Coal slime flotation control method based on ensemble learning and distributed mixed Bayesian network

Abstract

The invention discloses a coal slime flotation control method based on integrated learning and a distributed mixed Bayesian network. The coal slime flotation control method comprises the steps that firstly, offline modeling is conducted on the distributed mixed Bayesian network based on the integrated learning; s11, the coal slime flotation process is divided, and Bayesian network nodes are determined; s12, N corresponding mixed Bayesian network structures are determined; s13, determining parameters of N hybrid Bayesian networks by using a maximum likelihood estimation method; s14, completing distributed hybrid Bayesian network modeling based on integrated learning; 2, on-line safe operation of the coal slime flotation process is controlled; s21, judging whether an abnormal working condition occurs or not based on the current working condition data information, and making safety control decisions by using N hybrid Bayesian network control models; s22, integrating N decision results, and formulating a final safety control strategy of the current working condition; and S23, when the abnormal working condition is not eliminated, the step S21 is executed again. According to the method, a more reasonable and reliable safety control decision can be made.

Description

Technical Field

[0001] The invention belongs to the technical field of safe operation control of process industry, and specifically is a coal slime flotation control method based on integrated learning and distributed hybrid Bayesian network. Background Art

[0002] Coal slime flotation is an important coal processing technology. It combines heavy medium coal preparation technology with flotation technology. Through physical and chemical methods, it can efficiently separate impurities in coal slime from coal, improve the grade and quality of coal, remove harmful elements, improve combustion efficiency, and effectively reduce environmental pollution. At the same time, the coal slime flotation process can also recycle coal gangue and other valuable ores, realizing the sustainable utilization of resources. However, due to the interference of the actual production environment and the uncertainty of the quality of raw coal, the quality of coal produced by the coal slime flotation process may decline, and even abnormal operating conditions may occur. Therefore, it is necessary to design a timely and reliable safe operation control and product quality control method for the coal slime flotation process.

[0003] As an important probabilistic graphical model, Bayesian networks are widely used in fault diagnosis, reliability assessment, and prediction. In the production environment of coal slime flotation process, there are many uncertain factors, such as sensor errors and raw coal fluctuations. When abnormal working conditions occur, these factors bring certain difficulties to operators in making control decisions. In order to effectively solve this problem, Bayesian networks are widely used to model and process these uncertainties. Through probabilistic reasoning, Bayesian networks can accurately estimate and infer uncertain factors, thereby helping operators to make more reasonable control decisions. In addition, Bayesian networks can effectively combine expert knowledge and data information, represent the dependencies between process variables in coal slime flotation process through directed edges, and use conditional probability to describe the strength of the dependencies, graphically represent the interactions between variables, and improve the comprehensibility of control decisions. Therefore, it is urgent to provide a method for safe operation and product quality control of coal slime flotation process using Bayesian networks. Summary of the invention

[0004] In view of the problems existing in the above-mentioned prior art, the present invention provides a coal slime flotation control method based on ensemble learning and distributed hybrid Bayesian network. The method has a simple implementation process and low implementation cost. It can effectively combine the ensemble learning strategy to significantly improve the generalization ability of the distributed hybrid Bayesian network, and can formulate more reasonable and reliable safety control decisions by establishing multiple control models and weightedly fusing the decision results of each model.

[0005] In order to achieve the above-mentioned object, the present invention provides a coal slime flotation control method based on integrated learning and distributed hybrid Bayesian network, including a coal slime flotation industrial system, wherein the coal slime flotation industrial system includes a primary grading screen, a secondary grading screen, a conveyor, a mixing barrel, a pressure pump, a heavy medium cyclone, a first de-medium screen, a second de-medium screen, a qualified medium barrel, a circulation pump, a coal slime barrel, a curved screen, a thickener, a slurry preprocessor, a flotation tank and a controller;

