On the blackboard, the stamping operation objects are in a lower level than the stamping feature objects, and are used to define the manufacturing process from metal strip to the formed metal part. Stamping features constitute a stamped part, while stamp- ing operations are selected as elements of a stamping process plan. Essentially, the stamping process planning task is to trans- form a set of stamping features into a set of stamping opera- tions, and to describe the relations between these. The generic declaration of a stamping operation object includes stamping op- eration type, geometric shapes, geometric constraints, precision, roughness, relationships with stamping features, control param- eters, etc. Typical stamping operation objects include piercing, notching, cut-off, blanking, lancing, shaving, drawing, emboss- ing, coining, trimming, and other miscellaneous forming opera- tions. A stamping feature may be manufactured with a specific stamping operation (one-to-one mapping) or a combination of stamping operations (one-to-many mapping). Several stamping features may also be manufactured with a single stamping oper- ation (many-to-one mapping).

3.1.4 Graph-based stamping process plan

After the mapping from stamping features to a set of stamping operations, the remaining process planning task is to assign each stamping operation to the relevant die station according to an op- timal sequence of stamping operations. Stamping operations are sequenced in a progressive manner by creating stamping opera- tion relations and using them to form a stamping process plan. This formal description of operation relations forms the founda- tion of automatic strip layout design.

A graph-based approach is used to arrange the stamping op- eration objects in a stamping process plan. The graph consists of a set of nodes that store information about the stamping opera- tions, and a set of arcs that store information about the operation relations. Stamping operations are related to one another through two kinds of relationship, “cluster” or “precedence” relations. Cluster stamping operations are executed simultaneously and can be staged at the same die station. Stamping operations in prece- dence  must be performed  in sequence and so they are  staged in adjacent die stations. Cluster relation, and precedence rela- tion are represented by dashed ellipses and directed solid line, respectively, as shown in Fig. 2. Note that stamping  operations B and C work simultaneously, and are staged at the same die station, while stamping operation A precedes operation B, and is staged in a die station immediately prior to the one for the operation B.

The strip layout can be generated by a computer automat- ically using the graph-based stamping process plan, which is suited for computer implementation and leads to efficient formu- lation and solution procedures.

3.2 Knowledge sources (KSs)

3.2.1 Specialist KSs

The planning objects on the blackboard outlined above are not isolated data structures, but are interrelated to each other by a set of specialist KSs that resemble experts by embodying the prob- lem solving knowledge.

The specialist KSs are independent chunks of knowledge and do not communicate directly with each other. Instead, they participate in the problem solving process by contributing  their

Fig. 2. Graph-based partial stamping process plan

partial solutions to the blackboard, or updating the contents of the blackboard.

The perse specialist KSs related to stamping process plan- ning include, but are not limited to, unfolding knowledge to produce a flat pattern, nesting knowledge to produce a blank layout, mapping knowledge to transform stamping features into stamping operations, and staging knowledge to sequence the stamping operations. Methods such as structured interviews, ob- servation, structuring techniques [26] and knowledge modeling approach [27] can be used for eliciting the perse knowledge. Due to the modularity of the blackboard framework, it is con- venient for the experts to expand  the  KS space  in the system by integrating different methods of knowledge representation, such as procedures, rules and objects with the aid of knowledge engineers. Unfolding and nesting knowledge are represented as conventional procedures and will not be elaborated further be- cause this topic has been well-researched and is mature technol- ogy [4, 9, 13, 28]. Our work focuses on the mapping knowledge and staging knowledge.