Xavier Delorme

GRASP for set packing problems

Abstract The principles of the Greedy Randomized Adaptative Search Procedure (GRASP) metaheuristic are instantiated for the set packing problem. We investigated several construction phases, and evaluated improvements based on advanced strategies. These improvements include a self-tuning procedure (using reactive GRASP), an intensification procedure (using path relinking) and a procedure involving the diversification of the selection (using a learning process). Two sets of various numerical instances were used to perform the computational experiments. The first set contains randomly generated instances, while the second includes instances relating to real problems in railway planning. No metaheuristic has previously been applied to this combinatorial problem. Consequently, we have discussed GRASP’s performances both in relation to lower/upper bounds and to the results obtained with Cplex when such results are available. Our analysis, based on the average performances observed, shows the impact of the suggested strategies, and indicates the configuration that produces the best results.

A MIP approach for balancing transfer line with complex industrial constraints

This paper deals with a novel line balancing problem for flexible transfer lines composed of identical CNC machines. The studied lines are paced and serial, i.e. a part to be machined passes through a sequence of workstations. At least one CNC machine is installed at each workstation. The objective is to assign a given set of operations required for the machining of the part to a sequence of workstations while minimizing the total number of machines used. This problem is subject to precedence, exclusion and inclusion constraints. In addition, accessibility has to be considered. Moreover, the workstation workload depends on the sequence in which the operations are assigned because of setup times related to the change and displacement of tools, rotation of the part, etc. It is a novel line balancing problem, and we highlight its particularities by reviewing the close problems existing in the literature. Then, a mathematical model as a mixed-integer program is suggested. A procedure for computing ranges for variables is given. Experimental computations with ILOG Cplex are reported.

Stability evaluation of a railway timetable at station level

This research deals with a real-world planning problem in railway infrastructure operations. It is part of the RECIFE project, which seeks to develop a decision support software to help evaluate the capacity of a rail junction or station. To this end, the project is working on a timetable optimization model, as well as timetable evaluation modules. This paper presents a module for evaluating timetable stability, which uses an original method based on delay propagation and using shortest path problem resolution. A didactic example and a complete case study applying this method to the Pierrefitte-Gonesse junction are also presented.

Genetic algorithm for supply planning in two-level assembly systems with random lead times

This paper examines supply planning for two-level assembly systems under lead time uncertainties. It is supposed that the demand for the finished product and its due date are known. The assembly process at each level begins when all necessary components are in inventory. If the demand for the finished product is not delivered at the due date, a tardiness cost is incurred. In the same manner, a holding cost at each level appears if some components needed to assemble the same semi-finished product arrive before beginning the assembly at this level. It is assumed also that the lead time at each level is a random discrete variable. The expected cost is composed of the tardiness cost for finished product and the holding costs of components at levels 1 and 2. The objective is to find the release dates for the components at level 2 in order to minimize the total expected cost. For this new problem, a genetic algorithm is suggested. The proposed algorithm is evaluated with a variety of supply chain settings in order to verify its robustness across different supply chain scenarios. Moreover, the effect of a local search on the performance of the Genetic Algorithm in terms of solution quality, convergence and computation time is also investigated.

Multi-objective optimization for inventory control in two-level assembly systems under uncertainty of lead times

Supply planning for two-level assembly systems under lead time uncertainties is considered. It is supposed that the demand for the finished product and its due date are known. The assembly process at each level begins when all necessary components are in inventory. A holding cost at each level appears if some components needed to assemble the same semi-finished product arrive before beginning the assembly at this level. It is assumed also that the component lead time is a random discrete variable. The objective is to find the release dates for the components at level 2 in order to minimize the expected component holding costs and to maximize the customer service level for the finished product. For this new problem, we consider two multi-objective approaches, which are both based on genetic algorithms. They are evaluated with a variety of supply chain settings, and their respective performance is reported and commented. These two heuristics permitted to obtain interesting results within a reasonable computational time.