[0006] The main grading screen is used to separate the raw coal into coarse coal and clean coal by screening; the feed port of the secondary grading screen is connected to the top discharge port of the main grading screen, and is used to remove coal slime impurities in the coal powder by screening; the feed end of the conveyor is connected to the top discharge port of the secondary grading screen, and is used to output the screened coal powder to the feed port of the mixing barrel; the feed port of the mixing barrel is connected to the discharge port of the conveyor, and is used to mix the clean coal with the heavy medium suspension to form a mixture; the inlet end of the pressure pump is connected to the discharge port of the mixing barrel through a pipeline, and is used to output the mixture to the heavy medium cyclone; the feed port of the heavy medium cyclone is connected to the outlet end of the pressure pump, and is used to utilize the density difference of the materials to The low-density material and the high-density material are separated to obtain overflow and underflow products; the feed port of the de-medium screen 1 is connected to the high-density discharge port of the heavy medium cyclone, which is used to dehydrate and de-mediumize the high-density material, and the high-density discharge port discharges the gangue; the feed port of the de-medium screen 2 is connected to the low-density discharge port of the heavy medium cyclone, which is used to dehydrate and de-mediumize the low-density material, and the high-density discharge port discharges the clean coal; the feed port of the qualified medium barrel is respectively connected to the low-density discharge port of the de-medium screen 1, the low-density discharge port of the de-medium screen 2, the magnetic medium filling pipeline and the circulating water pipeline filling pipeline, which are used to complete the preparation of the heavy medium suspension; the inlet end of the circulating pump is connected to the discharge port of the qualified medium barrel through a pipeline. The outlet of the concentrator is connected to the feed port of the mixing barrel through a pipeline, and its outlet end is connected to the feed port of the mixing barrel through a pipeline, so as to transport the heavy medium suspension into the mixing barrel; the feed port of the coal slime barrel is connected to the high-density discharge port of the de-medium screen 2, so as to perform preliminary treatment on the clean coal; the feed port of the curved screen is connected to the discharge port of the coal slime barrel, so as to control the particle size of the subsequent flotation coal slime by screening and sorting, and the top discharge port discharges coarse particles of clean coal; the feed port of the concentrator is connected to the bottom discharge port of the curved screen, so as to increase the concentration of the coal slime and reduce the moisture content of the coal slime, so that the pulp density is kept within the range of stable operation of the process; the feed port of the slurry preprocessor is respectively connected to the circulating water filling pipeline, the flotation agent filling pipeline and the concentrator The high-density discharge port of the machine is connected to fully mix the flotation reagent with the slurry, so as to make the final preparation for the flotation of coal slime foam; the feed port of the flotation tank is respectively connected to the air filling pipeline and the discharge port of the slurry preprocessor, so as to obtain the final clean coal in the overflow product of the flotation tank by utilizing the different hydrophilicity of different solid particles in the slurry, and the top discharge port outputs clean coal, and the bottom discharge port outputs impurities; the controller is respectively connected to the main grading screen, the secondary grading screen, the conveyor, the mixing barrel, the pressure pump, the heavy medium cyclone, the de-mediation screen 1, the de-mediation screen 2, the qualified medium barrel, the circulation pump, the coal slime barrel, the arc screen, the concentrator, the slurry preprocessor, and the flotation tank, so as to control each component;

[0007] The coal slime flotation control method comprises the following steps:

[0008] Step 1: Offline modeling of distributed hybrid Bayesian network based on ensemble learning;

[0009] S11: Divide the coal slime flotation process and determine the Bayesian network nodes;

[0010] According to the actual situation of the coal slime flotation industrial system and combined with operating experience, the coal slime flotation process is divided into multiple local modules, and the control variables, process variables and target variables of each local module are determined. Then, each variable is mapped to the corresponding Bayesian network node, and the level state of each node is further divided.

[0011] S12: Perform N stratified sampling on the training set to obtain sub-training sets D1, D2...D N , determine N corresponding hybrid Bayesian network structures;

[0012] A plurality of sensors installed at a plurality of sampling points in the coal slime flotation industrial system are used to collect a plurality of abnormal operating condition data information; a first Bayesian network model is used to make control decisions for the plurality of abnormal operating conditions, and the control decision reasoning accuracy is calculated according to formula (1); when the control decision reasoning accuracy reaches 98%, it is considered that the Bayesian network model can provide sufficiently accurate safety control decisions, and there is no need to integrate subsequent Bayesian network models; otherwise, the decision result of the second Bayesian network model is continuously integrated to further judge the control decision reasoning accuracy after integration, and so on and so forth, until the control decision reasoning accuracy made by the Nth Bayesian network model integrated is greater than or equal to 98%, and the number of models N in the integrated learning strategy is determined;

[0013]

[0014] In the formula, TP represents an effective safety control decision that can eliminate abnormal conditions; TN represents an invalid safety control decision that cannot eliminate abnormal conditions;

[0015] S13: Determine the parameters of N hybrid Bayesian networks using maximum likelihood estimation method;

[0016] S14: Complete the distributed hybrid Bayesian network modeling based on ensemble learning;

