Electroplating anode and cathode material degradation prediction and health management method

Abstract

The application discloses a method for predicting and managing degradation of anode and cathode materials in electroplating, and particularly relates to the field of data analysis of electroplating production lines, which is used to solve the problem that existing electroplating production lines rely on manual experience and single voltage threshold to arrange anode and cathode maintenance and are difficult to identify the degree of material degradation in time. The method is achieved by acquiring time-labeled data to construct an original working condition track, screening a slowly changing load interval, calculating a theoretical reference quantity and a deviation track, identifying a degradation sensitive interval, extracting a deviation monotonic cumulative feature and a deviation and load decoupling feature to form a comprehensive judgment coefficient, and then combining anode and cathode replacement records and plating layer quality records to adaptively update a comprehensive judgment coefficient threshold and a degradation sensitive interval division parameter, so as to realize continuous prediction and fine management of the degradation process of the electroplating anode and cathode.

Description

Technical Field

[0001] This invention relates to the field of electroplating production line data analysis, and more specifically, to a method for predicting the degradation and health management of electroplating anode and cathode materials. Background Technology

[0002] In electroplating production lines, cathodes and anodes are often made of insoluble materials, whose surface condition and coating integrity gradually degrade with long-term energization, electrochemical reactions, and changes in operating conditions. To monitor the anode's condition without shutting down the line, some literature proposes a method combining process data and physical models. First, the theoretical voltage of the rectifier when the anode is healthy is calculated based on process parameters. Then, the measured voltage is subtracted from the theoretical value, and the difference is converted into parasitic resistance. The magnitude and changes in parasitic resistance are used to determine whether the anode is aging or has an installation abnormality, thus assisting in deciding whether the anode needs replacement without disassembly and inspection. This approach has strong engineering practicality and has initially achieved anode condition identification based on process signals, providing important reference for predicting degradation and managing the health of electroplating anode and cathode materials.

[0003] However, this approach of directly treating voltage difference as a single health indicator remains significantly insufficient for predicting and managing the degradation of electroplated anode and cathode materials. In actual operation, parasitic resistance often reflects the cumulative changes of multiple factors, and its numerical change does not solely correspond to the degree of degradation of the anode material itself. The model's output typically only provides a rough judgment of whether the anode is abnormal, making it difficult to distinguish between different degradation stages or accurately reflect the complete decay process of a single cathode or anode material from initial application to eventual scrapping. For on-site maintenance personnel, they can only replace an entire set of anodes based on whether the parasitic resistance exceeds their experience range, without obtaining more in-depth information such as the degradation trend and remaining service life of individual electroplated anode and cathode materials. Therefore, it is difficult to achieve true prediction and management of anode and cathode material degradation.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for predicting and managing the degradation of electroplated anode and cathode materials. This method constructs an original operating condition trajectory by acquiring time-stamped data, filters out slow-changing load intervals, calculates theoretical reference values ​​and deviation trajectories, identifies degradation-sensitive intervals, and extracts monotonic cumulative deviation features and deviation-load decoupling features to form a comprehensive judgment coefficient. Then, by combining anode and cathode replacement records and plating quality records, the threshold of the comprehensive judgment coefficient and the parameters for dividing degradation-sensitive intervals are adaptively updated. This enables continuous prediction and refined management of the degradation process of electroplated anode and cathode, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] S1: Obtain the process operation data of the electroplating production line and organize the process operation data into the original working condition trajectory in chronological order.

[0008] S2: Select the slow load change interval in the original working condition trajectory, calculate the theoretical reference value corresponding to the slow load change interval, and obtain the deviation trajectory by subtracting the actual observation value from the theoretical reference value. Then, mark the degradation sensitive interval by comparing the deviation changes of adjacent slow load change intervals.

[0009] S3: For each degradation-sensitive interval, an analysis index is constructed based on the cumulative characteristics of the deviation over time and its correlation with load fluctuations. In the feature space, the normalized distance between the analysis index and the preset degradation target point is used to determine the comprehensive judgment coefficient used to characterize the degree of degradation of the anode and cathode.

[0010] S4: During subsequent operation of the electroplating production line, establish a correspondence between the comprehensive judgment coefficient of the degradation sensitive area and the cathode and anode maintenance records and coating quality records. Use the correspondence to train positive and negative samples and gradually correct the degradation identification parameters and decision boundaries.

[0011] Furthermore, step S1 includes the following:

[0012] The process operation data with time stamps is acquired and arranged in chronological order to form the original operating condition trajectory. The difference between adjacent time stamps is used to obtain the sampling time interval. The maximum value in the sampling time interval is calculated to construct the sampling interval ratio sequence. The sampling interval ratio threshold range is determined based on historical stable operation data. Sampling points whose sampling interval ratios fall within the threshold range are selected to form a valid sampling point index set. The original operating condition trajectory is reconstructed according to the valid sampling point index set.

[0013] Furthermore, step S2 includes the following:

[0014] The load change is obtained by calculating the difference between the load current values ​​of adjacent sampling points in the original operating condition trajectory. The absolute load change is obtained by taking the absolute value of the load change. The dimensionless load change ratio is constructed using the maximum value of the absolute load change. The dimensionless load change ratio threshold value is set based on the average value of the dimensionless load change ratio of the stable operation historical data. The slow load change interval is divided on the time axis by combining the minimum number of sampling points threshold and a set of slow load change intervals is formed.

[0015] Furthermore, step S2 also includes the following:

[0016] Within the set of slow-changing load intervals, the theoretical reference voltage value is obtained based on the load current value and the process variable vector through the theoretical voltage calculation relationship. The deviation value sequence is constructed using the observed voltage value and the theoretical reference voltage value. The representative value of the interval deviation is calculated according to the slow-changing load interval. The dimensionless adjacent interval deviation change ratio is constructed based on the difference of the representative value of the deviation of adjacent intervals. The slow-changing load intervals are merged according to the deviation change ratio threshold value to generate a set of degradation sensitive intervals.

[0017] Furthermore, step S3 includes the following:

[0018] Within each degradation-sensitive interval, a deviation sign sequence is constructed based on the deviation values. The ratio of the length of the longest continuous segment with the same non-zero statistical sign to the number of sampling points in the degradation-sensitive interval yields the deviation sign continuity ratio. The deviation increment is obtained based on the difference in deviation values ​​between adjacent sampling points, and the sum of the absolute values ​​of the deviation increments yields the total absolute deviation increment. The net cumulative deviation ratio is obtained based on the ratio of the absolute value of the difference in deviation values ​​between the first and last sampling points in the degradation-sensitive interval to the total absolute deviation increment. The monotonic cumulative deviation exponent is obtained by taking the square root of the product of the deviation sign continuity ratio and the net cumulative deviation ratio.

[0019] Furthermore, step S3 also includes the following:

[0020] Within each degradation-sensitive interval, load current increment and deviation increment are constructed based on load current value and deviation value. A directional consistency mark is obtained by combining the signs of load current increment and deviation increment. The directional consistency mark is averaged within the time step in which load current increment and deviation increment are not simultaneously zero to obtain the directional consistency average value. The deviation load decoupling index is calculated based on the absolute value of the directional consistency average value.

[0021] Furthermore, step S3 also includes the following:

[0022] The deviation monotonic cumulative index and the deviation load decoupling index are combined to form an eigenvector. The normalized distance between the eigenvector and the ideal degenerate target point is calculated in the eigenplane, and the comprehensive judgment coefficient is obtained by subtracting the normalized distance value from it.

[0023] Furthermore, step S4 includes the following:

[0024] In the subsequent operation of the electroplating production line, based on the end time and time window length parameters of each degradation sensitive interval, cathode replacement records, anode replacement records, and coating quality records are retrieved on the time axis. Degradation sensitive intervals with records within the time window are marked as positive samples, and degradation sensitive intervals without records within the time window are marked as negative samples, forming a positive sample index set and a negative sample index set.

[0025] Furthermore, step S4 also includes the following:

[0026] Extract the comprehensive judgment coefficient from the positive sample index set and sort it from smallest to largest. Calculate the sorting position based on the alarm quantile parameter. Take the comprehensive judgment coefficient at the corresponding position from the sorted sequence as the comprehensive judgment coefficient alarm threshold. In subsequent operation, when the updated comprehensive judgment coefficient is not less than the comprehensive judgment coefficient alarm threshold, the corresponding degradation sensitive interval will be used as the alarm object.

