OVERVIEW. AVL Reliability Engineering & Load Matrix

Transcription

1 AVL Reliability Engineering and Load Matrix OVERVIEW January 2006 Page 1

2 Content The Reliability Challenge AVL s Reliability Engineering Approach The Load Matrix Conclusions January 2006 Page 2

3 The Reliability Challenge Highly innovative systems (e.g., DPF, SCR, hybrids) Increasing complexity of electronics and software Demanding customer expectations concerning reliability and availability More full service contracts Product Complexity Reliability Issues & Warranty Costs? Development Time Warranty Periods January 2006 Page 3

4 Content The Reliability Challenge AVL s Reliability Engineering Approach The Load Matrix Conclusions January 2006 Page 4

5 Reliability Engineering and its links to other disciplines Testing Production Engineering Quality Assurance Design & Development Reliability Engineering Customer Service AVL Reliability Engineering... Is focused on failure-free products in the field Includes a range of methods, Risk management Field and test data analysis Statistical methods Validation optimisation Is a comprehensive process throughout product development January 2006 Page 5

6 AVL s Reliability Engineering Services AVL is an established and reknown partner for... Risk assessments Reliability assessment with statistical methods Series defects analysis Warranty analysis and reliability prediction Reliability target setting Definition / optimisation of validation programs (Load Matrix, DoE) In-vehicle test program consulting January 2006 Page 6

7 AVL s s Reliability Engineering Process Concept Phase Prototype Development (Gen 1) Preproduction Development (Gen 2) Production Validation Volume Prod. Test Design Development time Concern System RA & FMEA Reliability Allocation Reliability Charts Weibull Analysis Warranty Cost Prediction Load Matrix January 2006 Page 7

8 Statistical Test Design Test Design is a statistical technique for test time calculation can be applied to test bench runs provided that the acceleration factor is known Factor (relative to B10 target) 2,40 2,20 2,00 1,80 1,60 1,40 1,20 1,00 Test Time Determination for Endurance Tests (based on 3-param. Weibull distr, beta=2, 70% confidence; no severe defect during the entire test procedure) example: B10 target = hrs; 4 endurance tests, acceleration factor = 5 test time for each prototype = / 5 * 1.5 = 3000 hrs Number of Prototypes can make use of failure time distributions (e.g. Weibull model) Benefits Solves the question: How many engines do we have to test for what time to ensure that the durability goal is reached with a certain probability? January 2006 Page 8

9 AVL s Project Risk Assessment Procedure Why? How? To get as early as possible a quick, clear and unbiased view on project risks To be able to act upon critical risks in an appropriate way The risks not to reach the project targets are rated Assessment is done similar to FMEA. Scoring system, moderated team Technical, organisational, financial, legal and social risks are covered Generation of action plan January 2006 Page 9

10 Risk Assessment Process Preparation RA Main Session RA Minutes / Report Work with RA Document RA Review RA Review Minutes PM & moderator RA team (incl PM&mod) Moderator PM RA team (incl PM&mod) Moderator Duration 1/2 day 1 day 1/2 day continuous 2-3 hours 1/2 day RA team = Risk Assessment team PM = project manager mod = moderator January 2006 Page 10

11 Risk Assessment Benefits A Project Risk Assessment... Provides an overview of the project risks Yields a team-based view of the relevance of risk topics Allows concentrating on the most critical topics Is especially useful in large and / or complex projects January 2006 Page 11

12 Smart FMEAs Still very useful when applied in an appropriate fashion Lean documentation to keep the protocols manageable Modified procedure to focus on durability and reliability Typical application in AVL 1 concept FMEA, often combined with Project Risk Assessment ~ 5 design FMEAs on the most critical subsystems, often in two stages (for layout and detailed design phase) Regular reviews Extension to Failure Mode - Parameter sheets (see Load Matrix ) January 2006 Page 12

13 Weibull Analysis Weibull Analysis is a statistical failure analysis tool for technical applications is based on the Weibull or related distributions canmakeuseof failure data from endurance test benches, field protos and customer field vehicles Benefits Cumulative Failure (%) Time ( hours) Weibull Failure Analysis Weibull analysis allows the estimation of reliability indicators such as the B10 value it provides information on the nature of failures (e.g., fatigue / wear-out failures) January 2006 Page 13

14 Reliability Allocation Benefits exhaust system RF target = 0.05 crank train RF target = 0.01 injection system RF target = 0.05 engine RF target = 0.2 cylinder head assembly RF target = 0.04 (simplified model) crank case RF target = 0.01 add-on parts RF target = 0.04 Reliability allocation provides a good overview of the whole system and therefore gives an instant view on reliability-critical parts of the system it helps in providing reliability targets for subsystems, an important input for supplier contracts it is the basis for life cycle cost calculations An engine is modelled as a block diagram consisting of subsystems reliability values (eg, B 10 and RF) are allocated to each subsystem values are derived from similar projects, prototypes or field data. Expert estimation is used where no data are available January 2006 Page 14

