Vertical electroplating feeding control method and system

Abstract

The invention relates to the technical field of electroplating process control, and discloses a vertical electroplating feeding control method and system.The method comprises the steps that anode current distribution and electrolyte parameters are monitored in real time, and the feeding rate matching degree is judged; the plating thickness uniformity risk level is evaluated based on a fuzzy control model, and the cumulative effect intensity of temperature fluctuation on anode passivation is quantified through multivariable regression analysis and a dynamic feedback mechanism; judging whether to trigger multi-stage feeding strategy recombination according to the risk level and the cumulative effect intensity; the current density and the electrolyte flow velocity are decoupled, a material consumption compensation coefficient is calculated, an adjustment instruction is generated, and iterative optimization can be carried out according to the execution effect. The system comprises a state monitoring module, a risk assessment module, an effect analysis module, a strategy decision module, a parameter decoupling module and a compensation execution module. According to the invention, the feeding control precision and the process stability are improved, and the method is suitable for precise control of the vertical electroplating process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electroplating process control, in particular to a vertical electroplating feeding control method and system. BACKGROUND

[0002] In the industrial fields of electronic manufacturing, precision machinery, etc., vertical electroplating process is widely used in the metallization processing of products such as printed circuit boards (PCB) and semiconductor packaging due to its high production efficiency and good plating layer uniformity. However, with the continuous development of the trend of miniaturization and precision of electronic devices, the requirements for plating layer thickness uniformity, anode passivation control and material feeding accuracy in the vertical electroplating process are becoming increasingly stringent. The traditional feeding control method gradually exposes many deficiencies and cannot meet the current needs of high-precision electroplating process.

[0003] Traditional vertical electroplating feeding control is mostly based on fixed parameter strategy, lacking real-time perception and adaptive adjustment ability for dynamic changes in the electroplating tank. Specifically, the anode current distribution state in the electroplating process is influenced by multiple factors such as electrolyte flow rate, temperature, and metal ion concentration. However, the traditional method fails to fully consider the interaction between these parameters. For example, when the current density of different sections of the anode is unevenly distributed, if a uniform feeding rate is still used, it may lead to insufficient or excessive material supply in some areas, resulting in uneven plating layer thickness, and even local overplating or underplating defects.

[0004] At the same time, the influence of the spatio-temporal distribution characteristics of electrolyte parameters on anode passivation has not been effectively analyzed. Anode passivation is a common problem in electroplating process. When the electrolyte temperature gradient fluctuates, it will accelerate or slow down the formation and dissolution process of the passivation film on the anode surface, thereby affecting the continuous progress of the electroplating reaction. The traditional control method often adjusts the feeding only by threshold judgment of a single temperature parameter, which cannot quantify the cumulative effect strength of temperature fluctuation on anode passivation, resulting in a lack of scientific basis for adjusting the feeding strategy and difficulty in effectively inhibiting anode passivation.

[0005] In addition, in terms of the correlation between processing current density and electrolyte flow rate, the traditional method usually assumes a linear relationship between the two, ignoring the possible nonlinear coupling effect in actual working conditions. This simplification may result in large calculation errors of material consumption compensation coefficients, which cannot accurately reflect the real material demand of each anode section, thereby affecting the accuracy of feeding control and the quality stability of electroplated products.

[0006] With the development of intelligent manufacturing technology, the demand for intelligent and fine control of electroplating process is increasingly urgent. How to monitor the multi-dimensional parameters in the electroplating process in real time, establish a dynamic correlation model between the parameters, and accordingly realize the adaptive adjustment of the feeding strategy, has become a key technical problem to be solved in the current vertical electroplating field. There is a lack of a feeding control method in the prior art that can comprehensively consider the anode current distribution, the interaction of electrolyte parameters and the cumulative effect of temperature, which is difficult to ensure the uniformity of the electroplated layer and the stability of the process under complex working conditions, and therefore a more advanced and accurate vertical electroplating feeding control scheme is urgently needed. SUMMARY

[0007] The purpose of the present application is to provide a vertical electroplating feeding control method and system to solve the problems raised in the background art.

