An industrial furnace working condition identification method and system fusing process variables and expert knowledge

Abstract

This invention relates to the field of industrial furnace and kiln condition identification technology, and discloses an industrial furnace and kiln condition identification method and system that integrates process variables and expert knowledge. The method establishes an objective evidence theory model based on process variables and a subjective evidence theory model based on expert knowledge, respectively. It adopts conflict redistribution and evidence secondary fusion algorithms to identify industrial furnace and kiln conditions in a data-knowledge joint-driven manner, providing a reliable guarantee for the stable operation and safe longevity of industrial furnaces and kilns.

Description

Technical Field

[0001] This invention relates to the field of industrial kiln identification technology, and in particular to a method and system for identifying industrial kiln operating conditions by integrating process variables and expert knowledge. Background Technology

[0002] Industrial furnaces and kilns, as core equipment in industrial production, directly affect product quality, output, safety, and service life. Industrial furnace and kiln condition identification refers to identifying their operating status by monitoring and analyzing their condition parameters, combined with professional knowledge, and assessing the risk of abnormal operating conditions. As one of the most crucial pieces of equipment in many industrial production processes, industrial furnaces and kilns experience high frequency of control and significant fluctuations in raw material composition, which can lead to abnormal situations and pose a significant challenge to their stable operation and long-term safety.

[0003] Data-driven and expert knowledge-based methods are two commonly used approaches for identifying the operating conditions of industrial furnaces and kilns. The former utilizes statistical analysis of input and output data generated during furnace and kiln operation to build a data-driven model. However, this method is limited by the need for a large number of samples for training to achieve a certain predictive accuracy. In reality, abnormal operating condition data is often scarce, making it difficult to train a model with high generalization performance. Expert knowledge-based methods integrate professional knowledge into the model, modeling the principles of the smelting process and calculating key indicators to detect operating conditions. However, this method requires numerous constraints, and some ideal assumptions may not hold true in practice. Furthermore, mechanistic models typically involve complex equations, making analytical solutions difficult to obtain, and many model parameters often cannot be updated online in a timely manner. This may result in the formed theory not effectively guiding the identification of industrial furnace and kiln operating conditions.

[0004] Meanwhile, in the process of identifying the operating conditions of industrial furnaces and kilns, when the information obtained and utilized by on-site personnel is insufficient to accurately describe the objective situation, cognitive uncertainty arises. This uncertainty can be summarized as ambiguity, polysemy, and conflicting evidence from multiple sources. Therefore, simply considering process variables or expert knowledge and attempting to directly integrate multiple pieces of contradictory evidence may lead to inaccurate operating condition identification, or even conclusions that contradict intuition. Summary of the Invention

[0005] This invention provides a method and system for identifying the operating conditions of industrial kilns by integrating process variables and expert knowledge, in order to solve the problem of low accuracy in existing industrial kiln operating condition identification methods.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] In a first aspect, the present invention provides a method for identifying the operating conditions of industrial furnaces and kilns by integrating process variables and expert knowledge, comprising:

[0008] S1: Screen the key process variables of the industrial furnace process variable set, and extract the dynamic characteristics of the key process variables within a preset time window;

[0009] S2: Combine the dynamic features with the NuSVC model and probability mapping function to construct an objective evidence theoretical model, and obtain the objective evidence source based on the objective evidence theoretical model;

[0010] S3: Construct logical rules and membership functions for key process variables based on expert knowledge of industrial furnaces and kilns. Based on the logical rules, membership functions, and dynamic characteristics of the key process variables, construct a subjective evidence theory model. Obtain a subjective evidence source containing multiple subjective evidences based on the subjective evidence theory model.

[0011] S4: First, merge the multiple subjective evidences to obtain a new source of subjective evidence, and then merge the new source of subjective evidence with the objective evidence source to obtain the identification result of the industrial furnace working condition.

[0012] Secondly, this application provides an industrial furnace condition identification system that integrates process variables and expert knowledge, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the method described in the first aspect above.

