Coal mill unbalanced data fault diagnosis method based on Bayesian network

Abstract

The invention discloses a coal mill unbalanced data fault diagnosis method based on a Bayesian network, and belongs to the field of generalized zero sample fault diagnosis of a thermal power generation feed pump set. Aiming at the problems of unbalanced fault data distribution, difficult minority class fault diagnosis and poor result interpretability in an industrial scene, sample distribution is collaboratively optimized through SMOTE data enhancement and Dirichlet prior smoothing, a three-layer causal topology network of'fault layer-attribute layer-observation layer 'is constructed, expert knowledge constraint and data-driven learning are combined, and the fault layer-attribute layer-observation layer-based fault diagnosis method is established. High-precision fault classification is realized; a dual-mode diagnosis mechanism is designed, fault node direct inference and attribute node indirect inference are synchronously supported, and explainable physical-level diagnosis guidance is provided while the fault diagnosis accuracy is guaranteed. The method is successfully applied to a coal mill system of a coal-fired unit, and can be expanded to transparent intelligent operation and maintenance of complex industrial equipment.

Description

Technical Field

[0001] The invention belongs to the field of generalized zero-sample fault diagnosis of a thermal power generation water supply pump group and provides a coal mill unbalanced data fault diagnosis method based on a Bayesian network. Background Art

[0002] The coal pulverizing system is an important auxiliary equipment of coal-fired power plants. Its main function is to grind raw coal into coal powder that meets the combustion requirements and transport it to the boiler through primary air. The operating status of the coal pulverizing system directly affects the boiler combustion efficiency, the service life of the equipment, and the overall operating performance of the power plant. Once a coal pulverizer fails, it may cause the coal powder fineness to fail to meet the standard, reduce the boiler combustion efficiency, and even cause serious equipment damage and economic losses. Therefore, the development of an efficient and accurate coal pulverizer fault diagnosis method is of great engineering significance for improving the safety, economy, and operating efficiency of power plants.

[0003] In actual industrial scenarios, coal mill fault diagnosis faces a significant challenge, namely, the uneven distribution of data for different fault categories. Specifically, some fault categories occur more frequently, so the corresponding fault data is relatively sufficient; while other fault categories have a low probability of occurrence or limited collection conditions, resulting in extremely scarce data. This imbalance in data distribution makes the diagnosis model more inclined to learn the characteristics of fault categories with sufficient data during training, while ignoring fault categories with scarce data, thereby reducing the diagnostic accuracy of a few types of faults. In addition, multiple fault types in the coal mill pulverizing system may occur simultaneously or switch rapidly, further increasing the complexity of fault diagnosis.

[0004] At present, methods such as deep learning, statistical analysis and rule-based expert systems have been widely used in the field of fault diagnosis, but there are still the following limitations in dealing with the above problems: Deep learning relies on large-scale data for training and is good at extracting complex feature patterns from massive data. However, it has high requirements for the balance of data volume and category distribution. In the scenario of imbalanced fault category data, deep learning models often have the problem of overfitting the majority class, resulting in insufficient diagnosis ability for minority class faults. In addition, the "black box" characteristics of deep learning models limit their interpretability in the industrial field, making it difficult to provide engineers with clear fault tracing information. Traditional statistical methods and rule-based diagnostic methods rely more on domain expert knowledge and can provide certain interpretability. However, these methods have high requirements for the balance of data distribution and sample size, and are difficult to adapt to the diverse fault diagnosis needs in industrial environments, especially when dealing with minority class faults with scarce data.

[0005] In order to solve the problem of uneven distribution of fault category data, some studies in recent years have introduced data enhancement technology and prior knowledge fusion technology: Data enhancement technology (such as Synthetic Minority Over-sampling Technique, SMOTE) expands the data distribution of minority class faults by generating virtual samples, thereby alleviating the problem of data imbalance. Although SMOTE technology has improved the diagnostic ability of minority class faults to a certain extent, the virtual samples it generates may not fully reflect the real fault mode, especially in high-dimensional and multi-fault scenarios, which may affect the diagnostic accuracy and robustness of the model. Bayesian networks combine domain knowledge with data-driven methods through causal relationship modeling, making the diagnostic model highly interpretable. However, in complex industrial scenarios, how to effectively construct a Bayesian network structure and take into account both the diagnostic accuracy of minority class faults and the robustness of majority class faults is still a technical challenge. In addition, the complexity of equipment operation in industrial production environments further exacerbates the difficulty of fault diagnosis. For example, factors such as equipment aging, environmental fluctuations, and improper operation may lead to diversity and uncertainty of fault modes and cause concurrent or rapid switching of multiple fault categories. Summary of the invention

[0006] In view of the technical problems existing in the prior art, the present invention provides a coal mill imbalance data fault diagnosis method based on a Bayesian network, which can balance the diagnostic performance of minority faults and majority faults in complex scenarios where multiple faults coexist, and achieve efficient interpretability of the diagnostic results.

