Feature-based product modelling

Top PDF Feature-based product modelling:

Beyond Boolean Product-Line Model Checking: Dealing with Feature Attributes and Multi-Features

Beyond Boolean Product-Line Model Checking: Dealing with Feature Attributes and Multi-Features

A. Running Example CFDP is a highly-configurable deep-space file transfer pro- tocol [16]. In the past, our team helped Spacebel, a Belgian company, to develop an implementation of CFDP as an SPL [17]. The original CFDP FM has 98 features. Here, we consider a small subset of the protocol, i.e., the ack modes it offers. The corresponding sub-FM has 14 features and yields 1,058 different valid products. We had to limit ourselves to this subset because in addition to the variability, we had to model the behaviour of the protocol’s features, which is far more complex. We did that based on the protocol specification [16] and experimented with various SPL be- havioural modelling languages. This turned out to be a difficult and time-consuming activity. The resulting models describe a communication scenario where an entity (e.g., a spacecraft) has to transfer a message to another one. Depending on the features of the protocol’s instance in each entity, properties like successful transmission may or may not be satisfied during the transaction. The final model has 1,812,652 states.
En savoir plus

11 En savoir plus

A DSL-based Approach to Product Derivation for Software Product Lin

A DSL-based Approach to Product Derivation for Software Product Lin

defined in the UML metamodel and well-formed rules. The transformation is realized in the Atlas Transformation Language (ATL). ArchFeature a recent work proposed by Gharibi et al. (2016 ) which is a PLA modelling approach equipped with a graphical environment. ArchFeature integrates feature specification, PLA, and their relationships in a single monolithic architecture model. This is enabled by extending an existing XML-based architecture description language (ADL), xADL that is mostly used for modeling a single system’s architecture consisting of components and connections. It includes a graphical modelling environment that can (1) automatically capture, maintain, and visualize the feature-PLA relationship, (2) encapsulate variability modelling from the user, and (3) support automatic derivation of architecture instances from the PLA. ArchFeature is integrated in ArchStudio, an Eclipse-based architecture development platform.
En savoir plus

6 En savoir plus

Model-based (Mechanical) Product Design

Model-based (Mechanical) Product Design

increased the involvement of several experts in the solution selection. The design process has then to be centred on experts knowledge. New issues are then related to the complexity of managing that knowledge via computer-supported tools: Knowledge synthesis versus form feature modelling For almost 30 years CAD systems have been developed and improved to currently reach very powerful features to support product’s shape modelling. They are actually presented and used as one of the central systems that make the design process a geometric cen- tric approach. This approach has shown its great interest in industry to tackle the problem of digitizing hand-done drawing, to improve the CAD-CAM links. Nowadays, the CAD model also finds an interest to improve the digital mock- up used during a decision making process. However current CAD systems are not able to manage all the knowledge related to the product definition. This information as mentioned in [9] has to be related to the whole lifecycle (from re- quirement specifications to dismantling information). The product, and its CAD model, is then defined, as far as possible, taken into account “X” constraints as assumed in a DFX (Design For X) approach. CAD model (i.e. form feature) has then to be generated from knowledge synthesis approach [24] seen as knowledge transformation in the proposed MDE environment.
En savoir plus

16 En savoir plus

DFM synthesis approach based on product-process interface modelling. Application to the peen forming process.

DFM synthesis approach based on product-process interface modelling. Application to the peen forming process.

Considering that the manufacturing domain can be extended to other product lifecycle phases (e.g. assembly, recycling, dismantling, etc.), the assumption is that the design process should then be centred on multiple-views product modelling and expert analyses instead of being CAD-centred. One of the main limits of that CAD-centred approach remains in the unique product breakdown that does not reflect the design intentions of every expert designers involved in the design group. Figure 3 shows the form features breakdown used to obtain the CAD model of a Wing Cover. Obviously, this breakdown gives the way to draw the entire form but does not represent what should or could be the real manufacturing process plan. It does not make any sense for the engineers in charge of the manufacturing activities. For instance, the three slots are designed using the “extrusion” feature based on a 2D sketch while they should be manufactured as three machining operations. The information structure should therefore include both the manufacturing and form breakdowns (i.e. multiple-views).
En savoir plus