[0013] Beneficial effects:

[0014] The industrial furnace and kiln operating condition identification method provided in this application, which integrates process variables and expert knowledge, extracts a set of key process variables that contain important furnace and kiln operating condition information and are non-redundant. Addressing the problem of one-sided identification of furnace and kiln operating conditions solely based on data or knowledge, this invention proposes a data-knowledge jointly driven method. It establishes an objective evidence theoretical model based on process variables and a subjective evidence theoretical model based on expert knowledge, extracting corresponding evidence conclusions and evidence credibility. When fusing multiple conflicting evidence sources, it sequentially fuses multiple subjective evidence sources through conflict redistribution and secondary evidence fusion, obtaining new subjective evidence. This new subjective evidence is then fused with objective evidence to resolve local contradictions between pieces of evidence, resulting in an accurate furnace and kiln operating condition identification result that integrates process variables and expert knowledge.

[0015] In a further proposed solution, the distance from the sample to the corresponding hyperplane calculated by the NuSVC model is used as an influencing factor for furnace and kiln condition identification. This distance is input into the furnace and kiln condition probability mapping function, establishing an objective evidence theoretical model. Furthermore, by fully integrating on-site survey experience and knowledge from industrial furnace and kiln experts, logical rules and membership functions for furnace and kiln condition identification are constructed, establishing a subjective evidence theoretical model. This yields a rough identification result for judging furnace and kiln conditions, namely, objective evidence and subjective evidence sources containing multiple subjective evidence. This provides evidence sources for subsequent fusion of subjective and objective evidence.

[0016] In a further proposed solution, during the secondary fusion process, the Euclidean distance, Bhattacharyya distance, and cosine similarity among the various pieces of evidence were fully considered. Based on these three calculation indicators, three corrected pieces of evidence were obtained. Then, by calculating the degree of local conflict among the corrected pieces of evidence, a conflict redistribution factor was obtained. Subsequently, the corrected evidence was fused to obtain a new source of evidence, which solved the problem of the difficulty in effectively fusing multiple sets of conflicting evidence. This provides a theoretical basis for the fusion of subjective and objective evidence in the identification of furnace operating conditions. Attached Figure Description

[0017] Figure 1 This is one of the flowcharts of a preferred embodiment of the present invention for identifying the operating conditions of an industrial furnace by fusing process data and expert knowledge.

[0018] Figure 2 The second flowchart is a preferred embodiment of the present invention of a method for identifying the operating conditions of an industrial furnace by fusing process data and expert knowledge.

[0019] Figure 3 This is a flowchart illustrating the key process variable selection in a preferred embodiment of the present invention.

[0020] Figure 4 This is a flowchart of the conflict redistribution and evidence secondary fusion algorithm of a preferred embodiment of the present invention;

[0021] Figure 5 (a) is a test effect diagram of the objective evidence theory model of the preferred embodiment of the present invention; 5(b) is a test result diagram of the subjective evidence theory model; 5(c) is a test result diagram of the fusion model. Detailed Implementation

[0022] The technical solution of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "connected" or "linked" and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up," "down," "left," "right," etc., are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship also changes accordingly.

[0024] Please see Figures 1-2 This application provides a method for identifying the operating conditions of industrial furnaces and kilns by integrating process variables and expert knowledge, including:

[0025] S1: Screen the key process variables of the industrial furnace process variable set, and extract the dynamic characteristics of the key process variables within a preset time window;

[0026] S2: Combine the dynamic features with the NuSVC (Nu-Support Vector Classifier) ​​model and probability mapping function to construct an objective evidence theoretical model, and obtain the objective evidence source based on the objective evidence theoretical model;

[0027] S3: Construct logical rules and membership functions for key process variables based on expert knowledge of industrial furnaces and kilns. Based on the logical rules, membership functions, and dynamic characteristics of the key process variables, construct a subjective evidence theory model. Obtain a subjective evidence source containing multiple subjective evidences based on the subjective evidence theory model.

[0028] In other words, before fusion, there were multiple sources of subjective evidence leading to a conclusion, but after fusion, there is only one source of evidence leading to a conclusion.

[0029] S4: First, merge the multiple subjective evidences to obtain a new source of subjective evidence, and then merge the new source of subjective evidence with the objective evidence source to obtain the identification result of the industrial furnace working condition.

[0030] The aforementioned method for identifying industrial furnace operating conditions by integrating process variables and expert knowledge addresses the problem that existing methods do not fully integrate data and knowledge. This method extracts a set of key process variables that contain important furnace operating condition information and are non-redundant. To address the issue of identifying furnace operating conditions solely through data or knowledge, this invention proposes a data-knowledge joint-driven approach. It establishes an objective evidence theory model based on process variables and a subjective evidence theory model based on expert knowledge, extracting corresponding evidence conclusions and evidence credibility. When fusing multiple conflicting evidence sources, it sequentially integrates multiple subjective evidence sources through conflict redistribution and secondary evidence fusion, obtaining new subjective evidence. This new subjective evidence is then fused with objective evidence to resolve local contradictions between pieces of evidence, resulting in accurate furnace operating condition identification results that integrate process variables and expert knowledge.

