Professional Documents
Culture Documents
197
CHAPTER 9
SUMMARY OF THE RESULTS
9.1 INTRODUCTION
In this thesis, the issues related to the operation and control of two flexible
manufacturing configurations are addressed. Algorithms for optimal resource utilisation in
those models are designed and analysed. The salient points along with their usefulness of the
various scheduling procedures for the above models are summarised and discussed in this
chapter.
2. The solution is closer to other methodologies and capable of providing solutions nearer
to optimal.
3. By changing the evaluation parameter of the genetic search process, the solution can
be obtained for other suitable objectives and is flexible.
4. The method can be extended with artificial neural work in order to find system specific
operating strategies for the operational management and can be used for the
development of an expert system by inducing scheduling knowledge in FMS
environment.
5. The heuristic search process can be regarded better than simulation in the sense that it
can guarantee near optimal solutions in actual cases, since they provide exact
solutions.
6. Rescheduling with regeneration guarantees the solution quality whatever be the state
of the system parameters and addresses the real time operation.
7. The extensions to handle alternate route choices and to revise the schedules in real
time operations lead to enhance the productivity to a great extent.
In the present work, four simple pdrs have been employed to resolve the conflicts. The
methodology can be extended with more number of pdrs and other objectives. Future research
needs to concentrate in the genetic operators to get convergence faster.
(Continuous Interactive Dichotomister3 algorithm extended with genetic program for weight
optimisation) and Classification Decision Tree (CDT) algorithm. The knowledge base
obtained through the above learning schemes generate robust and effective schedules
intelligently with respect to the part-mix changes in real time for makespan criterion.
The comparison made with a GA based scheduling methodology (Chapter 5) show
that WCAODSs provide solutions closer to optimum. Advantages of WCAODSs when
applied to dynamic scheduling are:
1. Learning is done off-line;
2. It has the ability to infer the rules quickly for real time control which is essential for
handling dynamic states of the system;
3. Giffler and Thompson algorithm with Dynamic WCwise-pdr set assures good
schedules and performance;
4. Embedded knowledge acquisition mechanism makes WCAODSs intelligent and
flexible in obtaining required scheduling heuristics which enlarge its application
domain;
5. Compared to other inductive learning methods reported in previous research, the
genetic CID3 minimises the layers and hyperplanes, which are used to formulate the
decision rules, and hence the number of decision rules gets confined and
6. The main advantage is the computational time. WCAODSs require less time than GA
method.
The crucial phase in the proposed methodologies is the acquisition of scheduling
knowledge base that is dependent on many factors such as the number of good training
examples, the number and characteristics of problem attributes, the classes of each attribute,
and the operation environment. Scheduling knowledge is defined with randomly generated
data. Any how that can be defined well for a specific system with its past part mixes data and
the attributes that affect the scheduling process.
One main disadvantage of genetic CID3 algorithm is it is designated to tell two
distinct classes apart; consequently, several neural networks are sequentially generated by the
genetic CID3 algorithm for a multi class classification algorithm. The range of the connection
weights influences markedly on the convergence of NN architecture and learning time in
GCID3 algorithm. So in future, attention can be given to this point. As far as WCAODS2 is
Scheduling FMS Using Heuristic and Search Techniques 201
concerned, better results are possible with more numbers of training examples, attributes and
attribute classes.
9.7 CONCLUSIONS
In this thesis, scheduling procedures to tackle various operational level issues in FMS
have been discussed. The GA based methodologies have much scope in FMS scheduling
because of their capability to handle any hard problems. Any how, they require refinement in
certain aspects as outlined below.
1. Development of specific genetic operators for faster convergence with less number of
iterations and to reduce the possibility of getting trapped in the local minima.
2. Combining heuristic knowledge with the GA to have still better solutions.
Scheduling FMS Using Heuristic and Search Techniques
203
The knowledge-based schemes are capable of addressing real time control and is
useful for rescheduling. However, the acquisition of good scheduling knowledge base depends
on many factors such as the number of good training examples, the number and characteristics
of problem attributes, the classes of each attribute, and the definition of the operation
environment. The future research requires to concentrate on the above lines.
The heuristic suggested to generate AGV schedule from the machine schedule that has
neglected MHS provides exact time table of both AGV and machine schedules. This makes
the scheduling an integrated one.
The allocation of WIP in specific locations in AS/RS enhances productivity by
minimising the movements of the AS/RS shuttle.