A method and device for locating abnormal operation of a cement clinker burning system

Abstract

The application discloses a cement burning system operation abnormality positioning method and device. The method first acquires real-time data segments of each measuring point in the cement burning system; for the real-time data segment of each measuring point, the acquired data segment in a target time interval is subjected to standardization processing; then, region division is performed; the data corresponding to the neighborhood of the abnormal data time interval after the region division is determined, and the data corresponding to the neighborhood is coded to obtain the characteristic value of the current measuring point; finally, the range corresponding to each measuring point is determined according to the characteristic value of each measuring point, and when the range corresponding to a preset number of measuring points exceeds a threshold value, it is determined that the system in the target time interval is abnormally operated. The method can accurately position the position and severity of the abnormality in the case that there is little abnormal data and no model is used.

Description

Technical Field

[0001] This invention relates to the field of abnormal detection technology for cement firing systems, and in particular to a method and apparatus for locating abnormalities in the operation of cement firing systems. Background Technology

[0002] The cement calcination system, as the most critical link in the cement production process, is characterized by numerous parameters, complex operating conditions, nonlinearity, large time lag, and strong coupling. Once an anomaly occurs during production, rapid analysis and root cause identification are crucial; otherwise, delayed handling may affect the normal operation of the entire system, and in severe cases, may even lead to catastrophic accidents. A good anomaly localization method can quickly and accurately detect the characteristics and magnitude leading to the anomaly, reduce its duration, prevent it from developing into a more serious fault, and avoid or minimize losses to the enterprise caused by the anomaly. With the rapid development of computer technology and communication storage technology, production units have accumulated massive amounts of historical production data. Data-driven algorithms such as machine learning and statistical analysis have emerged and are being applied to anomaly localization tasks.

[0003] Existing anomaly localization methods mainly fall into two categories: localization methods based on association rule mining and localization methods based on decision trees.

[0004] Association rule mining is a rule-based machine learning algorithm that uses some metrics to identify association rules in a dataset. It requires labeled data. Once the association rules are generated, they can be used to assist in correlation analysis. When an anomaly occurs, association rule matching can be used to locate the features that caused the anomaly.

[0005] Decision tree algorithms belong to supervised learning and require labeled data to train the model. They represent a mapping between features and values; each node in the tree represents a condition for a feature's judgment, and its branches represent data that meets that condition. Through model training, a mapping between anomaly types and data can be established. Feature importance scores can be used to determine the degree of influence of each feature on the result, thereby identifying the features that cause the anomaly.

[0006] However, both association rule-based and decision tree-based anomaly localization methods rely on labeled anomaly data and are widely used in the operations and maintenance industry. However, in actual engineering project implementation, production units do not summarize and statistically analyze anomaly data, and the amount of anomaly data is extremely small, making it difficult to support model training and apply to engineering projects. Summary of the Invention

[0007] Based on this, the present application provides a method and apparatus for locating anomalies in a cement firing system. The purpose is to accurately locate the location and severity of anomalies by using coding when there is very little abnormal data and no model is used. At the same time, it performs secondary detection on the basis of anomaly detection, reducing the false detection rate and having strong interpretability, improving the intelligence of production, reducing losses caused by anomalies in production units, and reducing the labor intensity of workers.

[0008] Firstly, a method for locating operational anomalies in a cement calcination system is provided. The cement calcination system includes a preheater assembly, a decomposition furnace assembly, a rotary kiln assembly, and a grate cooler assembly. The method includes:

[0009] Acquire real-time data segments from various measuring points in the cement firing system;

[0010] For each real-time data segment at each measuring point, the data segments acquired within the target time interval are standardized.

[0011] Divide the value range of the standardized data fragments into regions;

[0012] The data corresponding to the neighborhood of the abnormal data time interval after the region is divided is determined, and the data corresponding to the neighborhood is encoded to obtain the feature value of the current measurement point;

[0013] The range corresponding to each measuring point is determined based on the characteristic value of each measuring point. When the range corresponding to a preset number of measuring points exceeds the threshold, the system is determined to be abnormal during the target time interval. Among them, the measuring points corresponding to the range exceeding the threshold are abnormal measuring points.

[0014] Optionally, acquiring real-time data segments from various measuring points in the cement firing system also includes:

[0015] Measuring points are set in each component of the cement firing system; multiple measuring points can be set in each component.

[0016] The data measured at each measuring point is stored in the database, and real-time data fragments are retrieved from the database.

[0017] An abnormal data index is obtained through anomaly detection based on real-time data fragments from each measuring point.

