The RIF Team Analatom, Inc. MERC Resensys Georgia Tech Warner ...

The RIF Team Analatom, Inc. MERC Resensys Georgia Tech Warner ...

Corrosion Sensing, Modeling, Detection, Prediction and Decision Support Technologies George Vachtsevanos Georgia Institute of Technology PHM17 Conference Panel Session on Corrosion Corrosion Technologies: Gaps, Challenges Corrosion Sensing: Need for new on-board sensors monitoring accurately long-term local and global corrosion Corrosion Modeling: Methods to address high fidelity corrosion modeling Corrosion Detection/Prediction Corrosion Mitigation: Emphasis on coating The Integrated Methodology Al Alloy Panels Corrosion Monitoring/ Sensing Sensors Data Mining Pre-Processing Corrosion Modeling Corrosion Detection/Prediction Assessment/ Decision Support Diagnostics Reasoning Global PF Detection Routine: GAG =246 4.5 4 3.5 3 50 100 150 200 250 200 250 Probability of Failure 1 0.5 0 50 -3 x 10 100 150 Type I Error = 5%. Type II Error =1.9577% 4

2 0 2.5 Data Acquisition Feature Extraction Local 3 3.5 4 Fisher Disc riminant Ratio =12.1834 Prognostics 4.5 Action Optimum Aircraft Maintenance The Database Sensing, Temperature, Relative Humidity, Salinity, Mass Loss measurements Images of coupons from submersion test and Lap Joint Chamber tests Images of cracks and pits found in the literature Pictures from field inspection Need for on-platform long-term data 8 Linear Polarization Resistance [] 10 7 10 6 10 LPR1 LPR2 LPR3 LPR4 LPR5 LPR6 LPR7 LPR8 5 10 4 10 3 10 0 50 100 Time [hours] 150 200

Types of Corrosion Micro-structure corrosion Pitting Common denominator in almost all types of corrosion attack May assume different shapes Chlorides (Cl) Inter-granular corrosion Grain-boundaries Stress induced cracking http://www.nace.org/Pitting-Corrosion/ 5 Data Mining (Extracting Useful Information from Raw Data) Original Image Step 1: Apply Threshold to whole Image Step 2: Apply Threshold to local regions 0.25 M as s los s [g ] 0.2 0.15 0.1 0.05 0 0 20 40 60 80 Time [hours] 100 120 140 1 0.8 0.6 1.5hrs 3.5hrs 13hrs 15.5hrs 3 rd P rin c ip a l C o m p o n e n t 0.98 0.97 0.96 0.95 0.94 0.7 0.4

0.6 0.5 0.4 2nd Principal Component -0.4 -0.2 0 0.2 0.4 0.6 1st Principal Compone nt 0.2 0 11 9 17 Data set 25 10 9 f a u lt le v e l ( m a s s ) Feature value PCA on Images from one panel Overlap Region Bulk Regio n Coated Re gion 0.99 8 7 6 5 4 3 2 1 0 0 5 10 15 20 25 time (sec ) 30 35 40 45 Image Processing: Profile of Pit using Laser Confocal Microscope Surface Plot:

2D Image: m m Corrosion Detection and Prediction 0 .8 0 .6 0.7 0 .2 0.6 F e a tu re V a lu e ) 0 .4 0 - 0 .4 0 0.5 1 1.5 2 2.5 x 10 4 Sensor Data Feature 1 0.5 0.4 0.3 0.2 Int erpolation of feat ure value wit h noise Int erpolation of feat ure vlaue snapshot with ground trut h data 0.1 0 0 100 200 300 400 500 600 700 800 900 1000

Time (min) Features & performance Fault Detection xd (t 1) f b xd (t ) n (t ) xc (t 1) f t ( xd (t ), xc (t ), (t )) Features(t ) h ( x (t ), x (t ), v(t )) t d c Operating conditions and inputs PF Detection Routine: GAG =246 4.5 4 3.5 3 50 100 150 200 250 200 250 Probability of Failure 1 Diagnostic Model 0.5 0 50 -3 x 10 100 150 Type I Error = 5%. Type II Error = 1.9577% 4 2 0 2.5 3 3.5 4 Fisher Discriminant Ratio =12.1834 4.5 Failure Prognosis L(t 1) f (t , u, L(t )) Prognostic Model

Fault Level Estimate 20 noisy fault estimate PF-based estimate actual fault 15 10 5 0 -5 0 11 5 10 15 20 25 time (sec) 30 35 40 9 Confidence in Fault Presence 1 8 0.8 0.6 0.4 7 6 5 4 3 0.2 0 0 10 45 fault level (mass) - 0 .6 Feature Extraction Preprocessing - 0 .2 fault level (mass) Sensor Data

