T. Coen

Nonlinear system identification on a combine harvester

The traction system of a combine harvester contains considerable nonlinearities. The objective of this paper is to derive a model of the propulsion which can then be used for regulator design. First the nonlinearities are quantified by analyzing the output of the system excited by a multisine. Standard linear system identification techniques (such as ARX and ARMAX) are then compared to a more recent nonlinear state-space technique. Finally the derived models are validated on several alternative input signals

ENGINE SOUND COMFORTABILITY: RELEVANT SOUND QUALITY PARAMETERS AND CLASSIFICATION

Abstract In order to be able to shorten the design cycle, automobile manufacturers are interested in modelling the human perception of engine sounds. In the first part of the paper the relevant Sound Quality parameters for the prediction of engine sound comfortability are determined. The inputs are ordered with Automatic Relevance Determination and the obtained ranking is verified on the data. In the second part, models are presented to classify and compare cars on comfortability. Least Squares Support Vector Machines (LS-SVMs) is used for the classification. The influence of selecting the relevant inputs on the model performance is illustrated.

Nonlinear State Space Modelling Of Multivariable Systems

In this paper we propose a method to model nonlinear multivariable systems. We will use a state space approach since this is inherently compatible with Multiple Input, Multiple Output (MIMO) systems. The basic idea is to fit first a linear model on the measured data, and to extend that model so that it can grasp the nonlinear behaviour of the system. The results are applied to a physical system.

A Stochastic MPC approach to controlling biological variable processes

Abstract The main challenge when controlling agricultural machinery is the biological variability of the incoming crop. This variability renders any process model time-variable and uncertain. Most robust control techniques start with a computationally intensive initialisation method which has to be repeated each time the model changes. Moreover, these techniques mostly focus on process stability, and not process performance. In this paper a different approach is taken with focus on process performance and attention for the computational complexity. The Stochastic MPC framework is used to design a controller that responds as swiftly as possible at all times. Two models are defined, one for the process variables and one for the model error on each of the process variables. The actual configuration parameters of the Model-based Predictive Controller are then calculated in each time step based on the operator settings and the estimated model error. The model error is also used to convert the deterministic process constraints into stochastic constraints which are respected with a given accuracy. Finally this approach is implemented and validated on a capacity control system for a combine harvester.

Throughput control on a combine harvester using Model-based Predictive Control

Since it is no longer possible to increase the size of combine harvesters, increasing the machine efficiency is the only way to further raise the capacity. The most straightforward way to do so is to make every operator behave as the optimal operator, and this 24/7. This paper presents on a control system that regulates the speed of the combine harvester based on process variables such as feed rate, engine load and grain mass flow such that the capacity utilization is maximized at all times.

IDENTIFICATION OF SENSOR FAULTS ON COMBINE HARVESTERS USING INTELLIGENT METHODS

Abstract Process monitoring and fault diagnosis is of considerable interest from an industrial perspective. In this paper, the general applicability of intelligent methods, like self-organizing maps (SOM) and multilayer feedforward networks with backpropagation, for the identification of sensor failure on combine harvesters will be illustrated. Both neural network types showed comparable results in order to classify normal and faulty sensor conditions.