,” presented at the American Helicopter Society 61st Annual
Forum, Grapevine, TX, June 1–3, 2005.
4. McColl, J., “Overview of Transmissions HUM Performance in UK North Sea Helicopter Operations.
Rotorcraft Transmission Systems Health Monitoring Seminar,” UK Civil Aviation Authority,
Institution of Mechanical Engineers. London U.K., 18 Nov. 1977.
5. Zakrajsek, J.J., Dempsey, P.J., Huff, E.M., Augustin, M., Safa-Bakhsh, R., Duke, A. Ephraim, P.,
Grabill, P. and Decker, H.J., “Rotorcraft Health Management Issues and Challenges,” NASA/TM—
2006-214022.
6. Zakrajsek, J.J., “An Investigation of Gear Mesh Failure Prediction Techniques,” NASA TM–102340,
AVSCOM TM 89–C–005, November 1989.
7. Stewart, R.M., “Some Useful Data Analysis Techniques for Gearbox Diagnostics,” Machine Health
Monitoring Group, Institute of Sound and Vibration Research, University of Southhampton, Report
MHM/R/10/77, July 1977.
8. Federal Aviation Authority, Advisory Circular, “Airworthiness Approval of Rotorcraft Health Usage
Monitoring Systems (HUMS),” PS-ASW100-1999-00063, July 15, 1999.
9. Civil Aviation Authority, CAP 693, “Acceptable Means of Compliance Helicopter Health
Monitoring,” CAA AAD 001-05-99, London, May 1999.
10. Civil Aviation Authority, CAP 753, “Helicopter Vibration Health Monitoring (VHM) Guidance
Material for Operators Utilizing VHM in Rotor and Rotor Drive Systems of Helicopters,” June 2006.
11. McKenna, J.T., An Eye on Operations. Rotor & Wing. September 2005.
12. Hess, R., Duke, A. and Kogut, D., “The IMD HUMS as a Tool for Rotorcraft Health Management
and Diagnostics,” IEEE Aerospace Conference, 2001.
13. 13. Wright, J., “Emerging Results Using IMD-HUMS in a Black Hawk Assault Battalion,” AHS 61st
Annual Forum, Grapevine, Texas, June 1–3, 2005.
14. Mimnagh, M.L., Hardman, W. and Sheaffer, J., “Helicopter Drive System Diagnostics Through
Multivariate Process Control,” IEEE Aerospace Conference Proceedings, vol. 6 pp. 381–415, Big
Sky, Montana, March 2000.
15. Grabill, P. Brotherton, Berry, J. and Grant, L., “The US Army and National Guard Vibration
Management Enhancement Program (VMEP): Data Analysis and Statistical Results,” proceedings
from the American Helicopter Society 58th Annual Forum, Montreal, Canada, June 11–13, 2002.
16. Keller, J.A., Branhof, R.W., Dunaway, D., Grabill, P., “Examples of Condition Based Maintenance
with the Vibration Management Enhancement Program,” proceedings of the American Helicopter
Society 61st Annual Forum, Grapevine, Texas, June 1–3, 2005.
17. Larder, B., Callan, R., Salter, S., “Intelligent Management of HUMS Data: Development and
Demonstration of a Vibration Health Monitoring Anomaly Detection System,” presented at the
American Helicopter Society 62nd Annual Forum, Phoenix, AZ, May 9–11, 2006.
18. Safa-Bakhsh, R., Byington, C.S., Watson, M., and Kalgren, P., “Metrics Evaluation and Tool
Development for Health and Usage Monitoring System Technology,” presented at the American
Helicopter Society 59th Annual Forum, Phoenix, Arizona, May 6–8, 2003.
19. Duke, Alan and Eric Bechhoefer. Telephone interview on survey response. July 12, 2006.
20. Grabill, Paul. Survey response via email. September 1, 2006.
21. Heather, Andy. Telephone interview on survey response. October 19, 2006.
22. Brotherton, T., Grabill, P., Friend, R., Sotomayer, B. and Berry, J., “A Testbed for Data Fusion for
Helicopter Diagnostics and Prognostics,” proceedings of the 2003 IEEE Aerospace Conference, Big
Sky., Montana, March 2003.
23. Bechhoefer, E., Bernhard, A., “Setting HUMS Condition Indicator Threshold by Modeling Aircraft
and Torque Band Variance,” proceedings of the 2004 IEEE Aerospace Conference, Big Sky,
Montana, 2004a.
24. Bechhoefer, E., Bernhard, A., “HUMS Optimal Weighting of Condition Indicators to Determine the
Health of a Component,” proceedings of the AHS 60th Forum, Baltimore, MD, 2004b.

