Coal slime flotation process safety control method based on active learning and Bayesian network

Abstract

The invention discloses a coal slime flotation process safety control method based on active learning and a Bayesian network. The method comprises the following steps: analyzing mechanisms and operation experience of common main abnormal working conditions in a coal slime flotation process; determining a reason variable, a phenomenon variable and a control variable related to the two main abnormal working conditions; determining Bayesian network node division and node state levels; performing parameter learning and structure learning of the Bayesian network to obtain a final Bayesian network; when an abnormal working condition occurs, adding phenomenon variable data as evidence into the Bayesian network model for reasoning; making a corresponding safe operation control decision; and executing a control decision, and determining whether the abnormal working condition disappears in the production process. The method is simple in step and high in reliability, an operator can be assisted in making a response in time when the abnormal working condition occurs, a corresponding control decision is made to eliminate the abnormal working condition, safe and stable operation of the industrial process can be guaranteed, and automatic decision support can be provided for the abnormal working condition in the flotation process.

Description

Technical Field

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

[0002] Coal is the most widely distributed and largest fossil energy in the world, and is an important foundation for the rapid development of economy and industry. With the development and progress of information and automation technology, the coal preparation process has received more and more attention. The washing and processing of raw coal refers to the selection of raw coal by various methods such as physics and chemistry, the separation of impurities such as gangue and sulfur in raw coal, and the processing of the separated clean coal to obtain finished coal of different qualities and different uses. At present, the widely used coal preparation technologies in my country are jigging coal preparation, heavy medium coal preparation and flotation coal preparation. Among them, jigging coal preparation is often used to process easy-to-select coal, with a simple process flow and easy equipment maintenance, but it is difficult to achieve when encountering difficult-to-select coal. The heavy medium coal preparation process has been widely used in coal preparation plants due to its strong adaptability to coal quality and high separation efficiency. However, in the actual production process, due to the complex operating environment, low degree of equipment automation, poor condition of coal preparation equipment and other reasons, a series of abnormal conditions such as poor separation and abnormal operation are prone to occur. Now, with the continuous exploitation of domestic coal resources, high-quality coal resources are reduced, and low-quality coal resources have begun to receive attention. The coal slime flotation process is a coal preparation method that mainly utilizes the different wettability of tiny coal particles and useless mineral surfaces to water. The hydrophobic coal particles float up with the bubbles, while the hydrophilic mineral impurities remain at the bottom, thereby separating the coal from the mineral impurities. The coal slime flotation process can not only sort fine-grained and ultra-fine-grained raw coal and low-quality coal, but also efficiently recover the residual coal slime in the mineral processing cycle to effectively reduce the loss of resources in the production process, and the clean coal obtained by flotation has low ash content and high yield. Now various industries have higher and higher requirements for fine coal and clean coal, and the advantages of the coal slime flotation process are gradually emerging.

[0003] At present, in the actual production process of coal slime flotation process, the operator's operating experience is usually relied on to judge whether there is an abnormal condition. First, the type of abnormal condition is judged by observing the parameter changes of coal preparation equipment, and then the cause of the abnormal condition is analyzed. Finally, the corresponding safety control plan is formulated to eliminate the abnormal condition and maintain the parameters of the production process in a normal and stable state. This method of relying on the operator's experience and expert knowledge to discover and eliminate abnormal conditions has great disadvantages. When the operator is inexperienced and does not have a deep understanding of the abnormal condition, the cause of the abnormal condition cannot be quickly judged in the face of sudden abnormal conditions, and then the correct treatment cannot be made in time, which is easy to cause large economic losses. At the same time, due to the complexity of the environment and the mutual coupling between variables, the difficulty of manual process analysis is increased. When multiple abnormal conditions occur at the same time, the operator may not be able to make a perfect safety control plan due to his own limitations. Therefore, in order to improve the efficiency and reliability of decision-making, it is necessary to design a more intelligent safety operation control method to eliminate abnormal conditions in time, improve production efficiency and product quality, and reduce economic losses. Summary of the invention

