Prognostics and Health Management E-Book

Douglas Goodman

• Herausgeber: John Wiley & Sons
• Kategorie: Wissenschaft und neue Technologien
• Serie: Wiley Series in Quality and Reliability Engineering
• Sprache: Englisch

Seitenzahl: 501

Veröffentlichungsjahr: 2019

Beschreibung

A comprehensive guide to the application and processing of condition-based data to produce prognostic estimates of functional health and life. Prognostics and Health Management provides an authoritative guide for an understanding of the rationale and methodologies of a practical approach for improving system reliability using conditioned-based data (CBD) to the monitoring and management of health of systems. This proven approach uses electronic signatures extracted from conditioned-based electrical signals, including those representing physical components, and employs processing methods that include data fusion and transformation, domain transformation, and normalization, canonicalization and signal-level translation to support the determination of predictive diagnostics and prognostics. Written by noted experts in the field, Prognostics and Health Management clearly describes how to extract signatures from conditioned-based data using conditioning methods such as data fusion and transformation, domain transformation, data type transformation and indirect and differential comparison. This important resource: * Integrates data collecting, mathematical modelling and reliability prediction in one volume * Contains numerical examples and problems with solutions that help with an understanding of the algorithmic elements and processes * Presents information from a panel of experts on the topic * Follows prognostics based on statistical modelling, reliability modelling and usage modelling methods Written for system engineers working in critical process industries and automotive and aerospace designers, Prognostics and Health Management offers a guide to the application of condition-based data to produce signatures for input to predictive algorithms to produce prognostic estimates of functional health and life.

Leseprobe

Table of Contents

Cover

List of Figures

Series Editor's Foreword

Preface

Acknowledgments

1 Introduction to Prognostics

1.1 What Is Prognostics?

1.2 Foundation of Reliability Theory

1.3 Failure Distributions Under Extreme Stress Levels

1.4 Uncertainty Measures in Parameter Estimation

1.5 Expected Number of Failures

1.6 System Reliability and Prognosis and Health Management

1.7 Prognostic Information

1.8 Decisions on Cost and Benefits

1.9 Introduction to PHM: Summary

References

Further Reading

2 Approaches for Prognosis and Health Management/Monitoring (PHM)

2.1 Introduction to Approaches for Prognosis and Health Management/Monitoring (PHM)

2.2 Model‐Based Prognostics

2.3 Data‐Driven Prognostics

2.4 Hybrid‐Driven Prognostics

2.5 An Approach to Condition‐Based Maintenance (CBM)

2.6 Approaches to PHM: Summary

References

Further Reading

3 Failure Progression Signatures

3.1 Introduction to Failure Signatures

3.2 Basic Types of Signatures

3.3 Model Verification

3.4 Evaluation of FFS Curves: Nonlinearity

3.5 Summary of Data Transforms

3.6 Degradation Rate

3.7 Failure Progression Signatures and System Nodes

3.8 Failure Progression Signatures: Summary

References

Further Reading

4 Heuristic‐Based Approach to Modeling CBD Signatures

4.1 Introduction to Heuristic‐Based Modeling of Signatures

4.2 General Modeling Considerations: CBD Signatures

4.3 CBD Modeling: Degradation‐Signature Models

4.4 DPS Modeling: FFP to DPS Transform Models

4.5 FFS Modeling: Failure Level and Signature Modeling

4.6 Heuristic‐Based Approach to Modeling of Signatures: Summary

References

Further Reading

5 Non‐Ideal Data: Effects and Conditioning

5.1 Introduction to Non‐Ideal Data: Effects and Conditioning

5.2 Heuristic‐Based Approach Applied to Non‐Ideal CBD Signatures

5.3 Errors and Non‐Ideality in FFS Data

5.4 Heuristic Method for Adjusting FFS Data

5.5 Summary: Non‐Ideal Data, Effects, and Conditioning

References

Further Reading

6 Design: Robust Prototype of an Exemplary PHM System

6.1 PHM System: Review

6.2 Design Approaches for a PHM System

6.3 Sampling and Polling

6.4 Initial Design Specifications

6.5 Special RMS Method for AC Phase Currents

6.6 Diagnostic and Prognostic Procedure

6.7 Specifications: Robustness and Capability

6.8 Node Specifications

6.9 System Verification and Performance Metrics

6.10 System Verification: Advanced Prognostics

6.11 PHM System Verification: EMA Faults

6.12 PHM System Verification: Functional Integration

6.13 Summary: A Robust Prototype PHM System

References

Further Reading

7 Prognostic Enabling: Selection, Evaluation, and Other Considerations

7.1 Introduction to Prognostic Enabling

7.2 Prognostic Targets: Evaluation, Selection, and Specifications

7.3 Example: Cost‐Benefit of Prognostic Approaches

7.4 Reliability: Bathtub Curve

7.5 Chapter Summary and Book Conclusion

References

Further Reading

Index

End User License Agreement

List of Tables

Chapter 1

Table 1.1 Expectations and variances.

Table 1.2 Decision table for purchasing a machine.

Table 1.3 Utility table for purchasing a machine.

Chapter 2

Table 2.1 Load types and examples.

Table 2.2 Failure distributions and example applications.

Table 2.3 Examples of reliability procedures and applications (White and Bernste...

Table 2.4 Examples of temperature acceleration models (White and Bernstein 2008 ...

Table 2.5 Parametric and nonparametric methods.

Table 2.6 Kernel functions.

Table 2.7 Supervised and unsupervised classification and clustering.

Table 2.8 Some advantages and disadvantages of model‐based and data‐driven progn...

Table 2.9 Differences in focus of model‐based and heuristic‐based approaches to ...

Chapter 3

Table 3.1 Range of failure thresholds: Lot 1 and Lot 2.

Chapter 4

Table 4.1 List of decreasing signatures and models and corresponding increasing ...

Table 4.2 Models: decreasing signature, increasing signature, FFP‐to‐DPS, and FL...

Chapter 5

Table 5.1 FFS nonlinearity procedure.

Table 5.2 List of calculations, specifications, and results for Example 5.2.

Table 5.3 List of calculations, specifications, and results for Example 5.3.

Table 5.4 Example of lookup values for resistance‐temperature, platinum RTD (ITS...

Chapter 6

Table 6.1 Differences in focus of model‐based and heuristic‐based approaches to ...

Table 6.2 Table of terms and definitions for performance metrics.

Table 6.3 SMPS example: table of number of data points and times to converge to ...

Table 6.4 Performance measurements and metrics.

Chapter 7

Table 7.1 Performance measurements and metrics.

Table 7.2 Summarized list of TBF and TTF calculations for various periods of ope...

Table 7.3 Tabulated calculations for PD(EST).

Table 7.4 Tabulated calculations using enhanced program with PITTFF0 = 4800.

Table 7.5 Cost estimates for benefits evaluation of prognostic enabling.

Table 7.6 Summarized list of test results.

List of Illustrations

Chapter 1

Figure 1.1 Core prognostic frameworks in a PHM system.

Figure 1.2 Framework diagram for a PHM system.

Figure 1.3 Graph of the exponential CDF with

λ

 = 3.

Figure 1.4 Graph of the Weibull CDF with

β

 = 1.2

and

η = 5

...

Figure 1.5 Graph of the exponential PDF with

λ

 = 3

.

Figure 1.6 Graphs of gamma PDFs.

Figure 1.7 Graphs of Weibull PDFs.

Figure 1.8 Failure rates of Weibull variables.

Figure 1.9 Failure rates of gamma variables.