[0031] The steps of the above method will now be described in detail with reference to a complete embodiment.

[0032] 1. Screening of key process variables and extraction of their dynamic features

[0033] The key process variable extraction method based on prior classification information first performs a coarse classification of process variables using prior classification knowledge, and then performs correlation analysis between and within classes to extract key process variables. The process is as follows: Figure 3 As shown in the diagram. The specific steps of this method are illustrated in Algorithm 1. Algorithm 1 implements the extraction of key process variables based on prior classification information. Algorithm 2, as a sub-step within it, implements the correlation analysis and variable deletion operations for a given set of process variables. A detailed analysis follows:

[0034] Algorithm 1:

[0035] Step 1: Based on prior classification knowledge of process variables, divide the set K of process variables for industrial furnaces and kilns into k classes, namely:

[0036] K = {C1,C2,...,C} k};

[0037]

[0038]

[0039] In the formula, C j Represents the set of variables of type j; m j x is the number of variables in the j-th type of variable set; j_u Let be the u-th variable in the j-th variable set; n is the number of samples contained in each variable;

[0040] Step 2: Linear correlation analysis involves traditional statistical methods, such as Pearson correlation coefficient and Kendall correlation coefficient. Nonlinear analysis is based on the mutual information coefficient (MIC). Representatives within each class are selected using various correlation analysis methods to form a representative set π, calculated using the following formula:

[0041]

[0042]

[0043]

[0044]

[0045] π = {rep1,rep2,...,rep} k};

[0046] In the formula, 1≤u≤m j ;rep j That is, select the variable corresponding to the column (the u-th column) with the largest sum of column elements in the correlation coefficient matrix COR; (m j ,m j ) represents the matrix dimension; sgn is the sign function; x p and x q All are C j A column in the array; p,q∈[j_1,j_2,...,j_m] j ];x p and x q x p and x q A certain element in; and x p and x q The i-th element in the middle; P(x p ,x q (x) represents the simultaneous values ​​of variables U and V. p and x q The probability of P(x) p ) and P(x q x represents the individual values ​​of variables U and V respectively. p and x q The probability of;

[0047] Step 3: Set the threshold σ1 using domain expertise and expert experience, and execute Algorithm 2;

[0048] Step 4: Determine whether any elements in the representative set have been deleted in this round. If yes, go back to Step 1 and select a new representative for the class of the deleted representative. If no, form a set π' with the remaining undeleted representatives and go to Step 5.

[0049] Step 5: Set a threshold σ2, and execute Algorithm 2 for each class sequentially. At this point, the set of variables remaining in each class that have not been deleted is V = {ν1,ν2,...ν}. k In summary, the key variable set selected by the algorithm proposed in this invention is KV={π',ν1,ν2,...ν}. k}

[0050] Algorithm 2:

[0051] This implementation demonstrates the operation of performing correlation analysis and variable deletion on a given set of process variables and a correlation threshold. The specific steps are as follows:

[0052] Step 11: Calculate the correlation coefficient matrix of the set;

[0053] Step 12: Select the pair of variables (M, N) with the largest correlation coefficient that is greater than the threshold σ;

[0054] Here, σ1 and σ2 are the thresholds σ, and they are different in size.

[0055] Step 13: Calculate the average correlation coefficients of M and N with other variables to obtain α1 and α2;

[0056] Step 14: If α1 > α2, delete M; otherwise, delete N.

[0057] Step 15: Repeat Steps 2-4 until all correlation coefficients are less than the threshold σ.

[0058] Compared to correlation analysis methods based on random forests and Lasso regression, this method first performs coarse classification of variables and coarse screening of variables between classes based on prior knowledge, which helps reduce the complexity of the key variable extraction process. Coarse classification and screening provide a preliminary framework for further fine screening within classes, which is also beneficial for in-depth analysis of the influencing factors of abnormal states before addressing abnormal operating conditions. The advantage of this method is that it utilizes expert knowledge of industrial furnaces and kilns to obtain pre-classification results of process variables, making the model more interpretable and feasible.

