Failure prognostics in a particle filtering framework Application to a PEMFC stack

Size: px

Start display at page:

Download "Failure prognostics in a particle filtering framework Application to a PEMFC stack"

Raymond Bennett Cole
5 years ago
Views:

1 Failure prognostics in a particle filtering framework Application to a PEMFC stack Marine Jouin Rafael Gouriveau, Daniel Hissel, Noureddine Zerhouni, Marie-Cécile Péra FEMTO-ST Institute, UMR CNRS 6174, Besançon FCLAB Research Federation, FR CNRS 3539, Belfort marine.jouin@femto-st.fr

2 Motivations Fuel Cell : an alternative to traditional energies Several application fields Transportation, µ-cogeneration, Portable devices powering, Aerospace. No mobile parts = good reliability Marine Jouin Journées inter-gdrs 12/06/2014 2

3 Motivations Fuel Cell : an alternative to traditional energies Several application fields Transportation, µ-cogeneration, Portable devices powering, Aerospace. No mobile parts = good reliability Current limitations Efficency Performances Durability Current Necessary Current Necessary 25 30% 35 40% h 8000 h transportation h stationary Major limitation: lifespan still too short Socio economic aspects Cost reduction of PEMFC system Public acceptance Technological bolts Stable hydrogen supply with high purity Hydrogen storage Marine Jouin Journées inter-gdrs 12/06/2014 2

4 Motivations Fuel Cell : an alternative to traditional energies Several application fields Transportation, µ-cogeneration, Portable devices powering, Aerospace. No mobile parts = good reliability Current limitations Efficency Performances Durability Current Necessary Current Necessary 25 30% 35 40% h 8000 h transportation h stationary Major limitation: lifespan still too short Socio economic aspects Cost reduction of PEMFC system Public acceptance Technological bolts Stable hydrogen supply with high purity Hydrogen storage Prognostics and Health Management (PHM) : a solution? Object : taking decision at the right time to optimize system use and avoid failures Marine Jouin Journées inter-gdrs 12/06/2014 2

5 Failure prognostics in a particle filtering framework 1. Backgrounds 2. Feature extraction and aging modeling 3. Prognostics based on particle filters 4. Conclusion Marine Jouin Journées inter-gdrs 12/06/2014 3

6 Failure prognostics in a particle filtering framework 1. Backgrounds - Prognostics and Health Management - Prognostics: a key element - PHM of PEMFC - First work and its limitations 2. Feature extraction and aging modeling 3. Prognostics based on particle filters 4. Conclusion Marine Jouin Journées inter-gdrs 12/06/2014 4

future states of the system, RUL estimates Cause of failure, isolation et identification of the component responsible of the failure System

7 1. Backgrounds Prognostics and Health Management (PHM) Human-machine interface Data coming from sensors or transducers Recommended actions to accomplish the mission (maintenance, command) Signals transformations: extraction / selection / descriptors generation Prediction of the future states of the system, RUL estimates Cause of failure, isolation et identification of the component responsible of the failure System state of health, comparison of descriptors on-line / expected, detection and location of failures Marine Jouin Journées inter-gdrs 12/06/2014 5

8 1. Backgrounds Prognostics: a key element Pronostic RUL estimates (Remaining Useful Life) Norme ISO :2004 : " estimation of time to failure and risk for one or more existing and future failure modes" Main objectives Estimation of the Remaining Useful Life(RUL) 0 tc RUL t fail. Estimation of the probability of failure of the system at a given date state S1 0,5 Taking into account uncertainty is a major issue Uncertainty / system Uncertainty / its use Uncertainty / sensors Uncertainty / prognostic model defined S2 S3 0,3 0,2 prob Marine Jouin Journées inter-gdrs 12/06/2014 6

9 1. Backgrounds Prognostics: a key element Different approaches Prognostic Model-based approaches Data driven approaches Hybrid approaches Analytical models of nonlinear phenomena Need a small quantity of data High computational cost Default / system specific Models hard to develop Transformation of raw data into behavioral models (learning) No degradation model a priori Good ability to catch nonlinearities Require a huge amount of data Benefit from the advantages of both approaches Better model learning Better uncertainty management Can be complex and computationally expensive Marine Jouin Journées inter-gdrs 12/06/2014 7

10 1. Backgrounds Prognostics: a key element Prognostics objective: illustration Degradation level time Marine Jouin Journées inter-gdrs 12/06/2014 8

11 1. Backgrounds Prognostics: a key element Prognostics objective: illustration Degradation level Critical threshold before failure Failure threshold time Marine Jouin Journées inter-gdrs 12/06/2014 8

12 1. Backgrounds Prognostics: a key element Prognostics objective: illustration Degradation level Critical threshold before failure Failure threshold time Starting point of prediction : t p Marine Jouin Journées inter-gdrs 12/06/2014 8

