Generative CAPP Approach

The generative CAPP approach represents an alternative approach to

automated process planning.
It uses the knowledge of manufacturing and the decision making logic of
process planning, captured and encoded into computer programs, to automatically
create process plans for a new part from scratch without referring to existing plans.

Parts description
database

Manufacturing
Manufacturing Generative CAPP
knowledgebase /
database System
decision making logic

Generated
process plan
 Generative CAPP
Part description
Part description forms the major part of the information needed for
process planning.
Since the aim is to automate the system, the part description should be in
a computer readable format.
Traditionally engineering drawings have been used to convey part
descriptions and to communicate between design and manufacturing.
Understanding engineering drawings was a task suited for well trained
human beings and not suitable for direct input for process planning.
The developments in CAD have provided some means for creating
computerized storage of engineering drawings.
Several methods have been used for representing parts for generative
process planning such as GT based codes, special descriptive languages,
CAD models.
 Generative CAPP
Part description (continued)
Codes
The GT based coding scheme consists of a sequence of symbols that identify a
parts design characteristics and features, and/or its manufacturing attributes.
Several coding schemes are available such as OPITZ, DCLASS, MULTICLASS,
KK3, etc.
Variant process planning systems have been based exclusively on coding schemes.
Several generative process planning systems such as APPAS, GENPLAN,
COBAPP, M-GEPPS, CORE-CAPP, etc. have used coding schemes to describe
parts for generative process planning systems.
The degree of detail that can be included in the code depends on the resolution
allowed by the number and the type of digits used.
The code based representation is not suited for a completely automated process
planning system, since coding is a manual process, and a human interface is
needed between the design and process planning function.
 Generative CAPP
Part description (continued)
Special descriptive languages
Special descriptive languages have been used for part description in generative CAPP systems.
They are specially designed to contain all the information required for the necessary process
planning functions.
Several generative process planning systems such as CIMS/DEC, GARI, etc. have used special
descriptive languages to describe parts for generative process planning systems.
For example, the AUTAP system used one such
system.
Fig. shows an example of a part modeled using this
approach.
Besides using true geometric elements, such as
cylinders, cones, etc., it also uses subordinate
elements represented using keywords and attributes.
The geometric elements characterize the main
contour, and the subordinate elements provide the
details.
Process planning is performed by analyzing the
component description commands and relating them
to manufacturing.
 Generative CAPP
Part description (continued)
Special descriptive languages
In the CIMS/DEC part description system, the part shape is described using volumetric
elements obtained by revolving or parallel moving generating surfaces.
The generating surface is composed of a concatenation of profile elements, which are given
by directed line segments as shown in Fig.
Technological information such as surface finish is associated with each profile element.
This representation is suited for process planning tasks but the problem is the limitation of
this representation to a limited domain of parts.
 Generative CAPP
Part description (continued)
Special descriptive languages (continued)
GARI uses a part description based on a set of features, such as tapped holes, countersink
holes, bores, grooves, notches, faces, etc. to describe the part as shown in Fig.
These features are described using system words such as diameter, surface finish, distance,
perpendicularity, tolerance, and so on.
The relationships between features are
also specified as attributes,
starting_from and opening_into.
The parts description as a set of
features conveys some expertise about
its manufacturing when specified by this
method.
This representation is suited for process
planning tasks, but it has to be
manually prepared for input into the
system.
For complex components, this can be
tedious and difficult.
 Generative CAPP
Part description (continued)
Special descriptive languages (continued)
The majority of generative process planning systems that use expert system
approaches employ special descriptive languages for part description.
Although the special descriptive language simplifies the process planning, it requires
lot of effort from the user of the system who has to manually prepare the input data
and is not suited for a completely automated generative process planning system.
 Generative CAPP
Part description (continued)
CAD Models
Although the uses of GT based coding scheme and that of special descriptive
languages can provide information for several process planning functions, the part
design still needs to be manually converted into one of these representations.
Since the CAD model contains all the detailed information about a part design that are
necessary for process planning, the use of CAD models as input to the process
planning module has the potential to eliminate the human effort that is involved in
translating a part design into GT code or special descriptive language formats.
However, special feature recognition algorithms need to be used in order to extract
information about the different machining features and their attributes that are
necessary for performing the different process planning functions.
Different 2-D and 3-D CAD models have been used for the input part description in
generative CAPP systems.
Feature based design models have been also used for the input part description in
generative CAPP systems.
 Generative CAPP
Part description (continued)
CAD Models (continued)
In feature based modeling, there is no need to recognize or extract features from the
CAD model.
Predefined features that can be translated into manufacturing operations are used to
design the part. They can be used directly as input to the CAPP system.
One problem with the feature based modeling approach could be that there is not
enough flexibility for the designer in some cases.
Another problem is the fact that he is required to fill in the manufacturing details while
designing, instead of concentrating on the functionality of the part.
 Generative CAPP
Manufacturing database
In addition to the part description database, various manufacturing databases are required to
provide the process planning system with the information required to make decisions.
Examples of various process planning functions and supporting databases for process planning are
shown in Fig

