SnO2 electrode digital twinborn management and control system for special glass electric melting application

Abstract

The invention belongs to the technical field of electrode digital analysis, and discloses a SnO2 electrode digital twinborn management and control system for special glass electric melting application. Comprising the steps of obtaining cycle control signal sequences and electrode working condition data of multiple debugging cycles of an electrode; fitting electrode working condition data of the electrode in different debugging periods to obtain a performance index sequence of the electrode; performing signal type identification on the plurality of periodic control signal sequences in sequence, constructing signal type sets, and extracting a signal time-frequency feature sequence of each signal type set; a sensitive signal type set is extracted by analyzing the relevance of a signal time-frequency feature sequence and a performance index sequence on the sequence change trend; and according to aggregation characteristics of different sensitive signal type sets, an optimal signal sequence combination is extracted, a digital twinborn body is constructed, and real-time control signals of the electrodes are managed and controlled. The influence of time-frequency variation of different control signals on the electrode performance attenuation rate is captured, the signal regulation and control efficiency is improved, and the electrode signal management and control mode is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrode digital analysis, more specifically, the present application relates to SnO2 electrode digital twin control system for special glass electric melting application. BACKGROUND

[0002] Electric melting glass technology is a technology that uses the Joule heat generated by the resistance of the melt itself to realize glass melting by introducing electric current into the glass melt through electrodes. Compared with traditional fuel heating, it is more direct and precise in temperature control. When using SnO2 electrodes, they have excellent high-temperature resistance and can withstand the high-temperature environment required for glass melting. They are also chemically stable and less likely to react with the glass melt to ensure the purity of the glass. At the same time, their good electrical conductivity can efficiently conduct electric current to ensure a stable and efficient melting process.

[0003] The melting temperature, viscosity, and electrical conductivity of different special glasses differ significantly, and the control signal parameters of SnO2 electrodes (such as current intensity and voltage frequency) are different. For example, high-melting-point special glasses require higher power from the electrodes, and the time-frequency characteristics of their control signals differ fundamentally from those of low-melting-point glasses. Therefore, when using SnO2 electrodes for electric melting of special glasses, multiple control signal adjustments are required for adaptation. In existing signal control methods, it is difficult to capture the influence of time-frequency changes in different control signal types on the performance degradation rate of SnO2 electrodes during multiple electrode signal adjustment processes. It is also difficult to integrate the control signal sequences after adjusting different signal types in each adjustment process, thereby optimizing the overall control signal of the adjustment period and improving the control of SnO2 electrodes. This can accelerate the wear of SnO2 electrodes. SUMMARY

[0004] To overcome the above-mentioned defects of the prior art and achieve the above-mentioned purposes, the present application provides the following technical solutions: SnO2 electrode digital twin control system for special glass electric melting application, comprising: A data acquisition module acquires periodic control signal sequences and electrode operating condition data of multiple adjustment periods of the SnO2 electrode. A data fitting module fits the electrode operating condition data of different adjustment periods of the SnO2 electrode to obtain a performance index sequence of the SnO2 electrode. A feature extraction module sequentially identifies the signal types of multiple periodic control signal sequences, constructs a signal type set, and extracts a signal time-frequency feature sequence of each signal type set. A data analysis module extracts a sensitive signal type set by analyzing the correlation between the signal time-frequency feature sequence and the performance index sequence in terms of sequence trend. The digital twin construction module extracts the aggregation characteristics of the sensitive signal type set on the signal time-frequency characteristics, extracts an optimal signal sequence combination according to the aggregation characteristics of different sensitive signal type sets, and constructs a digital twin according to the optimal signal sequence combination. The signal control optimization module controls and manages the real-time control signal of the SnO2 electrode based on the digital twin.

[0005] Preferably, the signal type set is constructed, including: The signal types in the periodic control signal sequence are identified, and the signal types are spliced according to the time sequence relationship of different control signals in the periodic control signal sequence to obtain a signal type sequence. A time window-based pattern growth algorithm is used to extract a high-frequency subsequence in the multiple signal type sequences, which is denoted as a high-frequency subsequence. Different signal types in the high-frequency subsequence are combined to obtain a signal type set of the signal type sequence. The signal type sets of all signal type sequences are merged to obtain a combination library. The frequency of each signal type set in the combination library is obtained in sequence, and the signal type set with a frequency greater than a preset frequency threshold is retained.

[0006] Preferably, the signal time-frequency characteristic sequence of each signal type set is extracted, including: The probability distribution of the performance index is calculated, and multiple probability distribution intervals are determined. The debugging periods corresponding to different performance indexes in each probability distribution interval are combined to obtain a period set. The periodic control signal sequence corresponding to each debugging period in the period set is marked as a target signal sequence. The signal subsequence corresponding to each signal type set in the target signal sequence is identified, and the time-frequency characteristics of the extracted signal subsequence are used as the signal time-frequency characteristics of the signal type set in the target signal sequence. The multiple signal time-frequency characteristics of the signal type set are combined in the time sequence order of the debugging periods to obtain a signal time-frequency characteristic sequence.

