A silicon steel furnace roller fault identification and alarm system and method based on a gradient boosting tree method

Abstract

This invention discloses a fault identification and alarm system and method for silicon steel furnace rolls based on the gradient boosting tree method. Based on the gradient boosting tree method, it utilizes furnace roll operation data within the silicon steel furnace, including time-series data and dimensional space data on rotational speed and current, to accurately identify, classify, and alarm for faults such as slight bending, heavy bending, and carbon sleeve wear of the furnace rolls. The steps include capturing furnace roll data from the existing IBA system, analyzing and filtering data in the frequency domain using dimensionality algorithms and existing AI models, and identifying and alarming the operating conditions. This invention extracts data by adding a developed plug-in to the IBA PDA, and through analysis, statistical calculation, and fault analysis of this data, it provides a direct understanding of the furnace roll's performance, technical level, and potential improvements, and triggers pre-defined alarm events to alert workers.

Description

Technical Field

[0001] This invention relates to the field of metallurgical cold rolling technology, and in particular to a fault identification and alarm system and method for silicon steel furnace rolls based on the gradient boosting tree method. Background Technology

[0002] A silicon steel furnace is a piece of equipment used to produce silicon steel, and the furnace rollers are an important component of the furnace. During the operation of the silicon steel furnace, the furnace rollers play a crucial role. They not only need to support and conduct heat, but also withstand the weight of the silicon steel billet and the corrosive effects of the furnace atmosphere. Therefore, the furnace rollers need to possess high strength and rigidity to ensure that they do not deform or break during long-term operation within the furnace.

[0003] Time-series fault detection is an important task in engineering, aiming to detect and predict equipment or system failures in a timely manner by analyzing and modeling system sensor data. Machine learning methods are widely used in time-series fault detection, with supervised learning being a common approach. Supervised learning methods utilize labeled training data to build models and predict the category or label of new data. In time-series fault detection, supervised learning methods can be trained using known fault and normal data to build fault detection models.

[0004] In the prior art, patent application number CN202410631010.9 discloses a method for assessing and maintaining the health status of furnace rollers, with the following steps: S1: Classification of health status levels; S2: Acquiring vibration signals from the fixed and free ends of different health states; S3: Labeling all processed samples with health status tags; S4: Constructing a BP neural network model; S5: Initializing the weights and biases of the model's network layers, training the model with training data, saving the model parameters after convergence, testing the evaluation results with test data, and saving the model; S6: Testing; S7: Writing the trained evaluation program into the system; S8: Setting the inspection cycle; S9: Acquiring vibration signals from the target furnace roller; S10: Preprocessing, calculating 12 time-domain feature indicators at both ends to generate a one-dimensional vector, totaling 5; S11: Using the model to diagnose the one-dimensional vector. If 3 out of 5 consecutive values ​​correspond to a certain label, that label represents the current health status, and the deteriorated or faulty furnace rollers are replaced.

[0005] However, the generalization ability of BP neural networks is limited by the quality and quantity of training data. Insufficient training data can cause the model to fail to accurately identify new or unknown fault modes. In addition, the model's parameter settings, such as the number of network layers and neurons, also affect its generalization ability. Improper parameter settings may lead to overfitting or underfitting of the model.

[0006] Therefore, it is necessary to improve it to overcome the above-mentioned defects. Summary of the Invention

[0007] The purpose of this invention is to provide a silicon steel furnace roll fault identification and alarm system and method based on the gradient boosting tree method, which can be used to accurately identify early faults in silicon steel furnaces, avoid the expansion of losses, and thus improve production efficiency.

[0008] The above-mentioned technical objective of this invention has been achieved by the following technical solutions:

[0009] A fault identification and alarm system for silicon steel furnace rolls based on the gradient boosting tree method includes,

[0010] The IBAPDA server's data input terminal is connected to the signal source for communication. The IBAPDA server is used to collect and record the operating data of the furnace rollers.

[0011] The database server's data input end communicates with the IBAPDA server's data output end. The IBAPDA server transmits the acquired data to the database server for subsequent data analysis and processing.