13 1. Backgrounds Prognostics: a key element Prognostics objective: illustration Degradation level RUL pdf Learning Critical threshold before failure Failure threshold RUL time Starting point of prediction : t p Critical threshold reached End of life Marine Jouin Journées inter-gdrs 12/06/2014 8

14 1. Backgrounds Prognostics: a key element RUL Prognostics objective: illustration RUL 1 RUL 2 Degradation level RUL pdf t p1 t p2 time Learning Critical threshold before failure Failure threshold RUL time Starting point of prediction : t p Critical threshold reached End of life Marine Jouin Journées inter-gdrs 12/06/2014 8

15 1. Backgrounds PHM of PEMFC M. Jouin, R. Gouriveau, D. Hissel, M-C. Péra, and N. Zerhouni, Prognostics and health management of PEMFC state of the art and remaining challenges, International Journal of Hydrogen Energy, vol. 38, no. 35, , 2013 Marine Jouin Journées inter-gdrs 12/06/2014 9

16 1. Backgrounds PHM of PEMFC: challenges pointed out L7 Human-Machine Interface Decide Fault tolerant, self-adaptative and reconfigurable control system L6 Decision Support Verification and validation procedures L5 L4 Prognostics Diagnostics Model / Analyze Extended framework for detection and diagnostics approaches Advanced prognostics models L3 Condition Assessment L2 L1 Data processing Data Acquisition Observe Reliable, non-intrusive, Événement non-damaging - date observation techniques Easily implementable technology (cost, volume, online, etc.) Complex system Degradation, losses and behavior [1] M. Jouin, R. Gouriveau, D. Hissel, M-C. Péra, and N. Zerhouni, Prognostics and health management of PEMFC state of the art and remaining challenges, International Journal of Hydrogen Energy, vol. 38, no. 35, , 2013 Marine Jouin Journées inter-gdrs 12/06/

1. Backgrounds PHM of PEMFC: challenges pointed out L7 Human-Machine Interface Decide Fault tolerant, self-adaptative and reconfigurable control system L6 Decision Support Verification and validation

17 1. Backgrounds PHM of PEMFC: challenges pointed out L7 Human-Machine Interface Decide Fault tolerant, self-adaptative and reconfigurable control system L6 Decision Support Verification and validation procedures L5 L4 Prognostics Diagnostics Model / Analyze Extended framework for detection and diagnostics approaches Advanced prognostics models L3 Condition Assessment L2 L1 Data processing Data Acquisition Observe Reliable, non-intrusive, Événement non-damaging - date observation techniques Easily implementable technology (cost, volume, online, etc.) Complex system Degradation, losses and behavior [1] M. Jouin, R. Gouriveau, D. Hissel, M-C. Péra, and N. Zerhouni, Prognostics and health management of PEMFC state of the art and remaining challenges, International Journal of Hydrogen Energy, vol. 38, no. 35, , 2013 Marine Jouin Journées inter-gdrs 12/06/

18 1. Backgrounds First work and its limitations Prognostics based on particle filters with simple empirical 1000 Predicted RUL comparison FC 2 aging models to predict the voltage degradation Linear model Exponential model Logarithmic model Real RUL 90% confidence interval 3 models tested 1. Linear 2. Exponential 3. Linear + logarithmic RUL Promising results with the 3 rd but too much uncertainty on the results Time in hours Main limit = disturbances induced by characterizations not taken into account [2] M. Jouin, R. Gouriveau, D. Hissel, M-C. Péra, and N. Zerhouni, Prognostics of PEM fuel cell in a particle filtering framework, International Journal of Hydrogen Energy, vol. 39, no. 1, pp , 2014 Marine Jouin Journées inter-gdrs 12/06/

19 1. Backgrounds First work and its limitations Prognostics based on particle filters with simple empirical 1000 Predicted RUL comparison FC 2 aging models to predict the voltage degradation Linear model Exponential model Logarithmic model Real RUL 90% confidence interval 3 models tested 1. Linear 2. Exponential 3. Linear + logarithmic RUL Promising results with the 3 rd but too much uncertainty on the results Time in hours Main limit = disturbances induced by characterizations not taken into account PROBLEM ADDRESSED HERE [2] M. Jouin, R. Gouriveau, D. Hissel, M-C. Péra, and N. Zerhouni, Prognostics of PEM fuel cell in a particle filtering framework, International Journal of Hydrogen Energy, vol. 39, no. 1, pp , 2014 Marine Jouin Journées inter-gdrs 12/06/

20 Failure prognostics in a particle filtering framework 1. Backgrounds 2. Feature extraction and aging modeling - Principle - Modeling 3. Prognostics based on particle filters 4. Conclusion Marine Jouin Journées inter-gdrs 12/06/