[0004] In view of the problems existing in the above-mentioned prior art, the present invention provides a safety control method for a coal slime flotation process based on active learning and Bayesian networks. The method has simple steps and high reliability. It can combine data information and expert knowledge in the coal slime flotation industrial process, and use a small amount of abnormal operating condition data information to establish a more accurate safety control Bayesian network model. It can assist operators to respond in time when abnormal conditions occur, formulate corresponding control decisions to eliminate abnormal conditions, ensure the safe and stable operation of the industrial process, and provide automatic decision-making support for abnormal conditions in the flotation process.

[0005] In order to achieve the above-mentioned object, the present invention provides a method for safe control of a coal slime flotation process based on active learning and 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 circulating pump, a coal slime barrel, an arc screen, a concentrator, a slurry pre-processor, 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 safety control method of coal slime flotation process includes the following steps:

[0008] Step 1: Analyze the mechanism and operating experience of the main common abnormal conditions in the coal slime flotation process;

[0009] Combined with the actual situation of the coal slime flotation industrial system, the two main processing links of the coal slime flotation operation and the two main abnormal working conditions corresponding to the two main processing links are determined; the two main processing links are the heavy medium coal preparation link and the froth flotation processing link, and the two main abnormal working conditions are the enlargement of the bottom flow port diameter of the heavy medium cyclone and the low inlet pressure of the heavy medium cyclone;

[0010] Step 2: Determine the cause variables, phenomenon variables and control variables related to the two main abnormal conditions;

[0011] Conduct in-depth analysis of the coal slime flotation process to determine the cause variables, phenomenon variables and control variables that cause the abnormal operating condition of the heavy medium cyclone bottom flow port diameter to increase, and at the same time, determine the cause variables, phenomenon variables and control variables that cause the heavy medium cyclone inlet pressure to be too low;

[0012] Step 3: Determine the Bayesian network node division and node status level;

[0013] Based on the determined control variables, cause variables and phenomenon variables, simulation experiments were conducted on two main abnormal operating conditions: the enlarged diameter of the bottom flow port of the heavy medium cyclone and the too low inlet pressure of the heavy medium cyclone. A safety control Bayesian network model was established, and then the nodes of the Bayesian network and the state level of each node were determined.

[0014] Step 4: Perform parameter learning and structure learning of the Bayesian network;

[0015] S41: Based on the determined nodes of the Bayesian network and the production data of the coal slime flotation industrial system, the skeleton graph G is obtained using the conditional independence test ug ; In the structure learning of Bayesian network, mutual information and conditional mutual information are used to determine whether there is a dependency relationship between different variables, and then determine the edges between network nodes, where mutual information I(X;Y) is calculated according to formula (1);

[0016]

[0017] Where p(x,y) is the joint probability distribution of variables X and Y, p(x) is the marginal probability distribution of variable X, p(y) is the marginal probability distribution of variable Y, and x and y are the values ​​of variables X and Y respectively;

[0018] S42: Determine the skeleton graph G using the IGCI algorithm ug The direction of the edges between the nodes is obtained to obtain the initial Bayesian network structure G Init; Use formula (2) to determine the causal relationship C between variables x and y by calculating the density loss value between the marginal probability distributions p(x) and p(y) of variables x and y x→y ;

[0019] C x→y =H(y)-H(x) (2);

[0020] Where H(x) and H(y) are the information entropy of variables x and y respectively; when C x→y <0, the causal relationship between variables x and y is x→y; when C x→y > 0, the causal relationship between variables x and y is y→x;

[0021] S43: Calculate each experimental action Ac using the utility function in formula (3) i For the initial Bayesian network structure G Init The value of the effect, and find the experimental action Ac with the maximum utility function value best ;

[0022]

