IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

New real-time structural health monitoring microsystem for aircraft propeller blades MATHIEU LASTAPIS(1) CHRISTOPHE ESCRIBA(1) STEPHANE ANDRIEU(2) JEAN-YVES FOURNIOLS(1) (1)

CNRS ; LAAS ; 7 avenue du colonel Roche, F-31077 Toulouse, France Université de Toulouse ; UPS, INSA, INP, ISAE ; LAAS ; F-31077 Toulouse, France (2) RATIER Figeac; Avenue Ratier, BP. 2 46101 Figeac Cedex, France

Abstract – The monitoring of aircraft composite structures is a key challenge today within the research field of structural health monitoring (SHM). We have devised a microsystem able to monitor a propeller blade in real time. This monitoring improves blade safety and maintenance. The system works based on microsystem accelerometers that detect various events such as impacts, overspeed, and aerodynamic charges in order to diagnosis blade structural health. The diagnosis algorithm is the key point of this research. Index Terms – accelerometer, aircraft, autonomous embedded systems, blade, propeller, signal processing, structural health monitoring.

I. INTRODUCTION A. Blade brief history Blades are sophisticated requiring high engineering performance in terms of aerodynamics and mechanical design. Archimedes was the first to build an ancestor to the propeller when he constructed a screw for lifting water. Nineteen centuries later, the air screw was developed by Leonardo Da Vinci, who envisioned its use in a rudimentary helicopter. It was not until 1884 that mechanically controlled flight arrived on the scene: the airship “La France” was powered by an electric motor and a 7-meter propeller. The first controlled “heavier-than-air” airplane was constructed by the Wright brothers in 1903; they understood that a blade is like a twisted wing. The aeronautics industry was soon born and the number of aircraft types quickly multiplied. Made of wood or metal blade enhancements were developed to increase

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

performance in air battles: in-flight adjustable and reversible pitch, electrical blade de-icing. Last ten years more noticeable innovation was the development of composite structures: new Airbus A400M is propelled by four 8-blade composite propellers which counter-rotate (i.e. on the same wing, each propeller rotates in opposite directions). Today propellers are used almost exclusively by small planes or special purpose aircraft like the A400M. Yet with fuel prices increasing, there has been a renewed interest in propellers, particularly in the propfan, which is a mix between a propeller and turboprop that offers greater fuel economy while also surmounting the speed limitations of conventional propeller engines. B. Blade monitoring challenge A current challenge in propeller engineering is driven by the widespread use of composite propeller structures. Monitoring of composite propellers is necessary in order to know how the structure evolves over time and to save maintenance costs. The main problem with composites is the potential for internal damage which cannot be detected with an exterior inspection. Manufacturers thus need to know how and when internal damage can occur. Monitoring of blade structures is not only a current engineering challenge in the domain of aerospace [1-3]. In the energy industrial area, there is a need to monitor wind turbine blades [46]. Various methods of blade structural health monitoring has been proposed: piezoelectric wave actuator [5]; fiber optics integration into the blade [7]. A few companies have emerged to address this engineering challenge. One example is Insensys™, a company which focuses on wind turbines [8], Smart Fibres™ [9], which has developed so-called Fibre Bragg Grating (FBG) interrogators. Our technical research work is dedicated exclusively to aircraft propeller blades with industrial constraints: In order to continuously analyse blade health, it is necessary to develop a system that can monitor the structure over the lifetime of the blade thanks to embedding an electronic microsystem into the blade. This microsystem relies on an accelerometer that communicates to a central unit that can analyse signals and perform diagnostics. The detailed system that records data for post-processing is presented here. The completion of this first step in the project will enable the subsequent development of a real-time blade monitoring system. C. The aircraft blade: harsh environment for embedded electronics Numerous constraints, both exogenous (weather, temperature) and endogenous (blade engineering requirements) pose a major design challenge: Environment: temperature ranges per aircraft type can vary from -55°C to +90°C. This potentially