[0007] In order to achieve the above purpose, the technical solution adopted by the present invention is: 1. Offline data processing Step 1: Clean the faulty sample; Specifically, the original data were cleaned according to the standards of completeness, i.e., missing rate <5%, legality, i.e., values ​​within the physical range, and deletion of redundant records. The interquartile range (IQR) method was used to detect extreme outliers, and the anomaly thresholds were defined as Q1-1.5IQR and Q3+1.5IQR. Data out of the range would be eliminated.

[0008] Step 2: Imbalanced data enhancement; Specifically, according to the fault label, the majority class and the minority class are statistically distinguished, and the minority class samples are augmented using SMOTE: for each minority class sample x i , search for k nearest neighbor minority class samples from the feature space; randomly select one of the nearest neighbor samples x j , and perform linear interpolation to generate new sample points X new =x i +λ(x j -x i),λ~U(0,1), where λ is the random interpolation coefficient; repeat until the number of minority class samples is the same as the number of majority class samples.

[0009] Step 3: Data normalization and discretization; Specifically, the Min-Max method is used to map the continuous sensor data to the [0,1] interval; the normalized data is divided into 5 discrete intervals based on the equal-frequency binning method to ensure that the sample size in each interval is evenly distributed.

[0010] 2. Offline knowledge fusion Step 4: Fault-attribute mapping modeling; Specifically, we first define the key abnormal attributes of each fault category through domain knowledge; then we construct a fault-attribute matrix, which is specifically expressed by the following formula: ; where rows correspond to fault categories, columns correspond to attributes, m ij =1 indicates fault Will cause attributes , otherwise 0.

[0011] Step 5: Three-layer Bayesian network architecture design; The specific structure is as follows: Fault layer, i.e. top layer: consists of all fault category nodes, representing possible fault modes of the system; Fault attribute layer, i.e., the middle layer: fault attribute nodes extracted based on domain knowledge, describing typical characteristics of faults; Fault variable layer, i.e. bottom layer: contains monitoring variable nodes, whose values ​​are derived from sensor data after discretization, directly reflecting the real-time status of the system.

[0012] 3. Offline model learning Step 6: Bayesian network structure learning (1) Constructing the initial skeleton: Forced connection: Fixed the directed edge of "fault node → fault attribute node" based on expert knowledge, corresponding to the logical dependency of the fault-attribute matrix; (2) Open search: Allows data-driven algorithms to explore the potential causal relationship between “fault attribute node → monitoring variable node” and attribute / variable nodes.

[0013] Greedy structure search: The Hill-Climbing algorithm is used to iteratively perform edge addition, edge deletion, and edge reversal operations, and the BIC scoring function is used to balance the model accuracy and complexity: ; where logP(D|G) is the log-likelihood; |θ| is the total number of network parameters, and N is the number of samples.

[0014] (3) Industrial mechanism constraints: Direction constraint: prohibits attribute / variable nodes from pointing in the opposite direction to the fault node; Whitelist: Forces the retention of directed edges determined by process logic, such as attribute A → variable B; Blacklist: prohibits connections between physically irrelevant nodes, such as temperature attribute → vibration variable; Step 7: Bayesian network parameter learning; (1) Calculate the joint distribution frequency of the node and its parent node based on the training data to generate the initial CPT; (2) To avoid the zero probability problem, Dirichlet prior smoothing is used and a moderate prior smoothing coefficient is introduced in each value combination: ; where N(X i ,Pa(X i )) is X i The joint frequency with its parent node, α is the prior smoothing coefficient.

[0015] Step 8: Dual-mode diagnostic strategy; (1) Direct inference of faulty nodes: Input real-time monitoring variable evidence and calculate the posterior probability of the faulty node; if the maximum posterior probability exceeds the set threshold τ1, the corresponding fault category is output; otherwise, it is marked as “unknown fault”.