[0007] Preferably, the sensitive signal type set is extracted, including: The sequence variation characteristics of the signal time-frequency characteristic sequence and the performance index sequence are analyzed respectively, and a feature similarity sequence and an index attenuation amount sequence are constructed; the feature similarity sequence corresponds to the signal time-frequency characteristic sequence; and the index attenuation amount sequence corresponds to the performance index sequence. The deflection angle of the index attenuation amount sequence at each sequence point is extracted, denoted as a first deflection angle. The deflection angle of the feature similarity sequence at each sequence point is extracted, denoted as a second deflection angle; the second deflection angle corresponds to the first deflection angle one by one. Calculate the absolute difference between each second deflection angle and the corresponding first deflection angle, denoted as the relative deflection angle difference of the second deflection angle; Compare the relative deflection angle differences of different second deflection angles, and mark the second deflection angle corresponding to the minimum relative deflection angle difference as the target deflection angle; Compare the total number of target deflection angles in each feature similarity sequence, and retain the feature similarity sequence whose total number of target deflection angles is greater than the preset number threshold; Integrate the retained feature similarity sequences, and mark the signal type set corresponding to the feature similarity sequence as a candidate signal type set of the cycle set; Combine the candidate signal type sets of different cycle sets to obtain a candidate library, and count the frequency proportion of each candidate signal type set; Compare the frequency proportion of the candidate signal type set with the preset frequency threshold, and mark the candidate signal type set whose frequency proportion is greater than the frequency threshold as a sensitive signal type set.

[0008] Preferably, the construction of the feature similarity sequence and the index attenuation amount sequence comprises: Traverse the signal time-frequency feature sequence, and perform similarity measurement on two adjacent signal time-frequency features to obtain a feature similarity; Combine the feature similarities of multiple adjacent signal time-frequency features to obtain a feature similarity sequence; Traverse the performance index sequence, and calculate the difference value of two adjacent performance indexes to obtain an index attenuation amount; Combine the index attenuation amounts of multiple adjacent performance indexes to obtain an index attenuation amount sequence.

[0009] Preferably, the extraction of the optimal signal sequence combination comprises: Combine multiple signal time-frequency features of the same sensitive signal type set to obtain a sensitive feature set; Extract the feature similarity of any two sensitive features in the sensitive feature set, and cluster the corresponding debugging cycles according to the feature similarity to obtain multiple cluster centers; Determine each cluster center as a candidate node of the sensitive signal type set; Extract the control timing of the sensitive signal type set, and sequentially select any candidate node of each sensitive signal type set for connection according to the control timing sequence of different sensitive signal type sets; According to the multiple possibilities of connecting the candidate nodes corresponding to different sensitive signal type sets, exhaust the candidate node paths to obtain several candidate node paths; Comprehensively evaluate different candidate nodes in the candidate node path by combining the path attenuation degree and the cycle matching degree to determine the optimal node path; A plurality of debugging periods in each candidate node in the selected optimal node path are marked as selected periods corresponding to the sensitive signal type; The signal subsequences of the sensitive signal type set in different selected periods are combined to obtain an optimal signal sequence combination of the sensitive signal type set.

[0010] Preferably, the determining of the optimal node path comprises: The path attenuation degree of the candidate node path is extracted by analyzing the stability of different candidate nodes in the index attenuation amount set; The period matching degree of the candidate node path is extracted by performing debugging period matching measurement on the candidate nodes in the candidate node path; The path attenuation degree and the path period matching degree are weighted and fused, and the weighted fusion result is taken as the comprehensive index of the candidate node path; The comprehensive index of different candidate node paths is compared, and the candidate node path corresponding to the maximum value of the comprehensive index is marked as the optimal node path.

[0011] Preferably, the extracting of the path attenuation degree of the candidate node path comprises: The frequency proportion of different sensitive signal type sets in the candidate library is normalized; The result of the normalization processing is taken as the preset weight of the sensitive signal type set; According to the performance index of different debugging periods in the candidate node, the average index attenuation amount is extracted, and the reciprocal of the average index attenuation amount is taken as the index attenuation degree of the candidate node; The index attenuation degrees of different candidate nodes and the preset weight are weighted and calculated, and the weighted calculation result is taken as the path attenuation degree.

[0012] Preferably, the extracting of the period matching degree of the candidate node path comprises: The debugging periods of adjacent candidate nodes are subjected to intersection operation, the debugging periods of the intersection operation are marked as overlapping periods, and the total number of overlapping periods in the candidate node path is counted; The total number of debugging periods of all candidate nodes is counted, and the ratio calculation is performed on the total number of overlapping periods and the total number of debugging periods; The ratio calculation result is taken as the period matching degree of the candidate node path.