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

prevents the use of batteries, which do not work efficiently at very low temperatures and which cannot be recharged at -55°C. Moreover, even if the above temperatures represent outliers, temperature transitions are important. An aircraft can take off at 30°C and reach -30°C within a few minutes. We performed thermal tests by ramping temperature up/down in 10°C increments per minute. Rotational forces: a blade rotates at approximately 1000 rpm. On the A400M, blade speed is 842 rpm under normal conditions, but can increase to 1200 rpm if necessary. These rotating speed create a centrifugal acceleration amplified by blade length: values of 4000 g can occur at the blade tips! Non intrusive integration: microsystem must be embedded inside the blade without changing its performance characteristics.. We surmise that approximately 200 grams is a reasonable weight for this embedded system and the system should be able fit into a space of approximately 10 cm3. All electronics must fit into this space, and they must be immune to strong electromagnetic waves (field emission lower than 70 dB [µV/m], according to MIL-STD461 E standards) as well as water/solvent resistant. Blade impacts: impacts are a frequent occurrence, in most cases birds impacts during taking off and landing; in military aircraft, bullets or stones from provisional runways. All these shocks can obviously damage the blade, and may also damage the embedded electronics. This implies the need for a robust design that increases with rising weight and volume.

II. TOP-DOWN MODELING APPROACH Our microsystem is conceived for aircraft blades in the broadest sense, i.e. any rotating blade, including that on drone aircraft. Thus, while the system has the potential for use with wide range of propeller aircraft – such as the Aérospatiale C-160 Transall, Airbus A400M, ATR 72, and Northrop Grumman E-2 Hawkeye – the design we have implemented here could not match all blade variants with their specific requirements. As shown in Fig. 1, the design constraints are various and large in number. The environment is a key difficulty to overcome, and includes the need for robustness to temperature variation as well as resistance to humidity, fluids and dust. The specific design of the microsystem also needs to be adapted to the usage setting, which includes factors such as flight duration; operational time (which makes energy storage more or less important); and civilian or military use (bullet impacts, for example, are not considered in the former case). Also, these constraints imply the need to choose different electronic sensors. The diagnosis algorithm will also vary in relation to storage capacity and the energy source. Finally, the system geometry depends on the

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

specific design of each blade in terms of size, form and material. -------------------Insert Figure 1 here --------------------

The aim of the foregoing model is to identify the parameters that need to be varied when adapting the system to alternate usage settings. Similar to the above model, we devised a model to aid electronic design decisions. The model delineates the five main microsystem design parameters (see Fig. 2). Power supply is the first: the system design depends on the energy sources available. Are they available continuously (harvesting) or only periodically (scavenging)? How much power can they deliver? Which components are used to store that energy? Answering these questions leads to a specific power supply design. The other challenging issue is sensor choice, which depends on the energy source, space available, the sensor’s mechanical strength and measurement range. The remaining design parameters deal with timestamping, memory, interface (human-computer interaction) and diagnosis (microcontroller choice) issues. -------------------Insert Figure 2 here --------------------

III. MACRO AND MINI BLADE DESIGNS To apply these models, we developed a microsystem for two applications: a drone blade (mini-blade) and a Transall blade (macro-blade). Both blade designs contain commonalities, particularly with regard to electronic design specifications, although their fields of use are completely different. The drone blade recorder was our blade-monitoring prototype prior to the development of a blade microsystem for a fullsized aircraft. Our drone, which is 95 cm in length and 140 cm wide, has a brushless motor that powers two 25 cm propellers. Rotating speeds can reach 6000 rpm. The Transall blade, which has a length of 2.5 meters, depth of 40 cm, and thickness of 4 cm, is of course quite different. It rotates at 750 rpm with a maximum overspeed of 1400 rpm. Here the environment is harsher. The microsystem must handle temperatures of -40°C to +85°C and be resistant to water, oil and dust. IP65 was necessary as the plane can fly in various conditions. Both systems need precise data timestamping, large capacity non-erasable memory, fast interface link to extract recorded data, a microcontroller with analog-to-digital converters and without great calculatory

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

power, low-power and low-volume sensor, mechanically robust, with a frequency range less than 400 Hz for impact detection. On the other hand, electronic design parameters differ in the following areas: power source capacity and operating temperatures, memory capacity because of different flight times.

