V. G. Valeev and A. F. Avdeeva

Источник: http://www.who.int/hinari/en

This device consists of a capacitance transducer and a circuit with parametric modulation; the transducer capacitance is a complicated function of several factors. For instance, the dielectric constant is affected by the grain-size distribution, particle density, contact resistance between particles, rock concentration, chemical composition, and so on. Also, the structure of the material is indirectly affected by the method of consolidation, on account of inelastic strain in the dispersion, together with any effects from mechanical processing of the material before measurement (mode of grinding), and so on. Many factors are difficult to estimate, which results in additional error in water-content measurement. The error may be reduced by determining the effects of the perturbing factors.

The major monitored parameters in simulating the device were the water content W, temperature t, the compressibility h of the material at some specified pressure, and the conductance of tile device at 50 Hz.

The latter parameter indicates indirectly the contact resistance between the coal particles. The relevant functions are the capacitance C and conductance g of the device at 1.5 MHz.

The functions and parameters were varied during the experiment within the following limits:

W: 3.6-13%; t: 4-30"C; C: 40-80 pF; h: 9-28 mm; g: 0.3-6.5 mmho; 050:0.7-5.6 mmho.

We calculated 19 correction curves, which had r experimental points in each. In the first stage of approximation we took the functions in general form up to the quadratic and mixed terms, which enable us to determine the predominant factors, and also to estimate the nonlinearity and convert to a simplified form in which only the major terms were incorporated.

Stochastic approximation allowed us to improve the model from measurement data. The coefficients in the approximating equation were corrected in accordance with the formula of [1]:

where ,  are the values of coefficient i at steps N and N+1 in the correction, with ,  the value of the function at step N+1, and   the value of monitored parameter i at step N+1.
In this way we determined the coefficients in the simplified model via stochastic approximation, which gave the following mathematical description for the device:

C=54.8+6.7W+3.6h+1.1t +9.1+1.8+0.6+2.3W;

g = 3.1+1.23W+0.94h+0.5t+1.58+0.01-0.26.

On repeating the refinement and eliminating terms with small coefficients, we obtained

C=54+6.8W+4.5h+1.24t+8.8.

Fig. 1

Similarly, one can obtain a simplified expression for g. The two equations may be solved together to eliminate either parameter that is difficult to measure.

One can obtain a simple scheme for compensating the perturbing factors on the basis that C and g as corrupted by these factors are linearly related; then apart from constant component we have

,                                                                (1)

where ,  are constants.

One can judge the conductance from the amplitude of the high-frequency voltage on the tunnel circuit (1), may be derived by correlation means during calibration under production conditions.

Figure 1 shows the system with perturbing-parameter compensation PP; the device uses parametric modulation. If   is obeyed, the system is at balance. If the capacitance of the transducer undergoes some change , the high-frequency output voltage is amplitude modulated, and after the high-pass filter HPF it is detected and passes to the low-pass filter LPF, whereafter it is amplified and drives the asynchronous reversible motor M, which adjusts the capacitor  and brings the circuit to balance.

From  we get the capacitance of the transducer and hence the water content of the material.

The electrical losses in the transducer determine the amplitude of the high-frequency voltage at the tuned circuit; the signal after detection passes to an indicator, which shows the magnitude of the correction to the meter reading. The compensating circuit is adjusted by means of resistor R1.

The device has been tested at the dressing plant of Pit No. 25 of the Vorkuta Coal Plant. Samples for measurement and calibration were taken from the output of the washing plant. Tests show that the instruments had good sensitivity in the water-content range 6-12%. The basic error of measurement was ±0.5%.

Individual preliminary calibrations are required for coals from other deposits.

Literature

1. V.L. Myakishev and V. Z. Kozin, VUZ Gorn. Zh. No.2 (1971).