An Ant Colony Optimisation Algorithm for the Set Packing Problem

In this paper we consider the application of an Ant Colony Optimisation (ACO) metaheuristic on the Set Packing Problem (SPP) which is a NP-hard optimisation problem. For the proposed algorithm, two solution construction strategies based on exploration and exploitation of solution space are designed. The main difference between both strategies concerns the use of pheromones during the solution construction. The selection of one strategy is driven automatically by the search process. A territory disturbance strategy is integrated in the algorithm and is triggered when the convergence of the ACO stagnates. A set of randomly generated numerical instances, involving from 100 to 1000 variables and 100 to 5000 constraints, was used to perform computational experiments. To the best of our knowledge, only one other metaheuristic (Greedy Randomized Adaptative Search Procedure, GRASP) has been previously applied to the SPP. Consequently, we report and discuss the effectiveness of ACO when compared to the best known solutions and including those provided by GRASP. Optimal solutions obtained with Cplex on the smaller instances (up to 200 variables) are indicated with the calculation times. These experiments show that our ACO heuristic outperforms the GRASP heuristic. It is remarkable that the ACO heuristic is made up of simple search techniques whilst the considered GRASP heuristic is more evolved.

Heuristics for railway infrastructure saturation

This research concerns the problem of the evaluation of the railway infrastructure capacity. It is an important question when railway authorities have to choose between different infrastructure investment projects. We developped independently two heuristic approaches to solve the infrastructure saturation problem. The first is based on a constraint programming model which is solved using a greedy heuristic. The second approach identifies the saturation problem as a unicost set packing problem and its resolution is ensured by an adaption of GRASP metaheuristic. Currently, both resolution techniques are not in competition. The goal is to grasp the resolution ability of the heuristics and to analyse the kind of solutions produced. The Pierrefitte-Gonesse junction has been used as experimental support. A software environment allows to simulate several timetables involving TGV, Inter City and Freight trains.

Ergonomics in assembly line balancing based on energy expenditure: a multi-objective model

In many assembly systems, ergonomics can have great impact on productivity and human safety. Traditional assembly systems optimisation approaches consider only time and cost variables, while few studies include also ergonomics aspects. In this study, a new multi-objective model for solving assembly line balancing problem is developed and discussed in order to include also the ergonomics aspect. First, based on main features of assembly workstations, the energy expenditure concept is used in order to estimate the ergonomics level, thanks to a new technique, called Predetermined Motion Energy System, which helps rapidly estimate the energy expenditure values. Then, a multi-objective approach, based on four different objective functions, is introduced in order to define the efficient frontiers of optimal solutions. To complete the study, a simple numerical example for a real case is presented to analyse the behaviour of Pareto frontiers varying several parameters linked to the energy and time value.

A Reactive GRASP and Path Relinking for balancing reconfigurable transfer lines

A line balancing problem for reconfigurable transfer lines with sequence-dependent setup times and parallel machines was studied. These lines are paced and serial, i.e. a part to be machined passes through a sequence of stations. Stations are composed of CNC (Computer Numerical Control) machines. At least one CNC machine is installed at each station. These CNC machines are mono-spindle head machines, hence setup times between operations have to be taken into account. The origins of setup times are various, for example, the necessity to rotate the part, change and displace the tool, etc. Because of setup times, the station workload depends on the sequence in which the operations are assigned to the station. In addition, accessibility constraints have to be considered. The objective consists of assigning a given set of operations as well as machines to a sequence of workstations in order to minimise the total cost of the line. Keeping in mind the industrial importance of this problem and the lack of available methods in the literature tackling it efficiently, we propose a new heuristic based on GRASP combined with Path Relinking. A MIP approach is used to select the sequences of operations on workstations. Numerical experiments are presented and show that the proposed heuristic can provide good solutions even for large-sized instances while requiring a computational time that is fully compatible with a practical application. An industrial case study is also described.

Genetic algorithm for balancing reconfigurable machining lines

We consider the problem of designing a reconfigurable machining line. Such a line is composed of a sequence of workstations performing specific sets of operations. Each workstation is comprised of several identical CNC machines (machining centers). The line is required to satisfy the given precedence order, inclusion, exclusion and accessibility constraints on the given set of operations. Inclusion and exclusion are zoning constraints which oblige or forbid certain operations to be performed on the same workstation. The accessibility constraints imply that each operation has a set of possible part positions under which it can be performed. All the operations performed on the same workstation must have a common part position. Workstation times are computed taking into account processing and setup times for operations and must not exceed a given bound. The number of CNC machines at one workstation is limited, and the total number of machines must be minimized. A genetic algorithm is proposed. This algorithm is based on the permutation representation of solutions. A heuristic decoder is suggested to construct a solution from a permutation, so that the output solution is feasible w.r.t. precedence, accessibility, cycle time, and exclusion constraints. The other constraints are treated with a penalty approach. For a local improvement of solutions, a mixed integer programming model is suggested for an optimal design of workstations if the order of operations is fixed. An experimental evaluation of the proposed GA on large scale test instances is performed.