Papers by Jean-Claude Hennet
Springer series in advanced manufacturing, Oct 16, 2013
This chapter analyzes the process of forming a coalition within a corporate network. The objectiv... more This chapter analyzes the process of forming a coalition within a corporate network. The objective of the partner companies is to create a multi-stage manufacturing system, which generates a chain of increased value from raw materials to end-user market. This process is studied by cooperative game theory, through the key problems of maximizing the total profit and distributing it among the members of the coalition. To construct a pay-off policy that is both stable and fair, the study proposes to represent the productive resources of the firms not only by their capacity, but also by the WIP generated by product flows. The proposed profit sharing rule is then constructed from the dual of the profit maximization problem. It is both efficient and rational, with more fairness than the Owen set policy of classical Linear Production Games.
IFAC Proceedings Volumes, Apr 1, 2004
This study analyzes an inventory management problem within a supply chain. An enterprise whose st... more This study analyzes an inventory management problem within a supply chain. An enterprise whose stock is replenished by outside suppliers faces random demands for one of its products. The objective is to minimize the expected total cost for the enterprise, which includes the holding and shortage costs. The enterprise must choose the suppliers to whom the orders are sent while minimizing the costs. If the enterprise sends all the orders to the same supplier during the planning period, the supplier who has the maximum service rate must be chosen. In the case of choosing possibly different suppliers for different orders, the supplying system is modelled as an open queuing network and explicit expressions are obtained for the average total cost as a function of the decision variables. Numerical results measure the advantage for the producer to use different suppliers instead of a single one.
International Journal of Logistics Systems and Management, 2016
The purpose of this paper is to compare the effectiveness of two approaches for inventory managem... more The purpose of this paper is to compare the effectiveness of two approaches for inventory management in a network of retailers. In the individual approach, each retailer independently orders products from the central warehouse at the beginning of the commercial season, when demand is still uncertain. In the integrated approach, the first stage of ordering from the central warehouse is followed by a process of exchange of products between the retailers once real demands are known with certainty. In terms of modelling and methodology, the basic newsvendor model specialises in the case of discrete probability distributions. The compound distribution generates scenarios that allow solving the stochastic optimisation problem of expected profit maximisation as a deterministic problem. A real industrial example provides a numerical application in this study. The results indicate that the integrated approach increases the total expected profit of the network. It also improves the service level for customers and the efficiency of transport.
Nonlinear Analysis-theory Methods & Applications, Nov 1, 2005
This paper addresses the problem of symbolic optimization of functions defined over finite discre... more This paper addresses the problem of symbolic optimization of functions defined over finite discrete sets. In connection with the main goal of the paper, several interesting properties of boolean functions are discovered, related to a particular function representation, over the vertices of hyper-cube {−1, 1} k. The symbolic algorithm which is proposed has essentially the same computational complexity as the classical search methods, with the advantage of its possible application to parametric optimization.
IFAC Proceedings Volumes, Jul 1, 2001
Eugimio n. Castelan {.CM{/DAS/{Tf'SC, 88040-(}OO-Plorirmopolz8 (S.C)nmzil, Pax +55 (48) 3.11 .9.9... more Eugimio n. Castelan {.CM{/DAS/{Tf'SC, 88040-(}OO-Plorirmopolz8 (S.C)nmzil, Pax +55 (48) 3.11 .9.9 :14, e-mail eugenio~lcmi.ufsc.br Ahstrad. The st.udy addresses t.hf' prohlem of ('ollt.rolling a linear discret.e-time system suh.iecl 1.0 bounded addil.ive distllrhances , st.epwise-ill-time constraint.s and hounds on performance requiremenl.s. Tll(' object.iw is t.o propose a design t.echnique which can he efficient.ly applied to large scale systems. The design prohlem is solved locally hy ("ollst.ruct.ing a domaill of initial stat.es which is hot.h posit.ively invariant and illc!udf:d ill t he domain of sat.isfactory performance. This posit.ively invariant. domain is polyhedral. It. is const,ruded hy eig('nsl rllct.ure assignment, and an index of the size of t.he domaill is optimized hy Lin('ar Programming. Copyrz.qht @2001 {FAC
International Journal of Production Research, Feb 5, 2014
The paper addresses a supply network design problem in which a set of enterprises decides to orga... more The paper addresses a supply network design problem in which a set of enterprises decides to organize itself as a multi-stage supply network through resource sharing and production coordination. A biform game theory formulation of the problem is obtained by combining the cooperative subgame in the manufacturing network with the strategic subgame of the Stackelberg type between the retailer and the manufacturers' network. As a result of the interaction between these two subgames, a new type of cooperative game, the Quadratic Production Game (QPG), is formulated to describe the supply network design problem under anticipated price elastic demands from the market. The key problem of coalitional stability is addressed through the properties of rationality and fairness of the profit sharing agreement. We prove that the game is non convex in general and thus the fair solution given by the Shapley value allocation is not always rational. The main results indicate that the coalition stability can be reinforced by minimizing the number of partner enterprises achieving the maximal expected profit and applying a profit sharing policy with guaranteed fairness restricted to the member enterprises.
