Seamless steel tube production equipment state monitoring method and system

Abstract

The invention relates to the field of prediction data processing, in particular to a seamless steel tube production equipment state monitoring method and system.The seamless steel tube production equipment state monitoring method comprises the steps that data of all key parts of a continuous rolling rack are collected, feature vectors are constructed, and then the linear correlation degree of any two parameter time sequences is obtained to obtain a global linear correlation factor; further calculating the correlation degree of the harmonic energy ratio of the dominant frequency of the frequency spectrum corresponding to any two parameter time sequences, and obtaining a global dynamic coupling index so as to obtain a comprehensive correlation index which is used for optimizing a kernel function of an SVM classifier; and finally identifying the normal and abnormal states of the continuous rolling rack. According to the method, the SVM classifier based on the improved kernel function is adopted, and the state prediction model based on real-time data acquisition, feature extraction and optimization is adopted, so that efficient state monitoring and early warning of the continuous rolling rack are realized.

Description

Technical Field

[0001] The present invention relates to the field of predictive data processing, and in particular to a method and system for monitoring the state of seamless steel pipe production equipment. Background Art

[0002] The production process of seamless steel pipe mainly includes raw material preparation, heating, perforation, stretching, heat treatment, cooling and shaping, surface treatment and other links. In these processes, the continuous rolling mill is one of the key equipment, responsible for the continuous rolling of steel billets. The stable operation of the continuous rolling mill is crucial to the efficiency and quality of steel pipe production. Therefore, ensuring its normal operation is an important prerequisite for improving production efficiency and quality.

[0003] The operation status monitoring of the continuous rolling mill is of great significance for timely detection of equipment failures and reduction of production risks. Existing technologies usually use sensors to monitor the basic operation data of the equipment and determine whether the equipment is in normal working condition by setting thresholds; there are also related technologies that use models or algorithms for analysis and prediction to detect abnormal behavior of the equipment.

[0004] For example, a Chinese patent document with announcement number CN116562855B discloses an IoT maintenance management system for key equipment in the production of seamless steel pipes. It forms an equipment operation stability coefficient and provides intelligent maintenance suggestions through equipment data monitoring, historical maintenance analysis and status evaluation algorithms, solving the problems of extensive traditional equipment monitoring methods and low data utilization, and improving the scientificity and accuracy of equipment maintenance. However, this method has some technical defects: it focuses more on the historical data statistics and long-term maintenance management of equipment, and is suitable for the health assessment and optimization of overall production equipment. However, its real-time, predictive ability, intelligence and pertinence are insufficient, and it lacks the functions of real-time monitoring, early warning and dynamic adjustment of abnormal equipment status. Summary of the invention

[0005] In view of the above-mentioned lack of dynamic adjustment problem, in the first aspect, the present invention proposes a method for monitoring the state of seamless steel pipe production equipment, comprising: obtaining historical operation data of a continuous rolling mill stand, the historical operation data including the following operation parameters: temperature, amplitude, pressure and rotation speed; taking all historical operation data contained in a time window of a set length as feature vectors; adding labels to each feature vector, the labels and the corresponding feature vectors constitute sample data; training an SVM classifier based on the sample data; collecting the operation parameters in real time to obtain real-time operation data; taking all real-time operation data of the continuous rolling mill stand contained in the time window of a set length as real-time feature vectors, and inputting the real-time feature vectors into the trained SVM classifier to determine the current operation state of the continuous rolling mill stand; wherein the SVM classifier also includes optimized kernel function parameters, specifically: ; in Represents the optimized kernel function parameters; Represents the initial kernel function parameters; Represents the parameter adjustment factor; W represents the comprehensive correlation index, which is positively correlated with the global linear correlation factor and the global dynamic coupling index; the global linear correlation factor is positively correlated with the sum of the Pearson correlation coefficients of the operating parameters, and negatively correlated with the total number of non-repetitive paired parameters; the global dynamic coupling index is positively correlated with the sum of the correlations between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to each operating parameter, and negatively correlated with the total number of non-repetitive paired parameters; the dominant frequency refers to the frequency in the spectrum that corresponds to the maximum value of the harmonic energy ratio.

