Special Topics in Structural Dynamics & Experimental Techniques, Volume 5

Despite extensive research in prostheses development, amputees still have to cope with severe limits. Tasks, such as climbing stairs and running or walking on soft ground are demanding and represent obstacles in everyday life. Design verification of new devices helps to accelerate the development. However, current test procedures do not include the dynamic interaction between a prosthesis and the human. Real-time Substructuring (RTS) enables investigation of the dynamic behavior of a system, here human and prosthesis, by splitting it into numerically simulated components and one physical component. As this test imitates real dynamic conditions, foot prostheses can be improved during the development stage. In this preliminary study, a one-dimensional mass-spring-mass system is investigated. The upper mass, representing the human being, is simulated numerically on the computer. It is coupled virtually to a prosthesis, represented here as a spring-mass system, which is mounted on a Stewart Platform. Both systems exchange displacement and force information. The upper mass tries to follow a periodic desired trajectory, which is influenced by the coupling. This paper describes the experimental setup and the effect of delay compensation. In addition, it is shown how the accuracy and stability of the RTS simulation depends on the problem description, i.e. how much the system is governed by the mechanical properties of the numerical part. Although we are specifically considering the application of RTS to prosthetics, the current research tackles generic problems that will also help to enhance other applications involving contact, e.g. the docking of satellites.

Robots are widely used to support mission-critical, high-risk and complex operations. Human supervision and remote robot control are often required to operate robots in unpredictable and changing scenarios. Often, robots are controlled remotely by technicians via joystick interfaces which require training and experience to operate. To improve robot usage and practicality, we propose using augmented reality (AR) to create a more intuitive, less training-intensive means of controlling robots than traditional joystick control. AR is a creative platform for developing robot control systems, because AR combines the real world (the environment around the user, the physical robot, etc.) with the digital world (holograms, digital displays, etc.); it can even interpret physical gestures, such as pinching two fingers.

In this research, a Microsoft Hololens headset is used to create an AR environment to control a Yaskawa Motoman SIA5D robot. The control process begins with the user placing an interactable holographic robot in 3D space. The user can then select between two control methods: manual control and automatic control. In manual control, the user can move the end effector of the holographic robot and the physical robot will respond immediately. In automatic control, the user can move the end effector of the holographic robot to a desired location, view a holographic preview of the motion, and select execute if the motion plan is satisfactory. In this preview mode, the user is able to preview both the motion of the robot and the torques experienced by the joints of the manipulator. This gives the user additional feedback on the planned motion. In this project we succeeded in creating an AR control system that makes controlling a robotic manipulator intuitive and effective.

The logarithmic decrement method is perhaps the most common technique for estimating the damping ratio of linear systems with viscous damping. The approach directly relates the damping ratio to two samples collected from peaks of a recorded free oscillation. These peaks are separated by one or more oscillation periods and are inherently influenced by experimental uncertainty. Literature on the method indicates that improved estimates are sometimes obtained with more periods between samples. However, it is unknown when improvements can be expected for a given data set because there is a trade-off between the chosen number of periods and measurement noise. A guideline for selecting the number of periods which minimizes uncertainty in estimated damping is desired.

In this work, an analytical expression is derived for the optimal number of periods between peaks. This expression, obtained from an uncertainty analysis of the logarithmic decrement equation, is shown to be a function of only one system parameter: the damping ratio. This suggests that for linear systems with viscous damping there is a unique, damping-dependent period choice which guarantees minimum uncertainty in the estimated damping ratio. This result is used to obtain an optimal amplitude ratio which offers a simple, accurate, and easy to implement guideline for selecting a second sample. The derived expressions are applied to a set of numerical systems to confirm their validity.

This paper presents a novel arrangement of a current blocking shunt circuit for the mitigation of multiple structural resonances. The number of required electrical components is reduced compared to the previous versions of this circuit proposed in the literature. This paper also proposes a tuning methodology for the electrical parameters of this circuit based on the evaluation of the electromechanical coupling between the electrical circuit and the structure. Effective mitigation performance can be expected with little knowledge of the host structure. A comparison with the solutions in the literature demonstrates the efficiency of the proposed approach.

Reduced order modelling is of crucial importance for the dynamics of complex Finite Element structures. Thereby the overall deformation state is approximated by a superposition of weighted trial vectors, commonly called modes. The weighting factors (‘modal coordinates’) are obtained by numerical time integration of the reduced order model. In case of complex systems, the time integration normally dominates the overall simulation time. A multibody simulation of a flexible crankshaft interacting with pistons, con rods, fly wheel, hydrodynamic bearings and furthers for instance, takes at least several hours of CPU time. The modal coordinates can then be used for modal stress recovery in order to predict the fatigue lifetime. If a variant of the flexible body with small changes needs to be investigated, a new numerical time integration is necessary. In this paper a method is proposed where the modal coordinates of a flexible body will be projected unto another mode base. This will be done by using the key idea of the SEREP method where the modal coordinates are computed via the Pseudo-Inverse. One academic and one industrial example demonstrate that the time integration of the variant can totally be skipped without remarkable loss of accuracy, as long as the differences between the two flexible bodies are small enough.