[0016] (2) Indirect inference of attribute nodes: Calculate the posterior probability of the fault attribute node and binarize it into an abnormal pattern; perform similarity matching with the fault-attribute matrix, with cosine similarity > τ2, and output the best matching fault or "unknown fault".

[0017] Step 9: Joint optimization of parameters; Through grid search combined with K-fold cross validation, joint optimization is performed to select [α,τ 1, τ2] optimal configuration, taking the weighted score F1-score of diagnostic accuracy and false alarm rate as the optimization goal, and determining the optimal parameter combination.

[0018] Step 10: Model solidification and deployment; The optimal network structure, CPT parameters and diagnostic thresholds are locked and exported as a lightweight inference engine; integrated into the coal mill monitoring system to support real-time inference at the rate of hundreds of times per second.

[0019] 4. Online fault diagnosis Step 11: Online Diagnostic Process (1) Data preprocessing: Real-time sensor data is normalized, binned, and mapped to observation layer node values; (2) Dual-mode simultaneous inference: Direct inference mode: calculate the posterior probability of the faulty node and output the maximum probability fault category; Indirectly infer patterns: calculate the posterior probability of attribute nodes, generate abnormal patterns and match them with the fault-attribute matrix.

[0020] The beneficial effects of the present invention are: By balancing and discretizing the data in the offline stage, it can handle complex industrial scenarios where multiple faults exist in parallel or switch rapidly. The Bayesian network combines domain expert knowledge with data-driven, and the network structure itself reflects the causal relationship between faults, attributes, and monitoring variables, which is convenient for subsequent analysis and tracing. The use of Dirichlet prior smoothing effectively reduces the negative impact of unbalanced data on parameter learning and improves the accuracy and stability of diagnosis. Online reasoning is efficient: the reasoning mechanism based on the Bayesian network only needs to update and propagate the given evidence in the network, and the amount of calculation is relatively controllable, which can realize real-time or quasi-real-time online diagnosis of faults.

[0021] Through the above steps and method configuration, the recognition and diagnosis accuracy of multi-fault scenarios can be significantly improved, meeting the reliability and explainability requirements of industrial systems. This method not only performs well in scenarios with many types of faults and unbalanced data, but can also be extended to other similar industrial process monitoring and fault diagnosis tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The figure is a flow chart of a method for diagnosing coal mill imbalance data fault based on Bayesian network in an embodiment of the present invention.

[0023] Figure 2 Schematic diagram of the Bayesian network structure in an embodiment of the present invention.

[0024] Figure 3 The figure is a schematic diagram of the coal pulverizing process flow in the embodiment of the present invention.

[0025] Figure 4 Schematic diagram of coal mill fault-attribute matrix in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0027] In view of the existing problems in the prior art, this paper proposes a method for diagnosing coal mill unbalanced data faults based on Bayesian networks. In view of the problem of uneven distribution of fault category data in industrial scenarios, this paper combines the data enhancement technology SMOTE and domain prior knowledge to solve the technical pain point of difficulty in diagnosing minority class faults. Through offline data cleaning, sample balancing, normalization and discretization processing, and domain knowledge combing to construct a fault-attribute matrix, combined with Bayesian network causal modeling, a method is established. Figure 2 The three-layer Bayesian network model structure is shown in Figure 1, and the network structure is optimized by the Hill-Climbing algorithm. In the parameter learning stage, the Dirichlet prior smoothing method is used to alleviate the impact of small samples and zero probability problems on the model and improve the stability and robustness of the diagnosis model. Finally, the optimized model is deployed in an online environment, and real-time or quasi-real-time fault diagnosis is achieved by combining direct inference of fault nodes and indirect inference of attribute nodes, which significantly improves the accuracy and interpretability of diagnosis.

[0028] The core goal of this invention is to solve the challenges of uneven distribution of fault category data, difficulty in diagnosing a few types of faults, and coexistence of multiple faults in complex industrial environments in industrial fault diagnosis. By introducing SMOTE data enhancement and Dirichlet prior smoothing technology, the diagnostic capability of a few types of faults is improved; through the causal modeling capability of the Bayesian network, the high interpretability of the diagnostic results is guaranteed; and through the online reasoning mechanism, a balance between real-time performance and diagnostic robustness is achieved, ultimately providing accurate and reliable fault diagnosis solutions for industrial systems to ensure safe and efficient operation of equipment.

