Online coal ash content detection system and detection method based on X fluorescence analysis

Abstract

The invention discloses an on-line coal ash content detection system and method based on X fluorescence analysis. The system comprises a crushing module; a drying module; a tabletting module; a conveying module; the coal ash content analysis system is used for carrying out X fluorescence spectrum collection on the coal test sample at the detection position and analyzing to obtain the ash content value of the coal test sample; and a main control module. Automatic operation can be achieved in the whole detection process, lossless and rapid detection is achieved at the same time, and compared with a traditional scheme that a high-temperature firing method is adopted for ash content detection, the method has the advantages that time is greatly shortened, efficiency is improved, accuracy is high, errors caused by manual operation are reduced, and good application prospects are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal ash detection, and particularly relates to an on-line coal ash detection system and a detection method based on X-ray fluorescence analysis. Background Art

[0002] In the coal washing process, the ash content of coal has a crucial impact on coal quality and subsequent applications. Traditional flotation ash detection mainly relies on laboratory off-line analysis methods. For example, the commonly used coal ash determinator adopts the high-temperature burning method, that is, the weight difference before and after burning of the sample is measured and calculated respectively, and then divided by the weight of the sample before burning, and the obtained value is the ash content of the sample. This method has obvious drawbacks. Its detection process is cumbersome and time-consuming, and it cannot meet the real-time requirement of ash data in the production process. Moreover, there are many manual operation links in the traditional off-line detection process, and it is easy to introduce errors due to human factors, which is not conducive to accurate quality control and optimization of production. With the development of the coal industry towards automation and intelligence, there is an urgent need for a system that can detect coal ash on-line, quickly and accurately to achieve the efficient and stable operation of the coal washing process, improve the utilization efficiency of coal resources and economic benefits.

[0003] X-ray fluorescence spectrometry (XRF) has been widely used in the analysis of coal ash. Due to its advantages such as fast analysis speed, wide measurement range, and low anti-interference ability, it can analyze most elements and samples such as solids, powders, fused beads, and liquids. For coal, many studies have proven that X-ray fluorescence spectrometry can be used to analyze the ash in coal. Mujuru et al. determined the ash content by measuring the secondary radiation of coal in three different X-ray energy ranges (MUJURU M, MCCRINDLE RI, BOTHABM, et al. Multi-element determinations of N,N-dimethylformamide (DMF) coal slurries using ICP-OES[J]. Fuel, 2009, 88(4):719-724.). Kelloway calculated the main element oxides related to total ash and relative density through XRF technology (KELLOWAY S J, WARD C R, MARJO C E, et al. Calibration for ED-XRF profiling of coal cores for the Itrax Core Scanner[J]. Powder Diffraction, 2014, 29(S28-S34.)). Jia et al. established a distance correction method through iteration based on the relationship between XRF intensity and distance (JIAWB, ZHANGY, GU C G, et al. A new distance correction method for sulfur analysis in coal using online XRF measurement system[J]. Science China-Technological Sciences, 2014, 57(1):39-43.).

[0004] With the improvement of computing power, novel optimization methods, and large labeled datasets, data-driven machine learning and deep learning methods can obtain more elemental characteristic variables, providing another effective technical route for ash prediction. However, the elements in coal are diverse and complex, so the dimension of the dataset is also very high. Direct fitting may make it difficult to select and accurately quantify the influence of elemental content on ash content values, and it is difficult to formulate targeted and effective countermeasures.

[0005] Patent CN108489912B provides a method for analyzing coal components based on coal spectral data, which uses spectral data to analyze coal components and can be used for ash analysis. However, the principle of its ash analysis is that all components need to be analyzed, which will lead to a large computational amount of the model; and it lacks a verification function for the reliability of the collected samples or spectral data. When sampling problems or spectral collection problems occur, they cannot be detected in time, which easily leads to a large difference between the prediction result and the actual situation.

[0006] Therefore, it is necessary to improve the existing technology to provide a more reliable solution. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide an on-line coal ash detection system and detection method based on X-ray fluorescence analysis in view of the above deficiencies in the existing technology.

[0008] To solve the above technical problem, the technical solution adopted by the present invention is: in the first aspect of the present invention, an on-line coal ash detection system based on X-ray fluorescence analysis is provided, including:

[0009] A crushing module for crushing the coal to be detected;

[0010] A drying module for drying the coal to be detected;

[0011] A tablet pressing module for pressing the crushed and dried coal into a coal test sample;

[0012] A conveying module for conveying the coal test sample to the detection position;

[0013] A coal ash analysis system for collecting X-ray fluorescence spectra of the coal test sample at the detection position and analyzing to obtain the ash value of the coal test sample;

[0014] And a main control module for controlling the crushing module, the drying module, the tablet pressing module, the conveying module and the coal ash analysis system;

[0015] Wherein, the coal ash analysis system includes an X-ray fluorescence spectrum collection module and a coal ash analysis module. The coal ash analysis module includes a spectrum screening sub-module and an ash prediction model. The X-ray fluorescence spectrum collection module collects X-ray fluorescence spectrum data of the coal test sample. The spectrum screening sub-model screens out characteristic spectrum data from the collected X-ray fluorescence spectrum data, and the ash prediction sub-model analyzes to obtain the ash value of the coal test sample according to the characteristic spectrum data.

[0016] Preferably, the drying module includes an air pump, a heater connected to the air pump, a drying cylinder connected to the heater through an intake pipeline, and an intake valve provided on the intake pipeline.

[0017] Preferably, a crushing chamber and a drying chamber are sequentially arranged inside the drying cylinder from top to bottom. The crushing module is arranged in the crushing chamber. A feed valve for adding coal to be detected is arranged on the drying cylinder and communicated with the crushing chamber. The crushing chamber and the drying chamber are communicated through a crushing discharge port, and a crushing discharge valve is arranged on the crushing discharge port. The intake pipeline is communicated with the first side of the drying chamber;

[0018] A drying discharge port is formed on the second side of the drying chamber. An exhaust valve is further arranged on the drying chamber. The drying discharge port is communicated with the tablet pressing module through a drying discharge pipe, and a drying discharge valve is arranged on the drying discharge pipe.

[0019] Preferably, the coal ash analysis module further includes a spectrum data self-checking and adjusting sub-module. The spectrum data self-checking and adjusting sub-module determines whether the currently collected X-ray fluorescence spectrum data can meet the analysis requirements, and adjusts the characteristic spectrum data screened by the spectrum screening sub-model to obtain the verified characteristic spectrum data, and then uses the verified characteristic spectrum data as the input of the ash prediction sub-model to analyze and obtain the ash value of the coal test sample.

[0020] Preferably, the ash prediction model is constructed by the following method:

[0021] S1. Collect the original X-ray fluorescence spectrum data of the coal test sample through the X-ray fluorescence spectrum collection module, and analyze the coal components of the coal test sample through X-ray fluorescence spectrum analysis;

[0022] S2. Determine the ash value of the coal test sample corresponding to the original X-ray fluorescence spectrum data through coal proximate analysis;

[0023] S3. The spectrum screening sub-module calculates the importance scores of the relationship between each oxide and the ash value in the coal test sample by using the random forest algorithm according to the coal components and the corresponding ash values, and then sorts them from high to low according to the importance scores. Take all oxides ranked in the top M1 as the main related oxides, and take all oxides ranked from M1 + 1 to M2 as the secondary related oxides; both M1 and M2 are positive integers, and M1 < M2;

[0024] The spectrum screening sub-module further screens out the spectral data corresponding to each main related oxide from the original X-ray fluorescence spectrum data, denoted as the main related oxide spectral data; and screens out the spectral data corresponding to each secondary related oxide from the original X-ray fluorescence spectrum data, denoted as the secondary related oxide spectral data;

[0025] S4. The spectral screening sub-module uses all the spectral data of the main relevant oxides and all the spectral data of the minor relevant oxides as characteristic spectral data, inputs them into the support vector regression model for the first-stage training, optimizes the support vector regression model using the particle swarm algorithm during the training process, with the output being the ash content value of the coal test sample. After the training is completed, an initial ash content prediction model is obtained.

