K. Maertens

ON-LINE MEASUREMENT OF GRAIN QUALITY WITH NIR TECHNOLOGY

During the summer season of 2000, a feasibility study is carried out to measure grain quality (moisture and protein content) with Near Infrared Reflection (NIR) technology on a conventional TX64 New Holland combine harvester. A Zeiss Corona 45 NIR 1.7 sensor was installed on a bypass of the clean grain elevator. Parallel with the NIR measurement, grain samples were taken at the end of the bubble-up auger that transports the grain from the clean grain elevator to the grain bin. Measured signals at the diode array of the spectrophotometer are highly variable in time and since data is acquired at a sampling rate of 23 Hz, an appropriate low pass filter is designed to obtain every second a relevant spectrum. In addition, an optimal time shift is calculated between the location of the NIR measurement and the sampling spot, in order to make the comparison between the estimated spectra and grain samples as accurate as possible. The spectra are first spectrally converted to either absorbance or Kubelka-Munk values. The required spectra of a white and black standard are time-interpolated between the standard measurements executed closest before and after the selected on-line spectrum. Preprocessing, calibration and validation are executed using the PLSplus/IQ module included in the GRAMS/32 software package. After application of Mean Centering, the Multiple Scatter Correction, Standard Normal Variate and detrending algorithm are applied. The calibration models are developed using the PLS algorithm and validated through cross-validation.

Genetic polynomial regression as input selection algorithm for non-linear identification

The performance of non-linear identification techniques is often determined by the appropriateness of the selected input variables and the corresponding time lags. High correlation coefficients between candidate input variables in addition to a non-linear relation with the output signal induce the need for an appropriate input selection methodology. This paper proposes a genetic polynomial regression technique to select the significant input variables for the identification of non-linear dynamic systems with multiple inputs. Statistical tools are presented to visualize and to process the results from different selection runs. The evolutionary approach can be used for a wide range of identification techniques and only requires a minimal input and a priori knowledge from the user. The evolutionary selection algorithm has been applied on a real-world example to illustrate its performance. The engine load in a combine harvester is highly variable in time and should be kept below an allowable limit during automatic ground speed control mode. The genetic regression process has been used to select those measurement variables that have a significant impact on the engine load and that will act as measurement variables of a non-linear model-based engine load controller.

Flow Rate Based Prediction of Threshing Process in Combine Harvesters

The threshing section affects grain separation performance in combine harvesters. Proper configuration is critical to ensure an optimal economic working of combine harvesters. Predicting grain separation online is essential for an automatic machine tuning system. An indirect process–monitoring device was designed based upon feedrate measurements. Stationary tests provided information about the behavior of grain separation during the threshing process as a function of different feedrates and crop properties. Six different theoretical threshing models were analyzed. For three models, a comparative study has been carried out to choose an optimal predictor of the separation process. The most accurate prediction model is based on Rusanov’s exponential model. The confidence interval necessary to analyze the residual value of the predictor was calculated for each predicted separation curve.

Comfort improvement by passive and semi-active hydropneumatic suspension using global optimization technique

The arising health problems of operators of off-road vehicles point out that still a lot of effort has to be put into the design of effective seat and cabin suspensions. The comfort problem originates form the vibrations transmitted to the driver caused by the unevenness of the road or soil profile. The paper develops a method for optimizing suspension systems using the global optimization technique DIRECT. Evaluation of the comfort improvement of the suspensions is done using the objective comfort parameters used by the norms. The design procedure is tested on a hydropneumatic suspension that can serve as a cabin suspension for off-road machinery. The evaluation shows that the vibration levels can be reduced by 90%, resulting in a drastic comfort improvement.