[0027] Furthermore, step S4 also includes the following:

[0028] The arithmetic mean of the deviation monotonic cumulative exponent and deviation load decoupling exponent in the positive samples is used as the coordinates of the degradation target point. The upper limit of distance normalization is determined based on the maximum Euclidean distance from all degradation sensitive intervals to the degradation target point. The threshold of deviation change ratio is selected based on the sorted distribution of the deviation change ratio of dimensionless adjacent intervals within the positive samples. In subsequent operations, the updated degradation target point, the upper limit of distance normalization, and the threshold of deviation change ratio are used to calculate and update the comprehensive judgment coefficient and divide the degradation sensitive intervals.

[0029] The technical effects and advantages of the present invention's method for predicting and managing the degradation of electroplated anode and cathode materials are as follows:

[0030] This invention standardizes process operation data into original operating condition trajectories in chronological order, constructs theoretical reference values ​​and deviation trajectories within time intervals where load changes are limited, and then filters out degradation-sensitive intervals. This concentrates degradation analysis on time periods where the impact of load fluctuations and process switching is reduced. It separates the current regulation, start-up and shutdown disturbances, process anomalies, and changes in the performance of anode and cathode materials in complex electroplating processes, fundamentally alleviating the problem of signal mixing when relying on direct determination of entire voltage and current curves in the background technology. This provides a more stable and clean time and data foundation for the extraction of anode and cathode degradation trends.

[0031] Within the degradation-sensitive range, this invention utilizes the monotonic accumulation of deviation over time and the decoupling characteristics between deviation and load changes to construct analytical indicators. In the feature space, a comprehensive judgment coefficient is formed by the distance to the degradation target point. This maps the slow degradation of anode and cathode materials, a process that is difficult to observe directly, into a single, continuous state quantity. This not only retains information on multiple factors such as voltage deviation, load fluctuation, and time evolution, but also avoids the limitations of making rigid judgments based solely on instantaneous voltage thresholds or simple statistics in the prior art. This ensures that degradation judgment still has a clear quantitative basis when facing complex operating conditions, which is beneficial for uniformly comparing and ranking the degree of degradation between different time periods and different electrolytic cells.

[0032] In the long-term operation dimension, this invention aligns the comprehensive judgment coefficient corresponding to the degradation sensitive interval with the cathode and anode replacement records and coating quality records on the time axis, constructs a positive and negative sample set, and dynamically corrects the merging threshold of the degradation sensitive interval, the alarm threshold of the comprehensive judgment coefficient, and the location of the degradation target point based on the sample distribution. This allows the judgment logic to be continuously adjusted according to production line maintenance behavior and quality results, solving the problems of fixed thresholds failing over time and the difficulty in migrating processes between different batches in the background technology. Thus, without relying on a single empirical parameter, it provides quantitative support that fits the actual operating characteristics for anode and cathode replacement timing decisions, preventive maintenance arrangements, and quality risk control. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the method for predicting and managing the degradation of electroplated anode and cathode materials according to the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Example 1: Figure 1 This invention provides a method for predicting the degradation and managing the health of electroplated anode and cathode materials, comprising:

[0036] S1: Obtain the process operation data of the electroplating production line and organize the process operation data into the original operating condition trajectory in chronological order.

[0037] S2: Select the slow load change interval in the original working condition trajectory, calculate the theoretical reference value corresponding to the slow load change interval, obtain the deviation trajectory by subtracting the actual observation value from the theoretical reference value, and mark the degradation sensitive interval by comparing the deviation changes of adjacent slow load change intervals.

[0038] S3: For each degradation-sensitive interval, an analysis index is constructed based on the cumulative characteristics of the deviation over time and its correlation with load fluctuations. Within the feature space, the normalized distance between the analysis index and the preset degradation target point is used to determine the comprehensive judgment coefficient used to characterize the degree of degradation of the anode and cathode.

[0039] S4: During subsequent operation of the electroplating production line, establish a correspondence between the comprehensive judgment coefficient of the degradation sensitive area and the cathode and anode maintenance records and coating quality records. Use the correspondence to train positive and negative samples and gradually correct the degradation identification parameters and decision boundaries.

[0040] In the technical scenario of determining the degree of degradation of electroplating cathodes and anodes, it is necessary to identify the slow load change intervals on the time axis, calculate the theoretical reference quantity and deviation trajectory within these intervals, and then further divide the degradation-sensitive intervals and generate a comprehensive judgment coefficient.

[0041] However, the process operation data directly collected from the electroplating production line often has uneven sampling times, with some areas having excessively long or short sampling intervals. If this data is not organized and filtered in the initial step, it is impossible to construct an original operating condition trajectory with a clear time sequence and reasonable sampling intervals. Subsequent time-based analyses will also lack clear physical meaning. The purpose of step S1 is to organize the scattered process operation data into a continuous, regular original operating condition trajectory that can be directly referenced by subsequent steps.

[0042] Detailed implementation of step S1:

[0043] 1-1 Collection and basic structure of process operation data.

[0044] During the operation of the electroplating production line, process operation data is recorded at multiple discrete sampling moments.

[0045] Each sampling moment is recorded with a timestamp to indicate the specific time point when the sampling occurred, and a process operation data vector is also recorded.

[0046] Each process operation data vector contains at least the current quantity used to represent the electroplating load, as well as several process variables required for subsequent calculations of theoretical reference quantities, such as voltage, temperature, electrolyte state, etc.

[0047] By adopting a unified data acquisition strategy, we ensure that each piece of process operation data contains a timestamp and a complete data vector, with a one-to-one correspondence between the two.

[0048] 1-2 Process operation data organized in chronological order.

[0049] All collected time stamps and corresponding process operation data are treated as an ordered pair and rearranged from morning to night according to the time stamps.

[0050] After the arrangement is completed, the time stamp sequence strictly increases from the first sampling point to the last sampling point, and the process operation data vector sequence corresponds one-to-one with the time stamp sequence, so that each time stamp corresponds to a unique process operation data vector.

[0051] After processing, an initial sequence of process operation data arranged in chronological order is obtained, providing a unified time sequence basis for subsequent time interval calculations.

[0052] 1-3 Construction of the ratio of adjacent sampling time interval to sampling interval.

[0053] On the initial sequence of process operation data with a fixed time order, the adjacent sampling time intervals starting from the second sampling point are calculated sequentially.

[0054] Each adjacent sampling time interval is obtained by subtracting the time stamp of the previous sampling time from the time stamp of the current sampling time, representing the time span between two sampling times.

[0055] After obtaining all adjacent sampling time intervals, the one with the largest value is selected, and this largest time interval is used as the reference time interval for subsequent dimensionless processing.

[0056] Subsequently, each adjacent sampling time interval is divided by the reference time interval to obtain a sequence of dimensionless sampling interval ratios.

[0057] Each sampling interval ratio represents the proportion of the corresponding sampling time interval to the largest time interval among all adjacent sampling time intervals. The value ranges from greater than zero to less than or equal to one, and serves as the basis for subsequent judgment on whether the sampling interval is within a reasonable range.

[0058] Determination of the threshold range for the sampling interval ratio 1-4.

[0059] To eliminate sampling points with abnormally large time intervals and sampling points with abnormally short time intervals when constructing the original working condition trajectory, it is necessary to predetermine the effective threshold range of the sampling interval ratio.

[0060] To determine the threshold range, a period of time in a stable production state is selected from historical operating data, and the sampling interval ratio sequence is calculated in the manner described above within this period.

[0061] Arrange this sampling interval ratio sequence in ascending order of value, count the frequency of occurrence in different value intervals, find the continuous value interval with the highest frequency, and take this continuous value interval as the main concentration interval of the sampling interval ratio under stable production conditions.

[0062] Within this main concentration interval, a value is selected as the effective lower threshold of the sampling interval ratio, and another value is selected as the effective upper threshold of the sampling interval ratio. The value of the lower threshold is greater than zero, the value of the upper threshold is less than one, and the value of the lower threshold is less than the upper threshold.

[0063] For example, a lower threshold and an upper threshold can be selected within the main concentration interval so that the interval between the two thresholds covers most of the values ​​in the main concentration interval, so that the sampling interval ratio falling into this interval does not correspond to either overly dense sampling or overly sparse sampling.

[0064] 1-5 Determination of valid sampling points and construction of original working condition trajectory.