[0008] To achieve the above-mentioned purpose, the present application provides the following technical solution: a vertical electroplating feeding control method and system, the method comprising:

[0009] Real-time monitoring of the anode current distribution state and electrolyte parameters in the electroplating tank, and determining whether the matching degree of the current feeding rate and the electrolytic reaction is lower than the set threshold;

[0010] When the matching degree of the feeding rate and the electrolytic reaction is lower than the set threshold, the interactive relationship between the anode current density and the electrolyte flow rate is analyzed based on the fuzzy control model, and the potential risk level of the uniformity of the electroplated layer is evaluated;

[0011] The spatio-temporal distribution characteristics of the electrolyte temperature gradient are analyzed by combining multivariate regression analysis and dynamic feedback mechanism, and the cumulative effect strength of temperature fluctuation on anode passivation phenomenon is quantified;

[0012] According to the potential risk level of the uniformity of the electroplated layer and the cumulative effect strength of temperature fluctuation on anode passivation phenomenon, it is determined whether to trigger the dynamic reorganization of the multi-stage feeding strategy;

[0013] When it is determined to trigger the reorganization of the multi-stage feeding strategy, the current density and electrolyte flow rate of each anode section are decoupled in correlation, and the material consumption compensation coefficients of each section are calculated.

[0014] Preferably, the method real-time monitors the anode current distribution state and electrolyte parameters in the electroplating tank, and determines whether the matching degree of the current feeding rate and the electrolytic reaction is lower than the set threshold, specifically:

[0015] Collecting real-time current values of each anode section and current gradient difference values between adjacent sections;

[0016] Synchronously acquiring the spatial distribution data of electrolyte flow rate, temperature and metal ion concentration;

[0017] calculating a dynamic deviation index of the current feeding rate according to a matching deviation of the current gradient difference and the metal ion concentration distribution data;

[0018] when the dynamic deviation index exceeds a preset threshold, determining that the feeding rate and the electrolysis reaction matching degree is lower than a set threshold.

[0019] Preferably, the method analyzes the interaction between the anode current density and the electrolyte flow rate based on a fuzzy control model, and evaluates the potential risk level of the uniformity of the plating layer thickness, specifically:

[0020] The input variables of the fuzzy control model are defined as the current density change rate and the electrolyte flow rate fluctuation amplitude, and the output variable is the risk level membership degree;

[0021] The membership function of the input variable is set and the fuzzy rule base is established, and the rule form is: if the current density change rate is higher than the threshold and the flow rate fluctuation amplitude is in a specific interval, then the output is a high risk membership degree;

[0022] The potential risk level quantization value is obtained by defuzzification calculation, and the high risk, medium risk and low risk levels are divided according to the preset threshold.

[0023] Preferably, the method analyzes the spatio-temporal distribution characteristics of the electrolyte temperature gradient by combining multivariate regression analysis with a dynamic feedback mechanism, specifically:

[0024] A multivariate regression equation containing temperature, flow rate and ion concentration is constructed, and the contribution weight of the temperature gradient of each section to the anode passivation rate is calculated;

[0025] A dynamic feedback correction factor is established, and its expression is:

[0026]

[0027] where λ i is the correction factor, τ i is the temperature gradient change rate of the i-th section, ΔC i is the ion concentration difference between adjacent sections, and ε is a constant;

[0028] The coefficient matrix of the regression equation is iteratively updated according to the correction factor, and the cumulative effect strength of the temperature fluctuation is obtained.

[0029] Preferably, the method determines whether to trigger the dynamic reorganization of the multi-stage feeding strategy, specifically:

[0030] When the potential risk level is high risk or the cumulative effect strength exceeds a critical value, the global feeding strategy reorganization is triggered;

[0031] When the potential risk level is medium risk and the cumulative effect strength is in a warning interval, the local feeding parameter adjustment is triggered.