[0012] A method for using a silicon steel furnace roll fault identification and alarm system based on the gradient boosting tree method includes the following steps:

[0013] A1: Data connection, connecting to the data source via API to provide data support for the alarm system;

[0014] A2: Data capture, which involves periodically capturing furnace roll data from the data source;

[0015] A3: Frequency domain analysis, used to perform frequency domain analysis on the acquired furnace roll data;

[0016] A4: Data filtering is used to remove invalid, duplicate, or erroneous data to improve the accuracy of analysis results;

[0017] A5: Fault Identification and Classification, used to quickly identify and classify faults;

[0018] A6: Trigger alarm event, used for real-time monitoring of key indicators and determining whether to issue an alarm.

[0019] The data connection in step A1 specifically includes the following steps:

[0020] B1: Obtain the IBAPDA server's API key file and API password through the executable process, determine the IBAPDA server's IP address, configure network settings, and establish a connection through the data acquisition interface for communication with the IBAPDA server, which is used to realize data interaction with the silicon steel furnace unit.

[0021] The data scraping in step A2 specifically includes the following steps:

[0022] C1: Set up a continuous loop task and event triggering mechanism to obtain the operating data of the silicon steel furnace unit through the data acquisition interface on the IBAPDA server, and periodically retrieve furnace roller data from the IBAPDA server.

[0023] The frequency domain analysis in step A3 specifically includes the following steps:

[0024] A3: Frequency domain analysis. The captured furnace roller data is analyzed in the frequency domain, the time domain data is converted into frequency domain data, and the characteristic data of the furnace roller operating status is extracted. A three-dimensional algorithm based on the Fourier function is established, including data in three dimensions: peak-to-peak value, second peak value of the spectrum, and variance. By calculating the furnace roller speed and load current data in a two-minute time sequence, the characteristic data of the equipment cycle is obtained.

[0025] The data filtering in step A4 specifically includes the following steps:

[0026] A4: Data filtering. The gradient boosting tree model is used to train the furnace roller operation data to create the training set and test set required for model training. This enables the model to distinguish between normal and abnormal data and filter data according to preset filtering conditions.

[0027] The fault identification and classification in step A5 specifically includes the following steps:

[0028] A5: Fault identification and classification. The filtered data is input into the trained gradient boosting tree model for statistical analysis and fault analysis. The classifier model identifies and classifies fault results such as gradual deterioration of furnace rolls, severe bending of furnace rolls, and wear of the carbon sleeve of furnace rolls.

[0029] The triggering of the alarm event in step A6 specifically includes the following steps:

[0030] A6: Trigger alarm events, set alarm rules and conditions, and trigger the corresponding alarm event based on the fault identification results when a furnace roller fault is detected.

[0031] The derivation of the gradient boosting tree model in step A4 includes the following steps: First, input the dataset. and a differentiable loss function

[0032] Step 1: That is, constant Initialize the model F0(x);

[0033] Step 2: for m = 1 to M:

[0034] a) For i = 1, ..., n, calculate

[0035] b)Fit a regression tree to the r im values ​​and create terminal regionsR jm , where j = 1, ..., J m ;

[0036] c) For j = 1, ..., J m ,calculate

[0037] d) Update

[0038] Step 3: Output F M (x).

[0039] In summary, the present invention has the following beneficial effects:

[0040] 1. Accuracy: By utilizing the gradient boosting tree method and time series data analysis, the operating status of the furnace rolls is monitored in real time. By continuously capturing data and performing frequency domain analysis, potential faults can be detected in a timely manner. It can accurately identify and classify faults such as gradual deterioration of the furnace rolls, severe bending of the furnace rolls, and internal wear of the carbon sleeve of the furnace rolls, thereby improving the accuracy of fault diagnosis.

[0041] 2. Real-time performance: The system can monitor the operation of the furnace rollers in real time, and perform data filtering and analysis according to preset screening conditions, triggering alarm events in a timely manner so that workers can take measures quickly to reduce the loss of steel production.