Figure 1.10 Failure rate of the standard normal variable.

Figure 1.11 Failure rate of the lognormal variable with

μ = 0

...

Figure 1.12 Logistic failure rate with

μ

 = 0

and

σ = 1

...

Figure 1.13Figure 1.13 Gumbel failure rate.

Figure 1.14 Log‐logistic failure rate with

μ

 = 0

.

Figure 1.15 Integration domain.

Figure 1.16 High‐level block diagram of a PHM system.

Figure 1.17 A framework for CBM for PHM.

Figure 1.18 Taxonomy of prognostic approaches.

Figure 1.19 Example of an FFP signature – a curvilinear (convex), noisy charact...

Figure 1.20 Ideal DPS transfer curve superimposed on an FFP signature.

Figure 1.21 Ideal DPS, degradation threshold, and functional failure.

Figure 1.22 Normalized and transformed FFP and DPS transformed into FFS.

Figure 1.23 Ideal FFS – transfer curve for CBD.

Figure 1.24 Variability in DPS transfer curves.

Figure 1.25 FFS transforms of the DPS plots shown in Figure 1.23 .

Figure 1.26 FFS and prognostic information.

Figure 1.27 FFS transfer curve exhibiting distortion, noise, and change in degr...

Figure 1.28 Example plots of an ideal RUL and ideal PH.

Figure 1.29 Example plots of an ideal SoH transfer curve and PH accuracy.

Figure 1.30 Example of RUL with an initial‐estimate error of 100 days.

Figure 1.31 Random‐walk with Kalman‐like filtering solution for a high‐value in...

Figure 1.32 Random‐walk with Kalman‐like filtering solution for a low‐value ini...

Figure 1.33 Example of FFS data exhibiting an offset error, distortion, and noi...

Figure 1.34 Utility function of price.

Figure 1.35 Utility function of expected lifetime.

Figure 1.36 Shape of the objective function.

Chapter 2

Figure 2.1 Block diagram showing three approaches to PHM.

Figure 2.2 Precision and complexity: relative comparison of classical PHM appro...

Figure 2.3 Model‐based approach to development and use.

Figure 2.4 Model‐use diagram.

Figure 2.5 A framework for CBM for PHM (CAVE3 2015 ).

Figure 2.6 Transition diagram.

Figure 2.7 Example of a fault tree showing an RTD fault and an RTD‐usage model.

Figure 2.8 Example plots of Weibull distributions.

Figure 2.9 A special system structure.

Figure 2.10 HALT result – 30 of 32 FPGA devices failed (Hofmeister et al. 2006...

Figure 2.11 Family of failure curves, failure distribution, and TTF.

Figure 2.12 Diagram of data‐driven approaches.

Figure 2.13 A special neural network.

Figure 2.14 Comparison of model‐based (PoF) and data‐driven prognostic approach...

Figure 2.15 Relative comparison of PHM approaches – PoF, data‐driven, and hybri...

Figure 2.16 Example diagram of a heuristic‐based CBM system using CBD‐based mod...

Figure 2.17 Diagram comparison of model‐based and CBD‐signature approaches to P...

Figure 2.18 Simplified diagram of a switch‐mode power supply (SMPS) with an out...

Figure 2.19 Unreliability plots for three models for capacitor failures (Alan e...

Figure 2.20 Example of the output of a SMPS.

Figure 2.21 Example of a ringing response from an electrical circuit to an abru...

Figure 2.22 Relationship of prognostic specifications (PD and PDα) to RUL and P...

Figure 2.23 Simulated change in resonant frequency as filter capacitance degrad...

Figure 2.24 Experimental change in resonant frequency as filter capacitance deg...

Chapter 3

Figure 3.1 Diagram of classical and CBD prognostic approaches for PHM systems.

Figure 3.2 Functional block diagram for CBD signature data and processing flow.

Figure 3.3 Example of CBD containing feature data and noise (FD + Noise).

Figure 3.4 Example of signals at an output node of an SMPS.

Figure 3.5 Example of a damped‐ringing response (Judkins and Hofmeister 2007 ...

Figure 3.6 Modeling a damped‐ringing response.

Figure 3.7 Example of a CBD signature: FD is the resonant frequency of a damped...

Figure 3.8 CBD signature and levels for SMPS lot 1 (top) and SMPS lot 2 (bottom...

Figure 3.9 Functional block diagram for FFP signature and processing flow.

Figure 3.10 FFP signatures using a fixed value and a calibrated value for nomin...

Figure 3.11 FFP signature, calibrated value for nominal frequency, failure thre...

Figure 3.12 FFS signatures for FL = 0.6 (top) and FL = 0.7 (bottom).

Figure 3.13 DPS and FFP signatures for data shown in Figure 3.11 .

Figure 3.14 FFP and DPS Showing FL = 0.4 and FL = 0.5.

Figure 3.15 DPS‐based FFS (FL = 0.65) and FFP‐based FFS (FL = 0.7).

Figure 3.16 Examples of signatures: decreasing (top) and increasing (bottom) sl...

Figure 3.17 Other examples of signatures: decreasing (top) and increasing (bott...

Figure 3.18 Simulated CBD‐based signature: FD = CBD − NM.

Figure 3.19 Differences: experimental and simulated signatures.

Figure 3.20 Simulated (top) and comparison of experiment (bottom) FFP signature...

Figure 3.21 Simulated DPS (top) and experimental DPS (bottom).

Figure 3.22 Simulated FFS from simulated (top) and from experimental DPS (botto...

Figure 3.23 Illustration of point‐by‐point FNL comparison.

Figure 3.24 Illustration of total FNL

E

comparison.

Figure 3.25 Example plot of FFP‐based and DPS‐based

FNL

i

.

Figure 3.26 SMPS output and extracted damped‐ringing response.

Figure 3.27 CBD signature and FFP signature.

Figure 3.28 FFP‐based FFS and DPS.

Figure 3.29 DPS‐based FFS.

Figure 3.30 Procedural diagram for producing a DPS‐based FFS.

Figure 3.31 Linear DPS (left side) and nonlinear DPS (right side) data plots.

Figure 3.32 Simulated CBD plots using nonlinear and linear degradation rates.

Figure 3.33 Comparison of ideal DPS to DPS from data for a nonconstant degradat...

Figure 3.34 Node‐based framework for supporting failure progression signatures.

Chapter 4

Figure 4.1 Block diagram for offline modeling of CBD signatures.

Figure 4.2 Flow diagram for developing signature models.

Figure 4.3 Power function #1: increasing curves, decreasing slope angles.

Figure 4.4 Power function #1: increasing curves, increasing slope angles.

Figure 4.5 Power function #2: decreasing curves, decreasing slope angles.

Figure 4.6 Power function #2: decreasing curves, increasing slope angles.

Figure 4.7 Power function #3: increasing curves, vertically asymptotic,

dP

i

 < 

P

Figure 4.8 Power function #4: decreasing curves, vertically asymptotic,

dP

i

 < 

P

Figure 4.9 Power function #5: increasing curves, horizontally asymptotic.

Figure 4.10 Power function #6: decreasing curves, horizontally asymptotic.

Figure 4.11 Power function #7: increasing curves, slightly curvilinear, decreas...

Figure 4.12 Power function #7: increasing curves, slightly curvilinear, increas...

Figure 4.13 Power function #8: decreasing curves, slightly curvilinear, decreas...

Figure 4.14 Power function #8: decreasing curves, slightly curvilinear, increas...

