Getting More From Abaqus by Tackling Noisy and Challenging Data

Size: px

Start display at page:

Download "Getting More From Abaqus by Tackling Noisy and Challenging Data"

Claud Garrison
6 years ago
Views:

Smart-Tools for Analyzing Noisy & Challenging Problems Getting More From Abaqus by Tackling Noisy and Challenging Data Snap-Fit Learn how to turn this mess

.5. 3 4 5 Strain (%) Advanced Material Testing Overview Noisy Data The Disease Bodie Technology Offerings The Cure Customizable Training, Kornucopia

Transient impact analyses via Abaqus/Explicit Including Shock Response Spectrum (SRS) assessments 4.

1 Smart-Tools for Analyzing Noisy & Challenging Problems Getting More From Abaqus by Tackling Noisy and Challenging Data Snap-Fit Learn how to turn this mess Failure Equipment Performance into success! Drop Testing. Impact & Buckling Copyright 4 -. Stress (MPa) Strain (%) Advanced Material Testing Overview Noisy Data The Disease Bodie Technology Offerings The Cure Customizable Training, Kornucopia Software, Expert Consulting Technology Demonstrations. Abaqus/Standard quasi-static rolling. Abaqus/Explicit quasi-static snap-fit 3. Transient impact analyses via Abaqus/Explicit Including Shock Response Spectrum (SRS) assessments 4. Improving FEA-Test correlations of impact events via highpass filtering 5. Preparing and cleaning physical test data for use in FEA modeling Bodie Technology Development Roadmap Upcoming Kornucopia V.6 release Future Kornucopia platforms: Plug-ins for Abaqus/Viewer, Excel, and Python Copyright 4 -.

4 Noisy Data The Disease Representing or manipulating oscillatory data with digital methods. Sufficient Sampling.

2 3 Noisy Data The Disease Representing or manipulating oscillatory data with digital methods. Sufficient Sampling.5 amplitude time (msec) Insufficient Sampling Causes Aliasing..5 amplitude time (msec) Copyright Noisy Data The Disease Representing or manipulating oscillatory data with digital methods. Sufficient Sampling.5 amplitude time (msec) Insufficient Sampling Causes Aliasing..5 amplitude time (msec) Copyright 4 -.

3 5. Specializes in solving complex problems in nonlinear mechanics by employing a proven mix of computational and testing knowledge in novel ways Numerical Analytical Creativity + Pragmatism Experimental Kornucopia Software Customizable Training info@bodietech.com Expert Consulting Smart-Tools for Analyzing Noisy and Challenging Problems Bodie Technology provides engineers with excellent software, training, and consulting resources to help analyze complex nonlinear mechanics problems, especially those involving problematic or noisy datasets. Steve Levine, Chief Strategy Officer, SIMULIA Copyright Technology Demonstrations. Abaqus/Standard quasi-static rolling. Abaqus/Explicit quasi-static snap-fit 3. Transient impact analyses via Abaqus/Explicit Including Shock Response Spectrum (SRS) assessments 4. Improving FEA-Test correlations of impact events via highpass filtering 5. Preparing and cleaning physical test data for use in FEA modeling Copyright 4 -.

7 Abaqus/Standard Quasi-Static Rolling Analyzing Paper Motion in Copiers Understand media transport, influence of roller deformation, frictional effects, roller viscoelasticity FEA model of rotating

Results from same model, just 4 different output requests. Normalized Moment.5.5 Output every 7 inc. (37 total points) FEA output After lowpass filtering Normalized Moment.5.5 Output every inc.

4 7 Abaqus/Standard Quasi-Static Rolling Analyzing Paper Motion in Copiers Understand media transport, influence of roller deformation, frictional effects, roller viscoelasticity FEA model of rotating roller Copyright Abaqus/Standard Quasi-Static Rolling Nonlinear quasi-static model predicts the torque (moment) required to rotate foam roller. Results from same model, just 4 different output requests. Normalized Moment.5.5 Output every 7 inc. (37 total points) FEA output After lowpass filtering Normalized Moment.5.5 Output every inc. (4 total points) FEA output After lowpass filtering Pseudo Time Pseudo Time Normalized Moment.5.5 Output every inc. (6 total points) FEA output After lowpass filtering Normalized Moment.5.5 Output every 8 inc. (5 total points) FEA output After lowpass filtering Pseudo Time Copyright 4 -. Pseudo Time

9 Abaqus/Standard Quasi-Static Rolling Nonlinear quasi-static model predicts the torque (moment) required to rotate

(5 total points) FEA output After lowpass filtering.5..4.6.8 Pseudo Time Aliasing errors!

