A safe fault-tolerant control method for power battery aluminum foil rolling

Abstract

This invention discloses a safety-tolerant control method for rolling aluminum foil for power batteries, comprising: acquiring rolling force distribution data, exit thickness data, and tension data during the rolling process; extracting rolling force direction features, thickness fluctuation spectrum features, and tension oscillation features; constructing a multi-dimensional physical feature space; constructing an anisotropic safety envelope in the multi-dimensional physical feature space, wherein the length of each half-axis is negatively correlated with the corresponding rolling force direction feature, thickness fluctuation spectrum feature, and tension oscillation feature; taking the feature point corresponding to the optimal process parameter in the candidate point set as the current feature point; identifying the coupling risk between the thickness fluctuation spectrum feature and the tension oscillation feature, and dynamically adjusting the boundary of the anisotropic safety envelope according to the identification result; determining the correction direction and adjusting the rolling process parameters according to the positional relationship between the current feature point and the boundary; thereby solving the problem that single-variable control methods cannot fully reflect the multi-physics coupling risk in the rolling process.

Description

Technical Field

[0001] This invention relates to the field of battery aluminum foil rolling technology, and in particular to a safety-tolerant control method for rolling power battery aluminum foil. Background Technology

[0002] Aluminum foil for power batteries is one of the key materials for lithium-ion batteries. Its thickness is typically between 0.008 mm and 0.02 mm, falling into the category of ultra-thin foil. During the rolling process, due to its extremely thin thickness and relatively large width, aluminum foil is highly susceptible to factors such as fluctuations in rolling force, uneven thickness, and tension oscillations, leading to frequent strip breakage accidents. Strip breakage not only wastes raw materials and causes production interruptions but can also damage rolls and equipment, severely impacting production efficiency and economic benefits.

[0003] Traditional aluminum foil rolling control methods primarily rely on single-parameter monitoring and fixed-threshold alarms, such as monitoring whether the rolling force exceeds the upper limit, whether the thickness deviation exceeds the standard, or whether the tension is unstable. However, these methods have limitations: First, the normality of a single parameter does not guarantee system safety; the coupling effect of multiple parameters may lead to accidents. Second, fixed thresholds cannot adapt to the dynamic changes in the rolling process; if the threshold is set too wide, the warning will be delayed; if it is set too strict, false alarms will be frequent. Third, traditional methods lack the ability to identify the coupling relationship between parameters and cannot warn of "hidden risks," that is, situations where all individual parameters are within the normal range but their combined effect leads to system instability. In addition, existing safety controls are mostly passive response-based, that is, intervention is only carried out after an anomaly is detected, by which time the optimal adjustment opportunity has often been missed. For a high-speed dynamic process like rolling, passive response is difficult to prevent accidents from occurring.

[0004] Therefore, there is an urgent need for a safety fault-tolerant control method that can comprehensively consider multi-physics coupling, adaptively adjust safety boundaries, and actively identify and eliminate risk sources. Summary of the Invention

[0005] The technical problems solved by this invention are: existing control methods are difficult to reflect the anisotropic safety boundary of the true risk distribution; there is a lack of online identification and quantitative assessment methods for the coupled risks of thickness fluctuation and tension oscillation; when the rolling state approaches the safety boundary, it is impossible to determine the optimal correction direction and implement active fault-tolerant control based on the geometric relationship of the multidimensional feature space; and there is a lack of targeted phase compensation strategies for different disturbance sources such as roll eccentricity, tension control instability, and incoming material fluctuation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] A safety-tolerant control method for rolling aluminum foil for power batteries includes the following steps:

[0008] Acquire rolling force distribution data, exit thickness data, and tension data during the rolling process;

[0009] Based on the rolling force distribution data, rolling force direction features are extracted; based on the exit thickness data, thickness fluctuation spectrum features are extracted; based on the tension data, tension oscillation features are extracted.

[0010] Based on the rolling force direction characteristics, thickness fluctuation spectrum characteristics, and tension oscillation characteristics, a multidimensional physical feature space is constructed.

[0011] An anisotropic safety envelope is constructed in the multidimensional physical feature space. The anisotropic safety envelope has a superellipsoidal shape. The length of each half-axis is negatively correlated with the corresponding rolling force direction feature, thickness fluctuation spectrum feature and tension oscillation feature, and the rate of change of different half-axis lengths is independent of each other.

[0012] The multi-objective optimization model trained based on historical rolling data outputs a set of candidate process parameters. The set of candidate process parameters is projected onto the multi-dimensional physical feature space to form a set of candidate points. The feature point corresponding to the optimal process parameter is selected from the set as the current feature point with the goal of maximizing the safety margin.

[0013] Identify the coupling risk between the thickness fluctuation spectrum characteristics and the tension oscillation characteristics, and dynamically adjust the boundary of the anisotropic safety envelope based on the identification results;

[0014] When the current feature point in the multidimensional physical feature space approaches the boundary of the anisotropic safety envelope, the correction direction is determined according to the positional relationship between the current feature point and the boundary.

[0015] Adjust the rolling process parameters according to the correction direction so that the current feature point returns to the anisotropic safety envelope.

[0016] Preferably, the step of extracting the rolling force direction feature includes:

[0017] The three-dimensional distribution data of rolling force in the rolling direction is obtained by a multi-point sensor array installed on the roll bearing housing;

[0018] Calculate the gradient vector of the rolling force in the three-dimensional distribution data;

[0019] The modulus and principal direction angle are calculated based on the gradient vector, and the complex form of the modulus and the principal direction angle is combined as the rolling force direction feature.