Figure 4.15 Power function #9: increasing curves, vertically asymptotic,

dP

i

 < 

Figure 4.16 Power function #9: increasing curves, horizontally asymptotic,

dP

i

 ...

Figure 4.17 Power function #10: decreasing curves, vertically asymptotic,

dP

i

 <...

Figure 4.18 Power function #10: decreasing curves, horizontally asymptotic,

dP

i

Figure 4.19 Exponential function #11: increasing curves, vertically asymptotic.

Figure 4.20 Exponential function #12: decreasing curves, vertically asymptotic.

Figure 4.21 Exponential function #13: increasing curves, horizontally asymptoti...

Figure 4.22 Exponential function #14: decreasing curves, horizontally asymptoti...

Figure 4.23 Simulated FFP signatures: power function #1.

Figure 4.24 Simulated DPS from FFP signatures: power function #1.

Figure 4.25 Simulated FFP signatures: power function #3.

Figure 4.26 Simulated DPS from FFP signatures: power function #3.

Figure 4.27 Simulated FFP signatures: power function #5.

Figure 4.28 Simulated DPS from FFP signatures: power function #5.

Figure 4.29 Simulated FFP signatures: power function #7.

Figure 4.30 Simulated DPS from FFP signatures: power function #7.

Figure 4.31 Simulated FFP signatures: power function #9.

Figure 4.32 Simulated DPS from FFP signatures: power function #9.

Figure 4.33 Simulated FFP signatures: exponential function #11.

Figure 4.34 Simulated DPS from FFP signatures: exponential function #11.

Figure 4.35 Simulated FFP signatures: exponential function #13.

Figure 4.36 Simulated DPS from FFP signatures: exponential function #13.

Figure 4.37 Simulated FFP and DPS signatures: power function #1 for

n

 = 2.0.

Figure 4.38 Simulated FFP and DPS signatures: power function #3 for

n

 = 2.0.

Figure 4.39 Simulated FFP and DPS signatures: power function #5 for

n

 = 1.5.

Figure 4.40 Simulated FFP and DPS signatures: power function #7 for

n

 = 0.75.

Figure 4.41 Simulated FFP and DPS signatures: power function #9 for

n

 = 0.25.

Figure 4.42 Simulated FFP and DPS signatures: exponential function #11 for P

0

 =...

Figure 4.43 Simulated FFP and DPS signatures: exponential function #13 for P

0

 =...

Figure 4.44 Example plots: FFS (top) and FFP and DPS (bottom): power function #...

Figure 4.45 Example plots: FFS (top) and FFP and DPS (bottom): power function #...

Chapter 5

Figure 5.1 Example of a non‐ideal FFP signature and an ideal representation of...

Figure 5.2 Plots of a family of FFP signatures and DPS transfer curves.

Figure 5.3 Plots of a curvilinear FFP, the transform to a linear DPS (top), and...

Figure 5.4 Offline phase to develop a prognostic‐enabling solution of a PHM sys...

Figure 5.5 Diagram of an online phase to exploit a prognostic‐enabling solution...

Figure 5.6 Plot of a non‐ideal CBD signature data: noisy ripple voltage, output...

Figure 5.7 Example plots: non‐ideal FFP signature data (top) and transformed DP...

Figure 5.8 Example plots: non‐ideal and ideal FFS data (top) and FNL (bottom).

Figure 5.9 Example plots: non‐ideal and ideal FFS data (top) and FNL (bottom) a...

Figure 5.10 Example plot: non‐ideal ADC transfer curve.

Figure 5.11 Example plots: ideal input and non‐ideal output from an ADC.

Figure 5.12 Temperature (a), voltage (b), and current (c) plots.

Figure 5.13 Temperature‐dependent (a) and temperature‐independent (b) plots of ...

Figure 5.14 Example of a damped‐ringing response.

Figure 5.15 Sampling diagram for Example 5.8.

Figure 5.16 ADC example: saw‐tooth input, sampling, digital output value.

Figure 5.17 Simulated data before (top) and after (bottom) filtering of white (...

Figure 5.18 Degradation signature exhibiting a change in shape.

Figure 5.19 Experimental data: temperature measurements for a jet engine.

Figure 5.20 Differential signature from temperature measurements for each of tw...

Figure 5.21 Temperature data and differential signatures: four engines on an ai...

Figure 5.22 Composite differential‐distance signature.

Figure 5.23 Example of a noisy FFP signature.

Figure 5.24 Example of a noisy FFP signature: calculated nominal FD value, redu...

Figure 5.25 Example of a smoothed (3‐point moving average) FFP signature.

Figure 5.26 Example of a DPS from a smoothed FFP signature.

Figure 5.27 Example of an FFS from a smoothed FFP signature.

Figure 5.28 Example of a {FNLi} plot from a smoothed FFS.

Figure 5.29 Example random‐walk paths and FFS input.

Figure 5.30 Example plots of input FFS data and adjusted FFS data.

Figure 5.31 Example of a {FNLi} plot from an adjusted FFS.

Figure 5.32 Ripple voltage: plots of an unsmoothed (top) and smoothed (bottom) ...

Figure 5.33 Ripple voltage: plots of an unsmoothed (top) and smoothed (bottom) ...

Figure 5.34 Smoothed FFS: for NM = 3.0 mV (top) and for NM = 2.0 mV (bottom).

Chapter 6

Figure 6.1 Core prognostic frameworks in a PHM system.

Figure 6.2 A framework for CBM‐based PHM.

Figure 6.3 Random walk with Kalman‐like filtering solution for a high‐value ini...

Figure 6.4 Random walk with Kalman‐like filtering solution for a low‐value init...

Figure 6.5 Block diagram showing three approaches to PHM.

Figure 6.6 Model development and use.

Figure 6.7 Diagram: model‐based and CBD‐signature approaches to PHM.

Figure 6.8 Example diagram: heuristic‐based CBM system using CBD‐based modeling...

Figure 6.9 Procedural diagram for producing a DPS‐based FFS.

Figure 6.10 Examples of CBD, FFP, DPS, and DPS‐based FFS.

Figure 6.11 DPS‐based FFS and FFP‐based FFS.

Figure 6.12 FNL plots for the FFS shown in Figure 6.11 .

Figure 6.13 Plots of a family of FFP signatures and DPS transfer curves.

Figure 6.14 Offline phase to develop a prognostic‐enabling solution for a PHM s...

Figure 6.15 Online phase to exploit a prognostic‐enabling solution.

Figure 6.16 Multiple temperature signals: before and after differential‐distanc...

Figure 6.17 Extracted FD from fusing differential‐distance conditioned CBD.

Figure 6.18 Block diagram of an example EMA subsystem.

Figure 6.19 Offline modeling and development diagram.

Figure 6.20 Design and analysis diagram.

Figure 6.21 Continual (top) and periodic (bottom) sampling.

Figure 6.22 Period‐burst sampling: sampling period (TS) and burst period (TB).

Figure 6.23 Example of power supplies and EMA subsystems.

Figure 6.24 SMPS output showing ripple period (TR) and sampling period (TS).

Figure 6.25 Damped‐ringing response caused by an abrupt load change.

Figure 6.26 Example of burst of burst sampling.

Figure 6.27 Example of alerts issued using an unrealistic PHM system clock and/...

Figure 6.28 Steps in test data due to low‐resolution fault injection.

Figure 6.29 Diagram of a test bed to inject faults into a power supply.

Figure 6.30 Diagram of a test bed to inject faults into an EMA.

Figure 6.31 Sampled‐ and windowed‐phase currents: no load (top) and extra load ...

