Search

On 19 September 2022, the first workshop on AI for Manufacturing (AI4M Workshop) took place at ECML-PKDD, the European Conference on Machine Learning and Principles and Practice for Knowledge Discovery in Databases. The workshop brought together researchers and practitioners, from academia and industry, contributing their perspectives. This special section includes five articles in which Artificial Intelligence methods are used to address real problems in the manufacturing industry, ranging from the supply chain, to production, to quality insurance, and predictive maintenance. In this introduction, we present a high-level overview of the current state of the area: observed trends and the main open challenges. This overview is based on these papers, the keynote presentation, the panel discussion, and the discussion that emerged during the workshop.

Packaging represents the largest fraction of plastic waste in Europe. Currently, mechanical recycling schemes are mainly focused on the recovery of rigid packaging (like bottles), while for flexible packaging, also called films, recycling rates remain very low. Existing mechanical recycling technologies for these films are quite basic, especially in the case of complicated post-consumer flexible plastics (PCFP) waste, leading to regranulate qualities that are often subpar for renewed use in demanding film applications. In this study, the technical and economic value of an improved mechanical recycling process (additional sorting, hot washing, and improved extrusion) of PCFPs is investigated. The quality of the four types of resulting regranulates is evaluated for film and injection molding applications. The obtained Polyethylene-rich regranulates in blown films offer more flexibility (45-60%), higher ductility (27-55%), and enhanced tensile strength (5-51%), compared to the conventional mechanical recycling process. Likewise, for injection molded samples, they exhibit more flexibility (19-49%), enhanced ductility (7 to 20 times), and higher impact strength (1.8 to 3.8 times). An economic assessment is made between the obtained increased market value and the capital investment required. It is shown that the economic value can be increased by 5-38% through this improved recycling process. Overall, the study shows that it is possible to increase the mechanical recycling quality of PCFP in an economically viable way, thus opening the way for new application routes and overall increased recycling rates.

The interoperability of different virtual commissioning applications requires a lot of manual work, as they mainly use proprietary file formats. In this case, a generic virtual commissioning data model is highly demanded to realize interoperability among different virtual commissioning applications. Based on this model, the effort of exchanging data between different virtual commissioning applications is significantly reduced. A comprehensive generic virtual commissioning data model contains a variety of virtual commissioning-related information, such as geometry, physics, kinematics, sensors, actuators, and signal connections. For such an elaborate data model, no method is contemporarily indicated to evaluate its interoperability performance. Thus, in this paper, a case-based evaluation framework is introduced to fill in this gap. With this framework, the interoperability performance of a generic virtual commissioning data model can be evaluated by conducting data exchange of several emulation models developed in different virtual commissioning applications, and each of their specific virtual commissioning modeling functions is covered in at least one of these emulation models. As an example, a case study is presented by exchanging 10 carefully selected emulation models between Siemens NX and Visual Components software applications via AutomationML data format. Based on the case study, the applicability of this evaluation framework is analyzed and discussed. Finally, conclusion and outlook are illustrated in the end.

Preoperative clinical MRI protocols for gliomas, brain tumors with dismal outcomes due to their infiltrative properties, still rely on conventional structural MRI, which does not deliver information on tumor genotype and is limited in the delineation of diffuse gliomas. The GliMR COST action wants to raise awareness about the state of the art of advanced MRI techniques in gliomas and their possible clinical translation. This review describes current methods, limits, and applications of advanced MRI for the preoperative assessment of glioma, summarizing the level of clinical validation of different techniques. In this second part, we review magnetic resonance spectroscopy (MRS), chemical exchange saturation transfer (CEST), susceptibility-weighted imaging (SWI), MRI-PET, MR elastography (MRE), and MR-based radiomics applications. The first part of this review addresses dynamic susceptibility contrast (DSC) and dynamic contrast-enhanced (DCE) MRI, arterial spin labeling (ASL), diffusion-weighted MRI, vessel imaging, and magnetic resonance fingerprinting (MRF). Evidence Level: 3. Technical Efficacy: Stage 2.

Purpose To support acquisition of accurate, reproducible and high-quality preclinical imaging data, various standardisation resources have been developed over the years. However, it is unclear the impact of those efforts in current preclinical imaging practices. To better understand the status quo in the field of preclinical imaging standardisation, the STANDARD group of the European Society of Molecular Imaging (ESMI) put together a community survey and a forum for discussion at the European Molecular Imaging Meeting (EMIM) 2022. This paper reports on the results from the STANDARD survey and the forum discussions that took place at EMIM2022. Procedures The survey was delivered to the community by the ESMI office and was promoted through the Society channels, email lists and webpages. The survey contained seven sections organised as generic questions and imaging modality-specific questions. The generic questions focused on issues regarding data acquisition, data processing, data storage, publishing and community awareness of international guidelines for animal research. Specific questions on practices in optical imaging, PET, CT, SPECT, MRI and ultrasound were further included. Results Data from the STANDARD survey showed that 47% of survey participants do not have or do not know if they have QC/QA guidelines at their institutes. Additionally, a large variability exists in the ways data are acquired, processed and reported regarding general aspects as well as modality-specific aspects. Moreover, there is limited awareness of the existence of international guidelines on preclinical (imaging) research practices. Conclusions Standardisation of preclinical imaging techniques remains a challenge and hinders the transformative potential of preclinical imaging to augment biomedical research pipelines by serving as an easy vehicle for translation of research findings to the clinic. Data collected in this project show that there is a need to promote and disseminate already available tools to standardise preclinical imaging practices.