IV. DRONE BLADE RECORDER DEVELOPMENT Prior to making technology choices based upon the described models, we define a mechanical structure for the electronic board. As the flights conducted with the drone (Fig. 3) were quite short and took place in a controlled environment, in terms of electronics, power is provided by a CR2032 cylindrical cell. This powers the microcontroller, which has 6 analog-to-digital channels and an embedded USB controller to extract data. Prior to testing, data were previously recorded onto a 2 GB flash memory. Furthermore, the microcontroller uses an I2C bus to communicate with a real-time clock (RTC) to ensure precision time conservation. Concerning sensors, the wing-mounted accelerometer is a Measurement Specialities™ 832M1™ piezoelectric accelerometer able to measure accelerations up to 100g in three directions. It was chosen for its mechanical resistance and low power consumption (it functions with a maximum current consumption of 22 µA!). Detailed results have been analysed [10]. -------------------Insert Figure 3 here --------------------

V. DEVELOPMENT OF A TRANSALL BLADE EMBEDDED RECORDER A. Microsystem development Based on experiences made with the drone system, we then devised a blade parameter recorder to be embedded on a Transall military aircraft. The ultimate aim is to use these data to develop an algorithm for an autonomous flying system which will record only detected events. Like the drone system design, the aircraft blade recorder is divided into two parts to concentrate maximum weight as close as possible to the rotating centre to assume blade balance and avoid components ejection due to huge centrifugal acceleration. The two parts are a main board with primary functions, and a secondary board with the accelerometer

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

mounted on the blade. Recorder design follows the parameter considerations detailed in Section II (Fig. 4). The first technology choice concerns the power supply function: although flight tests are not very long (one to two hours), the embedded system must have the capacity to store energy while waiting for these tests. The wait period can last up to 8 weeks. Thus, the power source must have sufficient capacity to stay in stand-by mode and activate when blade rotation is detected. Moreover, the temperatures during these tests can vary from -40°C to +85°C. So an adapted cell was needed: a single Saft™ LiSoCl2 LS14250™ meets these requirements. This cell technology is effective from -60°C to +85°C [11]. Providing 3.6V without load, a buck-boost DC/DC levels this voltage to a continuous 3.3V, even when cell voltage drops. This cell supplies the entire board, including the blade-mounted accelerometer, with 1.2 Ah capacity. As with the drone monitoring system, the microcontroller has 6 analog-to-digital channels and USB controller features. In this case, however, data is recorded onto an 8 GB flash drive due to recording times which can reach be longer than 2 hours. An ultra-low power “control” accelerometer is implemented in this case, too. It is used to detect propeller starts and stops; useful to save energy during testing. Similar to our drone recorder, this system is based on an blade-mounted accelerometer. Here, however, the device – a three-axial, 100g Measurement Specialities™ 832M1™ piezoelectric accelerometer – is integrated into the blade composite fibres (Fig. 6). Wires link it to the main board (Fig. 5) at the blade foot. -------------------Insert Figure 4 here --------------------

-------------------Insert Figure 5 here --------------------

-------------------Insert Figure 6 here --------------------

This system is able to record the three axis signals from blade accelerometer at 2 kHz. It requires 20 mW in full-operational mode and only 1.3 mW in standby mode. The main electronic board is housed in an aluminium box to protect against environmental influences, including water, oil, etc. This housing also improves electromagnetic emissivity and susceptibility results. Its characteristics are detailed in Table 1.

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

B. Recorder testing for flight certification Prior to embedding such a system on a flying aircraft blade, it was necessary to run several tests. The following tests were conducted to confirm system compliance with specifications: -

thermal stresses,

-

electromagnetic emissivity,

-

electromagnetic susceptibility,

-

mechanical resistance to centrifugal acceleration,

-

mechanical resistance to vibrations.