[0042] 3. Automation: By adding the developed executable process to IBAPDA, the furnace roll data features are calculated using a three-dimensional algorithm, and the AI ​​machine learning model trained based on the gradient boosting tree model is called. This makes it easy to obtain furnace roll operation data, perform data analysis and fault diagnosis, and reduce the workload and errors of manual operation.

[0043] 4. Improve output: By comparing the operating data of the furnace rolls under different furnace conditions, the standard for the optimal operating condition of each furnace roll can be established, which can effectively improve the output level and reduce production costs. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the structure of the present invention.

[0045] Figure 2 This is a schematic diagram of the structure of the present invention.

[0046] Figure 3 This is a schematic diagram of the structure of the present invention. Detailed Implementation

[0047] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below with reference to the figures and specific embodiments.

[0048] In this embodiment, the present invention is applied to the task of fault detection and classification of furnace rolls in silicon steel furnaces. Specifically, it employs a gradient boosting tree ensemble learning algorithm, which can construct a powerful classifier or regressor by combining multiple weak learners. To adapt to the differences in furnace conditions of different units, this system considers the influence of factors such as furnace temperature, strip speed, and strip tension. By adding a developed executable process to the IBA series software, furnace roll operation data can be easily obtained. After algorithm calculation, model training result matching, and fault analysis, this data provides a direct understanding of the furnace roll's performance, technical level, and potential improvements, and triggers pre-set alarm events to remind workers.

[0049] like Figures 1 to 3 As shown, the present invention proposes a fault identification and alarm system for silicon steel furnace rolls based on the gradient boosting tree method, comprising:

[0050] The IBAPDA server is equipped with a data acquisition interface. The IBAPDA server communicates with the signal source through the data acquisition interface. The IBAPDA server is used to collect and record the operating data of the furnace rollers. The IBAPDA server also has an executable process, which is written in Python.

[0051] The IBA software is used to identify and classify furnace roll faults. Specifically, the IBA software is used to extract the furnace roll's operating data. By adding an executable process to the IBAPA server, the required information can be quickly extracted from the furnace roll data, and the data can be filtered and analyzed to obtain fault result characteristics and complete fault diagnosis.

[0052] The database server has a data input terminal that communicates with the data output terminal of the IBAPDA server. The IBAPDA server transmits the acquired data to the database server for subsequent data analysis and processing. The database server stores the algorithm program, which the IBAPDA server can call through an external program to execute the furnace roll detection algorithm.

[0053] When in use, the executable program calls the API key file and API password of the IBAPDA server, establishes an API connection through the IP address of the IBAPDA server, and periodically captures the real-time operating data of the furnace roller motor.

[0054] By employing the gradient boosting tree method and through continuous iteration and optimization, high-precision identification of furnace roll faults in silicon steel furnaces can be achieved, accurately capturing fault characteristics in the furnace roll operation data. By setting an executable process on the IBAPDA server, the furnace roll operation data can be collected and recorded in real time, and fault identification and classification can be performed quickly.

[0055] A method for using a silicon steel furnace roll fault identification and alarm system based on the gradient boosting tree method includes the following steps:

[0056] A1: Data connection, obtains the API key file and API password of the IBAPDA server through the executable process, determines the IP address of the IBAPDA server, configures network settings, and establishes a connection through the data acquisition interface for communication with the IBAPDA server, which is used to realize data interaction with the silicon steel furnace unit;

[0057] By automatically obtaining API key files and API passwords through an executable process, and automating network settings configuration and connection establishment, the efficiency of the entire data interaction process is improved, allowing remote monitoring of the operating status and performance of the silicon steel furnace unit.

[0058] A2: Capture data, set up continuous loop tasks and event triggering mechanisms, obtain the operating data of the silicon steel furnace unit through the data acquisition interface on the IBAPDA server, and periodically capture furnace roll data from the IBAPDA server;

[0059] Preferably, the data capture time is two minutes each time. The captured data includes time-series data of parameters such as furnace roller speed and load current. After two minutes, the system switches to the next furnace roller motor, and the data is captured in a loop to ensure the continuity and accuracy of the data. The system captures the operating data of the silicon steel furnace unit in real time, providing rich time-series data to support in-depth analysis of the performance, efficiency, and faults of the silicon steel furnace unit.

