BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:"DATA1#2 - CANCELLED - Multi-Source Information Fusion Fault Diagn
 osis for Rotating Machinery using Signal and Data Processing"
DTSTART;VALUE=DATE-TIME:20230712T122000Z
DTEND;VALUE=DATE-TIME:20230712T124000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-656@events.isae-supaero.fr
DESCRIPTION:Speakers: Makrouf Iman ()\nMachine fault diagnosis is crucial 
 in industrial systems to enhance reliability\, lifetime\, and service avai
 lability. Intelligent fault diagnosis (IFD) using artificial intelligence 
 (AI) techniques has emerged as a promising approach for automating machine
  health assessment and reducing labor costs. One approach to improve fault
  diagnosis accuracy\, which has become a highly relevant research topic\, 
 is to use multi-data fusion\, which combines information from multiple sou
 rces to make a more informed decision. However\, there is a lack of resear
 ch focusing on the detection of combined machinery faults from multiple se
 nsors. Indeed\, when combined (and emerging) faults happen in different pa
 rts of the rotating machines their features are deeply dependent and the s
 eparation of characteristics becomes complex\, while multi-sensor informat
 ion can provide more comprehensive fault features to deal with the diagnos
 is and identification of multiple combined faults. This paper presents a c
 omprehensive methodology for diagnosing combined faults using data fusion 
 and machine learning techniques. The proposed approach leverages multiple 
 types of sensor data\, including vibration\, current\, temperature\, and a
 coustic data\, sensors to provide a comprehensive picture of the machine's
  health. Our proposed methodology incorporates an ensemble learning approa
 ch and time-domain features to improve diagnostic accuracy. The proposed a
 pproach is tested on a publicly available dataset of rotating machinery wi
 th multiple faults. The results indicate that the method is viable and ach
 ieves good accuracy and efficiency.  Keywords€” Machinery Fault diagno
 sis\, Combined fault diagnosis\, multi-sensor data fusion\, Data and Signa
 l processing\, Broken rotor bar\, Bearing fault\n\nhttps://events.isae.fr/
 event/22/contributions/656/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/656/
END:VEVENT
BEGIN:VEVENT
SUMMARY:"DATA1#6 - Ball bearing diagnosis using a homogenous hybrid databa
 se in a supervised machine learning"
DTSTART;VALUE=DATE-TIME:20230712T134000Z
DTEND;VALUE=DATE-TIME:20230712T140000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-660@events.isae-supaero.fr
DESCRIPTION:Speakers: Sow Souleymane ()\nDigital twins (DT) are often desc
 ribed as a virtual and dynamic representation of a system. They guarantee 
 interaction between physical and virtual spaces. In the context of mainten
 ance 4.0\, the lack of historical data can be caused by an impossible inst
 rumentation for complex systems. To face it\, DT offers the possibility to
  simulate several operating modes which can serve for a diagnostic. This o
 peration can be made by using machine learning algorithm (MLA) through a d
 iagnosis by classification. But the challenge is to identify the best use 
 of both data historical and simulated on a hybridisation database to make 
 the most reliable diagnosis. In this paper\, a digital twin combining a di
 screte element model (DEM) and a finite element model (FEM) is developed t
 o generate data with an outer race default signature. These generated data
  with five sizes of defaults are also measured on the test bench. Accordin
 g to a percentage\, historical data are used to build the homogenous hybri
 d database. Two MLAs (Support Vector Machines and K-Nearest Neighbours) ar
 e used to perform a classification by training the homogenous hybrid datab
 ase and the test is realised by using the rest of historical data. The res
 ults of this approach show a better reliability than existing methods on t
 he tested datasets also it's allowed to evaluate the contribution of histo
 rical data in homogenous hybridisation process.\n\nhttps://events.isae.fr/
 event/22/contributions/660/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/660/
END:VEVENT
BEGIN:VEVENT
SUMMARY:"DATA1#5 - Airplane turbulence detection with hybrid deep learning
  model"
DTSTART;VALUE=DATE-TIME:20230712T132000Z
DTEND;VALUE=DATE-TIME:20230712T134000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-659@events.isae-supaero.fr
DESCRIPTION:Speakers: Dampeyrou Charles ()\nAtmospheric turbulence has a s
 ignificant impact on airplane motions and can induce excessive stress and 
 fatigue damage. The identification of turbulence in aircraft service phase
  is of particular interest to estimate actual structural fatigue or to def
 ine an improved maintenance plan. This article describes a new model to de
 tect turbulence from in-board instrumentation solely. Equations expressing
  the relationship between the load factors in the 3 directions and the for
 ces applied to the airplane are derived from a rigid body 6 degrees of fre
 edom model. Lift\, drag and lateral force coefficients\, required to compu
 te the aerodynamic forces\, are predicted by multi layer perceptrons. The 
 architecture of the model makes it possible to train the multi layer perce
 ptrons in an unsupervised way with common deep learning techniques\, using
  only sensors commonly present on airplanes. After the training phase\, th
 e model is able to predict the lift\, drag and lateral force coefficients 
 for various configurations. The error between the measured load factors an
 d the load factors predicted by the model is used to identify the presence
  of turbulence. The performances of the model to predict the lift and drag
  coefficient is first evaluated on simulated data. The turbulence detectio
 n is then evaluated on a dataset composed of hundreds of commercial flight
 s as well as on simulated data.\n\nhttps://events.isae.fr/event/22/contrib
 utions/659/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/659/
END:VEVENT
BEGIN:VEVENT
SUMMARY:"DATA1#4 - Anomaly Detection in Aircraft Engine Vibration Using De
 ep Convolutional Autoencoder"
DTSTART;VALUE=DATE-TIME:20230712T130000Z
DTEND;VALUE=DATE-TIME:20230712T132000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-658@events.isae-supaero.fr
DESCRIPTION:Speakers: El Hidali Abdallah ()\nThe useful life of aircraft e
 ngines depends on their operating environment (polluted areas\, harsh clim
 ate\, etc.). Detecting signs of degradation and aging can be difficult due
  to background noise measured on vibrational signals. Statistical methods 
 such as threshold-based monitoring may not be reliable enough. This paper 
 presents a promising method based on learning normal behavior on a populat
 ion of engines considered to be healthy\, such as newly produced engines. 
