بهینه سازی ترکیبی ازدحام ذرات با جهش برای بهینه سازی خطوط تولید صنعتی : کاربرد یک فضای راه حل آمیخته با توجه به متغیرهای طراحی گسسته و پیوسته
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|5822||2013||11 صفحه PDF||سفارش دهید||9929 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Industrial Marketing Management, Volume 42, Issue 4, May 2013, Pages 496–506
This article presents an artificial intelligence-based solution to the problem of product line optimization. More specifically, we apply a new hybrid particle swarm optimization (PSO) approach to design an optimal industrial product line. PSO is a biologically-inspired optimization framework derived from natural intelligence that exploits simple analogues of collective behavior found in nature, such as bird flocking and fish schooling. All existing product line optimization algorithms in the literature have been so far applied to consumer markets and product attributes that range across some discrete values. Our hybrid PSO algorithm searches for an optimal product line in a large design space which consists of both discrete and continuous design variables. The incorporation of a mutation operator to the standard PSO algorithm significantly improves its performance and enables our mechanism to outperform the state of the art Genetic Algorithm in a simulated study with artificial datasets pertaining to industrial cranes. The proposed approach deals with the problem of handling variables that can take any value from a continuous range and utilizes design variables associated with both product attributes and value-added services. The application of the proposed artificial intelligence framework yields important implications for strategic customer relationship and production management in business-to-business markets.
Product line design is a critical task that may determine a firm's survival. Product lines need to constantly evolve in response to market and technology changes. The process of developing a product line is costly, whilst the failure rates of new products are alarmingly high. As a result, the determination of optimal product lines has attracted considerable attention in the marketing literature. However, all optimization algorithms so far have been applied to products whose attributes are treated as discrete. Contrary to existing product line optimization studies, the present application searches for optimal solutions in a very large, mixed design space, which consists of both discrete variables and variables that can take any value from a continuous range. Our approach makes the optimization problem much more realistic, given that, in practice, the design space of most products can be very large, virtually infinite and includes both discrete and continuous design variables (Luo, 2011 and Michalek et al., 2011). In real life, many product attributes are described in terms of continuous, real numbers. This is especially common in business-to-business markets, in which industrial products are often specified in terms of continuous variables such as weight, length, speed, capacity, power, energy, time etc. Existing approaches virtually convert the continuous attributes to discrete ones by defining just a set of values (i.e., attribute levels), usually the upper and the lower limit of the attribute's continuous range, along with a few representative points within the range. Obviously, this constitutes a very restrictive assumption, which on the one hand may reduces the problem's complexity, but on the other hand may also lead to less than optimal solutions and thus, less than maximum profits and sales for the firm. Let's imagine for example a hypothetical product design scenario in which the researcher considers two continuous attributes. If both attributes are treated as discrete and range across three levels each, the solution space of the single-product line problem will consist of nine candidate optimal solutions. On the other hand, if the attributes are treated as continuous, the algorithm will search for the optimal product configuration across an “infinite” set of candidate solutions. The inclusion of continuous variables into the design space may increase the problem's complexity, but at the same time, the problem becomes much more realistic. Additionally, its applicability for practical engineering problems increases, whilst the researcher can manage complex relationships among attainable combinations of product attributes that could not be easily handled through attribute discretization. In this paper we propose a new method for optimizing a line of products which consist of both discrete and continuous attributes. Specifically, we design a hybrid particle swarm optimization (PSO) algorithm with mutation and implement it to a business-to-business context. The implementation of our mechanism to an industrial setting yields important managerial implications for strategic customer relationship and production management. The study contributes to the literature by proposing an optimization method that works in relatively large-scale design problems consisting of both discrete and continuous design variables and provides an efficient solution to the focal manufacturer's design variable configuration problem. To the best of our knowledge this is the first study that optimizes an industrial product line that incorporates continuous attributes. The remainder of the article is organized as follows. The next section discusses the theoretical background with particular emphasis on the product line design literature. The third section develops the proposed hybrid PSO algorithm, which is then implemented to a case study concerning industrial cranes. A concluding section summarizes the paper and provides useful implications for managers and researchers.
