Coal slime flotation process control method based on distributed pyramid dynamic Bayesian network

Abstract

A coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network comprises the steps that the coal slime flotation process is divided into three sub-processes, a global quality index variable and a local quality index variable are determined, and a global dynamic Bayesian network is established; key variables in the sub-processes are determined, correlation among the variables is analyzed, and process time delay in each process is determined; determining the number of time slices of the first layer of dynamic Bayesian network; performing structure learning, and establishing a network structure of a first layer of the model; combining the same nodes of adjacent time slices into one node, and constructing a second-layer dynamic Bayesian network structure; bayesian network parameter learning is carried out; actual working condition data information is collected, and when an abnormal working condition occurs, a sub-process causing abnormity is determined; a local pyramid dynamic Bayesian network is utilized to determine a root cause causing an abnormal working condition; making a control decision scheme according to the abnormal reasons, and eliminating abnormal working conditions; the method can eliminate abnormal working conditions in time.

Description

Technical Field

[0001] The invention belongs to the technical field of complex industrial process safety control, and specifically is a coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network. Background Art

[0002] As one of the typical complex industrial processes, the coal slime flotation process consists of multiple mutually coupled sub-processes, such as raw coal pretreatment, sorting, centrifugal separation, froth flotation, etc. Each sub-process contains many variables, and there are complex interactions and coupling relationships between these variables. With the rapid development of modern industry, the scale of the coal slime flotation process is also expanding, which not only increases the difficulty of process control, but also makes the system dynamics, time lag and uncertainty more significant. The expansion of the scale of the coal slime flotation process has brought a series of problems, including increased equipment maintenance costs, increased operational complexity, increased sensitivity to raw material fluctuations, and higher requirements for adaptability to environmental changes. At the same time, the current level of automation in the coal slime flotation process is low, resulting in frequent abnormal conditions in the coal slime flotation process. If the abnormal conditions are not eliminated in time, it will lead to a decrease in production efficiency, unstable product quality, and may even cause safety accidents. In addition, the insufficient level of automation will also make the operator's subjective experience overly dependent on the response to abnormal conditions, increase the risk of human error, and thus affect the stability and reliability of the entire coal slime flotation process. Therefore, taking effective measures to improve the automation level of the coal slime flotation process is essential to ensure its efficient, stable and safe operation while expanding its scale.

[0003] As the scale of the coal slime flotation process increases, it also makes it extremely difficult to model it centrally. In large-scale flotation plants, the dynamic behavior of the system is affected by many variables, including the properties of the raw coal, the operating conditions of the flotation machine, environmental factors, etc. The interaction and changes of these variables increase the complexity of modeling. Centralized modeling requires consideration of the global characteristics of the entire system, including the interaction and information transfer between the various subsystems. However, since the difficulty of data collection, processing and analysis increases significantly with the increase in system scale, this requirement is often difficult to achieve in actual operation. In addition, centralized models often require a lot of computing resources and time, which is impractical in real-time control and decision support.

[0004] Bayesian network is a probabilistic model used to represent the conditional dependency between variables. It consists of a set of random variables and the conditional probability distribution defined on these variables, usually represented in the form of a directed acyclic graph. Dynamic Bayesian network is an extension of Bayesian network. It models time series data by replicating the network structure on the time layer, thereby capturing the dynamic characteristics of variables over time. In order to reduce the complexity of the model, dynamic Bayesian network is usually required to satisfy the first-order Markov property, that is, the state of the variable at the current moment depends only on the state of the variable at the previous moment. However, in the process of coal slime flotation, due to the large amount of time delay between variables, the state of the variable at the current moment not only depends on the state of the variable at the previous moment, but also is affected by the state of the variable at one or several time moments in the past. This time delay characteristic violates the first-order Markov assumption, making it difficult for traditional dynamic Bayesian network to be directly applied to the coal slime flotation process. Therefore, it is necessary to provide a dynamic Bayesian network model that not only meets the first-order Markov assumption, but also takes into account the dynamic characteristics and time delays in the coal slime flotation process, so as to accurately identify the causes of abnormal conditions in the coal slime flotation process, and based on these diagnostic information, formulate scientific control strategies to eliminate abnormalities in a timely manner and ensure the safety and stability of the coal slime flotation process. Summary of the invention

[0005] In view of the problems existing in the above-mentioned prior art, the present invention provides a coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network. The method has a simple modeling process and high reliability. It can accurately identify the causes of abnormal conditions in the coal slime flotation process, and based on this, quickly and accurately formulate a scientific control strategy, thereby eliminating abnormal conditions in time, and effectively ensuring the safety and stability of the coal slime flotation process.

[0006] In order to achieve the above object, the present invention provides a coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network, including a coal slime flotation industrial system, wherein the coal slime flotation industrial system includes a raw coal processing subsystem, a heavy medium coal preparation subsystem and a concentrated flotation subsystem;

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

[0008] Step 1: Divide the coal slime flotation process into three sub-processes, determine the global quality index variables and local quality index variables, and establish a global dynamic Bayesian network;

[0009] S11: Through in-depth analysis of historical data of coal slime flotation process, combined with rich process knowledge and the actual situation of coal slime flotation industrial system, the coal slime flotation process is divided into three sub-processes: raw coal processing, heavy medium coal preparation and concentrated flotation by adopting distributed thinking. The three sub-processes correspond to the raw coal processing subsystem, heavy medium coal preparation subsystem and concentrated flotation subsystem in the coal slime flotation industrial system respectively;

[0010] S12: The overflow ash content S of the flotation tank is determined as the global quality indicator variable, and a quality indicator expected value is set for each sub-process as the respective local quality indicator variable. The local quality indicator variables of the three sub-processes are S1, S2, and S3 respectively;

[0011] S13: Establish a global Bayesian network based on the causal relationship between the global quality indicator variables and the local quality indicator variables;

[0012] S14: Connect each subprocess between different time slices to build a global dynamic Bayesian network;

[0013] Step 2: Determine the key variables in the three sub-processes, collect process data, analyze the correlation between variables, and determine the process delay T in each process;

[0014] S21: Combined with the knowledge of coal slime flotation process, the key variables in the three sub-processes are determined based on the ability to reflect product quality and ease of measurement, and are divided into operating variables and measurement variables; the changes in product indicators are reflected through the measurement variables in each sub-process, and the improvement of product quality is achieved by adjusting the operating variables;