[0026] S5. The spectral screening sub-module uses all the spectral data of the main relevant oxides as characteristic spectral data, inputs them into the initial support vector regression model for the second-stage training, with the output being the ash content value of the coal test sample. After the training is completed, a final ash content prediction model is obtained.

[0027] In the second aspect of the present invention, an on-line coal ash content detection method based on X-ray fluorescence analysis is provided. It uses the system as described above, and this method includes the following steps:

[0028] Step 1. The coal to be detected is conveyed through the feed valve to the pulverizing chamber of the drying cylinder, and the pulverized coal after being pulverized by the pulverizing module enters the drying chamber through the crushing discharge valve.

[0029] Step 2. The air pump inputs gas into the heater, and the heated gas enters the drying chamber through the intake valve to dry the pulverized coal in the drying chamber.

[0030] Step 3. After the drying is completed, open the drying discharge valve and close the exhaust valve. The dried pulverized coal is input into the tablet pressing module by the air pressure of the gas, and a coal test sample is obtained through tablet pressing by the tablet pressing module; then close the drying discharge valve, open the exhaust valve, discharge the remaining pulverized coal in the drying chamber, and then close the exhaust valve.

[0031] Step 4. The conveying module conveys the coal test sample to the detection position of the coal ash content analysis system.

[0032] Step 5. The coal ash content analysis system performs ash content detection on the coal test sample.

[0033] Preferably, the steps for the coal ash content analysis system to perform ash content detection include:

[0034] Step 5-1. The X-ray fluorescence spectrum acquisition module acquires the original X-ray fluorescence spectrum data of the coal test sample.

[0035] Step 5-2. The spectral screening sub-module screens out all the spectral data V1 of the main relevant oxides and all the spectral data V2 of the minor relevant oxides from the original X-ray fluorescence spectrum data.

[0036] Step 5-3: The spectral data self-check and adjustment sub-module determines whether the currently collected X-ray fluorescence spectral data can meet the analysis requirements based on the main related oxide spectral data V1 and the secondary related oxide spectral data V2:

[0037] When it is determined that the analysis requirements are met, the verified characteristic spectral data is output, and Step 5-4 is entered;

[0038] When it is determined that the analysis requirements are not met, the current analysis step ends, and the main control module controls the re-acquisition of the original X-ray fluorescence spectral data for the current coal test sample. If the re-acquired original X-ray fluorescence spectral data still cannot meet the analysis requirements, the coal test sample is re-prepared until the original X-ray fluorescence spectral data that meets the analysis requirements is obtained;

[0039] Step 5-4: Input the verified characteristic spectral data into the ash content prediction sub-model to analyze and obtain the ash content value of the coal test sample.

[0040] Preferably, Step 5-3 is specifically as follows:

[0041] Step 5-3-1: The spectral data self-check and adjustment sub-module counts the number of main related oxides corresponding to the main related oxide spectral data V1, denoted as M'1, and counts the number of main related oxides corresponding to the secondary related oxide spectral data V2, denoted as M'2;

[0042] Step 5-3-2: If It is determined that the currently collected X-ray fluorescence spectral data meets the analysis requirements, and all the main related oxide spectral data V1 is directly output as the verified characteristic spectral data, and Step 5-4 is entered;

[0043] Step 5-3-3: If It is determined that the currently collected X-ray fluorescence spectral data meets the analysis requirements, and all the main related oxide spectral data V1 and all the secondary related oxide spectral data V2 are combined and output as the verified characteristic spectral data, and Step 5-4 is entered; where α is a preset judgment threshold, and α < 1;

[0044] Step 5-3-4: If It is determined that the currently collected X-ray fluorescence spectral data does not meet the analysis requirements, the current analysis step ends, and the main control module controls the re-acquisition of the original X-ray fluorescence spectral data for the current coal test sample. If the re-acquired original X-ray fluorescence spectral data still cannot meet the analysis requirements, the coal test sample is re-prepared until the original X-ray fluorescence spectral data that meets the analysis requirements is obtained.

[0045] Preferably, where α = 0.75 - 0.90, M1 = 4 - 7, M2 = 6 - 12, and M1 < M2.

[0046] Preferably, α = 0.8, M1 = 5, M2 = 10; the main related oxides include SiO2, Al2O3, CaO, TiO2, Fe2O3, and the minor related oxides include P4O 10 , K2O, SO2, Na2O, Mn3O4.

[0047] The beneficial effects of the present invention are:

[0048] The present invention provides an on-line coal ash detection system and detection method based on X-ray fluorescence analysis. The entire detection process of the present invention can achieve automated operation, and at the same time achieve non-destructive and rapid detection. Compared with the traditional method of detecting ash content by high-temperature burning method, the present invention greatly shortens the time, improves the efficiency, has high accuracy, reduces the error caused by manual operation, and has good application prospects.

[0049] The coal ash analysis system of the present invention uses a machine learning method. By using the X-ray fluorescence spectrum measurement results, the coal ash content can be analyzed, which has the advantages of rapidity and accuracy. In the present invention, the importance of each oxide in coal to the ash content is ranked by the random forest algorithm, and then the oxides with great influence on the ash content are selected as characteristic oxides to be used to predict the ash value in the subsequent steps, which can significantly reduce the data processing volume and avoid the interference of some unimportant components on the prediction results at the same time.

[0050] In the coal ash analysis module of the present invention, the random forest algorithm (RF), support vector regression (SVR) and particle swarm optimization algorithm (PSO) are combined, which can maximize the advantages of each algorithm, solve the problems that may exist in a single algorithm, and thus improve the accuracy and stability of the model in the ash content prediction application of the present invention.

