River Rapids Conference Proceedings of the Society for Experimental Mechanics Series Chad Walber Patrick Walter Steve Seidlitz Proceedings of the 38th IMAC, A Conference and Exposition on Structural Dynamics 2020 River Publishers Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting & Dynamic Environments Testing, Volume 7
Conference Proceedings of the Society for Experimental Mechanics Series Series Editor Kristin B. Zimmerman, Ph.D. Society for Experimental Mechanics, Inc., Bethel, CT, USA
The Conference Proceedings of the Society for Experimental Mechanics Series presents early findings and case studies from a wide range of fundamental and applied work across the broad range of fields that comprise Experimental Mechanics. Series volumes follow the principle tracks or focus topics featured in each of the Society’s two annual conferences: IMAC, A Conference and Exposition on Structural Dynamics, and the Society’s Annual Conference & Exposition and will address critical areas of interest to researchers and design engineers working in all areas of Structural Dynamics, Solid Mechanics and Materials Research.
River Publishers Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting & Dynamic Environments Testing, Volume 7 Proceedings of the 38th IMAC, A Conference and Exposition on Structural Dynamics 2020 Chad Walber • Patrick Walter • Steve Seidlitz Editors
Published, sold and distributed by: River Publishers Broagervej 10 9260 Gistrup Denmark www.riverpublishers.com ISBN 978-87-4380-002-6 (eBook) Conference Proceedings of the Society for Experimental Mechanics An imprint of River Publishers © The Society for Experimental Mechanics, Inc. 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, or reproduction in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Preface Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting &Dynamic Environments Testing represent one of eight volumes of technical papers presented at the 38th IMAC, A Conference and Exposition on Structural Dynamics, organized by the Society for Experimental Mechanics, and held in Houston, Texas, February 10–13, 2020. The full proceedings also include volumes on Nonlinear Structures and Systems; Dynamic Substructures; Model Validation and Uncertainty Quantification; Dynamic Substructures; Special Topics in Structural Dynamics & Experimental Techniques; Rotating Machinery, Optical Methods & Scanning LDV Methods; and Topics in Modal Analysis &Testing. Each collection presents early findings from experimental and computational investigations on an important area within sensors and instrumentation and other structural dynamics areas. Topics represent papers on calibration, smart sensors, practical issues improving energy harvesting measurements, shock calibration and shock environment synthesis, and applications for aircraft/aerospace structures. The organizers would like to thank the authors, presenters, session organizers, and session chairs for their participation in this track. Depew, NY, USA Chad Walber Fort Worth, TX, USA Patrick Walter Minneapolis, MN, USA Steve Seidlitz v
Contents 1 Orion MPCV E-STA Nonlinear Dynamics Uncertainty Factors ................................................... 1 Matt Griebel, Adam Johnson, Brent Erickson, Andrew Doan, Chris Flanigan, Jesse Wilson, Paul Bremner, Joel Sills, and Erica Bruno 2 The Vibration and Acoustic Effects of Prop Design and Unbalance on Small Unmanned Aircraft............ 9 William H. Semke, Djedje-Kossu Zahui, and Joseph Schwalb 3 A Deformed Geometry Synthesis Technique for Determining Stacking and Cryogenically Induced Preloads for the Space Launch System................................................................................ 17 Joel Sills, Arya Majed, and Edwin Henkel 4 End-to-End Assessment of Artemis-1 Development Flight Instrumentation ..................................... 27 Andrew Doan, Adam Johnson, Tony Loogman, Paul Bremner, Joel Sills, and Erica Bruno 5 Space Launch System Mobile Launcher Modal Pretest Analysis.................................................. 35 James C. Akers and Joel Sills 6 Feasibility Study to Extract Artemis-1 Fixed Base Modes While Mounted on a Dynamically Active Mobile Launch Platform................................................................................................ 49 Kevin L. Napolitano 7 Challenges to Develop and Design Ultra-high Temperature Piezoelectric Accelerometers ..................... 57 Chang Shu, Neill Ovenden, Sina Saremi-Yarahmadi, and Bala Vaidhyanathan 8 Application of Quasi-Static Modal Analysis to an Orion Multi-Purpose Crew Vehicle Test.................... 65 Matthew S. Allen, Joe Schoneman, Wesley Scott, and Joel Sills 9 Using BB-gun or Acoustic Excitation to Find High Frequency Modes in Additively Manufactured Parts ... 77 Aimee Allen, Kevin Johnson, Jason R. Blough, Andrew Barnard, Troy Hartwig, Ben Brown, David Soine, Tristan Cullom, Douglas Bristow, Robert Landers, and Edward Kinzel 10 Parametric Analysis and Voltage Generation Performance of a Multi-directional MDOF Piezoelastic Vibration Energy Harvester............................................................................................ 85 Paulo S. Varoto, Elvio Bonisoli, and Domenico Lisitano 11 Are We Nearly There Yet? Progress Towards the Fusion of Test and Analysis for Aerospace Structural Dynamics .................................................................................................... 97 David Ewins 12 Feasibility Study of SDAS Instrumentation’s Ability to Identify Mobile Launcher (ML)/Crawler-Transporter (CT) Modes During Rollout Operations ............................................. 101 James P. Winkel, James C. Akers, and Erica Bruno 13 The Integrated Modal Test-Analysis Process (2020 Challenges) ................................................... 123 Robert N. Coppolino vii