 The learning is done by calculating spectrograms of the vibrational signal
 s\, normalizing them and treating them as images\, then using a convolutio
 nal autoencoder to learn normal behavior. This model can be used during sh
 op visits to detect early degradation by comparing vibrational signals of 
 in-use engines to the learned standard. Keywords: vibrational signals\, ba
 ckground noise\, normal behavior\, convolutional autoencoder.\n\nhttps://e
 vents.isae.fr/event/22/contributions/658/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/658/
END:VEVENT
BEGIN:VEVENT
SUMMARY:"DATA1#3 - Wind turbine drivetrain fault detection using physics-i
 nformed multivariate deep learning"
DTSTART;VALUE=DATE-TIME:20230712T124000Z
DTEND;VALUE=DATE-TIME:20230712T130000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-657@events.isae-supaero.fr
DESCRIPTION:Speakers: Jamil Faras ()\nVibration analysis is a prevalent te
 chnique in the predictive maintenance of wind turbines. It is an effective
  method for early fault detection and enables the creation of cost-effecti
 ve maintenance strategies. Commonly used vibration analysis methods in the
  literature rely on signal processing techniques such as time and frequenc
 y domain approaches. However\, the signal processing techniques require ma
 nual interpretation by domain experts. It is important to note that differ
 ent indicators exhibit sensitivity to specific faults. Manual analysis of 
 indicators can be avoided by fusing them to derive high-level wind turbine
  health status. It enables the learning of complex non-linear relationship
 s among the indicators. This research focuses on a multivariate deep learn
 ing model\, i.e.\, autoencoder\, which fuses different signal processing i
 ndicators to provide a single high-level health status. The proposed model
  is a normal behaviour model that learns the indicator's normal behaviour 
 and labels faults if it observes deviation from the normal behaviour. The 
 proposed fusion method of indicators is robust compared to individual indi
 cator models as it learns complex non-linear relationships among indicator
 s. The proposed method is tested for fleet-level fault detection both with
  and without fine-tuning for a specific wind turbine. Furthermore\, it dec
 reases the time required for wind farm health prognosis analysis and compu
 tation. Various autoencoder architectures have been compared\, including s
 imple feedforward neural networks\, convolutional neural networks\, and re
 current neural networks. The proposed method is demonstrated using real-li
 fe\, high-frequency condition monitoring data from offshore wind turbines 
 over several years\, including wind turbines observed faults. The method's
  effectiveness and performance were demonstrated through analysis of plane
 tary stage\, generator\, and high-speed stage failure cases.\n\nhttps://ev
 ents.isae.fr/event/22/contributions/657/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/657/
END:VEVENT
BEGIN:VEVENT
SUMMARY:"DATA1#1 - Vibration based milling diagnostics using Artificial In
 telligence"
DTSTART;VALUE=DATE-TIME:20230712T120000Z
DTEND;VALUE=DATE-TIME:20230712T122000Z
DTSTAMP;VALUE=DATE-TIME:20260308T103707Z
UID:indico-contribution-192-655@events.isae-supaero.fr
DESCRIPTION:Speakers: Knittel Dominique ()\nIn industrial machining proces
 ses\, tool failures may result in losses in surface and dimensional accura
 cy of a finished part\, or possible damage to both the work piece and the 
 machine. Consequently\, tool condition monitoring has become essential to 
 achieve high-quality machining as well as cost-effective production. Moreo
 ver\, cutting tool degradation may vary considerably under different opera
 tion conditions and materials behaviour. Therefore real time identificatio
 n of the tool state during machining\, before it reaches its failure stage
 \, is critical. In this study the vibrations of the cutting tool and of th
 e workpiece material are online measured. Features are then calculated in 
 time domain from the raw signals. Transient zones\, when the cutting tool 
 enters or exits the material\, can be considered or not. All the calculate
 d features are normalized and stored in a table. It is then necessary to m
 ake a dimensional reduction of that feature table in order to avoid overfi
 tting and to reduce the computing time of the learning algorithm. In this 
 study\, 54 milling experiments were conducted from which features were cal
 culated and then split into two groups: 80% for the machine learning model
  training\, 20% for the test phase. The first part of this study proposes 
 an analysis on the impact of the features on the robustness of the models\
 , and a second part focuses on a real-time data driven prognostics and hea
 lth management (PHM) approach for tool condition monitoring\, based on sup
 ervised machine learning techniques (i.e. the model training needs labelle
 d data). The fusion of decision coming from several machine learning algor
 ithms (kNN\, decision trees (DT) and random forest (RF)) is then used to p
 redict the tool quality in real time. All parameters and configurations of
  the algorithms are optimized in order to maximize the real time diagnosis
  accuracy. The experimental results show that our proposed approach achiev
 es good accuracy and real time performances in dry milling operations. Res
 ults of our study are implemented in real tool wear diagnosis\, and thus g
 ive new opportunities toward realizing Industry 4.0\n\nhttps://events.isae
 .fr/event/22/contributions/655/
LOCATION:Toulouse
URL:https://events.isae.fr/event/22/contributions/655/
END:VEVENT
END:VCALENDAR