[0015] S22: using the sensor groups arranged in the raw coal processing subsystem, the heavy medium coal separation subsystem and the concentration flotation subsystem to respectively collect process data in the raw coal processing subprocess, the heavy medium coal separation subprocess and the concentration flotation subprocess in real time, and obtain time series data;

[0016] S23: Use the maximum information coefficient method to analyze the time series data in the raw coal processing subprocess, the heavy medium coal preparation subprocess, and the concentration flotation subprocess, and determine the correlation between the variables;

[0017] S24: approximate the time delay between the local quality indicator variables and the key variables in the three sub-processes according to the correlation results, and take the maximum time delay as the process delay T of the process;

[0018] Step 3: According to the process delay T of each sub-process, use formula (1) to determine the number of time slices of the first layer of the dynamic Bayesian network in the local pyramid dynamic Bayesian network model;

[0019] W=2 (K-1) (1);

[0020] Where W represents the number of time slices of the first-layer dynamic Bayesian network of the pyramid; K = ceil(log2(T+1); ceil(·) represents the upward rounding function;

[0021] Step 4: Perform structural learning and establish the first-layer network structure of the pyramid dynamic Bayesian network model;

[0022] After determining the number of time slices of the first layer of the local pyramid dynamic Bayesian network in each sub-process, the first layer dynamic Bayesian network structure is learned using the Bayesian information criterion scoring function in formula (2);

[0023]

[0024] In the formula, D represents the given data set; G represents the given structure; N represents the number of training samples; n is the number of nodes; I P (X i ; Pa(X i )) represents the node X in the given structure G i The mutual information with its parent node; Dim[G] is the number of independent parameters of model G; H P (X i ) represents node X i Entropy of

[0025] Step 5: Use the variable length data window to merge the same nodes of adjacent time slices in the first layer of dynamic Bayesian network into one node to construct the second layer of dynamic Bayesian network structure;

[0026] The same nodes of adjacent time slices in the first layer of dynamic Bayesian network are merged into one node by using variable length data window, so that the nodes in the second layer of dynamic Bayesian network contain more data dynamic information; at the same time, the data window size is changed, and the Bayesian information criterion scoring function is reused for structural learning to obtain the second layer of pyramid dynamic Bayesian network structure;

[0027] Step 6: Repeat step 5 until the number of time slices of the dynamic Bayesian network is 2;

[0028] The time slice data is reduced by merging network nodes until the number of time slices is reduced to 2 and the merging process is stopped; finally, the number of time slices of the pyramid dynamic Bayesian network is 2 (K-1) ,2 (K-2) ,2 (K-3) ,...,2 1 ;

[0029] Step 7: Parameter learning of global dynamic network and local pyramid dynamic Bayesian network;

[0030] Perform parameter learning to enable the dynamic Bayesian network to capture the dynamic relationship between variables in time series data; then, infer the variable state that causes abnormal conditions through conditional probability distribution, and formulate the optimal safety control decision plan;

[0031] Step 8: Collect actual operating condition data information, determine whether an abnormal operating condition occurs, and determine the sub-process that causes the abnormality when an abnormal operating condition occurs;

[0032] The actual working condition data information of the global quality indicator variable is compared with the set safety threshold, where the safety threshold is set to 9.2%; when the flotation tank overflow ash content S≤9.2%, it means that the coal slime flotation process is in a safe operating state; when the flotation tank overflow ash content S>9.2%, it means that the coal slime flotation process is abnormal, and the online abnormal data is input into the global dynamic Bayesian network as evidence information, and the sub-process where the abnormality occurs is inferred based on the global network reasoning principle; in the global network reasoning process, the abnormal probability of each sub-process at the current moment and the previous moment is calculated, and the sub-process with the highest abnormal probability at the current moment and the previous moment is found; if the sub-process with the highest abnormal probability at the previous moment is adjacent to the sub-process with the highest abnormal probability at the current moment, and is before the sub-process with the highest abnormal probability at the current moment, it means that the abnormality of the current sub-process is caused by the abnormality of the previous sub-process, and the previous sub-process needs to be detected for abnormality, otherwise it is considered that the cause of the abnormality is the current sub-process;

[0033] Step 9: Use the local pyramid dynamic Bayesian network to determine the root cause of the abnormal working condition;

[0034] After determining the abnormal sub-process, the local online abnormal data is input as evidence information into each layer of the local pyramid dynamic Bayesian network corresponding to the current moment, and reasoning is performed layer by layer from top to bottom to determine the abnormal diagnosis results of each layer of the pyramid dynamic Bayesian network according to the principle of maximum a posteriori probability;

[0035] Step 10: Based on the abnormal causes diagnosed by the local pyramid dynamic Bayesian network, formulate and implement a reasonable control decision plan to eliminate abnormal conditions;

[0036] S101: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are the same, it means that the time delay in the process does not affect the fault location result and decision-making, and the control decision plan inferred based on the corresponding abnormal cause is directly implemented; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, it means that the abnormal condition is eliminated and the reasoning process ends; otherwise, execute step eleven;

[0037] S102: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are different, the control decision is implemented according to the following steps:

[0038] A1: First, use the adjustment plan inferred from the first level of the local pyramid to make operational adjustments. If the overflow ash content S of the flotation tank returns to the safety threshold, the reasoning process ends.

[0039] A2: After executing the adjustment plan of the first layer, if the coal slime flotation process has not returned to normal operation, the adjustment plan inferred by the second layer will be continued to be used for operation adjustment on the basis of the first layer operation adjustment; when implementing the adjustment plan of the second layer, only the newly emerged operation strategy will be adopted and executed, and the variables that have been adjusted in the first layer operation adjustment process will not be adjusted repeatedly; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, the reasoning process ends;

[0040] A3: After executing the adjustment plans of the first and second layers, if the coal slime flotation process has not resumed normal operation, the inference plans of the subsequent layers of the pyramid dynamic Bayesian network will continue to be adopted and executed. The execution process is the same as the execution process of the adjustment plan of the second layer. Only the newly appeared adjustment strategy will be executed, and the executed adjustment plan will not be repeated until the coal slime flotation process resumes normal operation. The inference process ends; when the adjustment plan inferred by the last layer is executed, if the coal slime flotation process has not resumed normal operation, execute step eleven.