15 Reliability Centered Maintenance Task Provide a strategy to keep the availability of a complex system on a given target value with minimum cost. Availability = Uptime / (Uptime + Downtime) Decision basis for preventive maintenance vs. reaction to failures. Simulation with BlockSim. Allows statistical simulation Costs /a cost/a caused by preventive maintenance total cost failure density cost/a caused by unforeseen failures failure density of complex systems composed of subsystems with optimal preventive maintenance time (simulation result) MTBF Time t statistical failure distributions and repair time distributions. January 2006 Page 15

16 Reliability Charts - Reliability Improvement Monitoring Monitoring technique - shows the durability and reliability status of an engine / powertrain / vehicle during product development One chart is made for the lifetime value, another one for Repair Frequency or MTBF value hours Benefits 1.Q target 2.Q Q Q Q Q B 10 Life 3.Q Q Q Q Q Q actual B10 value (calculated) prediction 20,0 18,0 16,0 14,0 12,0 10,0 8,0 6,0 4,0 2,0 0,0 Q Q Q Q Repair Frequency (RF) Q Q Q Q Q Q target Q Q actual RF values predicton Remark. 1) 3rd and 4th quarter 2000 reflect pilot and series production, respectively Shows current and historic values of reliability indices Illustrates the rate of improvement of these indices Provides a basis for prediction of lifetime and MTBF in the future January 2006 Page 16

17 Methods: Warranty Cost Prediction Warranty Cost Prediction illustrates in which way warranty costs depend on the B 10 and the MTBF/RF values of the product reflects 100%-repair campaigns due to serial defects requires as input repair costs and subsystem failure distributions (from field data or estimated from protos) Benefits costs (million currency units) 35,0 30,0 25,0 20,0 15,0 10,0 5,0 2 yrs to SOP: RF = 10 B10 = 2000 h 1 yr to SOP: RF = 7,5 B10 = 4500 h million currency units 30,0-35,0 25,0-30,0 20,0-25,0 15,0-20,0 10,0-15,0 5,0-10,0 0,0-5,0 makes the costs of unreliability transparent 0, B RF SOP RF = 2,5 B10 = 6500 h shows if the SOP date is realistic can be used for life cycle cost prediction January 2006 Page 17

18 AND FTA - Fault Tree Analysis Method: Fault Tree Analysis is a systematic, deductive methodology for determining all possible failures / failure combinations that cause a single undesired / catastrophic event is a top-down method (contrary to FMEA) Engine failure due to low oil pressure Benefits can be used as qualitative or quantitative method; in the latter application it provides the probability of the undesired event based on single event probabilities can also be employed as a trouble-shooting method Low oil pressure warning fails Oil pump defect OR PRV valve defect Extreme bearing wearout January 2006 Page 18

19 Additonal Statistical Methods Methods (examples) Exploratory Statistics Correlation Analysis Regression Analysis ANOVA Factor Analysis Control Charts Time Series Models Benefits can detect hidden relations in the data provide objective models based on mathematical grounds January 2006 Page 19

20 Dedicated Reliability Software Software Tools Weibull ++ (Weibull analysis, B10 calculation, test design, simulation) BlockSim (reliability modeling and simulation of complex systems) Load Matrix software (AVL developed) PRISM (planned for electronic components) Other Software APIS IQ-FMEA Software Statistica (correlation analysis, regression, ANOVA, factor analysis, time series, tests, nonparametric methods, graphical methods etc) Excel (for Load Matrix, tolerance calculations, graphics) January 2006 Page 20

21 Content The Reliability Challenge AVL s Reliability Engineering Approach The Load Matrix Conclusion January 2006 Page 21

22 Former Situation Acceleration factors of durability tests often based on experience (or even guessing). No subsystem specific consideration of acceleration. For new systems (e.g. Diesel Particle Filter) there is no proven validation method. Pressure to reduce the number and the duration of validation steps. January 2006 Page 22

23 The AVL Load Matrix The Load Matrix is... a methodology to optimise test & validation programs systematically The Load Matrix is applied to... optimise existing traditional durability & reliability validation programs design optimal validation programs for new systems (eg, DPF) January 2006 Page 23

24 The Load Matrix... Load Matrix Details is based on component and failure mode specific test acceleration factors uses these specific acceleration factors as weigth factors to compare test efficiency and life coverage uses damage models to calculate acceleration The Load Matrix is used for... minimising validation costs without jeopardizing product durability & reliability detecting hidden product failure modes that are not covered by traditional tests January 2006 Page 24

