A multi-objective optimization using distribution characteristics of reference data for reverse engineering

Abstract

As different industries produce similar products, engineers tend to analyze the products of competitors and adopt the excellent merits in their current products. This process is called reverse engineering. There can be multiple target characteristics in reverse engineering. In many cases, the improved design from reverse engineering usually keeps the data distribution characteristics of the competitors unless the developed product is a fully new creative design. The distribution should be considered in reverse engineering. Therefore, the reverse engineering process can be modeled as multi-objective optimization considering data distribution. Recently, Taguchi developed the Mahalanobis Taguchi system (MTS) technique to minimize the Mahalanobis distance (MD), which is defined by a multi-objective function with data distribution. However, the MTS technique has the limit that the new design is not better than the mean values of the competitors. In this research, a function named as the skewed Mahalanobis distance (SMD) is proposed to overcome the drawbacks of the MTS technique. SMD is a new distance scale defined by multiplying the skewed value of a design point to MD. SMD is used instead of MD and the method is named the SMD method. The SMD method can always give a unique Pareto optimum solution. To verify the efficiency of the SMD method, a non-convex mathematical example, a cantilever beam, and a practical automobile suspension system are optimized. Copyright (C) 2010 John Wiley & Sons, Ltd.