[0059] Considering the time-series characteristics of complex, multi-stage industrial process variables, we extract dynamic features from the dataset within a time window. These features reflect the dynamic characteristics of the operating conditions within the time window and mitigate the impact of uncontrollable transient disturbances. We extract the mean, standard deviation, trend index, volatility coefficient, autocorrelation coefficient, kurtosis, and skewness of the data within the time window. The mean reflects the average level of the data within the window; the standard deviation, trend index, and volatility coefficient characterize the data's changing trend and stability, enabling forward-looking judgments; the autocorrelation coefficient reflects the correlation of the data within the window; and kurtosis and skewness describe the data's kurtosis and distribution position, respectively. These dynamic features depict the changing characteristics of the process variables, providing a basis for subsequently establishing objective and subjective evidence-based theoretical models.

[0060] 2. Combining the NuSVC model and probability mapping function, construct an objective evidence theory model based on process data.

[0061] NuSVC, a widely used classification model with diverse kernel techniques, flexible parameter tuning, robustness to outliers, and suitability for small datasets, balances the number of support vectors and training error by adjusting the marginal error score nu∈(0,1]. Based on the classic principle of structural risk minimization, it explores the optimal match between the model's learning ability and complexity using limited training samples to enhance its adaptability. The optimization objective of the NuSVC model is to minimize the sum of squares of the weights while ensuring as many correct classifications as possible. The core formula and constraints are shown below:

[0062]

[0063] y i (w·x i +b)≥1-ξ i ,ξ i ≥0;

[0064] Where w is the weight; C is the regularization parameter, used to balance maximizing the margin and minimizing the error; N is the number of samples; ξ i y is a slack variable used to tolerate some training error during training, balancing model complexity and performance; i The classification result is processed using a one-to-many method, allowing it to take the value 1 or -1; x i is the input vector; b is the bias term.

[0065] When NuSVC is applied to sample classification problems, it only outputs a definite conclusion that the sample belongs to a certain class, i.e., 1 or 0, and does not support the probability analysis of the classification results. Therefore, this invention uses an improved Sigmoid probability model to calculate the probability of the occurrence of the working condition. The distance from each sample to the hyperplane in the trained NuSVC model is used as an influencing factor, and the parameters U and V in the following working condition probability mapping function P(dis,U,V) are determined by the cross-entropy loss function:

[0066]

[0067] Here, dis represents the distance from the sample to the corresponding hyperplane. By inputting dis into the corresponding operating condition probability mapping function, the probability of each operating condition occurring can be calculated, and this probability calculation result is the credibility of objective evidence, representing the degree of credibility of the identification result. This model obtains a rough identification result of the furnace operating condition judgment, i.e., objective evidence, providing an objective evidence source for subsequent evidence fusion.

[0068] 3. Construct a subjective evidence theory model based on expert knowledge

[0069] This invention systematizes and formalizes expert knowledge by constructing logical rules and membership functions, thereby establishing a set of scientific and reliable guiding principles. Specifically, the membership functions constructed using process variable eigenvalues ​​effectively characterize the fuzzy states of industrial furnace process variables and their changes in the industrial process, enabling uncertain reasoning about the operating state of industrial furnaces. Simultaneously, logical rules extracted from expert knowledge guide fuzzy expert reasoning, calculating the support degree of these rules for the operating state of industrial furnaces and extracting the credibility of subjective evidence.

[0070] To effectively articulate and store expert knowledge, this invention constructs logical rules as shown in the following formula based on expert knowledge and on-site survey experience. These rules utilize the membership function calculated above to perform reasoning operations and obtain subjective evidence and its credibility.

[0071] IF A And B And(C Or D)And E Then Category(CF∈[0,1]);

[0072] In the formula, (A, B, C, D, E) represent the premises of each rule (influence factors of the operating condition); And, Or, and Then are logical operators, including conjunction and disjunction relations; Category is the conclusion of the inference operation, including the identification type and its confidence level; CF∈[0,1] (Confidence Factor, CF, confidence coefficient) is the rule confidence level of the logical rule. This model obtains a rough identification result of the furnace operating condition, that is, a subjective evidence source containing multiple subjective evidences, providing a source of subjective evidence for subsequent evidence fusion.