[0032] The rest of the cases maintain the current feeding strategy.

[0033] Preferably, the method decouples the current density and the electrolyte flow rate of each anode section, specifically:

[0034] Calculate the covariance matrix of the current density and the flow rate, and identify the nonlinear coupling strength between the two;

[0035] Use orthogonal decomposition to eliminate the cross-interference terms in the covariance matrix to obtain independent influence factors and;

[0036] According to the independent influence factors, build a material consumption compensation coefficient calculation model for each section.

[0037] Preferably, the method calculates the material consumption compensation coefficient of each section, specifically:

[0038] Define the compensation coefficient, and its expression is:

[0039]

[0040] Wherein, K i is the compensation coefficient of the i-th section, μ I is the average current density, σ v is the standard deviation of the electrolyte flow rate, and θ i is the temperature gradient correction amount.

[0041] According to the compensation coefficient, generate the feeding rate adjustment instruction of each anode section.

[0042] Preferably, the method further comprises:

[0043] Real-time monitoring of the current distribution uniformity after the execution of the compensation coefficient;

[0044] When the uniformity improvement rate does not meet the expectation, recalculate the rule weight of the fuzzy control model and iteratively optimize the compensation coefficient.

[0045] Preferably, the present application includes a vertical electroplating feeding control system for implementing the above-mentioned vertical electroplating feeding control method, the system includes a state monitoring module, a risk assessment module, an effect analysis module, a strategy decision module, a parameter decoupling module and a compensation execution module.

[0046] State monitoring module: Real-time monitoring of the anode current distribution state and electrolyte parameters in the electroplating tank;

[0047] Risk assessment module: Based on the fuzzy control model, assess the potential risk level of the uniformity of the electroplated layer thickness;

[0048] Effect analysis module: quantifying the cumulative effect intensity of temperature fluctuation through multivariate regression analysis and dynamic feedback mechanism;

[0049] Strategy decision module: determining whether to trigger the dynamic reorganization of multi-stage feeding strategy;

[0050] Parameter decoupling module: decoupling the correlation between current density and electrolyte flow rate and calculating material consumption compensation coefficient;

[0051] Compensation execution module: generating feeding rate adjustment instructions according to compensation coefficient and monitoring execution effect.

[0052] Preferably, the compensation execution module further comprises:

[0053] The adaptive optimization unit is configured to dynamically adjust the rule base parameters of the fuzzy control model and the compensation coefficient calculation threshold according to the current distribution uniformity data.

[0054] Compared with the prior art, the present application has the following advantages:

[0055] The real-time monitoring and dynamic matching mechanism ensures the precise adaptation of the feeding rate to the electrolysis reaction. By collecting real-time current values, current gradient difference values, and spatial distribution data of electrolyte flow rate, temperature, and metal ion concentration of each anode section, the dynamic deviation index of the current feeding rate can be accurately calculated. When it is found that the matching degree of the feeding rate to the electrolysis reaction is lower than the set threshold, the subsequent analysis and adjustment process is triggered in time to avoid problems in the plating layer quality caused by improper material supply. This dynamic matching mechanism based on real-time data breaks the limitations of traditional fixed parameter control, enabling the feeding control to respond to the working condition changes in the electroplating tank in real time, thereby reducing the process fluctuations caused by material supply and demand imbalance from the root.

[0056] The introduction of the fuzzy control model and multivariate regression analysis realizes scientific evaluation of the risk of plating layer thickness uniformity and the cumulative effect of anode passivation. The fuzzy control model defines the current density variation rate and electrolyte flow rate fluctuation amplitude as input variables, combines with the preset membership function and fuzzy rule base, and can accurately evaluate the potential risk level of plating layer thickness uniformity. The combination of multivariate regression analysis and dynamic feedback mechanism quantifies the cumulative effect intensity of temperature fluctuation on anode passivation phenomenon by constructing a regression equation containing temperature, flow rate, and ion concentration, and iteratively updating the coefficient matrix using a dynamic feedback correction factor. The application of these two technologies enables the system to deeply analyze the interaction of multiple parameters in the electroplating process, providing a scientific and quantitative basis for the adjustment of the feeding strategy, and changing the blindness of traditional methods relying on experience.