[0006] The method for monitoring the state of seamless steel pipe production equipment provided by the present invention achieves the purpose of real-time monitoring and accurate fault warning by collecting key parameters such as temperature, amplitude, pressure and rotation speed in real time during the operation of the equipment and using an improved SVM support vector machine classifier to perform state prediction. Existing methods mostly use single features or traditional machine learning methods, which are usually difficult to effectively process multi-dimensional data and nonlinear characteristics of complex equipment states. The present invention integrates linear and nonlinear features and trains the SVM model through an improved kernel function, so that when faced with high-dimensional and complex industrial equipment data, the model can more accurately perform state classification and fault risk prediction.

[0007] Furthermore, the calculation method of the global linear correlation factor is specifically as follows: ; Where G represents the global linear correlation factor; Indicates Parameters and Parameters Pearson correlation coefficient between them; M represents the total number of unrepeated paired parameters.

[0008] Furthermore, the calculation method of the global dynamic coupling index is specifically as follows: ; Where H represents the global dynamic coupling index; Indicates The parameter corresponds to the harmonic energy ratio of the dominant frequency of the spectrum. With The parameter corresponds to the harmonic energy ratio of the dominant frequency of the spectrum. Q represents the total number of non-repeated pairing parameters.

[0009] Furthermore, the correlation between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to the operating parameters is specifically: ; in Indicates the harmonic energy ratio and degree of relevance; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The spectrum mean of Indicates Parameters corresponding to the spectrum The mean value of the spectrum.

[0010] Furthermore, the correlation between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to the operating parameters is specifically: ; in Indicates the harmonic energy ratio and degree of relevance; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum With Parameters corresponding to the spectrum The mutual information of Represents the maximum value function.

[0011] The correlation degree in the present invention is measured by combining frequency domain analysis with mutual information, which can capture the nonlinear dynamic characteristics of the equipment during operation and improve the ability to identify complex equipment failure modes. In particular, under equipment load changes or special working conditions, the present invention can more sensitively detect potential abnormal trends and significantly improve the accuracy and robustness of equipment monitoring.

[0012] Furthermore, the calculation method of the comprehensive correlation index is specifically as follows: ; Where W represents the comprehensive correlation index; G represents the global linear correlation factor; exp( ) represents the natural exponential function; H represents the global dynamic coupling index; Represents the tuning factor.

[0013] The present invention combines the global linear correlation factor and the global dynamic coupling index through a nonlinear combination function to construct a comprehensive correlation index, which effectively enhances the adaptability of the model to complex equipment states. In particular, when the equipment state changes drastically or multiple factors are coupled, the present invention can more accurately evaluate the equipment operation state and provide more accurate input for subsequent state classification.

[0014] Furthermore, it also includes preprocessing operations on historical operation data, specifically: smoothing the historical operation data using a moving average filter; performing linear interpolation on the smoothed data; using the 3σ principle to eliminate outliers and replace the interpolated data with the median; and using the Z-Score model to standardize the data.

[0015] The data quality and reliability are effectively improved through steps such as moving average filtering, linear interpolation, 3σ principle and Z-Score standardization. First, moving average filtering is used to remove noise, smooth data, and reveal the long-term trend of the equipment; then missing values ​​are filled by linear interpolation to ensure data continuity and avoid information loss. The 3σ principle is used to eliminate outliers, and extreme anomalies are further eliminated by median replacement to ensure the representativeness and reliability of the data. Finally, Z-Score standardization is used to normalize the data to eliminate the influence of different dimensions on the model, so that the contribution of each feature to the model is more balanced. The preprocessing operation significantly improves the quality of the data and the stability of the model, ensuring the accuracy of subsequent condition monitoring and predictive analysis.

