Method of Multi-objective Design of Strain Gauge Force Sensors Based on Surrogate Modeling Techniques

Abstract

The design of strain gauge force sensors is usually simulation-driven, requiring time-consuming finite element analyses to be conducted multiple times, limiting the efficacy of the simulation-driven approach. In this paper, we propose overcoming this limitation with a method of multicriteria optimization of strain gauge force sensors based on surrogate modeling techniques. Surrogate models approximate the relationship between the design variables and the elastic element stress-strain state's properties that determine the sensor characteristics. The proposed method's key feature is that a special surrogate model reconstructs the entire distribution of strain on the elastic element, which is required for determining the optimal placement of strain gauges. The surrogate models are neural network (NN)-based. We benchmarked the NN-based surrogate models against the RSM-based surrogate models in a case study. Based on the case study results, the RSM-based models are on par with the NN-based models in terms of predicting almost linear relationships between goal function and design variables, while the RSM-based model failed to reconstruct the strain distributions due to their nonlinearity accurately. The proposed method allows accelerating the design of strain gauge force sensors while retaining the possibility of determining the optimal placement of strain gauges on the elastic element.