Using proper DSP approach, steady state answer achievable with ½ the run-time! Copyright 4 -.

sensitivity to lower arm angle and influence of friction.

5 9 Abaqus/Standard Quasi-Static Rolling Nonlinear quasi-static model predicts the torque (moment) required to rotate foam roller. Correct model results using proper DSP techniques. Normalized Moment.5.5 Output every inc. (5 total points) FEA output After lowpass filtering Pseudo Time Aliasing errors! Solution noise caused by mesh discretization interacting with contact during rolling. Using proper DSP approach, steady state answer achievable with ½ the run-time! Copyright 4 -. Quasi-Static Snap-Fit Via Explicit Dynamics FEA Preliminary scouting analysis of a Snap-Fit design Understand sensitivity to lower arm angle and influence of friction. FEA via Abaqus/Explicit, Filtering via Kornucopia Simulate both insertion and retraction Implicit FEA will have difficulty with snap and retraction, so simulate with Explicit FEA. 3 angles: o, o +5º, o -5º Form of expected physical response during insertion force displacement Copyright 4 -.

6 Quasi-Static Snap-Fit Via Explicit Dynamics FEA Insertion Results Using Different Output Requests Every increment increments 8 increments (~35, incs) Unfiltered Force vs Disp (Insertion) Force vs Disp (Insertion) Force vs Disp (Insertion) 5 increments Force vs Disp (Insertion) Force Force Aliased Aliased Aliased 3 5 Force vs Disp (Insertion) 3 5 Lowpass filtered until smooth Force vs Disp (Insertion) 3 5 Force vs Disp (Insertion) 3 5 Force vs Disp (Insertion) Force.5 o +5º o o -5º.5 5 Force Copyright o o -5º o -5º o -5º Aliased Aliased Aliased.5 5 o +5º.5 o +5º o o.5 5 o +5º Quasi-Static Snap-Fit Via Explicit Dynamics FEA Retraction Results Using Different Output Requests Every increment increments 8 increments 5 increments (~35, incs) Unfiltered 5 Force vs Disp (Retraction) 5 Force vs Disp (Retraction) 5 Force vs Disp (Retraction) 5 Force vs Disp (Retraction) Aliased Aliased Aliased Force Force vs Disp (Retraction) Important part 5 4 Force vs Disp (Retraction) 5 4 Lowpass filtered until smooth o -5º Force vs Disp (Retraction) 5 4 Force vs Disp (Retraction) o -5º o +5º o +5º o +5º Force 4 o -5º o o +5º Copyright 4 -. o o -5º Aliased Aliased Aliased o o 6 5

3 Energy Quasi-Static Snap-Fit Via Explicit Dynamics FEA Energy Methods, Derivatives, and DSP to Improve Analysis Estimating Frictional Influence 5 4 3 Energy vs Disp (Insertion) External work

7 3 Energy Quasi-Static Snap-Fit Via Explicit Dynamics FEA Energy Methods, Derivatives, and DSP to Improve Analysis Estimating Frictional Influence Energy vs Disp (Insertion) External work Frictional energy Internal energy Kinetic energy "Friction Contribution Force" No filtering Case: o F fric U friction u Similar computations for other cases of o +5º and o -5º "Friction Contribution Force" With Filtering Filtered via bi-directional lowpass 6 th order butterworth, f c =.5 f s Copyright 4 -. Quasi-Static Snap-Fit Via Explicit Dynamics FEA Energy Methods, Derivatives, and DSP to Improve Analysis Estimating Frictional Influence.5 Orig. Simulation, COF =.3 4 o +5º Force o.5 o -5º Prediction equation F total COF new F total COF.3 F fric COF.3 COF new COF.3 Copyright 4 -.

Quasi-Static Snap-Fit Via Explicit Dynamics FEA Energy Methods, Derivatives, and DSP to Improve Analysis Estimating Frictional Influence.5 Orig.

5 (FEA model NOT re-run) 4 6 8 4 6 8 Prediction equation F total COF new F total COF.3 F fric COF.3 COF new COF.3 Validating predictions by running.

