An online soft measurement method for clean coal ash content in the process of lump coal shallow trough sorting

Abstract

The present invention discloses an online soft measurement method for clean coal ash in a shallow trough sorting process of lump coal, relates to the technical field of coal sorting and processing, and solves the problems of complexity, poor adaptability and high cost of the existing clean coal ash prediction method in the shallow trough sorting process of lump coal. The method comprises: real-time data collection, including clean coal output, tail coal output, qualified medium density and dual-energy X-ray image of clean coal; inputting the R value of each pixel point in the dual-energy X-ray image of clean coal collected in real time into a trained clean coal ash prediction main model, and obtaining the clean coal ash prediction value after processing; processing the real-time collected clean coal output and tail coal output to obtain the clean coal yield; and inputting the clean coal yield and the qualified medium density after normalization into the trained clean coal ash compensation model, and obtaining the clean coal ash error prediction value after processing; using the clean coal ash error prediction value to compensate the clean coal ash prediction value to obtain the final clean coal ash prediction result.

Description

Technical Field

[0001] The invention relates to the technical field of coal sorting and processing, and in particular to an online soft measurement method for clean coal ash content in a shallow trough sorting process of lump coal. Background Art

[0002] With the continuous advancement of the "dual carbon" strategy, the low-carbon and clean utilization of coal resources has become the top priority of the energy economy. However, the traditional shallow trough sorting and detection technology of lump coal has serious lags and cannot guide the adjustment of production indicators in a timely manner, resulting in a decrease in product quantity and a large loss of coal resources.

[0003] Shallow trough sorting of lump coal is often used for pre-discharging of gangue and classification. The existing sorting process is: the raw coal enters the shallow trough sorting machine through the raw coal belt, and the two products obtained by sorting are dehydrated and de-intermediated, and enter the clean coal belt and tail coal belt to become clean coal products and tail coal products. The existing clean coal ash content detection method is sample preparation class testing, which has serious lag, and the operating status of production equipment, production process parameters and product quality cannot be perceived in real time. Quality is often guaranteed by losing output, resulting in a large amount of clean coal lost in tail coal, resulting in low ash content, high calorific value and waste of coal resources.

[0004] Therefore, there is an urgent need to develop a simple, widely adaptable, low-cost online prediction method for clean coal ash content in the shallow trough sorting process of lump coal. Summary of the invention

[0005] In view of the above analysis, an embodiment of the present invention aims to provide an online soft measurement method for clean coal ash content in a shallow trough sorting process of lump coal, so as to solve the problems of complexity, poor adaptability and high cost of existing clean coal ash content prediction methods in a shallow trough sorting process of lump coal.

[0006] The present invention discloses an online soft measurement method for clean coal ash content in a lump coal shallow trough sorting process, comprising:

[0007] Real-time data collection, including clean coal production, tail coal production, qualified medium density and dual-energy X-ray images of clean coal;

[0008] The R value of each pixel point in the real-time collected dual-energy X-ray image of clean coal is input into the trained clean coal ash content prediction main model, and the clean coal ash content prediction value is obtained after processing;

[0009] Processing the real-time collected clean coal output and tail coal output to obtain the clean coal yield; and inputting the clean coal yield and the normalized qualified medium density into the trained clean coal ash compensation model to obtain the clean coal ash error prediction value after processing;

[0010] The clean coal ash content prediction value is compensated by the clean coal ash content error prediction value to obtain a final clean coal ash content prediction result.

[0011] On the basis of the above scheme, the present invention also makes the following improvements:

[0012] Further, the clean coal ash content prediction main model is trained in the following manner:

[0013] Acquire a first sample set, wherein each group of first sample data in the first sample set includes: a dual-energy X-ray image of clean coal and a measured value of ash content of the clean coal;

[0014] The R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of first sample data is used as input and the actual measured value of the clean coal ash content is used as a label, the clean coal ash content prediction main model is trained, the structure and parameters of the clean coal ash content prediction main model are determined, and the trained clean coal ash content prediction main model is obtained.

