Final Report Summary - SCOPE (Self-sensing Curved composite panel under Operational load: methodology Platform for prediction of damage Event)
Executive Summary:
This final report summarises the most significant work performed during the 2 year project SCOPE. In addition the validation of the methodologies against published results and experimental measurements are reported for the first time.
The main objective of the SCOPE project was to develop, verify and validate algorithms for Structural Health Monitoring (SHM) of a curved composite panel. The final deliverable reports on:
• the adopted methodologies for impact force magnitude and location,
• adopted methodologies for damage detection, identification and characterisation,
• self-diagnostic SHM system,
• methodologies for effective positioning of sensors and actuators,
• statistical analysis and probability of false alarms and
• a dedicated software platform
each of which have been detailed in separate deliverables. SCOPE’s work was dedicated to development of methodologies only. However, in this report the above methodologies are validated at coupon, simple stiffeners panel and a curved fuselage panel against numerical and experimental measurements carried out by ICL in other projects. Following the validation, it can be concluded that the methodologies developed perform with a high degree of accuracy and robustness for self-diagnostics, impact detection and damage detection.
Project Context and Objectives:
SCOPE project has resulted in an effective and efficient SHM methodology platform, for passive and active sensing together with optimisation algorithm applicable to curved sensorised fuselage panels under real load conditions. The project involved extending some of the methodologies developed in [3] for flat composite panels and developing new methods specifically to address curved panel of size 5m x 1.7m and radius 4.5 m. The significant new developments and innovations in developing the SCOPE platforms include:
• Impact detection and identification: The only reported method capable of accurately capturing any impact event (location, force magnitude, duration) on a composite panel, even if it causes damage in the structure and thus result in non-linearity in the response, is based on meta-modelling techniques [3, 5, 6].Therefore developing an accurate meta-model is imperative for a reliable impact detection algorithm. To achieve a 95% Probability of Detection (PoD) an order of 103 different impact scenarios (locations and energies) were required for generating the meta-model. To speed up the solution, a robust contact algorithm was developed to simulate all possible dynamic non-linear impact scenarios. MPI libraries were used to handle and process the large amount of simulated data in an efficient manner. Meta-models once generated were implemented on a small Laptop/PC to provide a rapid solution. For a fuselage panel, impacts can occur both from inside (e.g. tool drop during maintenance) and from outside (e.g. debris) which dues to the curvature of the panel can result in completely different impact dynamics. A novel impact categorization approach has been implemented to identify the type of impact (inside or outside) and thus estimate the location and level of contact force using the dedicated developed meta-model which resulted in higher probability of detection, POD.
• Sensor modelling: The proposed SMART sensor models incorporates the realistic effect of sensor on curved composite. The elements are capable of both sensing and actuating which has been validated against experimental results.
• Damage detection and characterisation: Damage detection algorithms suitable for large scale sensorised curved panels were developed based on probability based diagnostic method using different feature extraction techniques and damage indices of guided waves. SMART FE developed and validated for flat composite stiffened panels [3, 4] was extended to simulate sensing and actuating of Guided Ultrasonic Waves in curved panels. In complex curved panels, when PZT is used as an actuator, multiple modes of wave exist and their dispersion behaviour is complicated. Dispersion and attenuation of all active modes (different in curved panels) was included in designing appropriate actuation signal and data post-processing (WP2.1). Large scale FE analysis will be achieved through parallel dynamic simulations on High Performance Computing (HPC) units.
• Electro-Mechanical Impedance (EMI): EMI has proved highly effective for sensor self-diagnostics and detection of damage in the close vicinity of the transducer such as the foot or under the stringers [3, 7].The proposed EMI SMART model was used to directly measure the impedance response of the curved panel under dynamic load. The effectiveness of EMI method in terms of self-diagnostic, detectability range and effective frequency sweep range for curved composite panels were demonstrated for the first time.
• Optimisation of sensor layup: Optimisation was complex due to different characteristic of curved panel; that is the difference in sensor signals obtained by impacts from inside (tool drop) and outside (debris) of the panel is considered. Standard application of optimisation methods using exhaustive search algorithms for finding the optimum sensor layup of a 5m x 1.7m curved panel would require unrealistic number of simulations. The proposed parallel HPC multi-objective optimisation method based on Genetic Algorithm will significantly reduce the number of simulations and hence the computational cost.
An innovative optimisation approach based on maximum area coverage [9] was applied to the SCOPE panel. This method which is a physically based approach is shown to provide an effective optimisation strategy by minimising blind spots in the panel.
Considering all the above mentioned innovations, SCOPE platform offers the following specifics:
• Passive sensing through meta-models capable of detecting impact force and location;
• Active sensing through methodologies based on SMART FE, capable of detecting and characterising different types of impact damage;
• Self-diagnostic properties through “SMART sensors” models by utilising their Electro-Mechanical Impedance (EMI) properties;
• Optimal sensor position through a multi-objective optimisation algorithm based on the developed meta-models for impact and damage detection algorithms.
Finally, the developed methodologies were validated against in-house background experimental measurements through a building block approach (coupon, mono stringer and curved fuselage panel). Based on the findings of SCOPE, recommendations for test procedures are presented.
In this final deliverable a summary of the methodologies developed and reported throughout the project in different deliverables will be presented together with some results. Finally, the validation and verification of the developed methodologies against numerical and experimental results will be reported for the first time.
Project Results:
2 Methodologies for SCOPE platform
The main SHM work carried out in SCOPE was used to develop a platform through a Graphic User Interface (GUI) for impact and damage detection (BVID) of a curved fuselage panel, see Figure 1.
Figure 1: SCOPE SHM platform
The platform has three main capabilities:
• Active sensing through Ultrasonic Guided Wave excitation and interrogation resulting in damage detection and characterisation
• Self-diagnostics through EMI measurements to increase the reliability and robustness of the SHM system by detecting faulty sensors
• Passive sensing through developments of meta-models resulting detecting, locating and categorizing the type of impact event
The functionality of each option is presented in more detail below.
2.1 Self-diagnostic SHM system
The aim of the self-diagnostic capabilities for an SHM system is to reduce the probability of false alarms or missed-detection, which can be catastrophic, due to faulty sensors. EMI measures are sensitive to both faults within the sensor (damaged sensors/ bond quality) and damage in the vicinity of the transducer. The EMI is a complex spectrum which spreads over a frequency range. For self-diagnostic applications, previous studies have shown that the imaginary party of this spectrum is capable of identifying faults related to the transducer. Presence of damage can either shift the resonance frequency of the structure or reduce the amplitude or a combination of both depending on the type and severity of the fault. Therefore the self-diagnostic module in SCOPE platform results in three different damage metrics based on comparison of EMI measures with baseline values: RMSD, MAPD and CCD (each representing different change in the overall dynamics of the structure) which can result in damage detection and characterisation. The severity of damage is also shown to be directly correlated to the proposed damage indices. A threshold value must be set for each of the presented damage indices in order to detect damage with high reliability.