[0018] Optionally, the data segments acquired within the target time interval are standardized, including:

[0019] The z-score standardization method is used to standardize the obtained data fragments; the standardized result is determined according to a first formula, which specifically includes:

[0020]

[0021] Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

[0022] Optionally, the value range of the standardized data fragments can be divided into regions, including:

[0023] The standardized range of the current measuring point is determined to be [a, b]. This range is then divided into c regions, and the span of each region is... The final partitioning results include:

[0024]

[0025] The corresponding numbers for each region are as follows:

[0026] Optionally, the data corresponding to the neighborhood of the abnormal data time interval after the region is divided is determined, and the data corresponding to the neighborhood is encoded to obtain the feature value of the current measurement point, including:

[0027] Data within the time dimension neighborhood of abnormal data is extracted, and the time segment data is encoded. Each value of each data point is scaled and rounded proportionally, and each data point is encoded into a number combination. The number combination is used to represent the value of the current measurement point.

[0028] Secondly, a device for locating abnormalities in a cement firing system is provided, the device comprising:

[0029] The acquisition module is used to acquire real-time data fragments from various measuring points in the cement firing system.

[0030] The standardization module standardizes the data segments acquired within the target time interval for each real-time data segment at each measuring point.

[0031] The region partitioning module is used to partition the value range of standardized data fragments into regions.

[0032] The encoding module is used to determine the data corresponding to the neighborhood of the abnormal data time interval after the region is divided, and to encode the data corresponding to the neighborhood to obtain the feature value of the current measurement point.

[0033] The anomaly localization module is used to determine the range corresponding to each measuring point based on the characteristic values ​​of each measuring point. When the range corresponding to a preset number of measuring points exceeds a threshold, the system operation in the target time interval is determined to be abnormal. Among them, the measuring points corresponding to the range exceeding the threshold are abnormal measuring points.

[0034] Optionally, the acquisition module further includes:

[0035] Measuring points are set in each component of the cement firing system; multiple measuring points can be set in each component.

[0036] The data measured at each measuring point is stored in the database, and real-time data fragments are retrieved from the database.

[0037] An abnormal data index is obtained through anomaly detection based on real-time data fragments from each measuring point.

[0038] Optionally, the standardization module includes:

[0039] The z-score standardization device is used to standardize the acquired data fragments; wherein, the standardized result is determined according to a first formula, the first formula specifically including:

[0040]

[0041] Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

[0042] Optionally, the region division module includes:

[0043] The standardized range of the current measuring point is determined to be [a, b]. This range is then divided into c regions, and the span of each region is... The final partitioning results include:

[0044]

[0045] The corresponding numbers for each region are as follows:

[0046] Optionally, the encoding module includes:

[0047] Data within the time dimension neighborhood of abnormal data is extracted, and the time segment data is encoded. Each value of each data point is scaled and rounded proportionally, and each data point is encoded into a number combination. The number combination is used to represent the value of the current measurement point.

[0048] The technical solution provided in this application first acquires real-time data segments from various measuring points in the cement firing system; for each measuring point's real-time data segment, the data segments acquired within the target time interval are standardized; then, regions are divided; the data corresponding to the neighboring regions of the abnormal data time intervals after region division is determined, and the data corresponding to the neighboring regions is encoded to obtain the feature value of the current measuring point; finally, the range corresponding to each measuring point is determined based on the feature value of each measuring point, and when the range corresponding to a preset number of measuring points exceeds a threshold, the system operation in the target time interval is determined to be abnormal. It can be seen that the beneficial effects of this invention are:

[0049] (1) This method has no requirement for the amount of data and solves the problem of very little abnormal data;

[0050] (2) This method does not involve complex nonlinear models and solves the problem of unstable and difficult-to-interpret machine learning model outputs. Attached Figure Description

[0051] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0052] Figure 1 A flowchart of a method for locating operational anomalies in a cement firing system provided in this application embodiment;

[0053] Figure 2 This is a schematic diagram of a cement firing system provided in an embodiment of this application;

[0054] Figure 3 This is a schematic diagram illustrating anomaly localization in an embodiment of this application. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0056] In the description of this invention, the terms “comprising,” “having,” and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations of the inventive concept.

[0057] The purpose of this application is to solve the problem of accurately locating the position and severity of anomalies using coding methods when there is very little abnormal data and no model is used. It also performs secondary detection based on anomaly detection, reducing the false positive rate and providing strong interpretability, thereby improving production intelligence, reducing losses caused by anomalies in production units, and reducing the labor intensity of workers. For details, please refer to... Figure 1 The document illustrates a flowchart of a method for locating operational anomalies in a cement firing system, as provided in an embodiment of this application. This method may include the following steps:

[0058] S1, acquire real-time data segments from various measuring points in the cement firing system.

