A rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag

Abstract

The present invention discloses a method for rapidly determining the glass content of multi-component organic hazardous waste water-quenched slag, which matches the high temperature and gasification melting technology of multi-component organic hazardous waste and realizes the rapid determination of the glass content of water-quenched slag. The method determines the water-quenched slag particle size, color illumination L value, hardness data, moisture content, fine slag content and other indicators, combines and utilizes the BP neural network, uses the levenberg-marquardt algorithm to train the BP neural network, selects the neural network with the smallest result error among the neural networks with different hidden layer node numbers as the target neural network, and finally quickly obtains the glass content of the water-quenched slag.

Description

Technical Field

[0001] The invention relates to a method for quickly detecting the glass content of multi-component organic hazardous waste gasification water-quenched slag, and belongs to the field of hazardous waste resource utilization. Background Art

[0002] The high-temperature melting and gasification technology of multi-component organic hazardous waste can realize the gasification utilization of organic matter and the vitrification fixation of inorganic matter in organic hazardous waste, which is a technology for the resource utilization of solid waste. Since organic hazardous waste contains a certain amount of heavy metal harmful substances, the degree of vitrification of water-quenched slag is one of the important indicators for determining the solidification and harmlessness of heavy metals. The existing method for determining the vitreous content of water-quenched slag (GB / T18046-2017) is an X-ray diffraction technology. The XRD spectrum is used to calculate the proportion of the bulge peak area within the scanning angle range of 22°-38°, and finally determine the vitreous content of the water-quenched slag. This method requires expensive measuring instruments and delicate operations. It requires the measuring personnel to have a high level of technical skills and is not suitable for the requirements of fast, convenient, and large-scale data measurement in industrial production processes. Therefore, a fast, simple, and low-cost measurement method is needed.

[0003] The macroscopic structures of water-quenched slag with different glass contents are quite different. The particle size distribution, color illumination L value, particle hardness, moisture content, and fine slag content of water-quenched slag with different glass contents are quite different. The higher the glass content, the more uniform the particle size of the water-quenched slag, the higher the brightness of the color of the water-quenched slag, the greater the hardness of the water-quenched slag particles, the lower the moisture content, and the lower the fine slag content. By analyzing the macroscopic object characteristics of the water-quenched slag particles, a correlation model between the physical properties of the water-quenched slag and the glass content of the water-quenched slag can be established. The correlation model is optimized by combining the existing neural network method, and finally the glass content of the water-quenched slag is obtained. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a rapid detection method for the glass content of multi-component organic hazardous waste gasification water-quenched slag. Water-quenched slag is formed by quenching molten inorganic matter in a gasifier with quenching water. Water-quenched slag with different glass content has different physical properties. The formation of water-quenched slag with high glass content requires that the molten inorganic matter in the gasifier is completely melted and cooled by rapid flushing with quenching water. The melting condition of inorganic matter in the gasifier and the flushing amount of quenching water both affect the glass content of the water-quenched slag. If the molten inorganic matter is not completely melted, the viscosity of the melt increases. During the water quenching process, the melt cannot be completely brittle, resulting in different particle size distributions of the water-quenched slag and the appearance of large pieces of water-quenched slag particles. Glass is a solid in a stable state. The higher the glass content of the water-quenched slag, the denser the surface of the water-quenched slag particles, so that the color brightness and hardness of the water-quenched slag are improved, and at the same time, the water content and fine slag content of the water-quenched slag are reduced. For water-quenched slag with low glass content, the crystal material is wrapped in the glass structure during the water quenching process, which reduces the structural stability of the water-quenched slag and makes crystal particles exist on the surface of the water-quenched slag, thereby reducing the color brightness of the water-quenched slag and increasing the water content and fine slag content of the water-quenched slag. After comprehensive analysis, the particle size of the water-quenched slag is selected as a characteristic evaluation index reflecting the uniform brittle fracture of the molten state; the Vickers hardness is selected as a characteristic evaluation index reflecting the density of the glassy state of the water-quenched slag; the color illumination L value is selected as a characteristic evaluation index reflecting the reflective surface of the glassy state; the water content and fine slag content are selected as characteristic evaluation indexes reflecting the smoothness of the glassy particles.

[0005] The invention is based on the study of the formation process of water-quenched slag, takes the particle size, Vickers hardness, color Lab value, moisture content and fine slag content of the water-quenched slag as detection indicators, optimizes and calculates it using a neural network model, and finally obtains the glass content of the water-quenched slag.

