DE19540507A1

DE19540507A1 - Determining moisture level of sand

Info

Publication number: DE19540507A1
Application number: DE1995140507
Authority: DE
Inventors: Guenter Jahl; Reinhard Dr Fritz; Wolfgang Dr Schroer
Original assignee: FACHHOCHSCHULE ULM
Current assignee: FACHHOCHSCHULE ULM
Priority date: 1995-10-31
Filing date: 1995-10-31
Publication date: 1997-05-07

Abstract

The method involves inserting two electrodes into the sand whose degree of moisture is to be measured. The real and imaginary parts of the complex resistance between the electrodes are measured at two frequencies. From these values, four moisture dependent components of an electric model of the sand are calculated. These components are combined into five characteristics and are fed as input values to a trained neural network. The neural network provides an output signal indicative of the sand moisture. The neural network may be trained using a reference sand with a known moisture level, determined e.g. by drying and weighing the sand. The network may be a multilayer perceptron with back propagation training algorithms.

Description

Vorgestellt wird ein Verfahren, bei dem der komplexe Widerstand zwischen zwei in Sand eingetauchten Elektroden bei zwei verschiedenen Frequenzen gemessen wird. Aus diesen Messungen wird ein elektrisches Modell des feuchten Sandes mit vier feuchteabhängigen Bauelementen abgeleitet (Abb. 1). Aus den Bauelementen des Modells wird die Feuchtigkeit mit Hilfe eines neuronalen Netzes bestimmt, das mit einem Referenzsand trainiert wurde.A method is presented in which the complex resistance between two electrodes immersed in sand is measured at two different frequencies. An electrical model of the wet sand with four moisture-dependent components is derived from these measurements ( Fig. 1). Moisture is determined from the components of the model with the help of a neural network that has been trained with a reference sand.

Ein bisher eingesetztes Verfahren bestimmt die Feuchtigkeit von Sand durch Messen des ohmschen Widerstandes zwischen den Elektroden. Zuvor wird in einer Eichmessung die Widerstands-Feuchte-Kennlinie für einen Referenzsand ermittelt. Die Widerstandmessung ergibt vor allem bei hohen Feuchten eine geringe Genauigkeit, weil sich die Widerstands- Feuchte-Kennlinie zu hohen Feuchten hin abflacht. Außerdem hängt die ermittelte Sandfeuchte stark von der Sandsorte ab. Deshalb muß für jeden Sand eine gesonderte Eichmessung zur Bestimmung der Kennlinien durchgeführt werden.A previously used method determines the moisture of sand by measuring the ohmic resistance between the electrodes. Before this is carried out in a calibration measurement Resistance-moisture characteristic curve determined for a reference sand. The resistance measurement results in low accuracy, especially at high humidity, because the resistance The moisture characteristic flattens out towards high humidity. The determined sand moisture also depends strongly depends on the type of sand. Therefore, a separate calibration measurement must be carried out for each sand Determination of the characteristic curves can be carried out.

Bekannt geworden sind auch Verfahren, bei denen die Kapazität zwischen den Elektroden gemessen wird. Die Kapazitätsmessung liefert bei großen Sandfeuchten bessere Ergebnisse als die Widerstandsmessung, die Abhängigkeit von der Sandsorte ist aber noch größer als bei der Widerstandsmessung.Methods are also known in which the capacitance between the electrodes is measured. The capacity measurement gives better results than with large sand moisture the resistance measurement, the dependence on the type of sand is even greater than that of the Resistance measurement.

Mit dem vorliegenden Verfahren wird die Genauigkeit der Feuchtebestimmung verbessert und eine weitgehende Unabhängigkeit der Feuchtebestimmung von der Sandsorte erreicht. Das elektrische Modell des Sandes ist bei diesem Verfahren nicht mehr ein Widerstand oder eine Kapazität, sondern ein Netzwerk mit vier feuchteabhängigen Bauelementen, den Widerständen R1 und R2 und den Kapazitäten C1 und C2 (Abb. 1). Zur Bestimmung der vier Bauelemente werden Real- und Imaginärteil des komplexen Widerstandes bei zwei verschiedenen Frequenzen, z. B. bei 1 KHz und bei 1 MHz, gemessen und anschließend aus den vier Meßwerten vier Bauelemente berechnet.With the present method, the accuracy of the moisture determination is improved and the moisture determination is largely independent of the type of sand. With this method, the electrical model of the sand is no longer a resistor or a capacitance, but a network with four moisture-dependent components, the resistors R1 and R2 and the capacitors C1 and C2 ( Fig. 1). To determine the four components, the real and imaginary part of the complex resistance at two different frequencies, e.g. B. at 1 KHz and at 1 MHz, measured and then calculated four components from the four measured values.

