Fundamental Processes of Mechanical Designers Based on Empirical Data

Abstract

The fundamental processes of problem solving during the mechanical design activity are presented based on a study of human designers. The designers in this study solve problems by applying 10 specialized actions, called operators, which are grouped into the categories of 'generate', 'maluate' and 'decide'. These operators are applied in unique sequences of which 95% constitute the following four local methods: 'generate and test' (23%), 'generate and improve' (8%), 'deductive thinking' (33%) and 'means end analysis' (36%). Furthermore, cognitive activity is reduced to a hierarchical architecture of tasks and episodes which repect the goal structure of the designer. These descriptions demonstrate the abilities and limitations of the human designer's problem-solving performance, yielding an insight into how the man-machine interaction of design automation can be improved.

A Systematic Review of Protocol Studies on Conceptual Design Cognition

“…Stauffer and Ullman [23] suggest that during designing, the processor component of the IPS "accesses information from the LTM [long term memory] into the STM [short term memory, i.e. working memory] as it is needed by the operators" (p.117).…”

“…Operators may be applied in different combinations and sequences to reach a solution and manage the search process. These patterns of operator execution are described as search methods [23] and control strategies [22], respectively. Owing to the necessary space limitations of a conference paper, the full range of search methods, control strategies, and operators identified from the reviewed studies cannot be presented here.…”

“…Central to this viewpoint is the perspective that a designer may be modelled as an information processing system (IPS) [22,23]. As an IPS, a designer transforms information from input to output states via the execution of elementary information processes known as operators [23]. These transformations may be termed state transformations [2,20].…”

“…Design problem spaces are typically large owing to the illdefined nature of design problems; however, a designer may reduce the size of the space to be searched by implementing constraints [2,22]. Search processes are further managed with design goals specifing desired design states [22,23].…”

“…The problem space encompasses knowledge of the initial problem state, a goal state, and all possible design states in between these two states. Design as search may then be viewed as a sequence of state transformations, beginning with the problem state and proceeding through intermediate design states until the goal state is reached [2,[22][23][24]. Transformations may be (i) lateral, i.e.…”