[0059] In this application, the cement firing system can be as follows: Figure 2 As shown, the system mainly consists of four parts: a preheater, a decomposition furnace, a rotary kiln, and a grate cooler. The ground raw meal is fed into the preheater by an elevator, and then enters the rotary kiln for calcination under inertia. Next, under the influence of gravity and the constant speed rotation of the rotary kiln, the raw meal slowly moves from the kiln tail to the kiln head. During this movement, a series of physical and chemical reactions occur under high-temperature calcination. The high-temperature, red-hot clinker discharged from the kiln head then enters the grate cooler for rapid cooling under inertia. The grate cooler completes the cooling of the cement clinker and the recovery of the finished clinker product. Real-time data from various measuring points monitor the operation of the entire calcination system. When an anomaly occurs, the variable causing the anomaly is quickly located, and the severity of the anomaly is calculated.

[0060] Specifically, the measuring points in this application can be set in various components of the cement firing system, and multiple measuring points can be set in each component.

[0061] In this embodiment of the application, this step includes obtaining real-time data segments and abnormal data indexes. Specifically, after obtaining real-time data segments from the database, abnormal data indexes are obtained through anomaly detection.

[0062] S2 standardizes the data segments acquired within the target time interval for each real-time data segment at each measurement point.

[0063] This step mainly involves data standardization. Specifically, the z-score standardization method is used to standardize the acquired data fragments, eliminating differences in units, reducing the impact of outliers, and facilitating subsequent coding operations. The formula is shown below:

[0064]

[0065] Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

[0066] In optional embodiments of this application, the above standardization process can also be a normalization process, that is, the acquired data segments are normalized and scaled to a fixed range to facilitate subsequent encoding operations.

[0067] S3 divides the value range of the standardized data fragments into regions.

[0068] In this step, the specific process of dividing the region is as follows:

[0069] Suppose a data segment has k data points, where the standardized value range of a certain feature, b, is [a, b]. We need to divide this segment into c (where c is an odd number) regions. The span of each region is then... The final partitioning results are shown below:

[0070]

[0071] The corresponding numbers for each region are as follows:

[0072] S4. Determine the data corresponding to the neighborhood of the abnormal data time interval after the region is divided, and encode the data corresponding to the neighborhood to obtain the feature value of the current measurement point.

[0073] In this step, temporal neighborhood data is extracted and encoded: j data points within the temporal neighborhood of the outlier data are extracted. Assuming the index of this outlier data is i, the index range of the extracted outlier data's neighborhood data in the temporal dimension is... The time neighborhood data is encoded according to the region number obtained in the previous step. Each value of each data point is encoded according to the number corresponding to its respective range. In this way, each data point can be encoded as a combination of numbers, and each number represents the value of the feature corresponding to that data point.

[0074] S5. Determine the range corresponding to each measuring point based on the characteristic values ​​of each measuring point. When the range corresponding to a preset number of measuring points exceeds the threshold, it is determined that the system is malfunctioning in the target time interval.

[0075] Among them, the measurement points whose range exceeds the threshold are abnormal measurement points.

[0076] This step primarily focuses on anomaly localization and severity. The cause (feature) of the anomaly is identified by determining whether the uniformly encoded data features exhibit drastic fluctuations. The previous step encoded the extracted time-neighborhood data, and the range of each feature within that time-neighborhood is calculated. The range represents the maximum number of cross-regions for that feature, i.e., the maximum amplitude of fluctuation. If more than n features have a cross-region number greater than m, meaning more than n features have generated fluctuations with amplitudes greater than m, then an anomaly has indeed occurred within that time segment. These n or more features are the identified causes of the anomaly. The number of cross-regions represents the severity of the anomaly (curve fluctuation amplitude); a larger number of cross-regions indicates a more pronounced and severe anomaly.

[0077] As can be seen, this application mainly monitors the operation of the entire firing system through real-time data from various measuring points. When an anomaly occurs, the variable causing the anomaly is quickly located, and the severity of the anomaly is calculated.

[0078] In another embodiment of this application, an anomaly index is obtained after the real-time data fragment obtained from the database is processed by an anomaly detection method. The data fragment and the anomaly index are then passed to the anomaly localization program. The anomaly localization program first performs standardization and dimensionless processing on the fragment data to scale the data to a fixed range for subsequent encoding operations. Next, it performs region division to divide the data value range into regions.