[0041] Step 11: Select the sub-process with the second largest abnormal probability in the global reasoning process for reasoning, and re-execute step 9, and continue to use the online abnormal data as evidence information to infer the control decision plan until the abnormal condition is eliminated.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The coal slime flotation process is a key link in the field of coal washing. Its purpose is to improve the quality of coal through physical and chemical methods, thereby improving the utilization efficiency of coal resources. However, in the process of coal slime flotation, the fluctuation of raw material properties, frequent adjustment of operations, and complex and changeable working environment conditions make the coal slime flotation process have a large number of dynamic characteristics and time delays. Dynamic characteristics and time delays not only increase the difficulty of process control, but also lead to the frequent occurrence of abnormal conditions. If effective measures are not taken to timely identify the cause of the abnormality, make correct control decisions, and eliminate the abnormality, it will not only lead to waste of resources and environmental pollution, but also may cause damage to production equipment and cause safety accidents. Traditional dynamic Bayesian networks are a powerful tool for modeling and reasoning about uncertainty and dynamic characteristics in time series data. Dynamic Bayesian networks capture the dynamic characteristics of the process by extending the Bayesian network on the time axis to link the current state of the process with the previous state. However, dynamic Bayesian networks usually assume that the change of the variable state at the current moment depends only on the variable state at the previous moment. This first-order Markov assumption simplifies the complexity of the model in many cases, but also limits the model's ability to capture the causal relationship caused by time delay. In the process of coal slime flotation, due to the existence of time delay, relying solely on traditional dynamic Bayesian networks may not accurately predict and control the delayed relationship between process variables, resulting in the efficiency and accuracy of the control strategy being affected. In order to effectively deal with the impact of dynamic characteristics and time delay on abnormal cause identification in the process of coal slime flotation, the present invention provides a method for safe control of coal slime flotation industrial process based on a distributed pyramid dynamic Bayesian network. By establishing a dynamic Bayesian network with a decreasing number of time slices layer by layer, not only can the dynamic characteristics in the coal slime flotation process be accurately captured, but also the time delay in the process can be gradually covered. More importantly, the dynamic Bayesian networks established by this method all meet the first-order Markov assumption and will not increase the complexity of modeling. The main process of abnormal diagnosis and decision-making of the distributed pyramid dynamic Bayesian network is divided into four parts: the first part is data acquisition, sub-process division, variable determination, and establishment of a global dynamic Bayesian network. Connecting the three sub-processes between different time slices and establishing a global dynamic Bayesian network can effectively reflect the linkage between sub-processes, which can help to quickly solve the problem of abnormal information transmission. The second part is the correlation analysis between the variables in the sub-processes to establish the dynamic Bayesian network of the first layer of the pyramid. The maximum information coefficient method is used to analyze the correlation of variables in each sub-process, and the time delay of each variable is approximately estimated. The maximum time delay is used as the process delay T of the process, which can facilitate the determination of the number of time slices in the first layer of the pyramid dynamic Bayesian network and construct its network structure. The third part is to merge the same nodes of adjacent time slices in the first layer of the dynamic Bayesian network to obtain the dynamic Bayesian network of the subsequent layers.The data information of the same nodes in adjacent time slices in the first layer of the dynamic Bayesian network is merged by using a variable length data window, so that the network nodes in the second layer contain more data information. At the same time, the loss of node data information during the node merging process can be avoided. Furthermore, the structure learning and parameter learning are re-performed to complete the construction of the second layer of the pyramid dynamic Bayesian network model, which can facilitate the estimation of the conditional probability of variables in the network from historical data. The construction method of the dynamic Bayesian network model of the subsequent layers of the pyramid is the same as that of the second layer. When the number of dynamic Bayesian network time slices is reduced to 2, the modeling is completed. The fourth part is control decision reasoning to eliminate abnormal conditions. The online abnormal data is input into the global dynamic Bayesian network model as evidence information to infer the abnormal sub-process. Then, according to the local pyramid dynamic Bayesian network of the abnormal sub-process, the control strategy of the abnormal condition is inferred from top to bottom layer by layer. Finally, according to the inferred control strategy, the operation is carried out to eliminate the abnormal condition.

[0044] The modeling process of this method is simple and reliable. It can accurately identify the causes of abnormal conditions in the coal slime flotation process, and based on this, quickly and accurately formulate scientific control strategies, thereby eliminating abnormal conditions in a timely manner and effectively ensuring the safety and stability of the coal slime flotation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of the present invention;

[0046] Figure 2 It is a simplified principle block diagram of the coal slime flotation industrial system of the present invention;

[0047] Figure 3 It is a global dynamic Bayesian network structure diagram in the present invention;

[0048] Figure 4 It is the maximum information coefficient result diagram of subprocess 1 in the present invention;

[0049] Figure 5 This is the maximum information coefficient result diagram of subprocess 2 of the present invention;

[0050] Figure 6 This is the maximum information coefficient result diagram of subprocess 3 of the present invention;

[0051] Figure 7 It is a pyramid dynamic Bayesian network structure diagram of subprocess 1 of the present invention;

[0052] Figure 8 It is a pyramid dynamic Bayesian network structure diagram of subprocess 2 of the present invention;

[0053] Fig. 9 It is a pyramid dynamic Bayesian network structure diagram of subprocess 3 of the present invention;

[0054] Fig.10 is the global dynamic Bayesian inference result of case 1 in the present invention;

[0055] Fig.11 is the global dynamic Bayesian inference result of case 2 in the present invention;

[0056] Fig.12 It is the change of the overflow ash curve of the flotation tank in the coal slime flotation process of Case 1 of the present invention;

[0057] Fig.13 It is the change of the overflow ash curve of the flotation cell in the coal slime flotation process of Case 2 in the present invention. DETAILED DESCRIPTION

[0058] The present invention will be further described below in conjunction with the accompanying drawings.

[0059] Coal slime flotation is an efficient mineral processing process, which is mainly used to improve the quality of coal and separate the useful components from impurities in the coal slime by physical and chemical methods. This process involves several key steps, including crushing, mixing, pH adjustment, collector addition, stirring, aeration flotation, separation and concentration, dehydration, filtration and drying of raw coal. During the flotation process, coal particles adhere to bubbles and float up due to their hydrophobic properties, while hydrophilic impurities remain in the slurry, thus achieving separation. Coal slime flotation plays an important role in improving the utilization rate of coal resources, reducing environmental pollution and improving economic benefits. With the rapid development of automation and intelligent technology, the efficiency and quality control of coal slime flotation process are facing new opportunities for improvement. Automation technology can optimize the flotation effect and reduce human errors by accurately controlling various parameters in the flotation process, such as pulp concentration, pH value, and reagent addition. Intelligent technology can analyze historical data, predict and identify abnormal conditions in the production process, so as to achieve more accurate process control and decision support. The application of these technologies can not only improve the production efficiency of coal slime flotation and reduce operating costs, but also effectively reduce environmental pollution and promote the sustainable development of the coal industry. Therefore, integrating automation and intelligent technologies into the coal slime flotation process plays a key role in ensuring the safe and stable operation of the coal slime flotation process.

