RU2256166C1 - Method of determining oil content in water stream - Google Patents

Abstract

FIELD: water treatment.

SUBSTANCE: invention, which can be used for continuously controlling quality of water and to measure concentration of emulsions, resides in that, in the calibration step, one performs nephelometric measurements for several different scattering angles, e.g., 30, 60, 90, 120, 150, 180°, in a closed hydraulic circuit, while consecutively increasing oil content in circuit and, each time, dispersing oil in water over a period of time long enough to achieve maximally possible change in dispersity of the emulsion. Values of specified concentrations and current light scattering intensities for different angles are continuously recorded to be further used for educating artificial neuronet, on input neurons of which light scattering intensities for different angles are given and, on output neuron, searched concentration values are obtained. Nephelometric equipment with, contained therein, educated neuronet is used to continuously control content of oil in water stream.

EFFECT: increased accuracy and rapidity of measurements.

2 dwg

Description

The invention relates to the field of measuring technology and can be used in solving problems of continuous monitoring of oil or oil in water, environmental monitoring, measuring the concentration of emulsions.

There are various methods for measuring the concentration of emulsions. Many of them are specifically designed to determine the content of oil or oil in water. For example, this can be done by centrifuging the emulsion, followed by determining the number of separated phases by weighing or measuring the level of sediment [Belyakov V.L. Automation of oil and water field preparation. -M .: Nedra, 1988. - P.141]. This method requires a rather large investment of time and is only possible in laboratory conditions where it is possible to ensure thorough preparation of centrifuges before measurements. For conditions of in-line measurements, photometric methods of analysis are more applicable [Karpishchenko A.I. Medical laboratory technology. St. Petersburg: Intermedika, 1998. - P.64]. Among them, the most commonly used due to its high sensitivity, wide range of measurements and relative ease of implementation is the nephelometric method. This method is based on the transmission of the volume of liquid by probing radiation and registration of scattered radiation at a certain angle (usually 90 °) to the direction of the primary radiation [Belyakov V.L. Automation of oil and water field preparation. - M .: Nedra, 1988. - P.128]. The result of measuring the concentration by the nephelometric method (like any other photometric method) depends on the dispersion of the emulsion, since the radiation pattern of the scattered radiation varies greatly with the average size of the suspended droplets [Karpishchenko AI Medical laboratory technology. St. Petersburg: Intermedika, 1998. - P.77]. Almost scattered radiation is distributed approximately as shown in Fig. 1 for a low concentration oil-water emulsion at a wavelength corresponding to red radiation. External radiation R acts on the particle P, which scatters it in different directions. The directivity pattern of scattered radiation evolves from configuration 1 for the smallest globules of oil (with an average diameter of 0.1-1 μm) to configurations 2 for a medium-dispersed emulsion and 3 for coarsely dispersed ones (with a size of globes of the order of 100 microns). The influence of this factor can lead to the fact that during in-line measurements for the same concentration of suspended matter, the result can fluctuate within 10-40%. Therefore, the dispersion of the emulsion must either be stabilized or taken into account in some way during measurements.

Closest to the proposed method is a method for determining the oil content in water by mechanically dispersing a water sample in the presence of surface-active substances and subsequent indirect measurement of suspended particles by any photometric method [A.S. No. 317199 (USSR), IPC G 01 N 21/20. A method for the quantitative determination of petroleum products in water. Publ. 10/19/71, Byp. No. 31]. The essence of this method is to reduce the size of suspended droplets to a specific narrow range of values and stabilize the emulsion due to the action of surface-active substances. Dispersion is carried out using a gear pump, combining the functions of breaking the drops and mixing the emulsion. When implementing the method, the sample volume, temperature, pump drive rotation speed, surfactant flow rate, dispersion time and settling time must be fixed. The disadvantages of the method are:

- fuzzy fixation of the dispersion of the emulsion at a certain level, because it depends on a large number of hard to consider factors, and the associated error in determining the concentration;

- the relative complexity of the implementation of the method on the stream, due to the need for periodic sampling;

- a long analysis time (about 10 minutes), which makes it impossible to use in real-time systems;

- the reliability of the measurements during the implementation of the method on the stream, which is due to the possible unsatisfactory representativeness of the sample.