[0029] Specifically, the present invention provides a method for diagnosing coal mill imbalance data faults based on a Bayesian network. Figure 1 , including the following steps: 1. Offline data processing Step 1: Clean the faulty sample; Specifically, the original data were cleaned according to the standards of completeness, i.e., missing rate <5%, legality, i.e., the values ​​were within the physical range, and redundant records were deleted; the interquartile range IQR method was used to detect extreme outliers, and the abnormal thresholds were defined as Q1-1.5IQR and Q3+1.5IQR, and those beyond the range were eliminated; Step 2: Imbalanced data enhancement; According to the fault label, the majority class and the minority class are statistically distinguished, and the minority class samples are augmented using SMOTE: For each minority class sample x i , search for k nearest neighbor minority class samples from the feature space; randomly select one of the nearest neighbor samples x j , and perform linear interpolation to generate new sample points X new =x i+λ(x j -x i ),λ~U(0,1), where λ is the random interpolation coefficient; repeat until the number of minority class samples is the same as the number of majority class samples.

[0030] Step 3: Data normalization and discretization; Specifically, the Min-Max method is used to map the continuous sensor data to the [0,1] interval; the normalized data is divided into 5 discrete intervals based on the equal-frequency binning method to ensure that the sample size in each interval is evenly distributed.

[0031] 2. Offline knowledge fusion Step 4: Fault-attribute mapping modeling; Specifically, we first define the key abnormal attributes of each fault category through domain knowledge; then we construct a fault-attribute matrix, which is specifically expressed by the following formula: ; Where rows correspond to fault categories and columns correspond to attributes. ij =1 indicates fault Will cause attributes , otherwise 0.

[0032] Step 5: Three-layer Bayesian network architecture design; The specific structure is as follows: Fault layer, i.e. top layer: consists of all fault category nodes, representing possible fault modes of the system; Fault attribute layer, i.e., the middle layer: fault attribute nodes extracted based on domain knowledge, describing typical characteristics of faults; Fault variable layer, i.e. bottom layer: contains monitoring variable nodes, whose values ​​are derived from sensor data after discretization, directly reflecting the real-time status of the system.

[0033] 3. Offline model learning Step 6: Bayesian network structure learning (1) Constructing the initial skeleton: Forced connection: Fixed the directed edge of "fault node → fault attribute node" based on expert knowledge, corresponding to the logical dependency of the fault-attribute matrix; (2) Open search: Allows data-driven algorithms to explore the potential causal relationship between “fault attribute node → monitoring variable node” and attribute / variable nodes.

[0034] Greedy structure search: The Hill-Climbing algorithm is used to iteratively perform edge addition, edge deletion, and edge reversal operations, and the BIC scoring function is used to balance the model accuracy and complexity: ; where logP(D|G) is the log-likelihood; |θ| is the total number of network parameters, and N is the number of samples.

[0035] (3) Industrial mechanism constraints: Direction constraint: prohibits attribute / variable nodes from pointing in the opposite direction to the fault node; Whitelist: Forces the retention of directed edges determined by process logic, such as attribute A → variable B; Blacklist: prohibits connections between physically unrelated nodes, such as temperature attribute → vibration variable.

[0036] Step 7: Bayesian network parameter learning; (1) Calculate the joint distribution frequency of the node and its parent node based on the training data to generate the initial CPT; (2) To avoid the zero probability problem, Dirichlet prior smoothing is used and a moderate prior smoothing coefficient is introduced in each value combination: ; where N(X i ,Pa(X i )) is X i The joint frequency with its parent node, α is the prior smoothing coefficient.

[0037] Step 8: Dual-mode diagnostic strategy; (1) Direct inference of faulty nodes: Input real-time monitoring variable evidence and calculate the posterior probability of the faulty node; if the maximum posterior probability exceeds the set threshold τ1, the corresponding fault category is output; otherwise, it is marked as “unknown fault”.

[0038] (2) Indirect inference of attribute nodes: Calculate the posterior probability of the fault attribute node and binarize it into an abnormal pattern; perform similarity matching with the fault-attribute matrix, with cosine similarity > τ2, and output the best matching fault or "unknown fault".