[0073] 4. Integrate subjective and objective sources of evidence

[0074] When there is a high degree of conflict between pieces of evidence, traditional evidence fusion rules struggle to accurately describe the similarity measure between them, leading to counterintuitive results. Therefore, this invention proposes a conflict redistribution and secondary evidence fusion algorithm, the flowchart of which is shown below. Figure 4 As shown, the subjective evidence sources containing multiple subjective evidence sources are first fused together to obtain a new subjective evidence, and then this new subjective evidence is fused together with the aforementioned objective evidence a second time.

[0075] First, three types of corrected evidence sources are calculated based on Euclidean distance, Bhattacharyya distance, and cosine similarity, respectively. Euclidean distance intuitively depicts the distance between two points in high-dimensional space; Bhattacharyya distance measures the similarity between two discrete distributions; and cosine similarity analyzes the similarity between data from the perspective of the angle between two vectors, making it particularly effective for high-dimensional data similarity analysis, such as in complex industrial processes. Next, the conflict relationships among the three types of corrected evidence sources are analyzed, and the evidence conflict redistribution factor is calculated. Then, these three types of corrected evidence sources are fused to obtain new subjective evidence. Finally, the new subjective evidence and objective evidence are fused a second time to resolve local contradictions between the evidence, resulting in a comprehensive data and knowledge-based condition identification result.

[0076] The calculations for the three types of correction sources are as follows:

[0077] Corrected Evidence Source 1: Corrected Evidence Source Based on Euclidean Distance

[0078] Let the set of subjective evidence sources be...

[0079] M = {m i |i=1,2,...,n};

[0080] m i =(m i1 ,m i2 ,...,m iq );

[0081] And the mean of all elements in the j-th column (j=1,2,...,q) of M (possible results) is satisfy and For each source of evidence m i Calculate the m elements it contains. ij to the column mean of its column Euclidean distance:

[0082]

[0083] Therefore, the distance vector of the evidence source set M can be expressed as:

[0084] ED = [d1, d2, ..., d n ];

[0085] Considering that the smaller the distance between pieces of evidence, the higher the similarity, let... For the source of evidence m i credibility, d i express Refers to the i-th source of evidence m i In it, each element m is contained in ij Column mean of the column The Euclidean distance. Therefore, the corrected source of evidence 1 can be obtained as:

[0086]

[0087] Then, performing a Softmax operation on M1(j) will yield the corrected evidence source 1.

[0088] Corrected Evidence Source 2: Corrected Evidence Source Based on Bhattacharyya Distance

[0089] Let there be two sources of evidence m in the set of evidence sources M. i and m e The Bhattacharyya distance between (1≤i,e≤n) is:

[0090] BD(m i ,m e )=-ln(BC(m i ,m e ));

[0091]

[0092] In the formula, m ij Describes the i-th source of evidence m i The j-th element; m ej Indicates the e-th source of evidence m e The j-th element;

[0093] Therefore, we can obtain a symmetric distance matrix BDM with dimension (n,n) and all diagonal elements being 0, i.e., BD(m i ,m e ) = BD(m e ,m i ).

[0094]

[0095] make

[0096]

[0097]

[0098] Where, sum i Let Bhattacharyya distance represent the sum of the distances between the i-th evidence source and other evidence sources. The larger the distance, the smaller the similarity and the lower the support. Therefore, sup i With sum i They are negatively correlated and satisfy the following conditions: Therefore, the corrected source of evidence 2 can be obtained as follows:

[0099]

[0100] Modified Evidence Source 3: Modified Evidence Source Based on Cosine Similarity

[0101] Let there be two sources of evidence m in the set of evidence sources M. i and m e The cosine similarity between (1≤i, e≤n) is:

[0102]

[0103] make

[0104]

[0105] If a source of evidence is more similar to other sources of evidence, it will receive stronger support from those other sources, and vice versa. Therefore, let's assume...

