2015 HBM ncode Products User Group Meeting

Transcription

1 March 4-5, 2015 Livonia, MI (USA)

2 March 4-5, 2015 Livonia, MI (USA) GlyphWorks Accelerated Testing: Not Just for Developing PSD Based Shaker Profiles Presented By Phil Korth Technical Staff Engineer Harley-Davidson

3 March 4-5, 2015 Livonia, MI (USA) Intro to Harley-Davidson Harley-Davidson Motor Company produces cruiser and touring motorcycles and offers a complete line of Harley-Davidson motorcycle parts, accessories, riding gear and apparel, and general merchandise. For more information, visit Harley-Davidson's Web site at 3

4 March 4-5, 2015 Livonia, MI (USA) Challenge/project scope GlyphWorks Accelerated Test tool is recognized for developing PSD based vibration profiles but often overlooked for other uses Turning an invalid test into useful information Comparing alternate supplier test qualifications Duty Cycle/Time History Damage Editing Determination of a non-gaussian damage factor 4

5 March 4-5, 2015 Livonia, MI (USA) Why We Use GlyphWorks Accelerated Test Provides a documented repeatable process for developing vibration profiles Test profiles can be generated from time histories, PSD s or a mix of both Calculates fatigue based on accelerometer measurement Provides damage correlation Doesn t Require Component Response Measurement Test Profiles can be created prior to the existence of parts Test specification can be used for FEA based virtual test and real physical test (allows FEA to lab correlation) Provides a method to accelerate tests without extensive data editing Provides a method to verify that loads are not excessive One PSD represents the equivalent Damage of an entire Duty Cycle 5

6 March 4-5, 2015 Livonia, MI (USA) Why Other Uses Are Overlooked Accelerated Test Glyph is a canned process for producing a PSD or Sine Sweep People are familiar with what data goes into the glyph Time Histories PSD s People are familiar with what data comes out of the glyph PSD s Sine Sweeps People not as familiar with properties inside of the glyph Frequency Damage Spectrum (FDS) Shock Response Spectrum (SRS ) To most people the FDS & SRS are a data check With increased knowledge the FDS & SRS can be used for so much more For this presentation we will only focus our attention on the Frequency Damage Spectrum (FDS) 6

7 March 4-5, 2015 Livonia, MI (USA) Why Other Uses Are Overlooked FDS results are often unclear Not sure how to interpret data Is damage cumulative? Does more damage occur at low frequencies? How much damage difference exists?

8 March 4-5, 2015 Livonia, MI (USA) Divide & Conquer Method Better Method to Compare FDS Divide Results Produce Damage Factor Plot 8

9 March 4-5, 2015 Livonia, MI (USA) Improved FDS Interpretation When Comparing Data Sets Overlay FRF or Response PSD Compare Damage Factor at Damaging Frequency 9

10 March 4-5, 2015 Livonia, MI (USA) Case Study 1: Turning an Invalid Test Into Useful Information Test Scenario Previous Profile Used to Run Test Parts failing at 8 hr. duration need to last 30 hrs. Test profile suspected of being wrong Prove validity of test profile If wrong what do results mean? Available Information PSD data from ED Shaker Newly acquired valid test data from vehicle Process Create PSD test from new data Generate FDS from each PSD and compare 10

11 March 4-5, 2015 Livonia, MI (USA) Turning an Invalid Test Into Useful Information PSD Generation for Engine Excitation Step 1 - FDS & SRS Generation 11

12 March 4-5, 2015 Livonia, MI (USA) Turning an Invalid Test Into Useful Information PSD Generation for Engine Excitation Step 2 PSD Generation from FDS 12

13 March 4-5, 2015 Livonia, MI (USA) Turning an Invalid Test Into Useful Information From plot over test is obvious Was 8 hours enough at over test to validate the part? Minimum acceleration factor is 30, 8*30 = 240 hours PSD of response provides better confirmation of acceleration factor 13

14 March 4-5, 2015 Livonia, MI (USA) Case Study 2: Comparing Alternate Supplier Test Qualifications Test Scenario Vendor A part validated to motorcycle specification Proposed Vendor B part validated to automotive specification Does Vendor B part require validation? Available Information PSD data from Vendor A & B tests Process Generate FDS from each PSD and compare 14

15 March 4-5, 2015 Livonia, MI (USA) Comparing Alternate Supplier Test Qualifications PSD Vendor A vs PSD Vendor B Based on PSD overlay should be o.k. How much different is it? How much more damaging is it? 15

16 March 4-5, 2015 Livonia, MI (USA) Comparing Alternate Supplier Test Qualifications Convert PSD s to FDS s 16

17 March 4-5, 2015 Livonia, MI (USA) Comparing Alternate Supplier Test Qualifications FDS & Damage Factor FDS Worst case 2.5 X damage factor 17