[0079] like Figure 3The diagram illustrates anomaly localization. Assuming feature a, after standardization, has a value range of [-0.2, 2.3], dividing it into 5 regions yields the following partitions: [-0.2, 0.3), [0.3, 0.8), [0.8, 1.3), [1.3, 1.8), [1.8, 2.3]. Similarly, feature b, after standardization, has a value range of [0.4, 2.5], dividing it into 7 regions yields the following partitions: [0.4, 0.7), [0.7, 1.0), [1.0], [0.7, 1.0], [0.8, 1.3 ... The value range of feature c after standardization is [0.3, 2.1]. Dividing it into three regions, the resulting division is [0.3, 0.9), [0.9, 1.5), [1.5, 2.1]. Then, data within the time dimension neighborhood of the abnormal data (the part between the dotted lines on the right in the figure) is extracted and encoded into numerical combinations, completing the data encoding. Each numerical combination is a data point, and each number corresponds to a feature at a specific location. The range of each feature is calculated for the encoded numerical combinations. The range represents the maximum number of cross-domains. If more than n features have cross-domains greater than m, they are considered abnormal data. The feature corresponding to the range is the identified feature causing the anomaly. The number of cross-domains indicates the degree of anomaly; the larger the number of cross-domains, the more obvious and profound the anomaly.

[0080] like Figure 3 The data contains three features: a, b, and c. The data encoding part involves standardizing one of the data points (left dashed line) according to the above process, resulting in standard deviations of 2.12, 1.87, and 0.21. The encoded result is "210". After designing the encoding rules, encoding the data within the red dashed line range according to the same rules easily yields the following results: feature a has a maximum value of 2 and a minimum value of 0; feature b has a maximum value of 1 and a minimum value of -2; and feature c has a maximum value of 1 and a minimum value of 0. Therefore, the cross-domain numbers for features a, b, and c are 2, 3, and 1, respectively.

[0081] During online monitoring, this method performs a secondary detection of the anomaly detection results from the perspective of curve jumps, which significantly reduces the false alarm rate of the anomaly detection method. At the same time, it accurately locates the root cause features that cause the anomaly, and achieves anomaly localization without relying on abnormal data and models.

[0082] In summary, this application starts from the essence of anomalies, reconstructs feature data through encoding, reduces the false detection rate of anomaly detection methods, and achieves the quantification of anomaly severity. At the same time, it combines engineering practice to avoid dependence on anomalous data, avoids training of complex nonlinear models, and consumes less computing resources while achieving high computing speed.

[0083] This application also provides a device for locating abnormalities in a cement firing system. The device includes:

[0084] The acquisition module is used to acquire real-time data fragments from various measuring points in the cement firing system.

[0085] The standardization module standardizes the data segments acquired within the target time interval for each real-time data segment at each measuring point.

[0086] The region partitioning module is used to partition the value range of standardized data fragments into regions.

[0087] The encoding module is used to determine the data corresponding to the neighborhood of the abnormal data time interval after the region is divided, and to encode the data corresponding to the neighborhood to obtain the feature value of the current measurement point.

[0088] The anomaly localization module is used to determine the range corresponding to each measuring point based on the characteristic values ​​of each measuring point. When the range corresponding to a preset number of measuring points exceeds a threshold, the system operation in the target time interval is determined to be abnormal. Among them, the measuring points corresponding to the range exceeding the threshold are abnormal measuring points.

[0089] In an optional embodiment of this application, the acquisition module further includes:

[0090] Measuring points are set in each component of the cement firing system; multiple measuring points can be set in each component.

[0091] The data measured at each measuring point is stored in the database, and real-time data fragments are retrieved from the database.

[0092] An abnormal data index is obtained through anomaly detection based on real-time data fragments from each measuring point.

[0093] In optional embodiments of this application, the standardization module includes:

[0094] The z-score standardization device is used to standardize the acquired data fragments; the standardized result is determined according to a first formula, which specifically includes:

[0095]

[0096] Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

[0097] In an optional embodiment of this application, the region division module includes:

[0098] The standardized range of the current measuring point is determined to be [a, b]. This range is then divided into c regions, and the span of each region is... The final partitioning results include:

[0099]

[0100] The corresponding numbers for each region are as follows:

[0101] In an optional embodiment of this application, the encoding module includes:

[0102] Data within the time dimension neighborhood of abnormal data is extracted, and the time segment data is encoded. Each value of each data point is scaled and rounded proportionally, and each data point is encoded into a number combination. The number combination is used to represent the value of the current measurement point.