[0016] Furthermore, the real-time feature vector is input into a trained SVM classifier to determine the current operating status of the continuous rolling mill, and also includes: obtaining the real-time temperature, real-time amplitude, real-time pressure and real-time speed of the current time window of a set length; the last data of the current time window is the data at the current acquisition moment; all data in the current time window constitute a real-time feature vector; the real-time feature vector is input into a trained SVM classifier, and in response to the output result being an abnormal label, the staff is notified to inspect the continuous rolling mill.

[0017] Furthermore, the kernel function of the SVM classifier is a radial basis function.

[0018] In the SVM classifier training process, the present invention adopts an improved RBF radial basis function kernel and optimizes kernel parameters, which can effectively improve the classification accuracy of the classifier.

[0019] In a second aspect, the present invention provides a seamless steel pipe production equipment status monitoring system, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the seamless steel pipe production equipment status monitoring method of the present invention is implemented.

[0020] The technical effects of the present invention are: The present invention comprehensively reflects the equipment operation status by introducing multi-dimensional features such as linear correlation, nonlinear coupling and comprehensive correlation index, and optimizes the SVM model using an improved kernel function, thereby improving the classification accuracy and the adaptability of the model. Compared with the prior art, the present invention significantly improves the accuracy, real-time and robustness of fault prediction and equipment monitoring by adjusting kernel function parameters and comprehensive correlation indexes, reduces equipment failure rate, and optimizes production management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] 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: Figure 1 is a flow chart schematically showing a method for monitoring the state of a seamless steel pipe production equipment according to an embodiment of the present invention; Figure 2 is a color schematic diagram schematically showing a seamless steel pipe continuous rolling mill stand in an embodiment of the present invention; Figure 3 It is a structural block diagram schematically showing a seamless steel pipe production equipment status monitoring system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] 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.

[0023] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] Embodiment of the method for monitoring the state of seamless steel pipe production equipment: like Figure 1 As shown, a method for monitoring the state of a seamless steel pipe production equipment of the present invention comprises: S1. Collect historical operating parameters of the continuous rolling mill.

[0025] In the production process of seamless steel pipes, the tube blank needs to be pierced after being heated, and then it needs to enter the continuous rolling mill to roll the tube. Figure 2 As shown in the figure. As a key production equipment, the operation status of the continuous rolling mill directly affects the quality of steel pipes, production efficiency and service life of the equipment. Therefore, real-time monitoring of the operation status of the continuous rolling mill and mastering the changes in its key parameters are the basis for ensuring stable and efficient operation of the production process.

[0026] During the continuous rolling process, changes in parameters such as temperature, amplitude, pressure and speed can reflect the operating status of the equipment. For example, abnormal temperature may lead to a decline in steel performance, abnormal amplitude may indicate mechanical failure, pressure fluctuations may affect the stability of rolling force, and unstable speed may cause production rhythm disorder. Therefore, monitoring these key parameters can not only detect potential problems of equipment in a timely manner and prevent production interruptions, but also optimize production processes and improve product quality through data analysis.

[0027] Since the rollers on the continuous rolling mill stand are in direct contact with the seamless steel pipe, firstly, the TM750 infrared thermometer can be fixed on the outside of the continuous rolling mill stand to remotely collect the temperature data of the rollers; then, the 5860B vibration sensor probe can be installed at the connection between the rollers and the continuous rolling mill stand to collect the amplitude data during the operation of the continuous rolling mill stand; then, the SDB-600 hydraulic sensor can be installed in the hydraulic system of the continuous rolling mill stand to collect the hydraulic system pressure of the continuous rolling mill stand; finally, the XS12J4A speed sensor can be installed on the roller drive shaft of the continuous rolling mill stand to collect the speed data of the rollers. It should be noted that the acquisition frequency of the above-mentioned data acquisition devices is set to the empirical value of 20Hz, and the acquisition time is set to the empirical value of 2 hours; the implementer can also set the acquisition frequency to 10Hz, 30Hz or 60Hz according to the actual situation, and the above-mentioned data acquisition time period also includes the time period when the continuous rolling mill stand is in abnormal operation.

