Temperature monitoring system and temperature monitoring method for water jacket heating furnace

Abstract

The present invention relates to the field of data processing technology, and specifically to a temperature monitoring system and a temperature monitoring method for a water jacket heating furnace. The method comprises: obtaining historical temperature data and real-time temperature data of the water jacket heating furnace, randomly generating multiple binary trees using the historical temperature data, and evaluating the structural stability of each binary tree and the reliability of the root node selection. The Gini index is calculated by the temperature data of the binary tree and the inverse is taken to obtain the node purity of the binary tree. The product of the node purity and the reliability of the root node selection is used as the segmentation ability value, and then the weight of each binary tree is determined. The real-time temperature data is inserted into the binary tree, the change in the complexity of the binary tree is obtained, and the weighted sum is used to obtain the degree of abnormality, so as to judge whether the real-time temperature data is abnormal, so as to achieve the purpose of monitoring the temperature. This method comprehensively evaluates the performance of the binary tree based on multiple factors, improves the accuracy of temperature anomaly detection, and improves the accuracy and reliability of temperature monitoring of the water jacket heating furnace.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a temperature monitoring system and a temperature monitoring method for a water jacket heating furnace. Background Art

[0002] As an indirect fire barrel heater, the water jacket heating furnace has important applications in industrial production. Its working principle is that the fuel burns in the fire barrel at the bottom of the furnace body, and the flame and hot smoke transfer heat to the bath liquid through the fire barrel wall, and then the bath liquid heats the medium in the upper heating coil. Temperature monitoring is of great significance for water jacket heating furnaces. Accurate temperature monitoring can grasp the temperature changes in the furnace in real time, ensuring a stable and efficient heating process.

[0003] Traditional water jacket heating furnaces rely on manual control or simple automation systems for temperature monitoring. They are unable to cope with uneven temperature distribution and difficulty in adjustment in the furnace, which can easily lead to problems such as uneven heating and low thermal efficiency. In severe cases, temperature fluctuations can cause equipment overheating and damage.

[0004] There is a temperature monitoring method based on data analysis and processing technology. This method is more efficient than the traditional temperature monitoring method. For example, the existing Chinese patent document with announcement number CN118779814B provides an insulator temperature monitoring method and system. The temperature monitoring method uses RRCF (Robust Random Partition Forest) algorithm to generate multiple binary trees, screens out a part of the binary trees, and then inserts the temperature data into this part of the binary trees. The abnormality of the temperature data is obtained according to the change in the complexity of the binary trees, thereby realizing temperature monitoring.

[0005] However, when the temperature of the water jacket heating furnace is monitored using the above technical solution, the temperature change in the water jacket heating furnace is affected by many factors, such as heating speed, environmental changes, stability of the heating system, etc. These factors may cause the temperature of the water jacket heating furnace to fluctuate. The above technical solution selects part of the binary tree for insertion to determine abnormal temperature data, ignoring the characteristics of part of the binary tree, which is likely to cause the loss of some information. The technical solution focuses more on the classification of local temperature data, ignoring the overall temperature trend and fluctuation characteristics of the water jacket heating furnace, resulting in inaccurate abnormal identification results of temperature data. Summary of the invention

[0006] In order to solve the problem that the abnormal identification result of temperature data is not accurate enough when the temperature of a water jacket heating furnace is monitored, the present invention provides a temperature monitoring method of a water jacket heating furnace, comprising:

[0007] Obtain the historical temperature data and real-time temperature data of the water jacket heating furnace, randomly generate multiple binary trees based on the historical temperature data, and evaluate the structural stability of each binary tree and the reliability of the root node selection;

[0008] The Gini index is calculated based on all the temperature data corresponding to each binary tree, and the inverse of the Gini index is used as the node purity of each binary tree. The product of the node purity of each binary tree and the reliability of the root node selection of each binary tree is used as the segmentation ability value of each binary tree.

[0009] Calculate the weight of each binary tree: , For the The weight of a binary tree, and Respectively 、 The splitting ability value of a binary tree, and Respectively 、 The structural stability of a binary tree, is the total number of binary trees;

[0010] Insert each real-time temperature data into all binary trees, obtain the change in complexity of each binary tree, use the weight of each binary tree to perform weighted summation on the change in complexity of each binary tree, calculate the degree of abnormality of each real-time temperature data, and determine whether each real-time temperature data is abnormal based on the degree of abnormality of each real-time temperature data, so as to realize temperature monitoring of the water jacket heating furnace.