18 March 4-5, 2015 Livonia, MI (USA) Case Study 3: Duty Cycle/Time History Damage Editing Test Scenario Need to create a time history based test profile Data collected on a vehicle for various courses making up a duty cycle Only accelerometer data was collected. What can be done to edit the data to create a reasonable test time? Available Information Input and response time history accelerometer data Process Generate FDS from inputs and overlay FRF s to determine damage factors. Eliminate non damaging time histories Use visual amplitude editing of response acceleration and verify by comparing FDS s 18

19 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing Calculate FDS 6 Road Courses Unedited 444 hours of test time Test Name Time Duration DC Repeats DC Time % Time RC E % RC E % RC E % RC E % RC E % RC E % 1.60E Hours 19

20 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing Does FDS give ability to edit out non-damaging road content? Road course 2 appears to have overall least amount of damage 20

21 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing FDS damage factor compared to RC2 Can you easily edit out non-damaging road content? 21

22 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing FRF vs Response PSD Why is the FRF better? 22

23 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing Pick damage at frequency of interest from FDS Keep most damaging events Eliminating non-damaging events reduced test time to 170 hours Event DC DC % Test Name Damage Repeats Damage Damage RC1 5.15E E % RC2 1.00E E % RC3 1.65E E % RC4 4.10E E % RC5 7.90E E % RC6 3.11E E % Total Damage 3.57E+05 Time DC Test Name Duration Repeats DC Time % Time RC E % RC E % RC E % RC E % RC E % RC E % 6.12E Hours 23

24 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing Utilize Visual Editing Limit Detection Tool for further editing of individual time histories 24

25 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing Compare FDS of edited vs non-edited time history Verify that damage content is acceptable Test time now reduced to 80.5 hours Test Name Time Duration DC Repeats DC Time % Time RC E % RC E % RC E % RC E % RC E % RC E % 2.90E Hours 25

26 March 4-5, 2015 Livonia, MI (USA) Duty Cycle/Time History Damage Editing How Does The Method Compare To Strain Based Damage? Strain Based Damage Test Name Event Damage DC Repeats DC Damage % Damage RC1 3.05E E % RC2 4.95E E % RC3 1.98E E % RC4 8.99E E % RC5 5.44E E % RC6 2.67E E % Accelerated Test Based Damage Test Name Event Damage DC Repeats DC Damage % Damage RC1 5.15E E % RC2 1.00E E % RC3 1.65E E % RC4 4.10E E % RC5 7.90E E % RC6 3.11E E % 26

27 March 4-5, 2015 Livonia, MI (USA) Case Study 4: Determination Of A Non-Gaussian Damage Factor Test Scenario Using an Electro-Dynamic shaker to reproduce a non-gaussian based vibration profile GlyphWorks Accelerated Test produces equivalent Gaussian based PSD test profile How much more damaging is the test if its run in a non-gaussian mode with different Kurtosis? Available Information Time History Data from Shaker GlyphWorks Accelerated Test Process Collect a minimum of 3 minutes of accelerometer time history data from the Gaussian & non-gaussian profiles on the shaker Compare FDS s to determine increased damage factor 27

28 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor Compare 4 profiles 50 hour Gaussian 100 hour Gaussian 100 hour non-gaussian Kurtosis hour non Gaussian Kurtosis

29 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor Calculate & Compare FDS What can you determine from this plot? 29

30 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor Divide FDS s by 100 hour Gaussian FDS Compare Scale Factors Determine Damage Factor 30

31 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor How does strain response compare? 31

32 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor Calculate PSD of Strain to understand frequency content of damage 32

33 March 4-5, 2015 Livonia, MI (USA) Determination Of A Non-Gaussian Damage Factor Calculate Strain Damage Compare Strain Damage with FDS Damage Factor at Damaging Frequencies TestName Damage Damage Factor dc_50hr_gaussian_mi 2.84E dc_100hr_gaussian_mi 1.26E dc_100hr_kurt4pt64_mi 1.62E dc_100hr_kurt8pt48_mi 2.06E

34 March 4-5, 2015 Livonia, MI (USA) FDS Non-Gaussian (Kurtosis) Damage Factor Conclusions FDS provides a valid method to observe and calculate increased damage due to Kurtosis Knowledge of resonance frequency and bandwidth can provide a very accurate prediction of damage factor The calculated damage factor can be used to adjust original test time Though not demonstrated in this presentation it has been observed that not all vendors Kurtosis based vibration controllers produce the same damage factor across the frequency band Statistically they all produce the same Kurtosis when analyzing the time history Variations exists in damage factor and uniformity across the bandwidth The damage factor variation makes it difficult to create a canned process based on Kurtosis that can be used across all vibration controllers 34

35 March 4-5, 2015 Livonia, MI (USA) Summary GlyphWorks Accelerated Test tools were successfully utilized for the following: Turning an invalid test into useful information Comparing alternate supplier test qualifications Duty Cycle/Time History Damage Editing Determination of a non-gaussian damage factor A picture may be worth a thousand words, but sometimes to be worthwhile you have to look at the right picture Remember when comparing FDS s to divide and conquer 35

36 March 4-5, 2015 Livonia, MI (USA) Thank you! Phil Korth Harley-Davidson