In the following sections, we describe these tests and theirs results. Most of these tests were based on RTCA DO 160 [12] and EUROCAE ED14 norms [13]. 1) Thermal stress test Thermal stress tests were conducted to ensure the system is able to run at various extreme temperatures and during rapid temperature fluctuations. Particular attention was devoted to the testing of the embedded cell, as its proper performance under extreme temperature conditions is essential. At the same time, these tests provided information about electronic hysteresis and thermal sensitivity. In this way, the recorded accelerometer signals can be corrected for induced offset. Two kinds of thermal stress tests were conducted on the recorder. The first test was a thermal cycles test. For this test a running recorder was put inside a Temptronic™ Thermostream TP04300™ temperature test chamber. Six cycles were conducted, each one consisting of 60 minutes at -40°C followed by 60 minutes at +90°C. Temperature was modulated at 10°C intervals per minute. Under these conditions, the cell worked for 15h40m and the system stopped working during the 7 th cycle. Thus, the minimum of six cycles was successfully passed (see Fig. 7). Noteworthy is that cell performance degrades at -40°C. Voltage decreases quickly to keep sufficient current output. -------------------Insert Figure 7 here --------------------

Also noteworthy is that we did not test the temperature range all the way to -55°C but only to -40°C. The tested components were not designed for -55°C. The recorder is a first prototype with common electronic components, to be used for conducting a few hours of tests, and not permanently embedded. More robust

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

components will be used on the final system to assure compliance with a temperature range of -55°C to +90°C. The second stress test involved exposing the components to -40°C for 3h followed by +90°C for 3h. This test was successfully passed without any problems (Fig. 8). As with the thermal cycle test, cell voltage decreased more rapidly at temperatures under -20°C. -------------------Insert Figure 8 here --------------------

2) Electromagnetic emissivity and susceptibility Electromagnetic emissivity testing assures flight safety by verifying there is no interference between the embedded system and aircraft electronic systems. The required threshold varies between 30 and 70 dB (µV/m) depending on frequencies (see EUROCAE ED14/RTCA DO160 Section 8 on [12, 13] for more details). No problems were encountered during this test. We then conducted an electromagnetic susceptibility check to determine system immunity. Two tests were run to ensure our system can work under a wide range of electromagnetic fields:  test n°1: 30mA current at 5cm distance from main system,  test n°2: 7.5mA current at 5cm distance from main system. Various signals were analysed during these stress tests: power supply voltage, three axes of blademounted accelerometer, three axes of control accelerometer. -------------------Insert Figure 9 here --------------------

Several designers guidelines were extracted from this experience. First, the power supply chain is not affected at all by ambient electromagnetic fields. The same is true with regard to the three axes of the control accelerometer. However, with the blade-mounted accelerometer, perturbation occurred in the recorded signals (see Fig. 9). Two frequencies were involved: 500 MHz and 1 GHz. This corresponds to the length of the wires to the blade-mounted accelerometer: they act as an

antenna.

3) Mechanical resistance to centrifugal acceleration and to vibrations We must assume that the system would not be ejected or damaged during flight and keep alive its SHM

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

function. Battery cells contain electrolytes which can move under rapid acceleration, leading to reduced voltage output [14]. For the vibrations test, the goal was to check whether the components were wellsoldered and could handle the strong vibrations experienced by the blade. The first test conducted was the centrifugal force test. We placed the system on a rotating wheel for 24h under 900g of acceleration. The system passed this test: the cell was not affected by this acceleration. The second test was the mechanical vibrations test. The system was placed on a FH152 Transall aircraft blade. Alongside mechanical behaviour, we checked the system’s ability to record a wide spectrum of vibrations. The system also passed this test without problems. Fig. 10 is an extract of the test. More precisely, it is a window of a moment when the blade is excited at its eigenfrequency. The Y axis (the continuous green line) slowly increases before decreasing. -------------------Insert Figure 10 here --------------------

-------------------Insert Figure 11 here --------------------

Based on this test we conducted a frequency analysis to find blade eigenfrequencies (Fig. 11). Let us recall how such a graph is obtained: a fast Fourier transform (FFT) is performed using an equation (1). The absolute maxima are then determined above the half-FFT frequency range.

Where 1 ≤ k ≤ N, and N is the number of signal points.

VI. EMBEDDED PROCESSING The blade recorder allowed us to retrieve sufficient data to develop a processing routine able to detect impacts, overspeed, vibrations, general status. It was necessary to make this routine as simple as possible in order to choose a single microcontroller unit with the lowest power consumption.