Preoperative clinical magnetic resonance imaging (MRI) protocols for gliomas, brain tumors with dismal outcomes due to their infiltrative properties, still rely on conventional structural MRI, which does not deliver information on tumor genotype and is limited in the delineation of diffuse gliomas. The GliMR COST action wants to raise awareness about the state of the art of advanced MRI techniques in gliomas and their possible clinical translation or lack thereof. This review describes current methods, limits, and applications of advanced MRI for the preoperative assessment of glioma, summarizing the level of clinical validation of different techniques. In this first part, we discuss dynamic susceptibility contrast and dynamic contrast-enhanced MRI, arterial spin labeling, diffusion-weighted MRI, vessel imaging, and magnetic resonance fingerprinting. The second part of this review addresses magnetic resonance spectroscopy, chemical exchange saturation transfer, susceptibility-weighted imaging, MRI-PET, MR elastography, and MR-based radiomics applications.Evidence Level: 3Technical Efficacy: Stage 2

The credibility of threshold-based alarms in anesthesia monitors is low and most of the warnings they produce are not informative. This study aims to show that Machine Learning techniques have a potential to generate meaningful alarms during general anesthesia without putting constraints on the type of procedure. Two distinct approaches were tested – Complication Detection and Anomaly Detection. The former is a generic supervised learning problem and for this a simple feed-forward Neural Network performed best. For the latter, we used an Encoder-Decoder Long Short-Term Memory architecture that does not require a large manually-labeled dataset. We show this approach to be more flexible and in the spirit of Explainable Artificial Intelligence, offering greater potential for future improvement.

Field-Programmable Gate Array (FPGA)-based Software-Defined Radio (SDR) is well-suited for experimenting with advanced wireless communication systems, as it allows to alter the architecture promptly while obtaining high performance. However, programming the FPGA using a Hardware Description Language (HDL) is a time-consuming task for FPGA developers and difficult for software developers, which limits the potential of SDR. High-Level Synthesis (HLS) tools aid the designers by allowing them to program on a higher layer of abstraction. However, if not carefully designed, it may lead to a degradation in computing performance or significant increase in resource utilization. This work shows that it is feasible to design modern Orthogonal Frequency Division Multiplex (OFDM) baseband processing modules like channel estimation and equalization using HLS without sacrificing performance and to integrate them in an HDL design to form a fully-operational FPGA-based Wi-Fi (IEEE 802.11a/g/n) transceiver. Starting from no HLS experience, a design with minor overhead in terms of latency and resource utilization as compared to the HDL approach was created in less than one month. We show the readability of the sequential logic as coded in HLS, and discuss the lessons learned from the approach taken and the benefits it brings for further design and experimentation. The FPGA design generated by HLS was verified to be bit-true with its MATLAB implementation in simulation. Furthermore, we show its practical performance when deployed on a System-on-Chip (SoC)-based SDR using a professional wireless connectivity tester., 7 pages, extended version of poster accepted at FCCM 2023

There is a trend to go towards low gear-ratio or even direct-drive actuators in novel robotic applications in which high-torque density electric motors are required. The Yokeless and Segmented Armature Axial Flux Permanent Magnet Synchronous Machine is therefore considered in this work. In these applications, the motors should be capable to deliver high torque at standstill for long periods of time. This can cause overheating of the motors due to a concentration of the losses in a single phase; hence, it becomes necessary to derate the motor torque. In this work the influence of the slot/pole combination, the addition of a thermal end-winding interconnection and the equivalent thermal conductivity of the winding body on the torque performance at standstill will be studied both experimentally via temperature measurements on a prototype stator, and via a calibrated 3D thermal Finite Element model. It was found that both a good choice of the slot/pole combination and the addition of a thermal end-winding interconnection have a significant influence on the torque performance at standstill, and allow up to 8% increase in torque at standstill in comparison to a reference design.