Pre-processing Operating conditions and input stresses Diagnostic Model Features & performance Feature Extraction Fault diagnosis Prognostic Model Failure Prognosis detectionconfidence 2 1 5 10 15 20 25 time (sec) 30 35 40 45 0 0 5 10 15 20 25 time (sec) 30 35 40 45 Assessment/Decision Support/The Dynamic Case Based Reasoning Paradigm Assessing the severity of the corrosion state Severity Index-> determines the critical state of the aircraft/estimated from current corrosion state and prognostic information Exploits smart reasoning tools/methods Provides accurate and verifiable maintenance advisories A smart reasoning paradigm: The Dynamic Case Based Reasoning (DCBR) Architecture for Data Storage, Adaptation and Learning Dynamic Case Based reasoning-The smart knowledge base The Q-Learning Paradigm Expected reward: cost-to-go function Immediate reward Learning rate Discount

factor Q( s, a) Q( s, a) [r max Q( s' , a' ) Q( s, a)] a' Current state Current action Next state Next action From the Laboratory Environment to OnBoard The Aircraft-Moving Forward Sensor Sensing/ Data Acquisition Feature Extraction Corrosion Detection/ Prediction Stress Crack Corrosion Models spalling Hazard zone Spall/pit Prognosis for corrosion Prognosis for spalling 0. 4 0. 4 0. 25 0. 2 0. 2 0. 15 0. 15 0. 1 0. 1 0. 05 50 10 0 15 0 200 25 0 3 00 35 0 400 Spalling detected 0 5 0 -5

0 0. 3 0. 25 0. 2 0 .15 0 .05 0 0. 35 0. 3 0. 3 0 .25 10 0. 4 0. 35 0 .35 Corrosion detected noisy fault estimate PF-based estimate actual fault 15 Corrosion Hazard zone 0. 1 50 10 0 150 200 250 300 350 0. 05 0 40 0 Crack detected 50 100 150 2 0 2 5 0 3 0 350 400

time 1 0.8 0.6 0.4 Current Signal Assessment Fault Level Estimate 20 crack Prognosis for crack fault level (mass) crack Hazard zone corrosion ctionconfidence Fault dimensions for corrosion, spalling and cracking Corrosion parameter estimation 5 10 15 20 25 time (sec) 30 Confidence in Fault Presence 35 40 45 Reasoning: DCBR Severity Index A/C Maintenance From the Laboratory Environment to OnBoard The Aircraft-Moving Forward Damage detection provides a trigger for maintenance action Knowledge of damage and corrosion tolerance of the structure in adverse environments Parametric modeling allows for inclusion of stress crack profiles Moving forward: Accurate sensing; better corrosion assessment strategies; effective mitigation methods Living with corrosion?

Recently Viewed Presentations

  • Topics

    Topics

    Meter issue flag is set to 'N' in CAMS. ... If ISO determines real-time telemetry data is of good quality we will change the meter issue flag to 'N' effective the following day. ... or any positive load values will...
  • Elevating Service - Cultivating Leadership Strategic Plan 2013

    Elevating Service - Cultivating Leadership Strategic Plan 2013

    "We need to be more responsive and customize solutions to meet each agency's needs." ... recruitment, investigations and training needs." ... Created a worksheet for how to complete DHS situational assessments.
  • Deputy Assistant Director, Information Operations (DAD IO/J-6)

    Deputy Assistant Director, Information Operations (DAD IO/J-6)

    Deputy Assistant Director, Information Operations (DAD IO/J-6) Chief, Solution Delivery Division . COL Rich Wilson serves as the Division Chief for the Solution Delivery Division (SDD). SDD is responsible for the portfolio of enterprise applications, data solutions and clinical training...
  • Thermodynamic Modelling at NPL - Harry Bhadeshia

    Thermodynamic Modelling at NPL - Harry Bhadeshia

    Thermodynamic Modelling Thermodynamic modelling at the National Physical Laboratory Hugh Davies 8 March 2002 Content National Physical Laboratory NPL Materials Centre Materials Processing Team Thermodynamics and Process Modelling Group MTDATA and the calculation of phase equilibria Models, Materials and Databases...
  • gate.ac.uk

    gate.ac.uk

    Teamware: A Collaborative, Web-based Annotation Environment Kalina Bontcheva, Milan Agatonovic University of Sheffield
  • Unit 2 & 3 - WWW.MOES-PLACE.WEEBLY.COM

    Unit 2 & 3 - WWW.MOES-PLACE.WEEBLY.COM

    Unit 2 & 3. The Rise of Europe, the Middle Ages, and the Byzantine Empire. ... Manors included fields, a mill, a church, peasant huts, and the lord's manor house. Most serfs never traveled farther than a few miles away...
  • Diapositiva 1 - WordPress.com

    Diapositiva 1 - WordPress.com

    Secondo Bruner, accanto ad un dispositivo su base genetica che favorisce l'apprendimento del . linguaggio, è necessario postulare l 'intervento di un sistema di supporto per l'acquisizione . di questa abilità, il . LASS (Language. Acquisition. Support. System), che corrisponde...
  • The English Colonies - Kentucky Department of Education

    The English Colonies - Kentucky Department of Education

    1.Explain what this document tells you about life in the English Colonies. What is one research question that this document could help you answer? Explain how this document could help answer that question. 2. Paraphrase the most important information in...