[0020] Preferably, the step of extracting the thickness fluctuation spectral features includes:

[0021] By integrating X-ray thickness gauge data and grating displacement sensor data, continuous fluctuation data of the exit thickness can be obtained;

[0022] Spectral analysis was performed on the continuous fluctuation data to obtain the thickness fluctuation spectrum;

[0023] The thickness fluctuation spectrum is weighted and integrated according to a preset frequency domain sensitive weighting function to obtain the thickness fluctuation spectrum characteristics. The frequency domain sensitive weighting function is preset based on the frequency domain statistical results of historical band breakage accidents.

[0024] Preferably, the step of extracting tension oscillation features includes:

[0025] Acquire pre-tension and post-tension data at a sampling frequency of not less than 1 kHz;

[0026] Calculate the root mean square value of the fluctuation of the front tension data relative to its mean, and the root mean square value of the fluctuation of the back tension data relative to its mean, respectively.

[0027] The root mean square value of the fluctuation of the preceding tension data and the root mean square value of the fluctuation of the following tension data are combined to form the tension oscillation feature.

[0028] Preferably, the step of constructing the anisotropic safety envelope includes:

[0029] The rolling force direction characteristics, the thickness fluctuation spectrum characteristics, and the tension oscillation characteristics are respectively used as the first dimension, the second dimension, and the third dimension of the superellipsoid safety envelope;

[0030] Define the reference semi-axis lengths of the first dimension, the second dimension, and the third dimension;

[0031] The actual half-axis length of the first dimension is calculated based on the die length of the rolling force direction feature, and the actual half-axis length of the first dimension has a negative exponential relationship with the die length of the rolling force direction feature.

[0032] The actual half-axis length of the second dimension is calculated based on the numerical value of the thickness fluctuation spectrum feature. The actual half-axis length of the second dimension has a negative exponential relationship with the numerical value of the thickness fluctuation spectrum feature.

[0033] The actual half-axis length of the third dimension is calculated based on the numerical value of the tension oscillation characteristic, and the actual half-axis length of the third dimension has a negative exponential relationship with the numerical value of the tension oscillation characteristic.

[0034] The negative exponential decay rates of the first dimension, the second dimension, and the third dimension are independent of each other and are controlled by their respective sensitivity coefficients. The sensitivity coefficients are obtained by back-optimization based on historical belt breakage accident data, with the goal of minimizing the prediction error of belt breakage accidents.

[0035] Preferably, the step of determining the approximation of the anisotropic safety envelope boundary includes:

[0036] Calculate the magnitude of the current feature point in the multidimensional physical feature space;

[0037] Calculate the modulus of the corresponding point of the current feature point on the boundary of the anisotropic safety envelope;

[0038] Calculate the ratio of the magnitude of the current feature point to the magnitude of the boundary point;

[0039] When the ratio reaches a preset warning threshold, it is determined that the current feature point is approaching the boundary of the anisotropic safety envelope.

[0040] Preferably, the step of determining the correction direction includes:

[0041] Calculate the shortest distance from the current feature point to the boundary of the anisotropic safety envelope in the multidimensional physical feature space;

[0042] Determine the boundary point corresponding to the shortest distance;

[0043] Calculate the gradient direction from the current feature point to the boundary point;

[0044] The opposite direction of the gradient direction is determined as the correction direction.

[0045] Preferably, the step of adjusting the rolling process parameters includes:

[0046] The type of process parameter to be adjusted is determined according to the correction direction, and the type of process parameter includes at least one of rolling speed, front tension, back tension, and rolling force setting value;

[0047] The adjustment range is calculated based on the distance between the current feature point and the boundary point;

[0048] The adjustment range is filtered according to a preset smoothing filter coefficient, which is preset based on the dynamic response characteristics of the rolling mill.

[0049] The filtered adjustment value is output to the rolling mill actuator.

[0050] Preferably, the step of identifying coupling risks specifically includes:

[0051] Calculate the cross-phase spectrum of the thickness fluctuation spectral characteristics and the tension oscillation characteristics;

[0052] The phase lead relationship between thickness fluctuation and tension oscillation is determined based on the cross-phase spectrum, and the dominant source of disturbance is identified.

[0053] The integral of the energy product of the thickness fluctuation spectral characteristics and the tension oscillation characteristics in the overlapping frequency range is calculated as the coupling risk coefficient.

[0054] Calculate the nonlinear cooperative risk term of the thickness fluctuation spectrum feature and the tension oscillation feature. The nonlinear cooperative risk term is positively correlated with the square of the thickness fluctuation spectrum feature, the square of the tension oscillation feature, and the cosine of the phase difference between the two.

[0055] The coupling risk coefficient and the nonlinear cooperative risk term are used as additional modulation signals for the anisotropic safety envelope. When the coupling risk coefficient or the nonlinear cooperative risk term exceeds its corresponding risk threshold, the boundary of the anisotropic safety envelope is automatically tightened.

[0056] Preferably, it also includes phase compensation control:

[0057] When the coupling risk coefficient exceeds the risk threshold, the dominant frequency of the thickness fluctuation spectrum characteristic is changed by adjusting the rolling speed, causing it to deviate from the inherent frequency of the tension control system.

[0058] When the nonlinear collaborative risk term exceeds the risk threshold, the phase of the tension oscillation characteristic is changed by adjusting the pre-tension setting, so that the phase difference between the thickness fluctuation and the tension oscillation tends to 180 degrees.

[0059] When the main source of disturbance is identified as roll eccentricity, the roll eccentricity compensation algorithm is activated first.