[0011] The above technical solution can mine and learn the characteristics and patterns of historical temperature data from different angles by randomly generating multiple binary trees. The evaluation of the structural stability of each binary tree and the reliability of the root node selection can preliminarily screen and understand the quality and effectiveness of each binary tree, laying the foundation for the subsequent more accurate use of binary trees for temperature anomaly judgment. Furthermore, the node purity of the binary tree is determined by the Gini index, which can better reflect the degree of unity of the temperature data of each node of the binary tree. The node purity is multiplied by the reliability of the root node selection to obtain the segmentation ability value, which comprehensively considers the classification quality of the node and the importance and credibility of the root node, so that the segmentation ability value can fully reflect the potential ability of the binary tree to classify temperature data and judge anomalies. Furthermore, the weight of the binary tree is calculated by a specific formula, which realizes the quantification of the relative importance of each binary tree in the entire model system. The segmentation ability value and structural stability of each binary tree are comprehensively considered, which can highlight the influence of those binary trees with good segmentation effect and stable structure in the final result judgment. Furthermore, by inserting the real-time temperature data into the binary tree, its influence on the complexity of the binary tree can be observed. Temperature data with different abnormalities have different changes in the complexity of the binary tree. The degree of abnormality is obtained by weighted summing up the changes in the complexity of multiple binary trees. The previously constructed binary tree model, weight system and complexity change evaluation mechanism are fully utilized to realize the comprehensive judgment of whether the real-time temperature data is abnormal from multiple angles and multiple binary trees, thereby improving the accuracy and reliability of temperature monitoring. In short, the above technical scheme constructs a more comprehensive and intelligent temperature monitoring system, comprehensively considering the response of multiple binary trees to real-time temperature data, giving full play to the role of each binary tree and balancing their influence. Compared with the traditional scheme that only relies on some binary trees, it can more comprehensively and accurately capture the abnormal characteristics in the temperature data, and improve the accuracy of the temperature monitoring of the water jacket heating furnace.

[0012] Preferably, the structural stability of each binary tree is evaluated as follows:

[0013] The inverse of the variance value of the temperature data corresponding to the left subtree and the right subtree of each binary tree is taken as the stationarity of the left subtree and the right subtree;

[0014] Determine the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree according to the distribution characteristics of the temperature data corresponding to the left subtree and the right subtree;

[0015] Calculate the structural stability of each binary tree:

[0016]

[0017] In the formula, For the The structural stability of a binary tree, is the left subtree, is the right subtree, , Respectively The quantitative value of the data distribution characteristics of the left and right subtrees of a binary tree. , Respectively The stability of the left and right subtrees of a binary tree.

[0018] The above technical solution organically combines the quantitative values ​​of the data distribution characteristics of the left and right subtrees and the stability index. This quantitative calculation method that comprehensively considers multiple factors avoids the one-sidedness of single factor evaluation and can more objectively and accurately measure the stability of the overall structure of the binary tree, thereby providing a reliable basis for subsequent operations such as distinguishing the advantages and disadvantages of different binary trees in temperature monitoring and reasonably allocating weights.

[0019] Preferably, the reliability evaluation method of root node selection is:

[0020]

[0021] In the formula, For the The reliability of root node selection of a binary tree, is the left subtree, is the right subtree, For the The left subtree of a binary tree The value of the temperature data corresponding to each node, For the The value of the temperature data corresponding to the root node of the binary tree, For the The number of nodes in the left subtree of a binary tree, For the The right subtree of a binary tree The value of the temperature data corresponding to each node, For the The number of nodes in the right subtree of a binary tree, is a natural exponential function.

[0022] The above technical solution can obtain a specific reliability value for the root node of each binary tree, and realize the quantitative representation of the reliability of root node selection. This makes it possible to intuitively compare the reliability of root node selection of many binary trees, and clearly grasp the situation of each binary tree on this key indicator. It is convenient to fully consider the differences in the reliability of root node selection in subsequent segmentation capability value calculations, weight allocation and other links, so that the calculation and application of binary tree related indicators are more accurate.

[0023] Preferably, a method for determining the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree is:

[0024] The ratio of the number of nodes in the left subtree and the right subtree to their respective depths is used as the quantitative value of the data distribution characteristics of the left subtree and the right subtree, respectively.

[0025] The above technical solution takes into account that the number of nodes and depth are two important structural characteristics of binary trees. Combining the two can comprehensively reflect the shape and scale of the left subtree and the right subtree, thereby more accurately describing their data distribution characteristics. A subtree with a large number of nodes but a shallow depth and a subtree with a small number of nodes but a deep depth have different data distributions, and the quantitative value can effectively distinguish such differences.

[0026] Preferably, another method for determining the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree is:

[0027] The arithmetic square root of the sum of the squares of the number of nodes and the depth of the left subtree and the right subtree is used as the quantitative value of the data distribution characteristics of the left subtree and the right subtree, respectively.

[0028] The above technical solution actually amplifies the influence of the depth of the subtree in the calculation of the quantitative value by calculating the square sum of the number of nodes and the depth, and then taking the arithmetic square root. Compared with linear combinations, this method emphasizes the role of the overall structural form of the sub-binary tree in shaping the data distribution characteristics. For example, for a subtree with a deeper depth, the square of its depth will account for a larger proportion in the calculation, which means that when measuring the data distribution characteristics, more emphasis will be placed on the impact of depth, such as data stratification and progressive relationships, which helps to grasp the distribution of data in the subtree in a more comprehensive and three-dimensional way.