[0028] After collecting the relevant data of the rolling mill, the CH9121 industrial Ethernet interconnection module is used to collect all the data from the sensor and the thermometer and transmit them to the central database of the plant. Since the raw data directly collected may contain noise, missing values ​​and outliers, the accuracy of subsequent analysis is affected. Therefore, it is necessary to clean the collected raw data, including smoothing with a moving average filter, and then processing the smoothed data with a linear interpolation method, further using the 3σ principle to remove outliers and replace them with the median, and finally using the Z-Score model to standardize the data. The above-mentioned data cleaning and processing related algorithms belong to the known technology, and the implementation method will not be repeated here.

[0029] The temperature time series after data cleaning is recorded as C(t), the amplitude time series is recorded as V(t), the pressure time series is recorded as P(t), and the speed time series is recorded as R(t). So far, the time series set of relevant parameters of the continuous rolling mill stand is obtained: {C(t), V(t), P(t), R(t)}.

[0030] The data quality and reliability are effectively improved through steps such as moving average filtering, linear interpolation, 3σ principle and Z-Score standardization. First, moving average filtering is used to remove noise, smooth data, and reveal the long-term trend of the equipment; then missing values ​​are filled by linear interpolation to ensure data continuity and avoid information loss. The 3σ principle is used to eliminate outliers, and extreme anomalies are further eliminated by median replacement to ensure the representativeness and reliability of the data. Finally, Z-Score standardization is used to normalize the data to eliminate the influence of different dimensions on the model, so that the contribution of each feature to the model is more balanced. The preprocessing operation significantly improves the quality of the data and the stability of the model, ensuring the accuracy of subsequent condition monitoring and predictive analysis.

[0031] S2. Construct sample data of an SVM support vector machine based on the parameter time series set obtained in step S1.

[0032] In step S1, a time series set {C(t), V(t), P(t), R(t)} containing abnormal data related to the continuous rolling mill is obtained after data cleaning. In this embodiment, a feature vector is subsequently constructed based on the time series set to train an SVM support vector machine to predict the state of the continuous rolling mill and realize state monitoring.

[0033] For SVM, its sample data mainly includes feature vectors and labels. Therefore, the feature vector X is constructed first. The time series set {C(t), V(t), P(t), R(t)} can be divided into time periods according to the time window. In this embodiment, the length of the time window can be set to an empirical value of 5 seconds. Then The feature vector corresponding to the time window can be expressed as: Xω=[Cω(1), Cω(2),…, Cω(N), Vω(1), Vω(2),…, Vω(N), Pω(1), Pω(2),…, Pω(N), Rω(1), Rω(2),…, Rω(N)]. It should be noted that the feature vectors mentioned later in this embodiment are The eigenvectors refer to The feature vector corresponding to each time window.

[0034] Where Cω(1), Cω(2),…, Cω(N) represent the The feature vector contains a total of N temperature data; Vω(1), Vω(2), …, Vω(N) represents the first The feature vector contains a total of N amplitude data; Pω(1), Pω(2),…, Pω(N) represents the first The feature vector contains a total of N pressure data; Rω(1), Rω(2),…, Rω(N) represents the first The feature vectors contain a total of N rotation speed data. Since the length of the time window is set to 5 seconds and the acquisition frequency is set to 20 Hz in this embodiment, N can be obtained as 100.

[0035] After obtaining the feature vector of the SVM support vector machine, professional staff will label each feature vector according to the operating status of the continuous rolling mill in each corresponding time period. The feature vector corresponding to the time window is Xω, and the professional staff will label it according to the actual working status of the corresponding rolling mill stand in the time window. When there is an abnormal state in the time window corresponding to the time period, the label is marked as: "0"; when the continuous rolling stand is in the first If the operation status is normal within the time period corresponding to the first time window, the label is: "1"; The label corresponding to the feature vector is recorded as Yω. So far, the sample data of the SVM support vector machine has been obtained. Since the collection time set in this embodiment is 2 hours and the time window is 5 seconds, there are 2400 feature vectors and corresponding labels in this embodiment. Finally, 80% of all 2400 sample data are used as training sets, and the remaining 20% ​​are used as validation sets.

