Pierre Baptiste

Generalized Petri Net Reduction Method

A reduction method of generalized Petri nets is proposed. This method is a generalization of the reduction method which was previously given by Lee-Kwang and Favrel. The proposed method is defined not on the basis of the dynamic behavior but of the structure of the net, and thus the test of reducible subnet can be done by a deterministic approach. The reduction preserves the properties such as liveness, boundedness, and proper termination, and allows easy analysis of generalized Petri nets.

Resolution Of The Cyclic Multi-Hoists Scheduling Problem With Overlapping Partitions

AbstractThis paper deals with scheduling problem in electroplating facilities. To maximise the productivity of electroplating lines, a predictive approach can be used to find a sequence of hoist movements that can be cyclically repeated. This NP-hard problem is known as the Cyclic Hoist Scheduling Problem. An approach for the multi-hoists problem is suggested. The presented model uses Constraint Logic Programming for finding the optimal cyclic schedule for a given hoist assignment.

Hoist scheduling problem: an approach based on constraint logic programming

Scheduling the movements of a hoist in printed circuit board electroplating is known is the hoist scheduling problem (HSP). The objective of the HSP is to find an optimal cyclic sequence of moves for hoists. Results are already available for this problem, using classical operational research tools. They are briefly presented. Most of the proposed solutions are limited to the very specific initial model and are consequently difficult to apply in industrial problems. The constraint logic programming (CLP) approach to the problem is presented. A CLP model of the problem that allows modeling and solving more realistic cases is presented, and the results and the analysis of the solutions are outlined.<<ETX>>

Taking into account the rescheduling problem during the scheduling phase

Abstract The scheduling problem in production management has been studied for a considerable time, and several types of software are used. A problem arises in updating the production planning, or ‘rescheduling’, when an unexpected event occurs in the shop control. Solving this problem is difficult because the implications of such events are usually impossible to forecast. To prevent this problem, we propose to manipulate a set of equivalent schedules during the short time schedule. Then, if an unexpected event prevents realization of a given schedule, it will be possible to find an equivalent one, without full rescheduling. The primary requirement is to find a formal representation of a set of schedules. This has already been explored using CPM graphs with nodes associated to a set of tasks. We propose in this paper to use an extension of such graphs, PQR trees, that represent both precedence and group constraints. We first reiterate the notion of PQR trees. We present methods to take into acc...

A New Algorithm to Select Learning Examples from Learning Data

In almost every area of human activity, the formation of huge databases has created a massive request for new tools to transform data into task oriented knowledge. Our work concentrates on real-world problems, where the learner has o handle problems dealing with data sets containing large amounts of irrelevant information. Our objective is to improve the way large data sets are processed. In fact, irrelevant information perturb the knowledge data discovery process. That is why we look for efficient methods to automatically analyze huge data sets and extract relevant features and examples. This paper presents an heuristic algorithm dedicated to example selection. In order to illustrate our algorithm capabilities, we present results of its application to an artificial data set, and the way it has been used to determine the best human resource allocation in a factory scheduling problem. Our experiments have indicated many advantages of the proposed methodology.

Selection of a Relevant Feature Subset for Induction Tasks

The representation of problems dealt with by machine learning systems use many features, only a few of which may be related to concept designing. Feature selection is the problem of choosing an ideally small subset of necessary features that are sufficient to describe the target concept. It is important both to speed up learning and to improve concept quality. A huge amount of work has been done to select from input data, a subset of the most relevant features. In this paper, a new algorithm of feature pre-processing for induction methods, namely induction of decision trees, and applied to symbolic objects is suggested. It selects a subset of the more relevant features, taking into account feature interaction. It is based both on the DPGoal and ODPGoal of considered variables (features). It is evaluated using three benchmark artificial domains. Then it is compared with Relief [1].

Improving the productivity of electroplating lines by changing the layout of the tanks

The hoist scheduling problem deals with scheduling the movements of a hoist in electroplating facilities. In the mono-product case, cyclic schedules are searched. Various algorithms have been proposed up to now, they allow one to provide the optimal cyclic schedule of any lines. Therefore, the layout of the tanks is generally considered as a fixed data. In this paper the authors show how the layout of the tanks can influence the productivity of the line. The authors propose an approach that combines the research of the optimal cyclic schedule of the hoist movements and the optimal layout of the stations. The results show that this method allows one to really increase the productivity of the line.

Cognitive Science and Project Scheduling: More Realistic Representation

Project scheduling represents the biggest part of project modeling research. More and more hypotheses have to be done on the project to apply scheduling algorithms. From an other point of view, detailed scheduling is not the only problem in project modeling.

Sequences constraint solving in constraint logic programming

This paper deals with consistency techniques over sequences constraints embedded in Constraint Logic Programming CLPS. CLP Sequences constraints are defined over Hereditarily Homogeneous Finite Sets HHFS built on atomic elements to characterise a family of admissible sequences. The relations we are dealing with are classical sets relations (/spl isin/,/spl sub/,=,/spl ne/) and sequencing relations as potential, metric and range constraints. We define the semantics of these relations with a characteristic range function. The consistency techniques used are incremental reduction of the normal form based on a tree like representation called P-Q-R trees. This allows us to reduce the set of admissible sequences before generating solutions.<<ETX>>