[0039] Step 9: Joint optimization of parameters; Through grid search combined with K-fold cross validation, joint optimization is performed to select [α,τ 1, τ2] optimal configuration, taking the weighted score F1-score of diagnostic accuracy and false alarm rate as the optimization goal, and determining the optimal parameter combination.

[0040] Step 10: Model solidification and deployment; The optimal network structure, CPT parameters and diagnostic thresholds are locked and exported as a lightweight inference engine; integrated into the coal mill monitoring system to support real-time inference at the rate of hundreds of times per second.

[0041] 4. Online fault diagnosis Step 11: Online Diagnostic Process (1) Data preprocessing: Real-time sensor data is normalized, binned, and mapped to observation layer node values; (2) Dual-mode simultaneous inference: Direct inference mode: calculate the posterior probability of the faulty node and output the maximum probability fault category; Indirectly infer patterns: calculate the posterior probability of attribute nodes, generate abnormal patterns and match them with the fault-attribute matrix.

[0042] The key innovation of the present invention is that through the dual-stage diagnosis mode combining offline and online, SMOTE minority class enhancement, prior smoothing and three-layer Bayesian network structure are organically integrated, which can not only obtain a more balanced fault identification effect under the condition of small samples and data imbalance, but also make full use of industrial prior knowledge to construct a fault-attribute matrix, thereby achieving high-precision and strong interpretability diagnosis in multi-fault concurrent scenarios. It is particularly noteworthy that this method designs two paths, direct inference and indirect inference, for the fault layer, attribute layer and variable layer respectively, which can not only quickly output the posterior probability of the fault node, but also trace the specific abnormal mechanism based on the attribute node, making the diagnosis result more intuitive and transparent, and convenient for subsequent maintenance and troubleshooting.

[0043] At the same time, the present invention reduces manual parameter adjustment deviations through adaptive threshold search, i.e., grid search and cross-validation optimization parameter configuration, thus ensuring online diagnostic efficiency and model stability. Its versatility or adaptability is also quite outstanding: without changing the basic principles, it only needs to replace the domain knowledge and monitoring data in the offline stage, and can be widely used in different industrial scenarios such as chemical industry, metallurgy, electric power and machinery manufacturing, providing efficient and flexible solutions for the timely discovery and accurate positioning of diversified fault types. The present invention is further described in detail below through specific embodiments.

[0044] (1) Experimental scenario and data configuration The coal pulverizing system of a supercritical coal-fired unit is taken as the research object. The simplified principle diagram of the coal pulverizing process is as follows: Figure 3 As shown, the raw coal enters the coal feeder from the raw coal hopper through the coal drop pipe of the raw coal hopper, is transmitted to the coal drop pipe of the coal feeder, and enters the coal mill to be ground into coal powder. The cold and hot primary air are mixed by the regulating valve and blown into the coal mill. The mixed primary air carries the ground coal powder and is blown into the boiler.

[0045] Twelve key monitoring variables are selected, including the opening of the cold air valve, the opening of the hot air valve, the coal inlet flow rate, the moisture content of raw coal, the primary air volume, the primary air temperature, the amount of raw coal in the pulverizer, the amount of pulverized coal in the pulverizer, the pulverizer current, the amount of pulverized coal blown out, the outlet temperature, and the moisture content in the pulverized coal. Six types of typical faults are defined, as shown in Table 1, and eight fault attributes are defined, as shown in Table 2. The training set simulates the imbalanced industrial data scenario. The majority class faults contain 480 samples and the minority class faults contain 20 samples, with a ratio of 1:120. The test set uses balanced data to verify the generalization ability. Through ablation experiments and multi-scenario tests, the independent and joint effects of SMOTE data enhancement and Dirichlet prior smoothing are compared, and the effectiveness of the proposed method is verified. The schematic diagram of the pulverizer fault-attribute matrix is ​​shown in Figure 4 shown.

[0046] Table 1 Coal mill fault types and fault descriptions

[0047] Table 2 Coal mill fault attribute description table

[0048] On this basis, steps 1 to 11 are used to implement fault diagnosis of the coal mill process.

[0049] (2) Ablation Experiment Results In the ablation experiment, F6 is used as the minority fault and the other five are the majority faults to verify the effects of SMOTE data enhancement and Dirichlet prior smoothing on model performance. The experimental results are shown in Table 3.