[0057] The dynamic reorganization mechanism of the multi-stage feeding strategy enhances the adaptability of the system to complex working conditions. According to the results of risk assessment and effect analysis, the system can intelligently determine whether to trigger global feeding strategy reorganization, local feeding parameter adjustment or maintain the current strategy. When the potential risk level is high risk or the cumulative effect intensity exceeds the critical value, global reorganization is triggered, and the feeding strategy of the entire electroplating tank is comprehensively adjusted; when it is in the medium risk and the cumulative effect intensity is in the warning interval, local parameter adjustment is carried out, realizing the optimal allocation of resources. This hierarchical processing method not only ensures the rapid response capability of the system under emergency working conditions, but also avoids the process fluctuation caused by excessive adjustment, significantly improving the stability and reliability of the electroplating process.

[0058] The correlation decoupling of current density and electrolyte flow rate and the precise calculation of material consumption compensation coefficient improve the fine degree of feeding control. By calculating the covariance matrix to identify the nonlinear coupling strength, using orthogonal decomposition method to eliminate the cross interference term, independent influence factor is obtained, and based on this, a compensation coefficient calculation model is established, which can accurately reflect the actual material demand of each anode section. The expression of compensation coefficient considers multiple factors such as average current density, electrolyte flow rate standard deviation and temperature gradient correction, making the feeding rate adjustment instruction more targeted. In addition, the real-time monitoring and iterative optimization mechanism after compensation execution can dynamically adjust the fuzzy control model rule library parameters and compensation coefficient calculation threshold according to the current distribution uniformity data, forming a closed-loop control, further improving the control accuracy and adaptability of the system.

[0059] The vertical electroplating feeding control system realizes intelligent control of the whole process from parameter monitoring, risk assessment, effect analysis to strategy decision, parameter decoupling and compensation execution through the cooperative work of each functional module. The state monitoring module obtains key data in real time, the risk assessment module and the effect analysis module provide analysis support, the strategy decision module formulates adjustment strategy, the parameter decoupling module and the compensation execution module realize precise control, and each link is closely connected, forming an efficient and intelligent feeding control system. Compared with traditional systems, the present application can significantly improve the uniformity of electroplated layer thickness, reduce the risk of anode passivation, reduce material waste, improve production efficiency and product quality, and has significant economic benefits and industrial application value. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 The working principle diagram of the vertical electroplating feeding control method and system described in the present application;

[0061] Figure 2 The design diagram of the electroplating tank state monitoring module;

[0062] Figure 3 The flowchart of electrolyte temperature gradient analysis;

[0063] Figure 4 Reorganization decision flowchart for multi-stage feeding strategy;

[0064] Figure 5 Flowchart for decoupling current density and flow rate correlation. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0066] Please refer to Figures 1-5 The vertical electroplating feeding control method and system according to the present application specifically implement the following steps:

[0067] The system collects the current values of each anode section and the current gradient difference between adjacent sections in real time through the sensor array distributed in the electroplating tank, synchronously obtains the spatial distribution data of the electrolyte flow rate, temperature and metal ion concentration. The current gradient difference and the metal ion concentration distribution data are matched and analyzed to calculate the dynamic deviation index of the current feeding rate. If the dynamic deviation index exceeds the preset threshold value, it is determined that the matching degree of the current feeding rate and the electrolytic reaction is lower than the set threshold value, and the subsequent adjustment mechanism needs to be started.