[0017] Step 2: Control the online safe operation of the coal slime flotation process based on distributed hybrid Bayesian network modeling;

[0018] S21: using a number of sensors installed at a number of sampling points in the coal slime flotation industrial system to collect current operating condition data information, and judging whether an abnormal operating condition occurs based on the current operating condition data information; if S 38.5%, it is judged that an abnormal operating condition occurs, and the N hybrid Bayesian network control models established in the offline modeling process are used to respectively formulate safety control decisions;

[0019] S22: Integrate N decision results and formulate the final safety control strategy for the current working condition;

[0020] S22-1: Calculate the BIC score between the current operating condition data and the i-th (i￡N) hybrid Bayesian network control model,

[0021] This is used to evaluate the degree of fit between the current working conditions and each model;

[0022] S22-2: Assign corresponding weights to each control decision according to the degree of fit of each model;

[0023] S22-3: All decisions are weighted and integrated according to formula (2) to form the final safety control decision Adjustment for the current working condition

[0024]

[0025] In the formula, α i represents the weight of the control decision provided by the i-th model, in, The BIC scoring function representing the i-th model and the current operating data is used to indicate the degree of fit between the i-th model and the current operating data; Adjustment i represents the safety control decision made by the i-th model;

[0026] S23: Predict the overflow ash content after adjustment. If S<8.5%, it proves that the abnormal operating condition has been eliminated, execute the current safety control decision, and maintain the current normal operating mode. Otherwise, re-execute S21 and continue to use the current operating condition data information Bayesian network reasoning process.

[0027] Aiming at the problems of product quality degradation and abnormal working conditions caused by working condition fluctuations in the actual production of the coal slime flotation process, the present invention proposes a coal slime flotation process safety control method based on ensemble learning and distributed hybrid Bayesian network. Bayesian network is widely used in the safe operation control of complex industrial processes due to its powerful uncertainty reasoning ability. However, due to the fluctuation of raw materials and changes in operating parameters in the operating environment of the industrial process, the generalization ability of the Bayesian network control model may be insufficient, and it is impossible to provide accurate safe operation control decisions. At the same time, coal slime flotation is a complex industrial process with a high degree of scale and multivariate interaction. Directly establishing a Bayesian network model for the entire process may reduce the accuracy of the model. In order to meet this challenge, the present invention provides an effective solution through distributed modeling. By decomposing the coal slime flotation process into multiple local modules, each local module is modeled according to its own needs and characteristics, and each local module is adjusted and optimized according to the actual situation, reducing the complexity of calculation and reasoning. When abnormal working conditions occur, these models each generate corresponding safety control decisions. Then, the BIC scoring function is used to measure the degree of fit between the current working conditions and each model, and the weight method is used to weight the decision results of each model to form a final safety control decision. The currently widely used discrete Bayesian network model needs to discretize the data when modeling, resulting in a large amount of information loss. Therefore, in the process of reasoning and control decision-making, the model can only provide a rough adjustment direction, and cannot provide operators with accurate adjustment strategies, nor can it further improve the product quality of the coal slime flotation process. In order to solve this problem, the Bayesian network effectively improves the reasoning accuracy by introducing continuous nodes.

[0028] The implementation process of this method is simple and the implementation cost is low. It can effectively combine the integrated learning strategy to significantly improve the generalization ability of the distributed hybrid Bayesian network, and can formulate more reasonable and reliable safety control decisions by establishing multiple control models and weightedly integrating the decision results of each model. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of the present invention;

[0030] Figure 2 is the control decision reasoning accuracy corresponding to different numbers of models of the present invention;

[0031] Figure 3 These are three hybrid Bayesian network structures of the present invention, where left: structure 1; middle: structure 2; right: structure 3;

[0032] Figure 4 is the conditional probability distribution of some continuous nodes in model 1 of the present invention;

[0033] Figure 5It is the contribution diagram of four abnormal working conditions of the present invention;

[0034] Figure 6 is the overflow ash change curve of working condition 1 of the present invention;

[0035] Figure 7 It is a structural schematic diagram of the coal slime flotation industrial system in the present invention. DETAILED DESCRIPTION

[0036] The present invention will be further described below in conjunction with the accompanying drawings.