[0051] In the present invention, a spectral data self-checking and adjusting sub-module is set to judge whether the actually collected X-ray fluorescence spectrum data can meet the analysis requirements: when the original X-ray fluorescence spectrum data actually includes the spectral data of all the main related oxides, only the spectral data of the main related oxides is needed for ash content detection, thus reducing the data processing volume; when the spectral data of one of the oxides is missing, by combining the spectral data of the spare minor related oxides with the spectral data of the main related oxides, the analysis requirements can still be met; when the spectral data of the missing oxides reaches two or more, it is considered that there is a problem with the original X-ray fluorescence spectrum data and it cannot be used for ash content detection, and the X-ray fluorescence spectrum needs to be collected again or the coal test sample needs to be prepared again. Thus, through this self-checking scheme, it can be judged whether there are obvious problems with the obtained original X-ray fluorescence spectrum data, so as to better ensure the accuracy of the detection results. Brief Description of the Drawings

[0052] Figure 1 is a schematic structural diagram of the on-line coal ash detection system based on X-ray fluorescence analysis of the present invention;

[0053] Figure 2 is a schematic diagram of the construction process of the ash prediction model of the present invention;

[0054] Figure 3 is a flow chart of the on-line coal ash detection method based on X-ray fluorescence analysis of the present invention;

[0055] Figure 4 is a schematic diagram of the process of ash detection by the coal ash analysis system of the present invention;

[0056] Figure 5 is the result of the importance score ranking obtained in Example 1 of the present invention;

[0057] Figure 6 is the prediction result using five main related oxides in the test example of the present invention;

[0058] Figure 7 is the prediction result of missing the main related oxide SiO2 in the test example of the present invention;

[0059] Figure 8 is the prediction result of using four main related oxides: Al2O3, CaO, TiO2, Fe2O3 combined with 5 minor related oxides in the test example of the present invention;

[0060] Figure 9 is the prediction result of missing the main related oxide Al2O3 in the test example of the present invention;

[0061] Figure 10 is the prediction result of using four main related oxides: SiO2, CaO, TiO2, Fe2O3 combined with 5 minor related oxides in the test example of the present invention;

[0062] Figure 11 is the prediction result of missing the main related oxide CaO in the test example of the present invention;

[0063] Figure 12 is the prediction result of using four main related oxides: SiO2, Al2O3, TiO2, Fe2O3 combined with 5 minor related oxides in the test example of the present invention;

[0064] Figure 13 is the prediction result of missing the main related oxide TiO2 in the test example of the present invention;

[0065] Figure 14It is the prediction result of combining four main related oxides: SiO2, Al2O3, CaO, Fe2O3 with five minor related oxides in the test examples of the present invention;

[0066] Figure 15 It is the prediction result of lacking the main related oxide Fe2O3 in the test examples of the present invention;

[0067] Figure 16 It is the prediction result of combining four main related oxides: SiO2, Al2O3, CaO, TiO2 with five minor related oxides in the test examples of the present invention;

[0068] Explanation of reference numerals:

[0069] 1 - Crushing module;

[0070] 2 - Drying module; 21 - Air pump; 22 - Heater; 23 - Inlet gas pipeline; 24 - Drying cylinder; 25 - Inlet gas valve; 26 - Crushing chamber; 27 - Drying chamber; 28 - Crushing discharge port; 241 - Feed valve; 242 - Drying discharge pipe; 243 - Drying discharge valve; 261 - Auxiliary inlet gas valve; 262 - Auxiliary inlet gas pipe; 271 - Drying discharge port; 272 - Exhaust valve; 281 - Crushing discharge valve;

[0071] 3 - Tablet pressing module;

[0072] 4 - Conveying module;

[0073] 5 - Coal ash analysis system; 51 - X-ray fluorescence spectrum acquisition module; 52 - Coal ash analysis module. Detailed implementation manners

[0074] The following further elaborates on the present invention in conjunction with embodiments, so that those skilled in the art can implement it with reference to the text of the specification.

[0075] It should be understood that terms such as "having", "comprising", and "including" as used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0076] Embodiment 1

[0077] An on-line coal ash detection system based on X-ray fluorescence analysis, comprising:

[0078] Crushing module 1 for crushing the coal to be detected;

[0079] Drying module 2 for drying the coal to be detected;

[0080] Tablet pressing module 3 for pressing the crushed and dried coal into a coal test sample;

[0081] A conveying module 4 for conveying coal test samples to the detection position;

[0082] A coal ash analysis system 5 for collecting X-ray fluorescence spectra of coal test samples at the detection position and analyzing to obtain the ash content values of the coal test samples;

[0083] And a main control module for controlling the crushing module 1, the drying module 2, the tablet pressing module 3, the conveying module 4 and the coal ash analysis system 5.

[0084] Among them, the coal ash analysis system 5 includes an X-ray fluorescence spectrum collection module 51 and a coal ash analysis module 52. The coal ash analysis module 52 includes a spectrum screening sub-module and an ash content prediction model. The X-ray fluorescence spectrum collection module 51 collects X-ray fluorescence spectrum data of coal test samples. The spectrum screening sub-model screens out characteristic spectrum data from the collected X-ray fluorescence spectrum data. The ash content prediction sub-model analyzes to obtain the ash content values of the coal test samples based on the characteristic spectrum data.

[0085] In this embodiment, the drying module 2 includes an air pump 21, a heater 22 connected to the air pump 21, a drying cylinder 24 connected to the heater 22 through an intake pipeline 23, and an intake valve 25 provided on the intake pipeline 23.

[0086] In this embodiment, a crushing chamber 26 and a drying chamber 27 are sequentially arranged inside the drying cylinder 24 from top to bottom. The crushing module 1 is arranged in the crushing chamber 26. The drying cylinder 24 is provided with a feed valve 241 for adding coal to be detected and communicating with the crushing chamber 26. The crushing chamber 26 and the drying chamber 27 are communicated through a crushing discharge port 28. A crushing discharge valve 281 is provided on the crushing discharge port 28; the intake pipeline 23 is communicated with the first side of the drying chamber 27;

[0087] A drying discharge port 271 is opened on the second side of the drying chamber 27. The first side and the second side are on opposite sides. An exhaust valve 272 is further provided on the drying chamber 27. The drying discharge port 271 is communicated with the tablet pressing module 3 through a drying discharge pipe 242. A drying discharge valve 243 is provided on the drying discharge pipe 242.

[0088] Among them, the intake valve 25, the feed valve 241, the crushing discharge valve 281, the exhaust valve 272, and the drying discharge valve 243 are all connected to the main control module and are automatically controlled to open and close through the main control module.

[0089] The X-ray fluorescence spectrum acquisition module 51 is a conventional X-ray fluorescence spectrometer, which acquires the X-ray fluorescence spectrum data of the coal test sample and can analyze and obtain various components in the coal. The coal ash analysis module 52 is embedded in the computer and realizes the analysis of the ash content according to the X-ray fluorescence spectrum data collected by the X-ray fluorescence spectrum acquisition module 51, and outputs the analysis result with the aid of the computer.

[0090] Among them, the conveying module 4 is a conventional mechanical gripper, which can, under the control of the main control module, grab and move the coal test sample pressed by the tablet pressing module 3 to the detection position of the X-ray fluorescence spectrometer.

[0091] Among them, the pulverizing module 1 adopts a conventional coal pulverizing device, which is not limited in the present invention. It is integrally installed in the pulverizing chamber 26 to meet the use of the present invention. The tablet pressing module 3 adopts a conventional coal tablet pressing device, which is not limited in the present invention.