[0065] After the sampling interval ratio threshold range is determined, each sampling interval ratio in the initial sequence of the current process operation data to be processed is judged.

[0066] For each second and subsequent sampling point, if the corresponding sampling interval ratio value is greater than or equal to the lower limit threshold of the sampling interval ratio, and not greater than the upper limit threshold of the sampling interval ratio, then the sampling point is marked as a valid sampling point.

[0067] Since the first sampling point has no preceding sampling point, the sampling time interval cannot be calculated. To ensure the integrity of the starting point of the time series, the first sampling point is directly marked as a valid sampling point.

[0068] Through this determination process, all time stamp indices marked as valid sampling points are organized into a set of valid sampling point indices.

[0069] Subsequently, in chronological order, the corresponding time stamps and process operation data vectors are sequentially retrieved from the set of valid sampling point indices to construct the original operating condition trajectory.

[0070] The original operating condition trajectory consists of a series of time markers and process operation data vectors. The time order is strictly increasing, and the sampling interval ratios corresponding to the sampling intervals between adjacent time markers all fall within the pre-set effective threshold range. This provides a time-structured and stable operating condition data foundation for subsequently identifying slow load change intervals and calculating theoretical reference quantities on the original operating condition trajectory.

[0071] Step S1 arranges the time-stamped process operation data in chronological order, calculates the time interval between adjacent sampling points, and filters them using the sampling interval ratio threshold extracted from the historical stable operation phase. Only valid sampling points whose time intervals meet the stability requirements are retained to reconstruct the original operating condition trajectory. Compared with the practice in the background technology of directly using the full amount of time-series data collected on site without distinguishing between the commissioning phase, the shutdown phase, and the abnormal sampling loss phase, this effectively eliminates sampling points with abnormally long or excessively dense time intervals, eliminates the time axis distortion introduced by manual operation, maintenance shutdown, and communication jitter in the time-series data, and makes the subsequent analysis of load, voltage, and process variables based on the trajectory of uniform time scale and continuous operating conditions. This improves the reliability and comparability of the degradation identification process from the source and reduces the interference of time dimension anomalies on the judgment results.

[0072] Step S1 has yielded the original operating condition trajectory. This trajectory consists of a series of sampling points arranged chronologically, each containing a timestamp and corresponding process operation data. Subsequent degradation analysis does not require processing across all time ranges; instead, it prioritizes the time ranges where load current variations are limited. Within these time ranges, the deviation between the observed voltage and the theoretical reference voltage is calculated, and the intervals where the deviation level undergoes significant temporal shifts are further identified. Step S2 involves dividing the original operating condition trajectory into slow-change load intervals based on load variations, and then using these intervals as a basis to identify degradation-sensitive intervals, providing a clear time interval input for the degradation trend analysis in Step S3.

[0073] 2-1 Original working condition trajectory and variable decomposition.

[0074] In the original operating condition trajectory, each sampling point includes a time stamp and a process operation data vector.

[0075] In each process operation data vector, three types of data are extracted:

[0076] The first category is the load current value, which is used to represent the magnitude of the current passing through the electroplating electrolytic cell at that sampling moment;

[0077] The second category is the observed voltage value, which is used to represent the magnitude of the voltage measured at the electrolytic cell terminal at that sampling time;

[0078] The third category is process variable vectors, which are used to represent multiple parameters related to the electroplating process, such as electrolyte temperature, electrolyte concentration, electrode spacing, and electrode geometric parameters.

[0079] This split allows each sampling point in the original operating condition trajectory to be described using time stamps, load current values, observed voltage values, and process variable vectors. All subsequent calculations are based on this unified data structure.

[0080] 2-2 Construction of the ratio of load change to dimensionless load change.

[0081] In the original operating condition trajectory, in order to describe the change in load current between adjacent sampling points, each pair of adjacent sampling points is processed sequentially starting from the second sampling point.

[0082] For each pair of adjacent sampling points, first calculate the load current value of the latter sampling point minus the load current value of the former sampling point to obtain the load change corresponding to this pair of sampling points. The load change can be positive, negative, or zero. A positive value indicates an increase in current, a negative value indicates a decrease in current, and a zero value indicates no change in current.

[0083] Take the absolute value of each load change to obtain a sequence of absolute load changes. Then, find the absolute load change with the largest value in this sequence and take this largest absolute load change as the maximum load change.

[0084] If at least one absolute load change is not zero, then the maximum load change is a positive number. In this case, each absolute load change is divided by the maximum load change to obtain a sequence of dimensionless load change ratios. The numerical range of the dimensionless load change ratio is between zero and one, inclusive; where a value of zero indicates that the load current does not change within the corresponding time interval, a value of one indicates that the absolute value of the load current change within that time interval is equal to the maximum absolute load change that has occurred in all time intervals, and a value between zero and one indicates that the absolute value of the load current change within that time interval is between zero and the maximum absolute load change.

[0085] If all absolute load changes are equal to zero, it means that the load current values ​​in the original operating condition trajectory are completely consistent at all sampling points. At this time, it can be considered that the original operating condition trajectory satisfies the condition of slow load change in the load dimension as a whole. The subsequent division of the slow load change interval can directly use the entire time range without having to calculate the dimensionless load change ratio again.

[0086] 2-3 Obtaining the threshold for the dimensionless load change ratio and dividing the slow load change interval.

[0087] When at least one absolute load change is not zero, a slow load change interval needs to be constructed based on the dimensionless load change ratio. Therefore, a dimensionless load change ratio threshold value is introduced. The dimensionless load change ratio threshold value on the number axis is greater than zero and less than one.

[0088] The process for obtaining the threshold value of the dimensionless load change ratio is as follows:

[0089] First, select a stable operating period from historical operation records for the electroplating production line where the process formula is stable, the control settings have not been significantly adjusted, and the cathode and anode have not been replaced. During this stable operating period, calculate the load change, absolute load change, and dimensionless load change ratio according to the steps in 2-2 to obtain a sample of dimensionless load change ratios.

[0090] Second, sum all the dimensionless load change ratio samples obtained during the aforementioned stable operating period, and then divide the sum by the number of samples to obtain an average value. This average value lies between zero and one. When there is at least one non-zero absolute load change during the stable operating period, this average value is strictly greater than zero and strictly less than one.

[0091] Third, this average value is directly used as the threshold value for the dimensionless load change ratio. This threshold value is calculated from load change samples under stable operating conditions and reflects the typical level of the dimensionless load change ratio under stable conditions.

[0092] After determining the threshold value for the dimensionless load change ratio, each dimensionless load change ratio is sequentially evaluated starting from the second sampling point in the original operating condition trajectory. For each time interval, if the corresponding dimensionless load change ratio value is not greater than the threshold value, then the time interval from the previous time mark to the current time mark is considered to meet the slow load change condition.

[0093] To combine the time intervals that meet the conditions into a continuous time interval, a minimum sampling point threshold needs to be set. The minimum sampling point threshold is an integer not less than two, determined as follows:

[0094] First, calculate the time interval between any two adjacent sampling points based on the sampling cycle of the electroplating production line;

[0095] Second, a lower limit for the time length is determined based on the typical duration of a single electrochemical action or a single process control action in the electroplating process.

[0096] Third, divide the lower limit of the time length by the sampling point time interval and round up to obtain the minimum sampling point threshold.

[0097] After obtaining the minimum sampling point threshold, all consecutive time intervals that satisfy the condition that the dimensionless load change ratio is no greater than the threshold value are concatenated on the time axis to form candidate time intervals, and the number of sampling points contained in each candidate time interval is counted. For each candidate time interval, if the number of sampling points it contains is not less than the minimum sampling point threshold, this candidate time interval is defined as a slow load change interval, and its start and end sampling point indices are recorded. By traversing the entire trajectory, several slow load change intervals can be obtained, forming a set of slow load change intervals.

[0098] If, in case 2-2, all absolute load changes are equal to zero, then the starting and ending sampling point numbers of the original working condition trajectory can be directly used as the start and end of a slow load change interval, without performing the above-mentioned screening process based on the dimensionless load change ratio threshold.

[0099] 2-4 Theoretical reference voltage values ​​and deviation sequence construction.

[0100] After the set of slow load change intervals is determined, the theoretical reference voltage value needs to be calculated at each sampling point on the original operating condition trajectory to represent the expected voltage level when the cathode and anode are in ideal design condition.

