Real Use-case Research Articles

Iterative active learning strategies for subgraph matching

The subgraph matching problem arises in a variety of domains including pattern recognition for segmented images, meshes of 3D objects, biochemical reactions, and security applications. Large and complex solution spaces are common for this graph-based problem, especially when the world graph contains many more nodes and edges in comparison to the template graph. Researchers have focused on the task of finding one match or many matches, however a real use-case scenario can necessitate identifying specific matches from a combinatorially complex solution space. Our work directly addresses this challenge. We propose to introduce additional queries to the subgraph that iteratively reduce the size of the solution space, and consider the optimal strategy for doing so. We formalize this problem and demonstrate that it is NP-complete. We compare different quantitative criteria for choosing nodes to query. We introduce a new method based on a spanning tree that outperforms other graph-based criteria for the multichannel datasets. Finally, we present numerical experiments for single channel and multichannel subgraph matching problems created from both synthetic and real world datasets.

Learning Ordinal-Hierarchical Constraints for Deep Learning Classifiers.

Real-world classification problems may disclose different hierarchical levels where the categories are displayed in an ordinal structure. However, no specific deep learning (DL) models simultaneously learn hierarchical and ordinal constraints while improving generalization performance. To fill this gap, we propose the introduction of two novel ordinal-hierarchical DL methodologies, namely, the hierarchical cumulative link model (HCLM) and hierarchical-ordinal binary decomposition (HOBD), which are able to model the ordinal structure within different hierarchical levels of the labels. In particular, we decompose the hierarchical-ordinal problem into local and global graph paths that may encode an ordinal constraint for each hierarchical level. Thus, we frame this problem as simultaneously minimizing global and local losses. Furthermore, the ordinal constraints are set by two approaches ordinal binary decomposition (OBD) and cumulative link model (CLM) within each global and local function. The effectiveness of the proposed approach is measured on four real-use case datasets concerning industrial, biomedical, computer vision, and financial domains. The extracted results demonstrate a statistically significant improvement to state-of-the-art nominal, ordinal, and hierarchical approaches.

Data augmentation strategy for generating realistic samples on defect segmentation task

In the anomaly and defect detection tasks, the number of negative samples greatly exceeds the number of defective samples. As a result, a high-class imbalance exists among different classes in the detection task. In our work, we introduce a data-level solution for improving the generalization performance of the semantic segmentation of surface defects based on a data augmentation (DA) strategy. In particular, our DA approach comprised a generative stage to simulate synthetic defects and a validation stage to validate the synthetic image as close as possible to the real one. A Siamese network fully validates our synthetic samples to select only synthetic defects as close to the real ones. We demonstrated the effectiveness of our approach in a real-use case scenario to baseline DA approaches. Our DA approach allows balancing the minority classes while improving the overall generalization performance for semantic segmentation for defect detection.

Characterization of aluminum and boron carbide based additive manufactured material for thermal neutron shielding

We characterize the suitability of two additive manufactured materials for use in shielding thermal neutrons for neutron scattering instrumentation. We compare two different boron based additive manufactured masks to a traditionally manufactured boron nitride mask in a real-use case scenario. The first additive manufactured mask is a composite of aluminum and boron carbide. The second additive manufactured mask is a composite of boron carbide and cyanoacrylate. We report the thermal conductivity of the aluminum boron carbide mask in comparison to other relevant materials. The aluminum boron carbide material may find use in shielding applications where thermal conductivity, a low total neutron scattering cross section, and low toxicity are criteria.

Nuclear Quantum Effects Made Accessible: Local Density Fitting in Multicomponent Methods.

The simulation of nuclear quantum effects (NQEs) is crucial for an accurate description of systems and processes involving light nuclei, such as hydrogen atoms. Within the last years, the importance of those effects has been highlighted for a vast range of systems with tremendous implications in chemistry, biology, physics, and materials sciences. However, while electronic structure theory methods have become routine tools for quantum chemical investigations, there is still a lack of approaches to address NQEs that are computationally accessible and straightforward to use. To address this, we present the first combination of the nuclear-electronic orbital Hartree-Fock approach with both local and density fitting approximations (LDF-NEO-HF). This results in a low-order scaling approach that enables the inclusion of NQEs for large systems within a fraction of a day and for small to medium size systems in minutes. Moreover, we demonstrate the qualitative accuracy and robustness of our approach to retrieve NQEs for three real-use cases motivated by chemical, biological, and materials science applications.