This paper presents an enhanced artificial neural network (ANN) to predict the displacement in composite pipes impacted by a drop weight having different velocities. The impact response of fiber-reinforced polymer composite pipes depends on several factors including thickness, stacking sequence, and the number of layers. These factors were investigated in an earlier study using sensitivity analysis, and it was found that they had the most prominent effect on the impact resistance of the composite pipes. In this present study, composite pipes with a diameter of 54 mm are considered to explore the damages induced by low-velocity impact and the influence of these damages on their strength. To evaluate the effect of low-velocity, the pipes were exposed to impacts at different velocities of 1.5, 2, 2.5, and 3 m/s, and preliminary damage was initiated. Next, we used Jaya and E-Jaya algorithms to enhance the ANN algorithm for good training and prediction. The Jaya algorithm has a basic structure and needs only two requirements, namely, population size and terminal condition. Recently, Jaya algorithm has been widely utilized to solve various problems. Due to its single learning technique and limited population information, Jaya algorithm may quickly be trapped in local optima while addressing complicated optimization problems. For better prediction, an enhanced Jaya (E-Jaya) algorithm has been presented to enhance global searchability. In this study, ANN is enhanced based on the influential parameters using E-Jaya to test its effectiveness. The results showed the effectiveness of the E-Jaya algorithm for best training and prediction compared with the original algorithm.

This paper presents an in-depth comparison of the benefits and limitations of using a low-voltage DC (LVDC) microgrid versus an AC microgrid with regard to the integration of low-carbon technologies. To this end, a novel approach for charging electric vehicles (EVs) on low-voltage distribution networks by utilizing an LVDC backbone is discussed. The global aim of the conducted study is to investigate the overall energy losses as well as voltage stability problems on DC and AC microgrids. Both architectures are assessed and compared to each other by performing a power flow analysis. Along this line, an actual low-voltage distribution network with various penetration levels of EVs, combined with photovoltaic (PV) systems and battery energy storage systems is considered. Obtained results indicate significant power quality improvements in voltage imbalances and conversion losses thanks to the proposed backbone. Moreover, the study concludes with a discussion of the impact level of EVs and PVs penetration degrees on energy efficiency, besides charging power levels’ impact on local self-consumption reduction of the studied system. The outcomes of the study can provide extensive insights for hybrid microgrid and EV charging infrastructure designers in a holistic manner in all aspects.

Metal-halide perovskites (MHPs) exhibit excellent properties for application in optoelectronic devices. The bottleneck for their incorporation is the lack of long-term stability such as degradation due to external conditions (heat, light, oxygen, moisture, and mechanical stress), but the occurrence of phase transitions also affects their performance. Structural phase transitions are often influenced by phonon modes. Hence, an insight into both the structure and lattice dynamics is vital to assess the potential of MHPs. In this study, GIWAXS and Raman spectroscopy are applied, supported by density functional theory calculations, to investigate the apparent manifestation of structural phase transitions in the MHP CsPbBr3. Macroscopically, CsPbBr3 undergoes phase transitions between a cubic (alpha), tetragonal (beta), and orthorhombic (gamma) phase with decreasing temperature. However, microscopically, it has been argued that only the ? phase exists, while the other phases exist as averages over length and time scales within distinct temperature ranges. Here, direct proof is provided for this conjecture by analyzing both theoretical diffraction patterns and the evolution of the tilting angle of the PbBr6 octahedra from molecular dynamics simulations. Moreover, sound agreement between experimental and theoretical Raman spectra allowed to identify the Raman active phonon modes and to investigate their frequency as a function of temperature. As such, this work increases the understanding of the structure and lattice dynamics of CsPbBr3 and similar MHPs.

Additive manufacturing technologies have become at the forefront in tissue engineering, enabling the fabrication of complex tissues with intricate geometries that were not feasible using conventional manufacturing techniques. Due to the rapid progress in this field, it has become difficult not only to choose the most appropriate method, but also the optimal material, biological model (i.e., cells and bioactive compounds), and processing technique to fulfill the macro- and microstructural architecture and functions of biological tissues. The aim of this review is to describe recent advances in tissue engineering fabrication methods, from established electrospinning to emerging additive manufacturing technologies, with particular emphasis on tissues that exhibit hierarchically organized anisotropic architecture (skeletal muscle, tendons, and peripheral nerves). One of the current challenges is that the designs are usually dictated by the constraints imposed by the methods, rather than by criteria based on mechanical and biological requirements. Therefore, the review focuses on describing how the anatomical structure and function of muscles, tendons, and nerves should serve as the basis for an efficient three-dimensional design that considers both micro and macro aspects of the tissue. In addition, the individual factors that influence the fabrication strategy are discussed and related to the mechanical and biological properties of the three tissue types. The review highlights the advantages and limitations of each fabrication strategy and provides an overview of critical aspects relevant to future research strategies in this area.

