Swing angle tension bridge sensing signal calibration method of aluminum ingot continuous casting machine

Abstract

The invention discloses a swing angle tension bridge sensing signal calibration method for an aluminum ingot continuous casting machine, and relates to the technical field of sensing signal correction, and the method comprises the steps: taking the rotation rhythm of a traction wheel of a casting machine as a reference, screening a synchronous oscillation component from an original swing angle sensing signal, and obtaining a periodic characteristic component; performing dynamic correlation analysis on the phase track and the amplitude envelope of the periodic characteristic component to obtain a dynamic correlation relationship between the phase track and the amplitude envelope; performing active time disturbance on the periodic characteristic component based on the dynamic association relationship to obtain covariant response intensity of the phase track and the amplitude envelope under disturbance; on the basis of covariant response intensity, weighted reconstruction is carried out on a verified collaborative change mode between a phase track and an amplitude envelope of the periodic characteristic component; according to the invention, the problem of insufficient signal separation accuracy and calibration precision under working condition fluctuation is solved, and the signal precision required by product quality control is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of sensor signal correction technology, and in particular to a method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine. Background Technology

[0002] The pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine is a core data reflecting the equipment's operating status and the stability of the casting process. Its accuracy directly affects product quality control. Traditional signal calibration methods mostly rely on fixed rule processing methods. However, during the aluminum ingot casting process, changes in working conditions such as fine-tuning of casting speed and fluctuations in aluminum liquid supply flow will directly lead to dynamic adjustment of the rotation rhythm of the traction wheel, which in turn causes the periodic characteristics and phase-amplitude coordination mode of the pendulum angle sensor signal to evolve in real time.

[0003] Traditional fixed-rule processing methods cannot keep up with this dynamic evolution process and adjust the separation and calibration logic accordingly. It is difficult to accurately separate the effective signal components that are synchronized with the equipment's operating rhythm from the original signal with mixed interference, and it is also impossible to perform targeted calibration on dynamically changing signal deviations. Ultimately, the accuracy of signal separation and calibration precision cannot meet the actual production needs. Summary of the Invention

[0004] This invention provides a method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for calibrating the pendulum tension bridge sensor signal of an aluminum ingot continuous casting machine, comprising:

[0006] S1. Based on the operating status of the aluminum ingot continuous casting machine, the output signal of the swing angle tension bridge sensor is collected to obtain the original swing angle sensing signal;

[0007] S2. Using the rotational rhythm of the casting machine's traction wheel as a reference, the synchronous oscillation component is selected from the original swing angle sensing signal to obtain the periodic characteristic component;

[0008] S3. Perform dynamic correlation analysis on the phase trajectory and amplitude envelope of the periodic characteristic components to obtain the dynamic correlation relationship between the phase trajectory and amplitude envelope;

[0009] S4. Based on the dynamic correlation, the periodic characteristic components are actively perturbed in time to obtain the covariant response intensity of the phase trajectory and amplitude envelope under the perturbation.

[0010] S5. Based on the covariant response intensity, the verified covariance mode between the phase trajectory and amplitude envelope of the periodic characteristic components is reconstructed by weighting to obtain the dynamic correction mapping.

[0011] S6. The original swing angle sensing signal is processed by dynamic correction mapping to obtain the calibrated swing angle signal.

[0012] Preferably, the acquisition of the output signal of the swing angle tension bridge sensor based on the operating status of the aluminum ingot continuous casting machine to obtain the original swing angle sensing signal includes:

[0013] Based on the pulse signal of the rotary encoder of the casting machine traction wheel, the operating cycle and phase of the casting machine are analyzed in real time to obtain the state timing identifier synchronized with the traction action;

[0014] Based on the state timing identifier, the output signal of the pendulum tension bridge sensor is synchronously triggered and acquired to obtain a discrete signal sequence;

[0015] The discrete signal sequence is subjected to non-steady-state segment identification and steady-state segment extraction to obtain the original swing angle sensing signal.

[0016] Preferably, the step of filtering synchronous oscillation components from the original swing angle sensing signal, using the rotational rhythm of the casting machine traction wheel as a reference, to obtain periodic characteristic components includes:

[0017] The real-time rotation frequency of the traction wheel is obtained as the fundamental frequency, and the target analysis frequency band is determined based on the fundamental frequency;

[0018] Within the target analysis frequency band, the original swing angle sensing signal is transformed by time and frequency to obtain the time and frequency energy distribution of the signal;

[0019] Determine the coherence between the time-frequency energy distribution and the fundamental frequency and its harmonics, and generate a synchronization energy spectrum characterizing the strength of synchronization in each time period;

[0020] Based on the synchronization energy spectrum, signal segments corresponding to time periods when the synchronization intensity exceeds a preset threshold are extracted from the original swing angle sensing signal and combined to form a synchronization oscillation component;

[0021] By connecting the synchronous oscillation components according to the rotation period of the traction wheel, the periodic characteristic components are obtained.

[0022] Preferably, the dynamic correlation analysis of the phase trajectory and amplitude envelope of the periodic characteristic components to obtain the dynamic correlation relationship between the phase trajectory and amplitude envelope includes:

[0023] Based on the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope, the overlapping interval of the two in the time window is identified to obtain the coordinated oscillation interval.

[0024] Based on the coordinated oscillation interval, the order of occurrence of the extreme values ​​of phase trajectory angular velocity and amplitude envelope rate of change is tracked to obtain the dominant order relationship;

[0025] Based on the relationship between the cooperative oscillation interval and the dominant order, a cooperative mode is constructed for the phase trajectory and amplitude envelope to obtain a primary dynamic correlation.

[0026] Preferably, obtaining the dynamic correlation between the phase trajectory and the amplitude envelope further includes:

[0027] The stability of the duration of the coordinated oscillation interval in different traction wheel rotation cycles is evaluated to obtain the steady-state coordinated region and the transient coordinated region.

[0028] Repeatability analysis of the dominant order relationship was performed in the steady-state and transient cooperative regions to obtain the weighting factors;

[0029] Based on the weighting factor, the primary dynamic relationship is reconstructed by weighting to obtain the weighted dynamic relationship, which is then used as the dynamic relationship.

[0030] Preferably, the step of actively perturbing the periodic feature components based on dynamic correlation to obtain the covariant response intensity of the phase trajectory and amplitude envelope under perturbation includes:

[0031] Based on the dominant order relationship in the dynamic correlation, key time points in the periodic feature components are identified where the phase trajectory leads the amplitude envelope or the amplitude envelope leads the phase trajectory.

[0032] Apply a time offset that is inverse to the dominant order relationship to key time points to obtain the perturbation component with time reversal;

[0033] By observing the self-recovery process of the order of phase trajectory and amplitude envelope in the perturbation component, we can obtain the trend and speed measure of its convergence to the dominant order relationship.

[0034] Based on convergence trend and speed, the stability of dynamic correlation under time-reversal pressure is quantified to obtain the covariant response intensity.

[0035] Preferably, applying a temporal offset inverse to the dominant order relationship to key time points to obtain the perturbed component with temporal reversal includes:

[0036] Based on the weight factors associated with key time points in the dynamic correlation, key time points are divided into a high-weight time point subset and a low-weight time point subset.