Figure 6.32 FFP signature of fault‐injected SMPS.

Figure 6.33 Sampled‐phase currents: no load (top) and extra load (bottom).

Figure 6.34 Current magnitude: no degradation (high), degraded transistor (redu...

Figure 6.35 Shifted levels due to a degraded power‐switching transistor.

Figure 6.36 Illustration of using peak positive and negative threshold values.

Figure 6.37 Special rms: threshold and truncation.

Figure 6.38 Special rms applied to three phase currents.

Figure 6.39 FFP signatures due to loading.

Figure 6.40 Smoothed FFP signatures due to loading.

Figure 6.41 FFS: EMA loading.

Figure 6.42 Smoothed FFP signatures due to winding faults.

Figure 6.43 FFS: EMA winding.

Figure 6.44 Phase A currents: transistor fault in the positive Phase A portion.

Figure 6.45 Phase B currents: transistor fault in the positive Phase A portion.

Figure 6.46 Phase C currents: transistor fault in the positive Phase A portion.

Figure 6.47 Peak‐RMS currents: transistor fault in the positive Phase A portion...

Figure 6.48 Close‐up of the EMA motor winding.

Figure 6.49 FFP: H bridge fault, sum of both halves of the Phase A current.

Figure 6.50 Smoothed FFP: H bridge fault.

Figure 6.51 Smoothed FFS: H bridge fault.

Figure 6.52 Block diagram for an example of a robust PHM system.

Figure 6.53 Architectural block diagram for a node definition.

Figure 6.54 Example of a system node definition.

Figure 6.55 Block diagram of an example node definition.

Figure 6.56 NDEF: node status.

Figure 6.57 NDEF: sampling specifications.

Figure 6.58 NDEF: alert specifications.

Figure 6.59 NDEF: Special Files specifications.

Figure 6.60 NDEF: feature‐vector framework – (a) primary, (b) smoothing, (c) FF...

Figure 6.61 NDEF: Prediction Framework.

Figure 6.62 NDEF: Performance Services/Graphics.

Figure 6.63 NDEF: Input & Output Files.

Figure 6.64 NDEF: Checkpoint Library & File Name.

Figure 6.65 NDEF: Device Driver Program ID and Units of Measure.

Figure 6.66 NDEF: Other Program IDs.

Figure 6.67 NDEF: End of Definition.

Figure 6.68 NDEF updates to support node second SMPS (node 60).

Figure 6.69 NDEF updates to support EMA 1 (node 51).

Figure 6.70 Illustration and relationship of PH to BD, sample time, RUL, and EO...

Figure 6.71 Relationship of degradation times and an FFS.

Figure 6.72 Illustration of the uncertainty of determining exactly when functio...

Figure 6.73 Uncertainty: prognostic distance.

Figure 6.74 CBD at 1‐hour sampling (top) and at 24‐hour sampling (bottom).

Figure 6.75 Family of failure curves, failure distribution, and TTF.

Figure 6.76 Diagram of an initial estimate for PD.

Figure 6.77 Comparison: FFS and plots of RUL and PH estimates.

Figure 6.78 Prognostic bus (log file).

Figure 6.79 SoH plot from DXARULE.

Figure 6.80 Output file for SMPS using ARULEAV.

Figure 6.81 Plots of the RUL, PH, and SoH estimates produced by ARULEAV.

Figure 6.82 Output file for SMPS using DPS‐based FFS and ARULEAV.

Figure 6.83 Plots of the RUL, PH, and SoH estimates using DPS‐based FFS and ARU...

Figure 6.84 CBD, FFP, and FFS for EMA node 51 (friction/load).

Figure 6.85 EMA (load) plots: RUL, PH, and SoH estimates using DPS‐based FFS an...

Figure 6.86 CBD, FFP, and FFS for EMA node 61 (winding).

Figure 6.87 EMA (winding) plots: RUL, PH, and SoH estimates; DPS‐based FFS and ...

Figure 6.88 CBD, FFP, and FFS for EMA node 62 (power transistor).

Figure 6.89 EMA (transistor) plots: RUL, PH, and SoH estimates; DPS‐based FFS a...

Figure 6.90 PHM: high‐level control and data flow.

Figure 6.91 Initialization: system nodes.

Figure 6.92 System alerts, part 1.

Figure 6.93 System alerts, part 2.

Figure 6.94 Example of a GUI for a PHM system.

Chapter 7

Figure 7.1 Example of a broad view of an ecosystem.

Figure 7.2 Example of a five‐level model for health solutions.

Figure 7.3 Example of alerts for SoH at or below 25%.

Figure 7.4 Example of alerts for a damage‐detection approach.

Figure 7.5 MTTF, TTF, and PITTFF0: CBD signature and failure distribution.

Figure 7.6 Same MTTF for different failure distributions and signatures.

Figure 7.7 Failure plots with average values for TTF and PD = PITTFF0.

Figure 7.8 Estimated PD and actual PD: power supply (top) and EMA load (bottom)...

Figure 7.9 Estimated PD and actual PD: EMA winding (top) and EMA transistor (bo...

Figure 7.10 RUL and PH plot for PITTFF0 = 4800 and for PITTFF0 = 2290.

Figure 7.11 RUL and PH plots for SMPS: before and after PITTFADJ = 2.0.

Figure 7.12Figure 7.12 RUL and PH plots for EMA 51: before and after PITTFADJ =...

Figure 7.13Figure 7.13 RUL and PH plots for EMA 61: before and after PITTFADJ =...

Figure 7.14 RUL and PH plots for EMA 62: before and after PITTFADJ = 2.0.

Figure 7.15 Plots: test results for six power supplies (top) and 12 EMAs (botto...

Figure 7.16 Bathtub curve showing three regions, MTBF, and a prognostic trigger...

Figure 7.17 Possible relationship of bathtub curve to failure distribution and ...

Figure 7.18 Multiple instances of CBD signatures and trigger points.

Guide

Cover

Table of Contents

Begin Reading

Pages

iv

xi

xii

xiii

xiv

xv

xvi

xvii

xix

xx

xxi

xxii

xxiii

xxiv

xxv

xxvii

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

E1

Wiley Series in Quality & Reliability Engineering

Dr. Andre Kleyner

Series Editor

The Wiley Series in Quality & Reliability Engineering aims to provide a solid educational foundation for both practitioners and researchers in the Q&R field and to expand the reader's knowledge base to include the latest developments in this field. The series will provide a lasting and positive contribution to the teaching and practice of engineering.