[0006] The specific method of the present invention is as follows: A rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag, comprising the following steps:

[0007] 1) Take a water-quenched slag sample produced in a normal production process, analyze and measure the glass content of the water-quenched slag, and determine the glass content of the water-quenched slag;

[0008] 2) measuring or analyzing the water-quenched slag particle size, color illumination L value, Vickers hardness, moisture content, and fine slag content of the water-quenched slag sample in step 1) to obtain characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample;

[0009] Among them, the particle size of the water-quenched slag is determined by screening with a standard sieve, and the particle size of the water-quenched slag is divided into 6 levels: <0.125mm, 0.125mm~0.3mm, 0.3mm~1mm, 1mm~2mm, 2mm~4.75mm, 4.75mm<; after the water-quenched slag is screened with a standard sieve, the mass of the water-quenched slag of each particle size is normalized to determine the mass percentage of the water-quenched slag of each particle size; the mass percentage of the water-quenched slag of 1mm~2mm is taken as the characteristic particle size percentage data;

[0010] The color illumination L value of the water-quenched slag is analyzed by a color detector; the color value of the dried water-quenched slag is measured after being ground and pre-treated; the L value in the measured data is selected as a characteristic variable;

[0011] The fine residue content of the water-quenched slag is determined by washing and filtering the wet-based water-quenched slag with a 250-mesh sieve. The washing water is filtered through a Buchner funnel to obtain a filter cake. The filter cake and the washed water-quenched slag are dried to a constant weight to obtain the mass percentage of the fine residue of the water-quenched slag.

[0012] 3) Taking the characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample as characteristic variables and the corresponding vitreous content as the target variable, the training samples were constructed, and the abnormal samples caused by measurement errors and abnormal working conditions were removed. Then the data were normalized to remove the influence of the order of magnitude of the data itself;

[0013] 4) constructing a BP neural network, training the BP neural network, and obtaining a trained BP neural network, which is a water-quenched slag vitreous content detection model;

[0014] 5) Obtain characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample to be tested, and use the water-quenched slag vitreous content detection model obtained in step 4) to detect and analyze the sample to be tested to obtain the water-quenched slag vitreous content.

[0015] As a preferred embodiment of the present invention, the glass content in the water-quenched slag sample is determined by the method in Appendix C of "Granulated Blast Furnace Slag for Cement, Mortar and Concrete (GB / T18046-2017)".

[0016] As a preferred embodiment of the present invention, the Vickers hardness of the water-quenched slag is measured using water-quenched slag with a thickness greater than 2 mm, and the water-quenched slag is treated according to the Vickers measurement requirements; the composite used in the Vickers hardness measurement is 5 kg, the relative angle of the indenter surface is 136°, and the hardness of the water-quenched slag is measured according to the Vickers hardness measurement method.

[0017] As a preferred embodiment of the present invention, the moisture content analysis method is: using an oven at 105° C. to dry the water-quenched slag to a constant weight, and determining the moisture content of the water-quenched slag.

[0018] As a preferred embodiment of the present invention, the fine slag content analysis method is as follows: the wet-based water-quenched slag is placed in a 250-mesh screen, the water-quenched slag is rinsed with deionized water, and the washing water after rinsing is filtered through a Buchner funnel to obtain a filter cake. The filter cake and the washed water-quenched slag are dried to a constant weight, and the mass percentage of the filter cake in the water-quenched slag fine slag is calculated.

[0019] As a preferred embodiment of the present invention, during the color illumination L value analysis process, the particle size of the ground water-quenched slag should be less than 0.3 mm, preferably 0.1 mm to 0.3 mm.

[0020] In the analysis and determination of color illuminance L value, the instrument light source type should be D65 light source, spectral range: 360nm-750nm, wavelength interval 10nm; the same sample needs to be measured more than 3 times, and the average value of the measurement is taken as the characteristic color illuminance L value.

[0021] As a preferred embodiment of the present invention, in step 3), the method for removing abnormal samples caused by measurement errors and abnormal operating conditions is the box plot method.

[0022] As a preferred embodiment of the present invention, in step 4), the established neural network model should ensure stable working conditions. If the temperature fluctuation during the working condition exceeds 100°C, the model should be re-established.

[0023] The levenberg-marquardt algorithm is used to train the BP neural network, find the parameter variable with the smallest function value, and optimize the neural network.