Figure 2: Self-diagnostic module - SCOPE platform
2.2 Impact detection and characterisation
The challenge of developing a passive SHM sensing capable of detecting an impact event (location and force magnitude) in a curved composite panel is considerable difference in the dynamical response of the structure to impact occurring from inside and outside the structure, see Figure 3 for the difference between the contact force produces by impacts of the same energy and location but from opposite side.
(b) POD for impact location, outer panel
(a) Contact force (c) POD for impact location, inner panel
Figure 3: Examples of impact force caused from impacts on inside and outside of the panel
SCOPE platform overcomes this challenge by implementing a robust categorization algorithm to identify the side of an impact prior from sensor readings prior to detecting its location and force magnitude. Different meta-models were created for impacts occurring from inside and outside of the structure which resulted in noticeable reduction of the error. Meta-models were built by developing and training of Artificial Neural Networks (ANNs) which take into consideration non-linarites of the structural response. The input to the passive sensing module is recorded sensor data (experimental or numerical) in forms of discrete signals. After categorizing them in terms of the side of the impact, the platform will then perform the appropriate feature extraction technique to generate inputs for impact location and force magnitude detection. Having a set threshold for contact force magnitude, which is directly related to initiation of damage in composite structure, the impact event can then be labelled as alarming if it exceeds the value.
Figure 4: Passive sensing module - SCOPE platform
The required data (in the order of 103 simulations) for development of ANNs were generated following these steps:
• Developing a reliable FE model of the curved panel
• Developing a validated impact simulation together with a robust contact algorithm
• Developing a parametric code which can write different input files to submit analyses of the curved panel to an HPC platform
• Developing a code for post-processing of data (generating discrete signals for each sensor)
To assess the reliability and robustness of the passive sensing module in SCOPE, a statistical analysis was carried considering different impact scenarios (varying impact velocities, position: inside and outside, mass/size of impactor) and various network architecture. This study in particular led to:
• definition of POD for passive sensing (to identify the optimum network architecture)
• assessment of the passive sensing module in the presence of real operational load
• determination of the probability of false alarm
2.3 Damage detection and characterisation
The active sensing module developed for SCOPE panel is focused on damage detection algorithms based on UGWs applicable to curved composite panel. In particular it addresses challenges such as directionality of wave propagation, additional modes (longitudinal, torsional and flexural modes) and geometrical complexities (reflections from stiffener/frames, changing thickness, attenuation, superposition of various modes and reflections, noise). For development and validations of the active sensing modules, numerical simulations of actuation, wave propagation, interaction with damage and sensing have been carried out. As wave propagation is a non-linear dynamic analysis, a SMART FE was developed and implemented to model the actuation and sensing of the Lamb waves. Numerical procedures for appropriate bonding of flat PZT disks to curved surfaces of the panels were investigated and outlined. In addition, the SMART FE algorithm provides parameters of the numerical integration and minimum mesh size for a stable and converged representation of lamb waves and damage detection for different actuation frequencies and input signals.
Figure 5: Active sensing module - SCOPE platform
A total of 273 possible damage positions were included in the model. Damage is modelled as softening of the material in one or several layers. This type of damage, characterised through analyses on coupon level using the Hashin model, is the best representation of impact damage, which is the type of defect which has to be detected in the current project. Different sizes and severities of damage will change the damage index in the detection algorithm which results in damage characterization.
The developed damage detection algorithm is based on probabilistic imaging technique where the probabilistic damage indices measured from each transducer path is fused together for the whole structure and presented as a probability plot, with the highest value indicating the presence of damage, as shown in Figure 5.
The capabilities of the developed active sensing module were tested with statistical confidence against the following variables:
• Noise
• Type of bonding
• Failure of one or more sensors
• Sensitivity to sensor placement
For example the sensitivity of the damage detection algorithm to sensor failure was carried out for three various positions of damage: damage in the bay, damage close to the stringer and damage under the stringer. The results presented in Figure 6 illustrates the bar plots of error in damage detection due to sensor failure. It is concluded that damage under the stringer has the highest sensitivity to sensor failure. This information can be used to tailor the optimized sensor positions and redundancy of the developed network separately for various parts of the structure.
(a) damage in a bay (b) damage close to the stringer (c) damage under the stringer
Figure 6: Effect of sensor failure
The results of the statistical analysis were then used to in flight test procedures (deliverable D8).
2.4 Optimisation
A key aspect of any SHM system is the optimum number and placement of sensors to reach a certain level of reliability and POD. In SCOPE methodologies for effective positioning of sensors and actuators for both passive and active systems have been developed. A brief description of each optimisation methodology is outlined in the following sections.
2.4.1 Optimal sensor placement for impact detection and identification
The proposed optimization algorithm to determine the best sensor position is based on Genetic Algorithm (GA) by minimizing a fitness function. The fitness function for passive sensing is the performance of the developed meta-models for impact detection and identification (to minimize the error corresponding to 90% POD). The input to the GA in the first step is the minimum number of sensors for which the optimisation will find the best locations out of all the possible sensor locations. If the POD for impact detection related to that sensor combination is not satisfactory the number of sensors will be increased and the optimisation will repeat again. It is worth mentioning that the developed optimisation algorithm is tailored for the curved composite panel analysed in SCOPE which is not-symmetric due to directionality of the wave propagation velocity, change in thickness across various bays and different cross-sectional profiles. Therefore the symmetry of the problem cannot be used to simplify the optimisations procedure. Figure 7 illustrates the best 8 and 4 sensor locations out of 39 possible ones for a section of the fuselage panel; whereas Figure 8 represents the optimum sensor location within one bay only where the sensors are places symmetrically following the symmetry of the part of the structure.
Figure 7: Best senor positions for an investigation using 8 and 4 sensors in seven bays.
(a) Best 4 sensors placement in the central bay (b) Example of GA results
Figure 8: Example of GA results.
2.4.2 Optimal transducer number and location for damage detection and characterisation
A novel approach for optimal sensor positioning based on geometrical considerations has been adopted in SCOPE by adopting the methodology previously developed at ICL for a curved stiffened panel [9]. This approach is more general and always valid in the sense that it does not depend on the evaluation function which varies for each damage detection methodology. As long as the detection method is based on the time of arrival of damage reflected wave the proposed optimisation is effective. The methodology is based on maximum coverage area while minimizing the boundary reflections in the structure which can cause inaccuracies in the result. The fitness function to be maximized is defined as the sum of coverage index defined from each sensor-actuator path for every point in the structures. Therefore the result of the optimization will be an image illustrating the coverage area of the structures for a given number of transducers, see Figure 9.