NASA/TM—2007-214664 14
25. Bechhoefer, E., Mayhew, E., “Mechanical Diagnostics System Engineering in IMD-HUMS,”
proceedings of the 2006 IEEE Aerospace Conference, Big Sky, Montana, 2006.
26. Dempsey, P.J. Zakrajsek, J.J., “Minimizing Load Effects on NA4 Gear Vibration Diagnostic
Parameter,” NASA/TM—2001-210671, February 2001.
27. Mosher, Pryor, Huff, “Evaluation of Standard Gear Metrics in Helicopter Flight Operation,”
Proceedings of the 56th Meeting of the Society for Machinery Failure Prevention Technology,
April 15–19, 2002, Virginia Beach, Virginia.
28. Dempsey, P.J., Handschuh, R.F., and Afjeh, A.A., “Spiral Bevel Gear Damage Detection Using
Decision Fusion,” IEEE Aerospace and Electronics Systems Journal, vol. 18, Issue 9, September,
2003.
29. Dempsey, P.J., Mosher, M.M. and Huff, E.M, “Threshold Assessment of Gear Diagnostic Tools on
Flight and Test Rig Data,” NASA/TM—2003-212220, February 2003a.
30. Kacprzynski, Gregory J.; Roemer, M.J.; Orsagh, R.F., “Assessment of Data and Knowledge Fusion
Strategies for Diagnostics and Prognostics,” Society for Machinery Failure Prevention Technology,
H.C. Pusey, S.C. Pusey, and W.R. Hobbs, eds., Haymarket, VA, pp. 341−350, 2001.
31. Hall, D.L., “Mathematical Techniques in Multisensor Data Fusion,” Artech House, 1992.
32. Erdley, J.D.; and Hall, D.L., “Improved Fault Detection Using Multisensor Data Fusion,”
Proceedings of the 52nd Meeting of the Society for Machinery Failure Prevention Technology,
pp. 337−346, 1998.
33. Dempsey, P.J. and Afjeh, A.A. “Integrating Oil Debris and Vibration Gear Damage Measurement
Technologies Using Fuzzy Logic,” Journal of the American Helicopter Society, April 2004, Volume
49-Number 2.
34. Hall, David L.; Garga, A.K.; and Stover, J., “Machinery Fault Classification: The Case for a Hybrid
Fuzzy Logic Approach,” Society for Machinery Failure Prevention Technology, H.C. Pusey, S.C.
Pusey, and W.R. Hobbs, eds., Haymarket, VA, pp. 241−252, 1999.
35. McGonigal, Daniel Lewis, “A Comparison of Automated Reasoning Techniques for Condition-Based
Maintenance,” M.A. Thesis, Pennsylvania State University, 1997.
36. Zadeh, Lotfi A., “Outline of a New Approach to the Analysis of Complex Systems and Decision
Processes,” IEEE Trans. Syst. Man Cybern., vol. SMC−3, no. 1, pp. 28−44, 1973.
37. Mosher, M., Huff, E.M. and Barszcz, E., “Analysis Of In-Flight Measurements From Helicopter
Transmissions,” presented at the American Helicopter Society 60th Annual Forum, Baltimore, Md,
June 7–10, 2004.
38. Tumer, I.Y., Huff, E.M., “Using Triaxial Accelerometer Data For Vibration Monitoring of Helicopter
Gearboxes,” Proceedings of DETC’01 2001 ASME Design Engineering Technical Conferences
September 9–12, 2001, Pittsburgh, Pennsylvania.
39. Decker, H.J. and Lewicki, D.G., “Spiral Bevel Pinion Crack Detection in a Helicopter Gearbox,”
proceedings of the American Helicopter 59th Annual Forum, Phoenix, Arizona, May 6–8, 2003,
NASA TM—2003-212327, ARL–TR–2958.
40. Romero, R., Summers, H., and Cronkhite, J., “Feasibility Study of a Rotorcraft Health and Usage
Monitoring System (HUMS): Results of Operator’s Evaluation,” NASA CR–198446, Feb. 1996.

NASA/TM—2007-214664 15
 Form Approved
REPORT DOCUMENTATION PAGE
OMB No. 0704-0188
The public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the
data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this
burden, to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302.
Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB
control number.
PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
20-07-2007 Technical Memorandum
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Investigation of Current Methods to Identify Helicopter Gear Health
5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

Dempsey, Paula, J.; Lewicki, David, G.; Le, Dy, D.
5e. TASK NUMBER

5f. WORK UNIT NUMBER

WBS 877868.02.07.03.01.01
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
National Aeronautics and Space Administration REPORT NUMBER
John H. Glenn Research Center at Lewis Field E-15799
Cleveland, Ohio 44135-3191

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORS

National Aeronautics and Space Administration ACRONYM(S)
Washington, DC 20546-0001 NASA
11. SPONSORING/MONITORING
REPORT NUMBER
NASA/TM-2007-214664
12. DISTRIBUTION/AVAILABILITY STATEMENT
Unclassified-Unlimited
Subject Category: 07
Available electronically at http://gltrs.grc.nasa.gov
This publication is available from the NASA Center for AeroSpace Information, 301-621-0390

13. SUPPLEMENTARY NOTES

14. ABSTRACT
This paper provides an overview of current vibration methods used to identify the health of helicopter transmission gears. The gears are
critical to the transmission system that provides propulsion, lift and maneuvering of the helicopter. This paper reviews techniques used to
process vibration data to calculate conditions indicators (CI’s), guidelines used by the government aviation authorities in developing and
certifying the Health and Usage Monitoring System (HUMS), condition and health indicators used in commercial HUMS, and different
methods used to set thresholds to detect damage. Initial assessment of a method to set thresholds for vibration based condition indicators
applied to flight and test rig data by evaluating differences in distributions between comparable transmissions are also discussed. Gear
condition indicator FM4 values are compared on an OH58 helicopter during 14 maneuvers and an OH58 transmission test stand during
crack propagation tests. Preliminary results show the distributions between healthy helicopter and rig data are comparable and distributions
between healthy and damaged gears show significant differences.
15. SUBJECT TERMS
Gears; Transmissions; Health monitoring; Diagnostics

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF RESPONSIBLE PERSON
ABSTRACT OF Paula J. Dempsey
a. REPORT b. ABSTRACT c. THIS PAGES 19b. TELEPHONE NUMBER (include area code)
U U PAGE 21 216-433-3398
U
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std. Z39-18