An Approach to Surrogate Physical Systems Based on Dimensional Analysis and LASSO Regression Demonstrated on a Flexible Multibody System

Implementing model-based controllers using complex physical models on industrial control units is often challenging and sometimes impossible due to limited hardware resources. However, it is often sufficient to use simplified models for control design. In this paper, an approach based on dimensional analysis and LASSO regression (Least Absolute Shrinkage and Selection Operator) is presented, which is suited to surrogate complex physical models by simplified input-output relations. The approach is demonstrated for the natural frequency of a flexible multibody system representing a telescopic element, frequently used in vehicles transporting goods such as forklifts. The result is an exponential function which is linear in its only nine coefficients approximating the natural frequency with sufficient accuracy.