6 Severe Impact of Plastic Housing Experimental measurement Explicit Dynamics FEA Simulation Acceleration (kg) f s = 5 khz (original collection rate) raw data

8 Quasi-Static Snap-Fit Via Explicit Dynamics FEA Energy Methods, Derivatives, and DSP to Improve Analysis Estimating Frictional Influence.5 Orig. Simulation, COF =.3 Energy Derivative Predictions, COF = Force.5 o +5º o o -5º Force.5 o +5º o o -5º Three FAST predictions of COF change from.3 to.5 (FEA model NOT re-run) Prediction equation F total COF new F total COF.3 F fric COF.3 COF new COF.3 Validating predictions by running.5 actual full FEA models for all three o -5º angle cases with COF = Force Validation for COF=.5.5 o +5º o prediction actual FEA Copyright Severe Impact of Plastic Housing Experimental measurement Explicit Dynamics FEA Simulation Acceleration (kg) f s = 5 khz (original collection rate) raw data Cell phone 4 6 Time (msec) lens impact Why do engineers find completely different results starting with same data? Engineer A NO correlation Aliasing distorted analysis Engineer B Good Match Copyright 4 -.

Successful Analysis of Challenging Data: Accelerations and

assess impact severity from acceleration data.

assessing how a spectrum of Spring/Mass/Damper systems

questions such as Will a sensor or other component survive

9 Successful Analysis of Challenging Data: Accelerations and s Reusable, well documented Kornucopia analysis worksheet 7 Copyright 4 -. Shock Response Spectrum (SRS and PVSS) Methodology to assess impact severity from acceleration data. Applicable to FEA Physical tests Methodology based on assessing how a spectrum of Spring/Mass/Damper systems respond to transient acceleration input Useful to answer questions such as Will a sensor or other component survive the impact event? Examples include: Cameras Motion-sensing electronics IC chips Applicable to cell phones, smart-bombs, 8 Copyright 4 -.

SRS and PVSS Penetrator Example 9 Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock

5 msec Zoom from to 5 msec How to compare the following data to the specs?

10 SRS and PVSS Penetrator Example 9 Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec Zoom from to 5 msec How to compare the following data to the specs? Experiment and FEA Peak Accel = 7 G and 4, G Experiment and FEA data look totally different! How to relate. and.5 msec from spec to Test and FEA data? Copyright 4 -. SRS and PVSS Penetrator Example Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec First convert specs to Haversine shock pulses Easily done with Kornucopia functions Copyright 4 -.

SRS and PVSS Penetrator Example Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone:

11 SRS and PVSS Penetrator Example Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec Then compute SRS Abs Accel and PVSS Spectra on Haversine shock pulses Experimental data FEA data Copyright 4 -. SRS and PVSS Penetrator Example Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec Then compute SRS Abs Accel and PVSS Spectra on Haversine shock pulses Experimental data FEA data Copyright 4 -.

12 SRS and PVSS Penetrator Example 3 Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec Then compute SRS Abs Accel and PVSS Spectra on Haversine shock pulses Experimental data FEA data Need estimate of natural frequencies of device Copyright 4 -. SRS and PVSS Penetrator Example 4 Given Acceleration data from: Experiment Explicit Dynamics model Spec on absolute max shock for device to be placed near nose cone: (as stated by customer) 6, G for. msec, G for.5 msec Then compute SRS Abs Accel and PVSS Spectra on Haversine shock pulses Experimental data FEA data Need estimate of natural frequencies of device Copyright 4 -. SRS and PVSS New Capabilities in Upcoming Kornucopia V.6 Release Special Features to Properly Handle Unique Issues of Explicit Dynamics Data

Integrating Acceleration Data Another Big Challenge! Why does this measured acceleration produce a ridiculous displacement? 76 mm 57 mm Steel ball impacting aluminum plate.

13 Integrating Acceleration Data Another Big Challenge! Why does this measured acceleration produce a ridiculous displacement? 76 mm 57 mm Steel ball impacting aluminum plate.6 mm thick Answer: Low frequency distortion in accelerometer data Very common with measured data Accel (g) Disp (mm) Time (msec) 3 benchmark Integrated from accelerometer Time (msec) Implausible result! Copyright 4 -. Example Templates in Library Help Meet the Challenge! ALL Template Example Files Can Open LIVE in Mathcad Modify as desired by user Copyright 4 -.