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

A. Impacts Bird impacts are often a significant problem for blade manufacturers, as the impact strength can reach 38000J! Stone impacts are also a concern when planes use provisional runways. Most of the time, this is an issue in military applications. These impacts are less critical than bird impacts, and can be compared in force to the various impacts that are possible when the aircraft is stationary in the airport area. Bullets represent a final source of potential impacts. Of course, these types of impacts only occur to military planes during wartime. Yet they must be considered by the blade manufacturer and by us. 1) Signature detection -------------------Insert Figure 12 here --------------------

When an impact occurs to a blade, it resonates with the eigenfrequencies of the structure. These frequencies are a signature of the structure. They are well-known by the manufacturer and resonation with them must be avoided during normal use. Otherwise, damage to the blade could result. Thus, impact detection is performed by monitoring the blade’s particular eigenfrequencies. A power spectrum is processed from received accelerometer signals and the power of the frequencies is measured (see Fig. 12). If the power is above a predefined level, we can conclude that an impact occurred. Fig. 11 and Fig. 12 display frequency analyses on different blades. For this reason, the eigenfrequencies are not similar. Equation (1) shows how to compute a fast Fourier transform (FFT). Then, to get the energy density spectrum, equation (2) is applied. To compute the power equation (3) is added.

where fs is the signal sampling frequency. 2) Localization After having detecting an impact, the next step consists in localizing it. On a blade, impact localization is very important. The outer two-thirds of a blade are its weak points. This is where the blade delivers more power and where it can break most easily. To localize an impact, we work with the frequencies induced by the impact.

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

A blade is a partially constrained structure: the bottom is attached to the propeller hub and top is free. We find that an impact to the top of blade induces more low frequencies than one to the bottom (see Fig. 13), independent of impact energy. And this observation varies linearly from the top to the bottom of the structure. For example, we choose three characteristic eigenfrequencies, F1, F2 and F3, equivalent to 27 Hz, 63 Hz and 290 Hz, respectively. If F3>x.F2>x.F1 with x = 2, then the impact has occurred at the bottom of the structure. The value of x can be refined with each blade response. -------------------Insert Figure 13 (a and b) here --------------------

3) Lab demonstrator Our lab demonstrator is capable of detecting whether an impact has occurred in one of the three zones on the blade. The PC-based demonstrator processes accelerometer signals using the application National Instruments™ LabView™. It then displays the location of the impact in real time. -------------------Insert Figure 14 here --------------------

This demonstrator works based upon the following processing procedure: it first stores one second of data in its memory. After analysing blade eigenfrequencies, it detects whether an impact has occurred. It then computes a spectral energy analysis to compare various eigenfrequency levels. Finally, it indicates the location of the impact. Comparison between eigenfrequency energy is made using the three main frequencies at 27 Hz, 63 Hz and 290 Hz. Based upon which frequency is most energetic, the program can diagnose the impact location. B. Vibration detection A blade always vibrates when the propeller is rotating. In this section, we discuss “overvibrations”, which are caused by a blade’s mechanical inability to effectively overcome air resistance. This phenomenon can be amplified by a damaged structure: when the blade is less stiff, it vibrates more. Thus, the monitoring of vibrations is important in order to know:  if the blade has met adverse flight conditions,

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

 if the blade is damaged. The embedded accelerometer vibrates in unison with the blade. Thus, vibration detection can be performed through level detection. Each sample is compared to a pre-recorded threshold: if the voltage is higher or lower, an alarm is rung. -------------------Insert Figure 15 here --------------------

C. Overspeed Rotational blade speed is controlled by the pilot. Most of time, the propeller rotates at cruising speed. For example, A400M propellers are designed to rotate at 842 RPM under normal conditions. But sometimes the pilot needs extra power in particular situations, e.g. when taking off from a short runway, conducting an emergency flight manoeuvre, etc. One solution is to overspeed the propeller. This produces more power but subjects the blade to great stress. For this reason, overspeed must be monitored in order to know how the blade has been used and if further inspection is needed. -------------------Insert Figure 16 here --------------------