[0013] Preferably, the real-time control signal of the SnO2 electrode is controlled based on the digital twin, comprising: The real-time control signal of the SnO2 electrode is monitored, and the real-time signal type of the real-time control signal is identified; It is judged whether the real-time signal type belongs to the sensitive signal type set; If the real-time signal type belongs to the sensitive signal type set, a plurality of signal subsequences of the sensitive signal type set are sequentially marked as target subsequences; if the real-time signal type belongs to the sensitive signal type set, no operation is performed; A target subsequence fragment corresponding to the real-time signal type is extracted in different target subsequences; The real-time control signals are combined to obtain a real-time control signal sequence, and the real-time control signal sequence is time-aligned with the target subsequence fragment; The real-time control signals and the target subsequence fragment are respectively subjected to sliding window segmentation to respectively obtain real-time window signals and target fragment signals; The real-time window signals and the target fragment signals are subjected to similarity measurement to obtain a standard degree of the real-time window signals; The real-time window signals are controlled and optimized in the direction of the standard degree.

[0014] The technical effects and advantages of the SnO2 electrode digital twin control system for special glass electric melting application of the application are as follows: (1) By identifying a plurality of periodic control signal sequence signal type sets, determining a periodic set by statistical performance index probability distribution, extracting time-frequency features of a target signal sequence and constructing feature similarity and index attenuation sequence, determining a candidate signal type set by deflection angle analysis, and determining a sensitive signal type set based on frequency characteristics, the influence of time-frequency changes of different control signals on electrode performance attenuation rate can be captured through the correlation analysis of time-frequency features and performance attenuation, the signal regulation efficiency is improved, and the electrode signal control mode is optimized.

[0015] (2) By clustering and path optimization, the signal fragments of multiple debugging processes are integrated to comprehensively evaluate and ensure the overall optimization of the debugging periodic control signal, and the dynamic matching and control of the real-time signal and the optimal sequence are realized by combining the digital twin, the fragment integration problem is effectively solved, and the global optimization efficiency and debugging stability of the control signal sequence are improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The figure is a system structure schematic diagram of the SnO2 electrode digital twin control system for special glass electric melting application of the application.

[0017] Figure 2 The figure is a method flow schematic diagram of the SnO2 electrode digital twin control system for special glass electric melting application of the application.

[0018] Figure 3 The figure is a method flow schematic diagram of determining an optimal node path in the SnO2 electrode digital twin control system for special glass electric melting application of the application. DETAILED DESCRIPTION

[0019] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0020] The examples of the present application provide a SnO2 electrode digital twin management and control system for special glass electric melting application. By capturing the influence of time-frequency variation of different control signals on the electrode performance degradation rate, the signal control efficiency is improved, and the electrode signal management and control mode is optimized.