9 En savoir plus

A product feature-based user-centric product search model

A product feature-based user-centric product search model

The User Engagement Metric. The integration of the user engagement com- ponent is driven by the main aim of e-commerce application which consists in increasing the user conversion rate. Although the user engagement should be derived in accordance with the application goal, such metric emphasizes the positive aspect of the interaction [ 14 ]. In the setting of a Web application, the user’s engagement is often associated with his/her interactions including visits, clicks, comments, recommendations, etc. In accordance with the search scenario of this model, we propose to consider users’ interactions willing to be noticed after a product search. For instance, post-task evidence sources of interaction could be result clicks, product ratings, favorites, or users’ actions. The latter aim at bookmarking wishlist, adding to basket, and/or pushing the product. Unfor- tunately, these data are not available for this edition of LL4IR track. Thus, we estimate the user engagement by the number of social interactions, namely “Like” and “Share” actions, generated on the Facebook 5 social media platform. In order to get the social engagement toward a product, we first identify significant Web resources that represent a product, typically Web pages with technical description. In this aim, we used the product name as a query for exact search on a Web search engine. We assume that the set of top k resources significantly represent the product and their underlying users’ interactions may be associated with the product. With this in mind, let R p = {r 1 , r 1 · · · r k } be the set of resources that mention product p. likes(r i ), respectively shares(r i ), expresses the number of Facebook likes, respectively Facebook shares, obtained by a particular resource r i . Please, note that likes and shares are obtained by sending the URL of resource r i to the Facebook API.
En savoir plus

14 En savoir plus

Robust Object Pose Estimation From Feature-Based Stereo

Robust Object Pose Estimation From Feature-Based Stereo

O NE OF the main challenges inherent in using images from a large number of viewpoints is the issue of camera pose estimation. For three-dimensional (3-D) reconstruction to be possible, the location and orientation of the cameras at differ- ent capture instants must be accurately known. Several applica- tions can benefit from the knowledge of the position of a camera with respect to some rigid reference frame. Among them are virtual or augmented reality systems, scene reconstruction, object modeling, and robotics. In a video sequence in which a camera is moving inside a fixed environment, keeping track of the camera/object’s respective positions can be challenging. In the case of a moving camera, a workable solution consists of installing calibration targets, precisely registered with respect to a global reference frame. By having them visible inside the scene, it becomes possible to compute the camera position as the camera moves with respect to the global reference frame [5]. Alternately, in an object-based solution, a computer-aided design (CAD) model of the observed objects can be created, and by registering that model to the observations, the spatial relation between the scene and the camera can be determined [1]. In practice, however, these solutions are not always feasible. It is therefore desirable to develop a method to compute the camera
En savoir plus

13 En savoir plus

A Feature-based Survey of Model View Approaches

A Feature-based Survey of Model View Approaches

In this paper, we provide orientation via an extended study of the state-of-the-art in the area of model viewpoints/views. We believe this research line is becoming more and more crucial not only for the modeling community, but also from a more general software and systems engineering perspective. Thus, we motivate why many future applications of modeling will depend, one way or the other, on some kind of mechanism for dealing with viewpoints/views on models. Furthermore, we describe what are the solutions available at the time of writing and how they currently compare to each other. To this intent, we categorize them according to a detailed feature model that aims to shed more light on this topic.
En savoir plus

2 En savoir plus

ICA-based sparse feature recovery from fMRI datasets

ICA-based sparse feature recovery from fMRI datasets

Fig. 3. ICs estimated from fMRI data and thresholded using MELODIC’s mixture model, and our multivariate thresholding procedure. Top rows: IC detecting the primary visual areas. Bottom rows: IC representative of a vascular artifact. 5. CONCLUSION This contribution presents a procedure for thresholding ICA patterns of fMRI time series to recover sparse sources using a multivariate model of spatially-sparse brain activity that does not rely on correlating with external stimuli. From a practical point of view, the main improvement over existing ICA-based methods for fMRI is that non-neuronal patterns are rejected as they do not correspond to very salient features. We have val- idated on simulated data and resting-state fMRI data that the procedure can yield exact control of the false positive rates for p > 10 −2 and achieves better sensitivity/specificity trade- offs than the current state-of-art fMRI ICA support-selection procedures. Control of false detections and consistency of es- timation on noisy data is important for clinical and medical re- search applications of resting-state fMRI. Our procedure can be understood as outlier detection with projection pursuit, as proposed by Gnanadesikan and Kettenring [9], using ICA.
En savoir plus

5 En savoir plus

Multiscale modelling of low-pressure CVD of Silicon based materials in deep submicronic trenches: a continuum feature
scale model