[0029] Preferably, the change in complexity of each binary tree is determined by an RRCF algorithm.

[0030] Preferably, the method for determining whether each real-time temperature data is abnormal according to the degree of abnormality of each real-time temperature data is: setting a threshold value of the degree of abnormality, if the degree of abnormality of a certain real-time temperature data is greater than the threshold value of the degree of abnormality, the real-time temperature data is abnormal; if the degree of abnormality of a certain real-time temperature data is not greater than the threshold value of the degree of abnormality, the real-time temperature data is not abnormal.

[0031] Preferably, the calculation formula for the abnormality degree of each real-time temperature data is:

[0032]

[0033] In the formula, is the abnormality of the real-time temperature data, , They are all binary tree numbers. It is The weight of a binary tree, Insert the real-time temperature data into When the binary tree The change in complexity of a binary tree, is the total number of binary trees, Insert the real-time temperature data into When the binary tree The change in complexity of a binary tree.

[0034] The above technical solution measures the abnormality of real-time temperature data based on multiple binary trees, which reflects a comprehensive evaluation approach. Considering that a single binary tree reflects limited information, by summarizing the relevant changes of multiple binary trees, it can more comprehensively capture the degree of deviation from normal conditions presented by temperature data in different structures and different feature dimensions, avoiding one-sided conclusions caused by relying on single-aspect judgment.

[0035] The present invention also provides a temperature monitoring system for a water jacket heating furnace, the temperature monitoring system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any step of the temperature monitoring method.

[0036] The present invention has the following effects:

[0037] Through the synergistic effect of multiple links, the present invention comprehensively optimizes the temperature monitoring process and judgment mechanism of the water jacket heating furnace, and constructs a more complete and intelligent temperature monitoring system. When monitoring temperature data, the system comprehensively considers the response of multiple binary trees to real-time temperature data, gives full play to the role of each binary tree and balances its influence. Compared with the traditional solution that only relies on part of the binary tree, it shows a stronger ability to capture abnormal features, improves the accuracy of the temperature anomaly recognition results of the water jacket heating furnace, and the accuracy and reliability of temperature monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0039] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0042] Reference Figure 1 The present invention provides a method for monitoring the temperature of a water jacket heating furnace, comprising steps S1 to S6:

[0043] S1: Obtain historical temperature data and real-time temperature data of the water jacket heating furnace.

[0044] For the historical temperature data of the water jacket heating furnace, choose to obtain it from the database, because the historical temperature data of the water jacket heating furnace is stored in the database management system installed on the local server within the enterprise, such as MySQL, Oracle, SQLServer, PostgreSQL and other relational databases. This storage method is convenient for data management, maintenance and access control within the enterprise, and the data security is relatively high. It is suitable for use in the internal network environment of the enterprise. The corresponding database service can be accessed through the internal LAN to obtain data.

[0045] In relational databases, historical temperature data is often stored in the form of structured data tables. Generally speaking, each row in a data table represents a temperature data record collected at a specific time point, and each column corresponds to a different attribute field. Common fields include timestamp (accurately record the time when the temperature data was collected, usually stored in a specific time format such as year, month, day, hour, minute, and second), temperature value (record the actual temperature value of the water jacket heating furnace at that time, the data type may be floating point, etc.), equipment number (used to distinguish different water jacket heating furnace equipment, facilitating data management and analysis of multiple devices), collection location (for example, the data collected by temperature sensors in different parts of the furnace body will be marked with the corresponding collection location information), etc.

[0046] By accessing the database, the historical temperature data of the water jacket heating furnace is obtained, and then these historical temperature data are sorted in chronological order, so that the changing trend of the temperature data over time can be observed more clearly.

[0047] The real-time temperature data of the water jacket heating furnace is collected in real time through temperature sensors installed in key positions of the water jacket heating furnace, such as inside the furnace body close to the heating source, around the heating coil, and in the bath liquid, so as to accurately sense the temperature changes in different areas. The collection frequency of the temperature sensor is 1 Hz.

[0048] Through this step, the historical temperature data and real-time temperature data of the water jacket heating furnace can be obtained, providing a reliable data basis for subsequent further data analysis, modeling, temperature monitoring and other work.

[0049] S2: Randomly generate multiple binary trees based on historical temperature data.

[0050] Specifically, randomly generating multiple binary trees based on historical temperature data is a way to construct a binary tree by randomly selecting specific subsets from historical temperature data. This step randomly extracts 50 subsets of the same length from the historical temperature data, and each subset contains 100 data. Then, for each subset, the median of the subset is used as the split point, and the RRCF (Robust Random Partition Forest) algorithm is used to construct a binary tree corresponding to the subset, thereby obtaining multiple binary trees of historical temperature data. In each binary tree, each node stores a temperature data.

