The pioneering work of Boothroyd and Dewhurst [27] in developing the design-for-assembly guidelines has resulted in several automated assembly evaluation and advisory systems [16,20]. Swift [28] also presented a methodology similar to that of Boothroyd and Dewhurst. One of the earliest efforts in this direction was made by Jakiela et al [16]. They integrated a rule-based system with a CAD system to develop a design advisory system. This system provides a library of predefined features with which the designer can create a design. When new features are added to the design, the system makes use of production rules to evaluate the design and offer suggestions for improving it. In this approach, the designer designs the parts using the features offered by the library. This system works incrementally. as the design progresses,---offering advice at every design step. Hence, the design improvement suggestions are strongly influenced by the sequence in which the designer enters various features.
Sturges et al. [29] have developed a semi-automated assembly evaluation methodology that attempts to overcome some of the limitations of the scheme proposed by Boothroyd and Dewhurst [27]. Currently, while lacking geometric reasoning capabilities, their system serves as an interactive environment to study the effect of various design configurations on assembly difficulty.
Li and Hwang [30] did a study of design for assembly and developed a semi-automated system which closely follows the Boothroyd-Dewhurst methodology. The analysis of assembly difficulty and cost estimation modules are a direct computer implementation of the DFA rules. Their methodology considers multiple assembly sequences and calculates the time for all of the feasible sequences. They perform limited feature recognition for assembly and obtain from the user the non-geometric information that will affect the assembly. The final result is a table which is the roughly the same as a manual assembly worksheet. The authors argue that the assembly information developed quickly and in proper format will give the designer enough input to perform further analysis for design modification. The task of automated redesign is presented as a future goal.
One of the first efforts in to develop possible assembly sequences and selecting suitable ones using manufacturing information was done by De Fazio and Whitney [31,32]. Hsu et al. [20] developed an approach to design-for-assembly that examines and evaluates assembly plans using three criteria: parallelism, assemblability, and redundancy. They evaluate the plan to find the problems with the assembly. When possible, a better assembly plan is created by modifying the plan. If a better plan is found, the design is modified by splitting, combining or perturbing various components. Although limited in certain ways, this offers a new plan based approach.
Although the Hitachi Assemblability System [33,34] was not initially computerized, over time it served as a basis for development of automated assemblability system. The methodology is based on the principle of one motion per part; there are symbols for each type of assembly operation and penalties for each operation based on its difficulty. Finally, the method computes an assembly evaluation score and assembly-cost ratio. The assembly-cost ratio gives an indication of current assembly cost to previous cost. The methodology is common for manual, automatic and robotic systems. One of the early success stories of this method is highlighted in [35].
Miles et al. [36] also developed an assembly evaluation method in which parts are divided into two groups based on functional importance: ``category A'' parts are required from the design specification, and ``category B'' parts are accessories. The goal is to eliminate as many type B parts as possible through redesign. Analyses of feeding and fitting is carried out on the parts, with both results combined into a total score. This total is divided by the number of type A parts to obtain a final score. A proposed assembly sequence is used. to perform fitting analysis.
Warnecke and Bassler [37] studied both functional and assembly characteristics. Parts with low functional value but high assembly difficulty receive low scores, while parts with high functionality and low assembly cost receive high scores. The scoring is used to guide the redesign process.
Recently Jared et al. [38] presented mathematical models for the assembly operations and a DFA system that performs geometric reasoning based on the model. In this way, they rely less on user input. Their system calculates a manufacturability index for individual components and fitting index between the components.
Boothroyd presents a review of design for manufacture and assembly methodologies being employed by different companies in [39].