SUMMARY 

The design process is traditionally a time consuming, iterative business. 
First, a preliminary design must be created, which is then analysed, 
experimented or tested in use to determine its quality. The process of 
search and evaluation is repeated until the design is viewed as being 
acceptable. Computer aided-design software(s) and computer simulation 
and analysis tools are widely used today. In contrast automatic design 
or redesign techniques, are until now, less common. The recent success 
in design is due to the adaptive search techniques, in particular the 
genetic algorithms (GA). GAs are a powerful and broadly applicable 
stochastic search and optimisation technique. Perhaps, they are the most 
widely known type of evolutionary computation methods today.

Assembly lines are production systems composed of a succession of stations, 
connected by a conveyor, performing a set of tasks on the product passing 
through them. The design of an efficient assembly workshop is a problem 
of considerable industrial importance. A production workshop can be set up 
following various topologies-line, cells, combination of several lines. The 
line layout problem is composed of a logical and physical ones. The logical 
layout is defined in the literature as the assembly line balancing (ALB) 
and the resource planning (RP) problems. The ALB is used specially for 
manual assembly lines and aims to balance loads of stations. For hybrid 
(manual and automatic tasks) assembly lines, the RP helps to find an 
assignment of tasks to stations and an assignment of resources to each 
of them. The main objective is to minimise the total cost of the line by 
integrating design (stations space, cost, etc.) and operation issues (cycle 
time, precedence, availability, etc.). The physical layout module determines 
the space requirements taking into account station dimensions, material storage, 
etc. The main objective of assembly systems designers is to increase the 
efficiency of the line by maximising the ratio between throughput and the 
required costs.

A line design problem often has a complex structure due to multiple components, 
e.g. tooling, material handling facility, and so on. A number of design 
alternatives may exist. The problem can easily become unmanageable if the 
designer has to consider all these alternatives. Most real-world problems 
involve multiple, often conflicting, objectives. In assembly lines design 
problems, designers deal with objectives such as line efficiency, cost, 
imbalance, reliability, stations space. Thus, many practical search and 
optimisation problems are better posed as multiple objective problems (MOP) 
and ask for a compromise among conflicting objectives. Since it is impossible 
to replace designer's intelligence, experience and creativity, it is more 
important to provide him with a set of assistance tools. These tools just 
investigate and propose several solutions and make the necessary evaluations. 
Based on this information, the designer tests some alternatives and makes his 
decisions. Due to the difficulty of assembly line design problems, metaheuristics 
are often used to solve this kind of problems. The resulting solutions are adjusted 
to create an acceptable design. 

Applying GAs to solve MOP has to deal with the twin issues of searching large 
and complex solution space and dealing with multiple, potentially conflicting 
objectives. Selection of a solution from a set of possible ones on the basis of 
several criteria is considered as a difficult task. Classical methods reduce the 
problem to a mono-criterion one-the popular weighted-sum approach. Many studies 
adopted Pareto-based GA search to sample the entire Pareto-optimal solutions. 
The main drawback of Pareto approaches is the number of solutions the DM has to 
choose among them. The human cannot easily decide among more than few solutions. 
Few researchers have suggested ways of integrating the multi-criteria decision-aid 
(MCDA) and the GA search. The GA iteratively samples the tradeoff surface (Pareto 
solutions) while the MCDA method narrows the search.

In order to come out of the MOP stated by the cost function and especially in 
the case of grouping problems, we present a new multiple objective grouping 
genetic algorithm (MO-GGA). The method is based on the grouping genetic algorithm 
(GGA) a GA suited to grouping problems and an MCDA method called (PROMETHEE II). 
The choice of a solution over the others requires problem knowledge. It is the 
decision-maker's (DM) task to adjust the weights to help the algorithm to find 
good solutions. Optimising a combination of objectives has the advantage of producing 
a single solution, requiring no further interaction with DM.

In order to deal with line balancing, which is an important part of the assembly 
line design, we introduce a new algorithm called 'equal piles for assembly lines'. 
The hard constraint of the problem is the fixed number of stations (piles), the 
aim is to find the best balanced assembly system. The proposed approach is based 
on the so-called 'boundary-stones'. Many heuristics were also developed in this 
study: simple and multiple wheel, merge and split, pressure difference. This 
algorithm is used as an individual construction heuristic in the GGA. 

In the case of the RP, the aim is to select equipment to carry out the assembly 
tasks. We present a new method which is based on the MO-GGA, the branch-and-cut 
algorithm followed by the MCDA method. The accent is put on how to deal with the 
user's preferences in design problems.

In order to deal with the changes during operation phase of assembly lines, we 
introduce a new concept of balance for operation (BFO). This concept permits to 
treat the balancing and the scheduling model at the design stage. 
The balancing of assembly lines is most of the time uncoupled from the facility 
layout problem. This yields sub-optimal line layouts. An iterative procedure 
partially treating simultaneously the two problems is proposed. First, tasks 
performing alike activities are grouped together in a workcenter (a set of 
linked stations). This leads to a number of workcenters. Then, tasks are assigned 
to stations, which leads to a number of stations on each workcenter. The main 
concern of this approach is the quality of the resulting line in terms of balancing 
and its suitability to the material flow requirements of the production system. 

The last part of this thesis is dedicated to the integrated method to design assembly 
lines. A concurrent engineering approach is presented; it shows the way to help 
designers using iteratively and interactively the different modules.