[0036] Since there may be a complex nonlinear relationship between the state parameter data of the continuous rolling mill, the kernel function of the SVM support vector machine in this embodiment selects the radial basis function kernel, which is expressed as: ; in represents the i-th eigenvector and the vth eigenvector similarity; exp( ) represents the natural exponential function; Represents the feature vector With the eigenvector The square of the Euclidean distance between them; The training process of the above-mentioned SVM support vector machine is a well-known technology, and the specific implementation method will not be described in detail here.

[0037] S3. Obtain the Pearson correlation coefficient of any two parameter time series and construct a correlation matrix, and obtain the global linear correlation factor based on the correlation matrix; obtain the harmonic energy ratio of the dominant frequency of the spectrum corresponding to any two parameter time series, and obtain the global dynamic coupling index based on the harmonic correlation of each harmonic energy ratio; obtain the comprehensive correlation index based on the global linear correlation factor and the global dynamic coupling index, and optimize the kernel function of the SVM support vector machine based on the comprehensive correlation index.

[0038] In step S2, the SVM kernel function in this embodiment is set to be a radial basis kernel function, and the parameters of the kernel function are fixed values. However, due to the complex coupling relationship between the parameters of the continuous rolling mill, a single and fixed kernel function parameter cannot well improve the accuracy and robustness of the SVM prediction model. Therefore, in this embodiment, the operating parameters of the continuous rolling mill are analyzed to optimize the parameters in the SVM kernel function.

[0039] For the time series set {C(t), V(t), P(t), R(t)} obtained in step S1, first obtain the linear correlation between the time series of each parameter, which can be obtained by the Pearson correlation coefficient in this embodiment. Parameter timing and Parameter timing The Pearson correlation coefficient between , obtain the Pearson correlation coefficient of any two parameter time series in all parameter time series and construct the correlation matrix D, specifically: ; Where D represents the correlation matrix of the time series set, and then the global linear correlation factor is calculated based on the off-diagonal elements in the correlation matrix. The calculation method is: ; Where G represents the global linear correlation factor; Indicates Parameter timing and Parameter timing Pearson correlation coefficient between them; M represents the total number of feature pairs. In this embodiment, only the non-diagonal elements of the correlation matrix D and the symmetry are considered to obtain a total number of 6 feature pairs.

[0040] The global linear correlation factor G can reflect the coupling degree of linear parameters during the operation of the continuous rolling mill. For example, when there is a high positive correlation between temperature and amplitude, that is, , and when the correlation between other parameter time series is low, the global linear correlation factor G will be significantly improved. However, for the continuous rolling mill, its operating state is not only affected by the linear relationship between the time series of each parameter, but also by the complex nonlinear coupling effect. Exemplary explanation: The nonlinear increase in temperature may lead to a significant increase in the energy of certain specific frequency bands in the amplitude spectrum, thereby affecting the stability of the pressure and speed of the continuous rolling mill. The above-mentioned linear correlation metric cannot fully capture these nonlinear relationships. Therefore, in this embodiment, a dynamic coupling index is constructed to quantify the nonlinear correlation between the parameters.

[0041] First, perform FFT on each parameter time series to obtain the spectrum of each parameter time series and obtain the parameter spectrum set {C(f), V(f), P(f), R(f)}. For any parameter spectrum , any frequency component The calculation method of harmonic energy ratio is: ; in Representing spectrum Mid-frequency components Harmonic energy ratio; Represents the total energy of the spectrum. Further calculation of the spectrum The harmonic energy ratio of all frequency components in and obtain the parameter spectrum The frequency components with the highest energy ratio are: ; in represents the dominant frequency; Indicates that the harmonic energy ratio The frequency f corresponding to the maximum value. Then, the harmonic energy ratio corresponding to the dominant frequency is As a parameter spectrum represents the harmonic energy ratio and is recorded as .