[0015] Further, the clean coal ash compensation model is trained in the following manner:

[0016] Acquire a second sample set, wherein each set of second sample data in the second sample set includes: clean coal output, tail coal output, qualified medium density, dual-energy X-ray image of clean coal, and measured value of clean coal ash content;

[0017] Processing the clean coal output and tail coal output in each group of second sample data to obtain the clean coal yield; inputting the R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of second sample data into the trained clean coal ash content prediction main model to obtain the clean coal ash content prediction value;

[0018] The clean coal yield corresponding to each group of second sample data and the normalized qualified medium density are used as input, and the difference between the corresponding clean coal ash measured value and the clean coal ash predicted value is used as a label to train the clean coal ash compensation model to obtain a trained clean coal ash compensation model.

[0019] Furthermore, each set of second sample data satisfies the following requirements: the deviation between the theoretical value of clean coal ash content obtained by matching the clean coal output, tail coal output and the raw coal selectivity curve and the actual measured value of the clean coal ash content of the current second sample data does not exceed the set deviation threshold.

[0020] Furthermore, the method further comprises:

[0021] During the real-time data collection process, the actual measured value of clean coal ash content is also collected regularly;

[0022] If the error between the regularly collected actual measured value of clean coal ash content and the final clean coal ash content prediction result obtained based on the actual collected data at that moment is lower than the set error, the actual collected data at that moment and the actual measured value of clean coal ash content are used as a set of correction data, and multiple sets of correction data form a correction data set, and the clean coal ash compensation model is corrected online based on the correction data set.

[0023] Furthermore, the R value of each pixel in the dual-energy X-ray image of the clean coal is:

[0024]

[0025] Among them, μ l represents the attenuation coefficient of the pixel point in the dual-energy X-ray image under low-energy X-ray, μ h It represents the attenuation coefficient of the pixel in the dual-energy X-ray image under high-energy X-ray.

[0026] Further, the clean coal output and tail coal output collected in real time are processed to obtain the clean coal yield, including:

[0027]

[0028] Among them, r1 represents the clean coal yield, Q1 represents the clean coal output, and Q2 represents the tail coal output.

[0029] Furthermore, the main model for predicting clean coal ash content is implemented based on a convolutional neural network.

[0030] Furthermore, the clean coal ash compensation model is implemented based on least squares support vector regression.

[0031] Further, the clean coal output is measured by a clean coal belt scale;

[0032] The tail coal output is measured by a tail coal belt scale;

[0033] The density of the qualified medium is measured by a density meter in a qualified medium barrel;

[0034] The dual-energy X-ray image of the clean coal is acquired by an industrial X-ray machine arranged above the clean coal belt.

[0035] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0036] The present invention discloses an online soft measurement method for clean coal ash content in a lump coal shallow trough sorting process, which has the following advantages:

[0037] First, the clean coal ash content prediction value is determined based on the R value of each pixel point in the dual-energy X-ray image of the clean coal collected in real time, and the clean coal ash content error prediction value is determined based on the clean coal output, tail coal output and qualified medium density collected in real time. Then, the clean coal ash content prediction value is compensated by the clean coal ash content error prediction value to obtain the final clean coal ash content prediction result; this method can realize continuous measurement, and can realize the effective, accurate and rapid detection of clean coal ash content in the process of shallow trough sorting of lump coal, so as to guide the setting of feed amount, circulating suspension amount and suspension density based on the prediction result, reduce system fluctuations and increase clean coal output.

[0038] Second, it can realize real-time online continuous detection of the ash content of the sorted clean coal, and has a certain degree of self-adaptation, and can correct the predicted value according to the production situation. According to the predicted value, the sorting working condition can be adjusted in time to meet the new working conditions.

[0039] In the present invention, the above-mentioned technical solutions can also be combined with each other to achieve more preferred combination solutions. Other features and advantages of the present invention will be described in the subsequent description, and some advantages can become obvious from the description, or can be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the contents particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The drawings are only for the purpose of illustrating particular embodiments and are not to be considered limiting of the present invention. Like reference symbols denote like components throughout the drawings.

[0041] Figure 1 It is a schematic diagram of the online soft measurement process of clean coal ash content in the shallow trough sorting process of lump coal disclosed in an embodiment of the present invention;

[0042] Figure 2 A schematic diagram of the training process of the main model for predicting clean coal ash content in the sorting process provided by an embodiment of the present invention;

[0043] Figure 3 A schematic diagram of the training process of the clean coal ash content prediction compensation model for the sorting process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not used to limit the scope of the present invention.