[0101] Therefore, a theoretical voltage calculation relationship is established based on the electroplating process formula, electrolytic cell geometry parameters, and the target performance of the cathode and anode during the design phase. This calculation relationship takes the load current value at the sampling point and multiple process parameters from the process variable vector as input, and outputs a theoretical reference voltage value. The theoretical voltage calculation relationship can be derived through a circuit equivalent model, or it can be established by collecting multiple sets of load current, process variable, and observed voltage samples when the cathode and anode are in good condition, and then fitting the parameters using analytical functions. The design requirements for the theoretical voltage calculation relationship are: when the load change meets the condition of slow load change, the process variables are within the allowable process range, and the cathode and anode do not undergo significant degradation, the absolute value of the difference between the theoretical reference voltage value calculated by this relationship and the observed voltage value at the corresponding sampling point falls within a pre-set error control range. The error control range is set by the process designer based on the allowable fluctuation range of the electroplating voltage; specific values ​​do not need to be given in the instruction manual.

[0102] During operation, for each sampling point in the original operating condition trajectory, the theoretical reference voltage value of that sampling point is obtained using the load current value and process variable vector at that sampling point, through the aforementioned theoretical voltage calculation relationship. Subsequently, the theoretical reference voltage value is subtracted from the observed voltage value of that sampling point to obtain the deviation value of that sampling point.

[0103] The deviation values ​​of all sampling points are arranged in chronological order to form a deviation sequence. A positive deviation value indicates that the observed voltage is higher than the theoretical reference voltage, a negative value indicates that the observed voltage is lower than the theoretical reference voltage, and a value of zero indicates that the two are equal. The deviation sequence provides a basis for subsequent statistical analysis of the deviation level in the slow load change range and for comparing the deviation changes between different slow ranges.

[0104] 2-5 The ratio of the representative value of the deviation in the slow-changing load interval to the deviation change in the dimensionless adjacent interval.

[0105] After obtaining the set of slow-changing load intervals and the deviation sequence, it is necessary to compare the differences in deviation levels between different slow-changing load intervals. To do this, a representative deviation value is calculated for each slow-changing load interval.

[0106] For a given interval within the set of slowly changing load intervals, list the deviation values ​​corresponding to all sampling points within that interval. Sum these deviation values, then divide the sum by the number of sampling points within that interval to obtain the average deviation for that interval. Use this average deviation as the representative deviation value for this slowly changing load interval. Following this method, a sequence of representative deviation values ​​for all slowly changing load intervals can be obtained.

[0107] After obtaining the representative deviation value sequence, the slow-changing load intervals are numbered sequentially according to the time of the starting sampling point. Starting from the second slow-changing load interval, the representative deviation value of the current interval is calculated by subtracting the representative deviation value of the previous interval, resulting in a series of deviation differences between adjacent intervals. The absolute value of each deviation difference between adjacent intervals is taken to form a sequence of absolute deviation changes between adjacent intervals.

[0108] In this sequence, the term with the largest value is identified and called the maximum absolute adjacent interval deviation change. If at least one absolute adjacent interval deviation change is not equal to zero, then each absolute adjacent interval deviation change is divided by the maximum absolute adjacent interval deviation change to obtain a dimensionless adjacent interval deviation change ratio sequence. The dimensionless adjacent interval deviation change ratio ranges between zero and one, including zero and one. A value of zero indicates that the representative deviation values ​​of two adjacent slow load change intervals are exactly the same. A value of one indicates that the absolute value of the difference between the representative deviation values ​​of two adjacent intervals is equal to the maximum absolute value of the difference among all adjacent intervals. A value between zero and one indicates that the absolute value of the difference is between zero and the maximum absolute value of the difference.

[0109] If the variation in deviation between all absolutely adjacent intervals is equal to zero, it means that the representative values ​​of the deviations in all slowly changing load intervals are exactly the same. In this case, the ratio of the variation in deviation between all dimensionless adjacent intervals can be uniformly set to zero to indicate that there is no change in the level of deviation between intervals.

[0110] 2-6 Construction of the threshold for the ratio of deviation changes between dimensionless adjacent intervals and the degradation-sensitive interval.

[0111] To identify degradation-sensitive intervals from a sequence of dimensionless deviation change ratios between adjacent intervals, a deviation change ratio threshold value needs to be introduced. The deviation change ratio threshold value takes a value between zero and one; it can be equal to one, but cannot be less than or equal to zero.

[0112] The process for obtaining the threshold value of the deviation change ratio is as follows:

[0113] First, select a time period from historical operation records that includes the degradation process of both the cathode and the anode, and record the time of cathode replacement, anode replacement, or time when the coating quality shows obvious abnormalities within this time period.

[0114] Second, during this period, the original working condition trajectory, the set of slow load change intervals and the deviation sequence are constructed in the manner described in step S2, and the corresponding dimensionless adjacent interval deviation change ratio sequence is calculated.

[0115] Third, based on maintenance or quality records, select some adjacent intervals that are close to the time of the degradation event from the adjacent intervals corresponding to the dimensionless adjacent interval deviation change ratio sequence as degradation-related adjacent intervals. For example, several pairs of adjacent intervals can be selected before and after each actual cathode or anode replacement time point, and these adjacent intervals can be included in the degradation-related adjacent interval set. The remaining adjacent intervals that are not selected are included in the non-degradation adjacent interval set.

[0116] Fourth, extract all corresponding dimensionless adjacent interval deviation change ratio samples from the set of degenerate-related adjacent intervals, sum all these samples, and then divide the sample size by the sum to calculate an average value. Since the dimensionless adjacent interval deviation change ratio samples corresponding to the degenerate-related adjacent intervals are themselves between zero and one, and at least one sample is not equal to zero, this average value is between zero and one and is not equal to zero. Use this average value as the deviation change ratio threshold value.

[0117] After determining the threshold value for the deviation change ratio, the sequence of dimensionless adjacent interval deviation change ratios is traversed from front to back according to the interval number. For each dimensionless adjacent interval deviation change ratio, if its value is not less than the deviation change ratio threshold value, the two intervals with slow load changes corresponding to that ratio are considered as interval pairs that need to be merged.

[0118] During the traversal, if the deviation change ratios of several consecutively numbered dimensionless adjacent intervals are not less than the deviation change ratio threshold, then the corresponding intervals with slow load changes are merged into a single degradation-sensitive interval in chronological order. The starting sampling point number of the merged degradation-sensitive interval is taken as the starting sampling point number of the earliest interval among these intervals, and the ending sampling point number is taken as the ending sampling point number of the latest interval among these intervals.

[0119] By traversing and merging the deviation change ratios of all dimensionless adjacent intervals, several degradation-sensitive intervals can be obtained on the time axis. Each degradation-sensitive interval corresponds to a time range and a set of sampling point numbers. The degradation-sensitive intervals are numbered sequentially by time to form a set of degradation-sensitive intervals. Step S3 constructs degradation trend analysis indicators and comprehensive judgment coefficients based on the deviation sequences and load information within this set of degradation-sensitive intervals.

[0120] Through the processing of each sub-step in step S2, based on the original operating condition trajectory formed in step S1, a set of slow load change intervals is first divided using the dimensionless load change ratio threshold and the minimum number of sampling points threshold calculated from the stable operation samples, so that the subsequent analysis focuses on the time range within which the load current change amplitude is limited; then, in each slow load change interval, the theoretical reference voltage value is obtained according to the theoretical voltage calculation relationship, and a deviation sequence is formed with the observed voltage value, and then the difference in deviation level between different slow intervals is described by the interval deviation representative value and the dimensionless adjacent interval deviation change ratio; finally, through the deviation change ratio threshold and the continuous interval merging rule, a set of slow intervals with significant deviation changes is constructed into a set of degradation sensitive intervals.

[0121] In this way, the original operating condition trajectory is further refined into a set of time segments that are stable in the load dimension but have significant changes in the deviation dimension. This provides a data and time range basis for step S3 to extract degradation trend features around these degradation-sensitive intervals and construct the logic for determining the degree of degradation of the cathode and anode.