[0037] like Figure 7As shown, the present invention provides a coal slime flotation control method based on integrated learning and distributed hybrid Bayesian network, including a coal slime flotation industrial system, the coal slime flotation industrial system including a main grading screen, a secondary grading screen, a conveyor, a mixing barrel, a pressure pump, a heavy medium cyclone, a first de-mediating screen, a second de-mediating screen, a qualified medium barrel, a circulating pump, a coal slime barrel, an arc screen, a concentrator, a slurry preprocessor, a flotation tank and a controller; the main grading screen is used to separate raw coal into coarse coal and clean coal by screening; the feed port of the secondary grading screen is connected to the top discharge port of the main grading screen, and is used to remove coal slime impurities in coal powder by screening; the feed end of the conveyor is connected to the top discharge port of the secondary grading screen, and is used to output the screened coal powder to The feed port of the mixing barrel is connected to the discharge port of the conveyor, and is used to mix the clean coal with the heavy medium suspension to form a mixture; the inlet end of the pressure pump is connected to the discharge port of the mixing barrel through a pipeline, and is used to output the mixture to the heavy medium cyclone; the feed port of the heavy medium cyclone is connected to the outlet end of the pressure pump, and is used to utilize the density difference of the materials to separate the low-density materials from the high-density materials to obtain overflow and underflow products; the feed port of the de-medium screen 1 is connected to the high-density discharge port of the heavy medium cyclone, and is used to dehydrate and de-mediumize the high-density materials, and the high-density discharge port discharges the gangue; the feed port of the de-medium screen 2 is connected to the low-density discharge port of the heavy medium cyclone, and is used to dehydrate and de-mediumize the high-density materials. Dehydration and de-mediation operations are carried out, and the clean coal is discharged from its high-density discharge port; the feed port of the qualified medium barrel is respectively connected to the low-density discharge port of the de-mediation screen one, the low-density discharge port of the de-mediation screen two, the magnetic medium filling pipeline and the circulating water pipeline filling pipeline, so as to complete the preparation of the heavy medium suspension; the inlet end of the circulating pump is connected to the discharge port of the qualified medium barrel through a pipeline, and its outlet end is connected to the feed port of the mixing barrel through a pipeline, so as to transport the heavy medium suspension to the mixing barrel; the feed port of the coal slime barrel is connected to the high-density discharge port of the de-mediation screen two, so as to carry out preliminary treatment of the clean coal; the feed port of the curved screen is connected to the discharge port of the coal slime barrel, so as to control the particle size of the subsequent flotation coal slime by screening and sorting, and its top discharge The coarse clean coal is discharged from the outlet of the concentrator; the feed port of the concentrator is connected to the bottom outlet of the curved screen, which is used to increase the concentration of the coal slime and reduce the moisture content of the coal slime, so that the pulp density is kept within the range of stable process operation; the feed port of the slurry preprocessor is respectively connected to the circulating water filling pipeline, the flotation agent filling pipeline and the high-density outlet of the concentrator, which is used to fully mix the flotation agent with the slurry and make the final preparations for the froth flotation of the coal slime; the feed port of the flotation cell is respectively connected to the air filling pipeline and the outlet of the slurry preprocessor, which is used to obtain the final clean coal in the overflow product of the flotation cell by utilizing the different hydrophilicity of different solid particles in the slurry, and the top outlet outputs the clean coal, and the bottom outlet outputs the impurities;The controller is respectively connected with the main grading screen, the secondary grading screen, the conveyor, the mixing barrel, the pressure pump, the heavy medium cyclone, the first de-medium screen, the second de-medium screen, the qualified medium barrel, the circulation pump, the coal slime barrel, the curved screen, the thickener, the slurry pre-processor, and the flotation tank, and is used to control each component;

[0038] like Figure 1 As shown, in order to improve the generalization ability of the distributed hybrid Bayesian network, combined with the integrated learning strategy, multiple control models are established, and the decision results of each model are weighted and integrated to make more reliable safety control decisions. Specifically, the coal slime flotation control method includes the following steps:

[0039] Step 1: Offline modeling of distributed hybrid Bayesian network based on ensemble learning;

[0040] S11: Divide the coal slime flotation process and determine the Bayesian network nodes;

[0041] According to the actual situation of the coal slime flotation industrial system and combined with operating experience, the coal slime flotation process is divided into multiple local modules, and the control variables, process variables and target variables of each local module are determined. Then, each variable is mapped to the corresponding Bayesian network node, and the level state of each node is further divided.

[0042] S12: Perform N stratified sampling on the training set to obtain sub-training sets D1, D2...D N , determine N corresponding hybrid Bayesian network structures;

[0043] As the working conditions of the coal slime flotation process fluctuate, the Bayesian network structure may change. In order to build a Bayesian network control model that is more suitable for the current working conditions and make more accurate control decisions, the present invention adopts a scoring search-based method to determine the Bayesian network structure.

