From Prognostics and Health Systems Management to Predictive Maintenance 1

Rafael Gouriveau is Associate Professor (HDR) at the National Institute in Mechanics and Microtechnologies (ENSMM), Besançon, France, working with the AS2M (Automatic Control and Micro-Mechatronic Systems) department of the FEMTO-ST Institute. His research interests mainly concern the development of Prognostics and Health Management (PHM) solutions. He is also the scientific coordinator of the PHM research axis from the FCLAB (Fuel Cell Lab) Research Federation (CNRS).Kamal Medjaher is Professor at the Tarbes National School of Engineering (ENIT), France. He conducts his research activities within the Production Engineering Laboratory (LGP). He currently leads research works in the field of Prognostics and Health Management of industrial systems.Noureddine Zerhouni is Professor at the National Institute in Mechanics and Microtechnologies (ENSMM), Besançon, France, working with the AS2M (Automatic Control and Micro-Mechatronic Systems) department of the FEMTO-ST Institute. His current research interests include modeling, analysis and control of manufacturing systems, intelligent maintenance systems, e-maintenance, and Prognostics and Health Management (PHM).

• Introduction  ixChapter 1. PHM and Predictive Maintenance  11.1. Anticipative maintenance and prognostics 11.1.1. New challenges and evolution of the maintenance function 11.1.2. Towards an anticipation of failure mechanisms 31.2. Prognostics and estimation of the remaining useful life (RUL)  51.2.1. What is it? Definition and measures of prognostics  51.2.2. How? Prognostic approaches 61.3. From data to decisions: the PHM process  91.3.1. Detection, diagnostics and prognostics 91.3.2. CBM Architecture and PHM process  101.4. Scope of the book 12Chapter 2. Acquisition: From System to Data  152.1. Motivation and content  152.2. Critical components and physical parameters  162.2.1. Choice of critical components – general approach 162.2.2. Dependability analysis of the system and related tools  172.2.3. Physical parameters to be observed 192.3. Data acquisition and storage 202.3.1. Choice of sensors 222.3.2. Data acquisition  232.3.3. Preprocessing and data storage  242.4. Case study: toward the PHM of bearings  252.4.1. From the “train” system to the critical component “bearing” 252.4.2. Experimental platform Pronostia 262.4.3. Examples of obtained signals 302.5. Partial synthesis  30Chapter 3. Processing: From Data to Health Indicators 333.1. Motivation and content  333.2. Feature extraction 353.2.1. Mapping approaches 353.2.2. Temporal and frequency features  363.2.3. Time–frequency features 383.3. Feature reduction/selection  483.3.1. Reduction of the feature space  483.3.2. Feature selection . 543.4. Construction of health indicators 623.4.1. An approach based on the Hilbert-Huang transform 623.4.2. Approach description and illustrative elements  623.5. Partial synthesis  63Chapter 4. Health Assessment, Prognostics and Remaining Useful Life – Part A 674.1. Motivation and content  674.2. Features prediction by means of connectionist networks 694.2.1. Long-term connectionist predictive systems  694.2.2. Prediction by means of “fast” neural networks 774.2.3. Applications in PHM problems and discussion 844.3. Classification of states and RUL estimation 884.3.1. Health state assessment without a priori information about the data 884.3.2. Toward increased performances: S-MEFC algorithm 934.3.3. Dynamic thresholding procedure  954.4. Application and discussion  974.4.1. Tests data and protocol  974.4.2. Illustration of the dynamic thresholding procedure  1014.4.3. Performances of the approach  1044.5. Partial synthesis  105Chapter 5. Health Assessment, Prognostics, and Remaining Useful Life – Part B 1095.1. Motivation and object 1095.2. Modeling and estimation of the health state 1115.2.1. Fundamentals: the Hidden Markov Models (HMM)  1115.2.2. Extension: mixture of Gaussians HMMs  1175.2.3. State estimation by means of Dynamic Bayesian Networks 1185.3. Behavior prediction and RUL estimation  1245.3.1. Approach: Prognostics by means of DBNs 1245.3.2. Learning of state sequences 1245.3.3. Health state detection and RUL estimation 1265.4. Application and discussion  1295.4.1. Data and protocol of the tests 1295.4.2. Health state identification 1315.4.3. RUL estimation  1335.5. Partial synthesis  135Conclusion and Open Issues  137Bibliography 143Index  163