Multiscale modelling of low-pressure CVD of Silicon based materials in deep submicronic trenches: a continuum feature scale model

and the studied 2D trench initial profiles [(a), (b)] Figure 1 shows the two initial profiles of 2D trench structures we have studied. The structures (a) and (b) present respectively an AR of 17.5 and 30 with an initial aperture of 0.8µm and 1.5µ m in width. The structure (b) is deeper than the structure (a) (it has been subjected to a longer etching time), and presents a larger bowing (‘local’ enlargement) at its top and a sharper ending at its bottom. The study of bowings and sharp endings presents a strong interest because it is a remnant default of etching, occurring once one try simultaneously to reach deeper and deeper structures and shrink the whole surface structure area within a smaller and smaller substrate surface area. Usually, these phenomena are typical of very large AR structures of a hundred of microns in depth but, in order to reduce the time cost of the dynamic mesh procedures development we have simulated them for the 45µ m depth structure (b). In this conditions, our model must be able to manage dynamic geometrical domains evolving very strongly with time due to deposition, especially when they present particularly marked etching defects (as those previously discussed but also numerous other ones, intentional or not). So it was important that the dynamic triangular mesh of our continuum feature scale model can be subjected to important deforming processes without deteriorate the mesh quality and without generating negative volume cells near the trench surface. However, FLUENT
En savoir plus

8 En savoir plus

Modelling and Reasoning with Software Product Lines with Design Choices

Modelling and Reasoning with Software Product Lines with Design Choices

Sommaire Les gammes de produits logiciels (Software Product Lines)(SPLs) permettent de gérer la variabilité qui apparaît dans les familles de modèles logiciels connexes en raison des va- riations des besoins des clients. Durant la conception de leurs modifications, les ingénieurs doivent considérer plusieurs conceptions de SPLs alternatives. Cependant, sans informa- tions complètes sur les exigences de qualité souhaitées pour le SPL final, les ingénieurs sont confrontés à une incertitude quant au choix de la conception appropriée. Les formalismes et techniques existants ne conviennent pas à la modélisation et au raisonnement sur l’es- pace à deux dimensions défini par la variabilité et les choix conceptuels. Nous proposons une approche pour modéliser l’incertitude de conception dans les SPLs et, pour analyser et comprendre l’impact des choix conceptuels sur la qualité des SPLs, exprimé comme des propriétés. Nous définissons formellement les Gammes de produits logiciels avec des choix conceptuels (SPLDCs)(Software Product Lines with Design Choices) et nous décrivons une procédure pour les analyser et fournir une rétroaction appropriée aux ingénieurs basée sur l’ordre partiel des catégories de propriétés de SPLDC. Nous illustrons l’applicabilité de notre approche en utilisant un exemple entirément élaboré qui montre le type de rétroactions nuan- cées nécessaire pour des analyses significatives des SPLs en présence de choix conceptuels. Pour évaluer l’évolutivité de notre approche, nous utilisons notre approche sur de nombreux SPLDC et enregistrer des temps d’exécution.
En savoir plus

84 En savoir plus

A Feature-based Survey of Model View Approaches

A Feature-based Survey of Model View Approaches

Concerning our feature-based methodology for discussing different model view approaches, we follow a best-practice in publishing in the model-driven engineering domain. In previous works, several feature-based surveys have been already published [3, 22, 32]. In addition, related surveys are available which discuss prerequisites of model view approaches such as model transfor- mations in general [54] or particular model transformation approaches [33]. However, to the best of our knowledge, we are not aware of any survey about model view approaches (except a preliminary work on a tool-oriented taxon- omy for combining domain-specific languages with view-based modeling ap- proaches [30]). In particular, current language workbenches have been evalu- ated regarding the specification of views for domain-specific models [30]. Our work is orthogonal in the sense that we are interested in dedicated model view approaches which may build on language workbenches (such as EMF). How- ever, we do share small parts of definitions between our feature model and the taxonomy presented in this evaluation [30].
En savoir plus

33 En savoir plus

Video Watermarking Based on Interactive Detection of Feature Regions

Video Watermarking Based on Interactive Detection of Feature Regions

III. PROPOSED METHOD The study of watermarking and crowdsourcing recent works shows that these two areas can be combined to propose a new watermarking approach which presents a high level of robustness against the most important attacks. In fact, our work aims to develop new robust approaches to introduce signatures in videos. Our idea consists in a first step to understand the visual content of the original video and then to select feature regions to embed signature. To achieve our goal, crowdsourcing technique will be used. Although, the concept of crowdsourcing is based on sharing media to the public, this can cause confidentiality problems and can damage watermarking process. To avoid this problem, we thought to generate a video summary and to share it to a fixed number of selected users.
En savoir plus

5 En savoir plus

Feature in product engineering with single and variant design approaches. A comparative review

Feature in product engineering with single and variant design approaches. A comparative review