[0122] Step S2, on the already cleaned original operating condition trajectory on the time axis, introduces a dimensionless load change ratio and a minimum sampling point threshold to divide the slow load change intervals. Within these intervals, a deviation sequence is constructed using theoretical voltage calculations. Then, the representative values ​​of the interval deviations and the deviation change ratios of adjacent intervals are used to screen and merge the degradation-sensitive intervals. Compared with the background technology, which directly uses a single-point voltage threshold or global average value for coarse comparison over the entire time period, this method no longer allows large load adjustments, start-stop shocks, and short-term process switching to dominate the judgment results. Instead, it limits the analysis scope to the time period when the load current change amplitude is controlled and the deviation level shows a significant jump between intervals. This ensures that the subsequent degradation characteristic index is calculated only on time slices that are closer to the slow degradation process of the anode and cathode, thereby improving the separation between degradation signals and process disturbances, reducing misjudgments caused by load condition changes, and enhancing the pertinence and stability of degradation prediction for electroplated anode and cathode materials.

[0123] Step S2 identifies a set of degradation-sensitive intervals on the original operating condition trajectory, and provides a time-stamped sequence, load current value sequence, and deviation value sequence for each interval. While the degradation-sensitive intervals have already been filtered based on load and overall deviation level changes, they are still time-series data, making direct comparison with actual cathode / anode replacement records and plating quality records inconvenient. Step S3 requires compressing the evolution of deviation over time and the relationship between deviation and load into a set of indices within each degradation-sensitive interval. Based on this, a comprehensive judgment coefficient is constructed for subsequent analysis against maintenance records.

[0124] 3-1 Degradation-sensitive regions and basic data review.

[0125] The set of degradation-sensitive intervals consists of several degradation-sensitive intervals, each represented by a continuous range of sampling point numbers, extending from a starting sampling point number to an ending sampling point number. In the original operating condition trajectory, each sampling point corresponds to a time stamp and a set of process operation data, including load current values, observed voltage values, and process variable vectors. Step S2 has already calculated the theoretical reference voltage value for each sampling point using the theoretical voltage calculation relationship, and obtained the deviation value by subtracting the theoretical reference voltage value from the observed voltage value. For any degradation-sensitive interval, the number of sampling points within this number range can be obtained by subtracting the starting sampling point number from the ending sampling point number and adding one, which serves as the number of sampling points for that degradation-sensitive interval. Step S3 uses only three basic quantities—the deviation value sequence, the load current value sequence, and the number of sampling points—to complete subsequent calculations within each degradation-sensitive interval.

[0126] 3-2 Construction of the monotonic cumulative index of deviation.

[0127] Within each degradation-sensitive interval, to describe the directional consistency and cumulative trend of the deviation values ​​over time, a deviation sign sequence needs to be constructed first. For each sampling point within the degradation-sensitive interval, if the deviation value is greater than zero, it is recorded as a positive sign in the deviation sign sequence; if the deviation value is less than zero, it is recorded as a negative sign; and if the deviation value is equal to zero, it is recorded as a zero-value sign. Then, starting from the initial sampling point of the degradation-sensitive interval and proceeding chronologically to the final sampling point, a continuous segment with the same sign is recorded when a non-zero sign is encountered. Subsequent non-zero signs with the same sign are included in the same segment. The segment ends when the sign changes to another non-zero sign or becomes zero, and the next segment is identified. After scanning the entire degradation-sensitive interval, several continuous segments with the same sign are obtained. The length of each segment is equal to the number of sampling points it contains. The segment with the largest value among these segments is selected as the longest continuous segment with the same sign in this degradation-sensitive interval. Dividing this longest continuous segment with the same sign by the number of sampling points in the degradation-sensitive interval yields the deviation sign continuity ratio. The value of the continuous proportion of the deviation sign is between zero and one. A value of zero indicates that there is no non-zero deviation value in the interval, a value of one indicates that all deviation values ​​in the interval are completely consistent in sign and there is no zero value mixed in, and the value is between zero and one in other cases.

[0128] To characterize the net accumulation of deviation values ​​within the same degradation-sensitive interval, a deviation increment sequence needs to be constructed. For each sampling point from the next sampling point after the starting sampling point to the ending sampling point, the deviation value of the current sampling point is calculated by subtracting the deviation value of the previous sampling point, yielding the deviation increment within the corresponding time interval. The absolute value of each deviation increment is taken, and the absolute values ​​of all deviation increments within the entire degradation-sensitive interval are summed to obtain the total absolute deviation increment. The total absolute deviation increment represents the magnitude of the total change in deviation along the time axis within the entire degradation-sensitive interval. Simultaneously, the deviation value at the end sampling point of the degradation-sensitive interval is subtracted from the deviation value at the starting sampling point to obtain the first-to-last deviation difference. The absolute value of this difference is then taken to obtain the absolute value of the first-to-last deviation difference, which describes the overall change in deviation from the start to the end of the interval. If the sum of absolute deviation increments is greater than zero, the net cumulative deviation ratio is obtained by dividing the absolute value of the difference between the first and last deviations by the sum of absolute deviation increments. If the sum of absolute deviation increments is equal to zero, it means that the deviation value within the interval does not change between each adjacent time step, and in this case, the net cumulative deviation ratio is directly set to zero. The value of the net cumulative deviation ratio is also between zero and one. When the deviation frequently changes in opposite directions within the interval, the sum of absolute deviation increments is significantly greater than the absolute value of the difference between the first and last deviations, and the net cumulative deviation ratio is close to zero. When the deviation accumulates continuously in one direction within the interval, the sum of absolute deviation increments is close to the absolute value of the difference between the first and last deviations, and the net cumulative deviation ratio is close to one.

[0129] The continuous proportion of deviation signs reflects the sustained consistency of deviations in sign, while the net cumulative proportion of deviations reflects the unidirectional cumulative characteristic of deviations in value. To comprehensively express these two aspects of information in a single index, a deviation monotonic cumulative index is defined. Specifically, the continuous proportion of deviation signs is multiplied by the net cumulative proportion of deviations, and the square root of the product is taken to obtain the deviation monotonic cumulative index. The deviation monotonic cumulative index ranges from zero to one. When both the continuous proportion of deviation signs and the net cumulative proportion of deviations are close to one, the deviation monotonic cumulative index is close to one, indicating that deviations within the degradation-sensitive interval maintain a consistent sign over a long period and accumulate continuously in one direction in value. When either proportion equals zero, the deviation monotonic cumulative index is zero, indicating that deviations within the degradation-sensitive interval do not exhibit a clear monotonic cumulative pattern.

[0130] 3-3 Deviation Load Decoupling Index Construction.

[0131] Within each degradation-sensitive interval, to analyze whether the deviation change is mainly driven by the load current change, it is necessary to statistically analyze the directional relationship between the load current increment and the deviation increment. First, within the degradation-sensitive interval, starting from the next sampling point after the initial sampling point to the end sampling point, for each sampling point, the load current value of the current sampling point is calculated by subtracting the load current value of the previous sampling point to obtain the load current increment for the corresponding time interval; the deviation increment is obtained by subtracting the deviation value of the previous sampling point from the deviation value of the current sampling point in the same way as described above.

[0132] Within each time interval, a load increment sign is constructed based on the sign of the load current increment: a positive sign is recorded when the load current increment is greater than zero, a negative sign when the load current increment is less than zero, and a zero sign when the load current increment is equal to zero. Similarly, a deviation increment sign is constructed based on the sign of the deviation increment: a positive sign is recorded when the deviation increment is greater than zero, a negative sign when the deviation increment is less than zero, and a zero sign when the deviation increment is equal to zero. Subsequently, within each time interval, the combination of load increment and deviation increment signs is compared. If both are positive or both are negative, it is recorded as one in the direction consistency marker sequence; if one is positive and the other is negative, it is recorded as negative one; if at least one is a zero sign, it is recorded as zero.

[0133] To eliminate the impact of time intervals where both load current increment and deviation increment are zero on decoupling analysis, an effective time step set is constructed within the degradation-sensitive region. The effective time step set consists of all time steps where the sum of the absolute values ​​of the load current increment and deviation increment is greater than zero. In other words, if at least one of the load current increment or deviation increment is not zero, the time interval is included in the effective time step set. The number of effective time steps equals the number of elements in this set. If the number of effective time steps is zero, it indicates that the load current and deviation values ​​within the degradation-sensitive region have not changed between all adjacent sampling points. In this case, the deviation change cannot be analyzed in terms of its directional relationship with the load change, and subsequent directional correlation statistics are treated as zero.