4 sensors at the corners. 4 sensors at mid-edge position
Figure 9: Example of coverage area related to different transducer location
Once the fitness function to be optimised is defined, the next step is to develop an optimisation algorithm. For the purpose of SCOPE, GA has been applied to reduce the computational size of the problem. Parameters such as population size, percentage of elite genes, mutation and cross over was carefully chosen to result in a converged solution. An example of the best sensor placements for 6 transducers is illustrated in Figure 10. If the POD related to the optimal 6 transducers not acceptable, then the number of transducers will be increased and the optimisation procedure repeated.
Figure 10: Coverage and positioning for 6 sensors
3 Validation and calibration
In this section the validation and calibration of the developed methodologies against published results and in house experiments are reported. Sensor signals are calibrated with coupon test results. Calibrations of the contact force for impact detection and of Lamb-waves for damage detection have been performed and reported. A multi-scale approach has been followed for both passive and active sensing.
The performance of the SHM system with optimal sensor network has been compared to a random sensor network for the assessment of the optimisation technique for damage detection.
3.1 Validation for passive sensing
For passive sensing, the validation and calibration of signals and developed methodologies follows a building block approach shown in Figure 11.
Figure 11: Building block approach for validation of passive sensing.
The validation consists of the following steps:
• Experimental impacts were conducted at coupon level using a drop tower
• Contact force and PZT signals were acquired and compared to numerical results
• A multilevel approach (numerical and experimental) was conducted on coupon up to curved fuselage panel
The validation against published results was performed with respect to the papers listed in the reference section.
Figure 12: Drop tower used for the validation of passive sensing.
The experimental contact force at different impact energies was compared to numerical simulations, for impacts on both inner and outer surface of the panel. As it is possible to see from Figure 13, the agreement between numerical and experimental results is very good. Moreover the experiments confirm that the major complication arising when curved panels are involved is due to the different behaviour of the resulting contact force when the same impact occurs on the different sides of the panel.
Figure 13: Validation of contact force.
The response of the PZT was as well validated as presented in Figure 14. The time of arrival of the waves are very similar and also the magnitude of the first peaks, which are the parameters used for training the ANN, are in good agreement which suffices for the purpose of the develped activities in SCOPE. A frequency force reconstruction methodology, already presented in [1], was tested on the curved cupon panel and showed excellent resuts, see Figure 15.
Figure 14: Validation of the PZT signal.
Figure 15: Experimental force reconstruction.
3.2 Validation for active sensing
Validation of the active sensing methodologies developed for SCOPE followed a multi-level approach as well, Figure 16. Both Lamb-wave investigations and EMI methods were validated, in terms of signals (actuation, propagation and sensing) and damage detection capabilities.
Figure 16: Validation approach for active sensing.
In summary the validation followed the following steps:
• Experimental investigations were conducted at coupon level
• Numerical and experimental signals were compared and showed good agreement
• The damage detection methodology was able to detect BVID
• A multistage approach (numerical and experimental) was conducted
o Flat coupon
o Curved coupon
o Stiffened element
o SCOPE panel
3.2.1 EMI Technique
Self-diagnostics was performed at coupon level under different sensor failure conditions. Damage consisted of 2 different crack sizes on a PZT transducer. From the signals presented in Figure 18 it is evident that a difference in the spectrum of the signal is present due to damage. In order to quantify this difference, the damage metrics were evaluated and the results showed that the methodology is able to detect the damage in the sensor as well as its severity [2].
Different levels of debonding between transducers and plate were also tested. Three levels of debonding 25%, 50% and 75% were considered, Figure 20. Again results showed that the EMI approach is able to determine the extension of damage, showing a damage index which increases with the severity of the debonded area, Figure 21 [3].
Figure 17: Experimental setup for the EMI investigation.
Figure 18: Self-diagnostics via EMI for a cracked PZT.
Figure 19: Crack evaluation via RMSD.
Figure 20: Different debonding cases.
Figure 21: Debonding evaluation via EMI.
The capabilities of the EMI investigation to detect damage in the structure were evaluated with two different experiments, one at coupon level and one at sub-component level. In both cases the damage was a BVID created by impacting the plate with a drop tower. Figure 22 [3] and 23 show that the methodology is robust under different scales of investigations as the damage position can be identified in both cases by showing higher values of the DI for the transducers closer to the damage site. Moreover, different damage metrics introduced (RMSD, MAPD, CCD) is capable of giving information on type of damage as well (whether there is a shift in resonance frequency or amplitude difference).
Figure 22: Damage evaluation at coupon level via EMI.
Figure 23: Damage detection via EMI at sub-component level [14].
3.2.2 LAMB WAVES
The validation of Lamb wave signals was performed using the set up shown in Figure 24. Again the building block approach was used to go from coupon to sub-component level. The first step was to validate the developed SMART FE in terms of sensing and actuating in comparison to experimental signals. As it is presented in Figure 25, there is good agreement between the FE and experimental signals.
Figure 24: Experimental setup for Lamb wave investigations.
Figure 25: Validation of Lamb-wave signals.
Figure 26: Curved coupon panel under analysis.
The coupon panel shown in Figure 26 was impacted to produce BVID. The result of the impact was checked via C-scan and the presence of damage was evident, Figure 27a). The damage detection algorithm was tested on this example (both with 4 and 8 transducers) and the results obtained are in excellent agreement with the actual size and location of damage, see Figure 27b) [4].
Figure 27: a) C-scan of the plate. b) Damage detection evaluation with 8 transducers.
In the next step, the active sensing methodology was validated against tests for stiffened coupon panel. The details of this study is presented in [5]. Some of the results are highlighted in Figure 28 which shows the validation of the methodology for a more complex coupon.
Figure 28: a) Experimental panel. b) debonding detection.
Validation of damage detection was performed at sub-component level on a curved stiffened composite panel, for a BVID under the stringer. Also in this case damage was properly evaluated and located.
Figure 29: Sub-component specimen with PZTs installed on its inner surface [15].
Figure 30: Damage evaluation at sub-component level.
4 Optimisation
To validate the optimal sensor placement for damage detection, two numerical examples were carried out: One with the optimal sensor placement and the second one with random sensor placement. The aim of the optimisation in this case was to find the best sensor placements for the minimum required sensors (8 transducers in this case) which could detect damage in the correct bay in a large (~1.7m x 0.6m) section of a curved fuselage panel. The error in the damage detection with the optimal sensor network resulted to be smaller than the random sensor placements. More importantly, in both cases even though the error in locating the damage was high but it is worth mentioning that this results was obtained with to 8 transducers covering a large area and more importantly the optimal sensor network was able to detect damage in the correct bay. This optimization approach can be used in a multi-level manner to result in damage detection and identification in various levels to reduce the interrogation cost of the SHM system. For example, level 1 detection will utilise minimum number of sensors to detect damage and locate the bay in which the damage is present. Level 2 interrogation will consist of a more local and detailed analysis adopting a denser transducer network. The proposed optimisation algorithm is able to adapt to the multi-level requirements.