[0021] Please refer to Figure 1 , Figure 2 and Figure 3 In the embodiments of the present application, the SnO2 electrode digital twin management and control system for special glass electric melting application is realized by the following steps: The SnO2 electrode digital twin management and control system for special glass electric melting application specifically comprises a data acquisition module, a data fitting module, a feature extraction module, a data analysis module, a digital twin construction module, and a signal management and control optimization module. The data acquisition module acquires a cycle control signal sequence and electrode working condition data of the SnO2 electrode in multiple debugging cycles. It should be noted that the acquisition of the cycle control signal sequence is supplemented as follows. First, the control signal record of the SnO2 electrode in the debugging cycle is acquired, the same source control signal is extracted according to the control signal record, and the cycle control signal sequence under the debugging cycle is obtained according to the time stamp of the control signal of the same source. It should be noted that the control signal record is a structured or unstructured data set formed by collecting and storing the time sequence data of various control signals in the control system of the SnO2 electrode, which is obtained by calling a local or cloud data storage module. The debugging cycle refers to the process duration corresponding to the complete debugging process of the electric melting special glass. The electrode working condition data refers to a multi-dimensional performance parameter set obtained by detecting the SnO2 electrode after each debugging cycle. The electrode working condition data at least includes cross-sectional area, resistance, length, and thermal conductivity. The data fitting module fits the electrode working condition data of the SnO2 electrode in different debugging cycles to obtain a performance index sequence of the SnO2 electrode. The performance index of the electrode is a comprehensive index for measuring the service life and equipment operation and maintenance cost. The model expression for fitting the electrode working condition data is as follows: ; In the formula, is the performance index; is the initial value of the electrode working condition parameter; is the detection value of the electrode working condition parameter at the end of each debugging cycle; an index attenuation coefficient is used as an index for measuring electrode performance; i is the electrode working condition parameter number; i is a positive integer, i∈[1, n]; n is the total number of electrode working condition parameters; wherein, the index attenuation coefficient obtained according to big data testing of historical electrode performance; it should be noted that for different electrode working condition parameters in the parameter change interval, the electrode performance attenuation rate and the electrode working condition parameter may have a nonlinear relationship, therefore, the model needs to be fitted in segments according to the interval of different electrode working condition parameters to determine multiple segmented expressions; In the signal debugging process of multiple electrodes, it is difficult to capture the influence of time-frequency changes of different control signal types on the electrode performance attenuation rate, so the following modules are designed: The feature extraction module sequentially performs signal type recognition on multiple periodic control signal sequences, constructs a signal type set, and extracts a signal time-frequency feature sequence of each signal type set; In this embodiment, the signal type set is constructed, including: The signal type in the periodic control signal sequence is recognized, and the signal types in the periodic control signal sequence are spliced according to the time sequence relationship of different control signals to obtain a signal type sequence; wherein, the signal type in the periodic control signal sequence is recognized by extracting the time domain features and frequency domain features of the control signal, comparing the time domain features and frequency domain features with the template signals in the signal template library, and determining the type of the control signal. A corresponding signal type label can be generated for each signal type to identify a specific signal type; wherein the signal template is generated based on historical control signals; the signal type sequence is exemplarily illustrated, for the current signal, in the transition stage of typical debugging of glass melting, such as from the initial melting stage to the complete melting stage, the current needs to be increased to increase the temperature of the glass, and the control signal type of the current in the control logic is respectively the current increasing and the current maintaining; wherein, the current gradually increases, such as increasing by 20 amperes every 10 minutes until 500 amperes; the current is maintained, such as maintaining 500 amperes for 2 hours; the signal types of different debugging stages and transition stages are combined in time sequence to obtain a signal type sequence; Using a pattern growth algorithm based on a time window, a high-frequency appearing subsequence in multiple signal type sequences is extracted, denoted as a high-frequency subsequence; wherein, the pattern growth algorithm is one of the core methods of frequent pattern mining, aiming to find patterns (such as item sets, subsequences) with a frequency higher than a minimum support degree in a data set; in this embodiment, the pattern refers to a subsequence in the signal type sequence; the pattern growth algorithm is a prior art, which will not be described in detail here; Extracting a high-frequency appearing signal type subsequence in the signal type sequence includes: Setting a minimum support degree threshold, scanning the signal type sequence according to the minimum support degree threshold; wherein, the minimum support degree is obtained based on big data training of the frequency of the subsequence in the historical signal type sequence; All signal type subsequences satisfying the minimum support threshold are generated to determine signal type subsequences that occur frequently; the high-frequency subsequence is exemplified, for example, in the transition phase of typical debugging, the step-up current and the maintenance current occur in the transition phase of each debugging period, satisfying the minimum support threshold, the signal control types of the step-up current and the maintenance current are recorded as a1 and b1 respectively, and {a1, b1} is a high-frequency subsequence; Different signal types in the high-frequency subsequence are combined to obtain a signal type set of the signal type sequence; wherein, the signal type labels corresponding to different signal types of the high-frequency subsequence can be combined to obtain a label combination, which is recorded as the signal type set; The signal type sets of all signal type sequences are merged to obtain a combination library; The frequency of each signal type set in the combination library is obtained in sequence, and the signal type set with a frequency greater than a preset frequency threshold is retained; wherein, the frequency threshold is obtained based on the frequency big data of the historical signal type set; In this embodiment, the signal time-frequency feature sequence of each signal type set is extracted, including: The probability distribution of the performance index is counted to determine a plurality of probability distribution intervals; wherein, the probability distribution of the performance index includes setting a plurality of index intervals, counting the frequency histogram of the performance index in each index interval as the probability distribution of the performance index; the number of index intervals is obtained according to the performance index data of the historical debugging period; The debugging periods corresponding to different performance indexes of each probability distribution interval are combined to obtain a period set; The period control signal sequence corresponding to all debugging periods in the period set is marked as a target signal sequence; The signal sub-sequences corresponding to different signal type sets in the target signal sequence are identified, and time-frequency features of the extracted signal sub-sequences are taken as signal time-frequency features of the signal type sets in the target signal sequence; wherein the signal time-frequency features include statistical features, trend features, transient features and periodic features; the statistical features are obtained by acquiring signal intensity mean and standard deviation; the trend features are extracted by linear fitting of the signal sub-sequences, identifying the number of consecutive rising and falling sampling points in the signal time-frequency feature sequence, and calculating the proportion of the number of rising sampling points and falling sampling points as the trend features; the transient features are extracted by calculating the maximum value of the signal intensity difference between adjacent sampling points in the signal sub-sequences, taking it as the signal mutation threshold, counting the number of signal intensity differences greater than the signal mutation threshold in the signal sub-sequences to obtain the mutation number, and calculating the mutation frequency of the mutation number as the transient features; the periodic features are obtained by acquiring the signal length and sub-sequence proportion range of the signal sub-sequences in the target signal sequence; the sub-sequence proportion range is supplemented, for example, if the signal sub-sequence is in the 20%-45% segment of the whole target signal sequence, then the sub-sequence proportion range is 20%-45%; The plurality of signal time-frequency features of the signal type sets are combined in the time sequence order of the debugging period to obtain a signal time-frequency feature sequence; The data analysis module extracts the sensitive signal type set by analyzing the correlation between the signal time-frequency feature sequence and the performance index sequence in the sequence change trend; In this embodiment, the sensitive signal type set is extracted, including: The sequence change characteristics of the signal time-frequency feature sequence and the performance index sequence are analyzed respectively to construct a feature similarity sequence and an index attenuation sequence; the feature similarity sequence corresponds to the signal time-frequency feature sequence; the index attenuation sequence corresponds to the performance index sequence; The feature similarity sequence and the index attenuation sequence are constructed, including: The similarity of adjacent two signal time-frequency features is measured by traversing the signal time-frequency feature sequence to obtain the feature similarity; wherein the feature similarity is obtained by constructing the signal time-frequency feature vector corresponding to the signal time-frequency feature, and calculating the vector space distance of the two signal time-frequency feature vectors; The feature similarities of a plurality of adjacent signal time-frequency features are combined to obtain a feature similarity sequence; for example, there is a signal time-frequency feature sequence {x1, x2, x3, x4}, wherein the feature similarities of adjacent signal time-frequency features x1 and x2, x2 and x3, x3 and x4 are S1, S2 and S3 respectively, and the corresponding feature similarity sequence is {S1, S2, S3}; The difference values of adjacent two performance indexes are calculated by traversing the performance index sequence to obtain the index attenuation; The index attenuations of a plurality of adjacent performance indexes are combined to obtain an index attenuation sequence; extract a deflection angle of each sequence point in the index attenuation sequence, denoted as a first deflection angle; extract a deflection angle of each sequence point in the feature similarity sequence, denoted as a second deflection angle; the second deflection angle corresponds to the first deflection angle one by one; wherein the deflection angle of the sequence point is an index describing the direction change degree between adjacent points in the sequence, that is, the sequence segments formed by a certain sequence point and the previous and next sequence points, respectively, the turning amplitude of the sequence from one point to the next point is quantified by calculating the included angle between the two sequence segments; calculate the absolute difference value between each second deflection angle and the corresponding first deflection angle, denoted as the relative deflection angle difference of the second deflection angle; compare the relative deflection angle differences of different second deflection angles, and mark the second deflection angle corresponding to the minimum relative deflection angle difference as a target deflection angle; compare the total number of target deflection angles in each feature similarity sequence, and