[0037] A first time-off mode that is completely reversed from the dominant order relationship is applied to a subset of high-weight time-series points, and a second time-off mode that is partially reversed from the dominant order relationship is applied to a subset of low-weight time-series points, to obtain the offset scheme of partitioned modulation.

[0038] The periodic characteristic components are time-shifted according to the offset scheme of partitioned modulation to obtain the perturbed components with time reversal.

[0039] Preferably, the step of weighted reconstruction of the verified co-variation patterns between the phase trajectories and amplitude envelopes of periodic characteristic components based on co-variant response intensity to obtain a dynamic correction mapping includes:

[0040] Based on the comparison results of time-series verification intensity and partitioned response in covariant response intensity, a target mode feature set that remains stable under perturbation is extracted from the covariant variation mode.

[0041] Based on the covariant response intensity, the stability contribution of different features in the target mode feature set is weighted and assigned to obtain the weighted mode rule;

[0042] Based on the weighted mode rules, the correction relationship between the phase trajectory and the amplitude envelope is structured and encoded to obtain the dynamic correction mapping.

[0043] Preferably, the weighted pattern rule, which assigns weights to the stability contributions of different features in the target pattern feature set based on covariant response intensity, includes:

[0044] Based on the steady-state or transient cooperative region attributes corresponding to the target pattern feature set in the partition response comparison results, the features are divided into steady-state feature subsets and transient feature subsets.

[0045] For the steady-state feature subset, basic reconstruction weights are assigned based on its temporal verification strength; for the transient feature subset, adaptive decay weights are assigned based on its response divergence characteristics, resulting in a differentiated weight allocation scheme.

[0046] Based on the differentiated weight allocation scheme, the feature set of the target pattern is organized in a regularized manner to form weighted pattern rules.

[0047] Preferably, the step of processing the original swing angle sensing signal through dynamic correction mapping to obtain the calibrated swing angle signal includes:

[0048] Based on the correction relationship encoded in the dynamic correction mapping, feature decoupling and relationship correction are performed on the signal segments in the original swing angle sensing signal that match the mapping rules to obtain the preliminary correction signal segments;

[0049] For feature points in the initial correction signal segment that have been corrected according to the mapping rules, a smooth transition is performed between them and adjacent uncorrected feature points to obtain a continuous correction signal with a natural transition.

[0050] The amplitude of the continuous correction signal is normalized and the timing is aligned so that it is resynchronized with the rotation rhythm of the casting machine traction wheel to obtain the calibrated swing angle signal.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. Through a complete process of signal acquisition, periodic feature screening, dynamic correlation analysis, active disturbance verification, dynamic correction mapping construction, and signal calibration, taking the rotation rhythm of the casting machine's traction wheel as the core reference, the system dynamically captures the periodic characteristics and phase-amplitude co-evolution of the swing angle signal following changes in operating conditions. This breaks the limitations of traditional fixed rules, accurately separating the effective components in the original signal that are synchronized with the equipment's operating rhythm, and specifically correcting dynamically changing signal deviations. This effectively solves the problem of insufficient signal separation accuracy and calibration precision under fluctuating operating conditions, ensuring the signal accuracy required for product quality control.

[0053] 2. By leveraging the synergistic effect of active time perturbation, dynamic correlation analysis, and weighted pattern rule construction, the intensity of covariant response is quantified to screen stable covariant features. Weights are allocated according to the stability contribution to form a structured correction logic, enabling the dynamic correction mapping to have strong anti-interference capabilities and high adaptability. This not only avoids the misleading effect of transient interference on calibration results but also ensures that the calibration process always conforms to the actual operating rules of the equipment, significantly improving the reliability and adaptability of the pendulum tension bridge sensor signal calibration. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating a method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine according to an embodiment of the present invention.

[0055] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0056] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0057] This application provides a method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine. The execution subject of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0058] Example 1, referring to Figure 1 The diagram shown is a flowchart illustrating a method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine according to an embodiment of the present invention. In this embodiment, the method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine includes:

[0059] S1. Based on the operating status of the continuous aluminum ingot casting machine, the output signal of the swing angle tension bridge sensor is collected to obtain the original swing angle sensing signal, including:

[0060] Based on the pulse signal of the rotary encoder of the casting machine traction wheel, the operating cycle and phase of the casting machine are analyzed in real time to obtain the state timing identifier synchronized with the traction action;

[0061] Based on the state timing identifier, the output signal of the pendulum tension bridge sensor is synchronously triggered and acquired to obtain a discrete signal sequence;

[0062] The discrete signal sequence is subjected to non-steady-state segment identification and steady-state segment extraction to obtain the original swing angle sensing signal.

[0063] Specifically, the rotary encoder is coaxially mounted on the shaft of the casting machine's traction wheel to ensure that the encoder rotates synchronously with the traction wheel. The encoder outputs a pulse signal corresponding to the rotation angle in real time, and the pulse signal is transmitted to the signal analysis unit through the signal transmission line.

[0064] Furthermore, during analysis, the number of pulses per unit time is counted moment by moment, and the rotation cycle of the traction wheel is determined by combining the encoder resolution. At the same time, the real-time rotation angle of the traction wheel is calculated by accumulating the number of pulses, thereby resolving the running phase. Then, each rotation cycle is divided into multiple time segments according to the phase, and a unique timing code is assigned to each time segment. The rotation cycle information, phase information, and timing code are integrated to form a state timing identifier that is synchronized with the traction action.

[0065] Furthermore, the status timing identifier is transmitted to the data acquisition trigger module, and a specific phase moment in the status timing identifier is set as the trigger signal. When the trigger module detects the specific timing identifier, it immediately sends an acquisition command to the data acquisition hardware.

[0066] Specifically, the output of the pendulum tension bridge sensor is connected to the analog signal input of the data acquisition hardware via a shielded cable. Both ends of the shielding layer are grounded to shield against electromagnetic interference. After receiving a trigger command, the data acquisition hardware acquires the voltage signal output by the sensor at a preset sampling frequency. Each trigger acquisition obtains one data point, and continuous trigger acquisition forms a discrete signal sequence consisting of multiple data points arranged in chronological order.

[0067] Specifically, the preset signal fluctuation threshold setting method can be as follows: start the casting machine to run under standard stable conditions, collect multiple sets of normal signals, count the fluctuation range of the signal amplitude, and take the maximum value of the fluctuation range as the threshold.

[0068] Specifically, the discrete signal sequence is traversed segment by segment, and the amplitude difference between adjacent data points within each signal segment is calculated. When the difference continuously exceeds a preset threshold, the signal segment is identified as a non-steady-state segment. All identified non-steady-state segments are removed, and the remaining signal segments are spliced ​​together in their original time order to obtain the original swing angle sensing signal containing only the steady-state signal.