A novel real time system identification method for structural health monitoring in a multi-story buildings is presented. The goal is to identify structural parameters in order to improve damage detection methods. Here, structural dynamics is modeled by means of a wave equation with Kelvin damping, that relates the stiffness loss with reduction of shear wave velocity propagation of seismic movements in structural elements. The identification system requires only acceleration signals and is based on the least squares method with forgetting factor. Moreover, a new parameterization is presented based on linear integral filters, which eliminates constant disturbances and attenuate measurement noise. The experimental results through an RC building, validate the feasibility of the proposed method.

Decades of research have been conducted on vibration suppression, cancellation, and absorption methods. Recently, distributed arrays of resonators have been implemented in host structures creating devices termed mechanical metamaterials, also known in the literature as metastructures. The benefit of using mechanical metamaterials as opposed to traditional added absorbers is the structure is initially designed including the absorbers instead of adding them after creation, saving time and weight. Additionally, these structures remain capable of bearing loads without adding additional mass. Where past research has focused on designing and optimizing metastructures based on single degree of freedom excitations, the research presented in this paper focuses on a device capable of vibration suppression under excitation in three directions; longitudinal, transverse, and torsional. This accommodation is necessary for these devices to be implemented in non-laboratory settings to account for excitation from multiple directions.

This paper presents experimental data and a preliminary finite element model for a three-dimensional mechanical metamaterial vibration suppression system. Experimental results compare the structure with blocked absorbers to free absorbers to demonstrate vibration reduction bandwidths of 358 Hz in the longitudinal direction, 24 Hz in the transverse direction, and a frequency shift of 40 Hz in the torsional direction. These promising results show that a metastructure can be effectively designed to suppress vibrations in all three directions of excitation with further motivation to explore optimization of the absorber system for maximum suppression bandwidth.

The dynamic behavior of metamaterials and metastructures is often modeled using finite elements; however, these models can become quite large and therefore computationally expensive to simulate. Traditionally, large models are made smaller using any of the array of model reduction methods, such as Guyan or Craig-Bampton reduction. The regularized nature of metamaterials makes them excellent candidates for reduced-order modeling because the system is essentially comprised of a repeating pattern of unit cell components. These unit cell components can be reduced and then assembled to form a reduced-order system-level model with equivalent dynamics. The process is demonstrated using a finite element model of a 1-D axially vibrating metamaterial bar using Guyan, SEREP, and Craig-Bampton reduction methods. The process is shown to provide substantial reduction in the time needed to simulate the dynamic response of a representative metamaterial while maintaining the dynamics of the system and resonators.

Presented is a vision-based algorithm for extracting physical properties from melt pools. The bandwidth requirements for traditional high speed video are too high for real time analysis so silicon retinas are used. This method of imaging has a very fine temporal resolution, high dynamic range, and low bandwidth requirements. The ability to monitor melt pools in real time would improve the quality of laser printed parts and welds because it would allow automatic control systems to recognize and correct imperfections during the printing and welding processes. By measuring the change of intensity within a melt pool then applying blind source separation techniques, spatiotemporal data can be extracted. First a circular membrane model is evaluated to validate the technique. Then the separation technique is performed with a traditional camera on gallium pools of different depths and various lighting conditions. Finally, silicon retina data is used to show that the technique can be applied for this type of imager.

Noncontact methods of experimentally acquiring mode shapes and associated natural frequencies eliminate errors induced by mass-weighting of the structure by sensors. Traditional data acquisition methods require costly and delicate sensors such as accelerometers and strain gauges that are time-consuming to setup on each structure needing to be analyzed. Other non-contact data acquisition tools such as Laser Doppler Vibrometers (LDVs) and Digital Image Correlation (DIC) require expensive equipment and placement of speckle patters or high-contrast markers on the structure. Digital video cameras provide a relatively low-cost and portable method to measure a structure with high spatial resolution without needing to modify the structure. Previous work identified a novel variation on Operational Modal Analysis (OMA) to identify full-field mode shapes from video data. This work develops the algorithm’s robustness, investigating effects of camera motion, structure excitation type, and background intensity gradients. Camera motion and modal over-specification are shown to cause identification of modes that do not correspond to physical deformations of the structure. Previously, video stabilization algorithms have been used to eliminate camera motion from video data. These algorithms eliminate most camera motion, but residual motion remains and is identified in additional, spurious mode shapes. When the camera motion is oscillatory, these shapes can be correlated in the frequency domain to the spectrum seen by an accelerometer placed on the camera itself. Averaging techniques are implemented to improve mode shape quality and identify structural and camera modes from spurious modes identified from modal over-specification. When robustly understood, identification of full-field mode shapes and properties can cheaply and efficiently advance structural health monitoring, model verification and updating, change detection, load identification, and other fields of structural dynamics.