[0060] A3: Frequency domain analysis. The captured furnace roller data is analyzed in the frequency domain. The time domain data is converted into frequency domain data, and the characteristic data of the furnace roller operating status is extracted. A three-dimensional algorithm based on the Fourier function is established, including data in three dimensions: peak-to-peak value, second peak value of the spectrum, and variance. The characteristic data of the equipment cycle are obtained by calculating the furnace roller speed and load current data in a two-minute time sequence.

[0061] A4: Data filtering. Use the gradient boosting tree model to train the furnace roller's operating data, create the training set and test set required for model training, so that the model can identify the difference between normal data and abnormal data, and filter data according to preset filtering conditions.

[0062] The three-dimensional results of the furnace roll operation in step A3 are trained by using a gradient boosting tree model. The furnace roll operation data are classified according to the differences in different furnace conditions. The parameters of the training set and the test set are adjusted, and different screening conditions are set to meet the requirements of the model's judgment accuracy under actual working conditions.

[0063] A5: Fault identification and classification. The filtered data is input into the trained gradient boosting tree model for statistical analysis and fault analysis. The classifier model identifies and classifies fault results such as gradual deterioration of furnace rolls, severe bending of furnace rolls, and wear of the carbon sleeve of furnace rolls.

[0064] A6: Trigger alarm events. Set alarm rules and conditions. When a furnace roller fault is detected, trigger the corresponding alarm event based on the fault identification result to remind the staff.

[0065] In summary, by automating the data capture, processing, and fault identification process, manual intervention is reduced, and work efficiency and accuracy are improved. By using a gradient boosting tree model to perform deep learning and analysis on the furnace roll operation data, various furnace roll fault types can be accurately identified, such as gradual deterioration, severe bending, and internal wear of the carbon sleeve. The system can monitor the furnace roll operation status in real time and trigger an alarm immediately upon detecting a fault, providing strong support for timely maintenance and avoiding production interruptions.

[0066] The derivation steps of the gradient boosting tree model in step A4 include: first, inputting the dataset... and a differentiable loss function

[0067] Step 1: That is, constant Initialize the model F0(x);

[0068] Step 2: for m = 1 to M:

[0069] a) For i = 1, ..., n, calculate

[0070] b)Fit a regression tree to the r im values ​​and create terminal regionsR jm , where j = 1, ..., J m ;

[0071] c) For j = 1, ..., J m ,calculate

[0072] d) Update

[0073] Step 3: Output F M (x).

[0074] The working principle and process of this invention are as follows: A fault identification and alarm system for silicon steel furnace rolls based on the gradient boosting tree method is proposed. The alarm system utilizes the furnace roll operation data in the silicon steel furnace, including time sequence and dimensional space rotational speed and current, to accurately identify and classify faults such as slight bending of the furnace rolls, severe bending of the furnace rolls, and wear of the furnace roll carbon sleeve. Through automated data capture and processing, manual intervention is reduced and work efficiency is improved. By using the gradient boosting tree model to perform deep learning and analysis on the furnace roll operation data, it can accurately identify multiple types of furnace roll faults. The system can monitor the furnace roll operation status in real time and trigger an alarm immediately once a fault is detected, providing support for timely maintenance and avoiding production interruptions.

[0075] Specifically, the process includes: capturing furnace roll data from the existing IBA system; analyzing the frequency domain of furnace roll operation using dimensionality algorithms and gradient boosting tree models; filtering out data with standard symbols; determining the furnace roll's operating status based on the filtered data; analyzing, processing, calculating, and classifying the acquired data using IBA software to gain a direct understanding of the furnace roll's operating status; setting corresponding trigger alarm times to remind staff; and iteratively training multiple weak learners to gradually reduce prediction errors, thereby improving model performance.

