This proposed research provides funding for the development of efficient methods for measurement, control, and optimization of surfaces of manufactured products. The problem is challenging due to the complex nature of the profiles that exhibit nonstationarity and spatial correlation. Some type of surface measurements are time consuming making the online control infeasible, whereas some other types of surface measurements such as those based on images provide huge amounts of data making the estimation of the models computationally intractable. The need of modeling the manufacturing process variables for the purpose of optimization and adaptive control makes the problem even more challenging. These challenges are addressed in this project by proposing a novel easy-to-evaluate metamodel known as kernel-sum regression/interpolation and developing efficient engineering-driven sampling and modeling strategies.
If successful, the results of this research will help improve the quality and functionality of many manufactured products. The applications of the proposed research are not limited to engineered surfaces. The new metamodeling method can be applied in areas such as geology, environmental modeling, and astronomy. The new sampling strategies are also applicable in other areas such as in the large-scale computer and simulation experiments, which open its applications to a wide variety of problems in product design and manufacturing. The impacts of the research can also go beyond the manufacturing applications, such as improving the surface characteristics of solar cell for better energy conversion and improving the precise surface characteristics of hip/dental implants leading to better quality of life and the reduction of healthcare cost.