[0060] like Figure 1 As shown, the present invention provides a coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network, including a coal slime flotation industrial system, such as Figure 2As shown, the coal slime flotation industrial system includes a raw coal processing subsystem, a heavy medium coal preparation subsystem and a concentrated flotation subsystem; the raw coal processing subsystem mainly removes impurities such as rocks, soil and plant debris in the raw coal through screening, crushing and washing to improve the quality and availability of coal; the heavy medium coal preparation subsystem mainly utilizes the density difference between coal and gangue, and uses equipment such as heavy medium cyclones and separators to efficiently separate high calorific value coal while reducing transportation costs. The concentrated flotation subsystem mainly further purifies fine coal, removes sulfur and other harmful impurities through equipment such as flotation machines, agitators and reagent addition systems to improve coal combustion efficiency and reduce environmental pollution.

[0061] The raw coal processing subsystem includes a crusher, a double-layer screen and a single-layer screen. The crusher is used to crush the raw coal; the feed port of the double-layer screen is connected to the discharge port of the crusher, and is used to recover and reduce ash from the crushed coarse coal slime by screening, and output clean coal; the feed port of the single-layer screen is connected to the tailing discharge port of the double-layer screen, and is used to remove impurities in the raw coal by screening;

[0062] The heavy medium coal preparation subsystem includes a mixing box, a heavy medium cyclone, an arc screen 1, a de-medium screen 1, a clean coal centrifuge, a diversion device, a magnetic separator 1, a coal slime bucket, a arc screen 2, a centrifuge 1, a de-medium screen 2, a magnetic separator 2 and a centrifuge 2; the feed port of the mixing box is connected to the tail material discharge port of the single-layer screen, so as to mix the raw coal with the recovered liquid and then output the mixed material; the feed port of the heavy medium cyclone is connected to the discharge port of the mixing box, so as to separate the low-density material from the high-density material by utilizing the density difference of the material; the feed port of the arc screen 1 is connected to the low-density discharge port of the heavy medium cyclone, Used to realize pre-medium removal operation; the feed port of the medium removal screen 1 is connected to the fixed discharge port of the curved screen 1, which is used to realize secondary medium removal operation; the feed port of the clean coal centrifuge is connected to the solid discharge port of the medium removal screen 1, the solid discharge port of the clean coal centrifuge discharges clean coal, and the clean coal centrifuge is used to separate clean coal and liquid; the feed port of the diversion device is connected to the liquid discharge port of the curved screen 1, the concentrate discharge port of the diversion device is connected to the feed port of the mixing box, and the diversion device is used to realize the diversion of the top flow containing concentrate and the bottom flow containing tailings; the feed port of the magnetic separator 1 is connected to the liquid discharge port of the medium removal screen 1 and the tailings discharge port of the diversion device The low-density material discharge port of the magnetic separator one is connected to the feed port of the mixing box. The magnetic separator one is used to remove iron impurities and separate low-density materials from high-density materials; the feed port of the coal slime bucket is connected to the liquid discharge port of the clean coal centrifuge and the high-density material discharge port of the magnetic separator one, so as to achieve sufficient stirring and mixing of the coal slime; the feed port of the arc screen two is connected to the discharge port of the coal slime bucket for medium removal; the feed port of the centrifuge one is connected to the solid discharge port of the arc screen two, the liquid discharge port of the centrifuge one is connected to the feed port of the coal slime bucket, and the solid discharge port of the centrifuge one discharges clean coal. The centrifuge one is used To achieve separation of clean coal and liquid; the feed port of the second de-mediation screen is connected to the high-density discharge port of the heavy medium cyclone, the liquid discharge port of the second de-mediation screen is connected to the feed port of the mixing box, and the second de-mediation screen is used to achieve de-mediation operation; the feed port of the second magnetic separator is connected to the solid discharge port of the second de-mediation screen, the low-density material discharge port of the second magnetic separator is connected to the feed port of the mixing box, and the second magnetic separator is used to remove iron impurities and achieve separation of low-density materials and high-density materials; the feed port of the second centrifuge is connected to the liquid discharge port of the second de-mediation screen, the fixed discharge port of the second centrifuge discharges medium coal, and the second centrifuge is used to separate medium coal from liquid;

[0063] The concentrating and flotation subsystem comprises a coal slime concentrator, a pulp pre-processor, a flotation tank, a tailing coal concentrator 1, a clean coal filter, a drying device, a tailing coal concentrator 2, a concentrator 3 and a filter press; the feed port of the coal slime concentrator is connected to the liquid discharge port of the curved screen 1, so as to increase the concentration of the coal slime and reduce the moisture content of the coal slime; the feed port of the pulp pre-processor is connected to the solid discharge port of the coal slime concentrator, so as to pre-treat the pulp before flotation to improve the flotation performance; the feed port of the flotation tank is connected to the discharge port of the pulp pre-processor, so as to separate the coal and impurities; the feed port of the tailing coal concentrator 1 is connected to the tailings discharge port of the flotation tank, the solid discharge port of the tailing coal concentrator 1 discharges the tailings, the liquid discharge port of the tailing coal concentrator 1 discharges the circulating water, and the tailing coal concentrator 1 is used to increase the concentration of the tailings and reduce the moisture content of the tailings; the feed port of the clean coal filter is connected to the concentrate discharge port of the flotation tank, and the clean coal is discharged from the solid discharge port of the tailing coal concentrator 1. The liquid discharge port of the filter is connected to the feed port of the slurry preprocessor, and the clean coal filter is used to increase the concentration of the clean coal and reduce the moisture content of the clean coal; the drying equipment is connected to the solid discharge port of the clean coal filter, and is used to dry the clean coal and output the clean coal; the feed port of the tail coal concentrator II is connected to the high-density discharge port of the magnetic separator II and the liquid discharge port of the centrifuge II, and the liquid discharge port of the tail coal concentrator II discharges circulating water, and the tail coal concentrator II is used to increase the concentration of the tail coal and reduce the moisture content of the tail coal; the feed port of the concentrator III is respectively connected to the liquid discharge port of the coal slime concentrator and the fixed discharge port of the tail coal concentrator II, and the liquid discharge port of the concentrator III discharges circulating water, and the concentrator III is used to separate fixed impurities from liquid; the feed port of the filter press is connected to the solid discharge port of the concentrator III, the fixed discharge port of the filter press discharges filter cake, and the liquid discharge port of the filter press outputs circulating water, and the filter press is used for solid-liquid separation;