[0023] In the formula, P(y|G Init ,Ac i ,D) is the initial Bayesian network structure graph G Init Execute action Ac i The probability density of the variable y after , P(y|G Init ,D) is in the initial Bayesian network structure graph G Init The probability density of the observed variable y in it;

[0024] S44: The observation data set D Obs The experimental action Ac with the maximum utility function value best The corresponding experimental data set Mix and get a mixed data set

[0025] S45: Using BDe scoring function and hill climbing algorithm to extract the mixed data set D mix The final Bayesian network structure diagram is obtained;

[0026] S46: Calculate the conditional probability table of each node using the maximum likelihood estimation method;

[0027] Step 5: When an abnormal online working condition occurs, the phenomenon variable data is added as evidence to the Bayesian network model for reasoning;

[0028] S51: establishing an online abnormal operating condition event table according to the phenomenon variable nodes and the state levels corresponding to the phenomenon variable nodes;

[0029] S52: when an online abnormal condition occurs, determining online abnormal condition data information based on the online abnormal condition event table, inputting the online abnormal condition data information into the Bayesian network model as evidence, and calculating the posterior probabilities of other different control variables, and then selecting the control variable with the largest posterior probability;

[0030] Step 6: Make corresponding safe operation control decisions;

[0031] The control variable with the largest posterior probability is used as the decision-making basis. At the same time, the conditional probability table of each node is combined for auxiliary analysis to formulate corresponding safe operation control decisions;

[0032] Step 7: Execute the control decision and determine whether the abnormal conditions in the production process have disappeared;

[0033] The controller controls the operation of the coal slime flotation operation according to the obtained safe operation control decision, and uses several sensors installed at several sampling points in the coal slime flotation industrial system to collect online abnormal operating condition data information. If the online abnormal operating condition disappears, the operating state of the production process is maintained. If the online abnormal operating condition does not disappear, the current online abnormal operating condition data information is collected and input into the Bayesian network model as evidence, and the posterior probabilities of other different control variables are calculated. Then, the control variable with the largest posterior probability is selected, and steps six and seven are repeated until the abnormal condition disappears.

[0034] In the production process of coal slime flotation process, due to the influence of many uncertain factors such as environmental complexity, raw coal state and production conditions, the accuracy of the data information retained in the production process cannot be guaranteed, and the data information of abnormal working conditions is relatively scarce, which leads to the inability to establish an accurate safe operation control model simply relying on the abnormal working condition data in the actual production process. If only expert knowledge is used to determine the Bayesian network structure, when there are many nodes in the Bayesian network, it is difficult to determine the relationship between the Bayesian network nodes only by relying on expert knowledge. At the same time, the subjective limitations of expert knowledge are large, and an accurate Bayesian network model cannot be established. Therefore, the present invention provides a safe operation control method based on active learning and Bayesian network, and Bayesian network structure learning is performed in combination with experimental data and observation data. The method first obtains a skeleton graph by conditional independence test, then determines the direction of each edge of the skeleton graph to obtain the initial network structure, and then uses active learning to perform intervention experiments to obtain the final Bayesian network. Use less abnormal working condition data for structural learning, and finally use the online application strategy of the safe operation control model modeling method to establish an accurate safety control Bayesian network model.

[0035] The method has simple steps and high reliability. It can combine data information and expert knowledge in the coal slime flotation industrial process, and use a small amount of abnormal operating condition data information to establish a more accurate safety control Bayesian network model. It can assist operators to respond in time when abnormal conditions occur, formulate corresponding control decisions to eliminate abnormal conditions, ensure the safe and stable operation of the industrial process, and provide automatic decision-making support for abnormal conditions in the flotation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of the present invention;

[0037] Figure 2 It is a principle block diagram of the coal slime flotation industrial system of the present invention;

[0038] Figure 3 is the Bayesian network skeleton diagram obtained by the conditional independence test in the present invention;

