arXiv subjects

Maurice Pillet

Publications and source records attributed to Maurice Pillet.

9 recordsLinked to original sources

arXiv2020

Modal features for image texture classification

Feature extraction is a key step in image processing for pattern recognition and machine learning processes. Its purpose lies in reducing the dimensionality of the input data through the computing of features which accurately describe the original information. In this article, a new feature extraction method based on Discrete Modal Decomposition (DMD) is introduced, to extend the group of space and frequency based features. These new features are called modal features. Initially aiming to decompose a signal into a modal basis built from a vibration mechanics problem, the DMD projection is applied to images in order to extract modal features with two approaches. The first one, called full scale DMD, consists in exploiting directly the decomposition resulting coordinates as features. The second one, called filtering DMD, consists in using the DMD modes as filters to obtain features through a local transformation process. Experiments are performed on image texture classification tasks including several widely used data bases, compared to several classic feature extraction methods. We show that the DMD approach achieves good classification performances, comparable to the state of the art techniques, with a lower extraction time.

cs.CV↗

arXiv2019

Generative Adversarial Networks for geometric surfaces prediction in injection molding

Geometrical and appearance quality requirements set the limits of the current industrial performance in injection molding. To guarantee the product's quality, it is necessary to adjust the process settings in a closed loop. Those adjustments cannot rely on the final quality because a part takes days to be geometrically stable. Thus, the final part geometry must be predicted from measurements on hot parts. In this paper, we use recent success of Generative Adversarial Networks (GAN) with the pix2pix network architecture to predict the final part geometry, using only hot parts thermographic images, measured right after production. Our dataset is really small, and the GAN learns to translate thermography to geometry. We firstly study prediction performances using different image similarity comparison algorithms. Moreover, we introduce the innovative use of Discrete Modal Decomposition (DMD) to analyze network predictions. The DMD is a geometrical parameterization technique using a modal space projection to geometrically describe surfaces. We study GAN performances to retrieve geometrical parameterization of surfaces.

stat.ML↗

arXiv2018

Quality Prediction in Injection Molding

Injection molded part quality can be improved by precise process adjustment, which could rely on in-situ measurements of part quality. Geometrical and appearance quality (visually and sensory) requirements are increasing. However, direct measurement is often not feasible industrially. Therefore, process control must rely on a prediction of parts quality attributes. This study compares prediction performances of diverse neural networks architectures with "classical" regression algorithms. Dataset comes from inline industrial measurements. Regression was performed on 97 scalar statistical features extracted from multiple acquisitions sources: thermographic images and analog signals. Haralick features were extracted. Convolutional Neural Networks were trained on thermographic images and Long Short Term Memory networks were trained on raw signals. Although the dataset was small, neural networks show better predictions scores than other regression algorithms.

eess.SY↗

arXiv2017

Towards new methods for process adjustments based on parts quality measurements

Thermoplastics injection molding allows the production of complex parts in large series. Industrial quality requirements are increasing. The injection molding process needs to be regulate in order to maintain a working point. There is actually no method to adjust all the parameters of the process in order to optimize the final quality of the product. We can rely on the success of neural networks models to propose a robust self-adaptive adjustment method. Objective is to adjust machine parameters for each cycle, based on measured quality characteristics on the produced part. A classical industrial cycle time is often less than 30 seconds ; therefore challenges are: measuring, computing and setting up machine parameters within this short timeframe. In this presentation we establish a specific literature review, on which will base our experimental works.

eess.SY↗

arXiv2011

Optimisation de la taille de la série: illustration par un cas industriel de sous-traitance mécanique