[0064] The coal slime flotation process control method comprises the following steps:

[0065] Step 1: Divide the coal slime flotation process into three sub-processes, determine the global quality index variables and local quality index variables, and establish a global dynamic Bayesian network;

[0066] S11: Through in-depth analysis of historical data of coal slime flotation process, combined with rich process knowledge and the actual situation of coal slime flotation industrial system, the coal slime flotation process is divided into three sub-processes: raw coal processing, heavy medium coal preparation and concentrated flotation by adopting distributed thinking. The three sub-processes correspond to the raw coal processing subsystem, heavy medium coal preparation subsystem and concentrated flotation subsystem in the coal slime flotation industrial system respectively;

[0067] S12: The overflow ash content S of the flotation tank is determined as the global quality indicator variable, and a quality indicator expected value is set for each sub-process as the respective local quality indicator variable. The local quality indicator variables of the three sub-processes are S1, S2, and S3 respectively;

[0068] S13: Establish a global Bayesian network based on the causal relationship between the global quality indicator variables and the local quality indicator variables;

[0069] S14: Considering the abnormal information transmission between sub-processes, each sub-process is connected between different time slices to construct a global dynamic Bayesian network; the global dynamic Bayesian network structure constructed by learning the data information of the global quality indicator variables and the local indicator variables is as follows: Figure 3 shown.

[0070] Step 2: Determine the key variables in the three sub-processes, collect process data, analyze the correlation between variables, and determine the process delay T in each process;

[0071] The Maximum Information Coefficient (MIC) is a method for measuring the correlation between two variables, and is applicable to both linear and nonlinear variables. The principle of MIC is based on the concept of mutual information. The mutual information value is obtained by gridding the data points in two-dimensional space and calculating the joint probability distribution under different grid divisions. MIC looks for the grid division that can maximize the mutual information value, and normalizes the mutual information value so that it falls within the range of 0 to 1. The closer it is to 1, the stronger the correlation is. Conversely, the closer it is to 0, the more independent the two variables are. The maximum value of the normalized mutual information is the MIC. The results of the correlation analysis are as follows: Figures 4 to 6 shown.

[0072] S21: Combined with the knowledge of coal slime flotation process, the key variables in the three sub-processes are determined based on the ability to reflect product quality and ease of measurement, and divided into operating variables and measurement variables; there are 6 operating variables and 8 measurement variables; the changes in product indicators are reflected through the measurement variables in each sub-process, and the improvement of product quality is achieved by adjusting the operating variables;

[0073] The states of the operating variables and the measured variables are further divided into 1, 2, and 3, which respectively indicate that the corresponding variable values ​​belong to normal values, low values ​​(lower than normal values), and high values ​​(higher than normal values). For the global quality indicator variable S and the local quality indicator variables S1, S2, and S3, they only need to measure whether the whole process of the coal slime flotation process and the corresponding sub-processes are abnormal. Therefore, these four variables only need to be divided into two states, 1 and 2, which represent normal and abnormal, respectively. The node symbols of the measured variables and the operating variables and the corresponding state divisions are shown in Tables 1 and 2.

[0074] Table 1: Variables measured during coal slime flotation

[0075]

[0076]

[0077] Table 2: Slime flotation process operating variables

[0078]

[0079]

[0080] S22: using the sensor groups arranged in the raw coal processing subsystem, the heavy medium coal separation subsystem and the concentration flotation subsystem to respectively collect process data in the raw coal processing subprocess, the heavy medium coal separation subprocess and the concentration flotation subprocess in real time, and obtain time series data;

[0081] S23: Use the maximum information coefficient method to analyze the time series data in the raw coal processing subprocess, the heavy medium coal preparation subprocess, and the concentration flotation subprocess, and determine the correlation between the variables, such as Figures 4 to 6 As shown;

[0082] S24: Approximately estimate the time delay between the local quality indicator variables and the key variables in the three sub-processes based on the correlation results, and take the largest time delay as the process delay T of the process; the variables involved in each sub-process and the time delay of the variables and the process delay are shown in Table 3;

[0083] Table 3: Sub-process variable assignments and their time delays

[0084] Subprocess Variable time delay Process delay Raw coal processing <![CDATA[A(t-2),B(t-9),C(t-5),S1(t-9)]]> T＝9 Heavy medium coal preparation <![CDATA[D(t-3),E(t-1),F(t-4),G(t-14),H(t-12),I(t-9),S2(t-5)]]> T＝14 Concentration flotation <![CDATA[J(t-13),K(t-4),L(t-8),M(t-10),S3(t-7)]]> T＝13

[0085] Step 3: According to the process delay T of each sub-process, use formula (1) to determine the number of time slices of the first layer of the dynamic Bayesian network in the local pyramid dynamic Bayesian network model;

[0086] W=2 (K-1) (1);

[0087] Where W represents the number of time slices in the first layer of the pyramid dynamic Bayesian network; K = ceil(log2(T+1); ceil(·) represents the ceiling function; the purpose of using T+1 and the ceiling function is to ensure that the number of time slices can cover the time delay in the industrial process.

[0088] The parameter N is calculated by the process delay T of the sub-process, which is equal to 4. Therefore, the number of time slices of the first layer of the local pyramid dynamic Bayesian network of each sub-process is 2. N-1 =8.

[0089] Step 4: Perform structural learning and establish the first-layer network structure of the pyramid dynamic Bayesian network model;

[0090] After determining the number of time slices of the first layer of the local pyramid dynamic Bayesian network in each sub-process, the first layer dynamic Bayesian network structure is learned using the Bayesian information criterion scoring function in formula (2);

[0091]

[0092] In the formula, D represents the given data set; G represents the given structure; N represents the number of training samples; n is the number of nodes; I P (X i ; Pa(X i )) represents the node X in the given structure G i The mutual information with its parent node; Dim[G] is the number of independent parameters of model G; H P (X i ) represents node X i Entropy of

[0093] Step 5: Use the variable length data window to merge the same nodes of adjacent time slices in the first layer of dynamic Bayesian network into one node to construct the second layer of dynamic Bayesian network structure;

[0094] The same nodes of adjacent time slices in the first-layer dynamic Bayesian network are merged into one node using a variable-length data window, so that the nodes in the second-layer dynamic Bayesian network contain more data dynamic information. At the same time, the data window size is changed, and the Bayesian information criterion scoring function is reused for structural learning to obtain the second-layer dynamic Bayesian network structure of the pyramid. The use of a variable-length data window can avoid the loss of node data information during the node merging process.

