Portail HAL IMT Mines Albi
Not a member yet
5440 research outputs found
Sort by
Image Classification Applied to the Problem of Conformity Check in Industry
No full textInternational audienceThis paper shows the application of several learning-based image classification techniques to conformity check, which is a common problem in industrial visual inspection. The approaches are based on processing 2D images. First, a classification pipeline has been developed. An effort has been invested into choosing an appropriate classifier. First experiment was performed with HoG features (Histogram Of Gradient) and Support Vector Machine (SVM). Further, to improve accuracy, we employed a bag of visual words (BoVW) and ORB detector for extracting features that we further use to build our dictionary of visual words. The final solution uses features extracted by passing an image through a pre-trained deep convolutional neural network Inception. Using these features a SVM classifier was trained and high accuracy was obtained. To augment our image data set, different transformations such as zoom and shearing were applied. Promising results were obtained which shows that state-of-the-art deep learning classification techniques can be successfully employed in the visual industrial inspection field
An Intelligent Decision Support System Inspired by Newton’s Laws of Motion
No full textInternational audienceThe purpose of this study is to present a novel perspective on decision technology based on classical physics rules, considering risks and opportunities as physical forces deviating systems as an object from their stable states. The forces are created by changing the internal and external characteristics of the system. The ultimate objective is to propose a multi-criteria performance framework within the geometrical space of the system Key Performance Indicators (KPIs) based on classical physics rules by mapping management concepts onto physical notations. The present study is tuned to a model of interaction between the inventory management module and the workforce supply chain to present the main work. In addition, the significance of the study as an intelligent decision system to manage the given model through Newton laws is investigated
A Model Driven Approach to Transform Business Vision-Oriented Decision-Making Requirement into Solution-Oriented Optimization Model
No full textInternational audienceCurrently in our highly connected society, there is a strong requirement for decision-makers in organizations to coordinate and schedule their activities. Frequently, there are various uncertain factors, multiple objectives, many business knowledge and requirements, which heavily increase the difficulty of decision-making process regarding these issues. Therefore, a decision-maker will appreciate having control over the formulation of decision-making models and being able to adapt to highly dynamic situation. In this paper, we study a Model Driven Engineering (MDE) approach to link the business requirement defined by a model with solution-oriented logical models, which are codes that could be submitted to a combinatorial optimization solver. The design of our proposal follows the principles of three-levels Model Driven Architecture (MDA) and is based on a cognitive process for decision-making systems. Then, several transformation rules between models are explained to realize automatic Model to Model Transformation (M2M) with a special emphasis on the Platform Independent Model (PIM) to Platform Specific Model (PSM) part. To make a proof of our model transformation chain efficiency, a classical Travelling Salesman Problem (TSP) is chosen as a use case
Learning deep domain-agnostic features from synthetic renders for industrial visual inspection
No full textInternational audienceDeep learning has resulted in a huge advancement in computer vision. However, deep models require an enormous amount of manually annotated data, which is a laborious and time-consuming task. Large amounts of images demand the availability of target objects for acquisition. This is a kind of luxury we usually do not have in the context of automatic inspection of complex mechanical assemblies, such as in the aircraft industry. We focus on using deep convolutional neural networks (CNN) for automatic industrial inspection of mechanical assemblies, where training images are limited and hard to collect. Computer-aided design model (CAD) is a standard way to describe mechanical assemblies; for each assembly part we have a three-dimensional CAD model with the real dimensions and geometrical properties. Therefore, rendering of CAD models to generate synthetic training data is an attractive approach that comes with perfect annotations. Our ultimate goal is to obtain a deep CNN model trained on synthetic renders and deployed to recognize the presence of target objects in never-before-seen real images collected by commercial RGB cameras. Different approaches are adopted to close the domain gap between synthetic and real images. First, the domain randomization technique is applied to generate synthetic data for training. Second, domain invariant features are utilized while training, allowing to use the trained model directly in the target domain. Finally, we propose a way to learn better representative features using augmented autoencoders, getting performance close to our baseline models trained with real images
Pickering emulsion as template for porous bioceramics in the perspective of bone regeneration
No full textInternational audienceCalcium phosphate (CaP) based bioceramics are widely used as bone substitutes. The most encountered CaP ceramics are obtained from high temperature phases. However, their bioactivity and their association with biomolecules are limited, as well as their bioresorption in-vivo. The aim of this work is to develop biomimetic low temperature apatites ceramics with tunable porosity via biocompatible high internal phase Pickering emulsions. The biocompatible emulsions developed were stabilized by stoichiometric hydroxyapatite (HA) particles. Several parameters (mass of HA particles, oil/water weight ratio, electrolytes concentration in the aqueous phase) were investigated to define the optimized formulation conditions leading to a kinetically stable monodisperse emulsion with a minimum drop diameter of 200 µm and drops enough percolated to induce interconnected porosity. Two types of porous bioceramics were produced by low temperature processes with controlled composition and porosity, evidenced by X-ray microtomography: calcium phosphate monoliths from an apatitic gel, and silica-HA monoliths via a sol-gel process. These low temperature processes should provide bioceramics able to perform bioactivity and bio-resorption in-vivo, and could prefigure a drug or other therapeutic ions-delivery disposals for filling bone defects in maxillofacial or orthopedic surgery