Zur Bestimmung der Sandfeuchte werden folgende Merkmale verwendet: R1, C1, R2, die Zeitkonstante R1_*C1 und das Verhältnis R1/R2. Die Merkmale wurden so gewählt, daßThe following characteristics are used to determine the sand moisture: R1, C1, R2, the time constant R1 _* C1 and the ratio R1 / R2. The characteristics were chosen so that

- At least one characteristic has sufficient sensitivity for all relevant humidity levels (Change in the characteristic value vs moisture)
- A moisture determination largely independent of the sand type is made possible.

Der Zusammenhang zwischen diesen fünf Merkmalen und der Sandfeuchte wird empirisch ermittelt. Hierzu werden an einem Referenzsand bei verschiedenen, bekannten Feuchten im Bereich von 1 bis 12 Gewichtsprozent die vier Bauelemente des elektrischen Modells des Sandes bzw. die oben genannten fünf Merkmale bestimmt. Mit dem Ergebnis dieser Eichmessung wird ein neuronales Netz trainiert. Die fünf Merkmale bilden hierbei die Eingangswerte des neuronalen Netzes. Ausgangswert des Netzes ist die Sandfeuchte.The relationship between these five characteristics and sand moisture is empirical determined. For this purpose, a reference sand at different, known humidities in the Range from 1 to 12 percent by weight of the four components of the electrical model of the Sandes or the above five characteristics determined. With the result of this A neural network is trained in calibration measurement. The five characteristics form the Input values of the neural network. The initial value of the network is the sand moisture.

Beim Einsatz des Verfahrens berechnet das trainierte, neuronale Netz die Sandfeuchte aus fünf Merkmalswerten, die an dem zu messenden Sand durch Messung des komplexen Widerstandes bei zwei Frequenzen bestimmt wurden.When using the method, the trained neural network calculates the sand moisture from five Characteristic values on the sand to be measured by measuring the complex resistance were determined at two frequencies.

Claims

1. A method for determining the moisture of sand, characterized in that the real and imaginary part of the complex resistance between two electrodes immersed in the sand to be measured are measured at two frequencies and four moisture-dependent components R1, R2, C1, C2 of an electrical model ( Fig. 1) of the sand are calculated, which are combined into five characteristics, namely R1, R2, C1, R1 _* C1, R1 / R2, and fed to a trained neural network as input variables, the neural network providing the sand moisture as an output variable.

2. The method according to claim 1, characterized in that the neural network with a Reference sand with known sand moisture levels, e.g. B. by drying and weighing the sand were determined, is being trained.

3. The method according to claim 1 or 2, characterized in that as a neural network Multilayer Perceptron with back-propagation learning algorithm that is used in the Input layer five neurons, one in the hidden layer and one in the output layer Contains neuron.

4. Arrangement for determining the moisture of sand ( Fig. 2) for performing the method according to one of claims 1 to 3, characterized in that in a measuring circuit ( 2 ) real and imaginary part of the complex resistance between two electrodes immersed in sand ( 1 ) at two frequencies, e.g. B. measured at 1 KHz and 1 MHz, and the measured four values are read into a computer ( 4 ) via analog / digital converter ( 3 ), where the sand moisture is calculated from the four measured values and via a display unit ( 5 ) is issued.

5. Arrangement according to claim 4, characterized in that in the computer from the measured at two frequencies real and imaginary parts of the complex resistance, the components R1, R2, C1, C2 of the sand model ( Fig. 1) and the product R1 _* C1 and that Ratio R1 / R2 can be calculated.

6. Arrangement according to claim 4 and 5, characterized in that a neural network is programmed in the computer, which calculates the sand moisture after a training phase from the features R1, R2, C1, R1 _* C1, R1 / R2.

7. Arrangement according to claim 4 to 6, characterized in that for training the neural network the sand moisture is determined as described, the Electrodes are located in a sand of known humidity, and if the measured deviation Moisture from the known real sand moisture the neural network after a baking Propagation algorithm is corrected and this training process at five to nine various sand moisture levels in the range of 1 to 12 weight percent moisture becomes.