A unifying rank aggregation framework to suitably and efficiently aggregate any kind of rankings

The aggregation of multiple rankings into a consensus ranking is a crucial task in various domains such as search engine results or user-based ratings. This task poses significant challenges due to its inherent complexity. The complexity of the problem stems not only from the need for exactness and efficiency, but also from the diversity of real-world scenarios, which often involve incomplete rankings and ties.Most existing methods propose a specific way to aggregate rankings. However, these methods often do not take into account different real use-case scenarios which can impact the relevance of their final result, as the congruence between the aggregated output and the expected outcome inherently depends on the context. To address the issue of context-dependency in ranking aggregation, we introduce a unifying framework that subsumes a variety of generalizations of the Kemeny score for incomplete rankings with ties and enables the design of new ones if a specific context requires it. Our framework is parameterized, allowing for different behaviors depending on the specific use case.We provide a broader scope of application to the methods encompassed by our approach, augmenting them with a larger theoretical and algorithmic structure. We establish an axiomatic study to better understand each method within our framework and present an algorithmic approach that includes exact methods, partitioning algorithms, and heuristics.Finally, we demonstrate the practical relevance of our approach through an empirical study on both real and synthetic datasets. Notably, the synthetic datasets are generated based on devised real-world scenarios, highlighting the context-dependent applicability of different Kemeny-based rank aggregation methods within our framework.

LearnChain: Transparent and cooperative reinforcement learning on Blockchain

We consider multi-agent reinforcement learning (MARL) with the popular paradigm of centralized training and decentralized execution (CTDE). CTDE empowers sharing knowledge from agents in different environments for updating a shared model. A wide range of applications is supported through CTDE in MARL, such as self-driving vehicle coordination, traffic lights synchronization, or cooperation in various aspects of the Internet of Things (IoT), including resource management. Despite the drawbacks of relying on a central authority for handling model updates, incorporating multiple sources of data raises concerns about the trustworthiness of the process. For instance, participating agents could provide data in the favor of their experiences to shift the model towards certain behaviors. Similarly, sending falsified data for updates could lead to adversarial attacks. To overcome these challenges, it is essential to integrate the Ethereum Blockchain technology to handle model updates in the CTDE paradigm by achieving decentralized storage and consensus mechanism for model updates. In the literature, there exist multiple efforts that propose using reinforcement learning (RL) on Blockchain; however, none of them have considered updating MARL of CTDE on-chain, allowing transparent and auditable record of the training process. Therefore, we propose LearnChain, a framework that offers an integration between the CTDE mechanism and a Consortium Blockchain built between authorized participants, thus avoiding gas costs. At the core of LearnChain, RL is integrated with Quorum, offering separate smart contracts for deployment, data handling with incentive mechanisms, training, target update, and inference. Based on a real use-case entailing management of Vehicular Edge Computing tasks through multi-agent synchronization, we implement LearnChain and evaluate its performance and cost in different settings. Our results show the ability to improve learning from shared experiences and to adapt to environment changes on the Quorum BlockChain.

A multiobjective decision-making approach for modelling and planning economically and environmentally sustainable robotic disassembly for remanufacturing

Adopting initiatives to extend the useful life of products or recovering their components for reuse, remanufacturing or recycling is a key issue in the attempt to protect the environment and minimise the demand for natural resources. To contribute to the performance of automated disassembly practices, this paper presents a multiobjective decision-making approach based on the optimisation of three goals in a robotic disassembly cell framework: enhancing the economic performance of the process, reducing energy consumption and mitigating the environmental impact. Two real-use cases are presented as demonstrators, supported by appropriate, updated information from industry. The design model allowed the authors to obtain the best robotic disassembly sequence plan, the correct disassembly direction, the best recovery option for the disassembled components – reuse, remanufacturing, recycling or disposal – and the most appropriate disassembly tools, finding the optimal or near-optimal solution that best balances the three sustainability goals. An Enhanced Discrete Bees Algorithm with a mutation operator was employed to find the solution for the optimisation. Moreover, a multiobjective Bees Algorithm, a Non-dominated Sorting Genetic Algorithm II and a Pareto Envelope-based Selection Algorithm II were adopted to solve the multiobjective optimisation problem using different iteration numbers and population sizes. The results provide insights into robotic disassembly processes, encouraging firms to adopt more automated and sustainable remanufacturing strategies.