(a) Coverage area of the optimal 8 sensors (b) Optimal positioning for 8 sensors (c) Comparison between optimal and random placement for damage detection
Figure 31: Validation of Optimal sensor placement - Active sensing
5 Conclusions
An SHM platform was developed during SCOPE for impact and damage detection including self-diagnostic capabilities applicable to large curved composite fuselage panel. Self-diagnostic based on EMI technique was successfully shown to detect faulty sensors as well as damage in the vicinity of the transducers. The developed methodologies for passive sensing is based on ANN and is capable of detecting impact location and force magnitude with 90/95% POD. The meta-models were developed by a large library of data generated via FE simulations validated against experimental data. The FE models include material and geometrical non-linearity as well as a robust contact algorithm to allow large number of impact scenarios (varying mass, location, velocity) to be analysed. The analysis was carried out on HPC units and feature extraction applied to the output data. A reliable categorisation technique was proposed to classify impact events from the recorded sensor data and use the appropriate meta-model to characterise the impact event accordingly. This resulted in an increase in performance and reliability of the network.
Following a building block approach, all of the developed methodologies for active sensing were validated for detecting BVID in various locations (i.e. mid-bay, foot of the stringer, under the stringer) and with different severities. The performance of the platform was tested and validated for curved composite panels with complexities such as directional velocity profile, change in thickness, attenuation profile. It was shown that the developed methodologies are applicable for multi-level analysis of complex structures such as the SCOPE panel.
Optimisation procedures for sensor placement for both passive and active sensing were developed based on GA. For passive sensing the objective function was the performance of the ANN for impact detection and identification. The number and location of the sensors were optimised to reach 95% POD. For active sensing a physically base maximum area coverage was implemented for curved panel to ensure maximum detectability with minimum blind areas.
Statistical analysis was carried out for both passive and active sensing to investigate the sensitivity to sensor placement, sensor failure, real load conditions and noise. Guidelines for test procedures were provided. Recommendations for signal acquisition and data processing were provided.
6 References
[1] Thiene, M., M. Ghajari, U. Galvanetto, and M. Aliabadi, Effects of the transfer function evaluation on the impact force reconstruction with application to composite panels. Composite Structures, 2014. 114: p. 1-9.
[2] Sharif Khodaei, Z., Impact Damage Detection in Composite Plates using a Self-diagnostic Electro-Mechanical Impedance based Structural Health Monitoring System. SDHM, 2015.
[3] Sun, J., Structural Health Monitoring on composite panels with high grequency electromechanical impedence (EMI) technique, Imperial College London, final year MSc project, 2014, supervisor: M.H. Aliabadi.
[4] Thiene, M., Z. Sharif Khodaei, and M. Aliabadi, Damage detection in curved composite panels. Under review at SMS, 2015.
[5] Sharif-Khodaei, Z. and M. Aliabadi, Assessment of delay-and-sum algorithms for damage detection in aluminium and composite plates. Smart Materials and Structures, 2014. 23(7): p. 075007.
[6] Sharif Khodaei, Z., Aliabadi, M.H. Lamb-wave based damage detection in anisotropic composite plates (2015) Key Engineering Materials, 627, pp. 1-4.
[7] Meng, S., Sharif Khodaei, Z., Aliabadi, M.H. Localization of barely visible impact damage (BVID) in composite plates (2015) Key Engineering Materials, 627, pp. 217-220.
[8] Sharif Khodaei, Z., Bacarreza, O., Aliabadi, M.H. Lamb-wave based technique for multi-site damage detection (2014) Key Engineering Materials, 577-578, pp. 133-136.
[9] Thiene, M., Z. Sharif Khodaei, and M. Aliabadi, Optimal sensor placement for damage detection based on ultrasonic guided wave damage detection algorithm (2015), Key Engineering Materials, to appear.
[10] Thiene, M., Sharif-Khodaei, Z., Aliabadi, M.H. Comparison of artificial neural networks and the transfer function method for force reconstruction in curved composite plates (2015) Key Engineering Materials, 627, pp. 301-304.
[11] Thiene, M., Galvanetto, U., Ghajari, M., Aliabadi, M.H. A frequency analysis applied to force identification (2013) Research and Applications in Structural Engineering, Mechanics and Computation - Proceedings of the 5th International Conference on Structural Engineering, Mechanics and Computation, SEMC 2013, pp. 2193-2198.
[12] Ghajari, M, Sharif-Khodaei, Z., Aliabadi, M.H. Apicella, A. Identification of impact force for smart composite stiffened panels (2013) Smart Materials and Structures, 22 (8).
[13] Sharif Khodaei, Z., Ghajari, M., Aliabadi, M.H. Determination of impact location on composite stiffened panels (2012) Smart Materials and Structures, 21 (10)
[14] Thiene, M., Salmanpour, S., Sharif Khodaei, Z. and M.H. Aliabadi, Experimental verification of a damage detection methodology in real aeronautical structures based on the proper orthogonal decomposition of the electro-mechanical impedance (2015, to appear.
[15] Salmanpour, S., Sharif Khodaei, Z., Thiene, M. and M.H. Aliabadi, Multi-level damage detection and characterization in a sensorized composite fuselage panel (2015, to appear.
Potential Impact:
The topic that is addressed by the proposal is JTI-CS-2012-3-GRA-01-051 Modelling and simulation of self sensing curved composite panel to predict to predict/control damage evolution in real load condition.
The integration of sensors into composite structures is a structural health monitoring concept to overcome the actual constrains by certification as the demand for short inspection intervals. The innovative elements are smart methodologies for damage detection, location of impact and optimization of sensor layout. Design and sensorised panel and definition of test procedure for monitoring different impact events under real load conditions should have a significant impact in terms of the take up of SHM for aircraft panels.
These tools should assist in defining configurations of sensors and actuators in the design of smart structures that would allow for effective detection of impact events and possible subsequent damage. Structures designed in this way will be capable of immediate detection of impact and monitoring of subsequent behaviour during service to achieve the objectives of low weight configuration. The methodologies would through the optimum sensor layout not only ensure there are no “blind area” form damage detection, they will also minimise the extra weight due to the introduction of sensors.
The currently applied conventional inspection techniques demand an access to the area that has to be inspected. The removal from service to provide this access, leads to costly long down-times for aircrafts during maintenance. The time saving due to SHM is already an advantage, but the approach proposed in SCOPE goes far beyond that. The rapid measurement of the health of the structure allows inspection intervals which are much longer that actually possible.
In summary, SCOPE may enable, in long term, further reduction of overall weight and hence fuel consumption and operative costs by reduction of design allowables. In short to medium term, SCOPE will reduce inspection times (scheduled and unscheduled), hence decreasing the operative cost and time on the ground.
List of Websites:
Professor Ferri M.H.Aliabadi
Head of Department and Professor of Aerostructures
Department of Aeronautics
Imperial College, London
South Kensington Campus
London SW7 2AZ
Tel: +44 (0) 20759 45077
http://www3.imperial.ac.uk/aeronautics
PA: Miss Lisa Kelly
l.kelly@imperial.ac.uk
Tel: 020759 45056
This final report summarises the most significant work performed during the 2 year project SCOPE. In addition the validation of the methodologies against published results and experimental measurements are reported for the first time.
