Lars Duggen

Nowadays, intelligent models can correctly detect faults by analysing signals from rotating machinery. However, most of the studies are run in controlled environments and the performance in industrial real-world environments is not yet fully validated. Hence, a suitable tool to implement fault diagnosers is transfer learning, this topic is under development and challenges persist. This paper proposes a framework for creating accurate 1D-CNN based fault classifiers that can be transferred between different rotating machines and working conditions. Multiple Bayesian processes select architecture parameters and hyperparameters, which minimize a loss function related to their transferability to other machines and to the same machine under different operating conditions (such as load and engine speed). The resulting model is fitted to heterogeneous fault diagnosis data resulting in a 1D-CNN ensemble that improves the performance of the unitary model. Subsequently, the transfer learning capability of the ensemble is analyzed on two source data sets using function and parameter based transfer. The results are compared with classical fault diagnosis classifiers. Finally, additional transfer operations on five target domain datasets validate our framework on limited labeled samples and allow interpretation of the ensemble results. The ultimate goal is to find an ensemble that can generalize fault diagnosis on rotating machinery for easy implementation and update in industrial settings.

Additive manufactured smart wearables are establishing an increasingly firm foothold within the monitoring of physical activities. A tiler was the use case proposed by this study, which seeks to monitor changes in their typically kneeling work position with the goal of tracking behaviour and alleviating long-term damage due to incorrect posture. This paper has described the development process of a conformal kneepad with embedded sensing, obtained by means of interposing flexible tactile sensors between 3D printed elastomeric layers. Topology optimization is employed towards material reduction in the context of tailoring the kneepad design for the differentiated loading which the case study imposes. Considerations on the inherent challenges of overmolding flexible geometries are provided along with a design-to-fabrication process description which visits 3D printing, post-processing and 3D scanning. The use of the latter enables the creation of a fully conformal design, optimized for the perfect fit on each individual wearing the kneepad. The topology optimization is validated numerically and the ability of the kneepad to react to different pressure stimuli is tested with a robot arm, capable of simulating different contact angles and load cases. By means of dedicated electronics, the sensed information is modulestreamed over WiFi and tactile images are formed on a user-friendly terminal. The kneepad concept ultimately serves as a case study for extended additive manufactured smart wearables, with potential to be extended towards autonomous feedback and posture correction systems.

Photoacoustic (PA) measurements with open resonators usually provide poor detection sensitivity due to signal leakage at the resonator opening. We have recently demonstrated three different approaches for modelling the photoacoustic signal of open resonators. In this work, one of the approaches is applied for the optimization of the geometry of the T-shaped resonator for improved signal strength and thus sensitivity. The results from the numerical optimization show an increase in the photoacoustic signal by a factor of approximately 7.23. They are confirmed using numerical methods other than the one applied for the optimization and by experimental measurement. The measurement shows an increase in the photoacoustic signal by a factor of approximately 2.34.

The use of additive manufacturing within the field of mechatronics creates exciting oportunities for the integration of digital systems into advanced mechanical devices. This paper has investigated the design considerations inherent to the the 3D printing of a conductive interface between a microcontroller and a digital sensor. The leads of the sensor and the pins of the microcontroller were successfully coupled exclusively through a 3D-printed coupling consisting of both conductive and non-conductive FDM material, manufactured in a single print on an entry-level device. The coupling has been removed and repositioned several times with no functionality change, proving that the 3D printing considerations were valid in terms of maintaining sufficient pressure on the component terminals so as to create enough conductive paths within the material to drive the sensor properly. An impedance analysis and operational testing have been performed, and considerations concerning the robustness of the design have been drawn, all with respect to future improvements and the expansion towards related applications.

New durable elastomeric materials are commercially available for 3D printing, enabling a new class of consumer wearable applications. The mechanical response of soft 3D printed lattices can now be tailored for improved safety and comfort by (a) leveraging functional grading and (b) customizing the outer envelope to conform specifically to the anatomy of the subject (e.g. patient, athlete, or consumer). Furthermore, electronics can be unobtrusively integrated into these 3D printed structures to provide feedback relating to athletic performance or physical activity. A proposed sensor system was developed that weaves unjacketed wires at two distinct layers in a lattice to form a complex capacitor; the capacitance increases as the lattice is compressed and can detect lattice deformation. A structure was fabricated and demonstrated with both static compression as well as low-velocity impact to highlight the utility for wearable applications. This work is focused on improving the performance of American football helmets as highlighted by the National Football League (NFL) Helmet Challenge Symposium; however, the lattice sensing concept can be extended to metal and ceramic lattices as well - relevant to the automotive and aerospace industries.

This paper has explored the design considerations pertaining to the 3D printing of a fully autonomous mechatronic device, consisting of an on-board battery powering a standard DC motor without the need of wires or any other classical joining methods. The achieved goal was to make exclusive use of 3D printing to manufacture both the non-conductive geometry of the enclosure and the conductive paths necessary for transmitting the power through the circuit. A conductive connector was designed to provide a snap fit on the battery terminals while contacting the motor terminals at enough pressure to overcome the internal resistance and ensure current flow. The non-conductive holder which applies the required pressure to maintain the assembly in conductive contact was subjected to topology optimization, with future optimization strategies also discussed. The assembly was then manufactured and operational testing was performed, demonstrating a working product.

T-cell resonators have been used lately for non-invasive blood glucose measurements for photoacoustic spectroscopy on skin samples. A resonator has a significant role in determining the strength of the measured signal and the overall sensitivity of the sensor. Here we present results of the measurement of the photoacoustic signal of such a T-cell resonator. The signal is also modelled using the amplitude mode expansion method, which is based on eigenmode expansion and the introduction of losses in the form of loss factors. The measurement reproduced almost all the calculated resonances from the numerical models with fairly good agreement. The cause of the differences between the measured and the simulated resonances are explained. In addition, the amplitude mode expansion simulation model is established as a faster and computationally less demanding photoacoustic simulation alternative to the viscothermal model. The resonance frequencies from the two models differ by less than 1.8%. It is noted that the relative height of the amplitudes from the two models depends on the location of the antinodes within the different parts of the resonator. The amplitude mode expansion model provides a quick simulation tool for the optimization and design of macro resonators.

This paper presents an accurate programmable pulse generator implementable on low-cost embedded boards. The proposed method is software-based and real-time such that it can be used for real-time control applications. The range of frequencies that can be covered is from 1 Hz to 255 KHz with the resolution of 1 Hz. The convergence of the generated frequency to its desired value is proved theoretically and confirmed using simulations and experiments. The studied application is a stepper motor which should follow desired speed trajectories.