[0092] This embodiment also provides an on-line coal ash detection method based on X-ray fluorescence analysis, which uses the above system to perform on-line detection of coal ash. The method includes the following steps:

[0093] Step 1: The coal to be detected (such as flotation clean coal) is conveyed into the pulverizing chamber 26 of the drying cylinder 24 through the feed valve 241. The pulverizing module 1 pulverizes it into coal powder. Pulverization speeds up the drying speed and also ensures the uniformity of the coal sample. The pulverizing time can generally be 1-5 minutes and is adjusted conventionally according to the properties of the incoming coal. After pulverization, the main control module controls the opening of the crushing discharge valve 281 to make the coal powder enter the drying chamber 27;

[0094] In this embodiment, an auxiliary air inlet valve 261 is also connected above the pulverizing chamber 26. The auxiliary air inlet valve 261 is connected to the heater 22 through an auxiliary air inlet pipe 262. After pulverization, the auxiliary air inlet pipe 262 is opened. Through the action of air pressure, it promotes the coal powder to enter the drying chamber 27 more smoothly, and at the same time discharges the residual coal powder in the pulverizing chamber 26, so as to avoid the influence of the residual coal powder on the next detection;

[0095] Step 2: The air pump 21 inputs gas into the heater 22. The heated gas enters the drying chamber 27 through the air inlet valve 25 to dry the coal powder in the drying chamber 27. The drying time is generally 1-5 minutes and is adjusted conventionally according to the moisture content of the incoming coal. The gas input by the air pump 21 can be air, or nitrogen or other inert gases. In this embodiment, the input gas is air; the heated air can quickly dry the coal powder; the excess waste gas is discharged from the exhaust valve 272;

[0096] Step 3: After drying is completed, open the drying discharge valve 243 and close the exhaust valve 272. Through the air pressure of the gas, the dried pulverized coal is input into the tablet pressing module 3, and a coal test sample is obtained by pressing with the tablet pressing module 3. The pressing time is generally 60 s. Then close the drying discharge valve 243, open the exhaust valve 272, discharge the residual pulverized coal in the drying chamber 27, and then close the exhaust valve 272.

[0097] Step 4: The conveying module 4 conveys the coal test sample to the detection position of the coal ash analysis system 5.

[0098] Step 5: The coal ash analysis system 5 performs ash detection on the coal test sample, and the detection time is approximately 80 - 120 s.

[0099] It can be seen that the entire process can achieve automated operation, and at the same time achieve non-destructive and rapid detection. The total time consumption is approximately 12 minutes. Compared with the traditional coal ash analyzer that uses the high-temperature burning method for ash detection, the time consumption is approximately 3 hours. This embodiment greatly shortens the time, improves the efficiency, has high accuracy, and reduces the errors caused by manual operation.

[0100] In this embodiment, the coal ash analysis module 52 further includes a spectral data self-checking and adjusting sub-module. The spectral data self-checking and adjusting sub-module determines whether the currently collected X-ray fluorescence spectral data can meet the analysis requirements, and adjusts the characteristic spectral data screened by the spectral screening sub-model to obtain the verified characteristic spectral data, and then uses the verified characteristic spectral data as the input of the ash prediction sub-model to analyze and obtain the ash value of the coal test sample.

[0101] 5. The on-line coal ash detection system based on X-ray fluorescence analysis according to claim 4, wherein the ash prediction model is constructed by the following method:

[0102] S1: Collect the original X-ray fluorescence spectral data of the coal test sample through the X-ray fluorescence spectral collection module 51, and obtain the coal components of the coal test sample through X-ray fluorescence spectral analysis.

[0103] X-ray fluorescence spectroscopy can obtain coal components. The X-ray tube generates incident X-rays to excite the sample to be measured. Each element in the excited sample will emit specific X-ray fluorescence of that element. The detection system measures the energy or wavelength of the emitted X-ray fluorescence and converts it into the type of corresponding element in the sample, and finally realizes the detection of coal components.

[0104] In this embodiment, a total of 250 groups of coal samples were taken from the production samples of Datun Coal and Power Group. After the samples were collected, all coal samples were screened, reduced, dried, etc. according to GB474—2008 "Preparation Method of Coal Samples". To prepare the pressed particles, the pulverized coal samples were pressed in a hydraulic molding press (YST-45T) with a diameter of 35 mm with a force of 25 tons for 60 seconds to obtain coal test samples for X-ray fluorescence spectroscopy data. The diameter of the sample tablets was 35 mm and the thickness was about 4 mm. The X-ray fluorescence spectrum was collected using a desktop EDXRF analyzer (EDX 4500H), with dimensions of 65 cm (H) × 80 cm (W) × 60 cm (D), a maximum power of 50 W, a maximum voltage of 50 kV, and a maximum current of 2 mA for analyzing the samples.

[0105] The process of collecting the original X-ray fluorescence spectrum data is as follows:

[0106] First, the coal samples were dried, and then ground to obtain pulverized coal samples; the pulverized coal samples were reduced. Then, tablets were pressed to obtain the required test coal cakes: coal test samples. The counting rate of the desktop EDXRF analyzer was set to 4000 seconds / time. The probe of the spectrometer was 480 mm away from the surface of the pulverized coal sample tablet and perpendicular to the surface of the pulverized coal sample tablet. The halogen lamp was 320 mm away from the pulverized coal sample tablet and formed an angle of 45 degrees with the surface of the pulverized coal sample tablet. Since X-rays are sensitive to temperature changes and will affect the accuracy of the final spectrum, the entire experiment was kept under constant temperature control at 25 °C. The desktop EDXRF (EDX 4500H) analyzer was used to perform three spectral tests on each pulverized coal sample tablet, and then the average value was taken as the X-ray fluorescence spectrum data of the pulverized coal sample tablet.

[0107] Referring to Table 1, the X-ray fluorescence spectrum analysis results (wt%) of some coal samples are as follows:

[0108] Table 1

[0109]

[0110]

[0111] S2. Determine the ash content of the coal test sample corresponding to the original X-ray fluorescence spectrum data through industrial analysis of coal;

[0112] In this embodiment, a coal ash analyzer was used to determine the ash content by the high-temperature ignition method, that is, the difference in weight before and after the sample was ignited was measured and calculated respectively, and then divided by the weight of the sample before ignition. The obtained value was the ash content of the sample; after each coal test sample was subjected to X-ray fluorescence spectrum data collection in step S1, its ash content was obtained by this method (tested three times, and the average value was taken as the result). This ash content was used as the gold standard for subsequent construction steps.

[0113] S3. The spectral screening sub-module calculates the importance scores of the relationship between each oxide and the ash value in the coal test sample using the random forest algorithm based on the coal composition and the corresponding ash values, and then sorts them from high to low according to the importance scores. All oxides ranked among the top M1 are taken as the main relevant oxides, and all oxides ranked from M1 + 1 to M2 are taken as the secondary relevant oxides; both M1 and M2 are positive integers, and M1 < M2.

[0114] The spectral screening sub-module then screens out the spectral data corresponding to each main relevant oxide from the original X-ray fluorescence spectral data, denoted as the main relevant oxide spectral data; and screens out the spectral data corresponding to each secondary relevant oxide from the original X-ray fluorescence spectral data, denoted as the secondary relevant oxide spectral data.

[0115] In the present invention, the coal composition of the coal test sample obtained by X-ray fluorescence spectroscopy usually includes more than 20 oxides. The relationships between different oxides and their ash contents are different. By using the random forest algorithm to rank the importance of the influence of each oxide in the coal on the ash content, and then screening out the oxides with a large influence on the ash content as the characteristic oxides to predict the ash value in the subsequent steps, it can significantly reduce the data processing volume and avoid the interference of some unimportant components on the prediction results.

[0116] The random forest algorithm evaluates the feature importance by assessing the reduction in the Gini index when each feature is split in the decision tree. Each decision tree in the random forest is generated by training on the input data set. During the training process, the data set is randomly divided into multiple subsets, and each subset will be used to train a single decision tree. Whenever a tree uses a certain feature for splitting, the reduction in impurity contributed by that feature will be calculated. The splitting criterion adopted is the Gini index:

[0117]

[0118] where: p i is the proportion of class i in the data set, and C is the number of classes.