[0068] When the matching degree is lower than the threshold value, the system calls the fuzzy control model, takes the current density change rate and the electrolyte flow rate fluctuation amplitude as input variables, performs logical reasoning through the preset membership function (such as Gaussian function, trapezoidal function) and fuzzy rule base (such as "if the current density change rate is higher than the threshold value and the flow rate fluctuation amplitude is in the high interval, then output high-risk membership degree"), obtains the potential risk level quantization value through defuzzification processing (such as barycenter method), and compares it with the preset threshold value to divide the risk level into high, medium and low.

[0069] A multivariate regression equation containing temperature, flow rate and ion concentration is constructed, and the equation coefficient matrix is iteratively optimized through dynamic feedback correction factor. The correction factor expression is:

[0070]

[0071] Wherein, λ i is the correction factor of the i th section, τ i is the temperature gradient change rate, ΔC i is the ion concentration difference between adjacent sections, and ε is a constant. The model analyzes the spatiotemporal distribution characteristics of the electrolyte temperature gradient, and quantifies the cumulative effect intensity of temperature fluctuation on anode passivation phenomenon.

[0072] The system comprehensive risk level and cumulative effect intensity determine whether to trigger strategy reorganization: if the risk level is high risk or the cumulative effect intensity exceeds the critical value, trigger global feeding strategy reorganization; if the risk level is medium risk and the cumulative effect intensity is in the warning interval, trigger local feeding parameter adjustment; otherwise, maintain the current strategy.

[0073] When the strategy reorganization is triggered, first calculate the covariance matrix of the current density and the electrolyte flow rate of each anode section, identify the nonlinear coupling strength, use orthogonal decomposition method to eliminate the cross interference term, and obtain the independent influence factor. Based on the independent influence factor, build the material consumption compensation coefficient calculation model, and the compensation coefficient expression is:

[0074]

[0075] Where, K i is the compensation coefficient of the i-th section, μ I is the average current density, σ v is the standard deviation of the electrolyte flow rate, and θ i is the temperature gradient correction; according to the compensation coefficient, generate the feeding rate adjustment instruction of each section.

[0076] Example 1:

[0077] Real-time monitoring and matching degree judgment as a basic link, through multi-dimensional data acquisition and analysis to realize the initial evaluation of the feeding rate.

[0078] The system realizes data acquisition function through the sensor array integrated in the anode structure of the electroplating tank. The anode array is composed of multiple independent sections, and each section is equipped with a Hall current sensor. The sensor adopts closed-loop magnetic balance principle, the measurement range is 0-50A, the accuracy can reach ±0.5%FS, and the sampling frequency is set to 20Hz to ensure the high frequency update of real-time current value. The current gradient difference between adjacent sections is realized through the difference calculation module. The module synchronously samples the real-time current values of adjacent sections, and calculates the gradient value using the first-order difference algorithm, the formula is:

[0079]

[0080] Where, is the current gradient difference of the j-th adjacent section pair, I j and I j+1 are the real-time current values of the j-th and j+1-th sections, respectively.

[0081] The synchronous collection of electrolyte parameters is realized by multiple types of sensors distributed in different spatial positions of the tank. The ultrasonic flowmeter adopts the time difference method principle, and a group of transceiving probes is arranged every 10 cm in the horizontal direction of the tank, and a layer of monitoring planes is set every 5 cm in the vertical direction, and 4 probes are uniformly distributed in each layer to realize three-dimensional spatial scanning of the electrolyte flow rate. The flow rate measurement range is 0-0.5 m / s, and the accuracy is ±0.01 m / s. The thermocouple array is composed of T-type thermocouples and is distributed in a grid shape in the tank. The distance between adjacent thermocouples is 8 cm, the temperature measurement range is 0-100 DEG C, and the accuracy is ±0.5 DEG C. The metal ion concentration monitoring adopts an ion selective electrode, and a corresponding electrode is configured for the target metal ion (such as copper ion, nickel ion) in the electroplating solution. Each electrode corresponds to a detection unit, the detection range is 0-1000 ppm, the accuracy is ±5 ppm, and each detection unit is connected to the electroplating solution through a flow cell to realize dynamic concentration monitoring.

