1. Fatigue Analysis of Machines Using Strain Gauging, SCADA and FEA
Machine Group Meeting, Sydney
8 Feb 2016
Mark Biggs
Tatsuya Uno

2. Outline Parker Point Case Study
Links between AS load cases and SCADA data
Approach to fatigue loads and cycles
Correlation of FEA with strain gauge and SCADA
Fatigue findings
Tips

3. Parker Point Fatigue Life Case Study
Parker Point Case Study
Client wants an “end of life” fatigue study to determine when to replace the machine
Study inputs are:
Strain gauging ~30 members on machine for 3 months
Accelerometers on the boom tip, mast, portal and counterweight boom
SCADA data drive, position and throughput data
FEA model

4. Simple Fatigue Analysis – Slew Cuts / MT
SCADA: count times boom passes given slew angle, normalise by tonnes reclaimed

5. Simple Fatigue Analysis – Reference Point
Cycles in OEM fatigue calculations
~2-3 million
Now
~Life
Production records

6. AS4324. 1 Load Combination Table 3
AS Load Combination Table 3.7 – Related to Fatigue and Strength Design Loads, and SCADA Data
Loads used to calculate fatigue life of machine structure
Conveyor
Digging
Rare loads and accidents – strength design
e.g.
Flooded Belt
Collisions
Grounding
Cyclone
Operational loads used for strength design
Dead weight
Conveyor
Digging
Dynamics
Abnormal operation strength design
High digging loads
Slew bearing jacking
Many of the loads can be related to drive torque, or production rate which are data logged by the control system (SCADA system)

7. Fatigue Analysis - Method
AS Table 3.6 has fatigue modification factors
Basically want more accurate/applicable factors for the specific machine to use in fatigue calculations
Use strain gauging and SCADA to achieve this

8. What is SCADA Data BW drive power AS4324.1 Counterweight
Normal digging (U)
Abnormal digging (UU)
Counterweight
Belt weigher tonnes/hr:
AS4324.1
Conveyor live load (F)
Flooded belt (FF)
Luff cylinder forces:
AS4324.1
Dead load balance
Slew brg loads
Combined live loads
Collisions
Slew drive power and angle:
AS4324.1
Lateral digging (S)
Lateral collision (FS)
No. Slew cuts for fatigue
Bucketwheel boom

9. Example – Luff Cylinder Loads
The combined effect of the live loads can be seen in the luff cylinder pressures/loads.
We can also infer information about the dead load balance e.g. counterweight trim compared to design.

10. Example – Luff Cylinder Loads
Peak tensile luff cylinder loads approach the AS III/6 Flooded Belt and Blocked Chute Combination but are inside the strength envelope (red dotted line)
Loads exceed the fatigue design assumptions quite often (blue line)
Machine is slightly discharge boom heavy compared to design balance (green line)
Conclusions:
- Strength OK.
- May need CW trim
- Fatigue life may be shorter than design

11. Example – Luff Cylinder Loads
3D shows the E+V balance better (high frequency)

12. Parker Point RC2P Dataset – 3 months, 6MT

13. SCADA Load Analysis – Digging Load Compared to AS4324.1

14. SCADA – Equivalent Digging Fatigue Load (CU.U)
Allow for no load BW current e.g. 30A = zero digging load.
FLC on nameplate is the current at rated shaft power including power factor
Current is ~ linear with load above about 50% FLC. Below 50%FLC power factor drops.

15. Equivalent Loads and Stresses
Equivalent stress or load gives the same damage as a spectrum for a set number of cycles.
S-N slope is 3 up to 5 million cycles, then 5.
Therefore equivalent loads or stresses are weighted averages to power 3 or 5.

16. SCADA – Equivalent Digging Fatigue Load (CU.U)
Allow for no load BW current e.g. 30A = zero digging load.
FLC on nameplate is the current at rated shaft power including power factor
Current is ~ linear with load above about 50% FLC. Below 50%FLC power factor drops.

17. SCADA Load Analysis – Conveyor Load Compared to AS4324.1

18. SCADA – Equivalent Conveyor Fatigue Load (CF.F)

19. SCADA Load Analysis – Slew Torque Compared to AS4324.1
Same data on a log chart. Use if interested in peak loads

20. SCADA – Equivalent Lateral Digging Load (CS.S)

21. FEA Model Correlation
Show graph of FEA equivalent stress vs strain gauges
Show spreadsheet
Example strain gauge locations (red) – use to correlate model
Other gauges near critical welds – calculate damage directly
Accelerometers (yellow) – use to understand if dynamics are significant

22. Strain Gauging Analysis – Calculate Equivalent Stress
Matlab
Strain gauge data
Slew cuts / MT
Produce a single equivalent stress range value for each location that can be compared to FEA model for correlation.
Equivalent stress x cycles = same fatigue damage as strain gauge
~ cubic weighted average because S-N curve has slope of 3 below 5e6 cycles
Rainflow histogram

23. FEA Model Correlation – Suggested Approach
Want CF, CU and CS values  FEA Model  Stress at strain gauge locations = measurements
Try the SCADA C values as a starting point. Then vary factors to get best match between FEA and gauges using least cube error function
Show graph of FEA equivalent stress vs strain gauges
Show spreadsheet

24. FEA Model Correlation – Results
SCADA C values input into FEA model
Excel solver varying C values using least cubes
Strain gauge equivalent stress

25. Fatigue Load Conclusions
Digging load is higher than AS4324.1
Conveyor load and lateral digging load are lower
S is around 25% of U for Iron Ore

26. Fatigue Load Conclusions
Chute load is significant e.g. 17t, and is not in AS4324.1
Back wall of chute flat in bottom benches

27. Problem Areas

28. Problem Areas

29. Problem Areas

30. Problem Areas

31. Presentation of Results

32. Other Tips BS7608-2014 has guidance on hotspot weld categories
AS4100 is for nominal stresses
Permanent dynamics were found to be less than 0.02g in all directions
Similar to drive ramp rates e.g. g/100 – g/50
Much lower than AS values for fatigue
Use a lowpass filter on accelerometers that filters out high frequency noise before the datalogger
Record accelerometer data at about 10 times the highest frequency you think you will be interested in e.g. 30 Hz or so.
It is probably best to keep the analysis as simple as possible
Suggest using the model to find the worst welds, then strain gauge the weld toes and calculate the life directly rather than using FEA model correlation

33. Thank you!