[0045] The embodiment of the present invention discloses an online soft measurement method for clean coal ash content in a shallow trough separation process of lump coal. The flow chart is as follows: Figure 1 As shown, including:

[0046] Step S1: real-time data collection, including clean coal output, tail coal output, qualified medium density and dual-energy X-ray images of clean coal;

[0047] Step S2: inputting the R value of each pixel point in the real-time collected dual-energy X-ray image of clean coal into the trained clean coal ash content prediction main model, and obtaining the clean coal ash content prediction value after processing;

[0048] Step S3: Processing the real-time collected clean coal output and tail coal output to obtain the clean coal yield; and inputting the clean coal yield and the normalized qualified medium density into the trained clean coal ash compensation model to obtain the clean coal ash error prediction value after processing;

[0049] Step S4: compensating the clean coal ash content prediction value with the clean coal ash content error prediction value to obtain a final clean coal ash content prediction result.

[0050] Specifically, the data in this embodiment are obtained in the following ways: the clean coal output is measured by a clean coal belt scale; the tail coal output is measured by a tail coal belt scale; the qualified medium density is measured by a densitometer in a qualified medium barrel; the dual-energy X-ray image of the clean coal is acquired by an industrial X-ray machine arranged above the clean coal belt.

[0051] There is a direct relationship between the R value of the dual-energy X-ray image of clean coal and the prediction of clean coal ash content. Specifically, the R value of the material attribute is related to the effective atomic number of the material. The R value of each pixel in the dual-energy X-ray image of clean coal is calculated according to formula (1):

[0052]

[0053] Among them, μ l represents the attenuation coefficient of the pixel point in the dual-energy X-ray image under low-energy X-ray, μ h Represents the attenuation coefficient of the pixel point in the dual-energy X-ray image under high-energy X-ray;

[0054] The Z value of the material property satisfies:

[0055] Z = -6.596 × 10 5 ×e -9.815R +4.685×e 0.6783R (2)

[0056] The relationship between Z value and material classification is shown in Table 1:

[0057] Table 1 Relationship between Z value and substance classification

[0058]

[0059] In the clean coal obtained by sorting, the higher the clean coal ash content, the more inorganic matter (such as gangue, etc.) it contains; conversely, the lower the clean coal ash content, the more organic matter it contains; therefore, there is a direct relationship between the R value of each pixel point in the dual-energy X-ray image of the clean coal and the clean coal ash content, and the clean coal ash content can be predicted based on the R value of each pixel point in the dual-energy X-ray image of the clean coal.

[0060] In addition, the clean coal yield and qualified medium density also have different degrees of influence on the prediction of clean coal ash content, and the prediction results of clean coal ash content can be compensated based on the clean coal yield and qualified medium density.

[0061] Before implementing the above scheme, the training of the main model for clean coal ash content prediction and the compensation model for clean coal ash content prediction must be completed. The specific implementation method is introduced as follows:

[0062] (1) The clean coal ash content prediction main model is trained in the following manner. The training process is as follows: Figure 2 As shown:

[0063] Step A1: obtaining a first sample set, wherein each group of first sample data in the first sample set includes: a dual-energy X-ray image of clean coal and a measured value of ash content of the clean coal;

[0064] Step A2: Take the R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of first sample data as input and the measured value of the clean coal ash content as a label, train the main model for predicting the clean coal ash content, determine the structure and parameters of the main model for predicting the clean coal ash content, and obtain the trained main model for predicting the clean coal ash content.

[0065] In the process of training the main model for clean coal ash content prediction, a mapping relationship between the R value of each pixel in the dual-energy X-ray image of clean coal and the measured value of clean coal ash content can be established. This mapping relationship is reflected by the structure and parameters of the main model for clean coal ash content prediction. Therefore, in the real-time processing process, the R value of each pixel in the real-time collected dual-energy X-ray image of clean coal can be input into the trained main model for clean coal ash content prediction, and the clean coal ash content prediction value can be obtained after processing.

[0066] Preferably, the main model for clean coal ash content prediction uses a main model for clean coal ash content prediction based on a convolutional neural network. During the training process of the main model for clean coal ash content prediction, the convolutional neural network back-propagates the error obtained by gradient descent, updates the parameters of each layer of the convolutional neural network layer by layer, and finally determines the structure and parameters of the main model for clean coal ash content prediction after multiple rounds of iterative training.