[0082] After data collection is completed, the matching deviation analysis link is entered. The system first performs spatial registration on the current gradient difference and metal ion concentration distribution data, and establishes a three-dimensional data mapping relationship based on the Cartesian coordinate system. Specifically, each anode section is corresponded to a specific coordinate region in the tank space, and the current gradient difference and the metal ion concentration data of the region and the adjacent region form a correlation data set. The matching deviation calculation adopts the Euclidean distance algorithm, and the formula is:

[0083]

[0084] Wherein, D is the matching deviation value, G k is the current gradient normalized value of the kth spatial position, C k is the metal ion concentration normalized value of the corresponding position, and n is the number of spatial points participating in the calculation. The normalization process adopts the Z-score standardization method, and the formula is:

[0085]

[0086] Wherein, X ′ is the normalized value, X is the original data, μ is the mean value of the data set, and σ is the standard deviation.

[0087] The calculation of the dynamic deviation index is based on the matching deviation value and the statistical characteristics of the historical data. The system pre-collects 100 groups of sample data under normal working conditions, calculates the matching deviation value of each group of data, and constructs a probability density distribution model. The dynamic deviation index is defined as the deviation multiple of the current matching deviation value and the mean value under normal working conditions, and the formula is:

[0088]

[0089] Wherein, E is the dynamic deviation index, μ0 is the normal operating condition matching deviation mean, σ0 is the normal operating condition matching deviation standard deviation. The preset threshold is determined by the 95% quantile of historical data, for example, when the statistics show that 95% of the matching deviation values under normal operating conditions are less than 0.6, the preset threshold is set to 0.8 (reserving a 20% safety margin).

[0090] When the dynamic deviation index E exceeds the preset threshold, the system determines that the feeding rate and the electrolysis reaction matching degree are insufficient, triggering the subsequent risk assessment process. At this time, the data acquisition frequency is automatically increased to 50Hz to enhance the monitoring resolution of abnormal state. At the same time, the system starts the data caching mechanism, and stores all the monitoring data of the past 10 minutes into the solid state disk, so as to facilitate subsequent trace analysis.

[0091] In terms of hardware implementation, the sensor array is connected with the data acquisition module through RS-485 bus, the bus uses shielded twisted pair to reduce electromagnetic interference, and the communication baud rate is set to 115200bps. The data acquisition module integrates a 24-bit A / D converter to perform high-precision digital conversion on the analog signal, and realizes real-time data preprocessing (such as digital filtering and outlier rejection) through FPGA. The preprocessed data stream is transmitted to the host computer through Ethernet, and the computer is equipped with an industrial-grade CPU (such as Intel Core i7-12700) and a real-time operating system (such as QNX) to ensure the timeliness and stability of data processing.

[0092] On the software level, the monitoring and analysis functions are realized through the control program developed by C++ language. The program adopts a multi-thread architecture, which is divided into data acquisition thread, data processing thread and logic judgment thread. The data acquisition thread is responsible for communication with the hardware device, and uses a ring buffer to store the raw data; the data processing thread performs real-time calculation on the buffer data, including gradient difference, normalization, deviation calculation, etc.; the logic judgment thread executes matching degree judgment according to the processing results, and generates corresponding control signals. The synchronization between threads is realized through the semaphore mechanism to ensure the determinacy of program running.

[0093] It should be noted that the matching deviation of current gradient difference and metal ion concentration distribution reflects the influence of feeding rate on the mass transfer of electrolysis reaction. When the feeding is insufficient, the metal ion concentration distribution may show the characteristics of low near the anode and high far from the anode, and the current gradient may increase due to the intensification of polarization phenomenon, and the matching deviation of the two will expand. Conversely, when the feeding is excessive, the uniformity of ion concentration distribution is improved, but it may cause abnormal electrolyte conductivity, leading to uneven current distribution, also showing the increase of matching deviation. Therefore, the dynamic deviation index can comprehensively reflect the dynamic matching state of feeding rate and electrolysis reaction.