Feature-based design has had more influence on the detailed design and process planning activities of product development processes. Currently, feature-based CAx systems are considered the state-of-art technology for product modeling [10, 12]. At first, features were used to easy the modeling of similar components by developing CAx libraries of pre-defined elements (application feature), but their potential to integrate multiple CAx applications was soon recognized. From this perspective, a feature, like a hole or a pocket, is an engineering meaningful set of related surfaces in a computer model associated to some parameters. Later, features were used to anchor other kinds of qualitative information useful for modeling tasks including non-geometric (product) properties. By developing further this view, today’s features also include qualitative characteristics like colors.
En savoir plus

9 En savoir plus

Intelligent feature based resource selection and process planning

Intelligent feature based resource selection and process planning

Mohsen.Sadeghi@proctonlabs.org , William.ziz@proctonlabs.org Abstract: This paper presents an intelligent knowledge-based integrated manufacturing system using the STEP feature-based modeling and rule based intelligent techniques to generate suitable process plans for prismatic parts. The system carries out several stages of process planning, such as identification of the pairs of feature/tool that satisfy the required conditions, generation of the possible process plans from identified tools/machine pairs, and selection of the most interesting process plans considering the economical or timing indicators. The suitable processes plans are selected according to the acceptable range of quality, time and cost factors. Each process plan is represented in the tree format by the information items corresponding to their CNC Machine, required tools characteristics, times (machining, setup, preparatory) and the required machining sequences. The process simulation module is provided to demonstrate the different sequences of machining. After selection of suitable process plan, the G-code language used by CNC machines is generated automatically. This approach is validated through a case.
En savoir plus

13 En savoir plus

Requirement Mining for Model-Based Product Design

Requirement Mining for Model-Based Product Design

1.3.1 An NLP pipeline to extract requirements As we will be showing in Section 2.1, although modal verbs are reliable lexical features to identify requirements, the literature review shows that legal terms, as well as the verbs require, need and their derived forms, must be considered too. However, we contend that the solution is still sub-optimal. Indeed, most companies are not aware of, or do not want to follow, systems engineering principles defined in the INCOSE guidelines or IEEE standards, but instead, prefer ‘to colour outside the line’. It is therefore usual for them to use verbs that not only express a requirement or a need, but also an expectation, a wish, a hope, a desire, etc. Then, verbs including, but not limited to, expect, need, wish, want or desire also need to be considered. In addition, as there are diverse writing styles, the use of syntactic rules to identify requirements is very likely to be too restrictive in some cases. To follow the principle of parsimony, we propose to enhance the existing lists of prescriptive verbs and to stick to a rule-based classifier that does not sacrifice recall for the sake of precision.
En savoir plus

29 En savoir plus

Curb-intersection feature based Monte Carlo Localization on urban roads

Curb-intersection feature based Monte Carlo Localization on urban roads

surface. In other studies, researchers refer to local features for high precision localization. In [14], lane markers are extracted to reduce localization error. But due to the fact that lane makers just carry lateral position information, longitudinal localization error can be reduced only when the road cur- vature is big. In [8], a novel “Virtual 2D Scans” method is proposed by making use of the building outlines as features. A simplified 2D line feature map is generated beforehand as prior knowledge. However, possible lack of building features and slow update rate limits its effectiveness.
En savoir plus

8 En savoir plus

Feature-based map merging with dynamic consensus on information increments

Feature-based map merging with dynamic consensus on information increments

The consensus filters literature is greatly wide. A review of the most relevant results can be found in [46] and the references therein. Many recent works consider specific variations of the consensus problem to cope with communication delays [49], or stochastic commu- nication noises [34]. Most of the works in distributed consensus address the static case, i.e., consensus is achieved on a value that depends on the initial conditions of the system. Fewer works [10, 18, 43, 45, 48, 57, 59] con- sider the dynamic case, where nodes measure a variable along time, and the goal is to track the average of this variable. In map merging scenarios, dynamic consen- sus strategies are more appealing, since the local maps of the robots will change, and it would be desirable to track the global merged map. Several dynamic con- sensus methods [18, 43, 45, 48] consider continuous-time systems, and thus they are better suited for systems based on the observation of the states of the neigh- bors, instead of on communicating the states (in our case, the maps). [59] uses discrete-time communication, but it considers that nodes measure a local continuous physical process. On the other hand, [10, 57] track the average of inputs that change in a discrete-time way, us- ing discrete-time communications. Thus, they are bet- ter suited to the problem of map merging, where the local maps are modified at discrete time instances.
En savoir plus