Due to constant advancements in materials research, conductive textile-based materials have been used increasingly in textile-based wearables. However, due to the rigidity of electronics or the need for their encapsulation, conductive textile materials, such as conductive yarns, tend to break faster around transition areas than other parts of e-textile systems. Thus, the current work aims to find the limits of two conductive yarns woven into a narrow fabric at the electronics encapsulation transition point. The tests consisted of repeated bending and mechanical stress, and were conducted using a testing machine built from off-the-shelf components. The electronics were encapsulated with an injection-moulded potting compound. In addition to identifying the most reliable conductive yarn and soft–rigid transition materials, the results examined the failure process during the bending tests, including continuous electrical measurements.

The components of an operational wind turbine are continuously impacted by both static and dynamic loads. Regular inspections and maintenance are required to keep these components healthy. The main bearing of a wind turbine is one such component that experiences heavy loading forces during operation. These forces depend on various parameters such as wind speed, operating regime and control actions. When a wind turbine provides frequency containment reserve (FCR) to support the grid frequency, the forces acting upon the main bearing are also expected to exhibit more dynamic variations. These forces have a direct impact on the lifetime of the main bearing. With an increasing trend of wind turbines participating in the frequency ancillary services market, an analysis of these dynamic forces becomes necessary. To this end, this paper assesses the effect of FCR-based control on the main bearing lifetime of the wind turbine. Firstly, a control algorithm is implemented such that the output power of the wind turbine is regulated as a function of grid frequency and the amount of FCR. Simulations are performed for a range of FCR to study the changing behaviour of dynamical forces acting on the main bearing with respect to the amount of FCR provided. Then, based on the outputs from these simulations and using 2 years of LiDAR wind data, the lifetime of the main bearing of the wind turbine is calculated and compared for each of the cases. Finally, based on the results obtained from this study, the impact of FCR provision on the main bearing lifetime is quantified and recommendations are made, that could be taken into account in the operation strategy of a wind farm.

The DuRSAAM 2023 Symposium celebrates the end of an exciting and successful EU-project, titled ‘PhD Training Network on Durable, Reliable and Sustainable Structures with Alkali-Activated Materials (DuRSAAM)’ and coordinated by Ghent University. In line with the scope of this European project, that has run from 2018 till 2023, the DuRSAAM2023 Symposium focusses on new developments in all aspects of alkali-activated concrete, sometimes also referred to as geopolymer concrete. These open access proceedings collect the short papers, as presented by the participants during the symposium, and provides researchers, building professionals and stakeholders recent insights on advancing alkali-activated materials.

Digital microfluidics (DMF) holds great potential for the alleviation of laboratory procedures in assisted reproductive technologies (ARTs). The electrowetting on dielectric (EWOD) technology provides dynamic culture conditions in vitro that may better mimic the natural embryo microenvironment. Thus far, EWOD microdevices have been proposed for in vitro gamete and embryo handling in mice and for analyzing the human embryo secretome. This article presents the development of the first microfluidic chip utilizing EWOD technology designed for the manipulation of bovine embryos in vitro. The prototype sustains the cell cycles of embryos manipulated individually on the chips during in vitro culture (IVC). Challenges related to the chip fabrication as well as to its application during bovine embryo IVC in accordance with the adapted on-chip protocol are thoroughly discussed, and future directions for DMF in ARTs are indicated.

Natural marine soundscapes are being threatened by increasing anthropic noise, particularly in shallow coastal waters. To preserve and monitor these soundscapes, understanding them is essential. Here, we propose a new method for semi-supervised categorization of shallow marine soundscapes, with further interpretation of these categories according to concurrent environmental conditions. The proposed methodology uses a nonlinear mapping of short-term spectrograms to a two-dimensional space, followed by a density-based clustering algorithm to identify similar sound environments. A random forest classifier, based on additional environmental data, is used to predict their occurrence. Finally, explainable machine learning tools provide insight into the ecological explanation of the clusters. This methodology was tested in the Belgian part of the North Sea, and resulted in clearly identifiable categories of soundscapes that could be explained by spatial and temporal environmental parameters, such as distance to the shore, bathymetry, tide or season. Classifying soundscapes facilitates their identification, which can be useful for policy making or conservation programs. Soundscape categorization, as proposed in this work, could be used to monitor acoustic trends and patterns in space and time that might provide useful indicators of biodiversity and ecosystem functionality change.

Glass as a construction material has become indispensable and is still on the rise in the building industry. However, there is still a need for numerical models that can predict the strength of structural glass in different configurations. The complexity lies in the failure of glass elements largely driven by pre-existing microscopic surface flaws. These flaws are present over the entire glass surface, and the properties of each flaw vary. Therefore, the fracture strength of glass is described by a probability function and will depend on the size of the panels, the loading conditions and the flaw size distribution. This paper extends the strength prediction model of Osnes et al. with the model selection by the Akaike information criterion. This allows us to determine the most appropriate probability density function describing the glass panel strength. The analyses indicate that the most appropriate model is mainly affected by the number of flaws subjected to the maximum tensile stresses. When many flaws are loaded, the strength is better described by a normal or Weibull distribution. When few flaws are loaded, the distribution tends more towards a Gumbel distribution. A parameter study is performed to examine the most important and influencing parameters in the strength prediction model.