[0042] For each parameter spectrum in the parameter spectrum set {C(f), V(f), P(f), R(f)}, based on the above representative harmonic energy ratio The calculation method to obtain the representative harmonic energy ratio of each parameter spectrum is: , , as well as The specific calculation method will not be described here.

[0043] Accordingly, for any Parameter timing and Parameter timing Get parameter spectrum and , further obtain the parameter spectrum and The representative harmonic energy ratio and , then calculate the harmonic energy ratio and Harmonic correlation of , in one embodiment, the harmonic correlation It can also be recorded as the correlation degree, and the specific calculation method is: ; in Indicates the harmonic energy ratio and The harmonic correlation of Indicates Time series The harmonic energy ratio corresponding to the dominant frequency of the spectrum; Indicates Time series The harmonic energy ratio corresponding to the dominant frequency of the spectrum; Indicates Time series Corresponding spectrum The spectrum mean of Indicates Time series Corresponding spectrum The mean value of the spectrum.

[0044] In another embodiment, the calculation method of harmonic correlation is specifically as follows: ; in Indicates the harmonic energy ratio and The harmonic correlation of Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum With Parameters corresponding to the spectrum The mutual information of represents the maximum value function; the calculation process of the above mutual information belongs to the well-known technology, and the specific calculation method will not be repeated here.

[0045] Finally, the global dynamic coupling index H is calculated. The specific calculation process is as follows: ; Where H represents the global dynamic coupling index; Indicates the harmonic energy ratio and Pearson correlation coefficient; Q represents the total number of non-repeated pairing parameters. In this embodiment, there are 4 parameters, so the total number that can be obtained through permutations and combinations is 6. .

[0046] The correlation degree in the present invention is measured by combining frequency domain analysis with mutual information, which can capture the nonlinear dynamic characteristics of the equipment during operation and improve the ability to identify complex equipment failure modes. In particular, under equipment load changes or special working conditions, the present invention can more sensitively detect potential abnormal trends and significantly improve the accuracy and robustness of equipment monitoring.

[0047] So far, the global linear correlation factor G and the global dynamic coupling index H have been obtained, and the comprehensive correlation index W is further calculated. The specific calculation method is: ; Where W represents the comprehensive correlation index; G represents the global linear correlation factor; exp( ) represents the natural exponential function; H represents the global dynamic coupling index; It represents the parameter adjustment factor, which is used to control the amplifying effect of the global dynamic coupling index H on the global linear correlation factor G. In this embodiment, it can be set to an empirical value of 0.5. The implementer can also set it to 0.4, 0.6 or 0.7 according to the actual situation.

[0048] The present invention combines the global linear correlation factor and the global dynamic coupling index through a nonlinear combination function to construct a comprehensive correlation index, which effectively enhances the adaptability of the model to complex equipment states. In particular, when the equipment state changes drastically or multiple factors are coupled, the present invention can more accurately evaluate the equipment operation state and provide more accurate input for subsequent state classification.

[0049] The parameters of the optimized kernel function are finally defined as , then the calculation method in this embodiment is: ; in Represents the parameters of the optimized kernel function; represents the initial parameters of the kernel function, which can be obtained by cross-validation in this embodiment, and the specific implementation method will not be repeated here; W represents the comprehensive correlation index; It represents the parameter adjustment factor, which is used to control the influence of the comprehensive correlation index W on the parameters of the kernel function. In this embodiment, it can be set to an empirical value of 0.5.

[0050] When the comprehensive correlation index W is larger, it means that the data distribution in the feature space is more complex, and there are more nonlinear correlations between the operating parameters. At this time, the parameters of the kernel function should be reduced to expand the influence range of the RBF radial basis function kernel, make the decision boundary of the model smoother, reduce overfitting of local complexity, and thus improve the generalization ability of the model. When the comprehensive correlation index W is smaller, it means that the data distribution in the feature space is relatively simple and the correlation between features is weak. At this time, the parameters of the kernel function should be increased to reduce the influence range of the RBF radial basis function kernel and enhance the sensitivity of the model to local features, thereby improving the accuracy of classification. The bigger.