[0103] The cement firing system operation anomaly location device provided in this application embodiment is used to implement the above-described cement firing system operation anomaly location method. Specific limitations of the cement firing system operation anomaly location device can be found in the limitations of the cement firing system operation anomaly location method described above, and will not be repeated here. Each component of the above-described cement firing system operation anomaly location device can be implemented entirely or partially through software, hardware, or a combination thereof. Each of the above modules can be embedded in or independent of the processor in the device in hardware form, or stored in the memory of the device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0105] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for locating operational anomalies in a cement calcination system, the cement calcination system comprising a preheater assembly, a decomposition furnace assembly, a rotary kiln assembly, and a grate cooler assembly, characterized in that, The method includes: Acquire real-time data segments from various measuring points in the cement firing system; For each real-time data segment at each measuring point, the data segments acquired within the target time interval are standardized. Divide the value range of the standardized data fragments into regions; The data corresponding to the neighborhood of the abnormal data time interval after the region is divided is determined, and the data corresponding to the neighborhood is encoded to obtain the feature value of the current measurement point; The range corresponding to each measuring point is determined based on the characteristic value of each measuring point. When the range corresponding to a preset number of measuring points exceeds the threshold, the system operation is determined to be abnormal in the target time interval. Among them, the measuring points corresponding to the range exceeding the threshold are abnormal measuring points. Acquiring real-time data segments from various measuring points in the cement firing system also includes: Measuring points are set in each component of the cement firing system; multiple measuring points are set in each component. The data measured at each measuring point is stored in the database, and real-time data fragments are retrieved from the database. An abnormal data index is obtained through anomaly detection based on real-time data fragments from each measuring point.

2. The method for locating operational anomalies in a cement firing system according to claim 1, characterized in that, The data segments acquired within the target time interval are standardized, including: The z-score standardization method is used to standardize the obtained data fragments; the standardized result is determined according to a first formula, which specifically includes: Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

3. The method for locating operational anomalies in a cement firing system according to claim 1, characterized in that, The value range of the standardized data fragments is divided into regions, including: The standardized range of the current measuring point is determined to be [a, b]. This range is then divided into c regions, and the span of each region is... The final partitioning results include: The corresponding numbers for each region are as follows:

4. The method for locating operational anomalies in a cement firing system according to claim 1, characterized in that, After identifying the data corresponding to the neighborhood of the abnormal data time intervals after the region is divided, the data corresponding to the neighborhood is encoded to obtain the feature values ​​of the current measurement point, including: Data within the time dimension neighborhood of abnormal data is extracted, and the time segment data is encoded. Each value of each data point is scaled and rounded proportionally, and each data point is encoded into a number combination. The number combination is used to represent the value of the current measurement point.

5. A device for locating abnormalities in a cement firing system, characterized in that, The device includes: The acquisition module is used to acquire real-time data fragments from various measuring points in the cement firing system. The standardization module standardizes the data segments acquired within the target time interval for each real-time data segment at each measuring point. The region partitioning module is used to partition the value range of standardized data fragments into regions. The encoding module is used to determine the data corresponding to the neighborhood of the abnormal data time interval after the region is divided, and to encode the data corresponding to the neighborhood to obtain the feature value of the current measurement point. The anomaly localization module is used to determine the range corresponding to each measuring point based on the feature values ​​of each measuring point. When the range corresponding to a preset number of measuring points exceeds a threshold, the system operation is determined to be abnormal in the target time interval; among them, the measuring points corresponding to the range exceeding the threshold are abnormal measuring points. The acquisition module also includes: Measuring points are set in each component of the cement firing system; multiple measuring points are set in each component. The data measured at each measuring point is stored in the database, and real-time data fragments are retrieved from the database. An abnormal data index is obtained through anomaly detection based on real-time data fragments from each measuring point.

6. The cement firing system malfunction positioning device according to claim 5, characterized in that, The standardization module includes: The z-score standardization device is used to standardize the acquired data fragments; wherein, the standardized result is determined according to a first formula, the first formula specifically including: Where, x i Represents real data. y represents the mean, s represents the standard deviation, and y represents the mean. i This represents the standardized result.

7. The cement firing system malfunction positioning device according to claim 5, characterized in that, The region division module includes: The standardized range of the current measuring point is determined to be [a, b]. This range is then divided into c regions, and the span of each region is... The final partitioning results include: The corresponding numbers for each region are as follows:

8. The cement firing system malfunction positioning device according to claim 5, characterized in that, The encoding module includes: Data within the time dimension neighborhood of abnormal data is extracted, and the time segment data is encoded. Each value of each data point is scaled and rounded proportionally, and each data point is encoded into a number combination. The number combination is used to represent the value of the current measurement point.

Images