[0134] When the number of effective time steps is greater than zero, the directional consistency markers for each effective time step are summed, and the summation is divided by the number of effective time steps to obtain the average directional consistency value. The average directional consistency value ranges from -1 to 1. When the directional consistency markers for all effective time steps are 1, the average directional consistency value is 1, indicating that in the degradation-sensitive region, the load current increment and the deviation increment are in the same direction in each effective time step. When the directional consistency markers for all effective time steps are negative 1, the average directional consistency value is negative 1, indicating that the two are in completely opposite directions in each effective time step. When the number of cases in the same direction and opposite direction in the effective time steps is similar, the average directional consistency value is close to zero.

[0135] To construct an index reflecting the degree of decoupling from the average directional consistency, a deviation load decoupling index is defined. First, the absolute value of the average directional consistency is taken to obtain the degree of directional consistency. Then, this absolute value is subtracted from one to obtain the deviation load decoupling index. The deviation load decoupling index ranges from zero to one. A value of zero indicates that the direction of deviation increment within the degradation-sensitive region either completely follows the direction of load current increment or is always opposite to it, with deviation changes highly dependent on load changes. A value of one indicates that the relationship between deviation increment and load current increment in direction within the degradation-sensitive region is symmetrically distributed, with similar numbers of cases in the same and opposite directions. The deviation change is not dominated by load changes in direction and is more likely to reflect the changes in the state of the cathode and anode materials themselves.

[0136] 3-4 Construction of comprehensive judgment coefficients.

[0137] Within each degradation-sensitive region, the monotonic cumulative deviation index and the deviation-load decoupling index together constitute a two-dimensional characteristic quantity. The monotonic cumulative deviation index, as the horizontal component, and the deviation-load decoupling index, as the vertical component, can be represented as a point in a Cartesian coordinate system. Ideally, the degradation-sensitive region dominated by cathode and anode degradation should exhibit continuous unidirectional accumulation of deviation over time, and the directional relationship between deviation changes and load changes should not have a fixed dependence. Therefore, in the aforementioned coordinate system, the horizontal and vertical coordinates of the ideal degradation target point are both taken as one.

[0138] To convert the proximity between a feature point and the ideal degradation target point within any degradation-sensitive interval into a scalar, the Euclidean distance between them is first calculated. The Euclidean distance is calculated by summing the squares of the differences in the x-coordinates and y-coordinates on the plane, and then taking the square root of this sum. Since both the monotonic cumulative deviation exponent and the deviation load decoupling exponent vary between zero and one, the maximum distance occurs when both exponents are zero. The maximum distance is the distance between the point with both x and y coordinates equal to one and the origin, which is the square root of two. To obtain a dimensionless distance with a uniform range, the actual distance within a specific degradation-sensitive interval is divided by the maximum distance mentioned above to obtain the normalized distance. The normalized distance ranges from zero to one. A normalized distance of zero indicates that the feature point coincides with the ideal degradation target point, while a normalized distance of one indicates that the feature point is located at the origin.

[0139] The comprehensive judgment coefficient is defined using the complement of the normalized distance. Specifically, the comprehensive judgment coefficient is obtained by subtracting the normalized distance value from one. The comprehensive judgment coefficient ranges from zero to one. A value close to one indicates that both the deviation monotonic cumulative index and the deviation load decoupling index are close to one, and the degradation-sensitive region exhibits strong degradation characteristics simultaneously in terms of sign continuity, net accumulation, and decoupling from load changes, with obvious degradation signals for both cathode and anode materials. A value close to zero indicates that at least one of the deviation monotonic cumulative index and the deviation load decoupling index is close to zero, and the deviation change pattern within the degradation-sensitive region is more similar to load regulation or random disturbances, with limited material degradation information.

[0140] For all degradation-sensitive intervals in the degradation-sensitive interval set, the corresponding deviation monotonic cumulative index, deviation load decoupling index, and comprehensive judgment coefficient are calculated one by one according to the above method to form a comprehensive judgment coefficient sequence. This comprehensive judgment coefficient sequence will be compared and analyzed with cathode and anode maintenance records and coating quality inspection results in subsequent steps to correct the judgment parameters and optimize the degradation identification strategy.

[0141] Step S3 constructs a monotonic cumulative deviation index and a deviation load decoupling index within the degradation-sensitive range and converges them into a comprehensive judgment coefficient. Compared with the background technology that relies on instantaneous voltage or simple statistics combined with fixed empirical thresholds for coarse judgment, this step considers whether the deviation direction is consistently consistent and whether the deviation value accumulates in a single direction within the same time window. It also uses the relationship between deviation changes and load changes to identify voltage fluctuations caused by load adjustments. This creates a clear distinction between anode and cathode material degradation signals and process operation disturbances, reducing false alarms and missed alarms caused by single-point voltage deviations. This provides a continuous and unified quantitative basis for adjusting maintenance rhythm and threshold strategies based on the comprehensive judgment coefficient, improving the adaptability and stability of judgment results in the anode and cathode degradation prediction and management process when facing complex production conditions.

[0142] Step S3 constructs a deviation monotonic cumulative index and a deviation load decoupling index within each degradation-sensitive interval, and calculates a comprehensive judgment coefficient based on the distance between these two indices and the preset degradation target point. This coefficient characterizes the strength of the anode and cathode degradation signals within each degradation-sensitive interval. During long-term operation, the electroplating production line continuously accumulates cathode replacement records, anode replacement records, and coating quality anomaly records. These records have a one-to-one correspondence or proximity relationship with the degradation-sensitive intervals and their comprehensive judgment coefficients on the time axis. Step S4 addresses this correspondence by constructing labeled samples, updating the comprehensive judgment coefficient alarm threshold, correcting the normalized upper limit of the degradation target point and distance, and updating the dimensionless adjacent interval deviation change ratio threshold. This gradually brings the degradation identification logic closer to the actual degradation behavior of the production line.

[0143] 4-1 Annotated sample construction and time alignment.

[0144] During the continuous operation of the electroplating production line, the newly added process operation data is processed according to steps S1 to S3 to expand the original operating condition trajectory and obtain a new set of degradation-sensitive intervals. Each degradation-sensitive interval is determined by the starting sampling point number and the ending sampling point number, and has a time range from the starting time mark to the ending time mark. Step S3 has already calculated the deviation monotonic cumulative index, deviation load decoupling index, and initial comprehensive judgment coefficient for each degradation-sensitive interval.

[0145] Meanwhile, the actual replacement time of the cathode, the actual replacement time of the anode, and the time when serious defects occurred in the coating quality inspection were uniformly organized into an event sequence. Each event record contains a timestamp and an event type, used to identify whether the cathode was replaced, the anode was replaced, or a serious deviation in coating quality occurred.

[0146] To establish the link between degradation-sensitive intervals and events, a time window length parameter is introduced. This time window length parameter is a positive time length, selected during the electroplating production line design phase based on the typical time interval between the obvious accumulation of anode and cathode degradation signals and the time between operator replacement or quality inspection discovering quality abnormalities. This ensures the time window length covers this decision-making and response time.

[0147] For any degradation-sensitive interval in the set of degradation-sensitive intervals, first read the timestamp corresponding to the end sampling point and treat this timestamp as the end time of the degradation-sensitive interval. Check in the event sequence whether any event time falls within the time period from the end time of the degradation-sensitive interval to the end time plus the time window length. If at least one event record's timestamp falls within this time period, the degradation-sensitive interval is marked as a positive sample, indicating that the degradation signal within this interval has been confirmed by anode / cathode replacement or coating quality anomalies in subsequent time windows. If no event time falls within this time period, the degradation-sensitive interval is marked as a negative sample, indicating that there are no degradation-related maintenance or quality anomalies in the records within this observation time range.

[0148] Based on the above rules, positive and negative sample labels are assigned to all degradation-sensitive intervals, forming a positive sample index set and a negative sample index set. Each degradation-sensitive interval in the positive sample index set corresponds to a comprehensive judgment coefficient and a set of degradation behaviors that have been verified by maintenance records or quality records. Each degradation-sensitive interval in the negative sample index set corresponds to a comprehensive judgment coefficient and a segment of operation that was not identified as abnormal by maintenance records or quality records within the observation time window.

[0149] 4-2 Update the alarm threshold of the comprehensive judgment coefficient.

