Parameter estimation of an ion-exchange process based on a pseudo probability measure

Abstract

The manufacturing of integrated optical waveguide components can be complex if the design space is highly multivariate or constraints introduced by the process itself are not well-known. Numerical simulation helps to create a basic understanding of the process and the resulting components with their features but is costly with respect to time and hardware resources. The estimation approach in this work introduces a precise, numerically robust and controllable method without any black boxes to estimate the input parameters of the manufacturing ion-exchange process. Therefore, small number of simulation samples is used to determine basic dependencies of desired waveguide component features on the input parameters. A pseudo probability measure is then used to estimate the combination of process parameters that is most likely to result in the required features. The feature data is detached from the sample size of the simulation data with a multidimensional curve fit in order to debloat the estimation model. The implications of the estimation are wide-ranged and not restricted to the physics of the ion-exchange process, i.e. easy to be abstracted to other models. The estimation enables parameter subspace optimization or clustering of solution. Model benchmarks show a very high estimation rate for all types of considered applications.