[0069] In summary, this embodiment uses the pulse signal of the traction wheel rotary encoder as a reference to accurately analyze the operating cycle and phase and generate a synchronization status timing identifier. This ensures that signal acquisition is strictly synchronized with traction actions, avoids signal distortion caused by timing deviations in acquisition, and ensures that the original signal can truly reflect the correlation between the pendulum tension bridge and the core operating rhythm of the equipment, thus improving signal specificity. Then, the shielded cable grounding design blocks electromagnetic interference in the industrial field, and the non-steady-state segments in the discrete signal sequence are removed by a preset threshold. This effectively filters invalid signals caused by operating condition fluctuations and electromagnetic interference, retains stable and reliable steady-state signals, and improves the purity of the original signal.

[0070] Overall, the entire acquisition process, from synchronous triggering to steady-state extraction, is rigorously controlled at each stage. The output raw pendulum angle sensing signal combines synchronicity and reliability, laying a high-quality data foundation for subsequent periodic feature screening, phase-amplitude dynamic correlation analysis, and other steps, indirectly ensuring the accuracy and efficiency of the overall calibration process.

[0071] S2. Using the rotational rhythm of the casting machine's traction wheel as a reference, the synchronous oscillation component is filtered from the original swing angle sensing signal to obtain the periodic characteristic components, including:

[0072] The real-time rotation frequency of the traction wheel is obtained as the fundamental frequency, and the target analysis frequency band is determined based on the fundamental frequency;

[0073] Within the target analysis frequency band, the original swing angle sensing signal is transformed by time and frequency to obtain the time and frequency energy distribution of the signal;

[0074] Determine the coherence between the time-frequency energy distribution and the fundamental frequency and its harmonics, and generate a synchronization energy spectrum characterizing the strength of synchronization in each time period;

[0075] Based on the synchronization energy spectrum, signal segments corresponding to time periods when the synchronization intensity exceeds a preset threshold are extracted from the original swing angle sensing signal and combined to form a synchronization oscillation component;

[0076] By connecting the synchronous oscillation components according to the rotation period of the traction wheel, the periodic characteristic components are obtained.

[0077] The formula for calculating the synchronicity strength is as follows:

[0078]

[0079] In the formula, It is a time variable; For frequency variables; The real-time rotational fundamental frequency of the traction wheel is obtained through a speed sensor; The highest harmonic order under consideration is preset based on system characteristics; This refers to the frequency tolerance window width around each harmonic; The time-frequency energy distribution is obtained by performing a short-time Fourier transform on the original swing angle sensing signal; For a moment The synchronicity intensity index, and the sequence formed by it, constitutes the synchronicity energy spectrum.

[0080] Specifically, a speed sensor is installed at the end of the traction wheel shaft to collect the pulse signal of the shaft rotation in real time. The real-time rotation frequency of the traction wheel, i.e., the base frequency, is obtained by counting the number of pulses per unit time.

[0081] Specifically, the target analysis frequency band is determined based on the fundamental frequency, covering the frequency range from the fundamental frequency to the highest harmonic, to ensure that subsequent analysis can include all signal components related to the rotation of the traction wheel.

[0082] Furthermore, a short-time Fourier transform is used to perform a time-frequency transformation on the original swing angle sensing signal. The Hanning window can be selected as the time window to divide the original swing angle sensing signal into continuous and overlapping time segments. Each time segment is multiplied by the Hanning window and then subjected to a Fourier transform to obtain the frequency component and energy corresponding to each time segment. By sliding the time window to traverse the entire original swing angle sensing signal, the time-frequency energy distribution describing the signal energy at different times and frequencies is finally obtained.

[0083] Specifically, in the formula for calculating synchronicity intensity, where For continuous time variables, It is a continuous frequency variable; It is the real-time rotational fundamental frequency obtained by the speed sensor acquiring the pulse signal of the traction wheel shaft in real time and counting the number of pulses per unit time.

[0084] Specifically, in the formula for synchronicity strength, To set the preset highest harmonic order, the casting machine can be started and run stably. The K value can be gradually increased while observing changes in the synchronization energy spectrum. For example, when... Increase to 5, then continue to increase. The value of the spectral peak no longer changes, thus confirming the value. The value is 5, which can fully cover the rotationally correlated harmonic components.

[0085] Specifically, For the frequency tolerance window width, for example, it is fixed in the preset. Adjust to 5. The goal is to fully capture the peak harmonic energy without introducing any irrelevant components. This is the time-frequency energy distribution obtained from the aforementioned short-time Fourier transform.

[0086] Overall, by superimposing the energy quantization signal near each harmonic with the rotational rhythm using the formula, the trend is that the stronger the synchronization, the higher the value, and the value increases across all moments. The resulting sequence is the synchronous energy map.

[0087] It should be noted that when setting the synchronization intensity threshold, multiple sets of synchronization energy spectra are collected under the stable standard operating conditions of the casting machine, the distribution range of synchronization intensity is statistically analyzed, and the median of the distribution is taken as the threshold. This threshold can distinguish between synchronization signals and interference signals.

[0088] Furthermore, traversing the synchronization energy map for identification For time periods exceeding the threshold, signal segments corresponding to these time periods are accurately extracted from the original swing angle sensing signal and spliced ​​together in chronological order to form a synchronous oscillation component.

[0089] Furthermore, the rotation period of the traction wheel is determined by the speed sensor signal. Each signal segment in the synchronous oscillation component is matched according to the corresponding rotation period, and the waveforms are aligned and connected in the order of rotation time to ensure smooth connection of the endpoints of adjacent segments, forming a periodic feature component that continuously and repeatedly presents the characteristics of the rotation period.

[0090] In summary, this embodiment uses the rotation rhythm of the traction wheel as a reference to accurately anchor the signal components related to the core operating state of the casting machine, effectively eliminating irrelevant noise such as electromagnetic interference in the industrial field and improving the targeting of signal screening; it focuses on the fundamental frequency and harmonic frequency to determine the target analysis frequency band, avoids redundancy in the analysis range, and accurately captures the time-frequency energy distribution of the signal by combining time-frequency transformation, and quantifies the synchronicity of each time period through the synchronous energy spectrum, so as to achieve accurate extraction of effective signal segments.

[0091] In summary, by connecting waveforms according to the rotation period, a continuous periodic feature component that conforms to the operating pattern of the equipment is constructed. This provides a high-quality and highly correlated data foundation for the subsequent dynamic correlation analysis of phase trajectory and amplitude envelope, indirectly ensuring the accuracy of subsequent calibration steps and helping to improve the overall efficiency of sensor signal calibration.

[0092] S3. Perform dynamic correlation analysis on the phase trajectory and amplitude envelope of the periodic characteristic components to obtain the dynamic correlation relationship between the phase trajectory and amplitude envelope, including:

[0093] Based on the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope, the overlapping interval of the two in the time window is identified to obtain the coordinated oscillation interval.

[0094] Based on the coordinated oscillation interval, the order of occurrence of the extreme values ​​of phase trajectory angular velocity and amplitude envelope rate of change is tracked to obtain the dominant order relationship;

[0095] Based on the relationship between the cooperative oscillation interval and the dominant order, a cooperative mode is constructed for the phase trajectory and amplitude envelope to obtain a primary dynamic correlation.

