Frontiers of Risk and Reliability Engineering Research

Transcription

1 Frontiers of Risk and Reliability Engineering Research Mohammad Modarres Department of Mechanical Engineering Kececioglu Lecture April 14, 2016 Department of Aerospace and Mechanical Engineering University of Arizona Tucson, AZ

2 Outline A Timeline History of Reliability Engineering Frontiers in Reliability Engineering Frontiers in Risk Analysis Prognosis and Health Management (PHM): Reliability Science, Big Data, Damage Precursors Risk and Security of Complex Systems and Infrastructures Conclusions

3 Timeline Initiatives in 1950 s Weakest link Advisory Group on the Reliability of Electronic Equipment (AGREE) DoD Exponential life model Exponential Distribution Retreat in 1960 s Birth of Physics of Failure Uses of other distributions Reliability Growth Life Testing FMEA Birth of the Fault Tree Analysis in 1970 S Probabilistic Risk Assessment Common Cause Failures Uncertainty Analysis

4 Timeline (Cont.) Accelerated Life and Degradation Testing in1980 s Rebirth of Physics-of-Failure in 1990 s Probabilistic Physics of Failure Time Varying Accelerated Tests Highly Accelerated Life Testing (HALT) Hybrid Reliability and Prognosis Models in 2000 s Combined Logic Models, Physical Models and Probabilistic Models (e.g. BBN) PHM methods Powerful simulation tools (MCMC, Recursive Bayes and Bayesian filtering) Exploring Fundamental Sciences of Reliability in 2010 s Fundamental physical sciences Data science Big data Machine learning Autonomous systems and robots Infrastructure and cyber-physical safety and security

5 Key Areas of Research: Reliability Frontiers Probabilistic Physics-of-Failure (PPoF) More than 50-Years of History in PoF (More Recently PPoF) Accelerated Reliability Testing for PPoF Model Development Empirical Model for Unit-Specific Reliability Assessment Simulation-Based Reliability Hybrid Reliability Combined System Analysis Techniques: BBN, DBN, DFT, DET, Markov and Semi-Markov, FEM and FDM, FM, RBD, etc. Sensor-Based (Precursors) / Big Data Reliability Analysis Data Fusion, Machine Learning (GRP, SVM,..), Natural Language, Signal Processing, Detection Probability, Tensor-Based Computing Representative Sample-Based Approach and Data Clustering Massively Parallel Processing (MPP) Cyber-Physical Complex Systems and Structures Fundamental Science of Reliability Engineering

6 Key Areas of Research: Risk Frontiers Infrastructure Safety-Security-Resilience (SSR) Electronic Information Flow Embedded in Nearly Every Aspect of Life Integrity of Complex Systems and Networks: Cyber-Human-Software- Physical (CHSP) Systems Highly Connected Infrastructure Networks: Electricity, Gas, and Water Pose Major Societal Risks Through Cyberspace Attacks Risk Management and Resilience Societal Disruption, Health, Safety and Resilience Goals Life-Cycle Risks of Advanced Energy Systems Renewable Systems (Building, Environmental, Internal and External) Nuclear Energy (Fission and Fusion) Climate Change Risks of Disruptions in Sustained Energy Supply Health System Risks Estimates 1 Puts Medical Errors as the Third Leading Cause of Death in the U.S., Just Behind Cardiovascular Diseases and Cancer! Simulation-Based Dynamic Probabilistic Risk Assessment High Power Computing Leading to Less Inductive Risk Models More Deductive Computer-Assisted Risk Scenario Generation 1. A New, Evidence-based Estimate of Patient Harms Associated with Hospital Care, James, John T., Journal of Patient Safety, 2013 Vol. 9 - Issue 3 - p

7 Entropic Theory of Damage: A Fundamental Science of Reliability Failure mechanisms leading to degradation share a common feature at a deeper level: Dissipation of Energy Dissipation (or equivalently entropy generation) Damage Degradation mechanisms Damage Dissipation energies Entropy generation Failure 1 occurs when the accumulated total entropy generated exceeds the entropic-endurance of the unit Rudolf Clausius Entropic-endurance describes the capacity of the unit to withstand entropy Entropic-endurance of identical units is equal Entropic-endurance of different units is different Entropic-endurance to failure can be measured (experimentally) and involves stochastic variability 1. Defined as the state or condition of not meeting a requirement, desirable behavior or intended function

8 Thermodynamics as the Science of Reliability Past Future Present Why Entropy? ü Entropy can model multiple competing degradation processes leading to damage ü Entropy is independent of the path to failure ending at similar total entropy at failure ü Entropy accounts for complex synergistic effects of interacting degradation processes ü Entropy is scale independent Ludwig Boltzmann Statistical Mechanics Entropy

9 Thermodynamics as the Science of Reliability (Cont.) [4] σ = 1 T % J ' ( T + μ - T Thermal. -/0 + 1 T τ: ε T + v 8 A 8 : 8/0 + 1 T + > c < J < ψ </0 Diffusion Mechanical Chemical External field energy Product of thermodynamic forces and fluxes Entropy to Failure (MJ/m 3 K) 5 F=330 MPa 4.5 F=365 MPa 4 F=405 MPa 3.5 F=260 MPa 3 F=290 MPa Time (Cycle) 10 4 Fracture Fatigue Failure (MJ m -3 K -1 ) Number of Cycles to Failure [5] [4] Anahita Imanian and Mohammad Modarres, A Thermodynamic Entropy Approach to Reliability Assessment with Application to Corrosion Fatigue, Entropy (2015): [5] M. Naderi et al., On the Thermodynamic Entropy of Fatigue Fracture, Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, (2009): 1-16 [6] M. Naderi et al., Thermodynamic Analysis of Fatigue Failure in a Composite Laminate, Mechanics of Material 46 (2012):

10 Intersection of Data Science and Reliability: PHM Applications Damage Precursors: Any recognizable variation of materials/physical properties influenced by the evolution of the hidden/ inaccessible/ unmeasurable damage during the degradation process Heterogeneous Big Data / Information Sources Online and Offline Sensor Values Human Inspections Physical Model Predictions / Simulations

11 A Two-Stage Hybrid-Model PHM Approach Stage 1: Damage Precursor Model Microstructural Damage Mechanisms Stage 2: PPoF Models RUL Estimation Damage Precursors Sensor-Based Monitoring Extract Damage Indicators Machine Learning: Characterizing Precursor Indicator Failure Agents Model Parameters Initial Damage PPoF Models Future State of Health Current State of Health Partially Relevant Information Measurements Expert Judgments Measurement Errors Feature Extraction Measurement Errors Probability of Detection (POD) Built-in Sensors NDI Data Probability of Detection (POD)

12 Conclusions Thanks to Dr. Kececioglu s pioneering efforts reliability engineering education is rapidly growing in the US and abroad Reliability and Risk Analysis Now Forms an Integral Part of Modern Products, Systems and Infrastructures Design and Operation Exciting and Abundant Employment Opportunities Exist in Risk and Reliability Evidence: Number of Conferences, Educational Programs, Scholarly Journals, Human Resource Demands

13 Thank you for your attention!