Feature
CAD database
extraction
Stock
Stock database
selection
Process
Process Planning System

selection Process capability

Machine
Machine database
selection
Tool
Tool database
selection
Process
sequencing
Cutting Machinability
parameters
 Generative CAPP
Manufacturing knowledge base
Another major component of generative process planning system is the manufacturing
knowledge base containing the decision making logic and knowledge.
This includes the logic used by the process planner to make decisions on various
aspects of process planning, such as operations selection, machine tool selection,
cutting tool selection, operation sequencing, etc.
Several methods have been used to represent this planning logic and knowledge into
formats and structures which will facilitate program coding and documentation.
They include the use of decision trees, decision tables, algorithmic and graph theoretic
approaches, Artificial Intelligence (AI) and Soft Computing based approaches like expert
systems, neural networks, fuzzy logic, etc.
The decision logic has to be synthesized and structured in a clear manner, so that it is
complete, unambiguous, nonredundant, and capable of expansion.
 Generative CAPP
Decision Tree
A decision tree is a useful decision making tool.
Its structure resembles that of a tree with a single root and a set of branches emanating from
the root.
The root is the start node and each
branch emanating from the root can lead
to another node or terminate in an action. E7
A5
The nodes provide further branching External
possibilities. Root
E3
A4
Rotational Tolerance >0.0006mm
There is a condition statement specified parts
on each branch that must be satisfied in E1
N1
E4
A1
Drill
order to traverse that branch. Hole
Tolerance
<0.0006mm
N2 A2
Thus if the condition specified on a E2 E5
branch is true, then that branch can be
traversed to reach the next node, and so E6
A3
A Decision Tree
on until a terminal point on the tree is
reached.
Fig. shows a sample decision tree.
 Generative CAPP
Decision Table
A decision table is another useful decision making tool.
It organises the conditions, actions and decision rules in a tabular form.
Conditions and actions are placed in rows, while decision rules are identified by columns.
The upper part of the table
includes the conditions that must
Condition hole X X X
be met in order for the actions
represented in the lower part of dia <3.0mm X X X
the table to be taken. tol > 0.0006mm X
A sample decision table is shown 0.02 < tol < 0.0006mm X
in Fig.
tol < 0.0002mm X

sf > 0.005mm X

sf < 0.005mm X X

Action drill 1 1 1

bore 2

ream 2
 Generative CAPP
Expert System
An expert system is a class of intelligent computer programs designed to solve problems in a
specific domain of application, which would otherwise require significant human expertise for
their solution, by using domain specific knowledge and inference procedures.
The major parts of an expert system are:
a) Declarative knowledge about the problem
b) Problem solving knowledge or domain Procedural Knowledge
(Rules and facts about
knowledge or procedural knowledge, process planning)
Declarative Knowledge
c) Inference engine, and (facts about the part)

d) User Interface, Explanation Facility, Inference Engine

Knowledge Acquisition Facility.

User Interface Explanation facility

Knowledge
Acquisition/Learning

Block Diagram of an Expert System

 Generative CAPP
Declarative Knowledge
The declarative knowledge constitutes the input to the expert system.
It consists of facts about the problem being solved, which, in the case of process
planning, include detailed information about the part being planned, as contained in the
part design on a CAD system.
The part information necessary for process planning includes geometry, geometric
relationships, dimensions and tolerances, etc.
The above information needs to be converted from its original data format on the CAD
system into the representation format for declarative knowledge of the expert system.
A common form of representation of the declarative knowledge facts is frame.
 Generative CAPP
Declarative Knowledge (continued)
Frame
A frame system is a network of nodes and relations organized in a hierarchy.
A frame is a data structure which represents an entity type.
A frame consists of several slots. Each slot can store values,
pointers to other frames, or a procedure (See Fig.)
The slot value can be restricted to a certain range or from a
set of values. For example, the diameter of a hole is
restricted to a range which is machinable by the shop.
One of the important features of the frame is inheritance.
A structure of frames may be set up in which both
parent/child link and sibling link are allowed.
Parts of the slots can be inherited by the child from the
parent. For example, both the thru slot and blind slot belong
to class slot. The property of slot such as dimension,
dimensional tolerance, position, positional tolerance, surface
finish etc. can be inherited by both subclasses.
One frame can be linked to another frame by a pointer stored
in a slot.
Procedures or functions can be attached to a slot. The
procedure can be executed when a) new information is
placed in a slot, b) information is deleted from a lot.
 Generative CAPP
Declarative Knowledge (continued)
Frame (continued)
The following example shows the frame representation for a slot feature.
(slot (length 4.5) (width 0.8) (depth 0.5) (ref_face face 1) (surface_finish 250.0)
(length_tol_range ( -0.05, 0.05)) (width_tol_range ( -0.05, 0.05)) (depth_tol_range ( -0.05, 0.05))
(location_tol_range ( -0.05, 0.05)) (approach_direction (0.0 0.0 -1.0)))