Currently there is gaining interest in pre-compressed foam sealing tapes to seal joints watertight between different building envelope components. Little to no information is available on the parameters affecting the resistance of these foam tapes to driving rain. On the other hand, several research studies have shown that water leakages can be expected at relatively low-pressure differences and that drainage should be provided. Therefore, a study was designed to on the one hand assess the material and installation parameters that affect the watertightness of pre-compressed polyurethane foam sealing tapes impregnated with an acrylic polymer dispersion and on the other hand evaluate the potential of providing drainage possibilities, either as a two-barrier system or by means of integrated drainage cavities. It was found that the joint width, the presence of an airtight coating, and the position of the tape relative to the exterior surface affected the watertightness of the sealed joints. Notably, 87% of the evaluated foam tapes applied as a single barrier showed water leakages at pressure differences of 600 Pa or lower. Foam tapes with integrated drainage cavities, on the other hand, resulted in watertight joints up to an average pressure difference of 825 Pa.

A general anti-Baldwin radical 4-exo-dig cyclization from nitrogen-substituted alkynes is reported. Upon reaction with a heteroleptic copper complex in the presence of an amine and under visible light irradiation, a range of ynamides were shown to smoothly cyclize to the corresponding azetidines, useful building blocks in natural product synthesis and medicinal chemistry, with full control of the regioselectivity of the cyclization resulting from a unique and underrated radical 4-exo-dig pathway., The construction of four-membered rings via a 4-exo-dig cyclization was originally theorized to be unfavourable and only recently shown in sparse examples. Here the authors present a photochemical, radical 4-exo-dig cyclization of ynamides to form azetidines, promoted by copper photoredox catalysis.

Refereed/Peer-reviewed Video watermarking techniques can be used to prevent unauthorized users from illegally distributing videos across (social) media networks. However, current watermarking solutions are unable to embed a perceptually invisible watermark which is robust to the distortions introduced by camcording. These watermark-disrupting distortions include lossy compression, the addition of noise, frame-rate conversion and geometric distortions. In this paper, we present a novel video watermarking technique that is blind and robust to camcording attacks. The proposed approach uses the integration of the dual-tree complex wavelet transform (DTCWT) and singular value decomposition (SVD) to achieve robustness against geometric attacks. The experimental results validate our technique’s superior imperceptibility and robustness to several attacks when compared to existing peer mechanisms. In conclusion, the proposed technique can be used to protect against illegal distribution of video content.

In this paper, a new method in forecasting the horizontal displacement of diaphragm wall (D.W.) for high-rise buildings is introduced. A new stochastic optimizer, called Planet Optimization Algorithm (P.O.A.), is employed to assess how proper finite element (F.E.) simulation is against field data. The process is adopted for a real phased excavation measured at the field. To automatically run the iterative optimization tasks, a source code is constructed directly in the Geotechnical Engineering Software (PLAXIS) by using Python to ensure that the operation between optimization algorithm and F.E. simulations are smooth to guarantee the accuracy of the complex calculation for the soil problem. The proposed process consists of two steps. (1) The parameters will be optimized at the early phases of the excavation. (2) The responses of D.W. displacements are forecasted at the subsequent phases. The aim of the process is to predict the displacements of D.W. of the building from the result of the nearby excavation or to provide early warning about the risks of excavation that may happen under vital phases. The proposed procedure also provides an effective method for optimization-based soil parameters updating in real engineering practice.

This paper describes a novel automatic control toolbox, designed for non-experienced practitioners. Fractional order (FO) controllers are easily tuned with the main purpose of easy practical implementation. Experimental step data are required for the automatic FO controller tuning. An embedded system identification algorithm uses the step data to obtain a process model as a second order plus dead-time (SOPDT) system. Finally, the FO controller is computed based on the previously estimated SOPDT model in order to fulfil a set of user-imposed frequency domain performance specifications: phase margin, gain crossover frequency and gain margin maximization. Experimental step response data from a strongly nonlinear vertical take-off and landing unit have been used to design an FO controller using the toolbox. The experimental closed loop results validate the proposed toolbox. The end result is a user-friendly automatic fractional order controller tuning with endless possibilities of real-world applicability.