retain the feature similarity sequence whose total number of target deflection angles is greater than a preset number threshold; wherein the number threshold is obtained based on historical target deflection angle total number big data training; integrate the retained feature similarity sequences, and mark the signal type set corresponding to the feature similarity sequence as a candidate signal type set of the cycle set; combine the candidate signal type sets of different cycle sets to obtain a candidate library, and count the frequency proportion of each candidate signal type set; compare the frequency proportion of the candidate signal type set with a preset frequency threshold, and mark the candidate signal type set whose frequency proportion is greater than the frequency threshold as a sensitive signal type set; wherein the frequency threshold is obtained according to the frequency proportion big data training of a plurality of candidate signal type sets; By identifying the signal type set of the plurality of cycle control signal sequences, the performance index probability distribution is determined to determine the cycle set, the time-frequency features of the target signal sequence are extracted and the feature similarity and index attenuation sequence are constructed, the candidate signal type set is determined through deflection angle analysis, and the sensitive signal type set is determined based on the frequency characteristics. Through the correlation analysis of time-frequency characteristics and performance attenuation, the influence of time-frequency variation of different control signals on the electrode performance attenuation rate can be captured, which provides a quantitative basis for debugging and improves the signal debugging efficiency and performance optimization pertinence; For the adjustment of different types of control signals in each debugging process, it is difficult to integrate all the control signal sequence segments after the adjustment of different signal types in all the debugging processes, thereby solving the problem of optimizing the control signals of the whole debugging cycle; The digital twin construction module extracts the aggregation characteristics of the sensitive signal type set on the signal time-frequency characteristics, extracts the optimal signal sequence combination according to the aggregation characteristics of different sensitive signal type sets, and constructs a digital twin according to the optimal signal sequence combination; In the embodiment, the optimal signal sequence combination is extracted, including: The multiple signal time-frequency features of the same sensitive signal type set are combined to obtain a sensitive feature set; The feature similarity of any two sensitive features in the sensitive feature set is extracted, and the corresponding debugging period is clustered according to the feature similarity to obtain multiple clustering centers; wherein, according to the feature similarity, the feature similarity matrix of the sensitive feature set is constructed, and the multiple sensitive feature clusters are identified by using a hierarchical clustering algorithm, and the debugging periods corresponding to the different sensitive features in each sensitive feature cluster are combined to obtain the clustering centers; it should be noted that the construction of the feature similarity matrix of the sensitive feature set and the identification of the multiple sensitive feature clusters by using the hierarchical clustering algorithm are both prior art means, which will not be described in detail here; Each clustering center is designated as a candidate node of the sensitive signal type set; The control timing of the sensitive signal type set is extracted, and any candidate node of each sensitive signal type set is selected for connection in turn according to the control timing sequence of different sensitive signal type sets; wherein, the control timing sequence of the sensitive signal type set is supplemented, for example, in multiple process stages of the debugging process, the sensitive type sets A, B and C are respectively in the first, second and third process stages, and the control timing sequence of the three sensitive signal type sets is A, B and C in turn; the selection of any candidate node of each sensitive signal type set for connection is exemplarily described, the sensitive signal type sets A, B and C have corresponding candidate nodes A1 and A2; B1, B2 and B3; C1 and C2; the candidate nodes A1, B1 and C1 can be selected for connection; correspondingly, the candidate nodes A2, B3 and C2 can also be selected for connection; According to the multiple possibilities of connecting the candidate nodes of different sensitive signal type sets, the candidate node paths are exhausted to obtain a plurality of candidate node paths; for example, the sensitive signal type set A has two ways in the connection of the candidate nodes, the sensitive signal type set B has three ways, and the sensitive signal type set C has two ways, and 12 candidate node paths are obtained by exhaustion; The different candidate nodes in the candidate node path are comprehensively evaluated by combining the path attenuation degree and the period matching degree to determine the optimal node path; In the embodiment, the optimal node path is determined, including: The path attenuation degree of the candidate node path is extracted by analyzing the stability of different candidate nodes in the index attenuation set; The path attenuation degree of the candidate node path is extracted, including: The frequency proportion of different sensitive signal type sets in the candidate library is normalized; The normalized result is used as the preset weight of the sensitive signal type set; According to the performance indicators of different debugging periods in the candidate node, the average indicator attenuation amount is extracted, and the reciprocal of the average indicator attenuation amount is taken as the indicator attenuation degree of the candidate node; wherein, the greater the indicator attenuation degree is, the smaller the average indicator attenuation amount is, and the cycle signal control sequence corresponding to the debugging period in the candidate node is more optimal in reducing the performance indicator degradation of the electrode; the indicator attenuation degree of different candidate nodes is weighted with the preset weight, and the weighted calculation result is taken as the path attenuation degree; wherein, the significance of the path attenuation degree lies in quantifying the overall situation of the indicator attenuation degree of different candidate nodes, and the greater the path attenuation degree is, the more optimal the combination of the sensitive feature set corresponding to the different candidate nodes is in the feature distribution in the entire debugging period, which facilitates finding the best signal sub-sequence of each sensitive signal type in different debugging periods and slowing down the performance degradation rate of the electrode; The candidate nodes in the candidate node path are subjected to debugging period matching measurement, and the cycle matching degree of the candidate node path is extracted; The cycle matching degree of the candidate node path is extracted, including: The debugging periods of adjacent candidate nodes are subjected to intersection operation, the debugging periods of the intersection operation are marked as overlapping periods, and the total number of overlapping periods in the candidate node path is counted; The total number of debugging periods of all candidate nodes is counted, and the ratio calculation is performed between the total number of overlapping periods and the total number of debugging periods; The ratio calculation result is taken as the cycle matching degree of the candidate node path; wherein, the significance of the cycle matching degree lies in: quantifying the common influence degree of the best signal sub-sequence combination on the electrode performance degradation for the best signal sub-sequence of different sensitive signal type sets in each debugging process; if the cycle matching degree is low, it indicates that when the best signal sub-sequence of different sensitive signal type sets is determined, the debugging periods corresponding to different best sub-sequences have large differences in cycle span, there are obvious individual differences, there are accidental factors, and it is not conducive to reflecting the common influence degree of different signal sub-sequences on the electrode performance degradation; The path attenuation degree and the path cycle matching degree are weighted and fused, and the weighted fusion result is taken as the comprehensive indicator of the candidate node path; wherein, the weights of the path attenuation degree and the path cycle matching degree are obtained based on the debugging big data training of the historical debugging process; The comprehensive indicators of different candidate node paths are compared, and the candidate node path corresponding to the maximum value of the comprehensive indicator is marked as the optimal node path; A plurality of debugging periods in each candidate node in the optimal node path are selected and recorded as selected periods corresponding to the sensitive signal type; The signal sub-sequences of the sensitive signal type set in different selected periods are combined to obtain the optimal signal sequence combination of the sensitive signal type set; wherein, the optimal signal sequence combination is periodically updated; the update period can be an integer multiple of the adjustment period, such as 3 times the adjustment period; The signal management and optimization module manages and controls the real-time control signal of the SnO2 electrode based on the digital twin; In this embodiment, the real-time control signal of the SnO2 electrode is managed and controlled based on the digital twin, including: Monitoring the real-time control signal of the SnO2 electrode, and identifying the real-time signal type of the real-time control signal; Judging whether the real-time signal type belongs to the sensitive signal type set; If the real-time signal type belongs to the sensitive signal type set, a plurality of signal sub-sequences of the sensitive signal type set are sequentially marked as target sub-sequences; if the real-time signal type belongs to the sensitive signal type set, no operation is performed; Extracting target sub-sequence fragments corresponding to the real-time signal type in different target sub-sequences, combining the real-time control signal, obtaining a real-time control signal sequence, and performing time sequence alignment between the real-time control signal sequence and the target sub-sequence fragments; Respectively performing sliding window segmentation on the real-time control signal and the target sub-sequence fragments to obtain real-time window signals and target fragment signals; Measuring the similarity of the real-time window signals and the target fragment signals to obtain the standard degree of the real-time window signals; wherein the similarity of the real-time window signals and the target fragment signals is measured by extracting the time-frequency feature vectors of each real-time window signal and target fragment signal, and calculating the vector space distance between the time-frequency feature vectors; Controlling and optimizing the real-time window signals in the direction of the standard degree; wherein the standard degree in the direction of the standard degree can improve the standard degree of the real-time calculation window, a standard degree threshold is set in advance, the real-time control signal is adjusted, and the standard degree of the window is made to satisfy the standard degree threshold; By combining the signal time-frequency features of the same sensitive signal type set to form a sensitive feature set, the debugging period is clustered based on the feature similarity to obtain candidate nodes, the candidate nodes are connected in the control logic time sequence and the paths are exhausted, the optimal node path is determined based on the performance index distribution and the period matching degree evaluation, and then the optimal signal sequence combination of the sensitive signal type set is obtained; finally, the real-time signal type is monitored based on the digital twin, the signals belonging to the sensitive type are time sequence aligned with the optimal signal sequence fragments, and the management and optimization are realized through the sliding window similarity measurement; this method solves the problem of difficult integration of different signal adjustment sequence fragments in debugging, integrates the signal fragments of multiple debugging processes through clustering and path optimization, comprehensively evaluates to ensure the overall optimization of the debugging period control signal, realizes the dynamic matching and management of the real-time signal and the optimal sequence based on the digital twin, effectively solves the fragment integration problem, and improves the global optimization efficiency of the control signal sequence and the debugging stability.