[0096] Specifically, we first clarify the method for determining the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope. By continuously calculating the phase change rate of the phase trajectory at adjacent moments in the periodic characteristic component, when the change rate shows a continuous increasing trend, the corresponding time period is the acceleration segment of the phase trajectory. At the same time, we extract the amplitude envelope by comparing the amplitudes at adjacent moments. The time period when the amplitude at the next moment is greater than the amplitude at the previous moment and the trend is maintained continuously is the rising edge of the amplitude envelope.

[0097] Specifically, a fixed-duration sliding time window is set, and the time window is slid along the time axis of the periodic characteristic component unit by unit. Within each time window, the time range of the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope are compared synchronously. When the time ranges of the two overlap, the time interval corresponding to the time window is recorded. All time intervals that meet the conditions are integrated to form the coordinated oscillation interval.

[0098] Specifically, within the defined coordinated oscillation interval, the angular velocity of the phase trajectory is continuously calculated, which is the ratio of the phase difference between adjacent moments to the time interval. By comparing the angular velocity values ​​moment by moment, the maximum and minimum values ​​of the angular velocity within the interval are selected as the extreme values ​​of the angular velocity. At the same time, the rate of change of the amplitude envelope is calculated, which is the ratio of the amplitude difference between adjacent moments to the time interval. Similarly, the maximum and minimum values ​​of the rate of change within the interval are selected as the extreme values ​​of the rate of change of the amplitude envelope.

[0099] Furthermore, by sorting out the occurrence times of the two extreme values ​​in chronological order, and by directly comparing the order of the times, it is determined whether the extreme value of angular velocity or the extreme value of amplitude envelope change rate occurs first in the coordinated oscillation interval. This order is the dominant order relationship.

[0100] Furthermore, using the coordinated oscillation interval as the basic analysis unit, the phase trajectory change curve and the amplitude envelope change curve within each interval are correlated at each time step. The driving logic of the curve change is clarified by combining the dominant order relationship. That is, if the extreme value of angular velocity dominates, the change of phase trajectory will first trigger the corresponding change of amplitude envelope.

[0101] Finally, the curve correspondences of all cooperative oscillation intervals are integrated with the driving logic to form a set that can describe the cooperative change law of phase trajectory and amplitude envelope in different cooperative oscillation intervals. This set is the primary dynamic correlation.

[0102] In this embodiment of the invention, obtaining the dynamic correlation between the phase trajectory and the amplitude envelope further includes:

[0103] The stability of the duration of the coordinated oscillation interval in different traction wheel rotation cycles is evaluated to obtain the steady-state coordinated region and the transient coordinated region.

[0104] Repeatability analysis of the dominant order relationship was performed in the steady-state and transient cooperative regions to obtain the weighting factors;

[0105] Based on the weighting factor, the primary dynamic relationship is reconstructed by weighting to obtain the weighted dynamic relationship, which is then used as the dynamic relationship.

[0106] Weighting factors The calculation formula is

[0107]

[0108] In the formula, Index for collaborative patterns; For the first The variance of the duration of the cooperative oscillation interval corresponding to each cooperative mode; Indicates the first The square of the arithmetic mean of the duration of the cooperative oscillation interval corresponding to each cooperative mode. This represents the number of times the dominant order relationship can be repeated in this collaborative pattern. This represents the total number of analysis periods; For the first The weighting factor of each collaborative mode is such that the larger the value, the more stable and reliable the mode is.

[0109] Specifically, the duration of the coordinated oscillation interval in each traction wheel rotation cycle is first extracted, and data collection is completed by recording the duration value cycle by cycle.

[0110] It should be noted that the stability assessment criterion can be preset by continuously running the casting machine under standard stable operating conditions and collecting multiple sets of traction wheel rotation cycle data, statistically analyzing the dispersion of the duration of the coordinated oscillation interval within each cycle, and taking the critical value of the dispersion as the criterion. This critical value is determined by gradually adjusting the dispersion threshold and observing the stability performance of the coordinated interval. When the threshold is adjusted to a certain value, the coordinated interval corresponding to the dispersion exceeding the threshold will fluctuate in subsequent stable operating conditions, and this value is determined as the critical value.

[0111] Specifically, the duration of each periodic coordinated oscillation interval is compared with the critical value. The coordinated oscillation interval with a duration dispersion lower than the critical value is the steady-state coordinated region, and the one with a dispersion higher than the critical value is the transient coordinated region.

[0112] Specifically, in the formula for calculating the weighting factor, where This is an index for collaborative patterns, obtained by sequentially numbering each collaborative pattern in the primary dynamic association relationship; For the first The variance of the duration of the coordinated oscillation interval corresponding to each coordinated mode is calculated by first finding the arithmetic mean of the durations in all periods of the mode, then calculating the square of the difference between each duration and the mean, summing all the squares and dividing by the number of duration data.

[0113] Specifically, The square of the arithmetic mean of the duration of this pattern is obtained by first summing all durations and dividing by the number of data points, and then squaring the average.

[0114] Specifically, The number of times the dominant order relationship of this pattern can be repeated is obtained by iterating through all analysis periods and counting the number of times the dominant order relationship of each pattern is consistent with the dominant order of the pattern in the first period. This represents the total number of analysis cycles, which is the total number of traction wheel rotation cycles collected in this stability assessment.

[0115] Overall, the weighting factor can be combined with the stability of the duration of the cooperative interval and the repeatability of the dominant order to quantify the reliability of the pattern. The trend is that the smaller the variance, the more repetitions. The larger the value, the more stable and reliable the model.

[0116] Furthermore, the primary dynamic correlation characteristics of each collaborative mode are correlated with the corresponding weight factors. Multiplication enables weighted processing of different collaboration modes. The larger the weight factor of a collaboration mode, the higher the proportion of its corresponding associated features in the integration process.

[0117] Finally, all weighted collaborative mode correlation features are integrated, and collaborative modes with weight factors lower than the preset minimum value are eliminated. This minimum value is verified and determined under standard working conditions. The casting machine is started and run stably. The minimum value is gradually reduced and the stability of the correlation is observed. When the minimum value drops to a certain value, further reduction will cause the correlation to fluctuate. That is, this value is determined to be the preset minimum value. The correlation finally formed is the weighted dynamic correlation, which is used as the dynamic correlation between the phase trajectory and the amplitude envelope.

[0118] In summary, this scheme focuses on the overlapping interval between the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope, locks in the cooperative oscillation interval, excludes signal segments without cooperative significance, defines a precise range for correlation analysis, and avoids blind processing. Secondly, it tracks the order of extreme values ​​within the cooperative interval, clarifies the dominant order relationship, and gives the primary correlation a clear time-driven logic, thus solving the problem of pattern ambiguity in traditional correlation analysis.

[0119] In summary, by dividing the steady-state and transient synergistic regions through stability assessment, and combining the weight factor formula to quantify the repeatability of the dominant order, the primary correlation is reconstructed with weights to highlight the dominant position of highly stable and highly repeatable patterns and eliminate transient interference.