[0119] For each feature, accumulate the reduction in impurity contributed by that feature in all trees to obtain the importance score of that feature. Finally, by averaging the reduction in impurity of the same feature in all trees, the final importance of the feature is obtained:

[0120]

[0121] where: T is the number of trees in the random forest; n t is the number of nodes in tree t; ΔG ini (i, f) is the reduction in the Gini index of feature f at node i.

[0122] The random forest algorithm is currently widely used in the field of machine learning, especially in feature importance evaluation. The random forest can evaluate the relative importance of each oxide content to the coal ash value by constructing multiple decision trees. During the analysis process, the random forest can capture the complex non-linear relationship between the oxide content and the ash value, and at the same time quantify the contribution of each oxide (such as SiO2, Fe2O3, Al2O3, CaO, etc.) to the ash value through the feature importance score. Compared with traditional univariate analysis methods, the random forest algorithm not only has strong robustness and generalization ability under high-dimensional features, but also can provide more quantitative references on the characteristics of each oxide to the ash.

[0123] S4. The spectral screening sub-module takes all the main relevant oxide spectral data and all the secondary relevant oxide spectral data as the feature spectral data and inputs them into the support vector regression model for the first-stage training. During the training process, the particle swarm optimization algorithm is used to optimize the support vector regression model. The output end is the ash value of the coal test sample. After the training is completed, the initial ash prediction model is obtained;

[0124] S5. The spectral screening sub-module takes all the main relevant oxide spectral data as the feature spectral data and inputs them into the initial support vector regression model for the second-stage training. During the training process, the particle swarm optimization algorithm is also used to optimize the support vector regression model. The output end is the ash value of the coal test sample. After the training is completed, the final ash prediction model is obtained.

[0125] The theoretical function of the support vector regression algorithm is as follows:

[0126] f(x i )=ω T δ(x i )+σ

[0127] Where: f(x i ) is the predicted value of the input data x i . ω is the weight vector, which determines the direction of the hyperplane, and T represents the transpose operation. δ(x i ) is the mapping function. σ is the bias term, which adjusts the vertical position of the hyperplane to better fit the data.

[0128] The support vector machine realizes the optimization of the regression problem by minimizing the following objective function:

[0129]

[0130] The constraint condition is:

[0131]

[0132] Where: C is the penalty coefficient; ξi and are slack variables; ε is the deviation of the loss function; m represents the total number of samples in the training dataset (i.e., the number of data points), and x i is the input feature vector of the i-th sample, and y i is the actual target value corresponding to the input feature xi.

[0133] By introducing the Lagrange multiplier method to solve the objective function, the final weight ω and the function f(x) can be expressed as:

[0134]

[0135] ai represents the Lagrange multiplier corresponding to the upper bound constraint condition of the i-th sample, and bi represents the Lagrange multiplier corresponding to the lower bound constraint condition of the i-th sample;

[0136] where: K(x, x i ) is the kernel function used to map the input to a high-dimensional space, and its form is:

[0137] K(x, x i ) = exp(-g‖x - x i ‖ 2 )

[0138] where: g is the kernel parameter that determines the distribution shape of the kernel function.

[0139] During the training process, the particle swarm optimization algorithm is used to optimize the support vector regression model (SVR). First, the key parameters of the SVR are corresponding to the particle positions, and the SVR with the corresponding parameters is used for prediction. The fitness value is measured by the result. The particles update according to their own and the group's optimal positions. Then, the corresponding new parameters are used for training and prediction, and the better "parameter combination" is saved. The loop operation is performed until the termination condition is reached, and finally, the parameter values that make the SVR prediction performance optimal are obtained.

[0140] The particle swarm optimization algorithm (PSO) is a swarm intelligence optimization algorithm that simulates the foraging behavior of bird flocks. It searches for the optimal solution by a group of "particles" (i.e., candidate points of the solution) flying in the search space. Each particle has a position and a velocity in the search space, and during the search process, the particle adjusts its position according to its own experience and the experience of the group.

[0141] The particle swarm optimization algorithm was proposed by Kennedy and Eberhart in 1995 as a heuristic optimization algorithm, which mimics the behavior of bird flocks in the process of finding food. The core idea of the particle swarm optimization algorithm is to find the optimal solution through the current state of each particle and the experience of other particles in the group.

[0142] Each particle has a position X in the solution space i =(xi1 , x i2 ,..., x id ), representing a potential solution. i represents the index number of the i-th particle in the particle swarm, and d is the total number of particles.

[0143] Each particle also has a velocity v i = (v i1 , v i2 ,..., v id ), representing the moving direction and speed of the particle in the solution space.

[0144] Each particle updates its solution according to its position and velocity, and evaluates the quality of the current solution based on the fitness function.

[0145] The velocity and particle update equations used in PSO are as follows:

[0146] U i (t + 1) = w(t)·U i (t) + c1·r1·(p best,i - X i ) + c2·r2·(g best - X i )

[0147] X i (t + 1) = X i (t) + U i (t + 1)

[0148] Where: U i (t) is the velocity of the particle at iteration t, X i (t) represents its current position, P best,i i is its best-known position, and g best is the global best position among all particles. The inertia weight w(t) usually decreases linearly during the iteration process and can be calculated as:

[0149]

[0150] Where: w max and w min are the maximum and minimum inertia weights, t is the current iteration. The acceleration coefficients c1 and c2 and the random values r1 and r2 control the movement of the particle. MaxIt represents the maximum number of iterations for the algorithm to run.

[0151] The present invention combines the random forest algorithm (RF), support vector regression (SVR) with the particle swarm algorithm (PSO):

[0152] Random forest is an ensemble learning method that constructs multiple decision trees for classification or regression tasks, can effectively evaluate the importance of features and reduce overfitting. However, it may not be precise enough in capturing complex non-linear relationships.

[0153] Support vector regression model is good at dealing with non-linear regression problems, and solves complex problems that linear regression cannot handle by mapping to a high-dimensional space through kernel functions. But SVR is sensitive to the selection of hyperparameters and requires tuning. Particle swarm optimization algorithm is a global optimization algorithm that can efficiently search for the optimal solution and is especially suitable for optimizing the hyperparameters of SVR, such as kernel function type and penalty factor.

[0154] Based on these situations, combining the random forest algorithm, support vector regression and particle swarm optimization algorithm can maximize the advantages of each algorithm, solve the problems that may exist in a single algorithm, and thus improve the accuracy and stability of the model in practical applications.