[0067] The main model for clean coal ash prediction based on convolutional neural network can make a preliminary prediction of the clean coal ash content, but it is unable to obtain high-precision prediction results due to the complex sorting conditions, the error in the dual-energy X-ray image acquisition process, and the limited information reflected by the image. At the same time, other influencing factors of clean coal ash content were also analyzed above. Therefore, according to these influencing factors of clean coal ash content, the clean coal ash content prediction compensation model is trained to predict the deviation of the clean coal ash content prediction value of the main model for clean coal ash content prediction.

[0068] (2) The clean coal ash content prediction and compensation model is trained in the following manner: Figure 3 As shown:

[0069] Step B1: obtaining a second sample set, wherein each set of second sample data in the second sample set includes: clean coal output, tail coal output, qualified medium density, dual-energy X-ray image of clean coal, and measured value of clean coal ash content;

[0070] Step B2: Process the clean coal output and tail coal output in each group of second sample data to obtain the clean coal yield; specifically,

[0071]

[0072] Among them, r1 represents the clean coal yield, Q1 represents the clean coal output, and Q2 represents the tail coal output.

[0073] Step B3: inputting the R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of second sample data into the trained clean coal ash content prediction main model to obtain the clean coal ash content prediction value;

[0074] Step B4: Take the clean coal yield and the normalized qualified medium density corresponding to each group of second sample data as input, take the difference between the corresponding clean coal ash measured value and the clean coal ash predicted value as a label, train the clean coal ash compensation model, and obtain a trained clean coal ash compensation model. Preferably, the maximum and minimum normalization method is used to normalize the qualified medium density. After normalization, all data are converted to the [0,1] interval so that each indicator belongs to the same order of magnitude.

[0075] Exemplarily, the clean coal ash compensation model is implemented based on least squares support vector regression.

[0076] It should be noted that in order to ensure the accuracy of the data for the training model, each set of second sample data meets the following requirements: the deviation between the theoretical value of clean coal ash content obtained by matching the clean coal output, tail coal output and the raw coal selectivity curve and the actual measured value of the clean coal ash content of the current second sample data does not exceed the set deviation threshold.

[0077] Preferably, the method in this embodiment further includes:

[0078] During the real-time data collection process, the actual measured value of clean coal ash content is also collected regularly;

[0079] If the error between the regularly collected actual value of clean coal ash content and the final clean coal ash content prediction result obtained based on the actual collected data at that moment is lower than the set error, the actual collected data at that moment and the actual measured value of clean coal ash content are used as a set of correction data, and multiple sets of correction data form a correction data set. The clean coal ash content compensation model is corrected online based on the correction data set to achieve self-correction of model parameters and enhance the model's adaptability under various working conditions.

[0080] In summary, compared with the prior art, the online soft measurement method for clean coal ash in the shallow trough sorting process of lump coal provided in this embodiment has the following advantages: First, the clean coal ash prediction value is determined based on the R value of each pixel point in the dual-energy X-ray image of the clean coal collected in real time, and the clean coal ash error prediction value is determined based on the real-time collected clean coal output, tail coal output and qualified medium density, and then the clean coal ash error prediction value is used to compensate the clean coal ash prediction value to obtain the final clean coal ash prediction result; this method can realize continuous measurement, and can realize effective, accurate and rapid detection of clean coal ash in the process of lump coal shallow trough sorting, so as to guide the setting of feed amount, circulating suspension amount and suspension density based on the prediction result, reduce system fluctuations, and increase clean coal output. Second, it can realize real-time online continuous detection of clean coal ash, and has a certain adaptive ability, and can correct the prediction value according to production conditions. According to the prediction value, the sorting condition can be adjusted in time to meet the new working conditions.