A critical issue for data-based engineering is a lack of descriptive labels for the measured data. For many engineering systems, these labels are costly/impractical to obtain, and as a result, conventional supervised learning is not feasible. This article outlines a probabilistic framework for the investigation and labelling of engineering datasets. Two alternative probabilistic measures are suggested to define the most informative observations to investigate and annotate, in order to maximise the classification performance of a statistical model.

Multiple-Input Multiple-Output (MIMO) swept sine is nowadays acknowledged to be one of the best excitation techniques in applications where testing time is a constraint and high-quality Frequency Response Functions are compulsory. This is the case, for example, of testing large aerospace structures for model validation and updating. The high levels that can be reached during these tests can require a reliable MIMO closed-loop control strategy in order to guarantee that the response spectra will follow safe reference profiles (within defined tolerance limits). The development of a dedicated algorithm for these applications is however very challenging, especially due to the transient nature of the sweeps and the robustness of the MIMO controller. This paper proposes a steepest descent solution for the control of multiple inputs during a continuous sine-sweep, in order to simultaneously match specific response spectra for multiple control channels.

The cochlea, in the mammalian inner ear, transduces acoustic waves into electrical signals that are transmitted to the brain. One of the critical functions of the cochlea is its biological nonlinear behavior that amplifies faint sounds and compresses high sound levels. Previously, authors mimicked the aforementioned nonlinear characteristics in piezoelectric augmented structural cantilevers through nonlinear feedback controllers. The present effort is a continuation of the previous studies in the development of micro-electro-mechanical system (MEMS) scale artificial hair cells (AHCs).

The current research investigates the potential of transforming MEMS scale cantilevers, initially designed for use as scanning thermal microscopy probes, into micro-scale artificial hair cells. These cantilever structures are fabricated by employing electron beam- and photo-lithography, together with Low Pressure Chemical Vapor Deposition (LPCVD), metal evaporation, dry- and wet-etching on n-type silicon wafer substrates. In this work, dynamic characterization of these micro-structures is the focus. A series of dynamic tests are conducted on the MEMS micro-structure as it is subjected to base excitation. The dynamic characteristics of the MEMS system are investigated by varying the excitation levels and evaluating the limits of the structure’s linearity. Based on the experimental findings the potential of using these MEMS cantilevers as active artificial hair cells is evaluated.

This paper introduces a dynamic system parameter extraction approach using experimental data and a known characteristic equation for a single degree of freedom mass-spring-damper system. The efficacy of curve fitting to determine mass, damping, and stiffness for simulated underdamped, critically damped, and overdamped systems is explored. Two methods are investigated to obtain these characteristics for simulated systems; a step response approach in the time domain and a transfer function and dynamic stiffness approach in the frequency domain. Commonly available software, MATLAB, is used for curve fitting. An examination of the effect of inaccurate seeds, or starting values, for mass, damping, and stiffness is included.

A wind turbine foundation has been monitored for one year in order to assess the effectiveness of a repair work carried out in 2017. A number of displacement transducers have been installed together with strain gauges and accelerometers to understand the turbine dynamic behavior.

One year of data is now available coupled with the machine working parameters (power, wind intensity/direction). These data have been analyzed in order to identify any possible evolution in the turbine behavior and therefore highlight any deterioration in the foundation behavior (stiffness reduction). This paper describes the data analysis taking into account all environmental effects and the associated bias on the measured data. Results show a stable behavior of the turbine foundations through the whole year proving the effectiveness of the performed repairs.

Machine learning has been widely and successfully used in many Structural Health Monitoring (SHM) applications. However, many machine learning models can only make accurate predictions when the training and test data are measured from the same system; this is because most traditional machine learning methods assume that all the data are drawn from the same distribution. Therefore, to train a robust predictor, it is often required to recollect and label new training data every time when considering a new structure, which can be significantly expensive, and sometimes impossible in the SHM context. In such cases, the idea of transfer learning may be employed, which aims to transfer knowledge between task domains to improve learners. In this paper, a subfield of transfer learning i.e. domain adaptation, is considered, and its utility in SHM applications is briefly investigated.