viii Contents 14 Roadmap for a Highly Improved Modal Test Process............................................................... 161 Robert N. Coppolino 15 Using Low-Cost “Garage Band” Recording Technology for Acquiring High Resolution High-Speed Data.. 175 Randall Wetherington, Gregory Sheets, Tom Karnowski, Ryan Kerekes, Michael Vann, Michael Moore, and Eva Freer 16 Hybrid Slab Systems in High-rises for More Sustainable Design.................................................. 185 Katherine Berger, Samuel Benzoni, Zhaoshuo Jiang, Wenshen Pong, Juan Caicedo, David Shook, and Christopher Horiuchi 17 Ground Vibration Testing of the World’s Longest Wingspan Aircraft—Stratolaunch.......................... 193 Douglas J. Osterholt and Timothy Kelly 18 Using Recorded Data to Improve SRS Test Development........................................................... 205 Joel Minderhoud 19 Distributed Acquisition and Processing Network for Experimental Vibration Testing of Aero-Engine Structures................................................................................................................. 209 Michal J. Szydlowski, Christoph W. Schwingshackl, and Andrew Rix 20 Modal Test of the NASA Mobile Launcher at Kennedy Space Center ............................................ 213 Eric C. Stasiunas, Russel A. Parks, Brendan D. Sontag, and Dana E. Chandler 21 Using Deep-Learning Approach to Detect Anomalous Vibrations of Press Working Machine................. 229 Kazuya Inagaki, Satoru Hayamizu, and Satoshi Tamura 22 DAQ Evaluation and Specifications for Pyroshock Testing......................................................... 233 Erica M. Jacobson, Jason R. Blough, James P. DeClerck, Charles D. Van Karsen, and David Soine 23 Optimal Replicator Dynamic Controller via Load Balancing and Neural Dynamics for Semi-Active Vibration Control of Isolated Highway Bridge Structures ......................................................... 241 Sajad Javadinasab Hormozabad and Mariantonieta Gutierrez Soto 24 Forcing Function Estimation for Space System Rollout............................................................. 245 George James, Robert Grady, Matt Allen, and Erica Bruno
Chapter 1 Orion MPCV E-STA Nonlinear Dynamics Uncertainty Factors Matt Griebel, Adam Johnson, Brent Erickson, Andrew Doan, Chris Flanigan, Jesse Wilson, Paul Bremner, Joel Sills, and Erica Bruno Abstract NASA vibration testing of the European Service Module (ESM) Structural Test Article (E-STA) for the Orion Multi-Purpose Crew Vehicle (MPCV) program demonstrated significant nonlinear behaviors and response deviation from pre-test finite element analysis (FEA). A linear FEA correlation effort, previously performed in 2017, resulted in the creation of two finite element models (FEM) – one correlated to high-load level swept sine responses and one correlated to low-load level swept sine responses. Additional work was required to quantify the uncertainty introduced when applying these linear models to non-sinusoidal flight load cases. To do this, an additional nonlinear dynamics model was developed and correlated with sine sweep test responses for low load level and high load level load cases. Results showed that, when the appropriate linearized model was selected for each specific Coupled Loads Analysis (CLA) loading type (i.e. Liftoff, Transonic, etc. . . ), the linearized models closely matched predicted nonlinear responses with modest error (uncertainty factor). Results of this investigation have established a physics-based nonlinear dynamics approach that uses empirical test data to establish the credibility assumption for usage of linear FEM(s) in CLA. It is anticipated that the methodology employed can be extended for usage in correlation and flight loads analysis of subsequent spacecraft with major joint nonlinearities. Keywords Nonlinear Dynamics · Uncertainty · Orion MPCV · Coupled Loads Analysis (CLA) 1.1 Introduction MPCV E-STA sine vibration testing was performed at various flight-like load levels. Significant nonlinear behaviors were observed as well as response deviation from pre-test FEA. These nonlinearities were observable as significant frequency and damping shifts between the low-load level test cases (20% flight level loads) and the high-load level test cases (100% flight level loads). Sample sine sweep frequency response function (FRF) results are shown in Fig. 1.3. In addition, evidence of joint slipping onset was apparent in the test data FRFs, exhibited as nonlinear inflections in the ramp up to resonance. A previous investigation by Quartus and NESC identified and limited the primary nonlinearities to three major interface joints. A linear FEA correlation effort resulted in the creation of two linear FEMs intended for use in a CLA study. The purpose of this study was to inform the use of correlated linear FEM(s) in a comprehensive linear CLA of a truly nonlinear system. One linear FEM was correlated to high level E-STA load cases (HLL FEM) and the other was correlated to low level E-STA load cases (LLL FEM). These two linear FEMs effectively represented linearizations of the true E-STA response about two specific load levels. Figure 1.1 illustrates how these linearizations might not necessarily predict accurate responses of a nonlinear system at load levels about which they were not correlated. Because of this, a single nonlinear FEM was created to help quantify the error (or uncertainty) introduced from utilizing linearized models in a linear CLA. This nonlinear “truth” model was developed from the previously correlated linearized models and correlated to the same E-STA sine sweep test M. Griebel ( ) · A. Johnson · B. Erickson · A. Doan · C. Flanigan · J. Wilson Quartus Engineering Incorporated, San Diego, CA, USA e-mail: matt.griebel@quartus.com P. Bremner AeroHydroPLUS, Del Mar, CA, USA J. Sills NASA Engineering and Safety Center (NESC), Houston, TX, USA E. Bruno Analytical Mechanics Associates, Inc., Hampton, VA, USA © The Society for Experimental Mechanics, Inc. 2021 C. Walber et al. (eds.), Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting & Dynamic Environments Testing, Volume 7, Conference Proceedings of the Society for Experimental Mechanics Series, https://doi.org/10.1007/978-3-030-47713-4_1 1
2 M. Griebel et al. Fig. 1.1 Illustration of Linearization Uncertainty – A single nonlinear “truth” model can help quantify uncertainty introduced when the correlated HLL and LLL linear FEMs are used under a loading about which they were not correlated. (NOTE: Graphics generated in MATLAB 2018/2019 and MS Office 2016) Fig. 1.2 CLA Study Outline – CLA response was compared between the linearized models and the nonlinear model resulting in the selection of a best-fit linearized model and corresponding LUF for each CLA loading type. (NOTE: Graphics generated in MATLAB 2018/2019 and MS Office 2016) results. It included detailed representations of the nonlinear joints – primarily through the inclusion of Coulomb friction stick-slip conditions. The results of both the linear and nonlinear model correlation efforts are briefly summarized in this report; additional details of these correlation efforts can be found in their corresponding references [1, 2]. Figure 1.2 outlines the steps taken in a CLA study designed to assess the performance of the HLL and LLL FEM for a set of flight-like load cases. The study was performed using a subset of representative CLA load cases and compared the response of the HLL and LLL linearized models to the response of the nonlinear model. This subset of CLA load cases represented three distinct phases of flight: • 5 Liftoff cases • 6 Transonic cases • 5 Max Acceleration cases These CLA load cases (6 DOF acceleration time histories at the MPCV interface) were applied to all three