[0081] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0082] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. An online soft measurement method for clean coal ash content in the process of lump coal shallow trough sorting, characterized in that: include: Real-time data collection, including clean coal production, tail coal production, qualified medium density and dual-energy X-ray images of clean coal; The R value of each pixel point in the real-time collected dual-energy X-ray image of clean coal is input into the trained clean coal ash content prediction main model, and the clean coal ash content prediction value is obtained after processing; Processing the real-time collected clean coal output and tail coal output to obtain the clean coal yield; and inputting the clean coal yield and the normalized qualified medium density into the trained clean coal ash compensation model to obtain the clean coal ash error prediction value after processing; Compensating the clean coal ash content prediction value with the clean coal ash content error prediction value to obtain a final clean coal ash content prediction result; The R value of each pixel in the dual-energy X-ray image of the clean coal: Among them, μ l represents the attenuation coefficient of the pixel point in the dual-energy X-ray image under low-energy X-ray, μ h It represents the attenuation coefficient of the pixel in the dual-energy X-ray image under high-energy X-ray.

2. The online soft measurement method for clean coal ash content in the shallow trough sorting process of lump coal according to claim 1 is characterized in that: The clean coal ash content prediction main model is trained in the following way: Acquire a first sample set, wherein each group of first sample data in the first sample set includes: a dual-energy X-ray image of clean coal and a measured value of ash content of the clean coal; The R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of first sample data is used as input and the actual measured value of the clean coal ash content is used as a label, the clean coal ash content prediction main model is trained, the structure and parameters of the clean coal ash content prediction main model are determined, and the trained clean coal ash content prediction main model is obtained.

3. The online soft measurement method for clean coal ash content in the shallow trough sorting process of lump coal according to claim 1 is characterized in that: The clean coal ash compensation model is trained in the following way: Acquire a second sample set, wherein each set of second sample data in the second sample set includes: clean coal output, tail coal output, qualified medium density, dual-energy X-ray image of clean coal, and measured value of clean coal ash content; Processing the clean coal output and tail coal output in each group of second sample data to obtain the clean coal yield; Inputting the R value of each pixel point in the dual-energy X-ray image of the clean coal in each group of second sample data into the trained clean coal ash content prediction main model to obtain the clean coal ash content prediction value; The clean coal yield corresponding to each group of second sample data and the normalized qualified medium density are used as input, and the difference between the corresponding clean coal ash measured value and the clean coal ash predicted value is used as a label to train the clean coal ash compensation model to obtain a trained clean coal ash compensation model.

4. The online soft measurement method for clean coal ash content in the shallow trough sorting process of lump coal according to claim 3 is characterized in that: Each set of second sample data meets the following requirements: the deviation between the theoretical value of clean coal ash content obtained by matching the clean coal output, tail coal output and the raw coal selectivity curve and the actual measured value of the clean coal ash content of the current second sample data does not exceed the set deviation threshold.

5. The online soft measurement method for clean coal ash content in the shallow trough sorting process of lump coal according to claim 3 is characterized in that: The method further comprises: During the real-time data collection process, the actual measured value of clean coal ash content is also collected regularly; If the error between the regularly collected actual measured value of clean coal ash content and the final clean coal ash content prediction result obtained based on the actual collected data at that moment is lower than the set error, the actual collected data at that moment and the actual measured value of clean coal ash content are used as a set of correction data, and multiple sets of correction data form a correction data set, and the clean coal ash compensation model is corrected online based on the correction data set.

6. The online soft measurement method for clean coal ash content in the shallow trough separation process of lump coal according to any one of claims 1 to 5, characterized in that: The processing of the clean coal output and tail coal output collected in real time to obtain the clean coal yield includes: Among them, r1 represents the clean coal yield, Q1 represents the clean coal output, and Q2 represents the tail coal output.

7. The online soft measurement method for clean coal ash content in the shallow trough separation process of lump coal according to any one of claims 1 to 5, characterized in that: The main model for predicting clean coal ash content is implemented based on a convolutional neural network.

8. The online soft measurement method for clean coal ash content in the shallow trough separation process of lump coal according to any one of claims 1 to 5, characterized in that: The clean coal ash compensation model is implemented based on least squares support vector regression.

9. The online soft measurement method for clean coal ash content in the shallow trough separation process of lump coal according to any one of claims 1 to 5, characterized in that: The clean coal output is measured by a clean coal belt scale; The tail coal output is measured by a tail coal belt scale; The density of the qualified medium is measured by a density meter in a qualified medium barrel; The dual-energy X-ray image of the clean coal is acquired by an industrial X-ray machine arranged above the clean coal belt.

Images