Decision Making in Engineering Design
18 Multilevel Optimization for Enterprise-Driven Decision-Based Product Design
Download citation file:
- Ris (Zotero)
- Reference Manager
There is a growing recognition in the design community of the need for a rigorous design approach that considers the enterprise goal of making profits and the decision-maker's risk attitude, while also dealing adequately with engineering needs and various sources of uncertainty. Decision-based design (DBD)  is a collaborative design approach that recognizes the substantial role that decisions play in design and in other engineering activities. In recent years, we have seen many (DBD) related research developments in the field of engineering design (e.g., [2; 3; 4; 5; 6; 7; 8; 9; 10; 11; 12, 13; 14; 15; 16]). For profit-driven design under the DBD framework, ideally, all product design decisions, whether directly related to engineering or otherwise, are made simultaneously to optimize the enterprise-level design objective, i.e., to maximize the expected utility, expressed as a function of net revenue (profit), subject to various sources of uncertainty. The existing implementation of the profit-driven DBD approach [11; 12, 13] seeks to integrate enterprise product planning and engineering product development by using an all-in-one (AIO) approach and solving as a single optimization problem. The enterprise is defined here as the organization that designs and produces an artifact to maximize its utility (e.g., profit). Marketing, production planning and other enterprise-level activities are referred to as enterprise-level product planning; engineering-related design activities are referred to as engineering product development.
Designing a large-scale artifact typically involves multidisciplinary efforts in marketing, product design and production making. The AIO approach is often practically infeasible in such situations due to computational and organizational complexities. Optimization by decomposition, while alleviating the problem of having to deal with a large number of design variables and constraints at the same time, is made necessary by a number of factors. The decomposed approach helps enable simultaneous multidisciplinary optimization wherever possible and also addresses organizational needs to distribute the work over several groups of engineers∕analysts. Historical evolution of engineering disciplines and the complexity of the multidisciplinary design optimization (MDO) problem suggest that disciplinary autonomy is a desirable goal in formulating and solving MDO problems. In MDO, several design architectures have been developed to support collaborative multidisciplinary design using distributed design optimization, e.g., concurrent subspace optimization (CSSO) , bilevel integrated system synthesis (BLISS) [18, 19], collaborative optimization (CO) [20; 15], and analytical target cascading (ATC) [21; 22, 23, 24, 25; 27]. A comprehensive review of the MDO architectures is provided by Kroo . It should be noted that the choice of MDO formulations largely depends on whether the problem follows the hierarchical or non-hierarchical characteristics of decision flow. In most of the MDO approaches listed above, a complex engineering problem is non-hierarchically decomposed along disciplinary or other user-specified boundaries into a number of sub-problems. Then they are brought into multidisciplinary agreement by a system-level coordination process. In our opinion, the non-hierarchical MDO infrastructure is better suited to capture the interrelationships between multiple engineering disciplines in engineering-level product development; however, a hierarchical approach, such as ATC, is more appropriate in an enterprise-driven product design scenario where the enterprise decision-making is often done at a higher level to set up targets for engineering product development. To represent the organizational infrastructure in industry more accurately, the interrelationships between enterprise product planning and engineering product development, as well as the engineering product development itself at system, subsystem and component levels, should be treated as hierarchical. Such a hierarchical framework, as will be detailed later, is more representative of the hierarchical decision-making in the industry.
In this chapter, we present a DBD-based hierarchical approach to enterprise-driven design that treats enterprise-level product planning and engineering-level product development as two interrelated but separate optimization problems in a multilevel optimization framework. To fully integrate business and engineering decision-makings, we illustrate how a disaggregate probabilistic choice model can be used to establish the link between the decomposed enterprise product planning and engineering development models. Any hierarchical approach, like the one presented here, should ensure preference consistency. In other words, the optimization of the engineering objectives at the product development level needs to correspond to the maximization of the utility objective function at the enterprise product planning level. This is to guarantee that the solution from the multilevel optimization procedure will be close, if not identical, to the one that is obtained by solving the AIO-integrated enterprise and engineering problem. As will be discussed in Section 18.3, if the feasible domain imposed by the engineering product development is disconnected in the space of engineering performance attributes, achieving the design which corresponds to the maximum enterprise utility becomes more challenging. A search algorithm that can systematically explore attribute targets in the disconnected feasible domain to lead the engineering product design to feasible and optimal designs in the enterprise context is needed and such an algorithm is also presented here.