models as a time domain base excitation (fixed base analysis). The HLL and LLL response time histories were then compared to the nonlinear response time histories and a best-fit linear FEM was selected to represent each CLA loading type. Linearization Uncertainty Factors (LUF) were then computed for each CLA loading type using the selected best-fit linear FEMs. 1.2 Analysis A previous investigation performed by Quartus and NESC resulted in the creation of three correlated FEMs for use in a CLA study: • LLL linearized model: Linear NASTRAN FEM correlated to E-STA response at 20% flight load levels • HLL linearized model: Linear NASTRAN FEM correlated to E-STA response at 100% flight load levels • Nonlinear “Truth” FEM: Single nonlinear Abaqus model representing E-STA response at all load levels The results of both the linear and nonlinear model correlation efforts are briefly summarized in this report; additional details of these correlation efforts can be found in their corresponding references [1, 2]. The nonlinear Abaqus model is a combination of nonlinear Abaqus joints and a Hurty-Craig/Bampton (HCB) reduced model of the majority of the E-STA structure (reduction performed on NASTRAN model and HCB matrices converted to Abaqus format).
1 Orion MPCV E-STA Nonlinear Dynamics Uncertainty Factors 3 One of the primary metrics used in the model correlation process was comparing E-STA FEM and test FRFs based on the E-STA sine test drive inputs. The linearized model FRFs were directly computed in the frequency domain using NASTRAN (sol 111). Figure 1.3 shows a comparison between the E-STA measured sine test FRFs and the linearized model FRFs at the Launch Abort System (LAS) simulator (considered a representative response location for gauging primary mode responses). Results are shown for the high-load level sine sweep (100% flight loading shown in red) and low-load level sine sweep (20% flight loading shown in blue) load cases. In the case of the linear correlation, the high load level sine sweep was applied to the HLL FEM while the low load level sine sweep was applied to the LLL FEM. The two linearized models approximated the frequency and damping shift observed during the E-STA test, however effects of slipping joints observable in the test responses was not captured. The nonlinear model FRFs were computed by post-processing time histories from a nonlinear direct transient simulation of the E-STA high level and low level sine sweeps. These simulations were performed using Abaqus implicit dynamics. Figure 1.4 shows the same comparison with E-STA test FRFs using only a single correlated nonlinear model. In this case, both the high-load level and the low-load level sine sweep transient inputs were applied to the same correlated nonlinear model in direct transient simulations in Abaqus. The use of this single nonlinear model showed improved correlation over the use of the two separate linearized models and captured the slipping onsets observed in the measured test FRFs. Because of this, the nonlinear model was considered a “truth” model and acted as a surrogate for the E-STA true nonlinear response. A CLA response study was performed using the three correlated models discussed above. The purpose of this study was to determine a best-fit linearized model for each CLA loading type (i.e. Liftoff, Transonic, Max Acceleration) by comparing the linearized model responses to the response of the nonlinear “truth” model. Figures 1.5 and 1.6 show the ensemble of CLA response locations used in this study. Fig. 1.3 Linear FEM Correlation Results (LAS Simulator FRF) – Two linear FEMs were able to approximate the frequency and damping shifts observed in test; however effects of joint slipping (evident in some test FRF peaks) could not be captured using two correlated linear FEMs. (NOTE: Graphics generated in MATLAB 2018/2019 and MS Office 2016) Fig. 1.4 Nonlinear FEM Correlation Results (LAS Simulator FRF) – A single correlated nonlinear FEM was able to capture frequency and damping shifts as well as effects of joint slipping observed in test. (NOTE: Graphics generated in MATLAB 2018/2019 and MS Office 2016)
4 M. Griebel et al. Fig. 1.5 Nodal Response Locations – CLA transient acceleration response was recovered in all translation DOFs at 27 nodal locations for the HLL, LLL, and nonlinear FEM. (NOTE: Graphics generated in Siemens Femap 11.3) Fig. 1.6 Strain Response Locations – CLA transient strain response was recovered at 4 elemental locations for the HLL, LLL, and nonlinear FEM. (NOTE: Graphics generated in Siemens Femap 11.3) Due to the large number of response outputs (i.e. locations, degrees of freedom [DOF], and load cases), Pearson Correlation Coefficients were used to compare the response of the HLL and LLL model to the nonlinear model for each output. A Pearson Correlation Coefficient is a time-integrated measure of the linear dependence of two variables (Eq. 1.1). When this is applied to two response time histories, it provides a high-level frequency and phase comparison. These are the most important metrics when determining which linearized model best represents the nonlinear response for any CLA loading type. It should be noted that the Pearson Correlation Coefficient is not sensitive to magnitude discrepancies between two time histories. These are accounted for in a subsequent computation of uncertainty factors after the best-fit linear models have been selected for each CLA loading type. ρ(A,B) = 1 N−1 N i=1 Ai −uA σA Bi −uB σB (1.1) Pearson Correlation Coefficient – N: Sample Size, A&B: Random Variables, σ: sample standard deviation, μ: sample mean Pearson correlation coefficients were calculated between each linearized model and the nonlinear model for all response locations over all load cases and grouped by their respective CLA loading types. Figure 1.7 shows histogram distributions of Pearson Correlation Coefficients representing the LLL and HLL acceleration response comparison with the nonlinear “truth” response. In addition, a representative sample transient response comparison is shown at a single node on the E-STA LAS simulator for each CLA loading type. For Liftoff and Transonic type load cases, a strong correlation was observed between the HLL linearized model and the nonlinear model. On the other hand, Max Acceleration load cases were a lower-level load case; therefore a strong correlation was observed between the LLL linearized model and the nonlinear model. The same trend was observed for strain response comparisons, but is not shown explicitly in this report. These trends informed the selection of the best-fit linearized models for each CLA loading type. The best-fit linearized model for Liftoff and Transonic CLA cases was the HLL FEM, while the best-fit linearized model for Max Acceleration CLA cases was the LLL FEM.