[0050] Table 3 Ablation experiment results

[0051] According to the results, the recall rate and F1 score of the traditional MLE method for minority classes are only 0.79-0.84, indicating that it is difficult to deal with the problem of data imbalance. After introducing SMOTE alone, the F1 score of F6 increased to 0.90, proving that data enhancement effectively expanded the feature space of minority classes; when only Dirichlet prior smoothing was used, the F1 score did not increase significantly, and even decreased slightly, indicating that smoothing technology needs to work in synergy with data enhancement. When SMOTE and Dirichlet smoothing were applied together, the model's diagnostic performance for F6 reached the optimal level (F1=0.985), and the recall rate increased to 98.5%, significantly reducing the risk of missed detection.

[0052] Robustness verification of multiple minority fault scenarios As shown in Table 4, in a single minority class fault scenario (such as F1-F6 as minority classes), the model F1 score is stable above 0.92, and the fault diagnosis accuracy of both minority and majority classes can be guaranteed. As shown in Table 5, when multiple minority class fault complex scenarios are further set (such as 2-3 minority class faults), the model still maintains high robustness: F1>0.98 for 2 minority class faults, and F1=1 for 3 minority class faults, indicating that SMOTE can generate synthetic samples for each class independently to avoid feature confusion. It is worth noting that when the attribute features of multiple minority class faults are significantly different, such as F2 "bearing overheating" and F5 "powder outlet pipe blockage", the model achieves accurate distinction by strengthening their respective causal paths.

[0053] Table 4. Experimental results of single minority fault

[0054] Table 5 Experimental results of multiple minority fault classes

[0055] (3) Explainability verification and industrial application value The attribute layer diagnosis mode reveals the fault mechanism through the combination of abnormal attributes. For example, when "abnormal coal powder fineness A7 + increased outlet pressure A5" is detected, it is determined that the F5 powder outlet pipe is blocked, prompting to check the grinding parts or clean the pipeline; "abnormal bearing temperature A3 + increased vibration A2" indicates that the F2 bearing is overheated or damaged, and the lubrication system should be repaired first. Compared with the direct probability inference F1=0.985, the indirect inference mode F1=0.9287 has a slightly lower accuracy, but it provides physical-level attribution explanations to support engineers to quickly locate the root cause of the fault. False detection cases can be traced back to specific attribute abnormalities, such as "when F4 is misjudged, it is actually caused by A8 pressure difference fluctuations."

[0056] Comprehensive experimental results show that the method proposed in this study, which integrates SMOTE enhancement, Dirichlet smoothing and causal network, achieves higher accuracy in coal mill fault diagnosis under data imbalance scenarios, which is significantly improved compared with traditional methods. Its hierarchical diagnosis mechanism, rapid warning at the fault layer → physical attribution at the attribute layer, takes into account both real-time and explainability, and provides an innovative path from "black box prediction" to "transparent decision-making" for industrial scenarios. In practical applications, this method has successfully reduced 70% of unplanned downtime, verifying its engineering practical value.

[0057] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above invention concept. For example, the above features are replaced with the technical features disclosed in the embodiments of the present disclosure, but not limited to the technical features with similar functions.