[0051] Since traditional temperature anomaly detection has obvious limitations when applied to water jacket heating furnaces, it uses the method of screening part of the binary tree and inserting data to determine abnormal data points. This method focuses on the classification and processing of local data points. However, the temperature change of the water jacket heating furnace is affected by many complex factors, showing an overall temperature trend and fluctuation characteristics, and these changes are often nonlinear. Traditional detection methods that only focus on local data classification are difficult to effectively handle this complex overall temperature condition, which greatly reduces the accuracy of temperature anomaly detection results.

[0052] In view of this, the present invention proposes a new solution:

[0053] By comprehensively considering all binary trees, the change in the complexity of the binary trees caused by inserting real-time temperature data into all binary trees is weighted. Such a weighted operation can fully take into account the information contained in each binary tree and avoid the loss of some information caused by binary tree screening. Since the influence of all binary trees is fully considered and the temperature fluctuation characteristics of the water jacket heating furnace are taken into consideration, the detection dilemma caused by temperature fluctuations or nonlinear changes faced by the RRCF algorithm in the temperature monitoring of the water jacket heating furnace can be effectively improved, thereby improving the accuracy and reliability of temperature anomaly detection and providing a more powerful guarantee for the safe and stable operation of the water jacket heating furnace.

[0054] Therefore, determining the weight of each binary tree becomes a key link to achieve this optimization goal. The present invention determines the weight of each binary tree through steps S3 to S5. Specifically, in the process of obtaining the weight of each binary tree, the present invention proposes to comprehensively consider the structural stability of the binary tree and the segmentation ability of the binary tree to determine the weight of the binary tree. The specific reasons are as follows:

[0055] The partitioning ability of a binary tree refers to the ability of a binary tree to effectively partition the data space into more distinguishable subspaces when facing new nodes (new temperature data). Specifically, in the water jacket heating furnace temperature monitoring scenario, if the binary tree has a strong partitioning ability, it means that it can quickly and accurately partition nodes based on different characteristics of the temperature data, such as the temperature range in different time periods, the temperature change rate, etc., and construct a subspace with obvious differentiation. For example, in the heating furnace heating stage, the rapidly rising temperature data can be effectively distinguished from the relatively stable preheating temperature data. In the insulation stage, the normal small temperature data fluctuations can be clearly distinguished from the possible abnormal cooling data. When a binary tree can partition temperature data so quickly and effectively, it means that it can sensitively capture various changes in temperature data, whether it is a subtle fluctuation adjustment or a phased large change, it can accurately perceive it. At this time, the binary tree is particularly important in the entire temperature monitoring system, so the corresponding weight is higher, and it can play a greater role in comprehensively judging whether the temperature is abnormal or not.

[0056] The structural stability of the binary tree reflects the uniformity of data distribution and the balance between nodes. If the structural stability of the binary tree is higher, it means that the binary tree is more difficult to flexibly adapt to the addition of new nodes. The addition of new nodes may cause the complexity of the binary tree to fluctuate greatly, which in turn leads to a greater possibility that the new node will be judged as an abnormal node, which is easy to cause misjudgment. Therefore, it is chosen to reduce the weight of the binary tree. Specifically, because the temperature data of the water jacket heating furnace itself is volatile, if a certain temperature data is normal temperature data, but it is inserted into a binary tree with a very stable structure, then the insertion of the temperature data will cause a large change in the complexity of the binary tree, which will cause the temperature data to be misjudged as abnormal. Therefore, in order to reduce the misjudgment caused by the structural characteristics of the binary tree, when determining the weight, it is chosen to reduce the weight of the binary tree, thereby reducing its influence on the final judgment result in the overall temperature monitoring system and ensuring the accuracy of the temperature abnormality judgment.

[0057] S3: Evaluate the structural stability of each binary tree.

[0058] The present invention determines the structural stability by analyzing the stability of the left subtree and the right subtree of the binary tree and the distribution characteristics of the temperature data of the left subtree and the right subtree.

[0059] S31: Analyze the stability of the left subtree and the right subtree of the binary tree. Specifically, obtain all temperature data corresponding to the left subtree, and use the inverse of the variance of these temperature data as the stability of the left subtree. Similarly, obtain all temperature data corresponding to the right subtree, and use the inverse of the variance of these temperature data as the stability of the right subtree.

[0060] Variance is a classic statistic that measures the degree of data dispersion. By calculating its reciprocal to evaluate stationarity, the degree of data fluctuation can be quantified. In the case of water jacket heating furnace temperature monitoring, the fluctuation of temperature data directly reflects the stability of the heating furnace working state. A smaller variance means that the temperature data is relatively concentrated and has small fluctuations; after taking the reciprocal, this relationship is reversed, so that the larger the stationarity value, the more stable the temperature data. This intuitively quantifies the stability of the temperature data represented by the subtree, which meets the intuitive understanding of stationarity and subsequent calculation requirements.

