Determination of Achievable Process Tolerances with Respect to Manufacturing Process Capability - Manufacturing Systems Processes - page 59

TABLE 2.3 Summary of the Tolerance Design Expert System

Expert System Data Input Dimension Tolerance Method Notes

Proplan (1984) CAD

database

2D No Graphic search Inefficient for

complex parts

Excap (1984) Decision

tree

2D No Backward

chaining

No interface
with CAD

Sipp (1985) N/A 3D Yes Least-cost search Tolerance for process

selection

Tipps (1985) CAD

database

3D Yes Surface extraction Interactive tolerance

not detected

Turbo CAPP (1987) CAD

database

2D Yes Surface & bound

extraction

Improper tolerances
not detected

Xplan-R (1988) N/A 2D Yes Hybrid No interface with CAD

TVCAPP (1993) CAD

interface

3D Yes Interactive CAD Tolerance verified vs. CSA

standard

From Malck, L.A. and Asadathorn, N., Process and desigh tolerance in relation to manufacturing cost, The Engineering
Economist , 40(1), 1994. Reprinted with the permission of the Institute of Industrial Engineers, 25 Technology Park,
Norcross, GA 30092, 770-449-0461. Copyright©1994.

can perform a complete tolerance analysis without the designer's guidance. The technique of artificial
intelligence has also been utilized in tolerance allocation. An AI system usually establishes a knowledge
base defining the rules that are used to allocate the tolerance of component. Table 2.3 summarizes the
tolerance design expert systems using the AI technique [1].
Abdou and Cheng [26] indicated that few systems have developed the important linkage between CAD
and CAPP. Among the AI systems in Table 2.3 , TVCAPP can verify and generate both dimensional and
geometric tolerances, and produce alternatives for process plans with corresponding costs.
Most of the mathematical models described above are nonlinear programming problems, which are
frequently solved using either nonlinear optimization packages or heuristic algorithms. The heuristic
algorithms proposed include genetic algorithm, simulated annealing, and Taguchi method.

Genetic Algorithm [27]
GA uses three operators, the reproduction, the crossover, and the mutation. The reproduction operator
allows the highly productive chromosomes (strings) to live and produce offspring in the next generation.
The crossover operator, used with a specified probability, exchanges genetic information by splitting two
chromosomes at a random site and joining the first part of one chromosome with the second part of
another chromosome. Mutation introduces occasional changes of a random string position with a
specified mutation probability. The general procedure for genetic algorithms in tolerance optimization
problems is described as follows:

1. An appropriate chromosome representation should be defined to represent the combinations of
design variables that correspond to the fitness or objective function values. The representation
should be a one-to-one mapping in order to have a normal coding and decoding process.
2. The probabilities of crossover and mutation are specified. The population size and maximum
number of generations are selected, and an initial population in the genetic system is generated.
3. Evaluate the objective function value or fitness of each solution in the current generation.
4. Apply the three operators to those solutions in the current generation to produce a new population.
5. Repeat steps 3 and 4 until the maximum number of generations is reached.

Simulated Annealing [31]
Simulated annealing was proposed in early 1980s. It is a simulation of the physical annealing process
of metal heat treatment. It differs from iterative improvement algorithms in the sense that the latter
accepts only those moves that will lead to the improvement of the results. The inherent problems with