Fatigue and Accelerated Testing of Structural Components. Steffen Haslev Sørensen

Transcription

1 Fatigue and Accelerated Testing of Structural Components Steffen Haslev Sørensen May 20, 2014

2 Topics Introduction Testing as part of other verification activities Fatigue Damage of structural component Background and assumptions Implications for testing Challenges Page 2

3 Verification activities Analysis Using mathematical models and analytical techniques Inspection Visual examination Demonstration Test Basic confirmation of performance Operation of system, subsystem or component to obtain data to verify performance or provide information to verify performance through further analysis System Engineering Fundamentals, US Department of Defense Page 3

4 Analysis Aerodynamic Analysis Analysis of blade aerodynamics, and other minor areas such as cooling Verification Activities Aeroelastic analysis Determine loads on WTG during different wind conditions Aero servo elasticity Control simulation Optimize and evaluate control algorithms before implementing them into aeroelastic code Multi-body simulation Nonlinear contact analysis for instance of gearboxes Finite Element analysis Determine multi-axial stress situation in components Component specific numerical analysis For instance bearing analysis Strength Analysis Determine fatigue and extreme strength based on stress analysis and material Page 4

5 Test Full scale test of WTG Strain gauges on blades, main shaft tower etc. Purpose is to verify the aero elastic model Required in order to obtain IEC approval Since 20+ years. Verification Activities Calibrated Accelerated Lifetime Test, CALT of blades Extreme flap and edge loads Fatigue flap and edge loads Highly Accelerated Lifetime Test, HALT Slowly emerging since Identify Failure modes Electrical components: Cabinets, controllers, IGBTs and also larger components such as generator. Hydraulic components CALT/HALT of gearbox Calibrating the equivalent load of for instance the gearing. Evaluate strength of solutions Identify weak areas CALT of other structural components Page 5 Primarily focused on components that are expensive to replace as blade bearings and yaw ring.

6 Process Risk Analysis on sub system (D-FMEA, FTA) What are high risk areas? Evaluate risk mitigating actions Is analysis enough? Is testing actually simpler/cheaper than analysis? To what extent is it necessary that testing is used to support analysis models? What is the ambition level? Determine risk mitigating actions Execute Update risk evaluation Reevaluate

7 Fatigue Damage of Structural Components and Implications on Accelerated Testing Steffen Haslev Sørensen May 20, 2014

8 Objective of the structural test The objective of a structural test can be to: Verify the material assumption that is used for analysis on material level Verify the component in full scale test Verify the component by verifying a similar component (but maybe smaller) Verify component phenomena that may occur when the component is part of the system (hardware in the loop testing) such as vibrations or interactions between component, measurement system and control system Due to a wind turbine operating 24/7 for 20 years it is necessary to accelerate the testing.

9 Fatigue crack growth Initiation phase Very slow development where slip occurs in the lattice structure. Stage 1 fatigue crack, when the plastic tip is smaller than the grain size Stage 2 growth, More predictable direction. Crack growth is estimated by fracture mechanics analysis. Most mechanical components spends the majority of the life in the initiation and stage 1. Welded parts with flaws starts immediately in Stage 2 crack growth

10 Stress Range [MPa] Range Accelerated testing and SN-Curves 1000 SN-curve according to EC ; 71 MPa Reversed cycle E E E E E+09 Number of reversed cycles Curve 71 SN-curves or Wöhler slopes originates from train axle tests. The SN-curve provides the reliationship betveel the allovable stress range and a given number of cycles. The curves are typically given at a specific survival probability and confidence level. A partial factor shall be applied to reach an acceptable strength or more precisely theoretical failure probability within say 20 years The SN-curve is used in relation to numerical analysis of components and systems, but in relation to accelerated or calibrated testing it is the slope of the SN-curve that is of biggest importance.

11 The equivalent damage For a mechanical component the SN-curve (for load cycles above 10,000) is generally accepted to form a straight line in a log-log coordinate system. Such a curve can be represented by the following equations. σ m N= k (Eurocode 3) log(σ) = a 1/m log(n) (Straight line in log-log graph) L 10 = (C/P) m (Bearing analysis) The slope, m depends primarily on the material and the local geometry (notch and stress gradients). By looking at the Eurocode 3 notation, one can determine Equivalent load situations σ 1 m N 1 =σ 2 m N 2 If we do a test, where we double the stress compared to actual applications we can reduce the test cycles as follows: 1 3 N 1 =2 3 N 2 N 2 =1/8 N 1

12 Stress Range Multiple Slopes 1000 Different slopes is a fundamental challenge as the detail with a flat SNcurves is very sensitive to overloading. If this detail is used to calibrate the test any details with a more steep slope will not be tested sufficiently. The problem gets more severe with large differences in slopes Extending the test time to say one year instead of one month can help to reduce the problem but will not solve the fundamental problem. With different slopes it may be advisable to also do sub system testing of the detail with the steep SN-curve, alternatively be prepared to during the test exchange the part with the flat SN-curve SN-Curves m=4 m= ,00E+04 1,00E+05 1,00E+06 1,00E+07 1,00E+08 1,00E+09 Reversed Cycles

13 Calibrated accelerated testing Determine the fatigue spectrum and equivalent load, which is done by the wind turbine manufacturer using the aero elastic simulation model such as HawC, Bladed or Flex5. Determine approach towards different slopes of the SN-curve. Determine acceptable test time and equipment capacity Specify test and equipment Build and execute

14 Challenges for setting up calibrated accelerated tests Practical challenges for full scale testing Size Cost Power consumption, Planning Different slopes of the SN curves Overloading may cause failure modes that are not relevant in the actual application Sequence effects where early overloading might initiate a fault may have to be considered. Test conditions might be too ideal in relation to e.g. temperature, wear or lubrication conditions Understanding failure modes Some failure modes are linked to specific interactions or transient events Does the test actually provoke the relevant failure modes Evaluation of test Often there is a significant variation in component life, which means that to achieve e.g. 90% survival probability with 90% confidence would require a very large amount of tests.

15 Full scale testing of structural systems All the potential problems set aside there are offcourse a lot of benefits in doing full scale component testing. With proper control set-up transient effects and control system interaction can be taken into account. Very well controlled conditions enables possibilities for a lot of valuable post test analysis such as evaluation of load distributions through wear investigations and stress conditions through X-ray diffraction analysis. The benefits from the test as a calibration tool for Analysis models should not be neglected.

16 Extra May 20, 2014

17 Component specific examples Blade testing Coupon testing of blade details Lamina testing Gearbox test Pitch system testing Yaw and pitch drive testing Bearing testing (scaled and full scale) Slewing bearing testing

18 Stress Range [MPa] PM-sum and Stress/load Reserve Factor 1000 SN-curve according to EC3 For this welded section with m=3: SRF 90/70= ; 70 MPa PM sum / = E E E E E+09 Number of reversed cycles Curve 71 The Palmgren Miner Sum is directly related to the lifetime and the component with a SRF of 1.29 will have a safe life of 40 years For components with flat SN-curves the PM sum can be misleading as it is very sensitive to even small changes in stress levels.

19 High cycle vs. low cycle fatigue Low cycle fatigue basically deals with cyclic stresses for less than 10,000 cycles. A component tends to be able to withstand significantly more strain than predicted by the straight line in the log-log coordinate system. In principle the PM-sum can be used on this curved (Manson-Coffin) slope, but it is rarely done due in part to the risk of neglecting critical sequence effects.