Design and validation of mechanical multi-axis force/moment sensors

Abstract

Multi-axis force/moment sensors are capable of measuring three forces (Fx, Fy, Fz) and three moments (Mx, My, Mz) in three-dimensional space over time. They are mechanical sensors in which force and moment loads create deflection and strain. These physical changes are converted into electrical signals, producing a sensor that can measure forces and moments. Victor Scheinman introduces the first six-axis force/moment sensor that has a cross-beam compliant mechanism and strain gauges for the Stanford Arm in the late 1960s. Since then, these multi-axis force/moment sensors have been used in various robotics and automated systems. Over the years, various mechanical structures and sensing principles have been tested, and the sensors have been improved, becoming more efficient along with the technological development. Cross-beam and y-shaped-beam structures are the most common compliant mechanisms for the multi-axis force/moment sensors. The sensor structures are compact, easily machined with high precision, and have good sensitivity across all force and moment axes. The piezoresistive effect is the most common sensing technology implemented for multi-axis force/moment sensors. Strain gauges are used as sensing elements that are simple, reliable, and cost-effective. This thesis aims to design and optimize mechanical cross-beam and y-shaped-beam multi-axis force/moment sensors by proposing a new theoretical model that can rapidly predict sensor properties within an acceptable error margin compared to finite element method models. Current theoretical models lack the required accuracy and cannot estimate the equivalent stress and natural frequency. Moreover, the accuracy of these theoretical models is not validated by a comprehensive statistical analysis. The shortcomings of the theoretical models affect the sensor optimization. Multi-axis sensor optimization at the current state is stuck in a local domain of the initial sensor design, optimizing a few numbers of sensor dimensions. This thesis also deals with this optimization problem. Benefits of the proposed theoretical model are demonstrated by comparing the single-objective optimization cases from the literature. Moreover, a novel optimization problem formulation for multi-objective optimization without any initial design is suggested. This proposed optimization problem formulation is not possible with prior theoretical models because they cannot estimate the required sensor properties and is not feasible with finite element method models because the optimization problem is very time-consuming. The new theoretical model and optimization problem formulation are applied for cross-beam and y-shaped-beam sensor structures, and two compact sensor designs that have high structural reliability and natural frequency and are optimized for higher sensitivity are offered. Sensor prototypes are built in order to validate the proposed theoretical model and the problem formulation.