[0044] Before establishing a hybrid Bayesian network, it is necessary to clarify the number of models N in the integrated learning strategy; use a number of sensors installed at a number of sampling points in the coal slime flotation industrial system to collect multiple sets of abnormal operating condition data information. As a preferred method, 500 sets of abnormal operating condition data information are collected, and then the required number of models N is determined by evaluating whether the control decision reasoning accuracy meets the established threshold; specifically, use the first Bayesian network model to make control decisions for multiple sets of abnormal operating conditions, and calculate the control decision reasoning accuracy according to formula (1). When the control decision reasoning accuracy reaches 98%, it is considered that the Bayesian network model can provide sufficiently accurate safety control decisions and there is no need to integrate subsequent Bayesian network models. Otherwise, continue to integrate the decision results of the second Bayesian network model to further determine the control decision reasoning accuracy after integration, and repeat this process until the control decision reasoning accuracy made by the Nth Bayesian network model integrated is greater than or equal to 98%, and then determine the number of models N in the integrated learning strategy;

[0045]

[0046] In the formula, TP represents an effective safety control decision that can eliminate abnormal conditions; TN represents an invalid safety control decision that cannot eliminate abnormal conditions;

[0047] S13: Determine the parameters of N hybrid Bayesian networks using maximum likelihood estimation method;

[0048] S14: Complete the distributed hybrid Bayesian network modeling based on ensemble learning;

[0049] Step 2: Control the online safe operation of the coal slime flotation process based on distributed hybrid Bayesian network modeling;

[0050] S21: using a number of sensors installed at a number of sampling points in the coal slime flotation industrial system to collect current operating condition data information, and judging whether an abnormal operating condition occurs based on the current operating condition data information; if S 38.5%, it is judged that an abnormal operating condition occurs, and the N hybrid Bayesian network control models established in the offline modeling process are used to respectively formulate safety control decisions;

[0051] S22: Integrate N decision results, use the weight method as the integrated strategy of control decision, and formulate the final safety control strategy for the current working condition;

[0052] S22-1: Calculate the BIC score between the current operating condition data and the i-th (i￡N) hybrid Bayesian network control model,

[0053] This is used to evaluate the degree of fit between the current working conditions and each model;

[0054] S22-2: Assign corresponding weights to each control decision according to the degree of fit of each model;

[0055] S22-3: All decisions are weighted and integrated according to formula (2) to form the final safety control decision Adjustment for the current working condition

[0056]

[0057] In the formula, α i represents the weight of the control decision provided by the i-th model, in, The BIC scoring function representing the i-th model and the current operating data is used to indicate the degree of fit between the i-th model and the current operating data; Adjustment i represents the safety control decision made by the i-th model;

[0058] S23: Predict the overflow ash content after adjustment. If S<8.5%, it proves that the abnormal operating condition has been eliminated, execute the current safety control decision, and maintain the current normal operating mode. Otherwise, re-execute S21 and continue to use the current operating condition data information Bayesian network reasoning process.

[0059] Example:

[0060] Coal resources are an important pillar of my country's economic and social development, providing key impetus and support for achieving sustainable and rapid economic growth. my country has abundant coal reserves, with production ranking first in the world and occupying an important proportion in the energy structure. However, raw coal contains a large amount of impurities, and direct combustion will release harmful gases, causing serious damage to the ecological environment. Therefore, in order to improve the utilization efficiency of coal and reduce the harm of combustion to the environment, coal preparation technology has become a vital link in the process of coal utilization. As an important coal preparation technology, coal slime flotation process is widely used because of its fine separation particle size and high separation accuracy. The coal slime flotation process first uses the Archimedes principle to wash the coarse coal with larger particle size, and then recovers the finer particle clean coal after washing for froth flotation. Finally, the overflow of the flotation tank is scraped out by the scraper device, dehydrated and dried to become the product clean coal, and the gangue that cannot float is left at the bottom of the flotation tank with the slurry and discharged as tailings. However, the actual production environment of coal slime flotation is harsh and the raw coal properties change frequently, which may cause the coal preparation process to deviate from the set optimal operating point, resulting in a decline in the quality of the product coal, or even abnormal working conditions, which not only affects the production performance and causes huge economic losses, but also may pose a serious threat to the service life of the equipment and the personal safety of the operators. When the current coal slime flotation process deviates from the optimal operating point, resulting in a decline in product quality or even abnormal working conditions, the operator mainly relies on the current working conditions and the use of previous knowledge and experience to judge the reasons for the deviation of the coal slime flotation process from the optimal operating point, and then make corresponding safety control decisions and quality control decisions. However, this knowledge-driven method is highly subjective and the decision is not accurate enough, which cannot ensure that the coal slime flotation process can be restored to the optimal operating point, and may even lead to the deepening of abnormal working conditions. In addition, the existing control methods for coal slime flotation process are mostly concentrated on model-driven and data-driven methods. The coal slime flotation process includes multiple complex links such as raw coal processing, heavy medium separation and froth flotation, and it is difficult to establish an accurate mechanism model. Therefore, it is difficult for the model-driven method to effectively make corresponding control decisions. With the increasing maturity of big data technology, a large amount of working condition data has been accumulated in the actual operation of the coal slime flotation process. These data can be used to establish a control model. However, the output results of the data-driven method are less interpretable and difficult for operators to understand. Therefore, it is necessary to design a more intelligent safety and quality integrated control method for the coal slime flotation process.