1 Orion MPCV E-STA Nonlinear Dynamics Uncertainty Factors 5 Fig. 1.7 Pearson Correlation Coefficients for Acceleration Response – Pearson Correlation Coefficients were computed between each linear model (LLL and HLL) and the nonlinear FEM and represented by histogram distributions. This was done to determine the best fit linear model for each CLA loading type. (NOTE: Graphics generated in MATLAB 2018/2019 and MS Office 2016)
6 M. Griebel et al. Following the response comparison, Linearization Uncertainty Factors (LUF) were computed to represent the uncertainty that would be introduced using the selected best linear models in a comprehensive CLA. LUFs were computed for each CLA response location as a ratio of response magnitude between the nonlinear “truth” model and the selected best-fit linearized model (Eq. 1.2). Table 1.1 shows the selected response magnitude metrics for each CLA loading type. Because Liftoff and Max Acceleration load cases were non-stationary, peak-value was selected as the magnitude metric. Alternatively, since Transonic load cases were primarily random buffet loadings, root-mean-squared (RMS) was selected as the magnitude metric for these cases. LUFn = RNLin RLin (1.2) Linearization Uncertainty Factor – RLin: Linear response magnitude, RNLin: Nonlinear response magnitude By the definition in Eq. 1.2, the maximum LUF for any loading type was of primary interest as it represented instances in which the selected linear model was under-predicting the true nonlinear response. However, since this study was performed on only a subset of CLA load cases, a measure of statistical significance was desired to help characterize how well the computed max uncertainty factors can generalize a comprehensive CLA consisting of hundreds or thousands of load cases. Figure 1.8 shows Probability Density Functions (PDF) computed using LUFs from all response locations (including acceleration and strain). Statistical significance is typically given as a probability level (β) and confidence level (γ) associated with a particular maximum value. In this case, it was observed that the LUF distributions were irregular (not Normal Gaussian distributions) rendering the standard Normal Tolerance methodology unusable within the current analysis parameters [3]. Investigation revealed that the irregularities observed were primarily due to the use of peak value as a magnitude metric for non-stationary load cases. For example, the Liftoff load cases consist of two distinct events – one in which the HLL model over-predicts the true nonlinear response, and another in which the HLL model under-predicts true nonlinear response. This results in the bi-modal probability distribution observed in the Fig. 1.8 Liftoff PDF. A non-parametric, distribution-free methodology exists that does not assume any distribution of the sample [3]. The Distribution-Free Tolerance Limit (DFL) is defined as a value which exceeds all values for at least βfraction of a sample with a confidence γ defined by Eq. 1.3. The primary limitation of the DFL methodology is that it does not permit independent Table 1.1 Determination of Response Magnitude Metric (R) for Each CLA Loading Type Type of Loading Magnitude Metric RSS DOFs Liftoff Low Frequency Peak Value All Transonic Buffet +Thrust Osc. RMS YZLateral Max Acceleration Thrust Osc. Peak Value All Fig. 1.8 Probability Density Functions of LUFs Computed Using Best-fit Linear Models – Probability distributions were determined to be irregular (not Normal Gaussian distributions). (NOTE: Graphics generated in MS Office 2016)
1 Orion MPCV E-STA Nonlinear Dynamics Uncertainty Factors 7 Table 1.2 Max Uncertainty Factors for Each CLA Loading Type Liftoff Transonic Max Acceleration Max Uncertainty Factor 1.24 1.01 0.99 Selected Linear Model HighLoad Level HighLoad Level Low Load Level Probability 98% 98% 98% Confidence 96% 98% 96% selection of γ and β. Instead, β < βmax is chosen to achieve desired confidence (where βmax is the fractional portion of the data enveloped by the max value). Since the LUF PDFs in Fig. 1.8 were extremely irregular (particularly the Liftoff distribution), further development of this methodology should be done to arrive at more familiar LUF distributions. This would allow for a better estimation of statistical significance. γ =1−βN (1.3) Confidence of Distribution-free Tolerance Limit (DFL) – Confidence associated with a DFL that envelopes β fraction of a sample (sample size =N) Table 1.2 shows the max LUFs computed for each CLA loading type and an associated statistical probability and confidence (from the DFL methodology outlined above). Using the appropriate best-fit linear model for each loading type resulted in relatively modest maximum LUFs. It should be noted that inherent loads uncertainty as well as model uncertainty was not accounted for in this analysis. These additional sources of uncertainty would be combined with the LUFs to arrive at total CLA uncertainty. 