The experimentally observed interplay between catalytic activation of methane on Fe (c) SiO2 and gas-phase free radical methane coupling under non-oxidative conditions is analyzed by mechanistic modeling as well as by experiments. For the modeling, an off-the shelf gas-phase model, AramcoMech 3.0, was used unaltered to keep the number of adjustable parameters as low as possible. It was complemented by surface reactions specifically accounting for methane activation to methyl radicals. The model was validated against an independent set of experimental data and exhibited good accordance. The model accurately captured the significant contribution of gas-phase reactions responsible for methane conversion in the post-catalytic zone, indicative of gas-phase autocatalytic methane coupling. The low-activity induction period in gas-phase methane pyrolysis can effectively be overcome by adequate catalytic activation. Results show that the catalytic reaction only influences the activity of the system, with gas-phase reactions dictating the selectivity distribution. Simulations demonstrated that the optimum catalytic conversion roughly amounts to 4 % at 1000 degrees C and 1 atm. An equivalent effect can be reached by adding ca. 2 % of ethane or ethylene to the feed. Detailed reaction-path analyses were employed to corroborate these phenomena. Gas-phase reactions were found to be very rapid at 1000 degrees C, hence determining the product selectivity, without impact from either catalyst or C2 hydrocarbon addition. Current, freely available gas-phase models lack the required accuracy for detailed kinetic modeling of the product distribution, showing the requirement for the development of a dedicate non-oxidative methane coupling model.

Recently proposed methods in intrusion detection are iterating on machine learning methods as a potential solution. These novel methods are validated on one or more datasets from a sparse collection of academic intrusion detection datasets. Their recognition as improvements to the state-of-the-art is largely dependent on whether they can demonstrate a reliable increase in classification metrics compared to similar works validated on the same datasets. Whether these increases are meaningful outside of the training/testing datasets is rarely asked and never investigated. This work aims to demonstrate that strong general performance does not typically follow from strong classification on the current intrusion detection datasets. Binary classification models from a range of algorithmic families are trained on the attack classes of CSE-CIC-IDS2018, a state-of-the-art intrusion detection dataset. After establishing baselines for each class at various points of data access, the same trained models are tasked with classifying samples from the corresponding attack classes in CIC-IDS2017, CIC-DoS2017 and CIC-DDoS2019. Contrary to what the baseline results would suggest, the models have rarely learned a generally applicable representation of their attack class. Stability and predictability of generalized model performance are central issues for all methods on all attack classes. Focusing only on the three best-in-class models in terms of interdataset generalization, reveals that for network-centric attack classes (brute force, denial of service and distributed denial of service), general representations can be learned with flat losses in classification performance (precision and recall) below 5%. Other attack classes vary in generalized performance from stark losses in recall (−35%) with intact precision (98+%) for botnets to total degradation of precision and moderate recall loss for Web attack and infiltration models. The core conclusion of this article is a warning to researchers in the field. Expecting results of proposed methods on the test sets of state-of-the-art intrusion detection datasets to translate to generalized performance is likely a serious overestimation. Four proposals to reduce this overestimation are set out as future work directions.

In recent years machine learning has made extensive progress in modeling many aspects of mass spectrometry data. We brought together proteomics data generators, repository managers, and machine learning experts in a workshop with the goals to evaluate and explore machine learning applications for realistic modeling of data from multidimensional mass spectrometry-based proteomics analysis of any sample or organism. Following this sample-to-data roadmap helped identify knowledge gaps and define needs. Being able to generate bespoke and realistic synthetic data has legitimate and important uses in system suitability, method development, and algorithm benchmarking, while also posing critical ethical questions. The interdisciplinary nature of the workshop informed discussions of what is currently possible and future opportunities and challenges. In the following perspective we summarize these discussions in the hope of conveying our excitement about the potential of machine learning in proteomics and to inspire future research.

Recent advances in electrical machines and energy storage technologies make electric vehicles (EVs) feasible replacements to conventional internal combustion engines. One of the main challenges of high speed electrical machines is providing maximum output power with minimum energy losses, weight, and volume. At high frequency operation, the conductors of AC electrical machines can suffer from skin and proximity effects. This results in high AC losses in the machine windings and can eventually lead to machine failure. In this paper, a novel design for a semi-stranded coil is proposed to limit these undesirable effects. Enabled by additive manufacturing (AM) technology, this sophisticated design is 3D printed using ultralight aluminum alloy. Finally, the AC performance of this coil is measured and compared with conventional single-strand copper coil at different frequency levels. It is found that the proposed design can effectively limit the eddy current losses in the high frequency domain.

Data set acquisition and curation are often the most difficult and time-consuming parts of a machine learning endeavor. This is especially true for proteomics-based liquid chromatography (LC) coupled to mass spectrometry (MS) data sets, due to the high levels of data reduction that occur between raw data and machine learning-ready data. Since predictive proteomics is an emerging field, when predicting peptide behavior in LC-MS setups, each lab often uses unique and complex data processing pipelines in order to maximize performance, at the cost of accessibility and reproducibility. For this reason we introduce ProteomicsML, an online resource for proteomics-based data sets and tutorials across most of the currently explored physicochemical peptide properties. This community-driven resource makes it simple to access data in easy-to-process formats, and contains easy-to-follow tutorials that allow new users to interact with even the most advanced algorithms in the field. ProteomicsML provides data sets that are useful for comparing state-of-the-art machine learning algorithms, as well as providing introductory material for teachers and newcomers to the field alike. The platform is freely available at https://www.proteomicsml.org/, and we welcome the entire proteomics community to contribute to the project at https://github.com/ProteomicsML/ProteomicsML.