The main objective of the SCOPE project was to develop, verify and validate algorithms for Structural Health Monitoring (SHM) of a curved composite panel. The final deliverable reports on:
• the adopted methodologies for impact force magnitude and location,
• adopted methodologies for damage detection, identification and characterisation,
• self-diagnostic SHM system,
• methodologies for effective positioning of sensors and actuators,
• statistical analysis and probability of false alarms and
• a dedicated software platform
each of which have been detailed in separate deliverables. SCOPE’s work was dedicated to development of methodologies only. However, in this report the above methodologies are validated at coupon, simple stiffeners panel and a curved fuselage panel against numerical and experimental measurements carried out by ICL in other projects. Following the validation, it can be concluded that the methodologies developed perform with a high degree of accuracy and robustness for self-diagnostics, impact detection and damage detection.
Project Context and Objectives:
SCOPE project has resulted in an effective and efficient SHM methodology platform, for passive and active sensing together with optimisation algorithm applicable to curved sensorised fuselage panels under real load conditions. The project involved extending some of the methodologies developed in [3] for flat composite panels and developing new methods specifically to address curved panel of size 5m x 1.7m and radius 4.5 m. The significant new developments and innovations in developing the SCOPE platforms include:
• Impact detection and identification: The only reported method capable of accurately capturing any impact event (location, force magnitude, duration) on a composite panel, even if it causes damage in the structure and thus result in non-linearity in the response, is based on meta-modelling techniques [3, 5, 6].Therefore developing an accurate meta-model is imperative for a reliable impact detection algorithm. To achieve a 95% Probability of Detection (PoD) an order of 103 different impact scenarios (locations and energies) were required for generating the meta-model. To speed up the solution, a robust contact algorithm was developed to simulate all possible dynamic non-linear impact scenarios. MPI libraries were used to handle and process the large amount of simulated data in an efficient manner. Meta-models once generated were implemented on a small Laptop/PC to provide a rapid solution. For a fuselage panel, impacts can occur both from inside (e.g. tool drop during maintenance) and from outside (e.g. debris) which dues to the curvature of the panel can result in completely different impact dynamics. A novel impact categorization approach has been implemented to identify the type of impact (inside or outside) and thus estimate the location and level of contact force using the dedicated developed meta-model which resulted in higher probability of detection, POD.
• Sensor modelling: The proposed SMART sensor models incorporates the realistic effect of sensor on curved composite. The elements are capable of both sensing and actuating which has been validated against experimental results.
• Damage detection and characterisation: Damage detection algorithms suitable for large scale sensorised curved panels were developed based on probability based diagnostic method using different feature extraction techniques and damage indices of guided waves. SMART FE developed and validated for flat composite stiffened panels [3, 4] was extended to simulate sensing and actuating of Guided Ultrasonic Waves in curved panels. In complex curved panels, when PZT is used as an actuator, multiple modes of wave exist and their dispersion behaviour is complicated. Dispersion and attenuation of all active modes (different in curved panels) was included in designing appropriate actuation signal and data post-processing (WP2.1). Large scale FE analysis will be achieved through parallel dynamic simulations on High Performance Computing (HPC) units.
• Electro-Mechanical Impedance (EMI): EMI has proved highly effective for sensor self-diagnostics and detection of damage in the close vicinity of the transducer such as the foot or under the stringers [3, 7].The proposed EMI SMART model was used to directly measure the impedance response of the curved panel under dynamic load. The effectiveness of EMI method in terms of self-diagnostic, detectability range and effective frequency sweep range for curved composite panels were demonstrated for the first time.
• Optimisation of sensor layup: Optimisation was complex due to different characteristic of curved panel; that is the difference in sensor signals obtained by impacts from inside (tool drop) and outside (debris) of the panel is considered. Standard application of optimisation methods using exhaustive search algorithms for finding the optimum sensor layup of a 5m x 1.7m curved panel would require unrealistic number of simulations. The proposed parallel HPC multi-objective optimisation method based on Genetic Algorithm will significantly reduce the number of simulations and hence the computational cost.
An innovative optimisation approach based on maximum area coverage [9] was applied to the SCOPE panel. This method which is a physically based approach is shown to provide an effective optimisation strategy by minimising blind spots in the panel.
Considering all the above mentioned innovations, SCOPE platform offers the following specifics:
• Passive sensing through meta-models capable of detecting impact force and location;
• Active sensing through methodologies based on SMART FE, capable of detecting and characterising different types of impact damage;
• Self-diagnostic properties through “SMART sensors” models by utilising their Electro-Mechanical Impedance (EMI) properties;
• Optimal sensor position through a multi-objective optimisation algorithm based on the developed meta-models for impact and damage detection algorithms.
Finally, the developed methodologies were validated against in-house background experimental measurements through a building block approach (coupon, mono stringer and curved fuselage panel). Based on the findings of SCOPE, recommendations for test procedures are presented.
In this final deliverable a summary of the methodologies developed and reported throughout the project in different deliverables will be presented together with some results. Finally, the validation and verification of the developed methodologies against numerical and experimental results will be reported for the first time.
Project Results:
2 Methodologies for SCOPE platform
The main SHM work carried out in SCOPE was used to develop a platform through a Graphic User Interface (GUI) for impact and damage detection (BVID) of a curved fuselage panel, see Figure 1.
Figure 1: SCOPE SHM platform
The platform has three main capabilities:
• Active sensing through Ultrasonic Guided Wave excitation and interrogation resulting in damage detection and characterisation
• Self-diagnostics through EMI measurements to increase the reliability and robustness of the SHM system by detecting faulty sensors
• Passive sensing through developments of meta-models resulting detecting, locating and categorizing the type of impact event
The functionality of each option is presented in more detail below.
2.1 Self-diagnostic SHM system
The aim of the self-diagnostic capabilities for an SHM system is to reduce the probability of false alarms or missed-detection, which can be catastrophic, due to faulty sensors. EMI measures are sensitive to both faults within the sensor (damaged sensors/ bond quality) and damage in the vicinity of the transducer. The EMI is a complex spectrum which spreads over a frequency range. For self-diagnostic applications, previous studies have shown that the imaginary party of this spectrum is capable of identifying faults related to the transducer. Presence of damage can either shift the resonance frequency of the structure or reduce the amplitude or a combination of both depending on the type and severity of the fault. Therefore the self-diagnostic module in SCOPE platform results in three different damage metrics based on comparison of EMI measures with baseline values: RMSD, MAPD and CCD (each representing different change in the overall dynamics of the structure) which can result in damage detection and characterisation. The severity of damage is also shown to be directly correlated to the proposed damage indices. A threshold value must be set for each of the presented damage indices in order to detect damage with high reliability.