The problem solved in the invention is to increase the accuracy of determining the concentration of oil in the water stream and reduce the measurement time by excluding from the measurement cycle the stages of sampling and sample preparation and the use of neural network technology for processing photometric signals containing, in particular, information about the dispersion of the emulsion.

The problem is solved in that in the known method for determining the oil content in a water stream by mechanical dispersion of a water sample and subsequent indirect measurement of suspended particles by some photometric method, in contrast to the prototype, mechanical dispersion of samples is performed only at the stage of calibration of photometric equipment, during which in a closed hydraulic circuit make nephelometric measurements for several different scattering angles, for example, 30, 60, 90, 120, 150 °, 180 °, and at at this stage, the oil content in the circuit is successively increased, and for each new concentration value, dispersion is performed for a time sufficient for the maximum possible change in the dispersion of the emulsion, while the values of the given concentration and current light scattering intensities are continuously stored for various angles, which are then used to study artificial neural networks, on the input neurons of which they supply the values of light scattering intensities for different angles, and on the output neuron The value of the desired concentration is then determined, then the nephelometric equipment with an artificial neural network trained on it is used to continuously measure the oil content in the water stream.

Figure 1, discussed above, shows a change in the radiation patterns of the scattered radiation depending on the dispersion of the emulsion. Figure 2 schematically shows an example of a device with which the proposed method can be implemented.

An example of a specific implementation of the method.

The implementation of the method can be divided into 2 stages: graduation / training and the actual flow measurements.

At the graduation / training stage, a closed hydraulic circuit 1 is assembled, as shown in FIG. The circuit includes a dispenser 2 (containing a filler neck 3, a syringe inlet 4 and a drain neck 5), a gear pump 6 and a nephelometer 7 containing an emitter 8 and several photodetectors 9 located in a semicircle so that it is possible to measure the scattering intensities under different angles relative to the probe beam. The processing of the nephelometer signals is performed using the controller 10, which includes a microcomputer 11, a keyboard 12, an indicator 13, and an artificial neural network (ANN) 14.

The presentation of ANN 14 as a separate functional unit is conditional, because it can be implemented both hardware and software. The paradigm (type) of ANNs can be different. For example, it can be a multilayer perceptron with one hidden layer, as shown in figure 2 [Kruglov V.V., Borisov V.V. Artificial neural networks. Theory and practice. - M .: Hot line - Telecom, 2001. - P.53-58].

Graduation (training) is carried out as follows.

Through the filler neck 3 of the dispenser, circuit 1 is completely washed down with water, so that there are no air accumulations. The volume of the circuit is known and constant. They include a gear pump 6, the functions of which include pumping water along the circuit and dispersing the emulsion. The controller 10 is transferred to the calibration mode from the keyboard 12.

Nephelometric measurements are performed for several different scattering angles. To do this, by means of the microcomputer 11, the emitter 8 is turned on and the signals of the photodetectors 9 stored in the memory of the microcomputer 11 are automatically registered. In this example, the stored signals correspond to scattering angles of 30, 60, 90, 120, 150, 180 °. From the keyboard 12 enter the value of the true concentration (currently equal to zero), also displayed on the indicator 13, which is also stored.

Then, through the syringe input 4 of the dispenser 2, a small, defined value of the oil (s) of the kind with which it is supposed to work on the stream is introduced into the circuit. When this gear pump 6 continues to work, performing mixing and dispersion. From the keyboard 12 enter the next value of the current concentration and the start of the countdown (for example, 10 minutes). During this time interval, the dispersion of the emulsion will change, passing the entire range of its possible values (from large droplets of the order of 100 microns to small ones of the order of 0.1-1 microns). Moreover, it is essential that the dispersion (average particle size) is not directly measured; it is enough to know that during mixing the dispersion varies in the entire possible range of its values. In this case, the microcomputer 11 periodically, at equal intervals of time, initiates the reading of the values of the signals of the photodetectors 9. The arrays of values accumulated in this way will correspond to the same concentration but different emulsion states (dispersion values). The described procedure for introducing a new portion of oil (oil) is repeated for other calibration concentration values. Moreover, if the subsequent concentration values increase sequentially and are quite different from each other (for example, 0, 10, 20, 40, 80, 160 mg / l), then you can not change the fluid in the circuit before each new concentration value. If the calibration values lie quite close to each other and greater accuracy is required, then before setting each new concentration value, it is necessary to drain the previous sample through the drain neck with tap 5, flush the circuit and fill the circuit with clean water.