[0060] When tension control instability is identified as the dominant source of disturbance, the tension closed-loop control parameters should be adjusted first.

[0061] When the main source of disturbance is identified as fluctuation in incoming material, the rolling speed should be adjusted first.

[0062] The beneficial effects of this invention are:

[0063] First, the safety-tolerant control method for rolling aluminum foil for power batteries provided by this invention constructs a multi-dimensional physical feature space that integrates rolling force direction characteristics, thickness fluctuation spectrum characteristics, and tension oscillation characteristics, and builds an anisotropic safety envelope in this space. This allows for the simultaneous characterization of the spatial distribution of rolling force, the frequency domain characteristics of thickness fluctuation, and the dynamic oscillation characteristics of tension, thereby achieving an accurate representation of the overall risk state of the rolling system. In this way, it solves the problem that traditional single-variable control methods cannot fully reflect the multi-physical field coupling risks in the rolling process, and significantly improves the comprehensiveness and accuracy of safety assessment.

[0064] Secondly, the safety tolerance control method for rolling aluminum foil for power batteries provided by this invention designs the anisotropic safety envelope as a hyperellipsoid, so that the length of each half-axis is negatively correlated with the corresponding characteristic value and the rate of change of different half-axis lengths are independent of each other. It can adaptively adjust the safety boundary according to the risk characteristics of abnormal rolling force direction, severe thickness fluctuation and tension oscillation amplitude, so as to realize the refined and personalized dynamic adjustment of the safety envelope. In this way, it solves the problem that traditional fixed threshold and isotropic safety boundary cannot adapt to the differentiated risk sensitivity of different physical characteristics, and plays a significant role in improving the pertinence and effectiveness of safety control strategy. Attached Figure Description

[0065] Figure 1 This invention provides a basic flowchart of a safety-tolerant control method for rolling aluminum foil for power batteries. Detailed Implementation

[0066] To make the above-mentioned objects, features and advantages of the present invention more readily understood, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0067] Reference Figure 1 This invention provides a safety-tolerant control method for the rolling of aluminum foil for power batteries, applicable to aluminum foil rolling production lines, particularly for the rolling process of ultra-thin aluminum foil (typically 0.008mm-0.02mm thick) for power batteries. This method constructs a multi-dimensional physical feature space and establishes an anisotropic safety envelope to achieve safety-tolerant control of the rolling process, effectively preventing strip breakage accidents and improving production stability and product quality.

[0068] S100: Data Acquisition and Preprocessing

[0069] S110: Obtaining Rolling Force Distribution Data

[0070] This invention acquires three-dimensional distribution data of rolling force along the rolling direction using a multi-point sensor array mounted on a roll bearing housing. The multi-point sensor array includes at least four piezoelectric force sensors, respectively arranged at the four corners of the work roll bearing housing, with a sampling frequency of not less than 2 kHz, to capture the dynamic changes of rolling force in the rolling direction, transverse direction, and vertical direction. The three-dimensional distribution data includes the distribution of rolling force along the rolling direction (longitudinal), along the roll length direction (transverse), and the normal distribution perpendicular to the rolling plane.

[0071] S120: Obtaining Export Thickness Data

[0072] By fusing data from an X-ray thickness gauge and a grating displacement sensor, continuous fluctuation data of the exit thickness is obtained. The X-ray thickness gauge, installed on the exit side of the rolling mill, measures the absolute thickness of the aluminum foil with an accuracy of ±0.1 μm. The grating displacement sensor, installed in the roll bearing housing, measures the elastic deformation of the roll and the roll gap change, indirectly reflecting the thickness variation. A data fusion algorithm is used to time-align and weight the measurement data from both sensors, eliminating measurement noise and systematic errors from individual sensors and obtaining high signal-to-noise ratio thickness fluctuation data.

[0073] S130: Tension Data Acquisition

[0074] Tension data (entry-side tension) and post-entry tension data (exit-side tension) are collected using tension meters at a sampling frequency of no less than 1 kHz. The front tension sensor is mounted on the tension roll bearing housing at the mill inlet, and the post-entry tension sensor is mounted on the coiler drive shaft at the mill exit. The acquisition of tension data must be synchronized with the rolling force and thickness data, and time alignment of multi-source data is achieved through a unified timestamp.

[0075] S140: Data Preprocessing

[0076] The collected raw data underwent filtering, noise reduction, outlier removal, and normalization. A moving average filter was used to eliminate high-frequency noise, the 3σ criterion was used to remove outliers, and the Z-score normalization method was used to normalize the data across all dimensions to the same scale, laying the foundation for subsequent feature extraction.

[0077] S200: Multidimensional Physical Feature Extraction

[0078] S210: Rolling Force Direction Feature Extraction

[0079] S211: Construction of Three-Dimensional Distribution Data

[0080] Based on the three-dimensional distribution data of rolling force obtained from S110, a spatial distribution function F(x,y,z) of rolling force is constructed, where x represents the rolling direction, y represents the transverse direction, and z represents the normal direction.

[0081] S212: Gradient Vector Calculation

[0082] Calculate the gradient vector of rolling force in three-dimensional distribution data. ,Right now:

[0083] ;

[0084] in, It represents the rate of change of rolling force along the rolling direction, reflecting the longitudinal instability of the rolling process; This indicates the uneven distribution of rolling force along the roll body, reflecting the risk of sheet shape defects; It represents the normal fluctuation of the rolling force and reflects the vibration state of the rolls.