[0106]

[0107] And satisfy Therefore, the corrected source of evidence 3 can be obtained as follows:

[0108]

[0109] M d =(M d (1),M d (2),...,M d (q)), d = 1, 2, 3;

[0110] In the formula, M d This represents a modified evidence source of length q (the number of possible categories). Based on the three types of modified results of the original evidence, a conflict redistribution and secondary evidence fusion algorithm is proposed, which cleverly solves the evidence conflict problem and realizes the fusion of conflicting evidence, namely:

[0111]

[0112] Where ε1(Q) is the conflict redistribution term, calculated as follows:

[0113]

[0114] In the formula, M1 represents the evidence allocation function for the first type of evidence source, which is the confidence level of the evidence source with respect to the occurrence of its possible events. For example, if an evidence source contains three possible events X, Y, and Z, the confidence levels of these three possible events are 0.5, 0.3, and 0.2, respectively. Here, M1 refers to modified evidence source 1; M2 is modified evidence source 2; M3 is modified evidence source 3; ε1 is the conflict redistribution factor at the subsequent evidence fusion; M u For u,v,w∈[1,2,3], this is also a modified source of evidence. Here, it means replacing 1, 2, and 3 with u,v, and w, which have different combinations; M v Refers to the revised source of evidence v; M w The formula refers to modifying the source of evidence w; Q, E, and F each represent a possible event; γ is the threshold for conflict redistribution, which can be set based on prior knowledge, such as γ = 0.5. That is, when the difference in credibility between the evidences is greater than the threshold γ, the conflict between the evidences is assigned to the proposition with higher credibility; when the difference in credibility between the evidences is less than the threshold γ, the conflict between the evidences is shared by all sources of evidence.

[0115] Finally, normalizing M(Q)1 yields this new subjective evidence.

[0116] The above steps yield a new piece of subjective evidence, which is then fused with objective evidence to obtain the final accurate identification result. The fusion rules are as follows:

[0117]

[0118] In the formula, M1 represents the evidence allocation function for the first type of evidence source, which is the credibility of the evidence source with respect to the occurrence of its possible events. For example, if a certain evidence source contains three possible events X, Y, and Z, the credibility of these three possible events is 0.5, 0.3, and 0.2, respectively. Here, M1 refers to the subjective evidence source. M2 represents the evidence allocation function for the second type of evidence source, where M2 refers to the objective evidence source. ε2 represents the conflict redistribution factor during the second evidence fusion. X, Y, and Q are all possible events. X∩Y=Q indicates that the intersection of events X and Y is event Q; M(Q)2 is the result after fusion.

[0119] Wherein, ε2(Q) is the conflict redistribution term, calculated as follows:

[0120]

[0121] Finally, normalizing M(Q)2 yields the fused evidence result. This method integrates corrected multi-source evidence information, resolves evidence conflict issues, redistributes conflicting evidence, and improves the usability and scientific rigor of the fusion model. By combining objective and subjective evidence sources obtained in the above steps through a data-knowledge joint-driven approach, more accurate industrial furnace operating condition identification results can be obtained.

[0122] Below, we take a 2650m steel plant in China as an example. 3 Taking a blast furnace as an example, the above steps are described in detail: Specifically, a 2650m³ blast furnace in a domestic steel plant... 3 The data samples used for the blast furnace included normal operating conditions (7116), suspended charge conditions (3534), and piping conditions (1767). First, a key process variable extraction method based on prior classification information was used to select 15 key process variable sets from over fifty industrial furnace process variables. Then, the dynamic features of these 15 key process variables were extracted, a NuSVC model was trained, the parameters of the probability function were calculated, an objective evidence theoretical model was established, and the test results are as follows: Figure 5 As shown in (a). Then, based on the dynamic characteristics of 15 key variables representing the working process of industrial furnaces and kilns obtained from the above experiments, and integrating expert knowledge, five logical rules were created to judge three types of working conditions: normal working conditions, suspended material (abnormal working conditions), and pipeline (abnormal working conditions). Under the guidance and suggestions of on-site experts and scholars, the defined logical rules are as follows:

[0123] Rule1:IF A And B And C Then Category(CF1=0.85,CF2=0.87,CF3=0.90);

[0124] Rule2:IF(D Or E)And F And(J Or K)Then Category(CF1=0.90, CF2=0.80, CF3=0.95);

[0125] Rule3:IF L And M And(N Or O)Then Category(CF1=0.90, CF2=0.93, CF3=0.88);

[0126] Rule4:IF(G Or H)And I And T Then Category(CF1=0.95, CF2=0.90, CF3=0.95);

[0127] Rule5:IF(P Or Q)And R And S Then Category(CF1=0.85, CF2=0.87, CF3=0.92);