[0024] As a preferred embodiment of the present invention, in step 5), during the model prediction process, characteristic data of the same batch of samples are collected at least 3 times, and vitreous content prediction is performed at least 3 times to ensure the accuracy and stability of the prediction results.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The present invention aims at the changes in the characteristics of water-quenched slag during the formation of multi-component organic hazardous waste water-quenched slag, and proposes a simple and rapid determination method for the glass content of water-quenched slag. During the use of this method, the requirements for a simple, rapid and low-cost determination process in the industrial production process are met. The present invention uses a neural network model as a calculation means to reduce the workload of simulation analysis of the calculation formula while improving the accuracy of the glass content of the water-quenched slag. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a comparison chart of the predicted value and the true value of the method of the present invention.

[0028] Figure 2 This is a scatter plot of the error of each data. DETAILED DESCRIPTION

[0029] The present invention is further described and illustrated below in conjunction with specific embodiments. The technical features of each embodiment of the present invention can be combined accordingly without conflicting with each other.

[0030] The process of the method of the present invention is:

[0031] The water-quenched slag samples produced during the normal production process of the plant were taken, and the glass content was analyzed and determined to determine the glass content of the water-quenched slag. This example uses the method in Appendix C of "Granulated Blast Furnace Slag for Cement, Mortar and Concrete (GB / T18046-2017)" to determine the glass content in the water-quenched slag samples.

[0032] The particle size of the water-quenched slag is determined by screening with a standard sieve, and the particle size of the water-quenched slag is divided into 6 levels: <0.125mm, 0.125mm~0.3mm, 0.3mm~1mm, 1mm~2mm, 2mm~4.75mm, 4.75mm<. After the water-quenched slag is screened with a standard sieve, the mass of the water-quenched slag of each particle size is normalized to determine the mass percentage of the water-quenched slag of each particle size. The mass percentage of the water-quenched slag of 1mm~2mm is the characteristic particle size percentage data, which is used as one of the input characteristic variables of the neural network.

[0033] The color illuminance L value of the water-quenched slag is analyzed using a color detector. After the dry water-quenched slag is ground and pre-treated, its color value is measured. The illuminance L value in the measured data is the characteristic particle size percentage data, which serves as one of the input characteristic variables of the neural network. In the color illuminance L value analysis and measurement of this embodiment, the water-quenched slag is ground, and the particle size after grinding should be controlled at 0.1mm~0.3mm; the light source type of the color illuminance L value measurement instrument should use D65 light source, spectral range: 360nm-750nm, wavelength interval 10nm; it is necessary to measure the same sample more than 3 times, and take the measured average value as the characteristic color illuminance L value.

[0034] The Vickers hardness of the water-quenched slag is measured using water-quenched slag larger than 2mm, and the water-quenched slag is processed according to the Vickers measurement requirements. The composite used in the Vickers hardness measurement process is 5kg, and the angle between the relative faces of the indenter is 136°. The hardness of the water-quenched slag is measured according to the Vickers hardness measurement method. The Vickers hardness (HV) value of the water-quenched slag is one of the input characteristic variables of the neural network. The water content of the water-quenched slag is dried in an oven at 105°C to a constant weight, and the water content of the water-quenched slag is determined as one of the input characteristic variables of the neural network.

[0035] The water-quenched slag is placed in a 250-mesh screen and rinsed with deionized water. The washing water after rinsing is filtered through a Buchner funnel to obtain a filter cake. The filter cake and the washed water-quenched slag are dried to a constant weight, and the mass percentage of the filter cake in the water-quenched slag fine slag is calculated. This data is one of the input feature variables of the neural network.

[0036] The vitreous content of water-quenched slag and the above five characteristic variables of water-quenched slag are the basic data for neural network model training. The collected characteristic variables and their corresponding target variable data (vitreous content) are preprocessed, and the box plot method is used to remove abnormal data points caused by measurement errors and abnormal working conditions, and then the data is normalized to remove the influence of the order of magnitude of the data itself;

[0037] Establish a BP neural network model for measuring the temperature in the gasifier;

[0038] The established neural network model is trained with the processed data (i.e. the processed feature variable data and its corresponding target variable data). The training adopts the Levenberg-Marquardt algorithm to find the parameter variable with the smallest function value and optimize the neural network to determine the optimal number of hidden layer nodes and network weights of the BP neural network model, and obtain a model that can accurately measure the vitreous content of water-quenched slag; the vitreous content of water-quenched slag is finally determined using the neural network.