[0120] Overall, this process of precise targeting, clear logic, and weighted optimization ensures that the final dynamic correlation closely matches the mechanical operation of the aluminum ingot casting machine. It retains the core collaborative characteristics while filtering out unstable disturbances, providing a reliable basis for subsequent active time disturbance testing and ensuring the logical consistency and accuracy of subsequent calibration processes.

[0121] S4. Based on the dynamic correlation, the periodic characteristic components are actively perturbed in time to obtain the covariant response intensity of the phase trajectory and amplitude envelope under the perturbation, including:

[0122] Based on the dominant order relationship in the dynamic correlation, key time points in the periodic feature components are identified where the phase trajectory leads the amplitude envelope or the amplitude envelope leads the phase trajectory.

[0123] Apply a time offset that is inverse to the dominant order relationship to key time points to obtain the perturbation component with time reversal;

[0124] By observing the self-recovery process of the order of phase trajectory and amplitude envelope in the perturbation component, we can obtain the trend and speed measure of its convergence to the dominant order relationship.

[0125] Based on convergence trend and speed, the stability of dynamic correlation under time-reversal pressure is quantified to obtain the covariant response intensity.

[0126] In this embodiment of the invention, a temporal offset reversed to the dominant order relationship is applied to key time points to obtain a perturbed component with temporal reversal, including:

[0127] Based on the weight factors associated with key time points in the dynamic correlation, key time points are divided into a high-weight time point subset and a low-weight time point subset.

[0128] A first time-off mode that is completely reversed from the dominant order relationship is applied to a subset of high-weight time-series points, and a second time-off mode that is partially reversed from the dominant order relationship is applied to a subset of low-weight time-series points, to obtain the offset scheme of partitioned modulation.

[0129] The periodic characteristic components are time-shifted according to the offset scheme of partitioned modulation to obtain the perturbed components with time reversal.

[0130] Specifically, the dominant order relationship already determined in the dynamic correlation is retrieved first to clarify whether the phase trajectory leads the amplitude envelope or the amplitude envelope leads the phase trajectory, and this is used as the identification benchmark.

[0131] Then, the complete time series of the periodic feature components is traversed, and the starting time of the change of the phase trajectory and amplitude envelope is recorded moment by moment. By comparing the time sequence of the two starting points of change, the moment points that can directly reflect the dominant order relationship are selected. These moment points are the key time points, ensuring that each key time point can accurately correspond to the leading and lagging logic in the dominant order relationship.

[0132] Furthermore, the weight factors corresponding to the collaborative mode of each key time point in the dynamic relationship are extracted, and a one-to-one mapping between key time points and weight factors is established.

[0133] It should be noted that the weighting threshold is preset in the following way: the casting machine is started and continuously operated under standard stable conditions, multiple sets of weight factor data of key time points are collected, the distribution range of weight factors is statistically analyzed, the threshold is gradually adjusted and small-scale time offset tests are carried out. When the threshold is adjusted to a certain value, time points above the threshold can produce a significant time reversal effect after offset, and time points below the threshold will not excessively destroy the periodicity of the signal after offset. That is, the value is determined as the weighting threshold.

[0134] Then, the weight factor of each key time point is compared with a preset threshold. Those with a weight factor higher than the threshold are classified as a high-weight time point subset, and those with a weight factor lower than the threshold are classified as a low-weight time point subset.

[0135] Furthermore, taking the time flow of the dominant order relationship as a reference, the first time series offset mode is determined to be completely reversed to this flow. Specifically, the time position of each time series point in the high-weight time series point subset is shifted backward or forward as a whole, so that the phase trajectory and amplitude envelope change order of the time series points in the subset are completely reversed after the shift.

[0136] Furthermore, the second timing offset mode is set to partial inversion. First, a preset shift ratio is established experimentally. During the experiment, the proportion of shifted points in the low-weight timing point subset is gradually adjusted under standard operating conditions. The stability of the signal after perturbation is observed. When the proportion is adjusted to a certain value, it can generate timing reversal perturbation without disrupting the fundamental periodicity of the signal; this proportion is then determined as the shift ratio. A portion of timing points from the low-weight timing point subset are selected according to this ratio for inversion shifting. The two offset modes are then bound to their corresponding subsets and integrated to form a partitioned modulation offset scheme.

[0137] Furthermore, based on the offset scheme of partitioned modulation, the specific position of each time point in the periodic feature component of the high-weight time-series point subset is first located by the time axis, and these positions are completely reversed in time according to the first time-series offset mode; then the positions of time points of a preset proportion in the low-weight time-series point subset are located, and partial reverse shift is performed according to the second time-series offset mode.

[0138] Furthermore, during the translation process, the signal amplitude corresponding to each time point remains unchanged, and only the time sequence is adjusted. After the translation is completed, the endpoint values ​​of adjacent signal segments are compared point by point. By fine-tuning the values ​​near the endpoints, a smooth transition between segments is achieved, avoiding signal abrupt changes, and finally obtaining the perturbed component with time reversal.

[0139] Furthermore, the sequence changes of phase trajectories and amplitude envelopes in the components after the perturbation are continuously tracked, recording the complete process from the time-reversal state to the return to the dominant order relationship. By comparing the consistency between the perturbation order and the dominant order at each time step, the convergence trend of the order self-recovery is judged. If the consistency gradually increases over time, the convergence trend is positive. At the same time, the time length required from the occurrence of the perturbation to the complete recovery of the order to stability is recorded. This time length is the convergence speed measure. The shorter the time, the faster the convergence speed.

[0140] Furthermore, the covariant response intensity is calculated using the following formula:

[0141]

[0142] in It serves as an index for the disturbance period, that is, to sequentially number each period of a continuous disturbance; This represents the total number of periods of continuous perturbation. The preset method is to continuously apply perturbations under standard operating conditions until the observed covariant response intensity value tends to stabilize. The number of perturbation periods recorded at this point is the total number of periods. .

[0143] Specifically, For the first The original phase-amplitude order function without perturbation in each period is derived from the time series data of the corresponding period in the periodic characteristic components without perturbation. The first component after perturbation The order relationship function of each period is derived from the corresponding period data of the perturbed components.

[0144] Specifically, the formula quantifies the overall stability of dynamic correlations under time-reversal pressure by averaging the restorative measures over multiple perturbation cycles, with the trend being... The larger the value, the stronger the ability of the dynamic relationship to resist temporal disturbances and maintain its own stability.

[0145] Specifically, the calculation first calculates the numerator integral term, which is the cumulative total of the absolute deviation between the perturbed order and the original order over the entire period, and then calculates the denominator integral term, which is the cumulative total of the absolute value of the original order function over the period. The ratio of the two is the degree of perturbation deviation for that period. Subtracting this ratio from 1 yields the restorative measure for a single period. Sum of the restorative measures over each period and divide by The average value obtained is the covariant response intensity.

[0146] In summary, this scheme identifies key timing points based on the dominant order relationship, accurately pinpointing the nodes that core the phase-amplitude correlation, avoiding invalid perturbations, and improving the targeting of the test. Secondly, by dividing the timing point subsets into high-weight and low-weight subsets according to weighting factors, and adopting a partitioned modulation offset scheme with both fully inverse and partially inverse directions, it ensures that the perturbation strength is sufficient to trigger an effective response while avoiding excessive damage to signal integrity, thus balancing the effectiveness of the perturbation and the stability of the signal.