[0095] Step 6: Repeat step 5 until the number of time slices of the dynamic Bayesian network is 2;

[0096] The time slice data is reduced by merging network nodes until the number of time slices is reduced to 2 and the merging process is stopped; the dynamic Bayesian network with 2 time slices is the simplest dynamic Bayesian network model, so when the number of time slices is reduced to 2, the network nodes are no longer merged.

[0097] Figure 7 , Figure 8 , Fig. 9 The local pyramid dynamic Bayesian network structures are respectively the raw coal processing subsystem, the heavy medium coal preparation subsystem and the concentrated flotation subsystem. Since the data information contained in the same nodes in each layer of the network structure is different, the dependencies between variables in time slices may be different. The dynamic Bayesian network can capture the dynamic characteristics between variables. Secondly, by changing the amount of data information in the node, the causal relationship caused by time delay can be gradually covered, thereby improving the accuracy of the model.

[0098] Finally, the number of time slices of the pyramid dynamic Bayesian network is 2 (K-1) ,2 (K-2) ,2 (K-3) ,...,2 1 ;

[0099] Step 7: Parameter learning of global dynamic network and local pyramid dynamic Bayesian network;

[0100] After the model structure learning is completed, parameter learning is required. The purpose of parameter learning is mainly to estimate the conditional probability table of variables in the network from historical data. Through parameter learning, the dynamic Bayesian network captures the dynamic relationship between variables in time series data; then, the variable state that causes abnormal working conditions is inferred through conditional probability distribution, and the optimal safety control decision plan is formulated; the conditional probability distribution of global quality indicator variables and local quality indicator variables in the last layer of the pyramid structure in each sub-process is shown in Tables 4 to 7.

[0101] Table 4: Conditional probability distribution of node S1 in subprocess 1

[0102]

[0103] Table 5: Conditional probability distribution of node S2 in subprocess 2

[0104]

[0105] Table 6: Conditional probability distribution of node S3 in subprocess 3

[0106]

[0107] Table 7: Conditional probability distribution of global quality indicator variable S

[0108]

[0109] In the table, the measured variable status is divided into 1, 2, and 3, which represent the corresponding variable value within the normal range, below the normal range, and above the normal range. Since the quality variables S, S1, S2, and S3 are used to measure whether abnormal conditions occur in the entire process or the corresponding sub-process, it is only necessary to divide their status into 1 and 2, representing normal and abnormal conditions, respectively. When abnormal conditions occur in the coal slime flotation process, the online abnormal data is input into the corresponding local pyramid dynamic Bayesian network model, and the abnormal evidence status is set to 2 to infer the root cause of the abnormality. Based on this, a scientific control decision plan is formulated.

[0110] Step 8: Collect actual operating condition data information, determine whether an abnormal operating condition occurs, and determine the sub-process that causes the abnormality when an abnormal operating condition occurs;

[0111] The actual operating data information of the global quality indicator variable (flotation tank overflow ash content S) is compared with the set safety threshold, where the safety threshold is set to 9.2%; when the flotation tank overflow ash content S ≤ 9.2%, it indicates that the coal slime flotation process is in a safe operating state; when the flotation tank overflow ash content S> 9.2%, it indicates that the coal slime flotation process is abnormal, and the online abnormal data is input into the global dynamic Bayesian network as evidence information, and the sub-process where the abnormality occurs is inferred based on the global network reasoning principle;

[0112] In order to verify the effectiveness of the distributed pyramid dynamic Bayesian network security control strategy, two cases can be simulated on the coal slime flotation process simulation platform, and the abnormal data can be input into the distributed pyramid Bayesian model. First, the abnormal sub-process is inferred by the global dynamic Bayesian network. In the global network reasoning process, the abnormal probability of each sub-process at the current moment and the previous moment is calculated, and the sub-process with the highest abnormal probability at the current moment and the previous moment is found; if the sub-process with the highest abnormal probability at the previous moment is adjacent to the sub-process with the highest abnormal probability at the current moment, and is before the sub-process with the highest abnormal probability at the current moment, it means that the abnormality of the current sub-process is caused by the abnormality of the previous sub-process, and the previous sub-process needs to be detected for abnormality, otherwise it is considered that the cause of the abnormality is the current sub-process;

[0113] The global network inference results are as follows Fig.10 and Fig.11 As shown. Taking Case 1 as an example, the sub-process with the highest probability of abnormality at the current moment is sub-process 3, and the sub-process with the highest probability of abnormality at the previous moment is sub-process 2. Therefore, the abnormality of sub-process 3 is caused by the abnormality of sub-process 2. It is necessary to detect the cause of the abnormality of sub-process 2 and formulate a reasonable control decision plan.

[0114] Step 9: Use the local pyramid dynamic Bayesian network to determine the root cause of the abnormal working condition;

[0115] After determining the abnormal sub-process, the local online abnormal data is input as evidence information into each layer of the local pyramid dynamic Bayesian network corresponding to the current moment, and reasoning is performed layer by layer from top to bottom to determine the abnormal diagnosis results of each layer of the pyramid dynamic Bayesian network according to the principle of maximum a posteriori probability;

[0116] Since the same nodes in different layers contain different data information, the inferred abnormal causes may be the same or different. When the local online abnormal data is input into each layer of the pyramid dynamic Bayesian network, it is input into the network at the current moment, and the inferred control decision is the previous moment, in order to improve the response speed of the model.

[0117] Step 10: Based on the abnormal causes diagnosed by the local pyramid dynamic Bayesian network, formulate and implement a reasonable control decision plan to eliminate abnormal conditions;

[0118] S101: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are the same, it means that the time delay in the process does not affect the fault location result and decision-making, and the control decision plan inferred based on the corresponding abnormal cause is directly implemented; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, it means that the abnormal condition is eliminated and the reasoning process ends; otherwise, execute step eleven;

[0119] S102: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are different, the control decision is implemented according to the following steps:

[0120] A1: First, use the adjustment plan inferred from the first level of the local pyramid to make operational adjustments. If the overflow ash content S of the flotation tank returns to the safety threshold, the reasoning process ends.