[0039] The present invention is further described below in conjunction with a specific embodiment: The present invention is used to quickly determine and analyze the glass content of water-quenched slag samples from a factory.

[0040] First, 50 groups of water-quenched slag samples with known vitreous content were selected from the factory to measure the test data of the water-quenched slag. The test data operation process of the water-quenched slag is as follows:

[0041] 1. Weigh 300g of the original water-quenched slag sample and put it into an oven at 105℃ to dry it to constant weight. After drying, determine the mass of the water-quenched slag and calculate the moisture content of the water-quenched slag.

[0042] 2. Put the wet-based water-quenched slag into a 250-mesh screen, rinse the water-quenched slag with deionized water, and filter the washing water through a Buchner funnel to obtain a filter cake. The filter cake and the washed water-quenched slag are dried to constant weight, and the mass percentage of the filter cake in the water-quenched slag fine slag is calculated.

[0043] 3. Take 100g of dried water-quenched slag and sieve it with a standard sieve. The particle size of the water-quenched slag is divided into 6 levels: <0.125mm, 0.125mm~0.3mm, 0.3mm~1mm, 1mm~2mm, 2mm~4.75mm, 4.75mm<. After the water-quenched slag is sieved with a standard sieve, the mass of the water-quenched slag of each particle size is normalized to determine the mass percentage of the water-quenched slag of each particle size. Take the mass percentage data of 1mm~2mm.

[0044] 4. Take 10g of water-quenched slag sample and grind it in an agate mortar. The particle size of the water-quenched slag after grinding should be less than 0.3mm. Take the ground powder to measure the color illumination L value.

[0045] 5. Take the water-quenched slag larger than 2mm for Vickers hardness test. The composite used in the Vickers hardness test is 5kg, the relative angle of the indenter is 136°, and the hardness of the water-quenched slag is measured according to the Vickers hardness test method.

[0046] All sample data obtained after the measurement are divided into a training set and a validation set; the division ratio can be 7:3 or 8:2. Each sample data in the training set and the validation set includes the characteristic particle size percentage data, color Lab value characteristic variable, Vickers hardness data, water content, ignition rate data and vitreous content of the sample.

[0047] The neural network training was carried out. The BP neural network model was selected as the neural network. The neural network input variables were 5 groups (i.e., characteristic particle size percentage data, color Lab value characteristic variables, Vickers hardness data, water content, and ignition loss data), and the neural network output variables were 1 group, i.e., vitreous content. The neural network training was carried out through the training samples of the training set to determine the optimal number of hidden layer nodes and network weights of the BP neural network model, and to obtain a model that can accurately measure the vitreous content of water-quenched slag; the vitreous content of the water-quenched slag to be tested can be determined using the trained neural network.

[0048] The method of the present invention was verified by using a verification set. The average relative percentage error of the simulation results was calculated to be 1.457%. Figure 1 As shown, the scatter plots of the data errors are as follows Figure 2 As shown; through simulation training of the final data, a neural network model was obtained, and finally a rapid determination method for vitreous content was established.

[0049] The selection of five characteristic variables, namely, characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of water-quenched slag samples, has the following advantages and necessity.

[0050] 1. The measurement process of the five characteristic variables is simple and fast. The characteristic particle size percentage of water-quenched slag, color illumination L value, Vickers hardness, moisture content, and fine slag content are all relatively simple measurement data in the daily production process. The moisture content and fine slag content of the water-quenched slag can be obtained by simply drying and constant weighting the water-quenched slag. After that, the particle size percentage, color illumination L value, and Vickers hardness of the water-quenched slag can be obtained by simple screening and measurement of the water-quenched slag. The entire analysis and measurement process does not use large test instruments or consume long measurement time.

[0051] 2. The five characteristic variables are correlated with the vitreous content of water-quenched slag. The vitreous content of water-quenched slag is an important data to measure the vitrification of water-quenched slag. The higher the vitreous content of water-quenched slag, the higher the vitrification of water-quenched slag, the smoother the surface of water-quenched slag, the harder the texture, and the easier it is for the water-quenched slag to crack during the water quenching and washing process. The smoother the surface of the water-quenched slag, the lower the fine ash particles on the surface, the worse the water absorption of the water-quenched slag itself, and the higher the reflected light intensity. The five characteristic variables accurately correspond to the characteristics presented by the vitreous content of water-quenched slag, and have a very strong correlation /

[0052] 3. The selection of five characteristic variables reduces the computational difficulty while taking into account the accuracy of the neural network model. The more characteristic variables are input, the larger the amount of data required in the neural network training process, and the better the data simulation results. While minimizing the amount of data collected in the early stage, the present invention selects five characteristic variables as characteristic variables to ensure the accuracy of its simulation data results.