1.3 Conclusion NASA vibration testing of the Orion E-STA demonstrated significant nonlinear behaviors and response deviation from pretest analysis. Because of this, two correlated linear FEMs were developed that acted as linearizations of E-STA response about low-load level (LLL FEM) and high-load level (HLL FEM). Additionally, a single correlated nonlinear “truth” model was developed to act as a surrogate for the E-STA true nonlinear response. A CLA response comparison study revealed that the linearized models can approximate the true nonlinear response and informed the selection of a best-fit linearized model. To account for the uncertainty introduced from performing a comprehensive linear CLA on a nonlinear system, Linearization Uncertainty Factors (LUF) were computed for each loading type. Using the best-fit linear FEM for each CLA loading type, it was found that modest LUFs could be applied to a comprehensive linear CLA. Acknowledgements Special thanks to Joel Sills at NASA’s Johnson Space Center (JSC) for his guidance and continued support of this assessment. References 1. Doan, A., Erickson, B., Owen, T.: Orion MPCV E-STA Structural Dynamics Correlation for NASA NESC. El Segundo California, Spacecraft and Launch Vehicle (SCLV) Dynamic Environments Workshop (2018) 2. Griebel, M., Johnson, A., Erickson, B., Doan, A., Flanigan, C., Wilson, J., Bremner, P., Sills, J., Bruno, E.: Orion MPCV E-STA Nonlinear Correlation for NESC. El Segundo California, Spacecraft and Launch Vehicle (SCLV) Dynamic Environments Workshop (2019) 3. National Aeronautics and Space Administration: Computation of maximum expected environment. In: NASA Handbook 7005. Pasadena, CA: Jet Propulsion Laboratory, National Aeronautics and Space Administration (2017)
Chapter 2 The Vibration and Acoustic Effects of Prop Design and Unbalance on Small Unmanned Aircraft William H. Semke, Djedje-Kossu Zahui, and Joseph Schwalb Abstract The vibration and acoustic effects due to prop design, damage, and unbalance on a popular small unmanned aircraft systems (sUAS) is presented. The use of sUAS or drones is becoming ever more popular for hobbyists, as well as in commercial and military operations, and the influence of props on the vibration and acoustic environment are of interest. Many types of props are available for use on sUAS that promise extended run times, increased performance, and quieter operation. While many systems promise these results, few studies have been conducted to measure and evaluate their true performance. Therefore, this review provides data obtained by experimentally measuring the vibration levels onboard the host aircraft as well as the acoustic levels produced. The data is analyzed to gain further understanding of the vibration and acoustic properties and make predictions on prop performance. The fundamental frequencies of the aircraft are found along with the acoustic signature. These two outcomes are compared and studied to find the correlation between them. The analysis will utilize an airframes that is commonly used in the UAS community along with frequently used props that are balanced and unbalanced and with and without damage. The data is obtained from aircraft fully powered and airborne in a hovering or level flight configuration. This study provides sUAS operators the information required for choosing the most effective prop design to effectively reduce vibration and acoustic sound levels. Keywords Unmanned aircraft · Remote sensing · Vibration · Sound levels · Acoustic testing 2.1 Introduction Among the growing userbase of small Unmanned Air Systems (sUAS), the academic community takes a special interest in investigating their mechanical characteristics. Our focus exists in the mechanical characteristics and acoustic profile of the sUAS as a result to different styles of props and prop damage. There are multitudes of sellers carrying different types of propellers for the DJI Phantom 4 Pro, but we focus on select venders of propellers of two types: Standard and Low Noise. For both types of props, we examine the affect imbalance has on the vibration of the vehicle, as well as the change in acoustics introduced by the imbalance. The magnitude and spectral response of the system vibration and acoustic levels are presented. Many types of props are available for use on sUAS that promise extended run times, increased performance, and quieter operation. A maritime engineering reference book examines different types of propellers and provides performance metrics and much of the analysis done by Carlson [1] assumes that the propeller is in an undamaged state. This tends to be a common assumption between similar works. The work presented here does not offer a fault-analysis of a damaged propeller, but the vibration characteristics and acoustic profile changes causal to a damaged propeller, in a controlled manner. The need to reduce or eliminate unwanted vibration in imaging systems is crucial in many applications of UAS. Researchers have carried out investigations into vibration isolation and active control for many years. Several studies [2–4] provide numerous resources into the basic principles. Given the limited range and payload capacity of these vehicles, vibration control is often carried out through passive means with vibration isolation and low noise and vibration blade design. Many researchers have carried out research into vibration isolation and active control over a long period of time and the fundamentals are explained in vibrations texts [5–7]. W. H. Semke ( ) · D.-K. Zahui Department of Mechanical Engineering, University of North Dakota, Grand Forks, ND, USA e-mail: william.semke@und.edu J. Schwalb Department of Aerospace Sciences, University of North Dakota, Grand Forks, ND, USA © The Society for Experimental Mechanics, Inc. 2021 C. Walber et al. (eds.), Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting & Dynamic Environments Testing, Volume 7, Conference Proceedings of the Society for Experimental Mechanics Series, https://doi.org/10.1007/978-3-030-47713-4_2 9