[0022] The above merely describes preferred embodiments of the present application, and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, the technical solutions recorded in the foregoing embodiments can be modified or some technical features can be replaced by equivalents by those skilled in the art, without departing from the spirit and principle of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0023] It should be noted that, in this document, the terms "comprises", "comprising", or any other variation thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without more limitations, an element defined by the statement "comprising a" does not exclude the existence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0024] The formulas in the present specification are calculated by dimensionless values, the formulas are obtained by collecting a large amount of data to simulate the most real situation, and the preset parameters and threshold values in the formulas are set by those skilled in the art according to the actual situation.

[0025] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A digital twin control system for SnO2 electrodes for special glass electrofusion applications, characterized in that, include: The data acquisition module acquires the periodic control signal sequence and electrode operating condition data for multiple debugging cycles of the SnO2 electrode; The data fitting module fits the electrode operating data of the SnO2 electrode during different debugging cycles to obtain the performance index sequence of the SnO2 electrode. The feature extraction module sequentially identifies the signal type of multiple periodic control signal sequences, constructs a signal type set, and extracts the time-frequency feature sequence of each signal type set. The data analysis module extracts a set of sensitive signal types by analyzing the correlation between the time-frequency characteristic sequence of the signal and the performance index sequence in terms of sequence change trends. The digital twin construction module extracts the clustering characteristics of sensitive signal type sets in the time-frequency features of signals, extracts the optimal signal sequence combination based on the clustering characteristics of different sensitive signal type sets, and constructs a digital twin based on the optimal signal sequence combination. The signal control and optimization module manages the real-time control signals of the SnO2 electrode based on a digital twin.

2. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 1, characterized in that, The constructed signal type set includes: Identify the signal types in the periodic control signal sequence, and concatenate the signal types according to the temporal relationship of different control signals in the periodic control signal sequence to obtain the signal type sequence; Using a time window-based pattern growth algorithm, high-frequency subsequences are extracted from multiple signal type sequences and denoted as high-frequency subsequences. By combining different signal types in high-frequency subsequences, a signal type set of signal type sequence is obtained; Merge the signal type sets of all signal type sequences to obtain the combined library; The frequency of each signal type set in the combination library is obtained sequentially, and the signal type sets with frequencies greater than the preset frequency threshold are retained.

3. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 1, characterized in that, The extraction of the signal time-frequency feature sequence for each signal type set includes: The probability distribution of statistical performance indicators, determining multiple probability distribution intervals; By combining the debugging periods corresponding to different performance indicators for each probability distribution interval, a set of periods is obtained; Mark the cycle control signal sequence corresponding to each debugging cycle in the cycle set as the target signal sequence; Identify signal subsequences corresponding to different signal type sets in the target signal sequence, and extract the time-frequency features of the signal subsequences as the signal time-frequency features of the signal type sets in the target signal sequence; By combining multiple signal time-frequency features of the signal type set in the timing sequence of the debugging cycle, a signal time-frequency feature sequence is obtained.

4. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 3, characterized in that, The set of sensitive signal types extracted includes: The sequence variation characteristics of the signal time-frequency feature sequence and the performance index sequence are analyzed separately, and feature similarity sequence and index attenuation sequence are constructed. The feature similarity sequence corresponds to the signal time-frequency feature sequence, and the index attenuation sequence corresponds to the performance index sequence. Extract the deflection angle of the index decay sequence at each sequence point, and denote it as the first deflection angle; The deflection angle of the feature similarity sequence at each sequence point is extracted and denoted as the second deflection angle; the second deflection angle corresponds one-to-one with the first deflection angle. Calculate the absolute difference between each second deflection angle and the corresponding first deflection angle, and record it as the relative deflection angle difference of the second deflection angle; By comparing the relative deflection angle differences of different second deflection angles, the second deflection angle corresponding to the minimum relative deflection angle difference is marked as the target deflection angle; Compare the total number of target deflection angles in each feature similarity sequence, and retain the feature similarity sequences whose total number of target deflection angles is greater than a preset threshold; Integrate the retained feature similarity sequences and mark the signal type set corresponding to the feature similarity sequence as the candidate signal type set of the periodic set; By combining candidate signal type sets from different period sets, a candidate library is obtained, and the frequency proportion of each candidate signal type set is statistically analyzed. The frequency proportion of the candidate signal type set is compared with the preset frequency threshold, and the candidate signal type set with a frequency proportion greater than the frequency threshold is marked as the sensitive signal type set.

5. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 4, characterized in that, The construction of the feature similarity sequence and the index decay sequence includes: Traverse the time-frequency feature sequence of the signal, measure the similarity of the time-frequency features of two adjacent signals, and obtain the feature similarity. By combining the feature similarity of time-frequency features of multiple adjacent signals, a feature similarity sequence is obtained; Traverse the performance index sequence, calculate the difference between two adjacent performance indices, and obtain the index decay amount; By combining the attenuation of multiple adjacent performance indicators, an indicator attenuation sequence is obtained.

6. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 3, characterized in that, The extraction of the optimal signal sequence combination includes: By combining the time-frequency features of multiple signals from the same sensitive signal type set, a sensitive feature set is obtained; Extract the feature similarity between any two sensitive features from the sensitive feature set, and cluster the corresponding debugging cycles according to the feature similarity to obtain multiple cluster centers; Each cluster center is designated as a candidate node for the sensitive signal type set; Extract the control timing of the sensitive signal type set, and connect any candidate node of each sensitive signal type set according to the control timing order of different sensitive signal type sets. Based on the multiple possibilities of connecting candidate nodes corresponding to different sensitive signal type sets, the candidate node paths are exhaustively enumerated to obtain several candidate node paths; For different candidate nodes in the candidate node path, the joint path attenuation goodness and period matching degree are comprehensively evaluated to determine the optimal node path; Select multiple debugging cycles within each candidate node in the optimal node path, and record them as the selected cycles corresponding to the sensitive signal types; By combining the signal subsequences of the sensitive signal type set in different selected periods, the optimal signal sequence combination of the sensitive signal type set is obtained.

7. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 6, characterized in that, Determining the optimal node path includes: By analyzing the stability of the concentration of index decay among different candidate nodes, the path decay goodness of the candidate node path is extracted. Perform a periodic matching metric on the candidate nodes in the candidate node path and extract the periodic matching degree of the candidate node path; The path decay excellence and path period matching degree are weighted and fused, and the weighted fusion result is used as a comprehensive index of the candidate node path; By comparing the comprehensive indicators of different candidate node paths, the candidate node path corresponding to the maximum or minimum comprehensive indicator value is marked as the optimal node path.

8. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 7, characterized in that, The path decay quality of the extracted candidate node paths includes: The frequency proportions of different sensitive signal type sets in the candidate library are normalized. The result of normalization is used as the preset weight for the sensitive signal type set; Based on the performance indicators of candidate nodes in different debugging cycles, the average indicator decay is extracted, and the reciprocal of the average indicator decay is used as the indicator decay goodness of the candidate node. The index decay goodness of different candidate nodes is weighted and calculated with preset weights, and the weighted calculation result is used as the path decay goodness.

9. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 7, characterized in that, The periodic matching degree of the extracted candidate node paths includes: Perform an intersection operation on the debugging cycles of adjacent candidate nodes, mark the debugging cycles of the intersection operation as overlapping cycles, and count the total number of overlapping cycles in the candidate node paths. Calculate the total number of debugging cycles for all candidate nodes, and then calculate the ratio of the total number of overlapping cycles to the total number of debugging cycles. The ratio calculation result is used as the periodic matching degree of the candidate node path.

10. The SnO2 electrode digital twin control system for special glass electrofusion applications according to claim 3, characterized in that, The real-time control signal management of the SnO2 electrode based on the digital twin includes: Monitor the real-time control signal of the SnO2 electrode and identify the real-time signal type of the real-time control signal; Determine whether the real-time signal type belongs to the sensitive signal type set; If the real-time signal type belongs to the sensitive signal type set, then the multiple signal subsequences in the sensitive signal type set will be marked as the target subsequence in sequence; otherwise, no operation will be performed. Extract target subsequence segments corresponding to real-time signal types from different target subsequences; Combine the real-time control signals to obtain a real-time control signal sequence, and then align the real-time control signal sequence with the target subsequence segment in time. Sliding window segmentation is performed on the real-time control signal and the target subsequence segment to obtain the real-time window signal and the target segment signal, respectively. The similarity between the real-time window signal and the target segment signal is measured to obtain the standard degree of the real-time window signal; With the goal of improving the standard, we manage and optimize the real-time window signal.

Images