21 2.Feature extraction and aging modeling Principle 1. Observation of the data Different continuous aging parts separated by characterizations Recovery observed after characterizations Same trends of all the continuous aging parts but with an acceleration of the degradation 2. Selection of a model for continuous aging parts Characterizations Global model selected from previous work: P(t) = - a.ln(t) b.t + c Continuous aging parts Marine Jouin Journées inter-gdrs 12/06/

22 2.Feature extraction and aging modeling Principle 3. Identification of coefficients a & b of the model on each part by robust least square fitting ydata ydata xdata xdata Part 1: a 1 & b 1 ydata Part n: Événement a n & b n - date xdata Part 2: a 2 & b 2 Marine Jouin Journées inter-gdrs 12/06/

23 2.Feature extraction and aging modeling Principle 4. Feature extraction from a i & b i i є [1, n] 5. Extraction of the recovery 3.45 fitted curve Vrecup Vrecup temps Marine Jouin Journées inter-gdrs 12/06/

24 2.Feature extraction and aging modeling Modeling Main objective: choosing models that are close to the data but can be justified by phenomena occurring within the stack Global model for power aging: P(t) = - a.ln(t) b.t + c Models built thanks to feature extraction for coefficient a aging: a(t) = a 1.exp(a 2.t) + a 3.exp(a 4.t) for coefficient b aging: b(t) = b 1.exp(b 2.t) + b 3 for recovery aging: R(t) = r 1.exp(r 2.t) + r 3.exp(r 4.t) Marine Jouin Journées inter-gdrs 12/06/

25 Failure prognostics in a particle filtering framework 1. Backgrounds 2. Feature extraction and aging modeling 3. Prognostics based on particle filters - Data available - Development hypotheses - Problem formalization and adaptation - Particle filtering approach - Results 4. Conclusion Marine Jouin Journées inter-gdrs 12/06/

26 3. Prognostics based on particle filters Data available data set: power degradation through time aging tests on a stack of 5 cells, 100cm² FC : 1750h at constant current solicitation of 60 A i FC 0.6A/cm² t = 1750h Marine Jouin Journées inter-gdrs 12/06/

27 3. Prognostics based on particle filters Development hypotheses / FC aging Degradation Irreversible with a long time constant Not measurable directly (simply) deductible from another variable Examples of possible candidates Electrodes active surface area degradation H2 crossover through the membrane Aging observation through power evolution / Functioning Constant current solicitation Constant operating conditions / Study framework Opening applicative limits: model Non-exact (unknown coefficients) Non-stationary (time varying) Non-linear Non Gaussian noise Bayesian tracking Marine Jouin Journées inter-gdrs 12/06/

28 3. Prognostics based on particle filters Problem formalization Formulation Hidden state model Degradation state x,, k f xk 1 k k Observation model Available measurements z h x, k k k Optimal Bayesian solution Initial state distribution p(x 0 z 0 ) p(x 0 ) Obtaining of p(x k z 1:k ) in 2 steps p( x / z ) p( x / x ). p( x / z ). dx k 1: k1 k k1 k1 1: k1 k1 px ( / z ) k 1: k pz ( / x). px ( / z ) k k k 1: k1 pz ( / z ) k 1: k1 Marine Jouin Journées inter-gdrs 12/06/

29 3. Prognostics based on particle filters Problem formalization Problem adaptation Formulation Modeling Hidden state model Degradation state x,, k f xk 1 k k Observation model Available measurements z h x, k k k Aging models developed earlier Voltage and current measurements of the stack Optimal Bayesian solution Initial state distribution p(x 0 z 0 ) p(x 0 ) Obtaining of p(x k z 1:k ) in 2 steps p( x / z ) p( x / x ). p( x / z ). dx k 1: k1 k k1 k1 1: k1 k1 px ( / z ) k 1: k pz ( / x). px ( / z ) k k k 1: k1 pz ( / z ) k 1: k1 Solving : particle filtering Marine Jouin Journées inter-gdrs 12/06/

30 3. Prognostics based on particle filters Particle filtering approach Marine Jouin Journées inter-gdrs 12/06/

31 3. Prognostics based on particle filters Particle filtering approach Principle LEARNING PREDICTION Raw data Feature extraction Filters initialization Prognostics by PF Behavior prediction RUL Filters associations to include characterizations Filter 1: power aging P Filter 2: coefficient a Filter 3: coefficient b Filter 4: recovery R Marine Jouin Journées inter-gdrs 12/06/

32 3. Prognostics based on particle filters Particle filtering approach Filters interactions t = t+1 Threshold for learning or prognostics end not reached Is a characterization scheduled? Yes No Update P with particles from models a, b & R Filter 1 P, a, b, R Filter 2 Filter 3 Filter4 Marine Jouin Journées inter-gdrs 12/06/