[0061] The stability of the left and right subtrees is an important component of measuring the overall structural stability of the binary tree. In an ideal stable binary tree structure, the stability of the left and right subtrees should be close and both at a high level. By quantifying the stability of the left and right subtrees separately, it is possible to clearly judge the contribution and impact of each subtree on the overall stability in the subsequent comprehensive evaluation of the binary tree structure stability, thereby providing basic data support for building a more reliable and accurate temperature monitoring system, and helping to select binary trees with high structural stability for temperature anomaly judgment.

[0062] S32: Determine the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree according to the distribution characteristics of the temperature data corresponding to the left subtree and the right subtree, and reflect the distribution characteristics of the temperature data on the corresponding subtree through the quantized values.

[0063] There are two methods here:

[0064] The first method is to use the ratio of the number of nodes in the left subtree and the right subtree to their respective depths as the quantitative values ​​of the data distribution characteristics of the left subtree and the right subtree, respectively.

[0065] In one example, Taking the left subtree of a binary tree as an example, the quantitative value of its data distribution characteristics is:

[0066]

[0067] In the formula, is the left subtree, For the The quantitative value of the data distribution characteristics of the left subtree of a binary tree, For the The number of nodes in the left subtree of a binary tree, For the The depth of the left subtree of a binary tree.

[0068] In addition, The quantized value of the data distribution feature of the right subtree of a binary tree can also be obtained according to this method.

[0069] The second method is to use the arithmetic square root of the sum of the squares of the number of nodes and the depth of the left subtree and the right subtree as the quantitative values ​​of the data distribution characteristics of the left subtree and the right subtree, respectively.

[0070] In one example, Taking the left subtree of a binary tree as an example, the quantitative value of its data distribution characteristics can also be expressed as:

[0071]

[0072] In the formula, For the The quantitative value of the data distribution characteristics of the left subtree of a binary tree, For the The number of nodes in the left subtree of a binary tree, For the The depth of the left subtree of a binary tree.

[0073] In addition, The quantized value of the data distribution feature of the right subtree of a binary tree can also be obtained according to this method.

[0074] S33: Calculate the structural stability of each binary tree based on the stationarity of the left subtree and the right subtree of each binary tree and the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree.

[0075] The present invention proposes that: for a binary tree, the smaller the difference between the stability of the left subtree and the stability of the right subtree, the higher the structural stability of the binary tree, and the smaller the difference between the quantized value of the data distribution feature of the left subtree and the quantized value of the data distribution feature of the right subtree, the higher the structural stability of the binary tree. Because, ideally, the node distribution in the left and right subtrees of the binary tree and the stability of both sides are consistent.

[0076] Therefore, based on the above logic, the calculation formula for the structural stability of each binary tree is:

[0077]

[0078] In the formula, For the The structural stability of a binary tree, is the left subtree, is the right subtree, , Respectively The quantitative value of the data distribution characteristics of the left and right subtrees of a binary tree. , Respectively The stability of the left and right subtrees of a binary tree.

[0079] In this formula, It is The difference in the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree of a binary tree. The larger the difference, the greater the difference, indicating that there is a greater difference in the distribution characteristics of the temperature data on the left subtree and the right subtree, indicating that the structural stability of the binary tree is worse, and vice versa. Indicates The difference in stability between the left and right subtrees of a binary tree. The greater the difference in stability between the left and right subtrees, the worse the structural stability of the binary tree, and vice versa. Therefore, this formula combines the differences in these two aspects and can accurately evaluate the structural stability of each binary tree.

[0080] S4: Evaluate the partitioning capability of each binary tree.

[0081] When evaluating the partitioning capability of each binary tree, the present invention mainly analyzes from two aspects: the reliability of the root node selection of the binary tree and the node purity of the binary tree.

[0082] Among them, the reliability of the root node selection of the binary tree is of great significance to the binary tree. In the process of binary tree construction and data segmentation, the selection of the root node plays an extremely critical leading role. The reliability of the root node selection directly determines the direction and effect of the first data division. If the root node is selected appropriately, the temperature data can be reasonably divided into two parts, the left and right subtrees, based on the key features of the temperature data (such as the average temperature level, the turning point of the temperature change trend, etc. in the water jacket heating furnace temperature data), making the subsequent branch construction and data segmentation more logical and effective, thus laying the foundation for the good structure of the entire binary tree and greatly improving the segmentation effect. For example, in the start-up and heating stage of the water jacket heating furnace, if the root node can accurately divide the data according to a certain key temperature threshold, and divide the preheating data below the threshold and the rapid heating data above the threshold into the left and right subtrees respectively, then the subsequent analysis and anomaly detection of temperature changes at different stages can be carried out more accurately.