Solar Energy, 1984
Conception, evaluation and real time control of solar "power tower" systems require the use of fa... more Conception, evaluation and real time control of solar "power tower" systems require the use of fast and accurate computer programs for calculating the flux density distributions on the receiver. Since the classical methods of "cone optics" and "hermite polynomial expansion" have some limitations of speed and accuracy, we have built an analytical model for calculating the convolution of the solar brightness distribution with the principal image of a heliostat (i.e. the fictive image for a "point sun"). We first characterize a principal image of a focusing heliostat by its shape and its geometrical concentration factor. Then this image is projected back onto the central plane (which passes through the center of the mirror), and considered as a flat reflecting surface. And the problem is reduced to density calculation for a flat heliostat. For each point of the receiver, the density of flux reflected by a heliostat is obtained by direct resolution of a convolution integral. The different formulations used to express the density function correspond to the various types of intersections between the image of the solar disk for the considered point and the principal image of the heliostat. Confrontation of this method with a program based on "cone optics" shows a good concordance of results and a strong decrease of computation time. We want to apply this method to the existing "THEMIS" solar plant built in France and to compare our results with real observations. Our density calculation programs will help conceiving fields of focusing heliostats for a new generation of power systems (gaz turbine systems).
Computers in Industry, 2004
In continuous production networks such as oil fields, global optimization requires to optimize no... more In continuous production networks such as oil fields, global optimization requires to optimize not only production mechanisms, but also scheduling mechanisms for continuous flows of products through the different units that belong to the production complex. Two functions must be performed in order to efficiently operate a production network: coordination to ensure coherence and consistency of states, controls and decisions; and selection of the best configuration and set-points to run the process according to optimization criteria. The hybrid nature of this problem stems from the continuous processes involved and the use of discrete decision mechanisms to supervise the global behavior of the production process. To increase reactivity and reduce complexity in the global control scheme, each component of the system is described as a holon and the complex forms a holarchy. Each holon interacts with the other holons to contribute to the system's optimization. Interaction is based on negotiation processes performed by agents that represent each holon.
Method of relaxation applied to optimization of discrete systems
HAL (Le Centre pour la Communication Scientifique Directe), 2005
Logistics - 3
In this regular session we aim to highlight some different research areas with regards to the log... more In this regular session we aim to highlight some different research areas with regards to the logistics problems.
Resolution of a quadratic combinatorial problem by dynamic programming
Lecture Notes in Control and Information Sciences
Algoritmo SCDO en el Principio de Mínimo y la programación dinámica definida sobre dominios finitos
Proceedings of the 11th International Conference on Integrated Modeling and Analysis in Applied Control and Automation (IMAACA 2018)
This research concerns the formulation of models and methods for supply chains risk analysis. An ... more This research concerns the formulation of models and methods for supply chains risk analysis. An ontological approach using the KOD method (Knowledge Oriented Design) has been implemented to clearly identify relationships between the concepts of supply chain, risk, vulnerability and disturbances (critical scenarios). As a result, conceptual models of supply chains facing risk situations and critical scenarios are proposed. From the resulting conceptual models and mathematical models proposed in the literature, a multi-stage supply chain model using ARIMA models incorporating the randomness of the demand has been elaborated. In order to adapt this model to scenario criticality, constraints on orders and inventories have been taken into account. Under critical disturbances on information flows (demand) and physical flows (quality of the product supplied), constraints can be reached and supply chain behaviours can evolve toward critical dynamics or even become unstable. Supply chain vulnerabilities has been assessed and discussed.