[0150] In step S3, the comprehensive judgment coefficient has already been assigned a value between zero and one for each degradation-sensitive interval, representing the strength of the degradation characteristic within that interval. Actual operation and maintenance requires an alarm threshold to divide the comprehensive judgment coefficient sequence into alarm and non-alarm intervals, ensuring that degradation-sensitive intervals with comprehensive judgment coefficients not lower than the alarm threshold are included in maintenance decision-making. The alarm threshold should not be set entirely based on experience but should be adaptively updated based on the distribution of comprehensive judgment coefficients from the positive sample set.

[0151] Therefore, the initial comprehensive judgment coefficients corresponding to the degradation sensitive intervals of all positive samples are extracted from the positive sample index set to form a comprehensive judgment coefficient set for positive samples. The comprehensive judgment coefficients in the set are sorted from smallest to largest to obtain an ordered sequence, where each item is between zero and one. The first item of the sequence is the minimum comprehensive judgment coefficient in the positive samples, and the last item is the maximum comprehensive judgment coefficient in the positive samples.

[0152] An alarm quantile parameter is introduced. This parameter is a real number greater than zero and less than one, used to specify the position of the alarm threshold within the ranking sequence of positive sample comprehensive judgment coefficients. When the alarm quantile parameter is close to one, the alarm threshold is near the end of the ranking sequence, and the corresponding comprehensive judgment coefficient is close to a relatively high level among positive samples. An alarm will only be triggered in the degradation sensitive range where the comprehensive judgment coefficient is in the high-value region of positive samples. When the alarm quantile parameter is close to zero, the alarm threshold is near the beginning of the ranking sequence, and the corresponding comprehensive judgment coefficient is close to a relatively low level among positive samples. More positive samples will trigger the alarm. The actual value of the alarm quantile parameter is determined by operations personnel based on the acceptable number of false alarms and the risk of missed alarms on the production line.

[0153] Based on the number of positive samples and the alarm quantile parameter, the product of the number of positive samples and the alarm quantile parameter is rounded up to obtain a sequence number. If this position exceeds the number of positive samples, the position is adjusted to the number of positive samples. Then, the comprehensive judgment coefficient value at the corresponding sequence number is taken from the sorted sequence of comprehensive judgment coefficients of positive samples and used as the new comprehensive judgment coefficient alarm threshold.

[0154] In subsequent operation, when a new set of degradation sensitive intervals and an updated comprehensive judgment coefficient are obtained after executing steps S2 and S3 for new operating condition data, as long as the updated comprehensive judgment coefficient of a certain degradation sensitive interval is not less than the alarm threshold, it is determined that this degradation sensitive interval has reached the alarm condition in the degradation signal dimension and enters the subsequent maintenance strategy formulation process.

[0155] 4-3 Calculation of Degraded Target Points and Updated Comprehensive Judgment Coefficients.

[0156] In the initial design phase, a fixed degradation target point was selected for the comprehensive judgment coefficient. The x-axis of the degradation target point represents the ideal value of the preset deviation monotonic cumulative exponent, and the y-axis represents the ideal value of the preset deviation load decoupling exponent. As actual operating data and maintenance records accumulate, the positive sample distribution may indicate that the concentrated location of the actual degradation behavior in the feature space deviates from the preset target point. Therefore, it is necessary to update the degradation target point using positive sample data and simultaneously correct the distance normalization upper limit of the comprehensive judgment coefficient.

[0157] In the positive sample index set, each degradation-sensitive interval already has a corresponding monotonic cumulative index of deviation and a deviation load decoupling index, which take values ​​between zero and one. First, the arithmetic mean of all deviation monotonic cumulative indices in the positive samples is calculated. Specifically, the deviation monotonic cumulative indices of all degradation-sensitive intervals in the positive samples are added together, and then the sum is divided by the number of positive samples, resulting in an average value between zero and one. This average value is used as the coordinate of the degradation target point in the direction of the deviation monotonic cumulative index. Second, the arithmetic mean of all deviation load decoupling indices in the positive samples is calculated, and the sum of all deviation load decoupling indices is divided by the number of positive samples, resulting in another average value between zero and one. This average value is used as the coordinate of the degradation target point in the direction of the deviation load decoupling index. If the average calculation slightly exceeds zero or one due to numerical error, it is truncated in the implementation using zero and one as the boundary, ensuring that the degradation target point always lies within the closed interval between zero and one.

[0158] After the degradation target point is updated, the distance between each feature point and the degradation target point is recalculated for all degradation-sensitive intervals. For a given degradation-sensitive interval, first, the difference between the coordinates of the monotonic cumulative index of the interval deviation and the coordinates of the monotonic cumulative index of the degradation target point deviation is calculated, and this difference is squared; then, the difference between the coordinates of the coordinates of the interval deviation load decoupling index and the coordinates of the degradation target point deviation load decoupling index is calculated, and this difference is squared; the two squared terms are added together, and the square root of the sum is taken to obtain the Euclidean distance between this degradation-sensitive interval and the degradation target point in the feature plane.

[0159] Within all degradation-sensitive intervals, find the maximum distance value and use this maximum distance as the upper limit for distance normalization. If the maximum distance value is greater than zero, for any degradation-sensitive interval, divide its Euclidean distance value by the upper limit for distance normalization to obtain a normalized distance value between zero and one. If the maximum distance value is equal to zero, it means that the coordinates of all degradation-sensitive intervals in the feature plane coincide with the degradation target point, and in this case, all normalized distances are uniformly set to zero.

[0160] The updated comprehensive judgment coefficient is given in the form of the complement of the normalized distance. For each degradation-sensitive interval, the normalized distance value of that interval is subtracted from one to obtain the new comprehensive judgment coefficient. The new comprehensive judgment coefficient is still in the range of zero to one. When the feature point of a degradation-sensitive interval is closer to the updated degradation target point, the normalized distance value of that interval is smaller, and the new comprehensive judgment coefficient value is close to one, indicating that the characteristics of that interval are more consistent with typical degradation behavior; when the feature point is far away from the degradation target point, the normalized distance value is close to one, and the new comprehensive judgment coefficient value is close to zero, indicating that the characteristics of that interval differ significantly from typical degradation behavior. The alarm threshold is used in conjunction with the new comprehensive judgment coefficient to filter alarms for degradation-sensitive intervals.

[0161] 4-4 Update the threshold for the ratio of deviation changes between dimensionless adjacent intervals.

[0162] In step S2, to differentiate the degree of change in the representative values ​​of deviations between slowly changing load intervals, a dimensionless ratio of deviation changes between adjacent intervals is introduced. A threshold value for this ratio is then used to determine which slowly changing load intervals need to be merged into the same degradation-sensitive interval. The initial threshold value for the dimensionless ratio of deviation changes between adjacent intervals is set based on historical data experience. As positive sample data accumulates, the threshold value can be updated using the distribution of the dimensionless ratio of deviation changes between adjacent intervals used for merging within the positive samples, making the merging rules more closely reflect the actual degradation process.

[0163] During step S2, the absolute value of the difference between the representative deviation values ​​of each pair of adjacent slow-changing load intervals is calculated and normalized to the absolute value of the global maximum difference, forming a dimensionless adjacent interval deviation change ratio between zero and one. When performing the slow-changing load interval merging operation, if the dimensionless adjacent interval deviation change ratio corresponding to a pair of adjacent slow intervals is not less than the deviation change ratio threshold, it is determined that these two slow intervals need to be merged into the same degradation-sensitive interval on the time axis.

[0164] In the positive sample index set, each positive sample degradation-sensitive interval is formed by merging several intervals with slow load changes, and is accompanied by a set of dimensionless adjacent interval deviation change ratios used for merging. All dimensionless adjacent interval deviation change ratios used for merging within the positive sample degradation-sensitive interval are summarized to form the positive sample ratio set. Each value in the positive sample ratio set is between zero and one, and the size of the set is the number of ratios in the set.

[0165] Arrange all values ​​in the positive sample ratio set in ascending order to obtain an ordered ratio sequence. The first term of the sequence is the dimensionless ratio of the deviation change between adjacent intervals where the deviation change within the positive sample is relatively weak, and the last term of the sequence is the dimensionless ratio of the deviation change between adjacent intervals where the deviation change within the positive sample is more significant.