Figure 2: Self-diagnostic module - SCOPE platform
2.2 Impact detection and characterisation
The challenge of developing a passive SHM sensing capable of detecting an impact event (location and force magnitude) in a curved composite panel is considerable difference in the dynamical response of the structure to impact occurring from inside and outside the structure, see Figure 3 for the difference between the contact force produces by impacts of the same energy and location but from opposite side.
(b) POD for impact location, outer panel
(a) Contact force (c) POD for impact location, inner panel
Figure 3: Examples of impact force caused from impacts on inside and outside of the panel
SCOPE platform overcomes this challenge by implementing a robust categorization algorithm to identify the side of an impact prior from sensor readings prior to detecting its location and force magnitude. Different meta-models were created for impacts occurring from inside and outside of the structure which resulted in noticeable reduction of the error. Meta-models were built by developing and training of Artificial Neural Networks (ANNs) which take into consideration non-linarites of the structural response. The input to the passive sensing module is recorded sensor data (experimental or numerical) in forms of discrete signals. After categorizing them in terms of the side of the impact, the platform will then perform the appropriate feature extraction technique to generate inputs for impact location and force magnitude detection. Having a set threshold for contact force magnitude, which is directly related to initiation of damage in composite structure, the impact event can then be labelled as alarming if it exceeds the value.
Figure 4: Passive sensing module - SCOPE platform
The required data (in the order of 103 simulations) for development of ANNs were generated following these steps:
• Developing a reliable FE model of the curved panel
• Developing a validated impact simulation together with a robust contact algorithm
• Developing a parametric code which can write different input files to submit analyses of the curved panel to an HPC platform
• Developing a code for post-processing of data (generating discrete signals for each sensor)
To assess the reliability and robustness of the passive sensing module in SCOPE, a statistical analysis was carried considering different impact scenarios (varying impact velocities, position: inside and outside, mass/size of impactor) and various network architecture. This study in particular led to:
• definition of POD for passive sensing (to identify the optimum network architecture)
• assessment of the passive sensing module in the presence of real operational load
• determination of the probability of false alarm
2.3 Damage detection and characterisation
The active sensing module developed for SCOPE panel is focused on damage detection algorithms based on UGWs applicable to curved composite panel. In particular it addresses challenges such as directionality of wave propagation, additional modes (longitudinal, torsional and flexural modes) and geometrical complexities (reflections from stiffener/frames, changing thickness, attenuation, superposition of various modes and reflections, noise). For development and validations of the active sensing modules, numerical simulations of actuation, wave propagation, interaction with damage and sensing have been carried out. As wave propagation is a non-linear dynamic analysis, a SMART FE was developed and implemented to model the actuation and sensing of the Lamb waves. Numerical procedures for appropriate bonding of flat PZT disks to curved surfaces of the panels were investigated and outlined. In addition, the SMART FE algorithm provides parameters of the numerical integration and minimum mesh size for a stable and converged representation of lamb waves and damage detection for different actuation frequencies and input signals.
Figure 5: Active sensing module - SCOPE platform
A total of 273 possible damage positions were included in the model. Damage is modelled as softening of the material in one or several layers. This type of damage, characterised through analyses on coupon level using the Hashin model, is the best representation of impact damage, which is the type of defect which has to be detected in the current project. Different sizes and severities of damage will change the damage index in the detection algorithm which results in damage characterization.
The developed damage detection algorithm is based on probabilistic imaging technique where the probabilistic damage indices measured from each transducer path is fused together for the whole structure and presented as a probability plot, with the highest value indicating the presence of damage, as shown in Figure 5.
The capabilities of the developed active sensing module were tested with statistical confidence against the following variables:
• Noise
• Type of bonding
• Failure of one or more sensors
• Sensitivity to sensor placement
For example the sensitivity of the damage detection algorithm to sensor failure was carried out for three various positions of damage: damage in the bay, damage close to the stringer and damage under the stringer. The results presented in Figure 6 illustrates the bar plots of error in damage detection due to sensor failure. It is concluded that damage under the stringer has the highest sensitivity to sensor failure. This information can be used to tailor the optimized sensor positions and redundancy of the developed network separately for various parts of the structure.
(a) damage in a bay (b) damage close to the stringer (c) damage under the stringer
Figure 6: Effect of sensor failure
The results of the statistical analysis were then used to in flight test procedures (deliverable D8).
2.4 Optimisation
A key aspect of any SHM system is the optimum number and placement of sensors to reach a certain level of reliability and POD. In SCOPE methodologies for effective positioning of sensors and actuators for both passive and active systems have been developed. A brief description of each optimisation methodology is outlined in the following sections.
2.4.1 Optimal sensor placement for impact detection and identification
The proposed optimization algorithm to determine the best sensor position is based on Genetic Algorithm (GA) by minimizing a fitness function. The fitness function for passive sensing is the performance of the developed meta-models for impact detection and identification (to minimize the error corresponding to 90% POD). The input to the GA in the first step is the minimum number of sensors for which the optimisation will find the best locations out of all the possible sensor locations. If the POD for impact detection related to that sensor combination is not satisfactory the number of sensors will be increased and the optimisation will repeat again. It is worth mentioning that the developed optimisation algorithm is tailored for the curved composite panel analysed in SCOPE which is not-symmetric due to directionality of the wave propagation velocity, change in thickness across various bays and different cross-sectional profiles. Therefore the symmetry of the problem cannot be used to simplify the optimisations procedure. Figure 7 illustrates the best 8 and 4 sensor locations out of 39 possible ones for a section of the fuselage panel; whereas Figure 8 represents the optimum sensor location within one bay only where the sensors are places symmetrically following the symmetry of the part of the structure.
Figure 7: Best senor positions for an investigation using 8 and 4 sensors in seven bays.
(a) Best 4 sensors placement in the central bay (b) Example of GA results
Figure 8: Example of GA results.
2.4.2 Optimal transducer number and location for damage detection and characterisation
A novel approach for optimal sensor positioning based on geometrical considerations has been adopted in SCOPE by adopting the methodology previously developed at ICL for a curved stiffened panel [9]. This approach is more general and always valid in the sense that it does not depend on the evaluation function which varies for each damage detection methodology. As long as the detection method is based on the time of arrival of damage reflected wave the proposed optimisation is effective. The methodology is based on maximum coverage area while minimizing the boundary reflections in the structure which can cause inaccuracies in the result. The fitness function to be maximized is defined as the sum of coverage index defined from each sensor-actuator path for every point in the structures. Therefore the result of the optimization will be an image illustrating the coverage area of the structures for a given number of transducers, see Figure 9.
4 sensors at the corners. 4 sensors at mid-edge position
Figure 9: Example of coverage area related to different transducer location
Once the fitness function to be optimised is defined, the next step is to develop an optimisation algorithm. For the purpose of SCOPE, GA has been applied to reduce the computational size of the problem. Parameters such as population size, percentage of elite genes, mutation and cross over was carefully chosen to result in a converged solution. An example of the best sensor placements for 6 transducers is illustrated in Figure 10. If the POD related to the optimal 6 transducers not acceptable, then the number of transducers will be increased and the optimisation procedure repeated.
