Challenges and opportunities in reliability engineering: the big KID - PowerPoint PPT Presentation

Challenges and opportunities in reliability engineering: the big KID (Knowledge, Information and Data) Enrico Zio Chair on Systems Science and the Energy Challenge – CentraleSupelec, Fondation Electricité de France (EDF), France Energy Department, Politecnico di Milano, Italy Aramis Srl, Italy

2

Problem statement Failures Prevented by Design for Reliability Maintenance Normal Degraded Failure … Time 5

INDUSTRY 7

Industry 1-2-3-4 8

(SMART) Reliability Engineering 10

The Big KID 12

Big Knowledge(ID) 13

Big (K)Information(D) 14

Big (KI)Data 1110101001010001011100100101011000010101001 1101110111011101010010100010111001001010110 0001010100111011101110111010100101000101110 0100101011000010101001110111011101110101001 0100010111001001010110000101110101001010001 0111001001010110000101010011101110111011101 0100101000101110010010101100001010100111011 1011101110101001010001011100100101011000010 15

Can the Big KID become SMART for Reliability Engineering ? Application 16

Problem statement Failures Prevented by Design for Reliability Maintenance Normal Degraded Failure … Time 17

SMART Reliability Engineering – component Big KID opportunities Reliability analysis for Design for Reliability: From failure modeling to degradation-to-failure modeling 18

SMART Reliability Engineering – component Big KID opportunities Reliability analysis for Design for Reliability: From failure modeling to degradation-to-failure modeling Integrating physics-of-failure knowledge in reliability models • Multi-State Physic-Based Models 19

SMART Reliability Engineering – component 20 Challenges KID Reliability ? Model (Knowledge, Information, Data) Statistical models of time to failure Sufficient failure data 1 Stochastic process 0.999 models 0.998 Reliability 0.997 Physics-based 0.996 0.995 models Physics knowledge 0.994 0 20 40 60 80 Year Expert judgment Field data Highly reliable Multi-state models 20

SMART Reliability Engineering – component 21 Multi-State Physic-Based Models Alloy 82/182 dissimilar metal weld of piping in a PWR primary coolant system Physical laws Multi-state physics model of crack development in Alloy 82/182 dissimilar metal weld 21

SMART Reliability Engineering – component 22 Opportunities Random shocks Degradation process Random shock process Dependences in degradation processes Internal leak Initial state Failure state λ 32 λ 21 λ 10 3 2 1 0 22

SMART Reliability Engineering – component 23 Opportunities Maintenance b a Degradation process Preventive maintenance (a) Corrective maintenance (b) 23

SMART Reliability Engineering – component 24 Challenges Uncertainty Uncertain parameters in degradation models Internal leak Initial state Failure state λ 32 λ 21 λ 10 3 2 1 0 24

SMART Reliability Engineering – component 25 Challenges Degradation processes Internal leak Initial state Failure state λ 32 λ 21 λ 10 3 2 1 0 Piecewise-deterministic Markov process (PDMP) 25

SMART Reliability Engineering – component 26 Challenges Finite-volume scheme MC Simulation 26

SMART Reliability Engineering – component Big KID opportunities Reliability analysis for Design for Reliability: From failure modeling to degradation-to-failure modeling Integrating physics-of-failure knowledge in reliability models • Multi-State Physic-Based Models ?And the data? 27

Theory Assumptions about how things Refinements work ADT Procedure Conclusion Design Insight about what works, A blueprint of the procedure gained from analysis Evaluation Experiment Performance Assessment of the outcome of Trial to test hypothesis the experiment Life parameter distribution General testing procedure Time t 0 Threshold t 1 distribution t 2 AM 2 AM AM Performance 1 Time 3 distribution Stress s 0 s 1 s 2 Degradation Model: Acceleration Model: Degradation VS Time Stress VS Time Stochastic process or degradation-path: Physical or empirical models: Arrhenius: 𝑒(𝑇) = 𝐵𝑓 − 𝐹 𝑏 𝑙𝑇 Wiener process: 𝑍(𝑢) = 𝜏𝐶(𝑢) + 𝑒(𝑇)𝑢 31