[0061] The data for this example comes from the full-process simulation platform of the coal slime flotation process (registration number: 2020SR1160052). In order to improve the interpretability of control decisions, the process variables involved in the coal slime flotation process are further considered, and the influence relationship between the control variables and the target variables is explained through a clearer graph network structure. Based on the modeling method formulated above, in order to provide more timely and accurate safety control decisions, it is first necessary to determine the number of models N in the integrated learning strategy. This embodiment selects 500 sets of abnormal operating data, and uses the established hybrid Bayesian network to make safety control decisions. When the number of models in the integrated learning is different, the corresponding control decision reasoning accuracy is as follows Figure 2 The results show that when the number of Bayesian network models N=3, the control decision reasoning accuracy reaches 98.7%, which meets the set accuracy threshold requirement, and as the number of models continues to increase, the control decision reasoning accuracy does not improve much. Therefore, the present invention sets the number of models N=3 in the integrated learning strategy.

[0062] The three hybrid Bayesian network model structures established by the present invention are as follows: Figure 3 The physical meaning of each node is shown in Table 1, the conditional probability distribution table of discrete nodes is shown in Table 2, and the conditional probability distribution table of continuous nodes is shown in Figure 4 As shown in the figure. The initial value of the control variable is represented as a discrete node. According to the expert knowledge and the influence of each control variable on the target variable, it can be divided into five levels: 1. low level, 2. lower level, 3. medium level, 4. higher level, 5. high level. In order to ensure the reasoning accuracy of the control decision, the control variable adjustment value, process variable and target variable are represented as continuous nodes.

[0063] Table 1: Bayesian network nodes and their physical meanings

[0064]

[0065]

[0066] Table 2: Conditional probability distribution table of nodes A, B, C, H, I, M, N, O in model 1

[0067]

[0068] Four possible abnormal working conditions are randomly listed, and the corresponding initial values ​​of the control variables and the target variables are shown in Table 3. Next, the online application strategy is used to make safety control decisions for these four abnormal working conditions. First, the contribution graph algorithm is used to determine the abnormal module. The contribution graphs of the four abnormal working conditions are shown in Figure 5As shown. Taking working condition 1 as an example, the abnormal module is the flotation module. Next, the BIC score of the current working condition data and the i-th (i≤N) Bayesian network model is calculated to further obtain the degree of fit between the current working condition and each model, and assign corresponding weights to each model control decision according to the degree of fit. The corresponding weights of each model under four abnormal working conditions are shown in Table 4.

[0069] Table 3: Four possible abnormal operating conditions

[0070]

[0071] Table 4: Weights corresponding to each model for four abnormal conditions

[0072]

[0073] The initial values ​​of the control variables of all modules and the overflow ash safety threshold S = 8.5% are used as evidence information. Each model uses Bayesian network reasoning to obtain the corresponding safety control decision, that is, the control variable adjustment value of the abnormal module. Combined with the weights of each model in Table 4, the final safety control decision is made according to formula (7). The results are shown in Table 5. Taking working condition 1 as an example, the physical meaning of the control decision represents that the aeration volume of the flotation tank increases by 0.067m 3 / m 2 *h, the motor stirring speed increases by 0.752rad / min, and the reagent addition amount increases by 1.889kg / t. The same is true for other working conditions.