[0155] In this embodiment, the steps for the coal ash analysis system 5 to perform ash detection include:

[0156] Step 5-1, the X-ray fluorescence spectrum acquisition module 51 acquires the original X-ray fluorescence spectrum data of the coal test sample;

[0157] Step 5-2, the spectrum screening sub-module screens out all the actually obtained main relevant oxide spectrum data V1 and all the secondary relevant oxide spectrum data V2 from the original X-ray fluorescence spectrum data;

[0158] Step 5-3, the spectrum data self-checking and adjustment sub-module determines whether the currently acquired X-ray fluorescence spectrum data can meet the analysis requirements according to the main relevant oxide spectrum data V1 and the secondary relevant oxide spectrum data V2:

[0159] Step 5-3-1, the spectrum data self-checking and adjustment sub-module counts the number of main relevant oxides corresponding to the main relevant oxide spectrum data V1, denoted as M'1, and counts the number of main relevant oxides corresponding to the secondary relevant oxide spectrum data V2, denoted as M'2;

[0160] Step 5-3-2, if It is determined that the currently acquired X-ray fluorescence spectrum data meets the analysis requirements, and all the main relevant oxide spectrum data V1 are directly output as the verified characteristic spectrum data, and step 5-4 is entered;

[0161] Step 5-3-3, if Determine whether the currently collected X-ray fluorescence spectrum data meets the analysis requirements, and combine all the spectral data V1 of the main related oxides and all the spectral data V2 of the minor related oxides as the verified characteristic spectral data for output, and enter step 5-4; where α is a preset judgment threshold, and in this embodiment, α = 0.8;

[0162] Step 5-3-4, if Determine that the currently collected X-ray fluorescence spectrum data does not meet the analysis requirements, end the current analysis step, and the main control module controls the re-acquisition of the original X-ray fluorescence spectrum data for the current coal test sample. If the re-acquired original X-ray fluorescence spectrum data still cannot meet the analysis requirements, then re-prepare the coal test sample until the original X-ray fluorescence spectrum data that meets the analysis requirements is obtained.

[0163] Step 5-4, input the verified characteristic spectral data into the ash content prediction sub-model to analyze and obtain the ash content value of the coal test sample.

[0164] Refer to Figure 5 , for the oxides ranked in the top 11 in descending order of importance scores obtained by the random forest algorithm in this embodiment. Among them, the contents of SiO2, Al2O3, CaO, TiO2, and Fe2O3 contribute significantly to the ash content value. P4O 10 , K2O, SO2, Na2O, Mn3O4 also have a certain contribution, but the scores are significantly lower than the first 5 types and can be used as alternatives. Therefore, in this embodiment, the main related oxides selected include SiO2, Al2O3, CaO, TiO2, and Fe2O3, and the minor related oxides include P4O 10 , K2O, SO2, Na2O, Mn3O4. That is, in this embodiment, α = 0.8, M1 = 5, and M2 = 10. That is to say, in this embodiment, when the original X-ray fluorescence spectrum data actually includes the spectral data of all 5 main related oxides, only the spectral data of these 5 main related oxides are required for ash content detection, and these data are directly output as the verified characteristic spectral data, thereby reducing the data processing volume. When the spectral data of one of the oxides is missing, it can still be used as ash content detection data, but at this time, the spectral data V2 of the spare minor related oxides also needs to be combined with the spectral data of the main related oxides to ensure the analysis accuracy. When the spectral data of 2 or more missing oxides are missing, it is considered that there is a problem with the original X-ray fluorescence spectrum data (for example, there is a problem in the X-ray fluorescence spectrum collection process or the preparation of the coal test sample), and it cannot be used for ash content detection, and the X-ray fluorescence spectrum needs to be re-collected, or the coal test sample needs to be re-prepared. Thus, through such a self-verification step, it can be determined whether there are obvious problems with the obtained original X-ray fluorescence spectrum data, thereby better ensuring the accuracy of the detection results.

[0165] Test Case

[0166] 1. As above, 100 coal test samples were prepared from the production samples of Datun Coal and Electricity Group. According to the method of Example 1, the ash content prediction sub-model in Example 1 was used to predict the ash content, and the obtained value was the model prediction value; then, the coal test sample after fluorescence detection was measured by using a coal ash determination instrument and the ash content was determined by high temperature burning method, and the result was used as the actual value of the sample. The model performance evaluation index Coefficient of determination (R 2 ) and Mean Absolute Percentage Error (MAPE) were used to evaluate the model performance.

[0167]

[0168] R 2 The value range is usually between 0 and 1. The closer it is to 1, the better the model fits the data, and the closer it is to 0, the worse the fit.

[0169]

[0170] x(t) is the actual value of the sample, is the model prediction value, n is the number of samples

[0171] MAPE measures the average level of relative error between the predicted value and the true value in the form of a percentage. The smaller the value, the higher the relative accuracy of the prediction.

[0172] 2. Prediction results

[0173] 2-1. Prediction using five main related oxides:

[0174] The main relevant oxides include SiO2, Al2O3, CaO, TiO2, and Fe2O3. The spectral data of the five main relevant oxides are input into the ash prediction sub-model to predict the ash content. Figure 6 The ash content of the prediction set covers a range from 5.28% to 10.91%. It can be seen that the ash prediction sub-model can accurately predict the ash content in coal, and the main error in the prediction result is caused by the 43rd and 52nd samples. 2 The value reaches 0.9850 and MAPE is 1.4002, indicating that the model has relatively good prediction performance.

[0175] 2-2. Prediction of missing main related oxide SiO2:

[0176] Silicon dioxide (SiO2) is one of the main components in coal ash and usually exists in the form of quartz, clay or other silicate minerals. In the proximate analysis of coal, the determination of silicon dioxide content has an important impact on the combustion characteristics, ash-forming properties and chemical stability of coal. Since silicon dioxide is stable at high temperatures and is not easily transformed or volatilized, it usually remains in the coal ash in solid form during the coal combustion process.

[0177] In this example, four main related oxides: Al2O3, CaO, TiO2, Fe2O3 are used. The spectral data of these four main related oxides are input into the ash prediction sub-model to predict the ash content, referring to Figure 7 For the prediction results, the silicon dioxide values in the prediction set cover a range from 4.2% to 5.08%. R 2 The value is 0.9699 and the MAPE is 2.1778. It can be seen that the prediction accuracy decreases significantly at this time. Therefore, when silicon dioxide is missing, it will affect the accuracy of the prediction results.

[0178] To address this problem, in this example, the four main related oxides: Al2O3, CaO, TiO2, Fe2O3 are combined with 5 minor related oxides including P4O 10 , K2O, SO2, Na2O, Mn3O4. The spectral data of the main related oxides and the spectral data of the minor related oxides are input into the ash prediction sub-model together to predict the ash content, referring to Figure 8 For the prediction results, it can be seen that at this time, the R 2 value is 0.9714 and the MAPE is 2.0143. Compared with only using the spectral data of the four main related oxides, the prediction accuracy is improved at this time.

[0179] 2-3. Prediction of the missing main related oxide Al2O3:

[0180] Aluminum oxide (Al2O3) is one of the important components in coal ash and usually comes from the mixture of plant residues and surrounding sedimentary clay minerals (such as kaolin, montmorillonite, etc.). These minerals are rich in aluminum elements. In the proximate analysis of coal, the determination of aluminum oxide content has an important impact on the ash-forming characteristics, melting behavior and boiler operation performance of coal. Since aluminum oxide is relatively stable at high temperatures, it usually does not volatilize or decompose and can react with other mineral components to form complex compounds. Therefore, it remains in the coal ash in solid form during the coal combustion process.

[0181] In this example, four main related oxides: SiO2, CaO, TiO2, Fe2O3 are used. The spectral data of these four main related oxides are input into the ash prediction sub-model to predict the ash content, referring to Figure 9 For the prediction results, it can be seen that R 2The value is 0.9727 and the MAPE is 2.1187. It can be seen that the prediction accuracy drops significantly at this time, and the main errors in the prediction results are caused by the 39th and 53rd samples. Therefore, when there is Al2O3, it will affect the accuracy of the prediction results..