[0083] Among them, the node purity of the binary tree reflects the consistency of the data contained in the node, and plays a supporting role in the segmentation effect of the binary tree. A higher node purity means that the data in the node has a strong similarity and occupies a relatively independent and concentrated area in the data space. In the current scenario, if the purity of a node in the binary tree is high, it means that the temperature data in the node are all in a certain temperature range or have similar temperature change patterns, and the more likely it is that the temperature data is in the stable insulation stage. This indirectly proves that the binary tree can effectively distinguish temperature data with different characteristics during the data segmentation process, avoid data confusion and intersection, and thus improve the overall segmentation effect.

[0084] This step combines the reliability of root node selection with the node purity of the binary tree to evaluate the segmentation ability of the binary tree. This is a reasonable and effective evaluation method based on the binary tree structure and data processing logic. The reliability of root node selection focuses on controlling the rationality of the binary tree's segmentation of temperature data from the perspective of the starting point, while the node purity reflects the fineness and accuracy of the segmentation from the composition of the temperature data inside the binary tree. The two complement each other and jointly determine whether the binary tree can effectively divide the data space into subspaces with clear distinction when facing the complex temperature data of the water jacket heating furnace, so as to achieve accurate identification and classification of different temperature states and change trends. For example, a binary tree with reasonable root node selection and high node purity can clearly divide the temperature data of different periods such as the heating stage, the insulation stage, and the cooling stage when processing the temperature data of the water jacket heating furnace, and separate the abnormal temperature data from the normal temperature data, thereby providing strong support for temperature monitoring, anomaly detection, and equipment operation status evaluation.

[0085] Therefore, this step calculates the reliability of the root node selection of each binary tree according to the following formula:

[0086]

[0087] In this formula, For the The reliability of root node selection of a binary tree, is the left subtree, is the right subtree, For the The left subtree of a binary tree The value of the temperature data corresponding to each node, For the The value of the temperature data corresponding to the root node of the binary tree, For the The number of nodes in the left subtree of a binary tree, For the The right subtree of a binary tree The value of the temperature data corresponding to each node, For the The number of nodes in the right subtree of a binary tree, is a natural exponential function.

[0088] In this formula, For the All nodes of the left subtree of a binary tree are The degree of difference of the root nodes of a binary tree, For the All nodes of the right subtree of a binary tree are The difference between the root nodes of a binary tree is the closer the two differences are, that is, The smaller the value, the The more balanced the left and right subtrees on both sides of the root node of a binary tree are, the The higher the reliability of root node selection of a binary tree, the more negative correlation is constructed using negative exponential function.

[0089] Therefore, the method for obtaining the node purity of each binary tree in this step is: first obtain all temperature data corresponding to each binary tree, calculate the Gini index of each node based on these temperature data, perform weighted summation of the Gini indexes of all nodes to obtain the Gini index of the binary tree, and use the inverse of the Gini index of the binary tree as the node purity of the binary tree.

[0090] The method for obtaining the Gini index of each binary tree is as follows:

[0091] The Gini index is mainly used to measure the purity or uncertainty of the data related to a node. In a binary tree, the Gini index of a leaf node is 0, indicating that its data purity is the highest. For each node other than the leaf node, when calculating the Gini index, we cannot just look at the temperature data of each node in isolation, but we must combine the temperature data of each node and its associated child nodes for a comprehensive evaluation.

[0092] For each binary tree, a temperature data set corresponding to each node is obtained, including: temperature data corresponding to the node itself, and temperature data corresponding to the associated child nodes of the node.

[0093] Classify all temperature data in the temperature data set corresponding to each node:

[0094] The temperature data with the same value are divided into the same category. For example, if all the temperature data are 20, 30, 30, and 50, then 20 is divided into one category, recorded as the first category, 30 is divided into one category, recorded as the second category, and 50 is divided into one category, recorded as the third category, for a total of 3 categories.

[0095] Then, count the proportion of temperature data belonging to each category in the temperature data set. In the above example, the proportion of temperature data belonging to category 1 in the temperature data set is , the proportion of temperature data belonging to the second category in the temperature data set is , the proportion of temperature data belonging to the third category in the temperature data set is .

[0096] According to the Gini index calculation formula, calculate the Gini index of each node:

[0097]

[0098] In the formula, express The Gini index of the node, is the sequence number of the category of all temperature data in the temperature data set of this node, is the total number of categories, The value of All integers in the range, Indicates that it belongs to The temperature data of the category is The proportion of the temperature data set corresponding to the node, Indicates that the temperature data belonging to category 1 is The proportion of the temperature data set corresponding to the node, Indicates that the temperature data belonging to category 2 is The proportion of the temperature data set corresponding to the node.

[0099] In the above example, when , , hour, .

[0100] For each binary tree, the Gini index of each node can be calculated according to the above method, and then the ratio of the number of all temperature data corresponding to each node (the corresponding temperature data set) to the number of all temperature data corresponding to the entire binary tree is used as the weight of the node. Finally, the weighted sum of all node weights and all node Gini indexes is performed to obtain the Gini index of the binary tree, and the inverse of the Gini index of the binary tree is used as the node purity of the binary tree.