[0039] Figure 4 is the initial Bayesian network structure in the present invention;

[0040] Figure 5 is the utility function value of 50 groups of interference actions in the present invention;

[0041] Figure 6 is the final Bayesian network diagram in the present invention;

[0042] Figure 7 It is the curve of the change of the cyclone inlet pressure after the safety control decision is implemented for abnormal working condition 2 in the present invention. DETAILED DESCRIPTION

[0043] The present invention is further described below in conjunction with examples and drawings to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some examples of the present invention, rather than all embodiments. The following description of at least one embodiment is actually only illustrative and is not intended to limit the present invention and its application or use.

[0044] like Figures 1 to 7 As shown, the present invention provides a method for safe control of a coal slime flotation process based on active learning and Bayesian network, including a coal slime flotation industrial system, such as Figure 1As shown, the coal slime flotation industrial system includes a main 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 circulating pump, a coal slime barrel, an arc screen, a concentrator, 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 separate the clean coal from the heavy medium. The medium suspension is mixed to form a mixture; the inlet end of the pressure pump is connected to the discharge port of the mixing barrel through a pipeline, so as 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, so as 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, so as to dehydrate and de-medium 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, so as to dehydrate and de-medium the low-density materials, and the high-density discharge port discharges the clean coal; The feed port of the medium barrel is respectively connected with the low-density discharge port of the first de-media screen, the low-density discharge port of the second de-media screen, 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 with the discharge port of the qualified medium barrel through a pipeline, and its outlet end is connected with 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 with the high-density discharge port of the second de-media screen, so as to perform preliminary treatment on the clean coal; the feed port of the curved screen is connected with 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 the coarse clean coal; the concentrated The feed port of the compressor is connected to the bottom discharge port of the curved screen to increase the concentration of the coal slime and reduce the moisture content of the coal slime, so that the pulp density is maintained within the range of stable process operation; the feed port of the pulp preprocessor is respectively connected to the circulating water filling pipeline, the flotation agent filling pipeline and the high-density discharge port of the concentrator to fully mix the flotation agent with the pulp 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 discharge port of the pulp preprocessor 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 pulp, and the top discharge port outputs the clean coal, and the bottom discharge port 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;

[0045] Among them, the typical coal slime flotation process flow of the coal slime flotation industrial system usually includes raw coal screening, coal-medium mixing, cyclone separation, coal slime de-mediuming, coal slime concentration, flotation pretreatment, foam flotation and other main processing links, which ultimately achieves the effective screening of small-particle difficult-to-select raw coal.

[0046] The safety control method of coal slime flotation process includes the following steps:

[0047] Step 1: Analyze the mechanism and operating experience of the main common abnormal conditions in the coal slime flotation process;

[0048] Heavy medium coal preparation uses the Archimedes principle to sort raw coal and finally obtain mineral impurities such as clean coal and gangue. The pressure pump pumps the coal-medium mixture into the heavy medium cyclone. The clean coal with lower density will be captured by the upward inner spiral flow in the cyclone, and will continue to move upward and finally be discharged from the overflow port (low-density discharge port) of the heavy medium cyclone; while the gangue impurities with higher density will be captured by the downward outer spiral flow, precipitate downward and be discharged from the bottom flow port (high-density discharge port) of the heavy medium cyclone. As the process continues, the bottom flow port of the heavy medium cyclone will be continuously washed by the gangue with higher density, which causes the bottom flow port diameter of the heavy medium cyclone to increase, and abnormal condition 1 (the bottom flow port diameter of the heavy medium cyclone increases) occurs. If abnormal condition 1 occurs, the suspension density is usually adjusted to reduce the amount of coal carried by the gangue.