[0074] In order to ensure the safe and stable operation of the coal slime flotation process, it is necessary to verify the obtained safety control decision and predict whether the overflow ash content after the execution of the decision meets the requirements. If S ≥ 8.5%, this decision will not be executed. After updating the working condition information, the Bayesian network reasoning process will continue; otherwise, this decision will be executed. The initial values ​​of the control variables of all modules and the adjusted values ​​of the control variables of the abnormal modules are used as evidence information, and the adjusted overflow ash content is inferred as shown in Table 6. The results show that the overflow ash content of all abnormal working conditions drops below the safety threshold, indicating that this method can effectively eliminate abnormal working conditions in the coal slime flotation process.

[0075] Table 5: Security Control Decisions

[0076]

[0077] Table 6: Overflow ash after implementing safety control decisions

[0078]

[0079] Taking working condition 1 as an example, the obtained safety control decision is applied to the coal slime flotation process simulation experimental platform. The overflow ash safety threshold is set to S = 8.5%, and the sampling interval is 1 second. The overflow ash change curve after executing the control decision is as follows: Figure 6 The operator collected the abnormal operating data information at the 943th moment as evidence, used Bayesian network reasoning to obtain the safety control decision, and after executing the control decision, the coal slime flotation process returned to normal at the 984th sampling point. Figure 6 The results show that the method proposed in this invention can effectively eliminate abnormal conditions in the coal slime flotation process and provide reliable safety control decisions. Figure 1 The results show that compared with the single Bayesian network model, the control decision made by the method proposed in the present invention has higher accuracy.

[0080] The present invention proposes a safe operation control method for coal slime flotation process based on ensemble learning and distributed hybrid Bayesian network. The method adopts the Bagging strategy in ensemble learning and constructs a hybrid Bayesian network model group by stratified sampling of the training set. When abnormal working conditions occur, these models each generate corresponding safety control decisions. Then, the BIC scoring function is used to measure the degree of fit between the current working conditions and each model, and the decision results of each model are weighted and fused using the weight method to form a final safety control decision. This process significantly improves the accuracy and robustness of the control decision. Finally, the effectiveness of the proposed method is verified through a coal slime flotation simulation experimental platform.