[0128] Wherein, Category∈{1,2,3} represents normal operating conditions, suspended material, and pipeline, respectively; CF1, CF2, and CF3 are the credibility of the criterion when used to judge the three types of operating conditions; influencing factors A to T refer to the mean CO, mean H2, mean CO2, standard deviation of top temperature, top temperature fluctuation coefficient, mean top pressure, mean permeability index, standard deviation of permeability index, mean total pressure difference, autocorrelation coefficient of actual pulverized coal injection in this hour, kurtosis of actual pulverized coal injection in this hour, mean oxygen-enriched flow rate, mean oxygen-enriched pressure, mean theoretical combustion temperature, standard deviation of theoretical combustion temperature, cold air temperature trend index, cold air temperature fluctuation coefficient, mean actual wind speed, mean cold air flow rate, and cold air flow resistance coefficient, totaling 20 influencing factors. A subjective evidence theoretical model was then constructed, and the test results are as follows: Figure 5 As shown in (b). Finally, based on the proposed conflict redistribution and secondary evidence fusion algorithm, the subjective evidence sources containing five pieces of subjective evidence are first fused to obtain new subjective evidence sources. Then, these new subjective evidence sources are fused with the objective evidence sources to obtain the final fusion result. Figure 5 (c) shows the test results of the fusion model.

[0129] In summary, the test results of the fusion model are significantly better than those of the objective evidence theory model and the subjective evidence theory model. The objective evidence theory model, based on extensive data and statistical patterns, provides objective and accurate predictions, while the subjective evidence theory model fully considers expert knowledge and is better suited to complex real-world scenarios. The fusion model effectively combines these two models. Experimental results show that the proposed method achieves high accuracy in the operating condition identification problem, meeting the needs of on-site work and providing reliable decision support for on-site personnel to monitor furnace and kiln operating status and adjust equipment parameters in real time.

[0130] This application also provides an industrial furnace condition identification system that integrates process variables and expert knowledge, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described method. This industrial furnace condition identification system that integrates process variables and expert knowledge can implement various embodiments of the above-described method and achieve the same beneficial effects; therefore, further details are omitted here.

[0131] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for identifying the operating conditions of industrial furnaces and kilns by integrating process variables and expert knowledge, characterized in that, include: S1: Screen the key process variables of the industrial furnace process variable set, and extract the dynamic characteristics of the key process variables within a preset time window; S2: Combine the dynamic features with the NuSVC model and probability mapping function to construct an objective evidence theoretical model, and obtain the objective evidence source based on the objective evidence theoretical model; S3: Construct logical rules and membership functions for key process variables based on expert knowledge of industrial furnaces and kilns. Based on the logical rules, membership functions, and dynamic characteristics of the key process variables, construct a subjective evidence theory model. Obtain a subjective evidence source containing multiple subjective evidences based on the subjective evidence theory model. S4: First, integrate the multiple subjective evidences to obtain a new source of subjective evidence, and then integrate the new source of subjective evidence with the objective evidence source to obtain the identification result of the industrial furnace working condition. S4 includes: S41: Three types of corrected evidence sources are calculated based on Euclidean distance, Bhattacharyya distance and cosine similarity, respectively; S42: Analyze the conflict relationship among the three types of modified evidence sources, and calculate the evidence conflict redistribution factor based on the conflict relationship; In the formula, The evidence allocation function represents the first type of evidence source. Refer to the revised source of evidence 1; To correct source of evidence 2; To correct source of evidence 3; The conflict redistribution factor during the first evidence fusion; for This also means correcting the source of evidence. Here, it means replacing 1, 2, and 3 with u, v, and w, which can be combined in different ways. Refers to the revised source of evidence v; The source of evidence w is modified; X, Y, Z, Q, E, and F all represent events. S43: Based on the fusion rules and conflict redistribution factors, the three types of modified evidence sources are fused to obtain new subjective evidence sources; S44: The new subjective evidence sources and objective evidence sources are fused a second time to obtain the identification result of industrial furnace and kiln operating conditions based on comprehensive data and knowledge.

2. The industrial furnace and kiln condition identification method based on the fusion of process variables and expert knowledge according to claim 1, characterized in that, S1 includes: S11: The method for extracting key process variables based on prior classification information firstly classifies process variables by prior classification knowledge, and then performs correlation analysis between and within classes to extract key process variables. S12: Extract dynamic features of the process dataset within the time window to describe the changing characteristics of process variables.