Degradation           Y t ( ) B t ( ) d S ( ) t y d S , ( ) exp a b ( S ) l l 0 l l Process Model   Data Analysis 2        n m 1  x d S ( ) t  k 1  l l    lij l lij   L ( , , ) a b exp     2 2 t 2   Maximum 2 t    l 1 i 1 j 1   lij lij Likelihood  k n m 1      l       x exp a b ( S ) t 0 lij l    l 1 i 1 j 1 Trend analysis & Accelerability Verification : LSE      ( ) E Y t ( ) d S ( ) t y d S Degradation 0,07 l l 0 l Fitting 0,06   y = 4E-06x + 0,0014    Degradation Percentage d S ( ) exp a b ( S ) LSE l l ˆ R² = 0,9837  ˆ 0,05 a b     ln ( d S ) a b ( S ) l l 0,04 Parameter  k n m 1 1 2   l  ˆ            ˆ 2 ˆ x exp a b ( S ) t Estimation     0,03   lij l n ( m k ) t    l 1 i 1 j 1 0,02 0,01 Parameter estimation: 0 -4000 1000 6000 11000 16000 𝜏 2 𝑏 𝑐 -0,01 Time/Min - 36.961 13.112 8.278e-07 Degradation VS Time -9 Reliability Prediction: 1,8 1,9 2 2,1 -9,5 1 Ln(degradation Rate) y = 13,693x - 37,955 0.95 -10 R² = 0,9494 0.9 -10,5 0.85 0.8 -11 0.75 0.7 -11,5 R(t) 0.65 -12 0.6 0.55 -12,5 0.5 0.45 -13 0.4 Ln(Voltage) 0.35 0.3 Time VS Stress 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 Time/Min 4 x 10 33

Challenges in ADT  The whole trend is defined Degradation  Incomplete knowledge due trend (linear, exponential, etc.) Epistemic to limited information uncertainty  Inherent randomness  Interval, possibility, etc. Aleatory  Probability uncertainty Stochastic Process – some revised models: Challenges:  Traditional methods mainly DBM model Revised model I* Revised model II model degradation trend 𝑍 𝑢 = 𝑍 𝑢 = 𝑒 𝑇 ⋅ 𝑢 + 𝜏 𝐶 𝐶 𝑢 𝑍 𝑢 = 𝑒(𝑇)𝑢 + 𝜏 𝐶 𝐶 𝑢 𝑒(𝑇)𝑢 + 𝜏 𝐶 𝐶 𝑢 and aleatory uncertainty. 𝑒 𝑇 ： a definite value 𝑒(𝑇) ∼ 𝑂 𝜈, 𝜏 2 𝜈, 𝜏 2 𝑒(𝑇) ∼ 𝑂  Failing to consider Degradation trend epistemic uncertainty may Degradation trend Degradation trend Aleatory uncertainty Aleatory uncertainty Aleatory uncertainty cause serious reliability Epistemic uncertainty evaluation problems. 𝝉 𝑪 𝑪 𝒖 & 𝝉 𝑪 𝑪 𝒖 𝒆(𝑻) 𝝉 𝑪 𝑪 𝒖 & 𝒆(𝑻) 34

Problem statement Failures Prevented by Design for Reliability Maintenance Normal Degraded Failure … Time 35

SMART Reliability Engineering – component Big KID opportunities Maintenance: Integrating physics knowledge and data: • Prognostics and Health Management (PHM) 36

Maintenance Maintenance 1950 1980 2000 2016 Corrective Planned Periodic Condition Based Predictive Maintenance Maintenance Maintenance (CBM) Maintenance (PrM) Prognostics and Health Management (PHM) PHM is fostered by advancements in: Sensor Algorithm Computation power 38

PHM for what? PHM in support to CBM and PrM Normal Fault Conditions Detection Abnormal Vibration Conditions No t Maintenance Anomaly of Type 1 … Fault Anomaly of Type 2 Diagnostics Maintenance Temperature Anomaly of Type 3 Equipment Decision Maintenance Maker Decision t Fault Remaining Useful Sensors Prognostics Life (RUL) measurements 40

PHM: why? (Industry)  Increase maintainability, availability, safety, operating performance and productivity  Reduce downtime, number and severity of failure and life-time cost 41

PHM: how? (Fault detection) Signal Real reconstructions measurements MODEL OF 80 80 75 PLANT BEHAVIOR 75 ≠ 70 70 IN NORMAL OPERATION 65 65 0 500 1000 Nominal Range-based 20 20 Physics-based 10 10 Data-Driven (AAKR, PCA, 0 0 0 500 1000 RNN,…) Abnormal Condition 43

PHM: how? (Fault diagnostics) • Signal measurements representative of the fault classes: « x 1 , x 2 ,… x n , class» Norm Node 14 Wavelet Norm Node 5 Peak Value Wavelet C 1 = Inner race x 1 x Empirical 2 C 2 = Balls x 3 Classifier • Empirical classification methods: C 3 = Outer race • Support Vector Machines • K-Nearest Neighbours • Multilayer Perceptron Neural Networks • Supervised clustering algorithms • Ensemble of classifiers • … 44