Reducing costs of manufactured products is one of the key issues of companies. Bar turning companies (mechanical subcontracting companies) are faced with the following dilemma: use a pull strategy or use a push strategy. Instinctively these companies produce more than demand required by customers. This strategy allows them to respond to requests forecasts and reduce their cost of changeover time. These companies make a bet on sales opportunities and think to realize an additional profit. We have tried to find in this study to provide elements to know the limits of this strategy. Our proposal focuses on developing a model to support the decision taking into account the mix of opportunities, economic constraints and mean constraints. This model features the particular importance of high rates of ownership and the risk of not selling. Réduire les coûts de revient des produits fabriqués est une des problématiques essentielles des entreprises d'aujourd'hui. Les entreprises de décolletage (entreprises de sous-traitance mécanique) sont confrontées au dilemme suivant : produire juste la demande client ou produire plus. Instinctivement ces entreprises, dont les temps de changement de série sont élevés, cherchent à produire plus que la demande exigée par le client. Cette stratégie leur permet de répondre à des demandes prévisionnelles et réduire ainsi le coût de revient des produits. Ces entreprises réalisent un pari sur les opportunités de vente et pensent réaliser un gain supplémentaire en réalisant des stocks. Nous avons cherché dans cette étude à fournir des éléments de décision pour connaître les limites de cette règle de gestion. Notre proposition porte sur le développement d'un modèle d'aide à la décision prenant en considération le mixte entre opportunités commerciales, contraintes économiques et contraintes de moyen. Ce modèle souligne l'importance particulière du taux de possession et du risque de non vente.

cs.OH↗

arXiv2011

A method to identify Resources of opportunity: illustration by an industrial case

The aim of this study is to develop a method that would enable the company to prioritize the means contributing the most to its performance. The proposed method is based on the profit margin (an economical performance measure of the company), the customer's risk, the costs of maintenance and the employee's safety. The prioritization method of resources was applied to the data obtained from a small subcontracting business in mechanics. The theoretical foundations of this method are based on a multi-criteria approach using the attribution of criticality indexes for nine criteria linked to the financial loss.

physics.gen-ph↗

arXiv2010

Inertial tolerancing and capability indices in an assembly production

Traditional tolerancing considers the conformity of a batch when the batch satisfies the specifications. The characteristic is considered for itself and not according to its incidence in the assembly. Inertial tolerancing proposes another alternative of tolerancing in order to guarantee the final assembly characteristic. The inertia I2 = \sqrt{δ^2 + σ^2} is not toleranced by a tolerance interval but by a scalar representing the maximum inertia that the characteristic should not exceed. We detail how to calculate the inertial tolerances according to two cases, one aims to guarantee an inertia of the assembly characteristic the other a tolerance interval on the assembly characteristic by a Cpk capability index, in the particular but common case of uniform tolerances or more general with non uniform tolerances. An example will be detailed to show the results of the different tolerancing methods.

cs.OH↗

arXiv2010

Modeling of 2D and 3D Assemblies Taking Into Account Form Errors of Plane Surfaces

The tolerancing process links the virtual and the real worlds. From the former, tolerances define a variational geometrical language (geometric parameters). From the latter, there are values limiting those parameters. The beginning of a tolerancing process is in this duality. As high precision assemblies cannot be analyzed with the assumption that form errors are negligible, we propose to apply this process to assemblies with form errors through a new way of allowing to parameterize forms and solve their assemblies. The assembly process is calculated through a method of allowing to solve the 3D assemblies of pairs of surfaces having form errors using a static equilibrium. We have built a geometrical model based on the modal shapes of the ideal surface. We compute for the completely deterministic contact points between this pair of shapes according to a given assembly process. The solution gives an accurate evaluation of the assembly performance. Then we compare the results with or without taking into account the form errors. When we analyze a batch of assemblies, the problem is to compute for the nonconformity rate of a pilot production according to the functional requirements. We input probable errors of surfaces (position, orientation, and form) in our calculus and we evaluate the quality of the results compared with the functional requirements. The pilot production then can or cannot be validated.

cs.OH↗

arXiv2010

A proposition of 3D inertial tolerancing to consider the statistical combination of the location and orientation deviations

Tolerancing of assembly mechanisms is a major interest in the product life cycle. One can distinguish several models with growing complexity, from 1-dimensional (1D) to 3-dimensional (3D) (including form deviations), and two main tolerancing assumptions, the worst case and the statistical hypothesis. This paper presents an approach to 3D statistical tolerancing using a new acceptance criterion. Our approach is based on the 1D inertial acceptance criterion that is extended to 3D and form acceptance. The modal characterisation is used to describe the form deviation of a geometry as the combination of elementary deviations (location, orientation and form). The proposed 3D statistical tolerancing is applied on a simple mechanism with lever arm. It is also compared to the traditional worst-case tolerancing using a tolerance zone.

cs.OH↗