Decision Support Systems XII: Decision Support Addressing Modern Industry, Business, and Societal Needs. ICDSST 2022-8th International Conference on Decision Support System Technology, Thessaloniki, Greece, from may 23-25, 2022: proceedings of the 8th International Conference on Decision Support System Technology
No full textInternational audienceThis book constitutes the proceedings of the 8th International Conference on Decision Support Systems Technologies, ICDSST 2022, held during May 23–25, 2022.The EWG-DSS series of International Conference on Decision Support System Technology (ICDSST) is planned to consolidate the tradition of annual events organized by the EWG-DSS in offering a platform for European and international DSS communities, comprising the academic and industrial sectors, to present state-of-the-art DSS research and developments, to discuss current challenges that surround decision-making processes, to exchange ideas about realistic and innovative solutions, and to co-develop potential business opportunities.The main aim of this year’s conference is to investigate the role DSS and related technologies can play in mitigating the impact of pandemics and post-crisis recovery. The 15 papers presented in this volume were carefully reviewed and selected from 46 submissions. They were organized in topical sections as follows: decision support addressing modern industry; decision support addressing business and societal needs, and multiple criteria approaches
Feasibility of a Full-Field Measurements-Based Protocol for the Biomechanical Study of a Lumbar Belt: A Case Study
No full textInternational audienceLow back pain represents a major economic and societal challenge due to its high prevalence. Lumbar orthoses are one of the recommended treatments. Even if previous results showed their clinical effects, the detailed mode of action is still poorly known, making the device design difficult. A renewed instrumentation and experimental protocol should bring better insight into the lumbar brace–trunk mechanical interaction. This instrumentation should give detailed information on the basic physical or geometrical parameters: the pressure applied on the trunk, the body shape and the strain in the belt. The principal objective of this study was to propose and validate a new measurement protocol, based on pressure mapping systems and full-field shape and strain measurement. The feasibility of the protocol was tested along with its validity and repeatability. The influence of various parameters, which could cause changes in the measurements, was tested with six different belt configurations on one subject. Measurements were also performed to study the impact of posture on pressure and strain. Both pressure and strain appeared to be asymmetric from left to right. The pressure applied by the lumbar belt on the back varies with breathing and with posture. This study showed that full-field measurements were necessary to render the high variability of pressure or strain around the trunk, under recommendations of their use to guarantee a satisfying repeatability
Fragmented Landscape Generator (flsgen): a neutral landscape generator with control of landscape structure and fragmentation indices
No full textInternational audience1. Neutral landscape models have many applications in ecology, such as supporting spatially-explicit simulations, developing, and evaluating landscape indices. However, current approaches provide few options to produce large landscapes with controlled composition and fragmentation indices. 2. We introduce flsgen (Fragmented Landscape Generator), a new neutral landscape generator that address this limitation by providing a high level of control over 14 landscape indices. The main novelty of flsgen is the decomposition of landscape generation into two steps: the solving of a constraint satisfaction problem and the generation of a landscape raster with a stochastic algorithm. The latter relies on a continuous environmental gradient that influences the landscape’s spatial configuration. 3. flsgen can generate fine-grained artificial landscapes in small amounts of time, which makes it suited to produce large landscape series systematically. We demonstrate the features of flsgen through three illustrative use cases. 4. flsgen is a practical and efficient tool that expand the current possibilities of neutral landscape models and widen their potential applications. To facilitate its uptake, flsgen is available as free and open-source software through a Java API, a command-line interface, or an R package
Process Parameters Optimization, Characterization, and Application of KOH-Activated Norway Spruce Bark Graphitic Biochars for Efficient Azo Dye Adsorption
No full textInternational audienceIn this work, Norway spruce bark was used as a precursor to prepare activated biochars (BCs) via chemical activation with potassium hydroxide (KOH) as a chemical activator. A Box–Behnken design (BBD) was conducted to evaluate and identify the optimal conditions to reach high specific surface area and high mass yield of BC samples. The studied BC preparation parameters and their levels were as follows: pyrolysis temperature (700, 800, and 900 °C), holding time (1, 2, and 3 h), and ratio of the biomass: chemical activator of 1: 1, 1.5, and 2. The planned BBD yielded BC with extremely high SSA values, up to 2209 m2·g−1. In addition, the BCs were physiochemically characterized, and the results indicated that the BCs exhibited disordered carbon structures and presented a high quantity of O-bearing functional groups on their surfaces, which might improve their adsorption performance towards organic pollutant removal. The BC with the highest SSA value was then employed as an adsorbent to remove Evans blue dye (EB) and colorful effluents. The kinetic study followed a general-order (GO) model, as the most suitable model to describe the experimental data, while the Redlich–Peterson model fitted the equilibrium data better. The EB adsorption capacity was 396.1 mg·g−1. The employment of the BC in the treatment of synthetic effluents, with several dyes and other organic and inorganic compounds, returned a high percentage of removal degree up to 87.7%. Desorption and cyclability tests showed that the biochar can be efficiently regenerated, maintaining an adsorption capacity of 75% after 4 adsorption–desorption cycles. The results of this work pointed out that Norway spruce bark indeed is a promising precursor for producing biochars with very promising properties