[0147] In summary, by observing the self-recovery process of components after perturbation and combining the covariant response intensity formula to quantify the convergence trend and speed, the stability of dynamic correlation is transformed from a qualitative description to a quantitative indicator, clearly demonstrating its ability to resist time-reversal perturbations.

[0148] Overall, this process not only verified the reliability of the dynamic correlation, but also provided core quantitative data for the subsequent screening of stable collaborative patterns. It avoided the blindness of relying on experience-based judgments in traditional methods, provided clear data support for the construction of subsequent calibration rules, and indirectly improved the anti-interference capability and accuracy of the final signal calibration.

[0149] S5. Based on the covariant response intensity, a weighted reconstruction is performed on the verified covariance patterns between the phase trajectories and amplitude envelopes of the periodic characteristic components to obtain a dynamic correction mapping, including:

[0150] Based on the comparison results of time-series verification intensity and partitioned response in covariant response intensity, a target mode feature set that remains stable under perturbation is extracted from the covariant variation mode.

[0151] Based on the covariant response intensity, the stability contribution of different features in the target mode feature set is weighted and assigned to obtain the weighted mode rule;

[0152] Based on the weighted mode rules, the correction relationship between the phase trajectory and the amplitude envelope is structured and encoded to obtain the dynamic correction mapping.

[0153] In this embodiment of the invention, based on the covariant response intensity, the stability contribution of different features in the target mode feature set is weighted to obtain a weighted mode rule, including:

[0154] Based on the steady-state or transient cooperative region attributes corresponding to the target pattern feature set in the partition response comparison results, the features are divided into steady-state feature subsets and transient feature subsets.

[0155] For the steady-state feature subset, basic reconstruction weights are assigned based on its temporal verification strength; for the transient feature subset, adaptive decay weights are assigned based on its response divergence characteristics, resulting in a differentiated weight allocation scheme.

[0156] Based on the differentiated weight allocation scheme, the feature set of the target pattern is organized in a regularized manner to form weighted pattern rules.

[0157] Specifically, the temporal verification intensity and the partition response comparison results in the covariance response intensity are first obtained. The temporal verification intensity is obtained by comparing the consistency between the covariance mode after perturbation and the original covariance mode on a cycle-by-cycle basis. When the feature matching degree of the two is higher than the preset matching degree, the temporal verification intensity is higher. The partition response comparison results are obtained by calculating the difference in covariance mode deviation between the perturbation region of the high-weight temporal point subset and the perturbation region of the low-weight temporal point subset.

[0158] Specifically, a stability screening threshold is set. The preset action of this threshold is to start the casting machine under standard stable operating conditions, collect multiple sets of covariance response intensity data under different disturbance intensities, statistically analyze the critical time-series verification intensity of whether the covariance mode is stable or not, gradually adjust the threshold and observe the screening effect. When the threshold is adjusted to a certain value, it can accurately screen out stable mode features with time-series verification intensity higher than that value and deviation difference in the partition response comparison results less than the preset deviation threshold. That is, the value is determined as the stability screening threshold.

[0159] Furthermore, the temporal verification strength of all cooperative change mode features is compared with the stability screening threshold. At the same time, combined with the partition response comparison results, features with temporal verification strength higher than the threshold and deviation difference less than the preset deviation threshold are extracted. These features are the features that remain stable under perturbation, and they are integrated to form the target mode feature set.

[0160] Furthermore, by tracing the original cooperative oscillation interval of each feature in the target mode feature set, and by retrieving the attribute records of the corresponding cooperative oscillation interval in the partition response comparison results, it is determined whether each feature corresponds to a steady-state cooperative region or a transient cooperative region.

[0161] Furthermore, the features corresponding to the steady-state coordinated region are classified into one category, ensuring that all features of this category originate from the coordinated oscillation intervals with stable durations in different traction wheel rotation cycles, and are integrated to form a steady-state feature subset; the features corresponding to the transient coordinated region are classified into another category, ensuring that all features of this category originate from the coordinated oscillation intervals with large duration fluctuations, and are integrated to form a transient feature subset, thus completing the accurate classification of features.

[0162] Furthermore, the allocation threshold for the basic reconstruction weight is preset in the following way: the casting machine is continuously operated under standard stable conditions, and multiple sets of steady-state feature time-series verification intensity data are collected. The influence of features on the overall collaborative mode stability under different time-series verification intensities is statistically analyzed, and the threshold range is gradually adjusted. When the threshold is set to a certain value, features with high time-series verification intensity can obtain higher weights and ensure the stability of the reconstructed mode. At the same time, the weights of features with low time-series verification intensity will not be too low, resulting in the loss of effective information. That is, this value is determined as the basic weight allocation threshold.

[0163] Furthermore, for each feature in the steady-state feature subset, its temporal verification strength is compared with a preset basic weight allocation threshold. Features with temporal verification strength higher than the threshold are assigned higher basic reconstruction weights, while features with temporal verification strength lower than the threshold are assigned lower basic reconstruction weights, ensuring that the weight allocation matches the temporal stability of the features.

[0164] Furthermore, the response divergence characteristics are determined by continuously monitoring the response change amplitude of transient features after disturbance. The larger the change amplitude, the stronger the response divergence. The attenuation coefficient of the adaptive attenuation weight is preset by experiment. During the experiment, the attenuation coefficient is gradually adjusted under standard operating conditions, while observing the impact of transient feature weight on the overall mode stability. When the attenuation coefficient is adjusted to a certain value, it can effectively suppress the interference of highly divergent transient features on mode stability, while retaining the effective information of weakly divergent transient features. That is, this coefficient is determined to be the adaptive attenuation coefficient.

[0165] Furthermore, an adaptive decay weight is assigned to each feature in the transient feature subset based on the adaptive decay coefficient. Features with stronger response divergence are assigned smaller weights, while features with weaker divergence are assigned larger weights. The basic reconstruction weight allocation rule and the adaptive decay weight allocation rule are integrated according to the feature subset category to form a differentiated weight allocation scheme.

[0166] Furthermore, based on the differentiated weight allocation scheme, a one-to-one correspondence is established between each feature in the steady-state feature subset and the transient feature subset and its corresponding weight, ensuring that the weight attribute of each feature is clear and unambiguous.

[0167] Furthermore, all features are sorted in descending order of weight value, prioritizing high-weight features and placing low-weight features in subsequent positions. The sorted features and their corresponding weights are then converted into standardized rule entries. Each entry clearly labels the weight value corresponding to the feature content and the collaborative region attribute to which the feature belongs, while also clarifying the scope of the feature's role in the subsequent construction of correction relationships.

[0168] Then, all rule entries are systematically integrated, and the logical relationships between the entries are sorted out to ensure that the rule entries corresponding to high-weight features are executed first, while the rule entries corresponding to low-weight features are supplemented to form a logically coherent and clearly defined set of structured rules, which is the weighted pattern rule.