14 Salvaging Data via Highpass Filtering Copyright 4 -. Salvaging Data via Highpass Filtering Copyright 4 -.

15 Salvaging Data via Highpass Filtering Laser Raw Filtered Copyright 4 -. Salvaging Data via Highpass Filtering Copyright 4 -.

3 Preparing & Cleaning Challenging Data via Kornucopia Processing Experimental Data File with beginning and ending markers

Trim analysis & clean worksheet datasets Read data file Unpack and separate datasets Grouping key: = specimen = info

3 Cleaning High-Rate Data With high-rate measurements, there is a greater likelihood that noise and oscillations appear in

16 3 Preparing & Cleaning Challenging Data via Kornucopia Processing Experimental Data File with beginning and ending markers for each dataset Challenging data to read Reusable, well (multiple datasets, header documented text and data, Kornucopia ) Trim analysis & clean worksheet datasets Read data file Unpack and separate datasets Grouping key: = specimen = info section = primary data Plot raw datasets Average datasets and plot final results Copyright Cleaning High-Rate Data With high-rate measurements, there is a greater likelihood that noise and oscillations appear in your data Often caused by excitation of mounting fixtures or testing frames. In many cases, the FEA analyst must deal with the data as is Copyright 4 -.

33 Cleaning High-Rate Data With high-rate measurements, there is a greater likelihood that noise and

In many cases, the FEA analyst must deal with the data as is Interactive Self-documenting Reusable

Enhancing Material Characterization from Raw Data Basic Steps for Elastic/Plastic Data A.

17 33 Cleaning High-Rate Data With high-rate measurements, there is a greater likelihood that noise and oscillations appear in your data Often caused by excitation of mounting fixtures or testing frames. In many cases, the FEA analyst must deal with the data as is Interactive Self-documenting Reusable Copyright 4 -. Enhancing Material Characterization from Raw Data Basic Steps for Elastic/Plastic Data A. Clean and average raw data B. Convert nominal stress/strain to true stress/strain C. Determine modulus D. Obtain yield stress vs plastic strain curve E. Output to ASCII file for FEA or other use Copyright 4 -.

18 35 Cleaning Hysteresis Data via Kornucopia Copyright 4 -. Key Benefits: Clear documentation of analysis Reusable & automated Saves times, Reduces errors Deployable to others 36 Free, Self-paced Kornucopia Training on BodieTech.com Copyright 4 -.

Analyzing Noisy Data via Filtering and DSP 3 or 4 Day Customizable Training Seminar Instructor: Ted Diehl, PhD 37 Training specially designed for FEA Users & Experimentalists Snap-Fit Learn how to

Motivation for Using DSP with Simulations and Experiments. DSP Fundamentals 3. DSP Using Various Software 4. Developing a DSP Strategy for a Given Problem 5. Using Energy to Enhance Analysis 6.

Working with Quasi-Static Models from Explicit FEA. Improving the Modeling of Failure. Creating Viable FEA Material Data from Noisy and Challenging Material Measurement Data Workshops.

19 Analyzing Noisy Data via Filtering and DSP 3 or 4 Day Customizable Training Seminar Instructor: Ted Diehl, PhD 37 Training specially designed for FEA Users & Experimentalists Snap-Fit Learn how to turn this mess Failure Equipment Performance into success! Drop Testing. Impact & Buckling Stress (MPa) Strain (%) Advanced Material Testing Copyright Course Content Lectures. Motivation for Using DSP with Simulations and Experiments. DSP Fundamentals 3. DSP Using Various Software 4. Developing a DSP Strategy for a Given Problem 5. Using Energy to Enhance Analysis 6. Working with Experimental Data and Validating Simulations/Experiments 7. Working with Transient-Dynamic Models 8. Shock Spectrum Analysis (SRS, PVSS) 9. Working with Quasi-Static Models from Explicit FEA. Improving the Modeling of Failure. Creating Viable FEA Material Data from Noisy and Challenging Material Measurement Data Workshops. Experiencing Common DSP-related Mistakes. Learning DSP Using Simple Signals 3. Transient Impact of a Structure 4. Transient Penetration - Comparing Models and an Experiment 5. Salvaging Shock Data via Highpass Filtering 6. Quasi-Static Cam Mechanism 7. Analyzing Failure and Crack Propagation 8. Computing Derivatives and Integrals From Noisy Data 9. Creating Elastic/Plastic Material Law From Slightly Noisy Experimental Data. Cleaning Problematic Hysteresis Data Copyright 4 -.