[0166] A bias change ratio quantile parameter is introduced. This parameter is a real number greater than zero and less than one, used to specify the position for selecting a new bias change ratio threshold in the positive sample ratio sorting sequence. When the bias change ratio quantile parameter is close to one, the ratio value corresponding to the new threshold is close to the end of the positive sample ratio sequence. In this case, only adjacent slow intervals exhibiting strong bias transitions within the positive samples will be merged in subsequent runs, and the coverage of degradation-sensitive intervals on the time axis is concentrated in the region with the most significant bias change. When the bias change ratio quantile parameter is close to zero, the ratio value corresponding to the new threshold is close to the beginning of the positive sample ratio sequence. In subsequent runs, more adjacent slow intervals will meet the merging condition, and the degradation-sensitive intervals will have a longer span on the time axis.

[0167] Based on the product of the number of positive sample ratios and the quantile parameter of the deviation change ratio, a position number is obtained by rounding up. If the position number exceeds the number of positive sample ratios, the position number is adjusted to match the number of positive sample ratios. Subsequently, the dimensionless adjacent interval deviation change ratio values ​​at the corresponding position are extracted from the ordered ratio sequence as the new deviation change ratio threshold.

[0168] When step S2 is executed again on subsequent operating data, a new deviation change ratio threshold is used to merge the slow load change intervals, generating a new set of degradation sensitive intervals. Then, the updated comprehensive judgment coefficient and alarm threshold are calculated through the aforementioned processes in steps S3 and S4, thus forming a closed loop in which the interval marking logic and degradation judgment logic are iteratively corrected based on actual operating data.

[0169] Step S4 aligns the comprehensive judgment coefficient of the degradation sensitive area with the records of cathode replacement, anode replacement, and coating quality anomalies on the time axis. It uses these operational segments with real result feedback to construct positive and negative samples, and adaptively updates the alarm threshold of the comprehensive judgment coefficient, the degradation target point, the distance normalization upper limit, and the dimensionless adjacent interval deviation change ratio threshold accordingly. Compared with the background technology, which uses manual experience to set fixed thresholds once and does not change with production line aging and process adjustments, this allows the degradation sensitive area division rules and degradation judgment scale to be synchronously corrected with maintenance behavior and quality results. This makes the judgment logic gradually converge with the actual degradation rhythm of the specific electroplating production line, reduces false alarms and false negatives caused by threshold distortion after long-term operation, and improves the adaptability and stability of anode and cathode degradation prediction and maintenance decisions in different operating conditions.

[0170] Specifically, the above description is only a preferred embodiment of this application and is not intended to limit this application.

[0171] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0172] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for predicting the degradation and managing the health of electroplated anode and cathode materials, characterized in that, Including the following steps: S1: Obtain the process operation data of the electroplating production line and organize the process operation data into the original working condition trajectory in chronological order. S2: Select the slow load change interval in the original working condition trajectory, calculate the theoretical reference value corresponding to the slow load change interval, and obtain the deviation trajectory by subtracting the actual observation value from the theoretical reference value. Then, mark the degradation sensitive interval by comparing the deviation changes of adjacent slow load change intervals. Step S2 includes the following: within the set of slow load change intervals, the theoretical reference voltage value is obtained based on the load current value and the process variable vector through the theoretical voltage calculation relationship; a deviation value sequence is constructed using the observed voltage value and the theoretical reference voltage value; the representative value of the interval deviation is calculated according to the slow load change intervals; a dimensionless adjacent interval deviation change ratio is constructed based on the difference of the representative values ​​of the adjacent interval deviations; and the slow load change intervals are merged according to the deviation change ratio threshold value to generate a set of degradation sensitive intervals. S3: For each degradation-sensitive interval, an analysis index is constructed based on the cumulative characteristics of the deviation over time and its correlation with load fluctuations. In the feature space, the normalized distance between the analysis index and the preset degradation target point is used to determine the comprehensive judgment coefficient used to characterize the degree of degradation of the anode and cathode. Step S3 includes the following: Within each degradation-sensitive interval, a deviation symbol sequence is constructed based on the deviation values. The ratio of the longest continuous segment with the same non-zero symbol length to the number of sampling points in the degradation-sensitive interval yields the deviation symbol continuity ratio. The deviation increment is obtained based on the difference in deviation values ​​between adjacent sampling points, and the sum of the absolute values ​​of the deviation increments yields the total absolute deviation increment. The net deviation accumulation ratio is obtained based on the ratio of the absolute value of the difference in deviation values ​​between the first and last sampling points in the degradation-sensitive interval to the total absolute deviation increment. The square root of the product of the deviation symbol continuity ratio and the net deviation accumulation ratio yields the deviation monotonic accumulation index. Within each degradation-sensitive interval, a load current increment and a deviation increment are constructed based on the load current value and the deviation value. A directional consistency mark is obtained based on the combination of the load current increment sign and the deviation increment sign. The directional consistency mark is averaged within time steps where the load current increment and the deviation increment are not simultaneously zero to obtain the directional consistency average value. The deviation load decoupling index is calculated based on the absolute value of the directional consistency average value. The deviation monotonic accumulation index and the deviation load decoupling index are combined to form a feature vector. The normalized distance between the feature vector and the ideal degradation target point is calculated in the feature plane, and the comprehensive judgment coefficient is obtained by subtracting the normalized distance value from one. S4: During subsequent operation of the electroplating production line, establish a correspondence between the comprehensive judgment coefficient of the degradation sensitive area and the cathode and anode maintenance records and coating quality records. Use the correspondence to train positive and negative samples and gradually correct the degradation identification parameters and decision boundaries.

2. The method for predicting and managing the degradation of electroplated anode and cathode materials according to claim 1, characterized in that, Step S1 includes the following: The process operation data with time stamps is acquired and arranged in chronological order to form the original operating condition trajectory. The difference between adjacent time stamps is used to obtain the sampling time interval. The maximum value in the sampling time interval is calculated to construct the sampling interval ratio sequence. The sampling interval ratio threshold range is determined based on historical stable operation data. Sampling points whose sampling interval ratios fall within the threshold range are selected to form a valid sampling point index set. The original operating condition trajectory is reconstructed according to the valid sampling point index set.

3. The method for predicting and managing the degradation of electroplated anode and cathode materials according to claim 2, characterized in that, Step S2 also includes the following: The load change is obtained by calculating the difference between the load current values ​​of adjacent sampling points in the original operating condition trajectory. The absolute load change is obtained by taking the absolute value of the load change. The dimensionless load change ratio is constructed using the maximum value of the absolute load change. The dimensionless load change ratio threshold value is set based on the average value of the dimensionless load change ratio of the stable operation historical data. The slow load change interval is divided on the time axis by combining the minimum number of sampling points threshold and a set of slow load change intervals is formed.

4. The method for predicting and managing the degradation of electroplated anode and cathode materials according to claim 1, characterized in that, Step S4 includes the following: In the subsequent operation of the electroplating production line, based on the end time and time window length parameters of each degradation sensitive interval, cathode replacement records, anode replacement records, and coating quality records are retrieved on the time axis. Degradation sensitive intervals with records within the time window are marked as positive samples, and degradation sensitive intervals without records within the time window are marked as negative samples, forming a positive sample index set and a negative sample index set.

5. The method for predicting and managing the degradation of electroplated anode and cathode materials according to claim 4, characterized in that, Step S4 also includes the following: Extract the comprehensive judgment coefficient from the positive sample index set and sort it from smallest to largest. Calculate the sorting position based on the alarm quantile parameter. Take the comprehensive judgment coefficient at the corresponding position from the sorted sequence as the comprehensive judgment coefficient alarm threshold. In subsequent operation, when the updated comprehensive judgment coefficient is not less than the comprehensive judgment coefficient alarm threshold, the corresponding degradation sensitive interval will be used as the alarm object.

6. The method for predicting and managing the degradation of electroplated anode and cathode materials according to claim 5, characterized in that, Step S4 also includes the following: The arithmetic mean of the deviation monotonic cumulative exponent and deviation load decoupling exponent in the positive samples is used as the coordinates of the degradation target point. The upper limit of distance normalization is determined based on the maximum Euclidean distance from all degradation sensitive intervals to the degradation target point. The threshold of deviation change ratio is selected based on the sorted distribution of the deviation change ratio of dimensionless adjacent intervals within the positive samples. In subsequent operations, the updated degradation target point, the upper limit of distance normalization, and the threshold of deviation change ratio are used to calculate and update the comprehensive judgment coefficient and divide the degradation sensitive intervals.

Images