[0085] S213: Calculation of module length and principal direction angle

[0086] Calculate its magnitude based on the gradient vector. and principal direction angles θ and φ:

[0087] ;

[0088] ;

[0089] ;

[0090] in, The principal direction angle in the horizontal plane. It is the principal direction angle in the vertical plane.

[0091] S214: Rolling force direction feature that combines the die length and principal direction angle into a complex form. :

[0092] ;

[0093] This complex form simultaneously contains both intensity information (die length) and direction information (angle) of the rolling force variation, enabling a comprehensive characterization of the spatial distribution characteristics of the rolling force. An increase in die length indicates increased unevenness in the rolling force distribution and a higher risk of strip shape deterioration. When the principal direction angle in the horizontal plane deviates laterally (approaching 90 degrees), it indicates prominent lateral strip shape problems. An increase in the principal direction angle in the vertical plane indicates intensified vertical vibration of the rolls.

[0094] In this way, the present invention achieves the organic integration of rolling force intensity and direction information. Compared with a single scalar feature, it can more sensitively capture the multidimensional abnormal state of the rolling process and improve the accuracy of fault early warning.

[0095] S220: Thickness Fluctuation Spectral Feature Extraction

[0096] S221: Acquisition of Continuous Fluctuation Data

[0097] Based on the S120 fused export thickness data, a continuous fluctuation sequence h(t) of thickness over time is obtained. This sequence reflects the dynamic change of thickness during the rolling process, including thickness variation information caused by various factors such as roll eccentricity, incoming material thickness fluctuation, and rolling force fluctuation.

[0098] S222: Spectrum Analysis

[0099] The thickness fluctuation spectrum is obtained by performing a fast Fourier transform on the continuous fluctuation sequence h(t). :

[0100] ;

[0101] The frequency range of the spectrum analysis covers 0.1Hz to 500Hz, with a frequency resolution of 0.1Hz. It can distinguish thickness fluctuation components in different frequency bands, such as roll eccentricity (usually corresponding to the fundamental frequency and its harmonics of the roll speed), hydraulic system oscillation (usually 10-50Hz), and mechanical vibration (usually 100-500Hz).

[0102] S223: Setting the frequency domain sensitive weighting function

[0103] Based on the frequency domain statistics of historical band breakage accidents, a frequency domain sensitivity weighting function is pre-defined. This function identifies characteristic frequency ranges highly correlated with the risk of band breakage through spectral analysis of historical band breakage incidents and assigns them higher weights. For example, if statistical analysis shows that thickness fluctuations are most correlated with band breakage incidents in the 20-30Hz frequency band, then this frequency band will be given higher weights. The value is set to 1.0, while other frequency bands are set to 0.5-0.8 in descending order of correlation.

[0104] S224: Weighted integral calculation

[0105] The thickness fluctuation spectrum characteristics are obtained by weighting and integrating the spectrum using a frequency-domain sensitive weighting function. :

[0106] ;

[0107] The larger this eigenvalue, the higher the thickness fluctuation energy within the sensitive frequency band, and the greater the risk of band breakage. Among these, the frequency domain sensitive weighting function... The construction method is as follows: Thickness fluctuation spectrum data within 10 seconds prior to historical band breakage accidents are collected; the energy distribution of each frequency band is statistically analyzed; the correlation coefficient between the energy of each frequency band and the occurrence rate of band breakage accidents is calculated; and the normalized correlation coefficient is used as the weight value for each frequency band, enabling feature extraction to focus on the critical frequency bands that truly affect safety. In other words, this invention introduces frequency-domain sensitive weights based on historical accident statistics, enabling feature extraction to focus on the critical frequency bands that truly affect safety, avoiding interference from non-sensitive frequency bands, and improving the targeting and accuracy of risk identification.

[0108] S230: Tension Oscillation Feature Extraction

[0109] S231: Acquisition of front and rear tension data

[0110] Based on the pretension data collected by S130 and post-tension data Calculate the mean of their time series respectively. , and standard deviation , .

[0111] S232: Calculation of RMS Value of Fluctuation

[0112] Calculate the root mean square value of the fluctuation of the pretension data relative to its mean. And the root mean square value of the fluctuation of the back tension data relative to its mean. :

[0113] ;

[0114] ;

[0115] in, The number of sampling points is used, and the sampling duration is usually 1-2 seconds to balance response speed and statistical stability.

[0116] S233: Feature Combination

[0117] The root mean square value of the fluctuation of the pre-tension and the root mean square value of the fluctuation of the post-tension are combined to form the tension oscillation characteristic P:

[0118] ;

[0119] Therefore, this invention employs a two-dimensional vector approach to monitor independent fluctuations in tension before and after the rolling process, while using a scalar approach to comprehensively evaluate overall tension stability. Tension oscillation is one of the main causes of breakage in aluminum foil rolling. By monitoring the intensity of tension fluctuations before and after the rolling process, the source of tension instability can be identified (inlet tension fluctuations are usually related to the uncoiler, while outlet tension fluctuations are usually related to the coiler), providing a basis for targeted control.

[0120] S240: Construction of Multidimensional Physical Feature Space

[0121] Rolling force direction characteristics extracted from S210, S220, and S230 Thickness fluctuation spectrum characteristics and tension oscillation characteristics Construct a four-dimensional or five-dimensional physical feature space This space uses each feature as a coordinate axis, and the state at each rolling moment corresponds to a feature point in the space. The feature point is...

[0122] ;

[0123] In this invention, the multidimensional physical feature space unifies various physical quantities of the rolling process under the same mathematical framework, realizing a comprehensive state representation across physical domains and overcoming the limitations of traditional single-parameter monitoring.