Procedural knowledge
The problem solving knowledge or domain knowledge or procedural knowledge of the expert system is
specific to a problem domain, which in the present case, is process planning.
The procedural knowledge is necessary to take decisions on various aspects of process planning such
as process selection, cutting tool selection, machine tool selection, set up planning, process parameter
selection, and so on.
The procedural knowledge contains mainly facts and rules.
A typical fact in a process plan may be The diameter of drill DX0021 is 0.25 in, or Machine ML001
can perform drilling, boring, and reaming operations.
Rules are quite often rules of thumb which human experts acquired over the years.
These rules of thumb are also called heuristics.
The process planning knowledge is embedded in those rules.
 Generative CAPP
Procedural knowledge (continued)
How good the knowledge is is crucial to the success of a process planning system.
However, in order to put it in an expert system, it must be represented in a usable form.
The most commonly used representation method for procedural knowledge is IF-THEN rules.
Rules
A rule has two parts:
The IF portion (also called antecedent) of the rule is a series of patterns, which specify the facts, which
cause the rule to be applicable.
The THEN portion (also called consequent) of the rule is a list of actions to be executed when the rule
fires.
For example, in order to develop a IF-THEN rule for performing process selection, the antecedent
IF part may contain feature type, dimension range, tolerances, surface finish, etc. information.
This information is also called process capability.
The consequent THEN part may describe the actions that need to be taken, if the patterns in the
antecedent IF part are satisfied by the facts in the database.
A simple rule written in pseudo language for a drilling process may be as follows:
If ((Feature_type is hole) (hole_diameter 3) (hole_diameter 1) (hole_l/d-ratio 8)
(hole_pos-dia-tol 0.003) (hole_neg-dia-tol 0.001) (hole_true-position 0.0004)
(hole_roundness 0.004) (hole_surf-fin 100))
then (select twist-drill))
 Generative CAPP
Procedural knowledge (continued)
The process selection module selects a set of feasible processes for each machining feature using
a set of knowledge base rules.
Then the most appropriate cutting tools are to be selected.
There are several decisions to be made in tool selection, as there are number of variables on a tool
including tool type (drill, mill, boring bar, etc.), tool geometry (e.g. nose radius, rake angles,
clearance angles, etc.), tool material (e.g. HSS or carbide, hardness/grade, etc.), tool dimension
(e.g. diameter, length, etc.), etc.
Tool type is determined by the process type, and the rest of the decisions have to be made based
on the workpiece material, hardness, feature dimension, etc.
Due to number of tools a shop may carry, the tool data requires to be stored in a relational
database.
In each tool data record, information on the tool number, tool material, tool geometry, dimension,
tool life left and operation types are stored.
Tool selection rules can be written to generate tool database queries to retrieve the desirable tools.
The query returns a set of feasible tools.
Next the most appropriate machine tools are to be selected.
The decision of machine selection is based on a number of characteristics of the machine such as
machine type, process type, work table size, work envelop, spindle horsepower, spindle taper type,
feed and speed range, axis torque, accuracy, cost, etc.
Rules can be written to match the above machine capabilities with the requirements.
 Generative CAPP
Procedural knowledge (continued)
Next it is necessary to perform set-up planning.
Set-up planning involves mainly the following three tasks: the set-up formation i.e. determination
of groups of machining operations that can be carried out in the same set-up, the determination
of sequence of operations within a set-up and the selection of the set-up datums.
The decision on set-up formation is made considering the tool access directions (TADs) of the
features present in the part to be machined and geometric tolerance relationship between
features.
For a given machine tool, the operations on features with the same TAD are grouped in the
same set-up.
Features having more than one TAD are assigned a unique TAD, depending on their tightest
tolerance relationship with other features, so that their corresponding machining operations can
be grouped along with other operations in one of the set-ups.
The above decision making tasks can be implemented using knowledge base rules.
The decision on determining the sequence of operations is based in accordance with the
constraints imposed by the precedence relationships between features and machining
operations to produce them.
A set of knowledge base rules can be written for determining the feasible sequence of
operations in accordance with the above constraints.
 Generative CAPP
Procedural knowledge (continued)
Finally the selection of datum features for locating and clamping is performed with the purpose of
obtaining the remaining critical tolerance relationships between features that could not be satisfied
during the set-up formation and leaving the inevitable tolerance chain errors to those unimportant
relationships with looser tolerances.
It is accomplished with a set of heuristic principles for datum selection which can be encoded in
the form of knowledge base rules.
It is further necessary to select process parameters for machining including feed and cutting