Wireless self-powered monitoring system for underground cable joints: a real use-case

Wireless self-powered monitoring system for underground cable joints: a real use-case

Integrating collective know-how for multicriteria decision support in agrifood chains-application to cheesemaking.

Agrifood chain processes are based on a multitude of knowledge, know-how and experiences forged over time. This collective expertise must be shared to improve food quality. Here we test the hypothesis that it is possible to design and implement a comprehensive methodology to create a knowledge base integrating collective expertise, while also using it to recommend technical actions required to improve food quality. The method used to test this hypothesis consists firstly in listing the functional specifications that were defined in collaboration with several partners (technical centers, vocational training schools, producers) over the course of several projects carried out in recent years. Secondly, we propose an innovative core ontology that utilizes the international languages of the Semantic Web to effectively represent knowledge in the form of decision trees. These decision trees will depict potential causal relationships between situations of interest and provide recommendations for managing them through technological actions, as well as a collective assessment of the efficiency of those actions. We show how mind map files created using mind-mapping tools are automatically translated into an RDF knowledge base using the core ontological model. Thirdly, a model to aggregate individual assessments provided by technicians and associated with technical action recommendations is proposed and evaluated. Finally, a multicriteria decision-support system (MCDSS) using the knowledge base is presented. It consists of an explanatory view allowing navigation in a decision tree and an action view for multicriteria filtering and possible side effect identification. The different types of MCDSS-delivered answers to a query expressed in the action view are explained. The MCDSS graphical user interface is presented through a real-use case. Experimental assessments have been performed and confirm that tested hypothesis is relevant.

ER-R: Improving regression by deep learning and prior knowledge utilization for fluorescence analysis

Linear regression is a dominant estimation technique in chemometrics, where there is a need for inexpensive and reliable sensors for water monitoring. However, most problems are nonlinear, such as the estimation of concentration in solution from an emitted fluorescence spectrum (EFS). Even if an estimation method gives desirable results, at some point it will be used under field conditions, where poor signal quality and less control over environmental effects are expected, leading to poor performance. In this study, we overcome these problems by implementing deep neural network (DNN) models and transfer learning technique for EFS analysis. The proposed models, R (Regression module) and ER (Encoder-Regression), outperformed linear methods and a naive DNN approach for high-quality laboratory-sampled data with a maximum mean relative error of ∼11%, vs. a minimum mean relative error of 184% for the linear methods. In the case of low-quality data, which were simulated based on a real-use case, the lowest error of the linear methods climbed to 263%, whereas the proposed ER model error remained at 9%. At low concentrations, ER gave the best results for all datasets: ∼3.46 ppb in the high-quality datasets, and 2.4 ppb in the low-quality datasets.

Comparing SYCL data transfer strategies for tracking use cases

The aim of this work is to compare the performance and ease of programming of the various data transfer strategies provided by SYCL 2020: buffers/accessors on one hand and the different storage types exposed by Unified Shared Memory (USM) on the other hand. We measured the relative performance of USM exclusively located either on the host (USM host) or on the device (USM device), or automatically managed and moved (USM shared). We also tried to quantify the impact of formatting data in a GPU-friendly manner as opposed to using a regular pointer-based structure in which the C++ allocator was replaced by the SYCL allocator sycl::malloc. We first made a PCIe-bound micro-benchmark to test the SYCL memory models with a simple access pattern. We then switched to a real use-case provided by Traccc, a research project associated to the ACTS particle tracking library. The algorithm of interest is SparseCCL, a clustering algorithm used on the first step of track reconstruction. For consistency, all tests were made on a single hardware architecture: a recent server having a discrete GPU with dedicated VRAM, representative of the hardware currently used by the ACTS team.