[0124] S300: Construction and Monitoring of Anisotropic Safety Envelope

[0125] S310: Definition of Security Envelope Geometry

[0126] In multidimensional physical feature space An anisotropic safety envelope is constructed, which has a hyperellipsoidal shape. For the five-dimensional feature space, the mathematical expression of the safety envelope is:

[0127] ;

[0128] in, , The real and imaginary parts corresponding to the rolling force direction characteristics. Corresponding thickness fluctuation spectrum characteristics , Corresponding to the front and rear tension oscillation characteristics; to This represents the length of the semi-axis in each dimension.

[0129] S320: Calculation of the length of anisotropic half-axis

[0130] The core characteristic of the anisotropic safety envelope is that the lengths of each semi-axis are negatively correlated with their corresponding eigenvalues, and the rates of change of the lengths of different semi-axiss are independent of each other. The specific calculation is as follows:

[0131] S321: Calculation of the half-shaft length in the first dimension (rolling force direction)

[0132] ;

[0133] in, The first dimension is the reference semi-axis length. The first dimension is the sensitivity coefficient. The die length represents the characteristic of the rolling force direction. This negative exponential relationship indicates that when the rolling force distribution is non-uniform ( As the value increases, the safety boundary of that dimension shrinks accordingly, and the tolerance for abnormal rolling forces decreases.

[0134] S322: Calculation of the semi-axis length in the second dimension (thickness fluctuation spectrum)

[0135] ;

[0136] in, The second dimension is the reference semi-axis length. The second dimension is the sensitivity coefficient. These are the characteristic values ​​of the thickness fluctuation spectrum.

[0137] S323: Calculation of the semi-axis length in the third dimension (tension oscillation)

[0138] ;

[0139] ;

[0140] in, , The third dimension is the reference semi-axis length. , The third dimension is the sensitivity coefficient. These are the characteristic values ​​of the thickness fluctuation spectrum.

[0141] The negative exponential decay rate of each dimension is determined by the sensitivity coefficient. , , , Independent control was implemented, and these coefficients were obtained using Bayesian optimization methods to minimize the prediction error of strip breakage accidents. The optimization was performed based on historical strip breakage accident data. Specifically, a strip breakage probability prediction model was constructed with sensitivity coefficients as variables. The distances between feature points prior to historical strip breakage accidents and the safety envelope boundary were used as training samples. A Gaussian process was used as a surrogate model, and parameter search was guided by a data acquisition function. Iterative optimization was performed until the prediction error converged. The optimization objective was to ensure that the safety envelope included normal rolling conditions while making the envelope boundary as close as possible to the danger zone, thereby improving the sensitivity of the early warning system.

[0142] Furthermore, this invention enables the safety envelope to adaptively adjust its shape according to the risk characteristics of different physical quantities through anisotropic design. For example, when tension fluctuates violently, the boundary of the tension dimension automatically contracts, and the system will issue an early warning even if other parameters are normal, embodying the safety concept of the "weakest link effect".

[0143] S330: Recommended Process Parameters and Feature Point Selection

[0144] A multi-objective optimization model trained on historical rolling data outputs a set of candidate process parameters. These parameters are then projected onto the multi-dimensional physical feature space to form a set of candidate points. The multi-objective optimization model employs a constrained Pareto optimization framework. The optimization objectives include maximizing rolling speed, minimizing thickness deviation, and minimizing tension fluctuation. The constraint condition is that the feature points must be located within a safe envelope. By evaluating the distance of each candidate point from the safe envelope boundary, its deviation from the current state, and the degree of achievement of each optimization objective, with maximizing the safety margin as the primary objective, the feature point corresponding to the optimal process parameter is selected as the current feature point. .

[0145] By introducing a process parameter recommendation model, process optimization under safety constraints is achieved, pursuing the optimal quality or efficiency while ensuring safety.

[0146] S340: Boundary Approach Judgment

[0147] Before performing boundary approach judgment, coupling risk identification and security envelope boundary modulation are performed first. See S500 for specific steps.

[0148] S341: Calculation of module length

[0149] Calculate the current feature point Magnitude in multidimensional physical feature space :

[0150] ;

[0151] S342: Calculation of the modulus of the corresponding boundary points

[0152] Calculate the corresponding point of the current feature point on the boundary of the anisotropic safety envelope. Length of the module For the boundary of the hyperellipsoid, the corresponding point is located at a point traversed from the origin. At the intersection of the ray and the boundary, its modulus satisfies:

[0153] ;

[0154] S343: Determination of degree of closeness

[0155] Calculate the ratio of the magnitude of the current feature point to the magnitude of the boundary point. :

[0156] ;

[0157] When the ratio When a preset warning threshold is reached (usually set to 0.8-0.9), the current feature point is determined to be approaching the anisotropic safety envelope boundary, triggering the safety fault-tolerant control program. Furthermore, by uniformly measuring the safety margin in different dimensions through the modulus ratio, the complexity of distance calculation in multidimensional space is avoided, while preserving anisotropic features.

[0158] S400: Safety and fault-tolerant control execution

[0159] S410: Correction direction determined

[0160] S411: Shortest distance calculation

[0161] Calculate the current feature point in a multidimensional physical feature space The shortest distance to the boundary of the anisotropic safety envelope and its corresponding boundary point This problem can be transformed into a constrained optimization problem: finding a match between the bounds of a hyperellipsoid and the desired shape. The point with the smallest Euclidean distance.