[0049] The medium inlet pressure determines the tangential speed and centrifugal force of the coal-medium mixture, and affects the separation effect of clean coal and gangue. During production, the medium inlet pressure of the cyclone is monitored by a pressure gauge, which can be used to determine whether abnormal working condition 2 (the medium inlet pressure of the heavy medium cyclone is too low) occurs. If an abnormality occurs, the medium inlet pressure can be increased by adjusting the liquid level of the qualified medium barrel and the speed of the pressure pump, which is mainly achieved by adjusting the water pump, the medium pump and the pressure pump.

[0050] Combined with the actual situation of the coal slime flotation industrial system, the two main processing links of the coal slime flotation operation and the two main abnormal working conditions corresponding to the two main processing links are determined; the two main processing links are the heavy medium coal preparation link and the froth flotation processing link, and the two main abnormal working conditions are the enlargement of the bottom flow port diameter of the heavy medium cyclone and the low inlet pressure of the heavy medium cyclone;

[0051] Step 2: Determine the cause variables, phenomenon variables and control variables related to the two main abnormal conditions;

[0052] Conduct in-depth analysis of the coal slime flotation process to determine the cause variables, phenomenon variables and control variables that cause the abnormal operating condition of the heavy medium cyclone bottom flow port diameter to increase, and at the same time, determine the cause variables, phenomenon variables and control variables that cause the heavy medium cyclone inlet pressure to be too low;

[0053] Among them, the speed of the dilute medium pump, the speed of the medium pump, the speed of the water pump, and the speed of the pressure pump are the control variables, the liquid level of the qualified medium barrel, the density of the suspension, the amount of coal carried by the gangue and the pressure are the phenomenon variables, and the cyclone inlet pressure is too small and the diameter of the cyclone bottom flow port becomes larger as the cause variables;

[0054] Step 3: Determine the Bayesian network node division and node status level;

[0055] Based on the determined control variables, cause variables and phenomenon variables, simulation experiments were conducted on two main abnormal conditions: the enlarged diameter of the bottom flow port of the heavy medium cyclone and the too low inlet pressure of the heavy medium cyclone, and a safety control Bayesian network model was established. Then, the nodes of the Bayesian network and the state level of each node were determined, as shown in Table 1.

[0056] As a preferred method, the production data of the coal slime flotation industrial system was deeply analyzed, and 9 Bayesian network nodes were determined, namely A dilute medium pump speed, B add medium pump speed, C add water pump speed, D pressure pump speed, E qualified medium barrel liquid level, F suspension density, G cyclone inlet pressure is too small, H cyclone bottom flow port diameter becomes larger, I gangue coal content and J pressure. The level of each node is divided, among which the node level of A dilute medium pump speed is divided into unchanged 1, medium and high-end 2 and high-end 3, the node level of B add medium pump speed is divided into unchanged 1, medium and high-end 2 and high-end 3, and the node level of C add water pump speed is divided into unchanged 1, medium and high-end 2 and high-end 3. The node level of D pressure pump speed is divided into unchanged 1, medium and high-end 2 and high-end 3, the node level of E qualified medium barrel liquid level is divided into non-occurrence 1, moderate 2 and severe 3, the node level of F suspension density is divided into non-occurrence 1, moderate 2 and severe 3, the node level of G cyclone inlet pressure is too small is divided into non-occurrence 1, moderate 2 and severe 3, the node level of H cyclone bottom flow port diameter enlargement is divided into non-occurrence 1, moderate 2 and severe 3, the node level of I gangue coal content is divided into non-occurrence 1, moderate 2 and severe 3, and the node level of J pressure is divided into non-occurrence 1, moderate 2 and severe 3.