International audience; Integrated photonics is an emerging technology for many existing and future tele- and data communication applications. One platform of particular interest is Silicon Nitride (SiN) thanks to - amongst others - its very low-loss waveguides. However, it lacks active devices, such as lasers, amplifiers and photodiodes. For this, hybrid or heterogeneous integration is needed. Here, we bring high-speed uni-travelling-carrier photodiodes (UTC PDs) to a low-loss SiN-platform by means of micro-transfer-printing. This versatile technology for heterogeneous integration not only allows very dense and material-efficient III-V integration, it also eases fabrication yielding high-performance detectors. The waveguide-coupled photodiodes feature a responsivity of 0.3 A/W at 1550 nm, a dark current of 10 nA and a bandwidth of 155 GHz at a low bias. At zero bias, a record bandwidth of 135 GHz is achieved. We further demonstrate that this integrated detector can be used for direct photomixing at terahertz frequencies. A back-to-back communication link with a carrier frequency around 300 GHz is set up, and data rates up to 160 Gbit/s with low error vector magnitude (EVM) are shown, showcasing near-identical performance at zero bias.

Ultrasonic welding (USW) is a solid-state welding process based on the application of high frequency vibration energy to the workpiece to produce the internal friction between the faying surface and the local heat generation required to promote the joining. The short welding time and the low heat input, the absence of fumes, sparks or flames, and the automation capacity make it particularly interesting for several fields, such as electrical/electronic, automotive, aerospace, appliance, and medical products industries. The main problems that those industries have to face are related to the poor weld quality due the improper selection of weld parameters. In the present work, 0.3 mm thick copper sheets were joined by USW varying the welding time, pressure, and vibration amplitude. The influence of the process variables on the characteristics of the joints and weld strength is investigated by using the analysis of variance. The results of the present work indicate that welding time is the main factor affecting the energy absorbed during the welding, followed by the pressure and amplitude. The shear strength, on the other hand, resulted mostly influenced by the amplitude, while the other parameters have a limited effect. Regardless the welding configuration adopted, most welds registered a failure load higher than the base material pointing out the feasibility of the USW process to join copper sheets.

Silicon photonics has emerged as a mature technology that is expected to play a key role in critical emerging applications, including very high data rate optical communications, distance sensing for autonomous vehicles, photonic-accelerated computing, and quantum information processing. The success of silicon photonics has been enabled by the unique combination of performance, high yield, and high-volume capacity that can only be achieved by standardizing manufacturing technology. Today, standardized silicon photonics technology platforms implemented by foundries provide access to optimized library components, including low-loss optical routing, fast modulation, continuous tuning, high-speed germanium photodiodes, and high-efficiency optical and electrical interfaces. However, silicon’s relatively weak electro-optic effects result in modulators with a significant footprint and thermo-optic tuning devices that require high power consumption, which are substantial impediments for very large-scale integration in silicon photonics. Microelectromechanical systems (MEMS) technology can enhance silicon photonics with building blocks that are compact, low-loss, broadband, fast and require very low power consumption. Here, we introduce a silicon photonic MEMS platform consisting of high-performance nano-opto-electromechanical devices fully integrated alongside standard silicon photonics foundry components, with wafer-level sealing for long-term reliability, flip-chip bonding to redistribution interposers, and fibre-array attachment for high port count optical and electrical interfacing. Our experimental demonstration of fundamental silicon photonic MEMS circuit elements, including power couplers, phase shifters and wavelength-division multiplexing devices using standardized technology lifts previous impediments to enable scaling to very large photonic integrated circuits for applications in telecommunications, neuromorphic computing, sensing, programmable photonics, and quantum computing.

Two-phase expansion devices can be applied in a multitude of cycles which allow for an increase of the overall system efficiency in currently used heat and power cycles. The expanders that are mainly investigated for the purpose of two-phase expansion are the Lysholm and the reciprocating expanders. From these two, the reciprocating expander is challenging to model as there is a need to describe the non-equilibrium state of the phases during the expansion process. This study aims to fill the knowledge gap that exists for two-phase reciprocating expanders, by constructing a predictive model to describe the expansion process and the non -equilibrium phenomena occurring. The simulation framework makes use of a modified homogeneous relaxation model to describe the thermal non-equilibrium between the phases within the isolated working chamber. The model results are verified and compared to experimental data in literature with water as working fluid. Subsequently, Cyclopentane is selected as a more suitable candidate and a thorough analysis is presented. The simulation results predict a 15% lower work and a 10% higher mass intake per stroke of the expander compared to equilibrium working conditions.