Design of a Friction Independent Mass Flow Sensor by Force Measurement on a Circular Chute

Abstract A mass flow sensor for bulk flow measurements is designed. The sensor consists of a curved chute and the force executed by the flow on the chute is a measure for the force. By properly selecting the projection angle of the force, i.e . the direction in which the force is measured, the sensor can be made almost independent of the friction coefficient of the bulk. The selection of the projection angle involves an optimisation problem. To carry out the optimisation, a physical model of the differential force is derived. The differential force is the force executed on the chute by an infinitesimal mass. Based on the model of the differential force, a starting value for the optimisation process is computed. After the optimisation the influence of friction on the measurement is less than 0·5% per 0·1 change of the friction coefficient. For some installation conditions, the effect of friction on the force measurement can be completely eliminated.

Design of a virtual combine harvester

Separation processes in agricultural machinery are typically non-linear, complex and uncertain. On-line measurements of input and output data are not available. Therefore, black-box modelling is not applicable and other techniques to describe the system must be used. In this paper, a dynamic separation model is constructed to predict the input-output behaviour of the separation section based on a combination of physical process knowledge and experimental results.When the hybrid separation model is implemented in a framework to simulate the total harvesting process, a virtual combine harvester is obtained, ready to process virtual fields that are created with standard software programs.

Development of a smart mass flow sensor based on adaptive notch filtering and frequency domain identification

A popular method for measuring mass flows of granular materials is to register the induced force on a circular chute. When properly designed, an accurate, friction independent and linear flow measurement is achieved, ready to install in agricultural machinery or process lines. Since this type of sensor is strongly subject to wear and is installed in inaccessible places, it is important to know when it provides wrong measurements without interrupting the process. In this study, an off-line frequency domain procedure is designed to estimate wear on the chute suspension based on normal grain flow variations without extra excitation devices. The uncertainty bounds on the dynamic parameters are iteratively estimated based on a non-linear transformation between the parameter and pole/zero space. As an alternative, a robust and stable method is presented to track the resonance frequency of the mechanical sensor and paddle rate of a feeding elevator on-line.

Using a Virtual Combine Harvester as an Evaluation Tool for Yield Mapping Systems

Different types of yield mapping systems for combine harvesters have been proposed in the literature. Each system is characterized by its composition of positioning system, grain flow measurement device, ground speed sensor and eventually cutting width sensors. The individual accuracy of each device has been optimised, without looking at their own impact on the accuracy of the global system.By developing a virtual harvesting process, it is possible to “harvest” virtual fields several times with different machine settings, ground speed strategies and measurement devices. Subsequently, the accuracy of different set-ups can be compared and the required accuracy for each individual measurement device can be quantified. In this way, it is possible to design a balanced composition of measurement devices for a predefined budget.

Noise cancellation in on-line acoustic impulse response measurements for the quality assessment of consumption eggs

Vibration analysis is a challenging technique to assess the quality of agro-products in a non-destructive way. Recent research shows that the technique is feasible for off-line (desktop) measurements. The step towards an on-line implementation requires research on the robustness of the parameters in the presence of surrounding noise, since a microphone is used to capture the vibration characteristics of the products. Two different methods are used to cancel environmental noise from acoustic impulse response measurements. As a practical example measurements were taken on consumption eggs. The first technique, called independent component analysis (ICA), makes uses of statistical principles. The main disadvantage is the fact that the algorithm is, in its present form, a batch algorithm rather than a recursive algorithm. The second algorithm is based on the adaptive interference cancelling principle. The major drawback is the instability of the parameter estimates that can occur under certain conditions.

Genetic polynomial regression as input selection algorithm for engine load prediction in combine harvesters

Abstract The performance of non-linear identification techniques is often determined by the selected input variables and the corresponding time lags. High correlation coefficients between candidate input variables in addition to a non-linear relation with the output signal induces the need for an appropriate input selection methodology. This paper proposes a genetic polynomial regression technique to select the significant input variables for the identification of non-linear dynamic systems with multiple inputs. An evolutionary measure is presented to visualize and to process the results from different selection runs. This evolutionary approach was applied to select an optimal set of variables for engine load prediction in combine harvesters.