[0169] Furthermore, the correction relationship between the phase trajectory and the amplitude envelope is clarified as the deviation adjustment law when the two change in tandem. This law is derived from the previously determined dynamic correlation relationship. That is, when the actual changes of the phase trajectory and the amplitude envelope deviate from the dynamic correlation relationship, they need to be adjusted to a state that conforms to the law through the correction relationship.

[0170] Furthermore, based on the weight order of each feature in the weighted pattern rules, starting from the correction relationship corresponding to the high-weight feature, the correction relationship of each feature is sequentially converted into a standardized coding entry. Each coding entry contains the feature recognition identifier correction action instruction and the corresponding weight ratio, ensuring that the coding entry can accurately correspond to the content in the weighted pattern rules.

[0171] Finally, all coded entries are structurally integrated to construct a hierarchical coding system. Coded entries with high-weight features are placed in the upper-level priority execution layer, while those with low-weight features are placed in the lower-level auxiliary layer. Simultaneously, connection logic is set between entries to ensure that coded entries at different levels can work synergistically. After integration, a structured coding set is formed that clearly defines the relationship between phase trajectory and amplitude envelope correction according to weight and priority; this set is the dynamic correction mapping.

[0172] In summary, this embodiment relies on the time-series verification strength of the covariant response strength and the comparison results of the partitioned response to screen out the target mode feature set that remains stable under disturbances, eliminate easily divergent transient interference features, ensure the effectiveness of the correction logic from the source, and solve the problem of interference features misleading in traditional correction.

[0173] In summary, this scheme implements differentiated weight allocation based on stability contribution, assigns basic reconstruction weights to steady-state features and adaptive decay weights to transient features, forming a weighted pattern rule with clear priorities, highlighting the dominant role of highly stable features, avoiding deviations in correction direction, and strengthening the mapping's targeting of core collaborative relationships.

[0174] In summary, the structured encoding of correction relationships is transformed into a hierarchical encoding system containing feature identification identifiers, correction instructions, and weight proportions. This converts fragmented collaborative rules into a system-executable mapping scheme, ensuring that subsequent signal correction follows a set pattern. The entire process, from feature selection and weight allocation to encoding integration, is progressive, guaranteeing the reliability of the correction logic while improving operational standardization and efficiency. This provides core technical support for the final accurate processing of the original signal and the output of a high-quality calibration signal.

[0175] S6. The original swing angle sensing signal is processed through dynamic correction mapping to obtain the calibrated swing angle signal, including:

[0176] Based on the correction relationship encoded in the dynamic correction mapping, feature decoupling and relationship correction are performed on the signal segments in the original swing angle sensing signal that match the mapping rules to obtain the preliminary correction signal segments;

[0177] For feature points in the initial correction signal segment that have been corrected according to the mapping rules, a smooth transition is performed between them and adjacent uncorrected feature points to obtain a continuous correction signal with a natural transition.

[0178] The amplitude of the continuous correction signal is normalized and the timing is aligned so that it is resynchronized with the rotation rhythm of the casting machine traction wheel to obtain the calibrated swing angle signal.

[0179] Specifically, the correction relationship encoded in the dynamic correction mapping is first defined as the phase trajectory and amplitude envelope adjustment rules sorted by weight priority. All signal segments of the original swing angle sensing signal are traversed. By comparing the phase and amplitude characteristics of each segment with the mapping rules in the dynamic correction mapping at each time step, signal segments with completely matching characteristics are selected.

[0180] Furthermore, feature decoupling is achieved by separating the phase trajectory features and amplitude envelope features within the matching segment. The two types of features are corrected separately according to the correction relationship. The correction process strictly follows the rule of prioritizing high-weight features to ensure that the corrected features conform to the law of coordinated change. All the matching segments that have been corrected are integrated to obtain the preliminary corrected signal segment.

[0181] Furthermore, the preliminary correction signal segment is traversed time-by-time, and the feature points that have been corrected according to the mapping rules are accurately identified by comparing the feature values ​​before and after the signal correction.

[0182] Specifically, the amplitude and phase information of the adjacent uncorrected feature points before and after each corrected feature point are extracted. According to the natural change trend of the adjacent uncorrected feature points, the values ​​of the corrected feature points are gradually fine-tuned so that the numerical changes between the corrected feature points and the uncorrected feature points before and after them present a continuous and gradual trend, avoiding abrupt changes. After the transition adjustment of all corrected feature points is completed, a continuous correction signal with a natural transition is obtained.

[0183] Specifically, the amplitude normalization reference value can be preset in the following way: start the casting machine and continuously collect multiple sets of normal swing angle signals under standard stable working conditions, count the amplitude distribution range of these signals, and take the maximum value within the distribution range as the normalization reference value.

[0184] Then, the amplitude values ​​at all times in the continuous correction signal are divided by the reference value to complete the amplitude normalization.

[0185] Finally, the timing alignment is based on the rotation rhythm of the traction wheel. The rotation cycle signal of the traction wheel is obtained in real time through the speed sensor. The time axis of the continuous correction signal is aligned and adjusted with the time axis of the rotation cycle signal unit by unit, so that the periodic change of the correction signal is completely synchronized with the rotation cycle of the traction wheel, and finally the calibrated swing angle signal is obtained.

[0186] In summary, in this embodiment, the covariance response strength originates from the convergence quantization analysis after active temporal perturbation. Its temporal verification strength is obtained by comparing the consistency between the perturbed and original cooperative modes. The partition comparison results reflect the deviation differences between the perturbation regions of high and low weight time series points. By using this data to screen the target mode feature set, transient interference features that are easily dispersed by perturbation can be accurately eliminated, while stable and reliable core features are retained. This provides high-quality data support for the construction of dynamic correction mapping, ensuring from the source that the correction logic conforms to the actual cooperative rules of the equipment, and significantly improving the effectiveness of correction.

[0187] In summary, the weight allocation is based on the contribution of feature stability. The document assigns basic weights to steady-state features according to the temporal verification intensity, and assigns adaptive decay weights to transient features according to the response divergence characteristics, forming a differentiated scheme. This approach highlights the dominant position of highly stable features, weakens the interference of low-contribution features, clarifies the priority of the weighting mode rules, avoids directional deviations caused by invalid features during the correction process, and strengthens the mapping's targeting of the core cooperative relationship of the swivel signal.

[0188] In summary, structured coding transforms fragmented collaborative rules into standardized entries containing feature identification markers, correction instructions, and weight proportions, constructing a hierarchical execution system. This allows the originally scattered correction logic to form a system-executable mapping, enabling subsequent signal correction to proceed in an orderly manner according to weight priority, thereby improving operational standardization, reducing logical conflicts, and accelerating correction efficiency.

[0189] In the several embodiments provided by this invention, it should be understood that the disclosed method can be implemented in other ways.