The series coverage will contain, but is not exclusive to,

Statistical methods

Physics of failure

Reliability modeling

Functional safety

Six‐sigma methods

Lead‐free electronics

Warranty analysis/management

Risk and safety analysis

Wiley Series in Quality & Reliability Engineering

Thermodynamic Degradation Science: Physics of Failure, Accelerated Testing, Fatigue and Reliability

by Alec Feinberg

October 2016

Design for Safety

by Louis J. Gullo, Jack Dixon

February 2018

Next Generation HALT and HASS: Robust Design of Electronics and Systems

by Kirk A. Gray, John J. Paschkewitz

May 2016

Reliability and Risk Models: Setting Reliability Requirements, 2nd Edition

by Michael Todinov

September 2015

Applied Reliability Engineering and Risk Analysis: Probabilistic Models and Statistical Inference

by Ilia B. Frenkel, Alex Karagrigoriou, Anatoly Lisnianski, Andre V. Kleyner

September 2013

Design for Reliability

by Dev G. Raheja (Editor), Louis J. Gullo (Editor)

July 2012

Effective FMEAs: Achieving Safe, Reliable, and Economical Products and Processes Using Failure Modes and Effects Analysis

by Carl Carlson

April 2012

Failure Analysis: A Practical Guide for Manufacturers of Electronic Components and Systems

by Marius Bazu, Titu Bajenescu

April 2011

Reliability Technology: Principles and Practice of Failure Prevention in Electronic Systems

by Norman Pascoe

April 2011

Improving Product Reliability: Strategies and Implementation

by Mark A. Levin, Ted T. Kalal

March 2003

Test Engineering: A Concise Guide to Cost‐Effective Design, Development and Manufacture

by Patrick O'Connor

April 2001

Integrated Circuit Failure Analysis: A Guide to Preparation Techniques

by Friedrich Beck

January 1998

Measurement and Calibration Requirements for Quality Assurance to ISO 9000

by Alan S. Morris

October 1997

Electronic Component Reliability: Fundamentals, Modelling, Evaluation, and Assurance

by Finn Jensen

Prognostics and Health Management

A Practical Approach to Improving System Reliability Using Condition‐Based Data

Douglas Goodman, James P. Hofmeister and Ferenc Szidarovszky Ridgetop Group, Inc., Arizona, USA

Copyright

This edition first published 2019

© 2019 John Wiley & Sons Ltd

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions.

The right of Douglas Goodman, James P. Hofmeister and Ferenc Szidarovszky to be identified as the authors of this work has been asserted in accordance with law.

Registered Offices

John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA

John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK

Editorial Office

The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK

For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com.

Wiley also publishes its books in a variety of electronic formats and by print‐on‐demand. Some content that appears in standard print versions of this book may not be available in other formats.

Limit of Liability/Disclaimer of Warranty

While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

Library of Congress Cataloging‐in‐Publication Data

Names: Goodman, Douglas (Industrial engineer), author. | Hofmeister, James

 P., author. | Szidarovszky, Ferenc, author.

Title: Prognostics and health management : a practical approach to improving

 system reliability using condition-based data / Douglas Goodman, Chief

 Engineer, Ridgetop Group, Inc., James P. Hofmeister, Distinguished

 Engineer, Ridgetop Group, Inc., Ferenc Szidarovszky, Senior

 Researcher, Ridgetop Group, Inc.

Description: Hoboken, NJ, USA : Wiley, [2019] | Includes bibliographical

 references and index. |

Identifiers: LCCN 2018060348 (print) | LCCN 2019000608 (ebook) | ISBN

 9781119356691 (AdobePDF) | ISBN 9781119356707 (ePub) | ISBN 9781119356653

 (hardcover)

Subjects: LCSH: Machinery--Reliability. | Equipment health monitoring. |

 Machinery--Maintenance and repair--Planning. | Structural

 failures--Mathematical models.

Classification: LCC TJ174 (ebook) | LCC TJ174 .G66 2019 (print) | DDC

 621.8/16--dc23

LC record available at https://lccn.loc.gov/2018060348

Cover Design: Wiley

Cover Image: © Sergey Nivens/Shutterstock

List of Figures

Figure 1.1

Core prognostic frameworks in a PHM system. Source: based on IEEE (2017).

Figure 1.2

Framework diagram for a PHM system. Source: based on CAVE3 (2015).

Figure 1.3

Graph of the exponential CDF with

λ

= 3.

Figure 1.4

Graph of the Weibull CDF with

β

= 1.2

and

η

= 5

.

Figure 1.5

Graph of the exponential PDF with

λ

= 3

.

Figure 1.6

Graphs of gamma PDFs.

Figure 1.7

Graphs of Weibull PDFs.

Figure 1.8

Failure rates of Weibull variables.

Figure 1.9

Failure rates of gamma variables.

Figure 1.10

Failure rate of the standard normal variable.

Figure 1.11

Failure rate of the lognormal variable with

μ

= 0

and

σ

= 1

.

Figure 1.12

Logistic failure rate with

μ

= 0

and

σ

= 1

.

Figure 1.13

Gumbel failure rate.

Figure 1.14

Log‐logistic failure rate with

μ

= 0

.

Figure 1.15

Integration domain.

Figure 1.16

High‐level block diagram of a PHM system.

Figure 1.17

A framework for CBM for PHM. Source: after CAVE3 (2015).

Figure 1.18

Taxonomy of prognostic approaches.

Figure 1.19

Example of an FFP signature – a curvilinear (convex), noisy characteristic curve.

Figure 1.20

Ideal DPS transfer curve superimposed on an FFP signature.

Figure 1.21

Ideal DPS, degradation threshold, and functional failure.

Figure 1.22

Normalized and transformed FFP and DPS transformed into FFS.

Figure 1.23

Ideal FFS – transfer curve for CBD.

Figure 1.24

Variability in DPS transfer curves.

Figure 1.25

FFS transforms of the DPS plots shown in Figure 1.23.

Figure 1.26

FFS and prognostic information.

Figure 1.27

FFS transfer curve exhibiting distortion, noise, and change in degradation rate.

Figure 1.28

Example plots of an ideal RUL and ideal PH.

Figure 1.29

Example plots of an ideal SoH transfer curve and PH accuracy.

Figure 1.30

Example of RUL with an initial‐estimate error of 100 days.

Figure 1.31

Random‐walk with Kalman‐like filtering solution for a high‐value initial‐estimate error.

Figure 1.32

Random‐walk with Kalman‐like filtering solution for a low‐value initial‐estimate error.

Figure 1.33

Example of FFS data exhibiting an offset error, distortion, and noise.

Figure 1.34

Utility function of price.

Figure 1.35

Utility function of expected lifetime.

Figure 1.36

Shape of the objective function.

Figure 2.1

Block diagram showing three approaches to PHM. Source: based on Pecht (2008).

Figure 2.2

Precision and complexity: relative comparison of classical PHM approaches.

Figure 2.3

Model‐based approach to development and use.

Figure 2.4

Model‐use diagram. Source: based on Medjaher and Zerhouni (2013).

Figure 2.5

A framework for CBM for PHM (CAVE3 2015).

Figure 2.6

Transition diagram.

Figure 2.7

Example of a fault tree showing an RTD fault and an RTD‐usage model.

Figure 2.8

Example plots of Weibull distributions.

Figure 2.9

A special system structure.

Figure 2.10

HALT result – 30 of 32 FPGA devices failed (Hofmeister et al. 2006).

Figure 2.11

Family of failure curves, failure distribution, and TTF.

Figure 2.12

Diagram of data‐driven approaches.

Figure 2.13

A special neural network.

Figure 2.14

Comparison of model‐based (PoF) and data‐driven prognostic approaches.

Figure 2.15

Relative comparison of PHM approaches – PoF, data‐driven, and hybrid.

Figure 2.16

Example diagram of a heuristic‐based CBM system using CBD‐based modeling.

Figure 2.17

Diagram comparison of model‐based and CBD‐signature approaches to PHM.

Figure 2.18

Simplified diagram of a switch‐mode power supply (SMPS) with an output filter.

Figure 2.19

Unreliability plots for three models for capacitor failures (Alan et al. 2011).

Figure 2.20

Example of the output of a SMPS.

Figure 2.21

Example of a ringing response from an electrical circuit to an abrupt stimulus.

Figure 2.22

Relationship of prognostic specifications (PD and PDα) to RUL and PH.

Figure 2.23

Simulated change in resonant frequency as filter capacitance degrades.

Figure 2.24

Experimental change in resonant frequency as filter capacitance degrades.