In this paper, we present a novel technique to model wide-band scattering parameter (S-parameter) curves of high-speed digital interconnects. The proposed technique utilizes a new kernel function with periodic components for Gaussian process (GP) models. After proper training, the GP models are able to predict the S-parameter values at arbitrary frequency points inside the trained interval. The performance of the proposed technique is reviewed by means of correlation with standard Gaussian Processes with squared exponential kernel and Matern kernel. Results for the proposed technique show an increased prediction accuracy when applied to interconnects.

Even without a separate tracking algorithm, the directions of arrival (DOAs) of moving talkers can be estimated with a deep neural network (DNN) when the movement trajectories used for training allow the generalization to real signals. Previously, we proposed a framework for generating training data with time-variant source activity and sudden DOA changes. Slowly moving sources could be seen as a special case thereof, but were not explicitly modeled. In this paper, we extend this framework by using small jumps between neighboring discrete DOAs to simulate gradual movements. Further, we investigate the benefit of a latency controlled bidirectional recurrent layer in the DNN architecture, such that the required strictly limited context of future frames may still be acceptable for real-time applications. Experiments with real recordings show that the revised data generation leads to more continuous DOA paths, whereas the future context enables a quicker detection of speech onsets and offsets.

This paper is concerned with the task of speaker verification on audio with multiple overlapping speakers. Most speaker verification systems are designed with the assumption of a single speaker being present in a given audio segment. However, in a real-world setting this assumption does not always hold. In this paper, we demonstrate that current speaker verification systems are not robust against audio with noticeable speaker overlap. To alleviate this issue, we propose margin-mixup, a simple training strategy that can easily be adopted by existing speaker verification pipelines to make the resulting speaker embeddings robust against multi-speaker audio. In contrast to other methods, margin-mixup requires no alterations to regular speaker verification architectures, while attaining better results. On our multi-speaker test set based on VoxCeleb1, the proposed margin-mixup strategy improves the EER on average with 44.4% relative to our state-of-the-art speaker verification baseline systems., Comment: proceedings of ICASSP 2023

Composite biopolymer/conducting polymer scaffolds, such as polylactic acid (PLA)/ polyaniline (PAni) nanofibers, have emerged as popular alternative scaffolds in the electrical-sensitive nerve tissue engineering (TE). Although mimicking the extracellular matrix geometry, such scaffolds are highly hydrophobic and usually present an inhomogeneous morphology with massive beads that impede nerve cell-material interactions. Therefore, the present study launches an exclusive combinatorial strategy merging successive pre- and post-electrospinning plasma treatments to cope with these issues. Firstly, an atmospheric pressure plasma jet (APPJ) treatment was applied on PLA and PLA/PAni solutions prior to electrospinning, enhancing their viscosity and conductivity. These liquid property changes largely eliminated the beaded structures on the nanofibers, leading to uniform and nicely elongated fibers having average diameters between 170 and 230 nm. After electrospinning, the conceived scaffolds were subjected to a N2 dielectric barrier discharge (DBD) treatment, which significantly increased their surface wettability as illustrated by large decreases in water contact angles for values above 125° to values below 25°. X-ray photoelectron spectroscopy (XPS) analyses revealed that 3.3% of nitrogen was implanted on the nanofibers surface in the form of C–N and N–C=O functionalities upon DBD treatment. Finally, after seeding pheochromocytoma (PC-12) cells on the scaffolds, a greatly enhanced cell adhesion and a more dispersive cell distribution were detected on the DBD-treated samples. Interestingly, when the APPJ treatment was additionally performed, the extension of a high number of long neurites was spotted leading to the formation of a neuronal network between PC-12 cell clusters. In addition, the presence of conducting PAni in the scaffolds further promoted the behavior of PC-12 cells as illustrated by more than a 40% increase in the neurite density without any external electrical stimulation. As such, this work presents a new strategy combining different plasma-assisted biofabrication techniques of conducting nanofibers to create promising scaffolds for electrical-sensitive TE applications.

Open-set recognition (OSR) has achieved significant importance in recent years. For a robust recognition system, we need to identify the right class from a myriad of knowns and unknowns. In this work, we build and compare OSR systems for patient activity recognition (PAR) using compact radar sensors in a hospital setting. Radar sensors are an important part of a privacy-preserving monitoring system. Specifically, the proposed approach is based on a deep discriminative representation network (DDRN) trained using the large margin cosine loss (LMCL) and triplet loss (TL). A probability of an inclusion model in the embedding space based on the Weibull distribution is able to separate knowns from unknowns. This overall approach limits the risk of open space and enables us to easily identify any unknown activities. Our experiments show that the proposed approach is significantly better for open-set human activity recognition (HAR) with radar when compared with the state-of-the-art open-set approaches.