Overspeed detection is based upon the monitoring of vibration frequency. The vibrations shown on Fig. 15 oscillate at the blade rotating speed. Thus, running a frequency analysis based on equation (1) gives us the rotating speed (Fig.16).. D. Real-time embedded algorithm computation Based on the data gathered in our tests, we were able to compute an embeddable real-time algorithm that operates using the four previously detailed parameters. As discussed, maximum frequencies occur during impact events, which can rise up to 300 Hz. Thus, the sampling frequency is set to 600 Hz to avoid losing data. In the same way, events such as impacts do not last more than 200 ms. The microcontroller has to store 120 samples to match the aforementioned specifications, thus yielding a 200 ms window at 600 Hz. This acquired window is processed in two ways: a power spectrum analysis and a minima/maxima search. The first processing, which is a frequency analysis, allows us to measure blade rotating speed. If this speed is too high, an event is recorded in the memory with a time stamp. Yet frequency spectrum measurement is

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

also needed to analyze eigenfrequencies. If detected, these blade specific frequencies show that an impact is happening on the blade. After analyzing eigenfrequency amplitudes, impact localization is determined and then stored in the memory. The last part of this algorithm concerns over-vibration detection. This is performed by a minima and maxima search of the signal. -------------------Insert Figure 17 here --------------------

Each of these presented events is stored in the memory with timestamping. In order to save memory space and also save energy while writing in memory and extracting data, events are not stored in detail. Embedded intelligence, which can be stored on only 3 bits, classifies events on an 8 level scale. If an insignificant event happens, no record is made. But if a small impact is detected, for example, an event of level 1 is stored. On the other hand, a very damaging impact which implies a non-operational blade is stored as a level 8 event. This classification is applied to any events that the system can detect: overspeed, vibrations, etc. It should be noted that such processing cannot be run using a single microcontroller. The recorder architecture should be evolved into a digital signal processor that can compute this algorithm in real time.

VII. CONCLUSION This paper presents work done on the project “Blade Parameters Recording Microsystem” funded by the French research program FUI. It resulted in a fully operational recorder that can be embedded on a plane. This system allows blade monitoring to be conducted using computing algorithms. Using a top-down modelling method, this recorder can be adapted to any kind of blade. In this article, recorder was devised for a Transall plane blade. This device was used to record data in order to subsequently compute a real-time blade monitoring algorithm. The algorithm takes several blade parameters into account. The top-down modelling easily allows the algorithm parameters to be adapted to various blades. The algorithm calculations performed thus far suffer from some limitations. As we could not conduct impact tests with a rotating blade, all of our results were obtained from a static test without rotation. These results need to be validated with a rotating blade. Similar limitations were encountered concerning the monitoring of blade damage. For safety reasons, a damaged blade could not be mounted on an aircraft. A related problem concerned the monitoring of general blade health. Static tests give good results but must be

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

confirmed during various flights phases. The devised system needs to be tested further to validate our results. Thus, our calculations are based on a mix between ground tests and hypotheses. Despite these limitations, we have confidence in the presented algorithms. The next step in this project concerns impact quantification. We now know when and where an impact occurred, but how energetic was the impact? The most interesting solution is to measure the power of a harmonic in correlation with the place of impact. This is the route we are pursuing at present.

-------------------Insert Table 1 here --------------------

VIII. REFERENCES [1]

[2] [3]

[4] [5]

[6] [7]

[8] [9] [10]

[11] [12] [13] [14]