[0121] A2: After executing the adjustment plan of the first layer, if the coal slime flotation process has not returned to normal operation, the adjustment plan inferred by the second layer will be continued to be used for operation adjustment on the basis of the first layer operation adjustment; when implementing the adjustment plan of the second layer, only the newly emerged operation strategy will be adopted and executed, and the variables that have been adjusted in the first layer operation adjustment process will not be adjusted repeatedly; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, the reasoning process ends;

[0122] A3: After executing the adjustment plans of the first and second layers, if the coal slime flotation process has not resumed normal operation, the inference plans of the subsequent layers of the pyramid dynamic Bayesian network will continue to be adopted and executed. The execution process is the same as the execution process of the adjustment plan of the second layer. Only the newly appeared adjustment strategy will be executed, and the executed adjustment plan will not be repeated until the coal slime flotation process resumes normal operation. The inference process ends; when the adjustment plan inferred by the last layer (2 time slice dynamic Bayesian network) is executed, if the coal slime flotation process has not resumed normal operation, step eleven is executed.

[0123] The abnormal data of the two cases are input into the distributed pyramid dynamic Bayesian network, and the inferred control decisions are shown in Table 8. Among them, "↑" indicates that the variable adjustment direction increases, "ˉ" indicates that the adjustment direction decreases, and "-" indicates that it remains unchanged.

[0124] Table 8: Control decisions corresponding to the two cases

[0125]

[0126]

[0127] As shown in Table 8, in case one, subprocess 2 is adjusted. The result of the first level of reasoning is to increase the coal flow rate into the mixed medium barrel, and the rest remains unchanged. The result of the second level of reasoning is to increase the coal flow rate into the mixed medium barrel, reduce the density adjustment opening of the qualified medium barrel, and the rest remains unchanged. The result of the third level of reasoning is to increase the coal flow rate into the mixed medium barrel, reduce the density adjustment opening of the qualified medium barrel, and increase the opening of the cyclone pressure adjustment valve. The reasoning results of each level in case one are different and need to be adopted layer by layer. Case two adjusts subprocess 3, and the corresponding operation is to reduce the underflow flow of the concentrator and increase the stirring speed of the flotation tank.

[0128] Step 11: Select the sub-process with the second largest abnormal probability in the global reasoning process for reasoning, and re-execute step 9, and continue to use the online abnormal data as evidence information to infer the control decision plan until the abnormal condition is eliminated.

[0129] Furthermore, the method in this application can be verified by observing the change of the overflow ash S of the flotation tank after the control decision is implemented:

[0130] In order to verify the effectiveness of the distributed pyramid dynamic Bayesian network modeling method and case adjustment strategy, the effects of the two cases after the control decision-making can be observed on the coal slime flotation process simulation platform. When the abnormal working condition occurs, observe whether the overflow ash content of the flotation tank can be restored to the normal range. Fig.12 and Fig.13 As shown in the figure, the safety threshold of the overflow ash content of the flotation tank is set to 9.2%. When the threshold is exceeded, it is input as evidence into the established distributed pyramid dynamic Bayesian network model to infer the control decision plan. Fig.12 and Fig.13 It can be seen that in each case study, when an abnormality occurs, the application of the model proposed by the present invention can not only infer the control decision plan layer by layer and quickly eliminate the abnormal working condition, but also take into account the dynamic characteristics and time delay in the industrial process on the control decision making, thus improving the decision accuracy.

[0131] Aiming at the problem that the dynamic characteristics and time delay in the coal slime flotation process lead to frequent abnormal working conditions, the present invention proposes a coal slime flotation industrial process safety control method based on a distributed pyramid dynamic Bayesian network. The method uses a distributed strategy to refine the coal slime flotation process into multiple sub-processes, analyzes historical data and real-time abnormal data, combines process knowledge, establishes a causal relationship between global and local quality indicator variables, and constructs a global dynamic Bayesian network. At the same time, by learning the sub-process historical data information, the structural relationship between the local quality indicator variables and the relevant variables in each sub-process is determined, and a local pyramid dynamic Bayesian network is established. When an abnormality occurs, the sub-process where the abnormality occurs is inferred by inputting online abnormal data into the global Bayesian network, and the root cause of the abnormality is determined by using the local pyramid dynamic Bayesian network, and a reasonable control decision plan is formulated and implemented to eliminate the abnormal working conditions. By collecting actual working condition data information and combining the reasoning mechanism of the dynamic Bayesian network, the method can quickly respond to and handle abnormal conditions in the coal slime flotation process to ensure the safety and stability of the process. The effectiveness of the proposed method is proved by implementation verification in the coal slime flotation process.