[0162] S412: Gradient Direction Calculation

[0163] Calculate the gradient direction vector from the current feature point to the boundary point. :

[0164] ;

[0165] This direction points to the boundary of the safety envelope, which is the direction in which risk increases.

[0166] S413: Correction direction determined

[0167] The opposite direction of the gradient direction is determined as the correction direction. :

[0168] ;

[0169] This direction points inward into the safety envelope, which is the direction of reduced risk.

[0170] The determination of the correction direction is based on the geometric relationship between the current state and the boundary, and has a clear physical meaning: adjusting the process parameters along this direction can increase the safety margin at the fastest speed.

[0171] S420: Process Parameter Adjustment

[0172] S421: Type of parameter to be adjusted determined

[0173] The candidate process parameter set output by at least one process parameter recommendation model built based on expert rules or machine learning algorithms, according to the correction direction The components in each dimension determine the type of process parameter to be adjusted:

[0174] when exist , When the directional component is large, adjust the rolling force setting or roll tilt control;

[0175] when exist When the directional component is large, adjust the rolling speed (which affects the main frequency of thickness fluctuation).

[0176] when exist When the directional component is large, adjust the front tension;

[0177] when exist When the directional component is large, adjust the tension.

[0178] S422: Adjustment range calculation

[0179] Based on the distance between the current feature point and the boundary point Calculate the adjustment range :

[0180] ;

[0181] in, This is the proportionality coefficient. The integral coefficient is used, and a PI control strategy is employed to ensure rapid response and steady-state accuracy. The adjustment range is proportional to the risk components in each dimension.

[0182] S423: Smoothing Filtering Process

[0183] The calculated adjustment range is then adjusted according to the preset smoothing filter coefficient. Filtering is performed, and the smoothing filter coefficients are preset based on the dynamic response characteristics of the rolling mill.

[0184] ;

[0185] Among them, filtering avoids drastic fluctuations in control quantities and prevents system oscillations caused by lag in the response of the actuator.

[0186] S424: Execution Output

[0187] The filtered adjustment is output to the rolling mill actuator, including the hydraulic pressing system, tension control system and main drive system, to realize closed-loop adjustment of process parameters and return the current feature point to the anisotropic safety envelope.

[0188] Therefore, this invention achieves safe, stable, and rapid adjustment of process parameters through a multi-level control strategy of direction-amplitude-filtering, which not only ensures the restoration of safety margin but also avoids the impact of the control process on product quality.

[0189] S500: Coupling Risk Identification and Phase Compensation

[0190] S510: Cross-phase spectrum calculation

[0191] Calculate the cross-phase spectrum of the thickness fluctuation spectral characteristic T and the tension oscillation characteristic P. :

[0192] ;

[0193] in, The spectrum of thickness fluctuations. It is the conjugate of the tension oscillation spectrum. This indicates taking the phase angle.

[0194] S520: Identification of Dominant Disturbance Sources

[0195] Determine the phase lead relationship between thickness fluctuations and tension oscillations based on the cross-phase spectrum:

[0196] If the phase of thickness fluctuation leads the tension oscillation ( > 0), indicating that thickness fluctuation is the disturbance source and tension fluctuation is the response;

[0197] If the phase of tension oscillation leads the thickness fluctuation ( < 0) indicates that tension instability is the source of disturbance and thickness fluctuation is the response.

[0198] By analyzing the phase relationship of the main frequency band, the main sources of disturbance were identified as roll eccentricity, tension control instability, or material fluctuation.

[0199] S530: Calculation of Coupling Risk Coefficient

[0200] Calculate the spectral characteristics of thickness fluctuation and tension oscillation in the overlapping frequency range. The energy product integral within the range serves as the coupling risk coefficient. This coefficient reflects the energy coupling strength of the two waves in the common frequency band.

[0201] ;

[0202] S540: Calculation of Nonlinear Cooperative Risk Term

[0203] Calculate the nonlinear collaborative risk term :

[0204] ;

[0205] in, This is the proportionality coefficient. The characteristics of the thickness fluctuation spectrum. For the characteristic mode of tension oscillation, This represents the phase difference between the two components in the dominant frequency band. This term is positively correlated with the squares of thickness fluctuations and tension oscillations, as well as the cosine of their phase difference. When the two components are in phase ( The risk is greatest when ≈0), and vice versa. The risk is minimized when the angle is approximately 180°.

[0206] S550: Secure Envelope Dynamic Modulation

[0207] Coupling risk coefficient and nonlinear collaborative risk terms As an additional modulation signal for the anisotropic safety envelope. When or When the risk threshold is exceeded, the boundaries of the anisotropic safety envelope are automatically tightened, i.e., the length of each half-axis is reduced proportionally. To improve the security level.

[0208] Therefore, by identifying the coupling risk of thickness and tension, this invention achieves a leap from single-parameter monitoring to multi-parameter collaborative monitoring, and can provide early warning of "hidden risks" where a single parameter is normal but the synergistic effect leads to an accident.

[0209] S560: Phase Compensation Control

[0210] S561: Frequency Avoidance Control

[0211] When the coupling risk coefficient When the risk threshold is exceeded, the rolling speed is adjusted. The dominant frequency that changes the spectral characteristics of thickness fluctuation :

[0212] ;

[0213] in, The radius of the roll. Speed ​​adjustment is used to... Deviation from the natural frequency of the tension control system To avoid resonant coupling.

[0214] S562: Phase Inversion Control

[0215] When nonlinear collaborative risk term When the risk threshold is exceeded, adjust the pre-tension setting. By altering the phase response characteristics of the tension control system, the phase difference between thickness fluctuations and tension oscillations tends to 180 degrees, and anti-phase interference is used to cancel out the fluctuation energy.