10 W. H. Semke et al. Part of the inspiration for this paper stems from the want to “cut the cord” that we use to measure mechanical vibration of the vehicle. Specifically, we wanted to see if good correlation between mechanical vibration and acoustic profile exists. Should a strong correlation exist, it would allow vibration analysis with a microphone as opposed to an accelerometer attached to a cable. In other works, similar ideas have been explored for different reasons. Sas et al. [8] introduces experimental and numerical analysis of a multirotor chassis, with the goal of isolating areas of lesser vibrations. Their goal was to identify places of lesser vibrations so they could mount the more sensitive electronics in those areas. Kloet et al. [9] has a journal publication identifying the ambient noise of a multirotor UAS. They map the sound profile in two planes, adjacent to the propellers, and under the propellers. Dissent on experimentation analysis was done in the scope of both U. S. A. and E. U. regulators and their views of noise pollution in civil aviation. In an interesting turn, Iannace et al. [10] offers a fault diagnosis method for UAVs using artificial neural networks. In the process of building a model, they examine different imbalances by placing makeshift (tape) weights to introduce vibration. The model was trained to recognize the acoustics of the imbalanced propeller, to a success rate of 97% in the environment they tested in. Previous work by Semke, [11], offers analysis of different types of mounting mechanisms onboard a quadcopter and the vibration characteristics of the sensor mounting mechanisms of both multirotor and fixed wing aircrafts. The vibration environment onboard fixed-wing and quadrotor sUAS and provides sensor data to assist in passive and active vibration control methodologies [12]. 2.2 Experimental Testing In contrast to many related works today, the goal is to investigate and better understand what happens when there is a damage to the propeller. A critical question we draw attention to is how much damage influences the acceleration and acoustic levels and ultimately the flight performance of the sUAS. Testing was performed on a DJI Phantom 4 Pro sUAS as shown in Fig. 2.1 Table 2.1 provides a key list of the aircraft specifications. In this paper four sets of blades were tested in the laboratory. For each type of blade, vibration and acoustic measurements were simultaneously collected. Undamaged blades were tested first, then one centimeter from a single blade tip, on one side Fig. 2.1 DJI Phantom 4 Pro UAS Table 2.1 DJI Phantom 4 Pro Specifications Weight 1388 g Diagonal Size 350mm MaxSpeed 2m/s Max Flight Time Approx. 30 minutes Hover Range Accuracy Vertical Accuracy: ±0.1 m (Vision Positioning), ±0.5 m (GPS Positioning) Horizontal Accuracy: ±0.3 m (Vision Positioning), ±1.5 m (GPS Positioning)
2 The Vibration and Acoustic Effects of Prop Design and Unbalance on Small Unmanned Aircraft 11 Fig. 2.2 Damaged propeller (left) and undamaged propeller (right) for the Standard 9450S and Low Noise 9455S Table 2.2 Summary of Acceleration Testing of the DJI Phantom 4 Pro with Various Propeller Configurations BladeType Amplitude: pk-pk (m/s2) Increase % Standard Deviation (m/s2) Increase% DJI OEM Standard 72 12 DJI OEM Standard Damaged 127 76 21 75 BTG Standard 67 11 BTG Standard Damaged 160 139 26 136 DJI OEM Low-Noise 70 11 DJI OEM Low-Noise Damaged 93 33 17 55 Helistar Low-Noise 138 22 Helistar Low-Noise Damaged 197 43 30 36 of the propeller, was removed to emulate damage. Figure 2.2 shows the damaged and undamaged props for the 4 types of props studied. 2.3 Vibration Testing To measure vibrations, a PCB Model 333B30 single-axis accelerometer (10.5 mV/m/s2) was attached to the base of the aircraft and the data was collected using a Data Physics Abacus 901 at a rate of 1536 Hz. The data was collected while the aircraft was in a hovering configuration approximately 1–2 meters above ground level. The peak-to-peak amplitude and the standard deviation for the recorded accelerations are shown in Table 2.2. Table 2.2 illustrates several interesting findings observed from the accelerometer data. First, damage had a more dramatic impact on standard prop designs than it did with the low noise prop configurations. In the undamaged state, all props showed similar peak-to-peak and standard deviation amplitudes, except for the Helistar low noise prop. This prop design had significantly higher peak-to-peak and standard deviation amplitudes compared to all other undamaged prop formats. The highest increase in vibration amplitudes due to prop damage was observed in the BTG standard prop model. The lowest increased in vibration amplitudes due to prop damage was observed in the DJI OEM low noise prop model. The Fast Fourier Transform (FFT) of each of the propeller configurations tests are shown in Figs. 2.3, 2.4, 2.5 and 2.6. The FFT plots show the frequency response of the aircraft for frequencies up to 600 Hz.
12 W. H. Semke et al. 0.00001 0.0001 0.001 0.01 0.1 1 0 100 200 300 400 500 600 Magnitude Frequency (Hz) Undamaged Damaged Fig. 2.3 DJI OEM 9450S Quick-Release Propellers FFT Comparison of Damaged and Undamaged Propellers 0.00001 0.0001 0.001 0.01 0.1 1 0 100 200 300 400 500 600 Magnitude Frequency (Hz) Undamaged Damaged Fig. 2.4 BTG 9450S Quick-Release Propellers FFT Comparison of Damaged and Undamaged Propellers 0.00001 0.0001 0.001 0.01 0.1 1 0 100 200 300 400 500 600 Magnitude Frequncy (Hz) Undamaged Damaged Fig. 2.5 DJI OEM 9455S Low-Noise Quick Release Propellers FFT Comparison of Damaged and Undamaged Propellers The DJI OEM 9450S quick-release propellers FFT show a higher baseline and peak amplitudes across all frequencies with a damaged prop. The largest increase in amplitude occurred at frequencies under 150 Hz. The BTG 9450S quickrelease propellers FFT shows a higher baseline and peak amplitudes across all frequencies when damaged as well. The highest increases occurred at frequencies below 150 Hz and between 200 and 450 Hz. The DJI OEM 9455S low-noise quick release propellers FFT did not show significant increase for most frequencies. Interestingly, there was a decrease in amplitude for most frequencies under 100 Hz along with an additional peak in the FFT plot around 50 Hz when the prop was damaged.