[0182] To address this issue, in this example, the four main related oxides: SiO2, CaO, TiO2, Fe2O3, are combined with 5 minor related oxides including P4O 10 , K2O, SO2, Na2O, Mn3O4. The spectral data of the main related oxides and the spectral data of the minor related oxides are input into the ash prediction sub-model together to predict the ash content. Refer to Figure 10 For the prediction result, it can be seen that at this time, the R 2 value is 0.9759 and the MAPE is 1.9768; compared with only using the spectral data of the four main related oxides, the prediction accuracy is significantly improved at this time.

[0183] 2 - 4. Prediction of the missing main related oxide CaO:

[0184] Calcium oxide (CaO) is one of the important components in coal ash and usually comes from calcareous minerals in coal, such as calcite, dolomite or gypsum. In the proximate analysis of coal, the determination of calcium oxide content is of great significance for the physical and chemical properties and total evaluation of ash. Since calcium oxide can react with other mineral components at high temperatures to form stable compounds such as calcium silicate or calcium aluminate, it has a significant impact on the fusion characteristics and chemical composition of coal ash. In addition, although calcium oxide itself does not participate in combustion heat release, it will directly affect the proportion of ash and the quality evaluation of coal.

[0185] In this example, four main related oxides: SiO2, Al2O3, TiO2, Fe2O3, are used. The spectral data of these four main related oxides are input into the ash prediction sub-model to predict the ash content. Refer to Figure 11 For the prediction result, the coverage range of calcium oxide values in the prediction set is from 0.2% to 0.38%. It can be seen that the main errors occur in the 37th, 39th, and 43rd samples, and the R 2 value is 0.9737 and the MAPE is 2.0183. It can be seen that the prediction accuracy drops at this time. Therefore, when CaO is missing, it will affect the accuracy of the prediction results.

[0186] To address this issue, in this example, the four main related oxides: SiO2, Al2O3, TiO2, Fe2O3, are combined with 5 minor related oxides including P4O 10 , K2O, SO2, Na2O, Mn3O4. The spectral data of the main related oxides and the spectral data of the minor related oxides are input into the ash prediction sub-model together to predict the ash content. Refer toFigure 12 For the prediction results, it can be seen that at this time, the R 2 value is 0.9778 and the MAPE is 1.7441; compared with only using the spectral data of the four main related oxides, the prediction accuracy is significantly improved at this time.

[0187] 2 - 5. Prediction of the absence of the main related oxide TiO2:

[0188] Titanium dioxide (TiO2) is one of the important trace components in coal ash, usually derived from titanium minerals in coal, such as rutile and anatase. In the proximate analysis of coal, the content of titanium dioxide has a certain influence on the physicochemical properties and determination results of ash. Due to the high melting point and stability of titanium dioxide, it will not volatilize or decompose during the coal combustion process and usually remains in the coal ash in a solid state, directly increasing the proportion of ash. In addition, titanium dioxide may react with other oxides (such as Fe2O3 or Al2O3) to form complex mineral phases, thus affecting the chemical composition and physical properties of coal ash.

[0189] In this example, four main related oxides: SiO2, Al2O3, CaO, and Fe2O3 are used. The spectral data of these four main related oxides are input into the ash prediction sub - model to predict the ash content. Refer to Figure 13 For the prediction results, the coverage range of titanium dioxide values in the prediction set is from 0.13% to 0.17%. The R 2 value is 0.9749 and the MAPE is 2.0437. It can be seen that the prediction accuracy decreases at this time. Therefore, when TiO2 is absent, it will affect the accuracy of the prediction results.

[0190] To address this problem, in this example, the four main related oxides: SiO2, Al2O3, CaO, and Fe2O3 are combined with 5 minor related oxides including P4O 10 , K2O, SO2, Na2O, and Mn3O4. The spectral data of the main related oxides and the spectral data of the minor related oxides are input into the ash prediction sub - model together to predict the ash content. Refer to Figure 14 For the prediction results, it can be seen that at this time, the R 2 value is 0.9795 and the MAPE is 1.5759; compared with only using the spectral data of the four main related oxides, the prediction accuracy is significantly improved at this time.

[0191] 2 - 6. Detection of iron(III) oxide:

[0192] Iron(III) oxide (Fe2O3) is an important component of coal ash, mainly derived from iron minerals in coal, such as hematite, pyrite, and rare earth minerals. In the proximate analysis of coal, the content of iron(III) oxide affects the composition and determination results of ash. Since iron(III) oxide is unstable at high temperatures and usually does not volatilize or dissolve, it remains in the coal ash in solid form, directly increasing the ash ratio. At the same time, Fe2O3 may react with other mineral components (such as SiO2 and Al2O3) to form compounds such as iron silicate or iron aluminate, thereby changing the chemical composition and physical properties of the ash.

[0193] In this example, four main related oxides: SiO2, Al2O3, CaO, TiO2 were used. The spectral data of these four main related oxides were input into the ash prediction sub-model to predict the ash content, referring to Figure 15 For the prediction results, the PSO-SVR model could basically accurately predict the ash value of coal. The R 2 value was 0.9753 and the MAPE was 1.8919. It can be seen that the prediction accuracy decreased significantly at this time. Therefore, when silica was missing, it would affect the accuracy of the prediction results.

[0194] To address this problem, in this example, the four main related oxides: SiO2, Al2O3, CaO, TiO2 were combined with 5 minor related oxides including P4O 10 , K2O, SO2, Na2O, Mn3O4. The spectral data of the main related oxides and the spectral data of the minor related oxides were input into the ash prediction sub-model together to predict the ash content, referring to Figure 16 For the prediction results, it can be seen that at this time, the R 2 value was 0.9837 and the MAPE was 1.4578; compared with only using the spectral data of the four main related oxides, the prediction accuracy was significantly improved at this time.

[0195] In this embodiment, an experiment on ash prediction using three main related oxides was also carried out. When any three of SiO2, Al2O3, CaO, TiO2, Fe2O3 were taken and input into the ash prediction sub-model to predict the ash content, the accuracy performance was too low (the R 2 values were all lower than 0.9010 and the MAPEs were all higher than 4.2432), and when the spectral data of the three main related oxides and the minor related oxides were input into the ash prediction sub-model together to predict the ash content, the accuracy still could not meet the requirements (the R 2 values were all lower than 0.9357 and the MAPEs were all higher than 3.261), indicating that the ash prediction accuracy was difficult to meet the requirements when only three main related oxides were selected.

[0196] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0197] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to specific details.

Claims

1. An online coal ash detection system based on X-ray fluorescence analysis, characterized in that: include: Crushing module, used to crush the coal to be tested; Drying module, used to dry the coal to be tested; A tabletting module, used to press the crushed and dried coal into coal test samples; A conveying module, used for conveying the coal test sample to the testing location; The coal ash analysis system is used to collect X-ray fluorescence spectra of the coal test samples at the detection position and analyze the ash values ​​of the coal test samples; and a main control module, which is used to control the crushing module, the drying module, the tableting module, the conveying module and the coal ash analysis system; Among them, the coal ash analysis system includes an X-ray fluorescence spectrum acquisition module and a coal ash analysis module, the coal ash analysis module includes a spectrum screening submodule and an ash prediction model, the X-ray fluorescence spectrum acquisition module collects X-ray fluorescence spectrum data of the coal test sample, the spectrum screening submodel filters out characteristic spectrum data from the collected X-ray fluorescence spectrum data, and the ash prediction submodel obtains the ash value of the coal test sample based on the characteristic spectrum data analysis.