[0057] Table 1: Physical meaning of Bayesian network nodes and their classification levels

[0058]

[0059] Step 4: Perform parameter learning and structure learning of the Bayesian network;

[0060] S41: Based on the determined nodes of the Bayesian network and the production data of the coal slime flotation industrial system, the skeleton graph G is obtained using the conditional independence test ug ,like Figure 3 As shown; in the structure learning of Bayesian network, mutual information and conditional mutual information are used to determine whether there is a dependency relationship between different variables, and then determine the edges between network nodes, wherein the mutual information I(X; Y) is calculated according to formula (1);

[0061]

[0062] Where p(x,y) is the joint probability distribution of variables X and Y, p(x) is the marginal probability distribution of variable X, p(y) is the marginal probability distribution of variable Y, and x and y are the values ​​of variables X and Y respectively;

[0063] S42: Determine the skeleton graph G using the IGCI algorithm ug The direction of the edges between the nodes is obtained to obtain the initial Bayesian network structure G Init ,like Figure 4 As shown; Formula (2) is used to determine the causal relationship C between variables x and y by calculating the density loss value between the marginal probability distributions p(x) and p(y) of variables x and y. x→y , the calculation results are shown in Table 2;

[0064] C x→y =H(y)-H(x) (2);

[0065] Where H(x) and H(y) are the information entropy of variables x and y respectively; when C x→y <0, the causal relationship between variables x and y is x→y; when C x→y > 0, the causal relationship between variables x and y is y→x;

[0066] Table 2: Edge density loss values ​​obtained by the IGCI algorithm

[0067]

[0068] S43: Calculate each experimental action Ac using the utility function in formula (3) i For the initial Bayesian network structure G Init The value of the effect, and find the experimental action Ac with the maximum utility function value best , the results are as follows Figure 5 As shown;

[0069]

[0070] In the formula, P(y|G Init ,Ac i,D) is the initial Bayesian network structure graph G Init Execute action Ac i The probability density of the variable y after , P(y|G Init ,D) is in the initial Bayesian network structure graph G Init The probability density of the observed variable y in it;

[0071] S44: The observation data set D Obs The experimental action Ac with the maximum utility function value best The corresponding experimental data set Mix and get a mixed data set

[0072] S45: Using BDe scoring function and hill climbing algorithm to extract the mixed data set D mix The final Bayesian network structure diagram is obtained, as shown in Figure 6 As shown;

[0073] S46: Calculate the conditional probability table of each node using the maximum likelihood estimation method, and the probability table distribution is shown in Tables 3 to 9 below;

[0074] Table 3: Conditional probability table of nodes A, B, C, and D

[0075]

[0076] Table 4: Conditional probability table of node E

[0077]

[0078] Table 5: Conditional probability table of node F

[0079]

[0080] Table 6: Conditional probability table of node G

[0081]

[0082] Table 7: Conditional probability table of node H

[0083]

[0084] Table 8: Conditional probability table of node I

[0085]

[0086] Table 9: Conditional probability table of node J

[0087]

[0088] Step 5: When an abnormal online working condition occurs, the phenomenon variable data is added as evidence to the Bayesian network model for reasoning;

[0089] S51: establishing an online abnormal operating condition event table according to the phenomenon variable nodes and the state levels corresponding to the phenomenon variable nodes;

[0090] S52: When an online abnormal working condition occurs, determine the online abnormal working condition data information based on the online abnormal working condition event table, then input the online abnormal working condition data information into the Bayesian network model as evidence, and calculate the posterior probabilities of other different control variables, and then select the control variable with the largest posterior probability; wherein the abnormal working condition event table is shown in Table 10, and the posterior probability distribution corresponding to each abnormal working condition is shown in Table 11;

[0091] Table 10: Nine possible operating events

[0092]

[0093] Table 11: Posterior probabilities of decisions corresponding to nine possible operating conditions

[0094]

[0095]

[0096] Step 6: Make corresponding safe operation control decisions;

[0097] Based on the reasoning results of step five, the control variable with the largest posterior probability is used as the decision-making basis. At the same time, the conditional probability table of each node is combined for auxiliary analysis to make corresponding safe operation control decisions;

[0098] Step 7: Execute the control decision and determine whether the abnormal conditions in the production process have disappeared;

[0099] The controller controls the operation of the coal slime flotation operation according to the obtained safe operation control decision, and uses several sensors installed at several sampling points in the coal slime flotation industrial system to collect online abnormal operating condition data information. If the online abnormal operating condition disappears, the operating state of the production process is maintained. If the online abnormal operating condition does not disappear, the current online abnormal operating condition data information is collected and input into the Bayesian network model as evidence, and the posterior probabilities of other different control variables are calculated. Then, the control variable with the largest posterior probability is selected, and steps six and seven are repeated until the abnormal condition disappears.