16 En savoir plus

A Novel Damage Sensitive Feature Based on State-Space Representation

A Novel Damage Sensitive Feature Based on State-Space Representation

INTRODUCTION For few years, vibration-based Structural Health Monitoring (SHM) process is presented in terms of statistical pattern recognition [1]. The extraction of Damage Sensitive Feature (DSF), in which high dimensional vibration data are compressed into low dimensional vectors, takes a central place in this process. The main challenge is to preserve information about structure condition. For this purpose, the paradigm of nonlinear dynamical systems offers some promising abilities [2, 3, 4, 5]. The information contained in time-series is extracted using reconstructed state-space representation [6]. Among the numerous candidate features, the Lyapunov Exponents (LE) have been particularly studied [7, 8, 9]. They are related to the long term predictability of dynamical system. To be more precise, two trajectories initially close in the state space will diverge as time evolves with an exponential rate proportional to LE. But two remarks can be formulated on their use as damage sensitive feature. First, the practical calculation of LE is very time consuming as trajectories need to be followed for several thousand of time steps [10]. Then, since damage manifests itself as local irregularities in the signal (due to opening and closing of cracks or loosened assembly) it should be more visible if short evolutions of trajectories are considered.
En savoir plus

9 En savoir plus

Dense Feature Matching Core for FPGA-based Smart Cameras

Dense Feature Matching Core for FPGA-based Smart Cameras

from Accelerated Segment Test) feature detector and the BRIEF [ 5 ] (Binary Robust Independent Elementary Features) feature descrip- tor in a customizable FPGA block. The developed blocks were designed to use hardware interfaces based on the AMBA AXI4 in- terface protocol and were connected using a DMA (Direct Memory Access) architecture. The proposed architecture computes feature matching over two consecutive HD frames coming from an exter- nal memory at 48 frames per second. In [ 26 ] a FPGA architecture of SIFT (Scale Invariant Feature Transform) visual descriptor as- sociated to an image matching algorithm was presented. For an efficient FPGA-SIFT image matching implementation (in terms of speed and hardware resources usage), the original SIFT algorithm was optimized as follows: 1) Upsampling operations were replaced with downsampling, in order to avoid interpolation operations. 2) Only four scales with two octaves were used. 3) Dimension of the visual descriptor was reduced to 72 instead of 128 in the orig- inal SIFT formulation. This implementation is able to detect and match features in 640×480 image resolution at 33 frames per sec- ond. More recently, Weberruss [ 25 ] have proposed a FPGA archi- tecture for ORB [ 16 ] (Oriented FAST) descriptor associated to a feature matching algorithm. An "harris corner" detection [ 10 ] was the feature extractor and ORB visual descriptors were computed at each "corner". Finally, the previous features (stored in a 2D Shift Register) and the current features were matched using the hamming distances as discrimination metric. In 2017, Vourvoulakis [ 23 ] pre- sented an FPGA-SIFT architecture for feature matching. In order to achieve high hardware parallelism, procedures of SIFT detec- tion and description were reformulated. At every clock cycle, the current pixel in the pipeline is tested and if it is a SIFT feature, its descriptor is extracted. Furthermore, every detected feature in the current frame is matched with one among the stored features of the previous frame, using a moving window, without breaking the "pixel pipeline". False matches are rejected using RANSAC (Random Sample Consensus) algorithm. The architecture was im- plemented on Cyclone IV. Maximum supported clock frequency was set as 25 MHz and the architecture was capable to process 81 frames per second, considering 640×480 image resolution.
En savoir plus

9 En savoir plus

Multi-domain product modelling: from requirements to cad and simulation tools.

Multi-domain product modelling: from requirements to cad and simulation tools.

Keywords: Functional Analysis, multi-domain product, multi-domain communication, mecatronic. 1 INTRODUCTION Today, in a very competitive industrial context, aggravated by the financial crisis, the fact to design and develop a product following a strict design process is always more and more important. There have been many attempts to draw up models of the design process in systematic steps (French [1]; Pahl and Beitz [2]; Suh [3]; Albano and Suh [4]). Product design is an iterative, complex, decision- making engineering process. It usually starts with the identification of a need, proceeds through a sequence of activities to seek an optimal solution to the problem, and ends with a detailed description of the product. Pahl and Beitz’ systematic approach to engineering design decompose the process in several sequential steps e.g.: product specification, conceptual design, embodiment design and detailed design. Design begins with identification of the customer needs and specification of the requirements. The conceptual design primary concern is the generation of physical solutions to meet the design specification. In the embodiment stage and the detail stage, an overall layout and then the definitive layout of the product are given. It includes the prescriptions of final dimensions and materials, arrangement and shapes of individual components etc. Thus engineering design is seen as an iterative and sequential process, i.e. the stages are achieved one by another.
En savoir plus

13 En savoir plus

Show all 10000 documents...