Claims

1. A coal slime flotation control method based on ensemble learning and distributed hybrid Bayesian network, comprising a coal slime flotation industrial system, wherein the coal slime flotation industrial system comprises a primary grading screen, a secondary grading screen, a conveyor, a mixing barrel, a pressure pump, a heavy medium cyclone, a first de-medium screen, a second de-medium screen, a qualified medium barrel, a circulation pump, a coal slime barrel, an arc screen, a thickener, a slurry pre-processor, a flotation tank and a controller; The main grading screen is used to separate the raw coal into coarse coal and clean coal by screening; the feed port of the secondary grading screen is connected to the top discharge port of the main grading screen, and is used to remove coal slime impurities in the coal powder by screening; the feed end of the conveyor is connected to the top discharge port of the secondary grading screen, and is used to output the screened coal powder to the feed port of the mixing barrel; the feed port of the mixing barrel is connected to the discharge port of the conveyor, and is used to mix the clean coal with the heavy medium suspension to form a mixture; the inlet end of the pressure pump is connected to the discharge port of the mixing barrel through a pipeline, and is used to output the mixture to the heavy medium cyclone; the feed port of the heavy medium cyclone is connected to the outlet end of the pressure pump, and is used to utilize the density difference of the materials to The low-density material and the high-density material are separated to obtain overflow and underflow products; the feed port of the de-medium screen 1 is connected to the high-density discharge port of the heavy medium cyclone, which is used to dehydrate and de-mediumize the high-density material, and the high-density discharge port discharges the gangue; the feed port of the de-medium screen 2 is connected to the low-density discharge port of the heavy medium cyclone, which is used to dehydrate and de-mediumize the low-density material, and the high-density discharge port discharges the clean coal; the feed port of the qualified medium barrel is respectively connected to the low-density discharge port of the de-medium screen 1, the low-density discharge port of the de-medium screen 2, the magnetic medium filling pipeline and the circulating water pipeline filling pipeline, which are used to complete the preparation of the heavy medium suspension; the inlet end of the circulating pump is connected to the discharge port of the qualified medium barrel through a pipeline. The outlet of the concentrator is connected to the feed port of the mixing barrel through a pipeline, and its outlet end is connected to the feed port of the mixing barrel through a pipeline, so as to transport the heavy medium suspension into the mixing barrel; the feed port of the coal slime barrel is connected to the high-density discharge port of the de-medium screen 2, so as to perform preliminary treatment on the clean coal; the feed port of the curved screen is connected to the discharge port of the coal slime barrel, so as to control the particle size of the subsequent flotation coal slime by screening and sorting, and the top discharge port discharges coarse particles of clean coal; the feed port of the concentrator is connected to the bottom discharge port of the curved screen, so as to increase the concentration of the coal slime and reduce the moisture content of the coal slime, so that the pulp density is kept within the range of stable operation of the process; the feed port of the slurry preprocessor is respectively connected to the circulating water filling pipeline, the flotation agent filling pipeline and the concentrator The high-density discharge port of the machine is connected to fully mix the flotation reagent with the slurry, so as to make the final preparation for the flotation of coal slime foam; the feed port of the flotation tank is respectively connected to the air filling pipeline and the discharge port of the slurry preprocessor, so as to obtain the final clean coal in the overflow product of the flotation tank by utilizing the different hydrophilicity of different solid particles in the slurry, and the top discharge port outputs clean coal, and the bottom discharge port outputs impurities; the controller is respectively connected to the main grading screen, the secondary grading screen, the conveyor, the mixing barrel, the pressure pump, the heavy medium cyclone, the de-mediation screen 1, the de-mediation screen 2, the qualified medium barrel, the circulation pump, the coal slime barrel, the arc screen, the concentrator, the slurry preprocessor, and the flotation tank, so as to control each component; It is characterized in that The coal slime flotation control method comprises the following steps: Step 1: Offline modeling of distributed hybrid Bayesian network based on ensemble learning; S11: Divide the coal slime flotation process and determine the Bayesian network nodes; According to the actual situation of the coal slime flotation industrial system and combined with operating experience, the coal slime flotation process is divided into multiple local modules, and the control variables, process variables and target variables of each local module are determined. Then, each variable is mapped to the corresponding Bayesian network node, and the level state of each node is further divided. S12: Perform N stratified sampling on the training set to obtain sub-training sets D1, D2...D N , determine N corresponding hybrid Bayesian network structures; A plurality of sensors installed at a plurality of sampling points in the coal slime flotation industrial system are used to collect a plurality of abnormal operating condition data information; a first Bayesian network model is used to make control decisions for the plurality of abnormal operating conditions, and the control decision reasoning accuracy is calculated according to formula (1); when the control decision reasoning accuracy reaches 98%, it is considered that the Bayesian network model can provide sufficiently accurate safety control decisions, and there is no need to integrate subsequent Bayesian network models; otherwise, the decision result of the second Bayesian network model is continuously integrated to further judge the control decision reasoning accuracy after integration, and so on and so forth, until the control decision reasoning accuracy made by the Nth Bayesian network model integrated is greater than or equal to 98%, and the number of models N in the integrated learning strategy is determined; In the formula, TP represents an effective safety control decision that can eliminate abnormal conditions; TN represents an invalid safety control decision that cannot eliminate abnormal conditions; S13: Determine the parameters of N hybrid Bayesian networks using maximum likelihood estimation method; S14: Complete the distributed hybrid Bayesian network modeling based on ensemble learning; Step 2: Control the online safe operation of the coal slime flotation process based on distributed hybrid Bayesian network modeling; S21: using a number of sensors installed at a number of sampling points in the coal slime flotation industrial system to collect current operating condition data information, and judging whether an abnormal operating condition occurs based on the current operating condition data information; if S 38.5%, it is judged that an abnormal operating condition occurs, and the N hybrid Bayesian network control models established in the offline modeling process are used to respectively formulate safety control decisions; S22: Integrate N decision results and formulate the final safety control strategy for the current working condition; S22-1: Calculate the BIC score between the current operating condition data and the i-th (i￡N) hybrid Bayesian network control model to evaluate the degree of fit between the current operating condition and each model; S22-2: Assign corresponding weights to each control decision according to the degree of fit of each model; S22-3: All decisions are weighted and integrated according to formula (2) to form the final safety control decision Adjustment for the current working condition In the formula, α i represents the weight of the control decision provided by the i-th model, in, The BIC scoring function representing the i-th model and the current operating data is used to indicate the degree of fit between the i-th model and the current operating data; Adjustment i represents the safety control decision made by the i-th model; S23: Predict the overflow ash content after adjustment. If S<8.5%, it proves that the abnormal operating condition has been eliminated, execute the current safety control decision, and maintain the current normal operating mode. Otherwise, re-execute S21 and continue to use the current operating condition data information Bayesian network reasoning process.

Images