Figure 3.1

Diagram of classical and CBD prognostic approaches for PHM systems. Source: based on Pecht (2008).

Figure 3.2

Functional block diagram for CBD signature data and processing flow.

Figure 3.3

Example of CBD containing feature data and noise (FD + Noise).

Figure 3.4

Example of signals at an output node of an SMPS.

Figure 3.5

Example of a damped‐ringing response (Judkins and Hofmeister 2007).

Figure 3.6

Modeling a damped‐ringing response. Source: based on Judkins and Hofmeister (2007).

Figure 3.7

Example of a CBD signature: FD is the resonant frequency of a damped‐ringing response.

Figure 3.8

CBD signature and levels for SMPS lot 1 (top) and SMPS lot 2 (bottom).

Figure 3.9

Functional block diagram for FFP signature and processing flow.

Figure 3.10

FFP signatures using a fixed value and a calibrated value for nominal frequency.

Figure 3.11

FFP signature, calibrated value for nominal frequency, failure threshold at 0.6 and 0.7.

Figure 3.12

FFS signatures for FL = 0.6 (top) and FL = 0.7 (bottom).

Figure 3.13

DPS and FFP signatures for data shown in Figure 3.11.

Figure 3.14

FFP and DPS Showing FL = 0.4 and FL = 0.5.

Figure 3.15

DPS‐based FFS (FL = 0.65) and FFP‐based FFS (FL = 0.7).

Figure 3.16

Examples of signatures: decreasing (top) and increasing (bottom) slope angles.

Figure 3.17

Other examples of signatures: decreasing (top) and increasing (bottom) slope angles.

Figure 3.18

Simulated CBD‐based signature: FD = CBD − NM.

Figure 3.19

Differences: experimental and simulated signatures.

Figure 3.20

Simulated (top) and comparison of experiment (bottom) FFP signatures.

Figure 3.21

Simulated DPS (top) and experimental DPS (bottom).

Figure 3.22

Simulated FFS from simulated (top) and from experimental DPS (bottom): FL = 0.65.

Figure 3.23

Illustration of point‐by‐point FNL comparison.

Figure 3.24

Illustration of total FNL

E

comparison.

Figure 3.25

Example plot of FFP‐based and DPS‐based

FNL

i

.

Figure 3.26

SMPS output and extracted damped‐ringing response.

Figure 3.27

CBD signature and FFP signature.

Figure 3.28

FFP‐based FFS and DPS.

Figure 3.29

DPS‐based FFS.

Figure 3.30

Procedural diagram for producing a DPS‐based FFS.

Figure 3.31

Linear DPS (left side) and nonlinear DPS (right side) data plots.

Figure 3.32

Simulated CBD plots using nonlinear and linear degradation rates.

Figure 3.33

Comparison of ideal DPS to DPS from data for a nonconstant degradation rate.

Figure 3.34

Node‐based framework for supporting failure progression signatures.

Figure 4.1

Block diagram for offline modeling of CBD signatures.

Figure 4.2

Flow diagram for developing signature models.

Figure 4.3

Power function #1: increasing curves, decreasing slope angles.

Figure 4.4

Power function #1: increasing curves, increasing slope angles.

Figure 4.5

Power function #2: decreasing curves, decreasing slope angles.

Figure 4.6

Power function #2: decreasing curves, increasing slope angles.

Figure 4.7

Power function #3: increasing curves, vertically asymptotic,

dP

i

<

P

0

.

Figure 4.8

Power function #4: decreasing curves, vertically asymptotic,

dP

i

<

P

0

.

Figure 4.9

Power function #5: increasing curves, horizontally asymptotic.

Figure 4.10

Power function #6: decreasing curves, horizontally asymptotic.

Figure 4.11

Power function #7: increasing curves, slightly curvilinear, decreasing slope angles.

Figure 4.12

Power function #7: increasing curves, slightly curvilinear, increasing slope angles.

Figure 4.13

Power function #8: decreasing curves, slightly curvilinear, decreasing slope angles.

Figure 4.14

Power function #8: decreasing curves, slightly curvilinear, increasing slope angles.

Figure 4.15

Power function #9: increasing curves, vertically asymptotic,

dP

i

<

P

0

.

Figure 4.16

Power function #9: increasing curves, horizontally asymptotic,

dP

i

<

P

0

.

Figure 4.17

Power function #10: decreasing curves, vertically asymptotic,

dP

i

<

P

0

.

Figure 4.18

Power function #10: decreasing curves, horizontally asymptotic,

dP

i

<

P

0

.

Figure 4.19

Exponential function #11: increasing curves, vertically asymptotic.

Figure 4.20

Exponential function #12: decreasing curves, vertically asymptotic.

Figure 4.21

Exponential function #13: increasing curves, horizontally asymptotic.

Figure 4.22

Exponential function #14: decreasing curves, horizontally asymptotic.

Figure 4.23

Simulated FFP signatures: power function #1.

Figure 4.24

Simulated DPS from FFP signatures: power function #1.

Figure 4.25

Simulated FFP signatures: power function #3.

Figure 4.26

Simulated DPS from FFP signatures: power function #3.

Figure 4.27

Simulated FFP signatures: power function #5.

Figure 4.28

Simulated DPS from FFP signatures: power function #5.

Figure 4.29

Simulated FFP signatures: power function #7.

Figure 4.30

Simulated DPS from FFP signatures: power function #7.

Figure 4.31

Simulated FFP signatures: power function #9.

Figure 4.32

Simulated DPS from FFP signatures: power function #9.

Figure 4.33

Simulated FFP signatures: exponential function #11.

Figure 4.34

Simulated DPS from FFP signatures: exponential function #11.

Figure 4.35

Simulated FFP signatures: exponential function #13.

Figure 4.36

Simulated DPS from FFP signatures: exponential function #13.

Figure 4.37

Simulated FFP and DPS signatures: power function #1 for

n

= 2.0.

Figure 4.38

Simulated FFP and DPS signatures: power function #3 for

n

= 2.0.

Figure 4.39

Simulated FFP and DPS signatures: power function #5 for

n

= 1.5.

Figure 4.40

Simulated FFP and DPS signatures: power function #7 for

n

= 0.75.

Figure 4.41

Simulated FFP and DPS signatures: power function #9 for

n

= 0.25.

Figure 4.42

Simulated FFP and DPS signatures: exponential function #11 for P

0

= 100

`

.

Figure 4.43

Simulated FFP and DPS signatures: exponential function #13 for P

0

= 150.

Figure 4.44

Example plots: FFS (top) and FFP and DPS (bottom): power function #1.

Figure 4.45

Example plots: FFS (top) and FFP and DPS (bottom): power function #5.

Figure 5.1

Example of a non‐ideal FFP signature and an ideal representation of that signature.

Figure 5.2

Plots of a family of FFP signatures and DPS transfer curves.

Figure 5.3

Plots of a curvilinear FFP, the transform to a linear DPS (top), and the transform to an FFS (bottom).

Figure 5.4

Offline phase to develop a prognostic‐enabling solution of a PHM system.

Figure 5.5

Diagram of an online phase to exploit a prognostic‐enabling solution.

Figure 5.6

Plot of a non‐ideal CBD signature data: noisy ripple voltage, output of a switched‐mode regulator.

Figure 5.7

Example plots: non‐ideal FFP signature data (top) and transformed DPS data (bottom).

Figure 5.8

Example plots: non‐ideal and ideal FFS data (top) and FNL (bottom).

Figure 5.9

Example plots: non‐ideal and ideal FFS data (top) and FNL (bottom) after using a NM of 3.0 mV.