39 Smart-Tools for Analyzing Noisy & Challenging Problems Without Kornucopia I really doubt the quality and quantity of work would have been what it was.

Kelly Laughlin, PhD US Army, Picatinny Arsenal Analyzing large-caliber ballistics from both Explicit FEA and tests Copyright 4 -.

This has improved my analysis and helped me deliver more useful FEA results on recent projects - your course is the best I have taken in a long time.

Army, AMSAA 4 Smart-Tools for Analyzing Noisy & Challenging Problems We congratulate the Kornucopia development team for creating such an effective and

20 39 Smart-Tools for Analyzing Noisy & Challenging Problems Without Kornucopia I really doubt the quality and quantity of work would have been what it was. The on instruction was invaluable Lt. Col. Kelly Laughlin, PhD US Army, Picatinny Arsenal Analyzing large-caliber ballistics from both Explicit FEA and tests Copyright 4 -. Your DSP course taught techniques that helped me create a standardized approach for processing noisy Explicit Dynamics simulations. This has improved my analysis and helped me deliver more useful FEA results on recent projects - your course is the best I have taken in a long time. Michael J. Iacchei, Mech. Eng. U.S. Army, AMSAA 4 Smart-Tools for Analyzing Noisy & Challenging Problems We congratulate the Kornucopia development team for creating such an effective and simple-to-use toolkit for engineers worldwide. Arun Nair, Ph.D. Becton Dickinson & Co. "Kornucopia has helped greatly in breaking down difficult data sets from experiments, in an easier, automatic, and understandable way. This has helped us setup, scrutinize and understand the experiments faster, which is priceless It helps make the organization move faster David McCalib, Mfg. Engineer - Blount International Copyright 4 -.

Bodie Technology Roadmap Upcoming Kornucopia V.

SRS for direct comparison to SRS of FEA or physical accelerometer data Specifically designed to handle Explicit Dynamics simulation data as well as

worksheet Enhancements to file reading, averaging, integration, & derivatives New example templates SRS data analysis, Slice/Dice data, and more

21 Bodie Technology Roadmap Upcoming Kornucopia V.6 Release 4 Shock Response Spectrum SRS Absolute Acceleration, PVSS, and more Easy-to-use Pulse Generation functions Convert component shock specs to SRS for direct comparison to SRS of FEA or physical accelerometer data Specifically designed to handle Explicit Dynamics simulation data as well as experimental data Example templates demonstrating SRS technology usage Other Enhancements Run/Loop other Mathcad worksheets from a Kornucopia worksheet Enhancements to file reading, averaging, integration, & derivatives New example templates SRS data analysis, Slice/Dice data, and more Copyright 4 -. Bodie Technology Roadmap Future Kornucopia Platforms 4 Independent Kornucopia Engine Allows Kornucopia technology to be connected to a variety of software packages Initial Platform Targets Abaqus/Viewer Microsoft Excel Python Mathcad Copyright 4 -.

Expert in Nonlinear Mechanics with + Years Experience Experimental, computational, and theoretical approaches Led nonlinear mechanics efforts at Kodak, Motorola, and DuPont Engineering Tools Mathcad

22 43 Smart-Tools for Analyzing Noisy & Challenging Problems Analytical Numerical Creativity + Perspective Experimental 44 Instructor s Profile Ted Diehl, Ph.D. Expert in Nonlinear Mechanics with + Years Experience Experimental, computational, and theoretical approaches Led nonlinear mechanics efforts at Kodak, Motorola, and DuPont Engineering Tools Mathcad & Kornucopia, Abaqus nonlinear FEA, and experimental methods Engineering Success in Industry NASA spacecraft Paper motion in copiers Nonlinear nip mechanics Cell-phone impact Simulating fabrics Ballistic protection Peeling mechanics Flexible structures Nonlinear materials Created unique DSP algorithms Enhance analysis of noisy data from experiments & Explicit Dynamics FEA Developer of Kornucopia and President of Bodie Technology Inc. Smart-Tools for Analyzing Noisy & Challenging Problems Copyright 4 -.

Applications of DSP to Explicit Dynamic FEA simulations of elastically-dominated impact problems

167 Applications of DSP to Explicit Dynamic FEA simulations of elastically-dominated impact problems Ted Diehl a, Doug Carroll b and Ben Nagaraj c a Mechanical Technology Center, PCS, Motorola, 8000 W.