speed.
The selection of process parameters depends on the tool material, workpiece material, tool
diameter and depth of cut.
Since there are very large number of materials, diameter and depth of cut combination, the
storage of process parameters requires a large database for which a relational database will be
the most appropriate.
Rules can be written to synthesize queries which will be sent to the database and the query will
return the suggested feed and speed from the database.
 Generative CAPP
Inference Engine
The third major part of the expert system is the inference engine that contains the general
problem solving knowledge.
It determines which rule antecedent, if any, is satisfied by the facts in the database.
Part of an inference engine is an interpreter. It interprets the procedural knowledge and resolves
the procedure embedded.
The inference engine also has built-in search algorithms to search among a large number of rules
and determine the sequence in which the rules are applied and facts found. This search method is
called chaining.
However, since the inference engine is supposed to be general, it is not necessary to develop a
special inference engine for a specific application. Many expert system shells provide a built-in
inference engine.
User Interface, Explanation Facility, Knowledge Acquisition Facility
An expert system also may have three other components namely, a user interface, an explanation
facility and a knowledge acquisition module.
The user interface provides the mechanism by which the user can communicate with the expert
system.
The user interface is meant to start the process planning by providing the necessary input data
files containing the complete part design information and then finally display the results of the
expert system based process planner.
 Generative CAPP
User Interface, Explanation Facility, Knowledge Acquisition Facility (continued)
The explanation facility displays the rationale behind applying (firing) a certain rule.
It traces the reasons an expert system concludes an answer or reconstructs the reasoning steps.
Since either erroneous rules may be acquired from a domain expert, or a rule may not be properly
entered into the system, the expert system may get incorrect answers.
Without an explanation facility to help debug the knowledge base, it would be extremely difficult to
pinpoint the error.
The purpose of having a knowledge acquisition module is to assist the user of the system to
update the knowledge base, because we know a manufacturing system is never static.
There are constant changes in machine tools, cutting tools, fixtures, etc. or new manufacturing
processes may be added to the system.
So we may want to modify some of the existing rules.
The knowledge acquisition module helps to ensure that the new rules, which are added to the
knowledge base, are consistent with the existing rules.
If there is any contradiction of the newly entered rule with the existing rules, it is detected by this
module.
 Generative CAPP
User Interface, Explanation Facility, Knowledge Acquisition Facility (continued)
There are many tools, which have been used to build expert systems.
These include Artificial Intelligence languages such as LISP and PROLOG and programming
languages such as FORTRAN and C. However writing a complete expert system using any of the
languages requires a tremendous amount of work.
Therefore, to ease the task of building an expert system, several expert system shells and expert
system building tools are available.
Expert system shell is a special purpose tool designed for certain types of application in which the
user must supply only the knowledge base. Examples of expert system shells are EMYCIN and
CLIPS.
Using expert system shells can speed up the expert systems development time.
However, shells were originally developed for other applications; therefore, for a different
application it must be fitted into the existing shell for which some compromise may have to be
made.
Expert system building tools, although similar in features to an expert system shell, provide the
users with a tool to build their own application. Examples of such tools are KEE, KnowledgeCraft.
 Advantages of Generative CAPP
- It helps to overcome limitations of variant CAPP.
- It can generate consistent process plans rapidly.
- It can be used to generate process plans for new parts with or without similarity to
pre-existing parts.
- It offers a structured knowledge representation.
- It is capable of automatically acquiring knowledge.
- It has the potential to eliminate human intervention in planning process by fully
automating the different decision making tasks.
 Semi-generative CAPP
It is combination of generative and variant CAPP approaches.
It involves interaction between a computer and a human process planner.
The computer prompts the process planner for inputs and generates a pre-process
plan using either existing standard process plans or predefined decision making logic
The human process planners responsibility could be:
Interpretation of the design data and / or an engineering drawing
To make suitable changes on a standard process plan for the corresponding
part family,
To manually complete portions of the process plan those that have been left
out because the computer is unable to resolve them
To draw up the final process plan
Advantages of Semi-generative CAPP
considerable time and cost savings,
reduction in process planners intervention compared to variant CAPP
quality of the process plan ensured.