Figure 5.10

Example plot: non‐ideal ADC transfer curve.

Figure 5.11

Example plots: ideal input and non‐ideal output from an ADC.

Figure 5.12

Temperature (a), voltage (b), and current (c) plots.

Figure 5.13

Temperature‐dependent (a) and temperature‐independent (b) plots of calculated resistance.

Figure 5.14

Example of a damped‐ringing response.

Figure 5.15

Sampling diagram for Example 5.8.

Figure 5.16

ADC example: saw‐tooth input, sampling, digital output value.

Figure 5.17

Simulated data before (top) and after (bottom) filtering of white (random) noise.

Figure 5.18

Degradation signature exhibiting a change in shape.

Figure 5.19

Experimental data: temperature measurements for a jet engine.

Figure 5.20

Differential signature from temperature measurements for each of two engines.

Figure 5.21

Temperature data and differential signatures: four engines on an aircraft.

Figure 5.22

Composite differential‐distance signature.

Figure 5.23

Example of a noisy FFP signature.

Figure 5.24

Example of a noisy FFP signature: calculated nominal FD value, reduced NM value.

Figure 5.25

Example of a smoothed (3‐point moving average) FFP signature.

Figure 5.26

Example of a DPS from a smoothed FFP signature.

Figure 5.27

Example of an FFS from a smoothed FFP signature.

Figure 5.28

Example of a {FNLi} plot from a smoothed FFS.

Figure 5.29

Example random‐walk paths and FFS input.

Figure 5.30

Example plots of input FFS data and adjusted FFS data.

Figure 5.31

Example of a {FNLi} plot from an adjusted FFS.

Figure 5.32

Ripple voltage: plots of an unsmoothed (top) and smoothed (bottom) FFP signature.

Figure 5.33

Ripple voltage: plots of an unsmoothed (top) and smoothed (bottom) FFS data.

Figure 5.34

Smoothed FFS: for NM = 3.0 mV (top) and for NM = 2.0 mV (bottom).

Figure 6.1

Core prognostic frameworks in a PHM system. Source: after IEEE (2017).

Figure 6.2

A framework for CBM‐based PHM. Source: after CAVE3 (2015).

Figure 6.3

Random walk with Kalman‐like filtering solution for a high‐value initial‐estimate error.

Figure 6.4

Random walk with Kalman‐like filtering solution for a low‐value initial‐estimate error.

Figure 6.5

Block diagram showing three approaches to PHM.

Figure 6.6

Model development and use.

Figure 6.7

Diagram: model‐based and CBD‐signature approaches to PHM.

Figure 6.8

Example diagram: heuristic‐based CBM system using CBD‐based modeling.

Figure 6.9

Procedural diagram for producing a DPS‐based FFS.

Figure 6.10

Examples of CBD, FFP, DPS, and DPS‐based FFS.

Figure 6.11

DPS‐based FFS and FFP‐based FFS.

Figure 6.12

FNL plots for the FFS shown in Figure 6.11.

Figure 6.13

Plots of a family of FFP signatures and DPS transfer curves.

Figure 6.14

Offline phase to develop a prognostic‐enabling solution for a PHM system.

Figure 6.15

Online phase to exploit a prognostic‐enabling solution.

Figure 6.16

Multiple temperature signals: before and after differential‐distance conditioning.

Figure 6.17

Extracted FD from fusing differential‐distance conditioned CBD.

Figure 6.18

Block diagram of an example EMA subsystem.

Figure 6.19

Offline modeling and development diagram.

Figure 6.20

Design and analysis diagram.

Figure 6.21

Continual (top) and periodic (bottom) sampling.

Figure 6.22

Period‐burst sampling: sampling period (TS) and burst period (TB).

Figure 6.23

Example of power supplies and EMA subsystems.

Figure 6.24

SMPS output showing ripple period (TR) and sampling period (TS).

Figure 6.25

Damped‐ringing response caused by an abrupt load change.

Figure 6.26

Example of burst of burst sampling.

Figure 6.27

Example of alerts issued using an unrealistic PHM system clock and/or polling method.

Figure 6.28

Steps in test data due to low‐resolution fault injection.

Figure 6.29

Diagram of a test bed to inject faults into a power supply.

Figure 6.30

Diagram of a test bed to inject faults into an EMA.

Figure 6.31

Sampled‐ and windowed‐phase currents: no load (top) and extra load (bottom).

Figure 6.32

FFP signature of fault‐injected SMPS.

Figure 6.33

Sampled‐phase currents: no load (top) and extra load (bottom).

Figure 6.34

Current magnitude: no degradation (high), degraded transistor (reduced amplitude).

Figure 6.35

Shifted levels due to a degraded power‐switching transistor.

Figure 6.36

Illustration of using peak positive and negative threshold values.

Figure 6.37

Special rms: threshold and truncation.

Figure 6.38

Special rms applied to three phase currents.

Figure 6.39

FFP signatures due to loading.

Figure 6.40

Smoothed FFP signatures due to loading.

Figure 6.41

FFS: EMA loading.

Figure 6.42

Smoothed FFP signatures due to winding faults.

Figure 6.43

FFS: EMA winding.

Figure 6.44

Phase A currents: transistor fault in the positive Phase A portion.

Figure 6.45

Phase B currents: transistor fault in the positive Phase A portion.

Figure 6.46

Phase C currents: transistor fault in the positive Phase A portion.

Figure 6.47

Peak‐RMS currents: transistor fault in the positive Phase A portion.

Figure 6.48

Close‐up of the EMA motor winding.

Figure 6.49

FFP: H bridge fault, sum of both halves of the Phase A current.

Figure 6.50

Smoothed FFP: H bridge fault.

Figure 6.51

Smoothed FFS: H bridge fault.

Figure 6.52

Block diagram for an example of a robust PHM system.

Figure 6.53

Architectural block diagram for a node definition.

Figure 6.54

Example of a system node definition.

Figure 6.55

Block diagram of an example node definition.

Figure 6.56

NDEF: node status.

Figure 6.57

NDEF: sampling specifications.

Figure 6.58

NDEF: alert specifications.

Figure 6.59

NDEF: Special Files specifications.

Figure 6.60

NDEF: feature‐vector framework – (a) primary, (b) smoothing, (c) FFP‐DPS transform.

Figure 6.61

NDEF: Prediction Framework.

Figure 6.62

NDEF: Performance Services/Graphics.

Figure 6.63

NDEF: Input & Output Files.

Figure 6.64

NDEF: Checkpoint Library & File Name.

Figure 6.65

NDEF: Device Driver Program ID and Units of Measure.

Figure 6.66

NDEF: Other Program IDs.

Figure 6.67

NDEF: End of Definition.

Figure 6.68

NDEF updates to support node second SMPS (node 60).

Figure 6.69

NDEF updates to support EMA 1 (node 51).

Figure 6.70

Illustration and relationship of PH to BD, sample time, RUL, and EOL.

Figure 6.71

Relationship of degradation times and an FFS.

Figure 6.72

Illustration of the uncertainty of determining exactly when functional failure occurs.

Figure 6.73

Uncertainty: prognostic distance.

Figure 6.74

CBD at 1‐hour sampling (top) and at 24‐hour sampling (bottom).

Figure 6.75

Family of failure curves, failure distribution, and TTF.

Figure 6.76

Diagram of an initial estimate for PD.

Figure 6.77

Comparison: FFS and plots of RUL and PH estimates.

Figure 6.78

Prognostic bus (log file).

Figure 6.79

SoH plot from DXARULE.