[0100] Verification analysis:

[0101] In order to verify the effectiveness of the safe operation control method of the coal slime flotation process based on active learning and Bayesian network, relevant abnormal operating conditions were simulated on the coal slime flotation simulation platform. When the abnormal operating condition occurred, the control decision obtained by inference using the established Bayesian network model was used to verify whether the decision provided by the Bayesian network can eliminate the abnormal condition. Figure 7 As shown in the figure, the inlet pressure threshold is set to 95Kpa, and sampling is performed every second. In the figure, the inlet pressure drops rapidly from the 974th sampling point, and drops to the set threshold at the 1017th sampling point. The abnormal operating condition data information is input into the Bayesian network model as evidence, and appropriate safety control decisions are made and implemented. The inlet pressure of the heavy medium cyclone basically returns to normal at the 1041st sampling point. It can be seen that the Bayesian network model is effective.

[0102] In the production process of coal slime flotation process, due to the influence of many uncertain factors such as environmental complexity, raw coal state and production conditions, the accuracy of the data information retained in the production process cannot be guaranteed, and the data information of abnormal working conditions is relatively scarce, which leads to the inability to establish an accurate safe operation control model simply relying on the abnormal working condition data in the actual production process. If only expert knowledge is used to determine the Bayesian network structure, when there are many nodes in the Bayesian network, it is difficult to determine the relationship between the Bayesian network nodes only by relying on expert knowledge. At the same time, the subjective limitations of expert knowledge are large, and an accurate Bayesian network model cannot be established. Therefore, the present invention provides a safe operation control method based on active learning and Bayesian network, and Bayesian network structure learning is performed in combination with experimental data and observation data. The method first obtains a skeleton graph by conditional independence test, then determines the direction of each edge of the skeleton graph to obtain the initial network structure, and then uses active learning to perform intervention experiments to obtain the final Bayesian network. Use less abnormal working condition data for structural learning, and finally use the online application strategy of the safe operation control model modeling method to establish an accurate safety control Bayesian network model.

[0103] The method has simple steps and high reliability. It can combine data information and expert knowledge in the coal slime flotation industrial process, and use a small amount of abnormal operating condition data information to establish a more accurate safety control Bayesian network model. It can assist operators to respond in time when abnormal conditions occur, formulate corresponding control decisions to eliminate abnormal conditions, ensure the safe and stable operation of the industrial process, and provide automatic decision-making support for abnormal conditions in the flotation process.