3. The industrial furnace and kiln condition identification method based on the fusion of process variables and expert knowledge according to claim 2, characterized in that, S11 includes: Step 1: Based on prior classification knowledge of process variables, classify the industrial furnace process variable set. Divided into Classes, with the following relation: (1)； (2)； (3)； In the formula, The first variable in the set of process variables for industrial furnaces and kilns Class variable collection, Representing the Class variable collection, For the first The number of variables in a class variable set. For the first The first in the class variable set One variable, This represents the number of samples included in each variable. The first variable in the set of process variables for industrial furnaces and kilns Class variable collection The first in One variable, The first variable in the set of process variables for industrial furnaces and kilns A collection of class variables The first in Variables The One value; Step 2: Use various correlation analysis methods to select representatives within each category and form a representative set. The calculation formula is as follows: (4)； (5)； (6)； (7)； (8)； In the formula, , That is, selecting the correlation coefficient matrix The variables corresponding to the middle column elements and the largest column. Indicates the size of the matrix dimension. For symbolic functions, and All One column in , and They are respectively and The Middle One element, For variables and variables Simultaneously take values and The probability, and Variables and variables Individual value and The probability, This represents the Pearson correlation coefficient. This represents the Kendall correlation coefficient. Indicates the maximum information coefficient. express The average of all elements in the set. Indicates from the first Class variable collection The variables selected from the data represent; Step 3: Set thresholds based on domain expertise and expert experience. As parameters in subsequent algorithm steps, and to perform correlation analysis and variable deletion operations on the target process variable set; Step 4: Determine whether any elements in the representative set have been deleted in this round. If yes, go back to Step 1 and select a new representative for the class containing the deleted representative. If no, form a set from the remaining representatives that have not been deleted in this round. Proceed to Step 5; Step 5: Set the threshold For each class, the steps of performing correlation analysis and variable deletion operations on the target process variable set are performed sequentially. At this point, the remaining set of variables that have not been deleted in each class is... Thus combining and The key process variable set is as follows: 。 4. The industrial furnace operating condition identification method integrating process variables and expert knowledge according to claim 3, characterized in that, The steps of performing correlation analysis and variable deletion on the target process variable set include: Step 11: Calculate the correlation coefficient matrix of the set; Step 12: Select the items with the highest correlation coefficient that is greater than the threshold. A pair of variables ; Step 13: Calculate separately , The average correlation coefficient with other variables is obtained. , ; Step 14: If , then delete Otherwise, delete. ; Step 15: Repeat Steps 12-14 until all correlation coefficients are below the threshold. .

5. The industrial furnace and kiln condition identification method integrating process variables and expert knowledge according to claim 1, characterized in that, The dynamic characteristics include: mean, standard deviation, trend index, volatility coefficient, autocorrelation coefficient, kurtosis, and skewness.

6. The industrial furnace operating condition identification method based on the fusion of process variables and expert knowledge according to claim 1, characterized in that, The NuSVC model in S2 satisfies the following relationship: (9)； (10)； in, It's weight. It is a regularization parameter. For the sample size, As slack variables, It is the classification result. It is the input vector. It is a bias term; The probability mapping function is determined in the following way: The probability of a given work condition occurring is calculated using an improved Sigmoid probability model. The distance from each sample in the trained NuSVC model to the hyperplane is used as an influencing factor. The following work condition probability mapping function is determined using the cross-entropy loss function. Parameters in and : (11)； in, The distance from the sample to the corresponding hyperplane is calculated by... The input is fed into the corresponding working condition probability mapping function to calculate the probability of each working condition occurring. The probability calculation result is used as the credibility of objective evidence, representing the credibility of the identification result.

7. The industrial furnace and kiln condition identification method integrating process variables and expert knowledge according to claim 1, characterized in that, S3 includes: S31: Constructing membership functions through the dynamic characteristics of process variables; S32: Logical rules are constructed based on expert knowledge of industrial furnaces and kilns, as shown in the following relation: (12)； In the formula, As a premise for each rule, , and For logical operators, The conclusion of the reasoning operation. The credibility of this logical rule; S33: Obtain subjective evidence sources based on the logical rules and the membership function.

8. The industrial furnace and kiln condition identification method based on the fusion of process variables and expert knowledge according to claim 1, characterized in that, The fusion rule for the secondary fusion of the new subjective evidence source and the objective evidence source satisfies the following relationship: ； In the formula, Indicates the subjective source of evidence. Indicates the objective source of evidence. This represents the conflict redistribution factor during the second evidence fusion. , , Both refer to events.

9. An industrial furnace condition identification system integrating process variables and expert knowledge, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 8.

Images