Eiji Adachi

Multiobjective Satisfactory Design with Sensitivity Coefficients Obtained by the Semantic Differential Method.

To realize actual designs in computer aided design, we introduced a design method by the attainability and simultaneous equations to determine structural dimensions in order to make all requirements accomplish their respective objective values. We also introduced an identification method of sensitivity coefficients without theoretical design formulae. We present a multiobjective satisfactory design method that includes non-technical objective functions. Design formulae of the functions are described by orthogonal polynominals, using panel evaluations by the semantic differential method and the regression analysis. Partial derivatives of the formulae give sensitivity coefficients. The iterative solving of the simultaneous equations with the coefficients yields converged or preferred solutions of the design. In a bath unit design with 6 objective functions and 9 design variables, this method produced more precise solutions than the current models in 2∼25 seconds.

Sensitivity Coefficient Identification by Trial Production and Multiobjective Satisfactory Design.

To materialize actual designs in computer-aided design, we presented a method to determine structural dimensions in order to make all design requirements accomplish their respective objective values, using the attainability and simultaneous equations with several unknowns. However, most of the designs have no theoretical design formula or sensitivity coefficients. This paper presents an approximation method to identify sensitivity coefficients by trial production evaluations of the designs. The coefficient matrix is specified between k th revised values of the attainabilities and k th revised values of the design variables, and made into a square or a triangular matrix by neglecting insignificant variables. The matrix is solved iteratively until all coefficients converge. This paper also presents a multiobjective satisfactory design using the identified coefficients to prove the effectiveness of the identification, comparing the identified coefficients with the theoretical ones.

Multiobjective Satisfactory Design by Simultaneous Equations with Several Unknowns.

To materialize the actual designs in computer-aided design, we presented a method to determine the structural dimensions (design variables) in order to make all the requirements (objective functions) accomplish their respective objective values, using a newly introduced attainability which can describe heterogeneous objective functions in the same form. This paper presents another method which uses the attainability and simultaneous equations with several unknowns. The equations are composed of (k-1) th attainabilities, kth sensitivity coefficients, and kth revised values of the design variables which number as many as the attainabilities; kth represents the kth iterative calculation. The solutions of the equations give kth revised variables and kth revised attainabilities. The equations are composed and solved iteratively until all attainabilities become zero. In an automotive disk brake design, this method yielded satisfactory results in less iteration number than the former method did.

Multiobjective Design Satisfaction with Heterogeneous Objective Functions

Actual design methods are multiobjective design satisfactions with heterogeneous objective functions such as performance and cost, and with dimensionaL design variables such as part size and distance. To materialize the actual designs in computer-aided design, this paper presents a new method which determines structural dimensions (design variables) in order to make all objective functions accomplish their respective functions. Heterogeneous objective functions can be described in the same form (a new concept termed attainability), which apparently indicates the satisfied functions by a positive sign or zero. The sensitivity coefficient matrix provides the most successful variables which can improve a minimum, as well as a negative, attainability to positive or zero, and simultaneously keep all positive attainabilities positive or zero. Values of the variables are revised iteratively for this purpose. In an agricultural transmission gear design, the method produced significant and practicable results in short CPU time.

Effects of Vibration Modes on the Viscoelastic Loss Factor Measured by the Half-Power Method.

The loss factor of a viscoelastic damping material is commonly measured by the half-power method using a composite beam with the viscoelastic layer. However, the method involves problems of large measurement errors and of the beam shape, which does not represent practical applications. This paper discusses the relatinship between vibration modes and measurement accuracy, and presents specimen dimensions which materialize the bending mode for the accurate loss factor. The vibration modes are classified into torsion, warping, torsion and warping, and bending. According to the discriminants this study induces, the beam shape has satisfy (l/b)(d1/b)≥0.06 to materialize the bending mode. The loss factor of the viscoelastic damping material obtained by the composite beam within the dimension limits and the Oberst theory had good agreement with the loss factor of the same material measured by the nonresonance forced vibration method, and was certified to be sufficiently accurate.

Satisfactory design of railway passenger cars using sensitivity analysis and empirical knowledge.

Railway passenger car bodies have been parametrically designed based on empirical knowledge of expert designers. To meet demands for systematic and satisfactory design, this paper presents a new computer aided design method to determine satisfactory structures and dimensions using sensitivity analysis and empirical knowledge. Data of a tentatively designed structure is inputted into a computer using a product modeling technique. The structure produces an evaluation model consisting of sections called blocks which in turn consist of components. Introducing sensitivity analysis we can evaluate the effectiveness of the weak parts of all blocks and all parameters of the components. Responding to the analysis in terms of empirical knowledge, the priority list of parameters to be modified is decided, since designers expect the new method to be as usable as conventional methods. On a railway passenger car we obtained a practical solution in less than 28 hours, after examining all possible cases of the modification.