33 3. Prognostics based on particle filters Results Behavior prediction (1/2) Learning of 500 hours a 0.5 b t t R P t t Prediction ended too early around 620 h Marine Jouin Journées inter-gdrs 12/06/

34 3. Prognostics based on particle filters Results Behavior prediction (2/2) 3 2 Learning of 1300 hours a b t t 220 R P t t Good prediction Marine Jouin Journées inter-gdrs 12/06/

35 3. Prognostics based on particle filters Results Behavior prediction: discussion MAPE during learning and prediction Marine Jouin Journées inter-gdrs 12/06/

36 3. Prognostics based on particle filters Results Behavior prediction: discussion Feature extraction change with the length of the learning: illustration on recovery and one coefficient of the power model Marine Jouin Journées inter-gdrs 12/06/

37 3. Prognostics based on particle filters Results RUL estimates Marine Jouin Journées inter-gdrs 12/06/

38 Failure prognostics in a particle filtering framework 1. Backgrounds 2. Feature extraction and aging modeling 3. Prognostics based on particle filters 4. Conclusion Marine Jouin Journées inter-gdrs 12/06/

39 4. Conclusion Motivations Challenges FC : technico-socio-economic stakes PHM : reliability / availability / costs thematic of growing interest Towards PHM of PEMFC : a lever to increase life duration Empirical modeling of aging Good way to represent power aging at constant current solicitation Allows integrating recovery induced by characterizations Prognostics results Better prediction of power behavior Less uncertainty in RUL estimates But poor results if the learning is too short Planned expansion Take into account mission profiles / variable conditions by including the current in the models Marine Jouin Journées inter-gdrs 12/06/

40 Failure prognostics in a particle filtering framework Application to a PEMFC stack Marine Jouin Rafael Gouriveau, Daniel Hissel, Noureddine Zerhouni, Marie-Cécile Péra FEMTO-ST Institute, UMR CNRS 6174, Besançon FCLAB Research Federation, FR CNRS 3539, Belfort marine.jouin@femto-st.fr

ANR PROPICE Summer School Diagnostics and Prognostics of Fuel Cell Systems 01-04 July 2014, FCLAB, Belfort, France https://propice.ens2m.fr/ecole-diag-pron-pac.

However, even if this technology is close to being competitive, it is not yet ready to be considered for large scale industrial deployment: FCS still must be optimized, particularly by increasing

Indeed, Proton Exchange Membrane Fuel Cell systems (PEMFC) usually have a life duration of around 2000 hours, whereas 6000 hours are required for some applications, including transportation.

development of improvements to those systems.

41 ANR PROPICE Summer School Diagnostics and Prognostics of Fuel Cell Systems July 2014, FCLAB, Belfort, France Motivations and objectives Fuel Cell Systems (FCS) appear to be a promising energy conversion device to face some of the economic and environmental challenges of modern society. However, even if this technology is close to being competitive, it is not yet ready to be considered for large scale industrial deployment: FCS still must be optimized, particularly by increasing their limited lifespan. Indeed, Proton Exchange Membrane Fuel Cell systems (PEMFC) usually have a life duration of around 2000 hours, whereas 6000 hours are required for some applications, including transportation... Enhancing FCS durability involves not only developing a better understanding of ageing phenomena but also requires the ability to emulate the behavior of the whole system to support the development of improvements to those systems. Prognostics and Health Management (PHM) of FCS is an emerging field of scientific and technological developments that has the potential to provide and enable improvements in the life management, use and support of Fuel Cell Systems. Objectives and program The aim of this summer school is to provide a forum for researchers and practitioners to discuss PHM of Fuel Cell Systems, and identify actual and future research challenges in the area. Topics of degradation mechanisms, diagnostic and prognostics of FCS, as well as aspects related to the social and economic challenges for a larger diffusion of FCS will be addressed. Courses will combine: Academic and industrial lectures given by experts in the field; Real case studies demonstrations with experimental manipulation on PEMFC platforms. Program (see reverse side for more details) Day 1: Introduction to Fuel Cell Technology Day 2: Diagnostics and prognostics - backgrounds Day 3: Socio-economic and industrial perspectives Day 4: Case studies and demonstrations Participants and registration The school is open to both academics (from University) and professionals (from Industry). Scientists and practitioners interest in PHM technologies and application to Proton Exchange Membrane Fuel Cell (PEMFC) are encouraged to register. Registration fee (online registration, 200 ) includes: Summer School facilities; Proceedings (hard copy); Coffee breaks, daily lunches and gala dinner.

A Data-driven Approach for Remaining Useful Life Prediction of Critical Components

GT S3 : Sûreté, Surveillance, Supervision Meeting GdR Modélisation, Analyse et Conduite des Systèmes Dynamiques (MACS) January 28 th, 2014 A Data-driven Approach for Remaining Useful Life Prediction of