25 Load Matrix Process Selection of critical parts and failure modes (based on FMEA, field data analysis, etc.) Definition of reliability targets for these critical parts / failure modes Collection of validation steps, test durations, determination of test acceleration factors Evaluation of validation program with respect to reliability & durability, weak point analysis January 2006 Page 25

26 Selection of critical parts & failure modes Existing reliability field data (eg, from previous engine) FMEAs and FP sheets of new subsystems Result: List of critical components and failure modes January 2006 Page 26

27 An Important tool: The FP Sheet FMEA, Risk Analysis, Field Data, Experience FP (Failure Mode - Parameter) Sheet = Extended FMEA with emphasis on parameters relevant for damaging and critical operating conditions Priorisation of critical failure modes Selection of damage model First completeness check of test program January 2006 Page 27

28 Single Sheet Top Ring example for passenger car engine tests acceleration factors equivalent kilometers sum & maximum of equivalent km January 2006 Page 28

29 Summary Sheet example Highlights where durability risks exist which cannot be detected within the planned test program Indicates to what extent the test program is adequate to demonstrate the desired reliability Indicates where additional actions are required to prove the required targets January 2006 Page 29

30 Calculation of Acceleration Factor (simplified) Test: Cyclic Load Test Vehicle (duty cycle) Frequency (or probability) h 1 (c) h 2 (c) h 3 (c) h 4 (c) h5 (c) Frequency Plot resulting from test procedure Frequency (or probability) h 1 (v) h 2 (v) h 3 (v) Frequency Plot resulting from statistics and invehicle measurements h (v) 4 h (v) 5 engine speed engine speed relative damage Damage profile resulting from damage model calculation d 1 d 2 d 3 d 4 d 5 identical relative damage Damage profile resulting from damage model calculation d 1 d 2 d 3 d 4 d 5 engine speed Relative damage per hour in Cyclic engine speed Relative damage per hour in vehicle Load Test: D R (c) = Σ h i (c) d i (duty cycle): D R (v) = Σ h i (v) d i Acceleration Factor = D R (c) / D R (v) January 2006 Page 30

31 Example for a Class B Model: Top Ring Wear January 2006 Page 31

32 Thin Layer Activation for Wear Prediction Top Ring Wear Map 14,00 12,00 10,00 8,00 6,00 4,00 Wear Rate [nm/h] 25% 40% 2,00 0,00 64% Speed 89% 115% Load [Nm] January 2006 Page 32

33 Example. Top Ring Wear Highway Rural Average Speed: Ring Wear: 131 km/h 0,018 nm/km Average Speed: Ring Wear: 85 km/h 0,013 nm/km Ring Wear: 2,4 nm/h Ring Wear: 1,1 nm/h Load Cycle Test: Ring Wear: 7,7 nm/h Acceleration: 7.7 / 2.4 = 3.2 Load Cycle Test: Ring Wear: 7,7 nm/h Acceleration: 7.7 / 1.1 = 7.0 Reference Duty Cycle = Highway, Rural + City: Acceleration = 5.1 January 2006 Page 33

34 Classes of Damage Models Class Method Example A Empirical model based Wear proportional to accumulated on general experience fuel consumption B simplified physical model L = (C/P) m for the lifetime of a ball bearing C full physical model FE-Analysis + Damage Accumulation Modell Hypothesis + Actual material data January 2006 Page 34

35 Data processing w.r.t. to damage time January 2006 Page 35

36 Content The Reliability Challenge AVL s Reliability Engineering Approach The Load Matrix Conclusion January 2006 Page 36

37 Load Matrix Benefits The Load Matrix... Generates complete and balanced validation plans, including analysis, component testing, test bed tests, vehicle tests. Shows how far durability and reliability targets can be demonstrated Helps to avoid unnecessary testing. Accelerates detection of problem areas and hidden failure modes Supports the exchange and proper use of key information from all involved partners, including suppliers. Supports optimised assessment procedures to make full use of all available information. Helps in deciding on the benefits of additional validation steps. Customers have adopted this procedure as a standard for engine and drivetrain validation! January 2006 Page 37

38 Examples of Load Matrix Projects Customer System Tasks European OEM HD Diesel Engine Setup of a warranty cost prediction model Assessment of Validation Program European OEM HD Diesel Engine Determination of sub-system specific acceleration factors Validation program optimisation Supplier Diesel Particle Filter Definition of a test program Definition of the LOAD MATRIX for Substrate, Mat and Canning European OEM DENOX System Definition and optimisation of a test program Japanese OEM SUV and LCV TCI Diesel Engine Determination of the effect of a different vehicle application (LCV instead of SUV) on engine life Definition of a durability test program European OEM Rear Axle Setup of Load Matrix Assessment of current test program European OEM Pass Car Gasoline Engine Comparison of two different validation programs European OEM LCV and HD TCI Diesel Engine Assessment of validation program for a DPF application January 2006 Page 38