Saliency detection for large-scale mesh decimation

Highly complex and dense models of 3D objects have recently become indispensable in digital industries. Mesh decimation then plays a crucial role in the production pipeline to efficiently get visually convincing yet compact expressions of complex meshes. However, the current pipeline typically does not allow artists control the decimation process, just a simplification rate. Thus a preferred approach in production settings splits the process into a first pass of saliency detection highlighting areas of greater detail, and allowing artists to iterate until satisfied before simplifying the model. We propose a novel, efficient multi-scale method to compute mesh saliency at coarse and finer scales, based on fast mesh entropy of local surface measurements. Unlike previous approaches, we ensure a robust and straightforward calculation of mesh saliency even for densely tessellated models with millions of polygons. Moreover, we introduce a new adaptive subsampling and interpolation algorithm for saliency estimation. Our implementation achieves speedups of up to three orders of magnitude over prior approaches. Experimental results showcase its resilience to problem scenarios that efficiently scales up to process multi-million vertex meshes. Our evaluation with artists in the entertainment industry also demonstrates its applicability to real use-case scenarios.

A User Perspective on HTR Methods for the Automatic Transcription of Rare Scripts: The Case of Codex Runicus

Recent breakthroughs in Artificial Intelligence, Deep Learning, and Document Image Analysis and Recognition have significantly eased the creation of digital libraries and the transcription of historical documents. However, for documents in rare scripts with few labelled training data available, current Handwritten Text Recognition (HTR) systems are too constraining. Moreover, research on HTR often focuses on technical aspects only, and rarely puts emphasis on implementing software tools for scholars in Humanities. In this article, we describe, compare, and analyse different transcription methods for rare scripts. We evaluate their performance in a real-use case of a medieval manuscript written in the runic script ( Codex Runicus ) and discuss advantages and disadvantages of each method from the user perspective. From this exhaustive analysis and comparison with a fully manual transcription, we raise conclusions and provide recommendations to scholars interested in using automatic transcription tools.

Reducing the Sim2Real-Gap in Extrusion Blow Molding using Random Forest Regressors

In plastics manufacturing, the extrusion of thermoplastic material and subsequent blow molding into a cavity is a common procedure to create hollow objects with arbitrary geometric shapes and sizes. To ensure reliability of the production process and satisfactory quality of the final product, several control parameters need to be adjusted to modify the process behavior. The appropriate parameter allocation is typically determined during product and process development using FEM simulations. The simulations provide the ability to estimate the expected geometric shape of the final product and adjust the control parameters accordingly. However, simulations are only as good as their underlying computational model and can reflect reality only to a restricted degree. Deviations between the simulated product quality and real quality measurements of the physically produced product, the so-called sim2real gap, cannot be entirely avoided and persist even after several optimization cycles. In this paper, we address this problem and propose a data-driven approach to correct the sim2real gap. At its core, our approach is based on a random forest regression model that predicts simulation deviations based on process and product characteristics. The input data for the model is composed of control parameters from the extrusion and blow molding process as well as data from intermediate simulation results. Given the multitude of different variables, we incorporate extensive feature selection and engineering to extract the most valuable information to predict the sim2real gap. We train and evaluate our approach in a real use-case in which several different product geometries are manufactured with differently configured control parameters. The results show that our approach successfully reduces the sim2real-gap significantly. In addition, we show that with carefully designed feature engineering and parameterization, a model trained on one set of products with certain geometrical characteristics can be transferred and applied to another set of products with distinctively different geometrical characteristics.

Embedding and Contacting of Electrical Components for Hybrid Additive Manufacturing

Integrating electrical functionality during additive manufacturing (AM) is a promising new field of application. In this article, different concepts for combining an extrusion-based AM process called 3-D dispensing (3DD) with the embedding of electrical components are developed and compared. Based on criteria, such as resulting surface quality or overall accuracy, the best strategy among six different approaches is identified. To further validate and confirm the choice, long-term reliability tests [humidity and thermal shock testing (TST)] with a particular focus on the electrical contacts between the printed tracks and the embedded electronic components have been conducted. It was shown that the embedding strategy and the cavity design have no influence on the long-term properties of the printed electrical tracks. Finally, an acceleration sensor is embedded in a 3-D housing to prove real use-case applicability and to show the improvements identified in the previous study.

Toward Mapping an NGSI-LD Context Model on RDF Graph Approaches: A Comparison Study.