2 The Vibration and Acoustic Effects of Prop Design and Unbalance on Small Unmanned Aircraft 13 0.00001 0.0001 0.001 0.01 0.1 1 0 100 200 300 400 500 600 Magnitude Frequency (Hz) Undamaged Damaged Fig. 2.6 Helistar 9455S Low-Noise Quick Release Propellers FFT Comparison of Damaged and Undamaged Propellers The Helistar 9455S low-noise quick release propellers FFT only shows a slight increase across the frequencies studied with prop damage. This case showed the least impact of damage on the FFT results as the undamaged and damaged FFT plots closely mirrored each other. 2.4 Acoustic Testing To measure noise, the ACO Pacific 4012 - 1/2 Inch directional microphone with a PS9200 2-channel power supply was used. The raw data was recorded using a TEAC LX-10 and later imported into MALAB® for analysis. Using a MATLAB® power spectrum analyzer object, a voltage spectrum was generated. The spectrum was then converted into Sound Pressure Lever (SPL) using the microphone sensitivity (50 mV/Pa). The results were converted into sound level3 at each frequency in Pascal (Pa) and subsequently in dB with a 20μPa reference pressure using Eq. (2.1). Li =20log10 pi/pref (2.1) where Li is the sound pressure in dB of the i th frequency and pi sound pressure in Pa. We also calculated the overall sound pressure level by combining the individual levels using Eq. (2.2). L=10log10 i 10Li/10 (2.2) Figures 2.7a, 2.8a, 2.9a, and 2.10a depict the noise level spectrum and the overall SPL for the eight tests. The DJI and BTG standard blades show a 4 dB in noise level increase from the undamaged to the damaged props. The increase in noise level can also be related to the increased in vibration level for these blades as shown in Figs. 2.3, 2.4, 2.5, and 2.6 and Table 2.1. There was no significant change in noise level between the undamaged and damaged blades for both types of the low noise props. However, the vibration level was increased for these props. The overall sound pressure level results are summarized in Table 2.3. There were 4-dB increases in noise level for the DJI and BTG standard blades, while the low noise props by DJI and Helistar only saw 1 and 0 dB increases, respectively. Further analysis of the data was carried out in the low frequency range (0–600 Hz) to better understand the correlation between the props radiated noise and vibration levels. Figures 2.7b, 2.8b, 2.9b, and 2.10b were thus generated. These figures represent the spectrum plot in Pascal (Pa) zooming in frequencies from 0 to 600 Hz. It can be seen on these sound pressure figures the frequency peaks correspond to the same frequencies as the vibration response frequency peaks. However, the peaks do not relate to the increase in sound pressure level from the undamaged to the damaged blades that were measured. For example, in Fig. 2.9b, the acoustic level peaks from the undamaged blade is higher than that of the damaged blade even though the overall sound pressure level indicates that the damaged blades were louder. The Figs. 2.7b, 2.8b, 2.9b, and 2.10b
14 W. H. Semke et al. Fig. 2.7 DJI OEM 9450S Quick-Release Propeller Comparison of Damaged and Undamaged Propellers. (a) Noise Level Spectrum (b) Sound Pressure Spectrum Fig. 2.8 BTG 9450S Quick-Release Propeller Comparison of Damaged and Undamaged Propellers. (a) Noise Level Spectrum (b) Sound Pressure Spectrum Table 2.3 Summary of Sound Pressure Level of the DJI Phantom 4 Pro with Various Propeller Configurations BladeType Sound Pressure Level (dB) Increase (dB) DJI OEM Standard 74 4 DJI OEM Standard Damaged 78 BTG Standard 76 4 BTG Standard Damaged 80 DJI OEM Low-Noise 74 1 DJI OEM Low-Noise Damaged 75 Helistar Low-Noise 74 0 Helistar Low-Noise Damaged 74 clearly show a correlation in frequency peaks between vibration and noise measurements. However, no definite conclusion can be drawn for the amplitude correlation between those two types of measurements.
2 The Vibration and Acoustic Effects of Prop Design and Unbalance on Small Unmanned Aircraft 15 Fig. 2.9 DJI OEM 9455S Low Noise Quick-Release Propeller Comparison of Damaged and Undamaged Propellers (a) Noise Level Spectrum (b) Sound Pressure Spectrum Fig. 2.10 Helistar 9455S Low Noise Quick-Release Propeller Comparison of Damaged and Undamaged Propellers (a) Noise Level Spectrum (b) Sound Pressure Spectrum 2.5 Conclusions The extent of the damage to the props in this study did not greatly impact the flight performance of the aircraft. When the damage was introduced, the DJI Phantom 4 Pro did experience slight yawing when first taking off, but the internal flight controller quickly resolved this issue after becoming airborne and nominal flight performance was experienced. More significant, or damage to multiple props, may impart more dramatic and potentially catastrophic consequences resulting in an aircraft that is not airworthy. This greater damage was not studied in this analysis, only the relatively small damage that may result from small prop strikes or an unbalanced situation. The observed vibration shows that the damage to the props increased both the peak-to-peak and standard deviation amplitudes for all configurations, but the damage had greater impact on the standard prop configurations. There was less impact in vibration due to the damage in the low noise configurations. The FFT amplitudes of the standard prop configurations showed a greater increase with damage. In general, the natural frequencies did not shift with damage except for the formation of an additional peak near 50 Hz in the damaged DJI OEM low noise prop.