[0051] So far, the optimized kernel function can be obtained: ; in represents the i-th eigenvector and the vth eigenvector similarity; exp( ) represents the natural exponential function; Represents the feature vector With the eigenvector The square of the Euclidean distance between them; Represents the parameters of the optimized kernel function.

[0052] S4. Training an SVM support vector machine based on the improved kernel function in step S3, and implementing the state monitoring of the continuous rolling mill stand based on the trained SVM support vector machine classifier.

[0053] During the training process, the SVM support vector machine is firstly supervised by learning based on the sample data obtained in step S2, and then optimized by using the improved kernel function in step S3 to improve the classification accuracy of the model.

[0054] The trained SVM support vector machine can process the newly collected feature vectors in real time. Therefore, the data collection method of step S1 is used to obtain the real-time parameters of the continuous rolling stand in the production process of seamless steel pipes, and input them into the trained SVM support vector machine for prediction and classification. The production status of the continuous rolling stand is monitored based on the classification results. Exemplary explanation: The real-time parameter time series set obtained by the real-time temperature time series, amplitude time series, pressure time series and speed time series is recorded as: , and obtain the data of the current time window of each real-time parameter time series, that is, the data corresponding to the current acquisition time node is taken as the last data in the current time window; then for the real-time temperature time series , the real-time temperature data of the current time window are: ; in Indicates the current collection time; It indicates the last temperature data of the current time window; represents the first temperature data in the current time window; N represents the total number of temperature data in the current time window. Here, the length of the time window and the device acquisition frequency are the same as those in step S1, and the corresponding total number N is 100.

[0055] In the same way, the real-time amplitude data, real-time pressure data and real-time speed data of the current time window are obtained, which are recorded as: ; ; ; The acquisition process of the above real-time amplitude data, real-time pressure data and real-time speed data is the same as the acquisition method of real-time temperature data, and the specific implementation method is not repeated here. So far, all real-time data of the current time window have been acquired and constructed as a real-time feature vector X * , input it into the trained SVM support vector machine classifier to obtain the prediction result of the real-time operating status of the continuous rolling stand. If the output label is "0", that is, the abnormal label, it is determined that the continuous rolling stand currently used to produce seamless steel pipes may have the risk of operating failure. The equipment automatically alarms and notifies relevant technical personnel to perform fault inspection and repair on the continuous rolling stand to ensure the production quality of seamless steel pipes and reduce the failure risk rate of the continuous rolling stand.

[0056] The method for monitoring the state of seamless steel pipe production equipment provided by the present invention achieves the purpose of real-time monitoring and accurate fault warning by collecting key parameters such as temperature, amplitude, pressure and rotation speed in real time during the operation of the equipment and using an improved SVM support vector machine classifier to perform state prediction. Existing methods mostly use single features or traditional machine learning methods, which are usually difficult to effectively process multi-dimensional data and nonlinear characteristics of complex equipment states. The present invention integrates linear and nonlinear features and trains the SVM model through an improved kernel function, so that when faced with high-dimensional and complex industrial equipment data, the model can more accurately perform state classification and fault risk prediction.

[0057] Example of seamless steel pipe production equipment status monitoring system: On the other hand, the present invention also provides a seamless steel pipe production equipment status monitoring system. Figure 3 As shown, the seamless steel pipe production equipment status monitoring system includes a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a seamless steel pipe production equipment status monitoring method according to the first aspect of the present invention is implemented.

[0058] The seamless steel pipe production equipment status monitoring system also includes other components familiar to those skilled in the art, such as a communication interface, and their settings and functions are known in the art, so they will not be described in detail here.