After graduation, the accumulated data arrays are stored in the memory of the microcomputer. An example of the structure of the stored data is given in table. 1.

Then, ANN training is performed on the accumulated data. This is done by means of the ANN training subroutine launched from the keyboard 12 and executed by the microcomputer 11. In this case, the signals of the corresponding photodetectors (FP1-FP6) are fed sequentially and repeatedly to the input neurons of the ANN from the memory of the microcomputer 11, and the training of the ANN is reduced to such adjustment of the weights of interneuron communications, at which the output neuron produces the correct values corresponding to the concentrations set during calibration.

For high-quality training of ANNs, it is necessary that the number of observations (the number of rows in Table 1) be as large as possible. To do this, increase the sampling frequency of the photodetector signals performed during mixing. The specific method for training ANNs may be different. For example, the so-called back propagation method of error can be used [Kruglov VV, Borisov VV Artificial neural networks. Theory and practice. - M .: Hot line - Telecom, 2001. - P.26-27, 43].

The parameters of the trained ANN (weight of interneuronal connections) are stored in the memory of microcomputer 11. At this point, the graduation / training process is completed.

The nephelometer 7 is removed from circuit 1 and, together with the controller 10, is placed on the test object (i.e., inserted into the pipeline rupture with the test liquid). After that, a measurement subroutine is launched from the keyboard 12, according to which the presentation of the current values of the signals of the photodetectors 9 to the inputs of the ANN 14 leads to an almost instantaneous response to the output of the ANN, which is displayed on the indicator 13 as the current emulsion concentration (oil content in water).

The described nephelometer 7 is presented here as simplified as possible. This can be either a serial nephelometer (for example, from the NASH model range 2100 (USA), or specially developed. Actually, it can have some improvements that eliminate the effects of emitter instability and contamination of photodetector windows. But these improvements do not change the essence of the proposed method.

The controller 10 can be implemented on the basis of a single-chip microcomputer (for example, РlС16F876 from MicroChip) or an embedded single-board computer. The author used a model 6052V from ICOR (Taiwan) with built-in flash memory. ANN can be created and saved as a file in the environment of the STATISTICA Neural Networks software package (product of StatSoft), and then transferred to the microcomputer memory.

Thus, with the help of ANNs and a nephelometer, measurements of the emulsion concentration are invariant, which are invariant with respect to the dispersion of the emulsion. ANN is here a convenient means of approximating the multidimensional concentration function of light scattering intensities at various angles. The same approximation can, in principle, be obtained by describing the indicated dependence by a function of some kind (for example, a power polynomial) and determining the coefficients using the least squares method. However, this approximation method is much less convenient, because the type of the approximated function is not known in advance. For a neural network approximation technology, the form of the function does not matter.

The proposed method in comparison with the prototype method has

- significantly reduced measurement time due to the exclusion from the measurement cycle of the stages of sampling and sample preparation and the use of a trained neural network for continuous processing of photometric signals;

- increased accuracy in determining the concentration of oil in the water stream due to the improved representativeness of the sample and the use of information about the dispersion of the emulsion;

- the relative simplicity of the implementation of measurements on the stream, because no preliminary preparation of the measurement object is required, it is not required to maintain the constancy of technological parameters, such as temperature, pressure, flow, ANN monitors all possible changes in dispersion, which can lead to a change in these parameters.

Claims (1)

A method for determining the oil content in a water stream by mechanical dispersion of water samples and subsequent indirect measurement of suspended particles by some photometric method, characterized in that the mechanical dispersion of samples is performed only at the stage of graduation of photometric equipment, during which nephelometric measurements are performed in a closed hydraulic circuit for several different scattering angles, for example 30, 60 90, 120, 150, 180 °, and at this stage, sequentially increase with the oil is kept in the circuit and for each new concentration value they are dispersed for a time sufficient for the maximum possible change in the dispersion of the emulsion, while the values of the given concentration and current light scattering intensities for various angles, which are then used to train the artificial neural network, are input to the input neurons which serves the values of light scattering intensities for different angles, and on the output neuron get the value of the desired concentration, then not elometricheskoe equipment with a trained artificial neural network it is used for continuous measurement of the oil content in the water stream.

Images