16 W. H. Semke et al. The observed acoustics depicted an increase in sound pressure level and can be qualitatively related to the increase in vibration levels. It is therefore our belief that based on these preliminary findings a microphone or an array of microphones could be used to monitor the vibrations on a sUAV during mission. In the testing conducted only 2 microphones were used and the exact positioning of the aircraft was not maintained throughout all the testing. While the position was not exact, only small variations of less than 1 m were observed. Even though the direct amplitude correlation between increase noise level and increased vibration level could not be determined in the current data sets further testing could be done in the future to establish a clear relationship between the noise level of a sUAV with damaged props and the onboard vibration levels. The amplitude of the various acoustic results across the frequency bands investigated do not necessarily directly correlate with the amplitudes of the vibration modes experienced onboard the aircraft. We believe further tests are needed to gain a better understanding of the behavior of sUAS props with a special emphasis on the low noise style. Greater prop damage magnitude and damage types will be investigated as well as a more detailed study on the correlation of vibration and acoustic responses. In this manner, we believe a non-invasive damage detection algorithm is possible using acoustic methods. This system would be part of robust and economical safety assessment procedures and protocols for preflight testing of remote autonomous sUAS systems that are planned for the future. References 1. Carlton, J.S.: In: marine propellers and propulsion,. Butterworth-Heinemann,. In: Chapter 2 - Propulsion Systems, Cambridge, MA, United States (2019) 2. Thomas, R., Zahui, M.: Small unmanned aerial vehicle fuselage dynamic model using wings vibrations, Proceedings of the 22th international congress on sound and vibration. Florence, 12–16 July, 2015 3. Katalin, A.: Engineering Society International Conference 2017, MESIC” 11th international conference interdisciplinarity in engineering, INTER-ENG 2017, Tirgu-Mures, Romania, 5–6 Oct, 2017 4. Barron, R.F.: Industrial Noise Control and Acoustics, 1st edn. CRC Press (2002) 5. Crede, C.E.: Vibration and Shock Isolation. Wiley, New York (1951) 6. Rao, S.S.: Mechanical Vibrations. Prentice Hall, New Jersey (2004) 7. Meirovitch, L.: Fundamentals of Vibrations. McGraw-Hill, New York, NY, United States (2001) 8. Sas, P., et al.: Vibration Analysis of a UAV Multirotor Frame. KU Leuven - Department Werktuigkunde (2016) 9. Kloet, N., et al.: Acoustic signature measurement of small multi-rotor unmanned aircraft systems. Int. J. Micro Air Veh. 3–14 (2017). https:// doi.org/10.1177/1756829316681868 10. Iannace, G., Ciaburro, G., Trematerra, A.: Fault diagnosis for UAV blades using artificial neural network. Robotics. 8, 59 (2019) 11. Semke, W.: Vibration reduction for camera systems onboard small unmanned aircraft, Proceedings of the international modal analysis conference (IMAC) XXXVII: a conference and exposition on structural dynamics, 2019 12. Semke, W., Dunlevy, M.: A review of the vibration environment onboard small unmanned aircraft, Proceedings of the international modal analysis conference (IMAC) XXXVI: a conference and exposition on structural dynamics, 2018
Chapter 3 A Deformed Geometry Synthesis Technique for Determining Stacking and Cryogenically Induced Preloads for the Space Launch System Joel Sills, Arya Majed, and Edwin Henkel Abstract The Space Launch System (SLS) stacking and Core Stage (CS) fueling induce significant preloads that contribute to the liftoff pad separation “twang”. To accurately capture this, an approach is required that can replicate the physics of all SLS physical stacking steps, CS cryogenic shrinkage, associated geometric nonlinearities, and the transient behavior and decay of the preloads with changing boundary conditions as the vehicle separates from the pad. The Deformed Geometry Synthesis (DGS) approach presented here satisfies the above requirements. DGS determines induced preloads by modeling components in their deformed geometry states and then enforcing compatibility by closing the resulting “deadbands”. DGS seamlessly integrates into the multibody modal synthesis framework and does not require the use of artificial external loads to enforce preloads or post-processing steps to remove their influence. Since DGS iterates to solve for the deformed state inclusive of geometric nonlinearities, running linearized parametrics to exercise different potential orientations of ball jointed struts that connect the CS to Boosters for cryo-shrinkage analyses is entirely avoided. Relative to the transient behavior and decay of stacking and cryo-induced preloads with SLS liftoff pad separation, this is an area of considerable interest to the SLS program. To capture this in the most accurate way possible, DGS algorithms are designed to work with Henkel-Mar nonlinear pad separation algorithms which operate on the separating longitudinal and lateral degrees of freedom (DoFs) between the vehicle and the pad. As the separating DoFs release, in whatever manner as dictated by the interface geometries, interface loads and interface flexibilities as well as the external loading on the vehicle, the subject preloads generate a complex twang/decay time-trace as dictated by the physics of the problem. This paper presents DGS numerical verification against the closed-form solution for Timoshenko’s 3 ball-jointed strut preload problem. This problem is then extended by the authors to the geometric nonlinear case where DGS is compared to the Newton-Raphson solution of the nonlinear equations. Next, DGS is utilized to solve the SLS stacking and cryogenic shrinkage coupled loads analyses. Finally, Henkel-Mar pad separation simulations are executed that isolate the impact of the induced preloads’ twang and decay characteristics. Keywords Space launch system · Deformed geometry synthesis · Henkel-Mar pad separation · Geometric nonlinear 3.1 Introduction A new approach for coupling deformed geometries and determining vehicle load indicators inclusive of prelaunch stacking and cryogenic induced preloads has been developed. This Deformed Geometry Synthesis (DGS) technique [1] is a specialized procedure in modal synthesis where components are coupled in their deformed geometry states by enforcing compatibility at the interfaces via a process of statically closing the resulting deadbands to lock-in the preloads. The procedure evokes a static version of the nonlinear deadband methodology first developed for the Space Shuttle Program (SSP). It has the advantage of being NASA verified and validated against test and utilized in the SSP nonlinear coupled loads analyses (CLAs) from 2005 until final flight in 2011. To accurately capture the SLS stacking and cryo-induced preloads, DGS replicates the physics of all SLS physical stacking steps, Core Stage (CS) cryogenic shrinkage, associated geometric nonlinearities (e.g., aft strut rotations – a set of 3 ball J. Sills ( ) NASA Engineering and Safety Center (NESC), Houston, TX, USA e-mail: joel.w.sills@nasa.gov A. Majed · E. Henkel Applied Structural Dynamics (ASD), Inc., Houston, TX, USA © The Society for Experimental Mechanics, Inc. 2021 C. Walber et al. (eds.), Sensors and Instrumentation, Aircraft/Aerospace, Energy Harvesting & Dynamic Environments Testing, Volume 7, Conference Proceedings of the Society for Experimental Mechanics Series, https://doi.org/10.1007/978-3-030-47713-4_3 17
RkJQdWJsaXNoZXIy MTMzNzEzMQ==