A considerable number of Internet of Things deployments are isolated from specific solutions, from devices to data platforms. Standardized data models were proposed to overcome the interoperability gap between these deployments. Next generation service interfaces-linked data (NGSI-LD) is one of the proposed platforms that exploits linked data and proposes an information model and an application programming interface (API) for easy use and standard management of context information. The NGSI-LD information model is based on JSON for Linked Data (JSON-LD) as a serialization format for context information. This efficiently exploits the potential of semantics and linked open data. However, the NGSI-LD graph API and query language are still theoretically defined and limited to some preliminary works. Consequently, current NGSI-LD implementations are mainly based on traditional databases, where the JSON-LD serialization is supported but not exploited owing to the difficulties in defining and implementing new NGSI-LD based Graph APIs. One of the basic solutions is the use of an RDF store for NGSI-LD payloads because these types of databases are well defined and maintained and will not need any added effort for JSON-LD based payloads. However, the main complication at this level is the use of reification to annotate relationships. This study focused on both aspects of exploiting the semantics of NGSI-LD by proposing standardized mapping mechanisms to RDF graphs without reifying JSON-LD payloads and with respect to the NGSI-LD context model and API. Our main proposals highlight that exploiting the RDF store for processing NGSI-LD data semantically is feasible and uncomplicated. We illustrated the proposed mapping approaches with real use-case examples and a possible exploitation of semantic approaches.

A Chip Antenna for Bluetooth Earphones with Cross-Head Interference Tested from Received-Signal Sensing.

In this paper, a novel chip antenna and its function in wireless connectivity are presented for Bluetooth (BLT) earphones. The chip antenna is a metamaterial so compact (<λ/8), as the size of 4.9 × 13.0 × 2.0 mm3, that when it is mounted on the realistic PCB, it can be held in the enclosure of the BLT earphone. This setting does not degrade the resonance (S11 < −10 dB) of the proposed antenna. As two earphones in a pair are demanded to communicate with each other, one shares an RF signal with the other and they take turns as the master and slave. The received signal sensing is conducted with the latest model of human head-ear-phantom located between the earphones to mimic the real use-case and cross-head interference. Electromagnetic simulation of the antenna is done and verified by fabrication and measurement. Particularly, received-signal strength indications between the proposed antennas in the earphones are experimentally obtained as −67.5 dBm and −70 dBm without and with the head-ear-phantom, respectively, much greater than −120 dBm, the limit of detection, and implying acceptable connectivity and invulnerability over cross-head-interference problems.

An embedded toolset for human activity monitoring in critical environments

In many working and recreational activities, there are scenarios where both individual and collective safety have to be constantly checked and properly signaled, as occurring in dangerous workplaces or during pandemic events like the recent COVID-19 disease. From wearing personal protective equipment to filling physical spaces with an adequate number of people, it is clear that a possibly automatic solution would help to check compliance with the established rules. Based on an off-the-shelf compact and low-cost hardware, we present a deployed real use-case embedded system capable of perceiving people’s behavior and aggregations and supervising the appliance of a set of rules relying on a configurable plug-in framework. Working on indoor and outdoor environments, we show that our implementation of counting people aggregations, measuring their reciprocal physical distances, and checking the proper usage of protective equipment is an effective yet open framework for monitoring human activities in critical conditions.

Methodology for data-driven predictive maintenance models design, development and implementation on manufacturing guided by domain knowledge

ABSTRACT The 4th industrial revolution has connected machines and industrial plants, facilitating process monitoring and the implementation of predictive maintenance (PdM) systems that can save up to 60% of maintenance costs. Nowadays, most PdM research is carried out with expert systems and data-driven algorithms, but it is mainly focused on improving the results of reference simulation data sets. Hence, industrial requirements are not commonly addressed, and there is no guiding methodology for their implementation in real PdM use-cases. The objective of this work is to present a methodology for PdM application in industrial companies by combining data-driven techniques with domain knowledge. It defines sequentially ordered stages, steps and tasks to facilitate the design, development and implementation of PdM systems according to business and process characteristics. It also facilitates the collaboration among the required working profiles and defines deliverables. It is designed in a flexible and iterative way, combining standards, state-of-the-art methodologies and referent works of the field. Finally, the proposed methodology is validated on two use-cases: a bushing testbed and a press machine of the production line. These use-cases aim to facilitate, guide and speed up the implementation of the methodology on other PdM use-cases.