[0101] The Gini index is an indicator that measures the degree of uneven distribution of categories in a data set. This step can deeply understand the distribution and purity of the water jacket heating furnace temperature data in the binary tree structure by calculating the Gini index of each node and the Gini index of the entire binary tree. The lower the Gini index of a binary tree, the higher the data purity of the node or binary tree, that is, the more concentrated the distribution of data in the high temperature and low temperature categories, which helps to accurately grasp the characteristics and laws of the data.

[0102] Finally, the reliability of the root node selection of each binary tree and the node purity of the binary tree are combined, and the product of the reliability of the root node selection and the node purity is taken as the partitioning ability value of the binary tree.

[0103] The partitioning ability value of a binary tree is used to reflect the partitioning ability of the binary tree. The higher the partitioning ability value of a binary tree is, the stronger the partitioning ability of the binary tree is. In the previous construction process, the tree has already partitioned the temperature data reasonably and effectively, forming a more representative node and branch structure. In this way, when faced with newly inserted temperature data, it can partition the newly inserted temperature data to accurate nodes based on the existing partitioning rules and node characteristics, giving a result that is closer to the actual situation.

[0104] S5: The weight of each binary tree is comprehensively evaluated based on its structural stability and segmentation capability.

[0105] The weight of the binary tree is determined by comprehensively considering factors such as its structural stability and segmentation ability. By setting the weight reasonably, we can give full play to the advantages of different binary trees when facing the volatility and complexity of the temperature data of the water jacket heating furnace, judge and process the inserted temperature nodes more accurately, and improve the analysis and decision-making ability of the entire system for temperature data.

[0106] As mentioned above, the temperature data of the water jacket heating furnace itself is volatile. The lower the structural stability of the binary tree, the more flexible the binary tree is and the more adaptable it is to the newly inserted temperature data. It can better adapt to this volatility and will not cause a significant change in the complexity of the binary tree. It is more tolerant of newly inserted temperature nodes and the possibility of misjudging normal temperature data as abnormal temperature data is smaller. It is precisely because the binary tree with low structural stability has the advantage of low possibility of misjudgment when processing volatile data such as water jacket heating furnaces, so it is given a higher weight. This reflects that in practical applications, more attention is paid to the adaptability and fault tolerance of the binary tree to normal temperature fluctuations, and it is hoped that by increasing its weight, it will play a more important role in overall decision-making or analysis.

[0107] As mentioned above, since the binary tree with high segmentation ability has higher accuracy in judging abnormal temperature data, its judgment results are more trustworthy in the entire analysis system, so it is given a higher weight. This makes the binary tree with high segmentation ability have a greater influence when considering the judgment results of multiple binary trees, thereby improving the reliability of the overall decision or analysis.

[0108] Therefore, this step calculates the weight of each binary tree according to the following formula:

[0109]

[0110] In this formula, For the The weight of a binary tree, and Respectively 、 The splitting ability value of a binary tree, and Respectively 、 The structural stability of a binary tree, is the total number of binary trees.

[0111] This formula comprehensively evaluates the performance of each binary tree by combining the splitting ability value with the structural stability. The splitting ability reflects the accuracy of the binary tree in classifying temperature data, and the structural stability reflects the adaptability of the binary tree to the insertion of new data. Both are indispensable.

[0112] The numerator of this formula represents the relative ratio of the partitioning ability and structural stability of the th binary tree, that is, the partitioning ability under unit structural stability. The denominator is the sum of the partitioning abilities of all binary trees under unit structural stability. By dividing the two, we can get the weight (normalized weight) of a binary tree, so that we can intuitively compare the relative importance of each binary tree in the whole.

[0113] For binary trees with strong segmentation ability and poor structural stability, The value of is relatively large, and it will get a higher proportion when calculating the weight, making it play a more important role in the overall decision. This helps to screen out the binary tree with better performance when processing the temperature data of the water jacket heating furnace and improve the accuracy of the overall judgment.

[0114] S6: Anomaly detection of real-time temperature data is performed based on the weight of each binary tree combined with the RRCF algorithm.

[0115] Each real-time temperature data is inserted into multiple pre-built randomly generated binary trees in turn. The RRCF algorithm is used to determine the change in complexity of each binary tree after the real-time temperature data is inserted. According to the weight corresponding to each binary tree determined previously, the change in complexity of each binary tree is weighted and summed, and for any real-time temperature data, the abnormality value of the real-time temperature data is calculated by the following formula:

[0116]

[0117] In the formula, is the abnormality of the real-time temperature data, , are all binary tree numbers, It is The weight of a binary tree, Insert the real-time temperature data into When the binary tree The change in complexity of a binary tree, is the total number of binary trees, Insert the real-time temperature data into When the binary tree The denominator is the total change in complexity of all binary trees caused by the real-time temperature data, and the numerator is the weighted sum of the change in complexity of each binary tree caused by the real-time temperature data. The fraction is constructed so that The value of Within the range.