Figure 6.80

Output file for SMPS using ARULEAV.

Figure 6.81

Plots of the RUL, PH, and SoH estimates produced by ARULEAV.

Figure 6.82

Output file for SMPS using DPS‐based FFS and ARULEAV.

Figure 6.83

Plots of the RUL, PH, and SoH estimates using DPS‐based FFS and ARULEAV.

Figure 6.84

CBD, FFP, and FFS for EMA node 51 (friction/load).

Figure 6.85

EMA (load) plots: RUL, PH, and SoH estimates using DPS‐based FFS and ARULEAV.

Figure 6.86

CBD, FFP, and FFS for EMA node 61 (winding).

Figure 6.87

EMA (winding) plots: RUL, PH, and SoH estimates; DPS‐based FFS and ARULEAV.

Figure 6.88

CBD, FFP, and FFS for EMA node 62 (power transistor).

Figure 6.89

EMA (transistor) plots: RUL, PH, and SoH estimates; DPS‐based FFS and ARULEAV.

Figure 6.90

PHM: high‐level control and data flow.

Figure 6.91

Initialization: system nodes.

Figure 6.92

System alerts, part 1.

Figure 6.93

System alerts, part 2.

Figure 6.94

Example of a GUI for a PHM system. Source: Ridgetop (2018).

Figure 7.1

Example of a broad view of an ecosystem.

Figure 7.2

Example of a five‐level model for health solutions.

Figure 7.3

Example of alerts for SoH at or below 25%.

Figure 7.4

Example of alerts for a damage‐detection approach.

Figure 7.5

MTTF, TTF, and PITTFF0: CBD signature and failure distribution.

Figure 7.6

Same MTTF for different failure distributions and signatures.

Figure 7.7

Failure plots with average values for TTF and PD = PITTFF0.

Figure 7.8

Estimated PD and actual PD: power supply (top) and EMA load (bottom).

Figure 7.9

Estimated PD and actual PD: EMA winding (top) and EMA transistor (bottom).

Figure 7.10

RUL and PH plot for PITTFF0 = 4800 and for PITTFF0 = 2290.

Figure 7.11

RUL and PH plots for SMPS: before and after PITTFADJ = 2.0.

Figure 7.12

RUL and PH plots for EMA 51: before and after PITTFADJ = 2.0.

Figure 7.13

RUL and PH plots for EMA 61: before and after PITTFADJ = 2.0.

Figure 7.14

RUL and PH plots for EMA 62: before and after PITTFADJ = 2.0.

Figure 7.15

Plots: test results for six power supplies (top) and 12 EMAs (bottom).

Figure 7.16

Bathtub curve showing three regions, MTBF, and a prognostic trigger point.

Figure 7.17

Possible relationship of bathtub curve to failure distribution and MTTF.

Figure 7.18

Multiple instances of CBD signatures and trigger points.

Series Editor's Foreword

As quality and reliability science evolves, it reflects the trends and transformations of the technologies it supports. A device utilizing a new technology, whether it be a solar power panel, a stealth aircraft, or a state‐of‐the‐art medical device, needs to function properly and without failure throughout its mission life.

In addition to addressing the reliability of new technology, the field of quality and reliability engineering has been going through its own evolution, developing new techniques and methodologies aimed at process improvement and reduction in the number of design‐ and manufacturing‐related failures. One of these disciplines is prognostics and health management/monitoring (PHM), a fast‐growing field intended to ensure safety and provide the state of health and estimate remaining useful life (RUL) of components and systems. PHM injects a more proactive approach into system reliability, where application of physics of failure (PoF), degradation analysis, and modern algorithms allow the prediction of failure time and, consequently, the ability to take actions preventing failures from happening.

The advancement and growing application of functional safety standards, along with the fast development of autonomous vehicles, increases the pressure to achieve exceptionally high system reliability, thus making PHM an indispensable tool to meet these expectations.

PHM can provide many advantages to users and maintainers, including financial benefits such as operational and maintenance cost reductions and extended lifetime. Despite the additional cost required to facilitate prognostics (monitoring systems, algorithms development, etc.), PHM has positive effects on the overall lifecycle cost of the system by avoiding costly and sometimes catastrophic failures.

This book has been written by the leading experts and state‐of‐the‐art practitioners in the field of prognostics and health management/monitoring. It discusses the many technical aspects of PHM along with its cost benefits and will be an excellent addition to this book series. The Wiley Series in Quality & Reliability Engineering aims to provide a solid educational foundation for researchers and practitioners in the field of quality and reliability engineering and to expand the knowledge base by including the latest developments in these disciplines.

Despite its obvious importance, quality and reliability education is paradoxically lacking in today's engineering curriculum. Few engineering schools offer degree programs or even a sufficient variety of courses in quality or reliability methods. Therefore, the majority of quality and reliability practitioners receive their professional training from colleagues, professional seminars, publications, and technical books. The lack of formal education opportunities in this field greatly emphasizes the importance of technical publications for professional development.

We hope that this book, as well as the whole series, will continue Wiley's tradition of excellence in technical publishing and provide a lasting and positive contribution to the teaching and practice of engineering.

Dr. Andre Kleyner

Editor of the Wiley Series in Quality & Reliability Engineering

Preface

A prognostics and health management/monitoring (PHM) system can be thought of as consisting of three major systems: a sensing system consisting of a sensor and a feature vector framework, a prognosis system comprising a prediction framework and a performance‐validation framework, and a health‐management framework. Although health management is probably the most complex and most expensive, this book presents topics related to sensing systems and prognosis. An important goal of those systems is to provide accurate prognostic information regarding the prognosis of the health of the system being monitored. This book begins by presenting approaches to reliability predictions based on traditional model‐driven, data‐driven, and hybrid‐driven approaches as necessary background to understanding the rationale for the signature‐driven approaches presented later in the book. Those traditional, or handbook, methods are evaluated as inaccurate and misleading when used for prognostic estimation of a future failure.

This book then develops approaches to modeling and data handling that take into account failure modes and operational environment and conditions, and presents an approach using signatures created by extracting leading indicators of failure/condition indicators as feature data that forms condition‐based data (CBD) signatures; such signatures can be normalized and converted into dimensionless ratios called fault‐to‐failure progression (FFP) signatures. FFP signatures form families of characteristic curves, and each family of such curves is dependent on the totality of the mechanisms of failures causing degradation, resulting in changes in monitored signals.

A set of degradation signature models is developed, each model representing a single mode of degradation, and it is shown how those models can be used to transform FFP signatures into degradation‐progression signature (DPS) data that can then be transformed into functional failure signature (FFS) data that is particularly amenable to processing by prediction algorithms. FFS data forms linearized transfer curves with values of 0 or less in the absence of degradation and 100 or larger when the component, assembly, or system being monitored (a prognostic‐enabled target) reaches or exceeds a level of degradation such that the prognostic target is no longer capable of operating within specifications. When all noise – defined as any change in data not due to degradation – is removed from monitored signals, an FFP signature can be transformed into DPS data that forms a linear straight‐line transfer curve, which can be converted into FFS data by defining and using a threshold value that defines when functional failure occurs.

However, since it is not practical to remove all noise, this book also presents some useful signal‐conditioning techniques to ameliorate and/or mitigate the effects of noise, including, but not limited to, the following: data fusion, data transforms, domain transforms, filtering, threshold margins, data smoothing, and data trending. Those techniques are illustrated using example sets of noisy data, including those that exhibit signal variations and changes in the shapes of curves caused by the effects of temperature and feedback.