[0076] In summary, this invention provides a novel method for fault detection in steel mills. It applies frequency domain analysis to the preprocessing of furnace roll operating data, extracting more meaningful feature data. By utilizing a gradient boosting tree model for deep learning and analysis of the furnace roll operating data, the accuracy and efficiency of fault identification are improved. Real-time monitoring of the furnace roll operating status and triggering alarm events provide strong support for timely maintenance and preventing production interruptions, enabling workers to detect damaged furnace rolls earlier and reducing steel production losses. Furthermore, it establishes standards for the optimal operating conditions of each furnace roll in the unit, and by comparing furnace roll operating data across different furnace conditions, it effectively improves output levels.

[0077] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," "left," and "right," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. These terms are used only for the convenience of describing this invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, terms such as "set" and "connect" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. In this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, which includes not only the elements listed but also other elements not expressly listed.

[0078] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A fault identification and alarm system for silicon steel furnace rolls based on the gradient boosting tree method, characterized in that, include, The IBAPDA server's data input terminal is connected to the signal source for communication. The IBAPDA server is used to collect and record the operating data of the furnace rollers. The database server's data input end communicates with the IBAPDA server's data output end. The IBAPDA server transmits the acquired data to the database server for subsequent data analysis and processing.

2. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 1, characterized in that, Includes the following steps: A1: Data connection, connecting to the data source via API to provide data support for the alarm system; A2: Data capture, which involves periodically capturing furnace roll data from the data source; A3: Frequency domain analysis, used to perform frequency domain analysis on the acquired furnace roll data; A4: Data filtering is used to remove invalid, duplicate, or erroneous data to improve the accuracy of analysis results; A5: Fault Identification and Classification, used to quickly identify and classify faults; A6: Trigger alarm event, used for real-time monitoring of key indicators and determining whether to issue an alarm.

3. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The data connection in step A1 specifically includes the following steps: B1: Obtain the API key file and API password of the IBAPDA server through the executable process, determine the IP address of the IBAPDA server, configure network settings, and establish a connection through the data acquisition interface for communication with the IBAPDA server, which is used to realize data interaction with the silicon steel furnace unit.

4. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The data scraping in step A2 specifically includes the following steps: C1: Set up a continuous loop task and event triggering mechanism to obtain the operating data of the silicon steel furnace unit through the data acquisition interface on the IBAPDA server, and periodically retrieve furnace roller data from the IBAPDA server.

5. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The frequency domain analysis in step A3 specifically includes the following steps: A3: Frequency domain analysis. The captured furnace roller data is analyzed in the frequency domain, the time domain data is converted into frequency domain data, and the characteristic data of the furnace roller operating status is extracted. A three-dimensional algorithm based on the Fourier function is established, including data in three dimensions: peak-to-peak value, second peak value of the spectrum, and variance. By calculating the furnace roller speed and load current data in a two-minute time sequence, the characteristic data of the equipment cycle is obtained.

6. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The data filtering in step A4 specifically includes the following steps: A4: Data filtering. The gradient boosting tree model is used to train the furnace roller operation data to create the training set and test set required for model training. This enables the model to distinguish between normal and abnormal data and filter data according to preset filtering conditions.

7. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The fault identification and classification in step A5 specifically includes the following steps: A5: Fault identification and classification. The filtered data is input into the trained gradient boosting tree model for statistical analysis and fault analysis. The classifier model identifies and classifies fault results such as gradual deterioration of furnace rolls, severe bending of furnace rolls, and wear of the carbon sleeve of furnace rolls.

8. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The triggering of the alarm event in step A6 specifically includes the following steps: A6: Trigger alarm events, set alarm rules and conditions, and trigger the corresponding alarm event based on the fault identification results when a furnace roller fault is detected.

9. The method for using a silicon steel furnace roll fault identification and alarm system based on gradient boosting tree method according to claim 2, characterized in that, The derivation of the gradient boosting tree model in step A4 includes the following steps: First, input the dataset. and a differentiable loss function Step 1: That is, constant Initialize the model F0(x); Step 2: for m = 1 to M: a) For i = 1, ..., n, calculate b) Fit a regression tree to the r im values and create terminal regions R jm , where j = 1, …, J m ; c) For j = 1, ..., J m ,calculate d) Update Step 3: Output F M (x).

Images