2. The online coal ash detection system based on X-ray fluorescence analysis according to claim 1 is characterized in that: The drying module comprises an air pump, a heater connected to the air pump, a drying cylinder connected to the heater via an air intake pipeline, and an air intake valve arranged on the air intake pipeline.

3. The online coal ash detection system based on X-ray fluorescence analysis according to claim 2 is characterized in that: The inside of the drying cylinder is provided with a crushing chamber and a drying chamber from top to bottom in sequence, the crushing module is arranged in the crushing chamber, the drying cylinder is provided with a feeding valve connected with the crushing chamber for adding the coal to be tested, the crushing chamber and the drying chamber are connected through a crushing discharge port, and the crushing discharge port is provided with a crushing discharge valve; the air intake pipeline is connected with the first side of the drying chamber; A drying discharge port is provided on the second side of the drying chamber, and an exhaust valve is also provided on the drying chamber. The drying discharge port is connected to the tablet pressing module through a drying discharge pipe, and a drying discharge valve is provided on the drying discharge pipe.

4. The online coal ash detection system based on X-ray fluorescence analysis according to claim 3 is characterized in that: The coal ash analysis module also includes a spectral data self-verification and adjustment submodule, which determines whether the currently collected X-ray fluorescence spectral data can meet the analysis requirements, and adjusts the characteristic spectral data screened out by the spectral screening submodel to obtain the verified characteristic spectral data, and then uses the verified characteristic spectral data as the input of the ash prediction submodel to analyze and obtain the ash value of the coal test sample.

5. The online coal ash detection system based on X-ray fluorescence analysis according to claim 4 is characterized in that: The ash content prediction model is constructed by the following method: S1. Collecting original X-ray fluorescence spectrum data of the coal test sample through the X-ray fluorescence spectrum acquisition module, and obtaining the coal composition of the coal test sample through X-ray fluorescence spectrum analysis; S2. Determine the ash value of the coal test sample corresponding to the original X-ray fluorescence spectrum data through coal industry analysis; S3, the spectral screening submodule uses the random forest algorithm to calculate the importance score of the relationship between each oxide and the ash value in the coal test sample according to the coal composition and the corresponding ash value, and then sorts them from high to low according to the importance score, taking all oxides ranked first M1 as the main related oxides, and taking all oxides ranked from M1+1 to M2 as the secondary related oxides; M1 and M2 are both positive integers, and M1<M2; The spectrum screening submodule further screens out spectrum data corresponding to each main relevant oxide from the original X-ray fluorescence spectrum data, which are recorded as the main relevant oxide spectrum data; and screens out spectrum data corresponding to each secondary relevant oxide from the original X-ray fluorescence spectrum data, which are recorded as the secondary relevant oxide spectrum data; S4, the spectrum screening submodule takes all the main relevant oxide spectrum data and all the secondary relevant oxide spectrum data as characteristic spectrum data, inputs them into the support vector regression model, and performs one-stage training. During the training process, the particle swarm algorithm is used to optimize the support vector regression model. The output end is the ash value of the coal test sample. After the training is completed, the initial ash prediction model is obtained; S5. The spectrum screening submodule takes all main relevant oxide spectrum data as characteristic spectrum data and inputs them into the initial support vector regression model for two-stage training. The output is the ash value of the coal test sample. After the training is completed, the final ash prediction model is obtained.

6. An online coal ash detection method based on X-ray fluorescence analysis, which uses the system as described in any one of claims 4-5, and the method comprises the following steps: Step 1: The coal to be tested is transported to the crushing chamber of the drying cylinder through the feed valve, and the coal powder crushed by the crushing module enters the drying chamber through the crushing discharge valve; Step 2: The air pump inputs the gas into the heater, and the heated gas enters the drying chamber through the air inlet valve to dry the coal powder in the drying chamber; Step 3: After the drying is completed, the drying discharge valve is opened and the exhaust valve is closed. The dried coal powder is input into the tablet pressing module by the gas pressure, and the coal test sample is obtained by pressing the tablet pressing module; then the drying discharge valve is closed and the exhaust valve is opened to discharge the residual coal powder in the drying chamber and then the exhaust valve is closed; Step 4: The conveying module conveys the coal test sample to the detection position of the coal ash analysis system; Step 5: The coal ash analysis system performs ash content detection on the coal test sample.

7. The on-line coal ash detection method based on X-ray fluorescence analysis according to claim 6 is characterized in that: The steps of the coal ash analysis system for ash detection include: Step 5-1, the X-ray fluorescence spectrum acquisition module acquires the original X-ray fluorescence spectrum data of the coal test sample; Step 5-2, the spectrum screening submodule screens out all the main relevant oxide spectrum data V1 and all the secondary relevant oxide spectrum data V2 from the original X-ray fluorescence spectrum data; Step 5-3, the spectrum data self-checking and adjustment submodule determines whether the currently collected X-ray fluorescence spectrum data can meet the analysis requirements based on the main relevant oxide spectrum data V1 and the secondary relevant oxide spectrum data V2: When it is determined that the analysis requirements are met, the verified characteristic spectrum data is output and the process goes to step 5-4; When it is determined that the analysis requirements are not met, the current analysis step is terminated, and the main control module controls the re-collection of original X-ray fluorescence spectrum data of the current coal test sample. If the re-collected original X-ray fluorescence spectrum data still cannot meet the analysis requirements, the coal test sample is re-prepared until the original X-ray fluorescence spectrum data that meets the analysis requirements is obtained; Step 5-4: input the verified characteristic spectrum data into the ash content prediction sub-model, and analyze and obtain the ash content value of the coal test sample.

8. The on-line coal ash detection method based on X-ray fluorescence analysis according to claim 7 is characterized in that: Step 5-3 is as follows: Step 5-3-1, the spectrum data self-checking and adjustment submodule counts the number of main relevant oxides corresponding to the main relevant oxide spectrum data V1, recorded as M'1, and counts the number of main relevant oxides corresponding to the secondary relevant oxide spectrum data V2, recorded as M'2; Step 5-3-2, if Determine whether the currently collected X-ray fluorescence spectrum data meets the analysis requirements, and directly output all the main relevant oxide spectrum data V1 as the verified characteristic spectrum data, and enter step 5-4; Step 5-3-3, if Determine whether the currently collected X-ray fluorescence spectrum data meets the analysis requirements, and combine all the main relevant oxide spectrum data V1 and all the secondary relevant oxide spectrum data V2 as the verified characteristic spectrum data output, and enter step 5-4; wherein α is a preset judgment threshold, and α<1; Step 5-3-4, if It is determined that the currently collected X-ray fluorescence spectrum data does not meet the analysis requirements, and the current analysis step is terminated. The main control module controls the re-collection of original X-ray fluorescence spectrum data of the current coal test sample. If the re-collected original X-ray fluorescence spectrum data still cannot meet the analysis requirements, the coal test sample is re-prepared until the original X-ray fluorescence spectrum data that meets the analysis requirements is obtained.

9. The on-line coal ash detection method based on X-ray fluorescence analysis according to claim 8, characterized in that: in, α=0.75-0.90, M1=4-7, M2=6-12, M1<M2.

10. The on-line coal ash detection method based on X-ray fluorescence analysis according to claim 9, characterized in that: in, α=0.8,M1=5,M2=10;Main related oxides include SiO2、Al2O3、CaO、TiO2、Fe2O3,Second related oxides include P4O 10 , K2O, SO2, Na2O, Mn3O4.

Images