Claims

1. A coal slime flotation process control method based on a distributed pyramid dynamic Bayesian network, comprising a coal slime flotation industrial system, wherein the coal slime flotation industrial system comprises a raw coal processing subsystem, a heavy medium coal preparation subsystem and a concentrated flotation subsystem; It is characterized in that The coal slime flotation process control method comprises the following steps: Step 1: Divide the coal slime flotation process into three sub-processes, determine the global quality index variables and local quality index variables, and establish a global dynamic Bayesian network; S11: Through in-depth analysis of historical data of coal slime flotation process, combined with rich process knowledge and the actual situation of coal slime flotation industrial system, the coal slime flotation process is divided into three sub-processes: raw coal processing, heavy medium coal preparation and concentrated flotation by adopting distributed thinking. The three sub-processes correspond to the raw coal processing subsystem, heavy medium coal preparation subsystem and concentrated flotation subsystem in the coal slime flotation industrial system respectively; S12: The overflow ash content S of the flotation tank is determined as the global quality indicator variable, and a quality indicator expected value is set for each sub-process as the respective local quality indicator variable. The local quality indicator variables of the three sub-processes are S1, S2, and S3 respectively; S13: Establish a global Bayesian network based on the causal relationship between the global quality indicator variables and the local quality indicator variables; S14: Connect each subprocess between different time slices to build a global dynamic Bayesian network; Step 2: Determine the key variables in the three sub-processes, collect process data, analyze the correlation between variables, and determine the process delay T in each process; S21: Combined with the knowledge of coal slime flotation process, the key variables in the three sub-processes are determined based on the ability to reflect product quality and ease of measurement, and are divided into operating variables and measurement variables; the changes in product indicators are reflected through the measurement variables in each sub-process, and the improvement of product quality is achieved by adjusting the operating variables; S22: using the sensor groups arranged in the raw coal processing subsystem, the heavy medium coal separation subsystem and the concentration flotation subsystem to respectively collect process data in the raw coal processing subprocess, the heavy medium coal separation subprocess and the concentration flotation subprocess in real time, and obtain time series data; S23: Use the maximum information coefficient method to analyze the time series data in the raw coal processing subprocess, the heavy medium coal preparation subprocess, and the concentration flotation subprocess, and determine the correlation between the variables; S24: approximate the time delay between the local quality indicator variables and the key variables in the three sub-processes according to the correlation results, and take the maximum time delay as the process delay T of the process; Step 3: According to the process delay T of each sub-process, use formula (1) to determine the number of time slices of the first layer of the dynamic Bayesian network in the local pyramid dynamic Bayesian network model; W＝2 (K-1) (1)； Where W represents the number of time slices of the first-layer dynamic Bayesian network of the pyramid; K = ceil(log2(T+1); ceil(·) represents the upward rounding function; Step 4: Perform structural learning and establish the first-layer network structure of the pyramid dynamic Bayesian network model; After determining the number of time slices of the first layer of the local pyramid dynamic Bayesian network in each sub-process, the first layer dynamic Bayesian network structure is learned using the Bayesian information criterion scoring function in formula (2); In the formula, D represents the given data set; G represents the given structure; N represents the number of training samples; n is the number of nodes; I P (X i ; Pa(X i )) represents the node X in the given structure G i The mutual information with its parent node; Dim[G] is the number of independent parameters of model G; H P (X i ) represents node X i Entropy of Step 5: Use the variable length data window to merge the same nodes of adjacent time slices in the first layer of dynamic Bayesian network into one node to construct the second layer of dynamic Bayesian network structure; The same nodes of adjacent time slices in the first layer of dynamic Bayesian network are merged into one node by using variable length data window, so that the nodes in the second layer of dynamic Bayesian network contain more data dynamic information; at the same time, the data window size is changed, and the Bayesian information criterion scoring function is reused for structural learning to obtain the second layer of pyramid dynamic Bayesian network structure; Step 6: Repeat step 5 until the number of time slices of the dynamic Bayesian network is 2; The time slice data is reduced by merging network nodes until the number of time slices is reduced to 2 and the merging process is stopped; finally, the number of time slices of the pyramid dynamic Bayesian network is 2 (K-1) ,2 (K-2) ,2 (K-3) ,...,2 1 ; Step 7: Parameter learning of global dynamic network and local pyramid dynamic Bayesian network; Perform parameter learning to enable the dynamic Bayesian network to capture the dynamic relationship between variables in time series data; then, infer the variable state that causes abnormal conditions through conditional probability distribution, and formulate the optimal safety control decision plan; Step 8: Collect actual operating condition data information, determine whether an abnormal operating condition occurs, and determine the sub-process that causes the abnormality when an abnormal operating condition occurs; The actual working condition data information of the global quality indicator variable is compared with the set safety threshold, where the safety threshold is set to 9.2%; when the flotation tank overflow ash content S≤9.2%, it means that the coal slime flotation process is in a safe operating state; when the flotation tank overflow ash content S>9.2%, it means that the coal slime flotation process is abnormal, and the online abnormal data is input into the global dynamic Bayesian network as evidence information, and the sub-process where the abnormality occurs is inferred based on the global network reasoning principle; in the global network reasoning process, the abnormal probability of each sub-process at the current moment and the previous moment is calculated, and the sub-process with the highest abnormal probability at the current moment and the previous moment is found; if the sub-process with the highest abnormal probability at the previous moment is adjacent to the sub-process with the highest abnormal probability at the current moment, and is before the sub-process with the highest abnormal probability at the current moment, it means that the abnormality of the current sub-process is caused by the abnormality of the previous sub-process, and the previous sub-process needs to be detected for abnormality, otherwise it is considered that the cause of the abnormality is the current sub-process; Step 9: Use the local pyramid dynamic Bayesian network to determine the root cause of the abnormal working condition; After determining the abnormal sub-process, the local online abnormal data is input as evidence information into each layer of the local pyramid dynamic Bayesian network corresponding to the current moment, and reasoning is performed layer by layer from top to bottom to determine the abnormal diagnosis results of each layer of the pyramid dynamic Bayesian network according to the principle of maximum a posteriori probability; Step 10: Based on the abnormal causes diagnosed by the local pyramid dynamic Bayesian network, formulate and implement a reasonable control decision plan to eliminate abnormal conditions; S101: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are the same, it means that the time delay in the process does not affect the fault location result and decision-making, and the control decision plan inferred based on the corresponding abnormal cause is directly implemented; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, it means that the abnormal condition is eliminated and the reasoning process ends; otherwise, execute step eleven; S102: When the abnormal causes diagnosed by each layer of the local pyramid dynamic Bayesian network are different, the control decision is implemented according to the following steps: A1: First, use the adjustment plan inferred from the first level of the local pyramid to make operational adjustments. If the overflow ash content S of the flotation tank returns to the safety threshold, the reasoning process ends. A2: After executing the adjustment plan of the first layer, if the coal slime flotation process has not returned to normal operation, the adjustment plan inferred by the second layer will be continued to be used for operation adjustment on the basis of the first layer operation adjustment; when implementing the adjustment plan of the second layer, only the newly emerged operation strategy will be adopted and executed, and the variables that have been adjusted in the first layer operation adjustment process will not be adjusted repeatedly; at the same time, observe whether the overflow ash content S of the flotation tank returns to the safety threshold. If the overflow ash content S of the flotation tank drops to the safety threshold, the reasoning process ends; A3: After executing the adjustment plans of the first and second layers, if the coal slime flotation process has not resumed normal operation, the inference plans of the subsequent layers of the pyramid dynamic Bayesian network will continue to be adopted and executed. The execution process is the same as the execution process of the adjustment plan of the second layer. Only the newly appeared adjustment strategy will be executed, and the executed adjustment plan will not be repeated until the coal slime flotation process resumes normal operation. The inference process ends; when the adjustment plan inferred by the last layer is executed, if the coal slime flotation process has not resumed normal operation, execute step eleven. Step 11: Select the sub-process with the second largest abnormal probability in the global reasoning process for reasoning, and re-execute step 9, and continue to use the online abnormal data as evidence information to infer the control decision plan until the abnormal condition is eliminated.

Images