[0053] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag, characterized in that: The steps include: 1) Take a water-quenched slag sample produced in the high-temperature gasification production process of organic hazardous waste, analyze and determine the glass content of the water-quenched slag, and determine the glass content of the water-quenched slag; in step 1), the glass content in the water-quenched slag sample is determined by the method in Appendix C of "Granulated Blast Furnace Slag for Cement, Mortar and Concrete (GB / T18046-2017)"; 2) measuring or analyzing the water-quenched slag particle size, color illumination L value, Vickers hardness, moisture content, and fine slag content of the water-quenched slag sample in step 1) to obtain characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample; Among them, the particle size of the water-quenched slag is determined by screening with a standard sieve, and the particle size of the water-quenched slag is divided into 6 levels: <0.125mm, 0.125mm~0.3mm, 0.3mm~1mm, 1mm~2mm, 2mm~4.75mm, 4.75mm<; After the water-quenched slag is screened with a standard sieve, the mass of the water-quenched slag of each particle size is normalized to determine the mass percentage of the water-quenched slag of each particle size; the mass percentage of the water-quenched slag of 1mm~2mm is taken as the characteristic particle size percentage data; The color illumination L value of the water-quenched slag is analyzed by a color detector; the color value of the dried water-quenched slag is measured after being ground and pre-treated; the L value in the measured data is selected as a characteristic variable; The Vickers hardness of the water-quenched slag is measured by using water-quenched slag with a thickness greater than 2 mm, and the water-quenched slag is treated according to the Vickers measurement requirements; the load used in the Vickers hardness measurement process is 5 kg, the relative angle of the indenter is 136°, and the hardness of the water-quenched slag is measured according to the Vickers hardness measurement method; 3) The characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample are used as characteristic variables, and the corresponding vitreous content is used as the target variable to construct training samples, remove abnormal samples caused by measurement errors and abnormal working conditions, and then normalize the data to remove the influence of the order of magnitude of the data itself; 4) constructing a BP neural network, training the BP neural network, and obtaining a trained BP neural network, which is a water-quenched slag vitreous content detection model; 5) Obtain characteristic particle size percentage data, color illumination L value, Vickers hardness data, moisture content, and fine slag content data of the water-quenched slag sample to be tested, and use the water-quenched slag vitreous content detection model obtained in step 4) to detect and analyze the sample to be tested to obtain the water-quenched slag vitreous content.

2. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that: The moisture content analysis method is: use an oven at 105°C to dry the water-quenched slag to constant weight, and determine the moisture content of the water-quenched slag.

3. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that: The analysis method of fine slag content is as follows: the wet-based water-quenched slag is placed in a 250-mesh sieve, the water-quenched slag is rinsed with deionized water, and the washing water after rinsing is filtered through a Buchner funnel to obtain a filter cake; the filter cake and the washed water-quenched slag are dried to a constant weight, and the mass percentage of the filter cake in the water-quenched slag fine slag is calculated.

4. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that: During the color illumination L value analysis process, the water-quenched slag needs to be ground, and the particle size after grinding should be controlled within 0.1mm~0.3mm.

5. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 4 is characterized in that In the analysis and determination of color illuminance L value, the instrument light source type should be D65 light source, spectral range: 360nm-750nm, wavelength interval 10nm; the same sample needs to be measured more than 3 times, and the average value of the measurement is taken as the characteristic color illuminance L value.

6. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that: In step 3), the method for removing abnormal samples caused by measurement errors and abnormal operating conditions is the box plot method.

7. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that In step 4), the established neural network model should ensure stable working conditions. If the temperature fluctuation during the working condition exceeds 100°C, the model should be re-established. The levenberg-marquardt algorithm is used to train the BP neural network, find the parameter variable with the smallest function value, and optimize the neural network.

8. The rapid detection method for the glass content of multi-component organic hazardous waste gasification water quenching slag according to claim 1 is characterized in that Step 5) During the model prediction process, feature data of the same batch of samples are collected at least 3 times, and vitreous content prediction is performed at least 3 times to ensure the accuracy and stability of the prediction results.

Images