[0059] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0060] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0061] 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 method for monitoring the state of seamless steel pipe production equipment, characterized in that: The method comprises: Acquire historical operation data of the continuous rolling mill stand, the historical operation data including the following operation parameters: temperature, amplitude, pressure and rotation speed; use all historical operation data contained in a time window of a set length as feature vectors; add labels to each feature vector, the labels and corresponding feature vectors constitute sample data; train an SVM classifier based on the sample data; The operating parameters are collected in real time to obtain real-time operating data; all real-time operating data of the continuous rolling mill frame contained in a set length time window are used as real-time feature vectors, and the real-time feature vectors are input into a trained SVM classifier to determine the current operating state of the continuous rolling mill frame; The SVM classifier also includes optimized kernel function parameters, specifically: ; in Represents the optimized kernel function parameters; Represents the initial kernel function parameters; represents a parameter adjustment factor; W represents a comprehensive correlation index, which is positively correlated with the global linear correlation factor and the global dynamic coupling index; the global linear correlation factor is positively correlated with the sum of the Pearson correlation coefficients of the operating parameters, and is negatively correlated with the number of non-repeated pairing parameters; The global dynamic coupling index is positively correlated with the sum of the correlations between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to each operating parameter, and negatively correlated with the total number of non-repeated paired parameters; the dominant frequency refers to the frequency in the spectrum that corresponds to the maximum value of the harmonic energy ratio.

2. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: The calculation method of the global linear correlation factor is specifically as follows: ; Where G represents the global linear correlation factor; Indicates Parameters and Parameters Pearson correlation coefficient between them; M represents the total number of unrepeated paired parameters.

3. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: The calculation method of the global dynamic coupling index is specifically as follows: ; Where H represents the global dynamic coupling index; Indicates The parameter corresponds to the harmonic energy ratio of the dominant frequency of the spectrum. With The parameter corresponds to the harmonic energy ratio of the dominant frequency of the spectrum. Q represents the total number of non-repeated pairing parameters.

4. A method for monitoring the state of seamless steel pipe production equipment according to claim 3, characterized in that: The correlation between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to the operating parameters is specifically: ; in Indicates the harmonic energy ratio and degree of relevance; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The spectrum mean of Indicates Parameters corresponding to the spectrum The mean value of the spectrum.

5. A method for monitoring the state of seamless steel pipe production equipment according to claim 3, characterized in that: The correlation between the harmonic energy ratios of the dominant frequencies in the spectrum corresponding to the operating parameters is specifically: ; in Indicates the harmonic energy ratio and degree of relevance; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum The harmonic energy ratio of the dominant frequency; Indicates Parameters corresponding to the spectrum With Parameters corresponding to the spectrum The mutual information of Represents the maximum value function.

6. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: The calculation method of the comprehensive correlation index is specifically as follows: ; Where W represents the comprehensive correlation index; G represents the global linear correlation factor; exp( ) represents the natural exponential function; H represents the global dynamic coupling index; Represents the tuning factor.

7. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: It also includes preprocessing operations on historical operation data, specifically: The historical operation data is smoothed using a moving average filter; Perform linear interpolation on the smoothed data; The interpolated data were subjected to outlier elimination using the 3σ principle and median replacement; The Z-Score model was used for data standardization.

8. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: The real-time feature vector is input into the trained SVM classifier to determine the current operating state of the continuous rolling mill stand, including: Get the real-time temperature, real-time amplitude, real-time pressure and real-time speed of the current time window of the set length; The last data of the current time window is the data at the current collection time; All data within the current time window constitute a real-time feature vector; The real-time feature vector is input into a trained SVM classifier, and in response to the output result being an abnormal label, a staff member is notified to perform maintenance on the continuous rolling mill stand.

9. A method for monitoring the state of seamless steel pipe production equipment according to claim 1, characterized in that: The kernel function of the SVM classifier is a radial basis function.

10. A seamless steel pipe production equipment status monitoring system, characterized in that: It comprises a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for monitoring the state of seamless steel pipe production equipment according to any one of claims 1 to 9 is implemented.

Images