|
Article Excerpt
1. Introduction
A production process materializes abstract design concepts into a physical product and includes a number of ordered operations (i.e., operations with precedence relationships). Each operation is associated with one or more input items, an output item, a machine, tools, fixtures, setups and cycle times. In discrete manufacturing, developing a process plan suitable for release to production is a tedious and time-consuming process (Huang et al. 2004). In practice, engineers plan production processes by trial and error based on individual "know how," experience and intuition rather than rigorous analysis (Huang et al., 2004; Tong et al., 2004). As a result, production processes planned for the same products often differ from one another. Moreover, it is not uncommon that engineers plan production processes starting from scratch due to, for example, the lack of appropriate information management systems (Forza and Salvador, 2002). Consequently, a number of avoidable variations are included in production processes, which, in turn, incur unnecessary changeovers on shop floors.
Manufacturing environments nowadays are characterized by a large number of customized products that are often required in small quantities and with short delivery times. The key to efficiently producing such product variety lies in an ability to maintain the corresponding production to be as stable as possible; and such production stability can only be achieved through fulfilling customized products using similar production processes (Schierholt, 2001; Williams et al., 2007; Zhang, 2007). The traditional practice in production process planning described above is unable to obtain such similar processes. This underscores the importance of devising robust models and well-structured mechanisms, which can be employed to plan similar production processes given a diversity of customized products.
Current design practice (e.g., platform-based product development) has led to the concept of product families, each of which consists of a set of different yet related products (Hegge, 1995). While family members assume a common product structure and contain same/similar component items, they differ from one another in optional items and features. With inherent design similarity and commonality, a product family is underpinned by a generic product (Van Veen, 1992). In relation to a product family, a process family is defined as a set of possible production processes which can fulfill customized products in the product family (Zhang, 2007). As with the generic product of a product family, fundamental to the corresponding process family is an underlying generic process. By including all possible process data of the product family by means of generic representation (Van Veen. 1992), a generic process entails a well-structured mechanism, with which companies are able to plan similar production processes for product families. In this respect, using generic processes to plan process families can help companies fulfill product variety, resulting in cost benefits. First, the complexities in material handling and routings on shop floors can be reduced since planning process families is anchored to a common platform (i.e., the generic process). Second, the similar production processes can free production personnel from frequently changing machines, tools, fixtures, etc. when producing new products.
Unlike product families, process families have not received much attention (Lu and Botha, 2006). The reason is twofold. First, planning production processes involves specific technical and managerial challenges, which are un-addressed by the existing bodies of literature (Pisano, 1997). Second, most efforts in academia and industry have been made to achieve design efficiency and effectiveness without considering the fact that it is at the production stage that product costs are actually committed and product quality and lead times are highlighted the most. In response to the lack of research, efforts should be made to develop process families, more specifically the underlying generic processes, from production experiences. The ultimate goal is to obtain stability of product family production.
Towards this end, in this study we address generic process construction based on large volumes of existing production data. To the best of our knowledge, this study is the first that: (i) identifies this practical problem in the context of product family production; and subsequently (ii) develops a method to construct generic processes. Since a production process is commonly represented by a tree, we propose a tree unification approach to constructing generic processes from past production practice. Martinez et al. (1995), Martinez et al. (2000) and others point out the advantage of using a binary tree to represent process plans, such as the simultaneous representation of operations parallelism and decrease of combinatory complexity. Another advantage of the binary tree representation is the elaboration of each individual operation of manufacturing a part, producing an assembly and assembling an end-product. Hence, we adopt binary tree representation in the proposed approach.
The tree unification approach takes advantage of the tree representation which allows a visualization of complex production processes and the resulting easier understanding. In addition, it exploits similarities among operations and precedence relationships exhibited by nodes and arcs, for which we develop formulations and algorithms. These formulations and algorithms are able to handle complex practical situations in which multiple operations and operations precedence along with component items and end-products are involved.
w The rest of this paper is organized as follows: the context of generic process construction is formulated in Section 2. Also discussed are the concept implications of a generic process. Following an illustrative example, the details of the proposed tree unification approach are provided in Section 3. An industrial example is presented in Section 4, and further discussions of the work developed in this paper are given in Section 5. Also included in Section 5 are concluding remarks and possible avenues for future research.
2. Problem formulation
Design similarity and commonality in a product family leads to process similarity and commonality in the corresponding process family, as reflected by the similar operations, operations types, operations precedence and routings. As with the generic product of a product family, a generic process underlying the corresponding process family is suggested by process similarity and commonality. The generic process contains process elements common to the process family. It also includes the set of specific process elements pertaining to each individual production process. Given large volumes of existing production data, this study focuses on the construction of such generic processes. Interested readers may refer to a text mining procedure and fuzzy clustering in Zhang (2007) for details of measuring process similarities and subsequently forming process families.
Definition 1. A process family, [SIGMA], corresponding to a product family is a set of specific production processes, i.e., [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], where [P.sub.p], p = 1,..., [N.sup.p] denotes a production process to fulfill a customized product in the family.
Definition 2. A production process, P [member of] [SIGMA], is defined as a two-tuple P = ([PSI], [GAMMA]), where [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is a set of operations and [GAMMA] is the set of precedence relationships between two operations, [O.sub.i], [O.sub.j] [for all]i [not equal to] j = 1,..., [N.sup.O], and describes the ordering of operations pairs.
Definition 3. An operations precedence is defined such that: (i) ([O.sub.i], [O.sub.j]) [member of] [GAMMA] indicates [O.sub.j] [member of] [PSI] can start if and only if [O.sub.i] [member of] [PSI] has been completed; (ii) [O.sub.j] follows [O.sub.i], immediately without any intermediate operations in between; and (iii) a transitive closure of [GAMMA] is irreflexive and ([O.sub.o], [O.sub.i]) [member of] [GAMMA] [conjunction] [O.sub.o], [O.sub.j] [member of] [GAMMA] [right arrow] [O.sub.i] = [O.sub.j], where [O.sub.o], [O.sub.i] and [O.sub.j] [member of] [PSI]. The implication is that along with [PSI], [GAMMA] forms a tree of a production process.
A binary tree is a tree, where each node has a maximum of two child nodes, as shown in Fig. 1. In a binary tree representation of a production process, nodes represent operations and are described by operations types, material inputs, outputs, machines and setups. Arcs between two nodes correspond to operations precedence. We note that by following the definitions in binary tree representation, we relate nodes to operations and arcs operations precedence. In other modeling/representation tools pertaining to manufacturing/assembly processes, such as AND/OR graphs (Benjaafar and Ramakrishnan. 1996; Homem de Mello and Sanderson. 1990), directed acyclic graphs (Borenstein, 2000) and state transition graphs (Borenstein and Becker. 2004), nodes and arcs are defined in a similar way. In general, two types of nodes can be identified: l-nodes and i-nodes. While l-nodes are nodes that do not...
|
