Self-Calibrating Mass Flow Sensor

Abstract

A comprehensive analysis was conducted for increased accuracy and self-calibration for a mass flow sensing system on a combine. This was undertaken as part of the John Deere Technology Innovation Center (JDTIC)-sponsored research program "Self-calibrating mass-flow sensor", in turn part of a John Deere Moline Technology Innovation Center (MTIC) effort toward optimization and closer integration of the components of the mass-flow sensing system in Deere harvesting combines. The long-term objective was to achieve a self-calibrating sensor system capable of adapting to varying input conditions due, for example, to changes in grain moisture content and aging of the system's elevator paddles.In analyzing the mass flow sensing system, a physics-based model was developed to describe the relationship between the rate of mass flow through the combine and the measured force imparted to the impact plate in terms of mechanical properties of the grains and the interior geometry of the combine. A computational realization of this model was constructed in Matlab. Accurate mass flow rate estimation was achieved through model-based estimation based on nonlinear regression applied to the physics-based model and data acquired through simulation and experimentation. Model-based estimation was also extended as a means for self-calibration of the sensing system. Through development of the physicsbased model, the dependence of the force imparted to the impact plate on the orientation of the impact plate was identified. By inducing known changes to the impact plate orientation and implementing model-based estimation, a means of self-calibration of the sensing system was achieved. Three methods of model-based estimation were successfully demonstrated using data generated from the physics-based model. Additionally, these were further verified using data collected from discrete element modeling simulations, and experimental data collected in two fashions: using a full-scale replica of the mass flow sensing system, and using a small-scale, benchtop testing apparatus. Furthermore, the ability of the developed algorithm to update theoretical model parameters while simultaneously estimating mass flow rate was shown to enable the system to self-calibrate. This was argued to allow the system to accommodate different operating conditions that may be encountered during combine harvesting, such as changes in crop moisture, grain variety, and aging of combine components.ii

A dynamic grain flow model for a mass flow yield sensor on a combine

“…Data was collected from simulations involving differing amounts of grain using the EDEM software. For a complete description of the simulations and modeling efforts, see Reinke (2010). Figure 9 shows the results of simulations for differing amounts of grain mass, each of which were repeated for a total of three simulations for each amount of mass.…”