Figure 10: Coverage and positioning for 6 sensors
3 Validation and calibration
In this section the validation and calibration of the developed methodologies against published results and in house experiments are reported. Sensor signals are calibrated with coupon test results. Calibrations of the contact force for impact detection and of Lamb-waves for damage detection have been performed and reported. A multi-scale approach has been followed for both passive and active sensing.
The performance of the SHM system with optimal sensor network has been compared to a random sensor network for the assessment of the optimisation technique for damage detection.
3.1 Validation for passive sensing
For passive sensing, the validation and calibration of signals and developed methodologies follows a building block approach shown in Figure 11.
Figure 11: Building block approach for validation of passive sensing.
The validation consists of the following steps:
• Experimental impacts were conducted at coupon level using a drop tower
• Contact force and PZT signals were acquired and compared to numerical results
• A multilevel approach (numerical and experimental) was conducted on coupon up to curved fuselage panel
The validation against published results was performed with respect to the papers listed in the reference section.
Figure 12: Drop tower used for the validation of passive sensing.
The experimental contact force at different impact energies was compared to numerical simulations, for impacts on both inner and outer surface of the panel. As it is possible to see from Figure 13, the agreement between numerical and experimental results is very good. Moreover the experiments confirm that the major complication arising when curved panels are involved is due to the different behaviour of the resulting contact force when the same impact occurs on the different sides of the panel.
Figure 13: Validation of contact force.
The response of the PZT was as well validated as presented in Figure 14. The time of arrival of the waves are very similar and also the magnitude of the first peaks, which are the parameters used for training the ANN, are in good agreement which suffices for the purpose of the develped activities in SCOPE. A frequency force reconstruction methodology, already presented in [1], was tested on the curved cupon panel and showed excellent resuts, see Figure 15.
Figure 14: Validation of the PZT signal.
Figure 15: Experimental force reconstruction.
3.2 Validation for active sensing
Validation of the active sensing methodologies developed for SCOPE followed a multi-level approach as well, Figure 16. Both Lamb-wave investigations and EMI methods were validated, in terms of signals (actuation, propagation and sensing) and damage detection capabilities.
Figure 16: Validation approach for active sensing.
In summary the validation followed the following steps:
• Experimental investigations were conducted at coupon level
• Numerical and experimental signals were compared and showed good agreement
• The damage detection methodology was able to detect BVID
• A multistage approach (numerical and experimental) was conducted
o Flat coupon
o Curved coupon
o Stiffened element
o SCOPE panel
3.2.1 EMI Technique
Self-diagnostics was performed at coupon level under different sensor failure conditions. Damage consisted of 2 different crack sizes on a PZT transducer. From the signals presented in Figure 18 it is evident that a difference in the spectrum of the signal is present due to damage. In order to quantify this difference, the damage metrics were evaluated and the results showed that the methodology is able to detect the damage in the sensor as well as its severity [2].
Different levels of debonding between transducers and plate were also tested. Three levels of debonding 25%, 50% and 75% were considered, Figure 20. Again results showed that the EMI approach is able to determine the extension of damage, showing a damage index which increases with the severity of the debonded area, Figure 21 [3].
Figure 17: Experimental setup for the EMI investigation.
Figure 18: Self-diagnostics via EMI for a cracked PZT.
Figure 19: Crack evaluation via RMSD.
Figure 20: Different debonding cases.
Figure 21: Debonding evaluation via EMI.
The capabilities of the EMI investigation to detect damage in the structure were evaluated with two different experiments, one at coupon level and one at sub-component level. In both cases the damage was a BVID created by impacting the plate with a drop tower. Figure 22 [3] and 23 show that the methodology is robust under different scales of investigations as the damage position can be identified in both cases by showing higher values of the DI for the transducers closer to the damage site. Moreover, different damage metrics introduced (RMSD, MAPD, CCD) is capable of giving information on type of damage as well (whether there is a shift in resonance frequency or amplitude difference).
Figure 22: Damage evaluation at coupon level via EMI.
Figure 23: Damage detection via EMI at sub-component level [14].
3.2.2 LAMB WAVES
The validation of Lamb wave signals was performed using the set up shown in Figure 24. Again the building block approach was used to go from coupon to sub-component level. The first step was to validate the developed SMART FE in terms of sensing and actuating in comparison to experimental signals. As it is presented in Figure 25, there is good agreement between the FE and experimental signals.
Figure 24: Experimental setup for Lamb wave investigations.
Figure 25: Validation of Lamb-wave signals.
Figure 26: Curved coupon panel under analysis.
The coupon panel shown in Figure 26 was impacted to produce BVID. The result of the impact was checked via C-scan and the presence of damage was evident, Figure 27a). The damage detection algorithm was tested on this example (both with 4 and 8 transducers) and the results obtained are in excellent agreement with the actual size and location of damage, see Figure 27b) [4].
Figure 27: a) C-scan of the plate. b) Damage detection evaluation with 8 transducers.
In the next step, the active sensing methodology was validated against tests for stiffened coupon panel. The details of this study is presented in [5]. Some of the results are highlighted in Figure 28 which shows the validation of the methodology for a more complex coupon.
Figure 28: a) Experimental panel. b) debonding detection.
Validation of damage detection was performed at sub-component level on a curved stiffened composite panel, for a BVID under the stringer. Also in this case damage was properly evaluated and located.
Figure 29: Sub-component specimen with PZTs installed on its inner surface [15].
Figure 30: Damage evaluation at sub-component level.
4 Optimisation
To validate the optimal sensor placement for damage detection, two numerical examples were carried out: One with the optimal sensor placement and the second one with random sensor placement. The aim of the optimisation in this case was to find the best sensor placements for the minimum required sensors (8 transducers in this case) which could detect damage in the correct bay in a large (~1.7m x 0.6m) section of a curved fuselage panel. The error in the damage detection with the optimal sensor network resulted to be smaller than the random sensor placements. More importantly, in both cases even though the error in locating the damage was high but it is worth mentioning that this results was obtained with to 8 transducers covering a large area and more importantly the optimal sensor network was able to detect damage in the correct bay. This optimization approach can be used in a multi-level manner to result in damage detection and identification in various levels to reduce the interrogation cost of the SHM system. For example, level 1 detection will utilise minimum number of sensors to detect damage and locate the bay in which the damage is present. Level 2 interrogation will consist of a more local and detailed analysis adopting a denser transducer network. The proposed optimisation algorithm is able to adapt to the multi-level requirements.