Claims

1. A method for diagnosing coal mill imbalance data faults based on Bayesian networks, characterized in that: Step 1: Offline data processing: preprocess the collected data, clean the fault samples, enhance the unbalanced data, and normalize and discretize the data; Step 2: Offline knowledge fusion: Model the fault-attribute mapping and construct a three-layer Bayesian network for fault classification; Step 2.1: Fault-attribute mapping modeling: Define the key abnormal attributes of each fault category through domain knowledge; then construct the fault-attribute matrix, which is expressed by the following formula: ; where rows correspond to fault categories, columns correspond to attributes, m ij =1 indicates fault Will cause attributes The obvious abnormality is 0 otherwise; Step 2.1: Three-layer Bayesian network architecture design: Fault layer, i.e. top layer: consists of all fault category nodes, representing the possible fault modes of the system; Fault attribute layer, i.e., the middle layer: fault attribute nodes extracted based on domain knowledge, describing typical characteristics of faults; Fault variable layer, i.e. bottom layer: contains monitoring variable nodes, whose values ​​are derived from sensor data after discretization, reflecting the real-time status of the system; Step 3: Offline model learning: Learn the Bayesian network structure and Bayesian network parameters, mark fault information through the dual-mode diagnosis strategy, optimize the parameter combination, and obtain the final model; 3.1: Bayesian network structure learning: 3.1.1: Build the initial skeleton: Forced connection: Fixed the directed edge of "fault node → fault attribute node" based on expert knowledge, corresponding to the logical dependency of the fault-attribute matrix; Open search: Allows data-driven algorithms to explore the potential causal relationship between "fault attribute node → monitoring variable node" and attribute / variable nodes; 3.1.2: Greedy structure search: The Hill-Climbing algorithm is used to iteratively perform edge addition, edge deletion, and edge reversal operations, and the BIC scoring function is used to balance the model accuracy and complexity: ; Where logP(D|G) is the log likelihood; |θ| is the total number of network parameters, and N is the number of samples; 3.1.3: Industrial mechanism constraints: Direction constraint: prohibits attribute / variable nodes from pointing in the opposite direction to the fault node; Whitelist: Forces the retention of directed edges determined by process logic; Blacklist: prohibits connections between physically unrelated nodes; 3.2: Bayesian network parameter learning: 3.2.1: Calculate the joint distribution frequency of the node and its parent node based on the training data to generate the initial CPT; 3.2.2: To avoid the zero probability problem, use Dirichlet prior smoothing and introduce a moderate prior smoothing coefficient in each value combination: ; where N(X i ,Pa(X i )) is X i The joint frequency with its parent node, α is the prior smoothing coefficient; 3.3: Dual-mode diagnostic strategy: 3.3.1: Direct inference of faulty nodes: Input real-time monitoring variable evidence and calculate the posterior probability of the faulty node; if the maximum posterior probability exceeds the set threshold τ1, the corresponding fault category is output; otherwise, it is marked as "unknown fault"; 3.3.2: Indirect inference of attribute nodes: Calculate the posterior probability of the fault attribute node and binarize it into an abnormal pattern; perform similarity matching with the fault-attribute matrix, i.e., cosine similarity>τ2, and output the most matching fault or "unknown fault"; 3.4: Joint Parameter Optimization: Through grid search combined with K-fold cross validation, joint optimization is performed to select [α,τ 1, τ2] optimal configuration, taking the weighted score F1-score of diagnostic accuracy and false alarm rate as the optimization target, and determining the optimal parameter combination; 3.5: Model solidification and deployment: The optimal network structure, CPT parameters and diagnostic thresholds are locked and exported as a lightweight inference engine; integrated into the coal mill monitoring system to support real-time inference at the rate of hundreds of times per second; Step 4: Perform online diagnosis.

2. A method for diagnosing coal mill imbalance data faults based on Bayesian network according to claim 1, characterized in that: In the step 1, the specific steps are: Step 1.1: Fault sample cleaning: Clean the original data according to the standards of completeness, i.e., missing rate <5%, legality, i.e., values ​​within the physical range, and deletion of redundant records; use the interquartile range IQR method to detect extreme outliers, define the anomaly thresholds as Q1-1.5IQR and Q3+1.5IQR, and exclude those beyond the range; Step 1.2: Imbalanced data enhancement: According to the fault label, statistically distinguish the majority class and the minority class, and use SMOTE to augment the minority class samples: For each minority class sample x i , search for k nearest neighbor minority class samples from the feature space; randomly select one of the nearest neighbor samples x j , and perform linear interpolation to generate new sample points X new =x i +λ(x j -x i ),λ~U(0,1), where λ is the random interpolation coefficient; repeat until the number of minority class samples is the same as the number of majority class samples; Step 1.3: Data normalization and discretization: Use the Min-Max method to map the continuous sensor data to the [0,1] interval; The normalized data were divided into five discrete intervals based on the equal frequency binning method to ensure that the sample size in each interval was evenly distributed.

3. A method for diagnosing coal mill imbalance data faults based on Bayesian network according to claim 1, characterized in that: In the step 4, the specific steps are: 4.1: Data preprocessing: Real-time sensor data is normalized, binned and mapped to observation layer node values; 4.2: Dual-mode simultaneous inference: Direct inference mode: calculate the posterior probability of the faulty node and output the maximum probability fault category; Indirectly infer patterns: calculate the posterior probability of attribute nodes, generate abnormal patterns and match them with the fault-attribute matrix.

Images