[0118] Next, the threshold of the abnormality is set to 0.9 (empirical value), and the abnormality of each real-time temperature data is compared with 0.9. If the abnormality of a certain real-time temperature data is greater than 0.9, the real-time temperature data is judged to be abnormal and marked; if the abnormality is not greater than 0.9, the real-time temperature data is judged to be not abnormal. When abnormal temperature data is detected, the system immediately triggers an alarm to notify relevant personnel, thereby realizing temperature monitoring of the water jacket heating furnace and ensuring safe and stable operation of the system.

[0119] The present invention also provides a temperature monitoring system for a water jacket heating furnace, the temperature monitoring system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any step of the temperature monitoring method.

[0120] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0121] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A temperature monitoring method for a water jacket heating furnace, characterized in that: include: Obtain the historical temperature data and real-time temperature data of the water jacket heating furnace, randomly generate multiple binary trees based on the historical temperature data, and evaluate the structural stability of each binary tree and the reliability of the root node selection; The Gini index is calculated based on all the temperature data corresponding to each binary tree, and the inverse of the Gini index is used as the node purity of each binary tree. The product of the node purity of each binary tree and the reliability of the root node selection of each binary tree is used as the segmentation ability value of each binary tree. Calculate the weight of each binary tree: , For the The weight of a binary tree, and Respectively 、 The splitting ability value of a binary tree, and Respectively 、 The structural stability of a binary tree, is the total number of binary trees; The structural stability of each binary tree is evaluated as follows: The inverse of the variance value of the temperature data corresponding to the left subtree and the right subtree of each binary tree is taken as the stationarity of the left subtree and the right subtree; Determine the quantized values ​​of the data distribution characteristics of the left subtree and the right subtree according to the distribution characteristics of the temperature data corresponding to the left subtree and the right subtree; Calculate the structural stability of each binary tree: ； In the formula, For the The structural stability of a binary tree, is the left subtree, is the right subtree, , Respectively The quantitative value of the data distribution characteristics of the left and right subtrees of a binary tree. , Respectively The stability of the left and right subtrees of a binary tree; The reliability evaluation method of root node selection is: ； In the formula, For the The reliability of root node selection of a binary tree, is the left subtree, is the right subtree, For the The left subtree of a binary tree The value of the temperature data corresponding to each node, For the The value of the temperature data corresponding to the root node of the binary tree, For the The number of nodes in the left subtree of a binary tree, For the The right subtree of a binary tree The value of the temperature data corresponding to each node, For the The number of nodes in the right subtree of a binary tree, is the natural exponential function; Insert each real-time temperature data into all binary trees, obtain the change in complexity of each binary tree, use the weight of each binary tree to perform weighted summation on the change in complexity of each binary tree, calculate the degree of abnormality of each real-time temperature data, and determine whether each real-time temperature data is abnormal based on the degree of abnormality of each real-time temperature data, so as to realize temperature monitoring of the water jacket heating furnace.

2. The temperature monitoring method of a water jacket heating furnace according to claim 1, characterized in that: A method for determining the quantitative values ​​of the data distribution characteristics of the left subtree and the right subtree is: The ratio of the number of nodes in the left subtree and the right subtree to their respective depths is used as the quantitative value of the data distribution characteristics of the left subtree and the right subtree, respectively.

3. The temperature monitoring method of a water jacket heating furnace according to claim 1, characterized in that: Another method for determining the quantitative values ​​of the data distribution characteristics of the left subtree and the right subtree is: The arithmetic square root of the sum of the squares of the number of nodes and the depth of the left subtree and the right subtree is used as the quantitative value of the data distribution characteristics of the left subtree and the right subtree, respectively.

4. The temperature monitoring method of a water jacket heating furnace according to claim 1, characterized in that: The change in complexity of each binary tree is determined by the RRCF algorithm.

5. The temperature monitoring method of a water jacket heating furnace according to claim 1, characterized in that: The method of determining whether each real-time temperature data is abnormal according to the magnitude of the abnormality of each real-time temperature data is as follows: setting a threshold of the abnormality, if the abnormality of a certain real-time temperature data is greater than the threshold of the abnormality, the real-time temperature data is abnormal; If the abnormality level of a certain real-time temperature data is not greater than the abnormality threshold, the real-time temperature data is not abnormal.

6. The temperature monitoring method of a water jacket heating furnace according to claim 1, characterized in that: The calculation formula for the abnormal degree of each real-time temperature data is: In the formula, is the abnormality of the real-time temperature data, , They are all binary tree numbers. It is The weight of a binary tree, Insert the real-time temperature data into When the binary tree The change in complexity of a binary tree, is the total number of binary trees, Insert the real-time temperature data into When the binary tree The change in complexity of a binary tree.

7. A temperature monitoring system for a water jacket heating furnace, characterized in that: The temperature monitoring system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the temperature monitoring method according to any one of claims 1 to 6.

Images