[0190] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0191] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, and technology that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine, characterized in that, The method includes: S1. Based on the operating status of the aluminum ingot continuous casting machine, the output signal of the swing angle tension bridge sensor is collected to obtain the original swing angle sensing signal; S2. Using the rotational rhythm of the casting machine's traction wheel as a reference, the synchronous oscillation component is selected from the original swing angle sensing signal to obtain the periodic characteristic component; S3. Perform dynamic correlation analysis on the phase trajectory and amplitude envelope of the periodic characteristic components to obtain the dynamic correlation relationship between the phase trajectory and amplitude envelope; S4. Based on the dynamic correlation, the periodic characteristic components are actively perturbed in time to obtain the covariant response intensity of the phase trajectory and amplitude envelope under the perturbation. S5. Based on the covariant response intensity, the verified cooperative variation mode between the phase trajectory and amplitude envelope of the periodic characteristic components is reconstructed by weighting to obtain the dynamic correction mapping. S6. The original swing angle sensing signal is processed by dynamic correction mapping to obtain the calibrated swing angle signal.

2. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 1, characterized in that, Based on the operating status of the continuous aluminum ingot casting machine, the output signal of the swing angle tension bridge sensor is collected to obtain the original swing angle sensing signal, including: Based on the pulse signal of the rotary encoder of the casting machine traction wheel, the operating cycle and phase of the casting machine are analyzed in real time to obtain the state timing identifier synchronized with the traction action; Based on the state timing identifier, the output signal of the pendulum tension bridge sensor is synchronously triggered and acquired to obtain a discrete signal sequence; The discrete signal sequence is subjected to non-steady-state segment identification and steady-state segment extraction to obtain the original swing angle sensing signal.

3. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 1, characterized in that, The process of filtering synchronous oscillation components from the original swing angle sensing signal, using the rotational rhythm of the casting machine's traction wheel as a reference, yields periodic characteristic components, including: The real-time rotation frequency of the traction wheel is obtained as the fundamental frequency, and the target analysis frequency band is determined based on the fundamental frequency; Within the target analysis frequency band, the original swing angle sensing signal is transformed by time and frequency to obtain the time and frequency energy distribution of the signal; Determine the coherence between the time-frequency energy distribution and the fundamental frequency and its harmonics, and generate a synchronization energy spectrum characterizing the strength of synchronization in each time period; Based on the synchronization energy spectrum, signal segments corresponding to time periods when the synchronization intensity exceeds a preset threshold are extracted from the original swing angle sensing signal and combined to form a synchronization oscillation component; By connecting the synchronous oscillation components according to the rotation period of the traction wheel, the periodic characteristic components are obtained.

4. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 1, characterized in that, The dynamic correlation analysis of the phase trajectory and amplitude envelope of the periodic characteristic components to obtain the dynamic correlation relationship between the phase trajectory and amplitude envelope includes: Based on the acceleration segment of the phase trajectory and the rising edge of the amplitude envelope, the overlapping interval of the two in the time window is identified to obtain the coordinated oscillation interval. Based on the coordinated oscillation interval, the order of occurrence of the extreme values ​​of phase trajectory angular velocity and amplitude envelope rate of change is tracked to obtain the dominant order relationship; Based on the relationship between the cooperative oscillation interval and the dominant order, a cooperative mode is constructed for the phase trajectory and amplitude envelope to obtain a primary dynamic correlation.

5. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 4, characterized in that, The obtained dynamic correlation between the phase trajectory and the amplitude envelope also includes: The stability of the duration of the coordinated oscillation interval in different traction wheel rotation cycles is evaluated to obtain the steady-state coordinated region and the transient coordinated region. Repeatability analysis of the dominant order relationship was performed in the steady-state and transient cooperative regions to obtain the weighting factors; Based on the weighting factor, the primary dynamic relationship is reconstructed by weighting to obtain the weighted dynamic relationship, which is then used as the dynamic relationship.

6. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 1, characterized in that, The method of actively perturbing the periodic feature components based on dynamic correlation to obtain the covariant response intensity of the phase trajectory and amplitude envelope under perturbation includes: Based on the dominant order relationship in the dynamic correlation, key time points in the periodic feature components are identified where the phase trajectory leads the amplitude envelope or the amplitude envelope leads the phase trajectory. Apply a time offset that is inverse to the dominant order relationship to key time points to obtain the perturbation component with time reversal; By observing the self-recovery process of the order of phase trajectory and amplitude envelope in the perturbation component, we can obtain the trend and speed measure of its convergence to the dominant order relationship. Based on convergence trend and speed, the stability of dynamic correlation under time-reversal pressure is quantified to obtain the covariant response intensity.

7. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 6, characterized in that, The process of applying a temporal offset inverse to the dominant order relationship to key time points to obtain the perturbed component with temporal reversal includes: Based on the weight factors associated with key time points in the dynamic correlation, key time points are divided into a high-weight time point subset and a low-weight time point subset. A first time-off mode that is completely reversed from the dominant order relationship is applied to a subset of high-weight time-series points, and a second time-off mode that is partially reversed from the dominant order relationship is applied to a subset of low-weight time-series points, to obtain the offset scheme of partitioned modulation. The periodic characteristic components are time-shifted according to the offset scheme of partitioned modulation to obtain the perturbed components with time reversal.

8. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 6, characterized in that, The dynamic correction mapping is obtained by weighting and reconstructing the verified co-variation patterns between the phase trajectories and amplitude envelopes of periodic characteristic components based on the covariant response intensity, including: Based on the comparison results of time-series verification intensity and partitioned response in covariant response intensity, a target mode feature set that remains stable under perturbation is extracted from the covariant variation mode. Based on the covariant response intensity, the stability contribution of different features in the target mode feature set is weighted and assigned to obtain the weighted mode rule; Based on the weighted mode rules, the correction relationship between the phase trajectory and the amplitude envelope is structured and encoded to obtain the dynamic correction mapping.

9. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 8, characterized in that, The weighted pattern rule, based on the covariant response intensity, assigns weights to the stability contributions of different features in the target pattern feature set to obtain the rule, including: Based on the steady-state or transient cooperative region attributes corresponding to the target pattern feature set in the partition response comparison results, the features are divided into steady-state feature subsets and transient feature subsets. For the steady-state feature subset, basic reconstruction weights are assigned based on its temporal verification strength; for the transient feature subset, adaptive decay weights are assigned based on its response divergence characteristics, resulting in a differentiated weight allocation scheme. Based on the differentiated weight allocation scheme, the feature set of the target pattern is organized in a regularized manner to form weighted pattern rules.

10. The method for calibrating the pendulum tension bridge sensor signal of a continuous aluminum ingot casting machine as described in claim 1, characterized in that, The process of processing the original swing angle sensing signal through dynamic correction mapping to obtain the calibrated swing angle signal includes: Based on the correction relationship encoded in the dynamic correction mapping, feature decoupling and relationship correction are performed on the signal segments in the original swing angle sensing signal that match the mapping rules to obtain the preliminary correction signal segments; For feature points in the initial correction signal segment that have been corrected according to the mapping rules, a smooth transition is performed between them and adjacent uncorrected feature points to obtain a continuous correction signal with a natural transition. The amplitude of the continuous correction signal is normalized and the timing is aligned so that it is resynchronized with the rotation rhythm of the casting machine traction wheel to obtain the calibrated swing angle signal.

Images