نتیجه گیری انگلیسی
This paper presents a new hybrid PSO approach and implements it to the optimal product line design problem. Due to the complexity of this optimization problem, all existing applications so far have traditionally searched for optimal solutions in discretized design spaces. However, in practice, the design spaces of most products are usually large, virtually infinite and consist of both discrete and continuous design variables. It is this important limitation of previous applications that the present study addressed by the use of swarm intelligence. Our approach, which was illustrated through an application to a simulated dataset pertaining to industrial cranes, yields important implications for strategic customer relationship and production management. 5.1. Managerial implications Most industrial products usually consist of attributes that are strictly specified in terms of continuous variables such as weight, length, speed, time etc. In such markets an inch or a kilogram deviation from the customer's specification may be deemed crucial for the relationship's continuance. Literature on industrial, relationship and business-to-business marketing has long emphasized the fact that in today's highly competitive industrial markets, one of the few ways left to gain differentiation from competitors is by offering tailor-made products and value-added services as a means to enhance relationships with customers (e.g., Morgan and Hunt, 1994 and Palmatier et al., 2007). Buyer–supplier relationships have evolved tremendously during the past decade, due to increased competition, account concentration, a reduction in the number of suppliers and tighter procurement and buying functions (Capon, 2001). As a result, key account management has gained relevance to supplier companies as a means to create value, by implementing specialized processes targeted to most important customers (Wengler, Ehret, & Saab, 2006). Customization of the product is one of the most effective ways to strengthen a key account relationship and offer more added-value (Ojasalo, 2002). Customization, of course, requires that the supplier knows the specific needs of the account and is willing to accommodate them. Swarm intelligence mechanisms could be particularly helpful in strengthening relationships with key accounts through customization. By allowing design variables to take any value from a continuous range, manufacturers have the ability to deliver more customized products to their customers, strengthening in that way their relationships with them. Customer needs have become so diversified and product variety has grown so much in many sectors that the need for more customization is dire. Especially in the case of industrial products the principle and methods of mass customization have been widely applied in their production. Mass customization is a concept created by Davis (1989) to describe a new paradigm in production management that aimed to fulfil the demand for tailor-made products and services. As an emerging paradigm, mass customization aims at satisfying individual customer needs while staying near mass production efficiency (Pine, 1993). In other words, it combines the cost-saving effectiveness of mass production with the value-added benefits associated with tailor-made product customization (Berman, 2002, Da Silveira et al., 2001, Duray, 2002 and Kotha, 1996). It could be argued that mass customization offers superior customer value compared to other strategies (e.g., Gilmore and Pine, 2000 and Tu et al., 2001). It is highly suggested that swarm intelligence mechanisms, as the one presented here, should become an integral part to specialized management information systems for mass customization production programmes in business-to-business markets. Of course, the integration of such AI-based approaches to mass customization production processes, implies that a) the selling company knows the specific needs of its customers and is willing to make those necessary investments to accommodate them, b) manufacturers assess their market demand in order to define the type and extent of mass customization they can really offer, and c) the supplier assesses the level of variety at which customers will still find its offerings attractive, as well as the level of complexity that will keep the production costs low (Child, Diederichs, Sanders, & Wisniowski, 1991). Undoubtedly, the proposed methodology can assist manufacturing firms in designing customer-driven industrial product lines, as a means to improve their relationships with key accounts, increase customer satisfaction and enhance loyalty levels. 5.2. Limitations and future research The proposed method, however, is not without limitations. First, in the present paper we demonstrate that the proposed optimization procedure provides an effective solution to the challenging product line design problem through a simulation study. Our mechanism should be also tested through empirical data and conjoint experiments. The literature in the area of conjoint analysis has grown at an impressive rate and considered different types of conjoint tasks (i.e., rankings, ratings, choices), alternative data collection techniques (i.e., compositional, decompositional and hybrid approaches), as well as techniques to estimate structural parameters (e.g., random utility, latent class and hierarchical Bayes estimation methods). It would be desirable to test our framework in empirical settings by the application of conjoint techniques, such as adaptive conjoint analysis and choice-based conjoint. Of particular note would be the use of hybrid conjoint and hierarchical Bayes conjoint techniques in this framework. Second, to keep the scope of the study within reasonable limits, the preceding analysis does not consider the cost structure of a multi-product, industrial line. The extension one might consider is to include cost variables and provide a more integrated analysis of the forces that shape optimal product line decisions. Third, developing a Product Family Architecture has been proposed as a strategy of implementing mass customization (Jiao and Tseng, 2000 and Tseng and Jiao, 1996). A product family architecture captures and utilizes commonality (similarity). Each new product instantiates and extends so as to anchor future designs to a common product line structure; therefore, a class of products can widely variegate their designs in response to diverse customization requirements. Due to the commonality (similarity) over the product line, the product family suggests itself as a natural technique to facilitate increasingly efficient and cost-effective product development (Meyer, Tertzakian, & Utterback, 1997). Designing a family of products, instead of single products, using a common platform approach, has gained momentum in various industrial markets. However, one of the main challenges faced by companies is to specify an optimal product family architecture to support the varieties. The implementation of swarm intelligence mechanisms could be particularly useful in that direction. PSO can be used as a means to achieve economies of scale and standardization of production by developing optimal product family designs. On the one hand, this will reduce the cost of extensive customization, since the products of the family will be sharing some component commonality indices, but, on the other hand, the characteristics of those products will be allowed to take any value from a continuous range, resulting in products that will better meet customer expectations. Finally, it would be interesting to examine whether heuristic methods presented in other optimization problems could be also extended to the optimal product line design problem. For example, future research could test the applicability of algorithms and techniques presented in the optimal product-mix problem (Bhattacharya & Vasant, 2007), and the industrial production planning problem (e.g., Vasant, 2011 and Vasant et al., 2012). In conclusion, this study introduced a new hybrid PSO mechanism to industrial marketing management and showed how this framework can be applied to identify optimal product portfolios. The approach stated above is integrated and comprehensive. We hope that the ideas presented here are demonstrative of the potential of artificial intelligence-based solutions to similar decisional situations and might motivate further developments in this emerging area.