IFAC Proceedings Volumes, 1990
Stable dynamic systems admit positively invariant domains associated to their Lyapunov functions.... more Stable dynamic systems admit positively invariant domains associated to their Lyapunov functions. Conversely, some domains can be made positively invariant for systems with state feedback controllers designed in such a way that some associated definite non-negative functions are bound to decrease. In particular, this approach can be used to establish conditions on the gain matrix for Linear Constrained Regulation Problems (LCRP). We propose a fixed and a variable regulator easy to compute through linear programming, for a class of con trained linear systems.
IFAC Proceedings Volumes, 2000
The study proposes to decompose a multistagc production planning problem into two subproblems : a... more The study proposes to decompose a multistagc production planning problem into two subproblems : a deterministic one over a short-term horizon, which can be solved by Linear Programming, and a stochastic one, over a longer time-horizon, solved as a closed-loop control problem through positive invariance of a polyhedral domain. This decomposition is based on different assumptions on demands in the two subproblems : perfect knowledge in the first subproblem , stationarity and randomnf'-ss in the second one. Copyright ©2DDD [FAC Keywords.
International Journal of Logistics Systems and Management, 2016
The purpose of this paper is to compare the effectiveness of two approaches for inventory managem... more The purpose of this paper is to compare the effectiveness of two approaches for inventory management in a network of retailers. In the individual approach, each retailer independently orders products from the central warehouse at the beginning of the commercial season, when demand is still uncertain. In the integrated approach, the first stage of ordering from the central warehouse is followed by a process of exchange of products between the retailers once real demands are known with certainty. In terms of modelling and methodology, the basic newsvendor model specialises in the case of discrete probability distributions. The compound distribution generates scenarios that allow solving the stochastic optimisation problem of expected profit maximisation as a deterministic problem. A real industrial example provides a numerical application in this study. The results indicate that the integrated approach increases the total expected profit of the network. It also improves the service level for customers and the efficiency of transport.
Proceedings of the 39th IEEE Conference on Decision and Control (Cat. No.00CH37187)
ABSTRACT This work extends the concept of self-bounded (A,B)-invariant subspaces to convex polyhe... more ABSTRACT This work extends the concept of self-bounded (A,B)-invariant subspaces to convex polyhedral sets. Self-bounded (A,B)-invariant polyhedra are defined and characterized. Necessary and sufficient conditions under which a given polyhedron is self-bounded are established in the form of linear matrix relations. It is then shown that the class of self-bounded sets contained in a given region has an infimum, that is, a self-bounded set which is contained in any set of this class. The infimal set is characterized and a numerical method is proposed for its computation in the polyhedral case. It is also shown how these results can be extended to systems subject to control constraints and bounded additive disturbances. A numerical illustrative example is finally presented
Proc. IFAC-SSC’97, 1997
The purpose of this paper is to extend some results recently obtained on the l 1 optimal control ... more The purpose of this paper is to extend some results recently obtained on the l 1 optimal control problem using the controlled invariance approach. Feasibility of a given l 1 perfor-mance bound is tested by checking the non-emptiness of the supremal controlled invariant domain with stability contained in the admissible performance domain.
This paper presents some results on the ability to control time-invariant linear systems by linea... more This paper presents some results on the ability to control time-invariant linear systems by linear state feedback under some constraints on the state vector or on the control vec-tor. The proposed unified framework provides solutions for continuous-time systems as well as discrete-time systems. This approach relies on the positive invariance properties of some domains of the state-space for stable closed-loop systems.
European Journal of Control, 1999
One simple way to prove it is to use Proposition 4.1 of the paper noting that the existence of &#... more One simple way to prove it is to use Proposition 4.1 of the paper noting that the existence of 'T > o such that 5 is (A, B)-invariant for the Euler Approximating System is equivalent to the invariance of the continuous-time system [4]. The conditions (1) and (2) apply under control constraints, because U is the control-at-vertices (2) AX + BU = XM, fM :5 0 The proposed conditions (7)-(9) of the paper consider the 'plane' representation of a polyhedral set 5 = R[G, p] and are both necessary and sufficient for its (A, B)-invariance. There exist dual conditions which involve the vertex representation of 5 as long as such a set is compact and includes the origin [1]. Assume that X is the matrix whose columns Xi are the vertices of 5:
Uploads
Papers by Jean-Claude Hennet