[0216] S563: Targeted control of disturbance sources

[0217] When the main source of disturbance is identified as roll eccentricity, the roll eccentricity compensation algorithm is activated first, and the thickness fluctuation caused by eccentricity is offset by periodic fine adjustment of the roll speed.

[0218] When the dominant source of disturbance is identified as tension control instability, priority should be given to adjusting the tension closed-loop control parameters (such as PID parameters) to improve the stability of the tension system.

[0219] When the main source of disturbance is identified as material inflow fluctuation, the rolling speed is adjusted first to smooth out the impact of material thickness fluctuation through variable speed rolling.

[0220] By using phase compensation and disturbance source identification, a shift from passive protection to active control is achieved, enabling the elimination of risk factors at the source, rather than merely responding passively after risks accumulate.

[0221] The following specific implementation example illustrates the overall processing flow of the method of the present invention:

[0222] A power battery aluminum foil rolling production line has a target rolling thickness of 0.012mm and an incoming material thickness of 0.03mm.

[0223] Three-dimensional distribution data of rolling force are acquired at a frequency of 2kHz using four piezoelectric sensors on the roll bearing housing.

[0224] The X-ray thickness gauge and the grating displacement sensor are fused to obtain the exit thickness data;

[0225] Tension data were collected by front and rear tension gauges at a frequency of 1 kHz.

[0226] The calculated rolling force gradient vector magnitude is 15.3 N / mm, with θ = 25° and φ = 5° in the principal direction angles, forming the rolling force direction characteristics;

[0227] FFT analysis showed a significant peak in thickness fluctuation at 25 Hz. After frequency-sensitive weighting, the spectral characteristic of the thickness fluctuation was 0.8. ;

[0228] Pretension fluctuation 12N, post-tension fluctuation The value is 8N, forming a tension oscillation characteristic. =(12,8).

[0229] S300 security envelope construction and monitoring:

[0230] Construct a five-dimensional feature space and set the reference semi-axis length. =20, =1.0, = =15;

[0231] Based on sensitivity coefficient =0.05, =2.0, = =0.03, calculate the actual half-shaft length:

[0232] = = 9.4、 = 0.20、 = 10.5、 =11.7.

[0233] Current feature point =(13.9, 6.4, 0.8, 12, 8), calculate the modulus ratio. =0.85, exceeding the warning threshold of 0.8, triggering fault tolerance control.

[0234] S400 safety-tolerant fault control:

[0235] Calculate the boundary point corresponding to the shortest distance and determine the correction direction, which mainly points to the tension dimension. , (Directional component is the largest)

[0236] Determine the tension before and after adjustment, and calculate the adjustment range. =-3N, =-2N;

[0237] After smoothing and filtering, the output is sent to the tension control system. After 3 seconds, the feature point modulus ratio drops to 0.65, returning to the safe zone.

[0238] S500 Coupling Risk Identification:

[0239] Cross-phase spectrum analysis showed that the thickness fluctuations preceded the tension oscillations by 30°, indicating that the incoming material fluctuations were the dominant source.

[0240] The coupling risk coefficient is 0.6, which does not exceed the threshold, but the nonlinear collaborative risk term is 0.45, which is close to the threshold.

[0241] By fine-tuning the rolling speed to 780m / min, the main frequency of thickness fluctuation was reduced from 25Hz to 24.4Hz, deviating from the natural frequency of the tension system by 25Hz, thus eliminating the risk of resonance.

[0242] Through the above process, safety and fault tolerance control of the rolling process is achieved, timely intervention is provided when risk signs appear, potential strip breakage accidents are avoided, and production continuity and product quality stability are guaranteed.

[0243] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0244] 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, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A safety fault-tolerant control method for rolling aluminum foil for power batteries, characterized in that, Includes the following steps: Acquire rolling force distribution data, exit thickness data, and tension data during the rolling process; Based on the rolling force distribution data, rolling force direction features are extracted; based on the exit thickness data, thickness fluctuation spectrum features are extracted; based on the tension data, tension oscillation features are extracted. Based on the rolling force direction characteristics, thickness fluctuation spectrum characteristics, and tension oscillation characteristics, a multidimensional physical feature space is constructed. An anisotropic safety envelope is constructed in the multidimensional physical feature space. The anisotropic safety envelope has a superellipsoidal shape. The length of each half-axis is negatively correlated with the corresponding rolling force direction feature, thickness fluctuation spectrum feature and tension oscillation feature, and the rate of change of different half-axis lengths is independent of each other. The multi-objective optimization model trained based on historical rolling data outputs a set of candidate process parameters. The set of candidate process parameters is projected onto the multi-dimensional physical feature space to form a set of candidate points. The feature point corresponding to the optimal process parameter is selected from the set as the current feature point with the goal of maximizing the safety margin. Identify the coupling risk between the thickness fluctuation spectrum characteristics and the tension oscillation characteristics, and dynamically adjust the boundary of the anisotropic safety envelope based on the identification results; When the current feature point in the multidimensional physical feature space approaches the boundary of the anisotropic safety envelope, the correction direction is determined according to the positional relationship between the current feature point and the boundary. Adjust the rolling process parameters according to the correction direction so that the current feature point returns to the anisotropic safety envelope.

2. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The step of extracting the rolling force direction features includes: The three-dimensional distribution data of rolling force in the rolling direction is obtained by a multi-point sensor array installed on the roll bearing housing; Calculate the gradient vector of the rolling force in the three-dimensional distribution data; The modulus and principal direction angle are calculated based on the gradient vector, and the complex form of the modulus and the principal direction angle is combined as the rolling force direction feature.

3. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The step of extracting the thickness fluctuation spectral features includes: By integrating X-ray thickness gauge data and grating displacement sensor data, continuous fluctuation data of the exit thickness can be obtained; Spectral analysis was performed on the continuous fluctuation data to obtain the thickness fluctuation spectrum; The thickness fluctuation spectrum is weighted and integrated according to a preset frequency domain sensitive weighting function to obtain the thickness fluctuation spectrum characteristics. The frequency domain sensitive weighting function is preset based on the frequency domain statistical results of historical band breakage accidents.

4. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The step of extracting tension oscillation features includes: Acquire pre-tension and post-tension data at a sampling frequency of not less than 1 kHz; Calculate the root mean square value of the fluctuation of the front tension data relative to its mean, and the root mean square value of the fluctuation of the back tension data relative to its mean, respectively. The root mean square value of the fluctuation of the preceding tension data and the root mean square value of the fluctuation of the following tension data are combined to form the tension oscillation feature.

5. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The steps for constructing the anisotropic safety envelope include: The rolling force direction characteristics, the thickness fluctuation spectrum characteristics, and the tension oscillation characteristics are respectively used as the first dimension, the second dimension, and the third dimension of the superellipsoid safety envelope; Define the reference semi-axis lengths of the first dimension, the second dimension, and the third dimension; The actual half-axis length of the first dimension is calculated based on the die length of the rolling force direction feature, and the actual half-axis length of the first dimension has a negative exponential relationship with the die length of the rolling force direction feature. The actual half-axis length of the second dimension is calculated based on the numerical value of the thickness fluctuation spectrum feature. The actual half-axis length of the second dimension has a negative exponential relationship with the numerical value of the thickness fluctuation spectrum feature. The actual half-axis length of the third dimension is calculated based on the numerical value of the tension oscillation characteristic, and the actual half-axis length of the third dimension has a negative exponential relationship with the numerical value of the tension oscillation characteristic. The negative exponential decay rates of the first dimension, the second dimension, and the third dimension are independent of each other and are controlled by their respective sensitivity coefficients. The sensitivity coefficients are obtained by back-optimization based on historical belt breakage accident data, with the goal of minimizing the prediction error of belt breakage accidents.

6. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The determination steps for approaching the anisotropic safety envelope boundary include: Calculate the magnitude of the current feature point in the multidimensional physical feature space; Calculate the modulus of the corresponding point of the current feature point on the boundary of the anisotropic safety envelope; Calculate the ratio of the magnitude of the current feature point to the magnitude of the boundary point; When the ratio reaches a preset warning threshold, it is determined that the current feature point is approaching the boundary of the anisotropic safety envelope.

7. The safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The step of determining the correction direction includes: Calculate the shortest distance from the current feature point to the boundary of the anisotropic safety envelope in the multidimensional physical feature space; Determine the boundary point corresponding to the shortest distance; Calculate the gradient direction from the current feature point to the boundary point; The opposite direction of the gradient direction is determined as the correction direction.

8. A safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The steps for adjusting the rolling process parameters include: The type of process parameter to be adjusted is determined according to the correction direction, and the type of process parameter includes at least one of rolling speed, front tension, back tension, and rolling force setting value; The adjustment range is calculated based on the distance between the current feature point and the boundary point; The adjustment range is filtered according to a preset smoothing filter coefficient, which is preset based on the dynamic response characteristics of the rolling mill. The filtered adjustment value is output to the rolling mill actuator.

9. A safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 1, characterized in that, The steps for identifying coupling risks specifically include: Calculate the cross-phase spectrum of the thickness fluctuation spectral characteristics and the tension oscillation characteristics; The phase lead relationship between thickness fluctuation and tension oscillation is determined based on the cross-phase spectrum, and the dominant source of disturbance is identified. The integral of the energy product of the thickness fluctuation spectral characteristics and the tension oscillation characteristics in the overlapping frequency range is calculated as the coupling risk coefficient. Calculate the nonlinear cooperative risk term of the thickness fluctuation spectrum feature and the tension oscillation feature. The nonlinear cooperative risk term is positively correlated with the square of the thickness fluctuation spectrum feature, the square of the tension oscillation feature, and the cosine of the phase difference between the two. The coupling risk coefficient and the nonlinear cooperative risk term are used as additional modulation signals for the anisotropic safety envelope. When the coupling risk coefficient or the nonlinear cooperative risk term exceeds its corresponding risk threshold, the boundary of the anisotropic safety envelope is automatically tightened.

10. A safety fault-tolerant control method for rolling aluminum foil for power batteries according to claim 9, characterized in that, It also includes phase compensation control: When the coupling risk coefficient exceeds the risk threshold, the dominant frequency of the thickness fluctuation spectrum characteristic is changed by adjusting the rolling speed, causing it to deviate from the inherent frequency of the tension control system. When the nonlinear collaborative risk term exceeds the risk threshold, the phase of the tension oscillation characteristic is changed by adjusting the pre-tension setting, so that the phase difference between the thickness fluctuation and the tension oscillation tends to 180 degrees. When roll eccentricity is identified as the dominant source of disturbance, the roll eccentricity compensation algorithm is activated first. When tension control instability is identified as the dominant source of disturbance, the tension closed-loop control parameters should be adjusted first. When the main source of disturbance is identified as fluctuation in incoming material, the rolling speed should be adjusted first.

Images