(a) Coverage area of the optimal 8 sensors (b) Optimal positioning for 8 sensors (c) Comparison between optimal and random placement for damage detection
Figure 31: Validation of Optimal sensor placement - Active sensing
5 Conclusions
An SHM platform was developed during SCOPE for impact and damage detection including self-diagnostic capabilities applicable to large curved composite fuselage panel. Self-diagnostic based on EMI technique was successfully shown to detect faulty sensors as well as damage in the vicinity of the transducers. The developed methodologies for passive sensing is based on ANN and is capable of detecting impact location and force magnitude with 90/95% POD. The meta-models were developed by a large library of data generated via FE simulations validated against experimental data. The FE models include material and geometrical non-linearity as well as a robust contact algorithm to allow large number of impact scenarios (varying mass, location, velocity) to be analysed. The analysis was carried out on HPC units and feature extraction applied to the output data. A reliable categorisation technique was proposed to classify impact events from the recorded sensor data and use the appropriate meta-model to characterise the impact event accordingly. This resulted in an increase in performance and reliability of the network.
Following a building block approach, all of the developed methodologies for active sensing were validated for detecting BVID in various locations (i.e. mid-bay, foot of the stringer, under the stringer) and with different severities. The performance of the platform was tested and validated for curved composite panels with complexities such as directional velocity profile, change in thickness, attenuation profile. It was shown that the developed methodologies are applicable for multi-level analysis of complex structures such as the SCOPE panel.
Optimisation procedures for sensor placement for both passive and active sensing were developed based on GA. For passive sensing the objective function was the performance of the ANN for impact detection and identification. The number and location of the sensors were optimised to reach 95% POD. For active sensing a physically base maximum area coverage was implemented for curved panel to ensure maximum detectability with minimum blind areas.
Statistical analysis was carried out for both passive and active sensing to investigate the sensitivity to sensor placement, sensor failure, real load conditions and noise. Guidelines for test procedures were provided. Recommendations for signal acquisition and data processing were provided.
6 References
[1] Thiene, M., M. Ghajari, U. Galvanetto, and M. Aliabadi, Effects of the transfer function evaluation on the impact force reconstruction with application to composite panels. Composite Structures, 2014. 114: p. 1-9.
[2] Sharif Khodaei, Z., Impact Damage Detection in Composite Plates using a Self-diagnostic Electro-Mechanical Impedance based Structural Health Monitoring System. SDHM, 2015.
[3] Sun, J., Structural Health Monitoring on composite panels with high grequency electromechanical impedence (EMI) technique, Imperial College London, final year MSc project, 2014, supervisor: M.H. Aliabadi.
[4] Thiene, M., Z. Sharif Khodaei, and M. Aliabadi, Damage detection in curved composite panels. Under review at SMS, 2015.
[5] Sharif-Khodaei, Z. and M. Aliabadi, Assessment of delay-and-sum algorithms for damage detection in aluminium and composite plates. Smart Materials and Structures, 2014. 23(7): p. 075007.
[6] Sharif Khodaei, Z., Aliabadi, M.H. Lamb-wave based damage detection in anisotropic composite plates (2015) Key Engineering Materials, 627, pp. 1-4.
[7] Meng, S., Sharif Khodaei, Z., Aliabadi, M.H. Localization of barely visible impact damage (BVID) in composite plates (2015) Key Engineering Materials, 627, pp. 217-220.
[8] Sharif Khodaei, Z., Bacarreza, O., Aliabadi, M.H. Lamb-wave based technique for multi-site damage detection (2014) Key Engineering Materials, 577-578, pp. 133-136.
[9] Thiene, M., Z. Sharif Khodaei, and M. Aliabadi, Optimal sensor placement for damage detection based on ultrasonic guided wave damage detection algorithm (2015), Key Engineering Materials, to appear.
[10] Thiene, M., Sharif-Khodaei, Z., Aliabadi, M.H. Comparison of artificial neural networks and the transfer function method for force reconstruction in curved composite plates (2015) Key Engineering Materials, 627, pp. 301-304.
[11] Thiene, M., Galvanetto, U., Ghajari, M., Aliabadi, M.H. A frequency analysis applied to force identification (2013) Research and Applications in Structural Engineering, Mechanics and Computation - Proceedings of the 5th International Conference on Structural Engineering, Mechanics and Computation, SEMC 2013, pp. 2193-2198.
[12] Ghajari, M, Sharif-Khodaei, Z., Aliabadi, M.H. Apicella, A. Identification of impact force for smart composite stiffened panels (2013) Smart Materials and Structures, 22 (8).
[13] Sharif Khodaei, Z., Ghajari, M., Aliabadi, M.H. Determination of impact location on composite stiffened panels (2012) Smart Materials and Structures, 21 (10)
[14] Thiene, M., Salmanpour, S., Sharif Khodaei, Z. and M.H. Aliabadi, Experimental verification of a damage detection methodology in real aeronautical structures based on the proper orthogonal decomposition of the electro-mechanical impedance (2015, to appear.
[15] Salmanpour, S., Sharif Khodaei, Z., Thiene, M. and M.H. Aliabadi, Multi-level damage detection and characterization in a sensorized composite fuselage panel (2015, to appear.
Potential Impact:
The topic that is addressed by the proposal is JTI-CS-2012-3-GRA-01-051 Modelling and simulation of self sensing curved composite panel to predict to predict/control damage evolution in real load condition.
The integration of sensors into composite structures is a structural health monitoring concept to overcome the actual constrains by certification as the demand for short inspection intervals. The innovative elements are smart methodologies for damage detection, location of impact and optimization of sensor layout. Design and sensorised panel and definition of test procedure for monitoring different impact events under real load conditions should have a significant impact in terms of the take up of SHM for aircraft panels.
These tools should assist in defining configurations of sensors and actuators in the design of smart structures that would allow for effective detection of impact events and possible subsequent damage. Structures designed in this way will be capable of immediate detection of impact and monitoring of subsequent behaviour during service to achieve the objectives of low weight configuration. The methodologies would through the optimum sensor layout not only ensure there are no “blind area” form damage detection, they will also minimise the extra weight due to the introduction of sensors.
The currently applied conventional inspection techniques demand an access to the area that has to be inspected. The removal from service to provide this access, leads to costly long down-times for aircrafts during maintenance. The time saving due to SHM is already an advantage, but the approach proposed in SCOPE goes far beyond that. The rapid measurement of the health of the structure allows inspection intervals which are much longer that actually possible.
In summary, SCOPE may enable, in long term, further reduction of overall weight and hence fuel consumption and operative costs by reduction of design allowables. In short to medium term, SCOPE will reduce inspection times (scheduled and unscheduled), hence decreasing the operative cost and time on the ground.
List of Websites:
Professor Ferri M.H.Aliabadi
Head of Department and Professor of Aerostructures
Department of Aeronautics
Imperial College, London
South Kensington Campus
London SW7 2AZ
Tel: +44 (0) 20759 45077
http://www3.imperial.ac.uk/aeronautics
PA: Miss Lisa Kelly
l.kelly@imperial.ac.uk
Tel: 020759 45056