Claims

1. A method for safe control of coal slime flotation process based on active learning and 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, a curved 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 safety control method of coal slime flotation process includes the following steps: Step 1: Analyze the mechanism and operating experience of the main common abnormal conditions in the coal slime flotation process; Combined with the actual situation of the coal slime flotation industrial system, the two main processing links of the coal slime flotation operation and the two main abnormal working conditions corresponding to the two main processing links are determined; the two main processing links are the heavy medium coal preparation link and the froth flotation processing link, and the two main abnormal working conditions are the enlargement of the bottom flow port diameter of the heavy medium cyclone and the low inlet pressure of the heavy medium cyclone; Step 2: Determine the cause variables, phenomenon variables and control variables related to the two main abnormal conditions; Conduct in-depth analysis of the coal slime flotation process to determine the cause variables, phenomenon variables and control variables that cause the abnormal operating condition of the heavy medium cyclone bottom flow port diameter to increase, and at the same time, determine the cause variables, phenomenon variables and control variables that cause the heavy medium cyclone inlet pressure to be too low; Step 3: Determine the Bayesian network node division and node status level; Based on the determined control variables, cause variables and phenomenon variables, simulation experiments were conducted on two main abnormal operating conditions: the enlarged diameter of the bottom flow port of the heavy medium cyclone and the too low inlet pressure of the heavy medium cyclone. A safety control Bayesian network model was established, and then the nodes of the Bayesian network and the state level of each node were determined. Step 4: Perform parameter learning and structure learning of the Bayesian network; S41: Based on the determined nodes of the Bayesian network and the production data of the coal slime flotation industrial system, the skeleton graph G is obtained using the conditional independence test ug ; In the structure learning of Bayesian network, mutual information and conditional mutual information are used to determine whether there is a dependency relationship between different variables, and then determine the edges between network nodes, where mutual information I(X;Y) is calculated according to formula (1); Where p(x,y) is the joint probability distribution of variables X and Y, p(x) is the marginal probability distribution of variable X, p(y) is the marginal probability distribution of variable Y, and x and y are the values ​​of variables X and Y respectively; S42: Determine the skeleton graph G using the IGCI algorithm ug The direction of the edges between the nodes is obtained to obtain the initial Bayesian network structure G Init ; Use formula (2) to determine the causal relationship C between variables x and y by calculating the density loss value between the marginal probability distributions p(x) and p(y) of variables x and y x→y ; C x→y =H(y)-H(x) (2); Where H(x) and H(y) are the information entropy of variables x and y respectively; when C x→y <0, the causal relationship between variables x and y is x→y; when C x→y > 0, the causal relationship between variables x and y is y→x; S43: Calculate each experimental action Ac using the utility function in formula (3) i For the initial Bayesian network structure G Init The value of the effect, and find the experimental action Ac with the maximum utility function value best ; In the formula, P(y|G Init ,Ac i ,D) is the initial Bayesian network structure graph G Init Execute action Ac i The probability density of the variable y after y, P(y|G Init ,D) is in the initial Bayesian network structure graph G Init The probability density of the observed variable y in it; S44: The observation data set D Obs The experimental action Ac with the maximum utility function value best The corresponding experimental data set Mix and get a mixed data set S45: Using BDe scoring function and hill climbing algorithm to extract the mixed data set D mix The final Bayesian network structure diagram is obtained; S46: Calculate the conditional probability table of each node using the maximum likelihood estimation method; Step 5: When an abnormal online working condition occurs, the phenomenon variable data is added as evidence to the Bayesian network model for reasoning; S51: establishing an online abnormal operating condition event table according to the phenomenon variable nodes and the state levels corresponding to the phenomenon variable nodes; S52: when an online abnormal condition occurs, determining online abnormal condition data information based on the online abnormal condition event table, inputting the online abnormal condition data information into the Bayesian network model as evidence, and calculating the posterior probabilities of other different control variables, and then selecting the control variable with the largest posterior probability; Step 6: Make corresponding safe operation control decisions; The control variable with the largest posterior probability is used as the decision-making basis. At the same time, the conditional probability table of each node is combined for auxiliary analysis to formulate corresponding safe operation control decisions; Step 7: Execute the control decision and determine whether the abnormal conditions in the production process have disappeared; The controller controls the operation of the coal slime flotation operation according to the obtained safe operation control decision, and uses several sensors installed at several sampling points in the coal slime flotation industrial system to collect online abnormal operating condition data information. If the online abnormal operating condition disappears, the operating state of the production process is maintained. If the online abnormal operating condition does not disappear, the current online abnormal operating condition data information is collected and input into the Bayesian network model as evidence, and the posterior probabilities of other different control variables are calculated. Then, the control variable with the largest posterior probability is selected, and steps six and seven are repeated until the abnormal condition disappears.

Images