I. Bovio, E. Monaco, M. Arnese, and L. Lecce, "Damage detection and health monitoring based on vibration measurements and recognition algorithms in real-scale aeronautical structural components," Damage Assessment of Structures, vol. 245(2), pp. 519-526, 2003. C. Paget, "Bonded thin metallic foil sensor applications for structural health monitoring of aeronautical structures," Physical Mesomechanics, vol. 11, pp. 308-313, Sep-Dec 2008. M. Lastapis, C. Escriba, Y. Grondin, G. Auriol, P. Amat, S. Andrieu, et al., "Blade recorder microsystem design and validation for aeronautical structural health monitoring," in European Workshop on Structural Health Monitoring, Sorrento, Italy, 2010, pp. 100-105. C. C. Ciang, J. R. Lee, and H. J. Bang, "Structural health monitoring for a wind turbine system: a review of damage detection methods," Measurement Science & Technology, vol. 19, pp. -, Dec 2008. A. Ghoshal, M. J. Sundaresan, M. J. Schulz, and P. F. Pai, "Structural health monitoring techniques for wind turbine blades," Journal of Wind Engineering and Industrial Aerodynamics, vol. 85, pp. 309-324, Apr 24 2000. M. A. Rumsey and J. A. Paquette, "Structural health monitoring of wind turbine blades," Smart Sensor Phenomena, Technology, Networks, and Systems 2008, vol. 6933, p. 450, 2008. A. Turner, T. W. Graver, M. A. Rumsey, and A. Mendez, "Wind turbine structural health monitoring using optical fiber based sensors," in International workshop on structural health monitoring, San Francisco, 2009, p. 1587. http://www.insensys.com, 2010 http://www.smartfibres.com, 2010 M. Lastapis, C. Escriba, G. Auriol, E. Albu, P. Berthou, J.-L. Boizard, et al., "Embedded blade microsystem and events recorder for drone structural health monitoring," in Asian-Pacific Workshop on Structural Health Monitoring, Tokyo, Japan, 2010. www.saftbatteries.com, 2010 www.rtca.org, 2010 www.eurocae.net, 2010 H. Cheng, K. Scott, and C. Ramshaw, "Intensification of water electrolysis in a centrifugal field," Journal of the Electrochemical Society, vol. 149, pp. D172-D177, Nov 2002.

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Table 1: Recorder characteristics

Environment -Temparature -Humidity -Liquids -Dust

Electronics

Use -Flight time -Operational time -Civilian or military

Microsytem constraints

-Measure -Diagnosis -Storage -Energy

Blade charateristics -Mass -Rotating speed -Material -Length -Geometry

Fig. 1: System constraints model

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE Power supply

Timestamping

-Harvesting? -Scavenging? -Power available -Storage support

Sensor choice -Energy -Volume -Mechanical strength -Frequency range

-Precision degree -Time sensitivity

Embedded electronic design

Intelligence

Memory -Capacity -Access frequency

Interface -Power available -Wired or wireless? -Wireless frequencies authorized -Range needed -Data rate

-Program memory -Complexity

Fig. 2: Electronic design parameters

Fig. 3: Drone picture with embedded system blow-up

LiSoCl2 cell Li battery Deicer power

Power supply conditioning Control accelerometer

RTC

Main board

Microcontroller

Deported accelerometer

RF link

Blade-mounted board

Operator

Fig. 4: Functional diagram of the blade recorder

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Fig. 5: Picture of the 63mm x 32mm microsystem, 8mm thick without power cell and 18mm thick with cell

Fig. 6: Picture of blade-mounted accelerometer on Transall blade

Fig. 7: Temperature and cell voltage during thermal cycles

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Fig. 8: Temperature and cell voltage at -40°C and +85°C.

Fig. 9: Electromagnetic susceptibility on the blade-mounted accelerometer – Z signal

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Fig. 10: Three axis recorded signals from blade-mounted accelerometer

Fig. 11: Blade spectral analysis during vibrations test (1- second FFT window) (blade type n°1)

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Fig. 12: Eigenfrequencies stimulated by an impact (blade type n°2)

(a)

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

(b) Fig. 13: Relative amplitude shift depending on impact localization – (a) indicates a top impact and (b) a bottom impact

Fig. 14: Screen capture of LabView™ localization program

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Fig. 15: Illustration of vibration and technique to detect “overvibrations”

Fig. 16: Measurement of the rotating speed based on frequency analysis

IEEE AEROSPACE AND ELECTRONIC SYSTEMS MAGAZINE

Acquisition N 120 samples acquisition (fs = 600Hz; Tacq = 200ms)

Eigenfrequencies?

Frequency analysis

Min and max measurement

Speed measurement

Above or under threshold? Y

Y

Localization by eigenfrequencies amplitude analysis

Overspeed? Y Y

Impact detection and localization

Overspeed detected

Overvibrations detected

Event memory storage with timestamping and stored by level

Fig. 17: Algorithm to monitor propeller blades