[0094] In the system debugging stage, the layout of the sensor array needs to be optimized. Through finite element simulation analysis of the flow field and electric field distribution in the plating tank, the key positions of sensor arrangement are determined, such as increasing the sensor density at the anode corner, the electrolyte vortex area and other positions where parameter mutations are prone to occur. At the same time, the synchronization of the data acquisition module needs to be calibrated, and a high-precision clock chip (such as DS3231) is used to provide a unified time reference for each sensor, ensuring that the sampling time deviation of current and electrolyte parameters is less than 1 ms.

[0095] Example 2:

[0096] The risk assessment module realizes the quantitative assessment of the potential risk of plating layer thickness uniformity through a fuzzy control model.

[0097] The construction of the fuzzy control model is based on the nonlinear influence characteristics of current density and electrolyte flow rate on the uniformity of the plating layer in the plating process. The input variables of the model are set as the current density change rate and the electrolyte flow rate fluctuation amplitude, and the output variable is the risk level membership degree. Among them, the current density change rate reflects the dynamic fluctuation of current distribution per unit time, which is calculated by the ratio of the current density difference between adjacent two sampling periods and the time interval, with the unit of A / dm 2 ·s; the electrolyte flow rate fluctuation amplitude is defined as the absolute value of the deviation of the current flow rate value from the historical average value, accounting for the percentage of the historical average value, which is used to represent the stability of the flow field.

[0098] The fuzzification process of the input variables is realized through the membership function. The definition domain of the current density change rate is divided into three fuzzy subsets: "low", "medium" and "high". The membership function of the "low" grade adopts the left half trapezoidal function, with the membership degree being 1 when the change rate is ≤0.1 A / dm 2 ·s, linearly decreasing to 0 between 0.1 and 0.3 A / dm 2 ·s; the "medium" grade adopts a triangular function, with the peak value at 0.2 A / dm 2 ·s, forming a symmetrical distribution between 0.1 and 0.3 A / dm 2 ·s; the "high" grade adopts the right half trapezoidal function, with the membership degree being 1 when the change rate is ≥0.3 A / dm 2 ·s, linearly rising to 1 between 0.2 and 0.3 A / dm 2 ·s. The electrolyte flow rate fluctuation amplitude is divided into three subsets: "small", "medium" and "large", with the "small" subset corresponding to a fluctuation amplitude ≤5%, the membership function being a left half trapezoidal; the "medium" subset corresponding to 5%15%, adopting a triangular function with the peak value at 10%; the "large" subset corresponding to a fluctuation amplitude ≥15%, the membership function being a right half trapezoidal.

[0099] The establishment of the fuzzy rule base is based on the electroplating process expert experience and historical data statistical analysis. The rule adopts the conditional sentence form of "if A and B, then C", and contains 9 rules in total, as follows:

[0100] If the current density change rate is low and the flow velocity fluctuation amplitude is small, the risk level membership degree is low;

[0101] If the current density change rate is low and the flow velocity fluctuation amplitude is medium, the risk level membership degree is low;

[0102] If the current density change rate is low and the flow velocity fluctuation amplitude is large, the risk level membership degree is medium;

[0103] If the current density change rate is medium and the flow velocity fluctuation amplitude is small, the risk level membership degree is low;

[0104] If the current density change rate is medium and the flow velocity fluctuation amplitude is medium, the risk level membership degree is medium;

[0105] If the current density change rate is medium and the flow velocity fluctuation amplitude is large, the risk level membership degree is high;

[0106] If the current density change rate is high and the flow velocity fluctuation amplitude is small, the risk level membership degree is medium;

[0107] If the current density change rate is high and the flow velocity fluctuation amplitude is medium, the risk level membership degree is high;

[0108] If the current density change rate is high and the flow velocity fluctuation amplitude is large, the risk level membership degree is high.

[0109] The antecedent (input variable) of each rule determines the trigger strength through logical "and" operation (minimum value method), and the membership function of the consequent (output variable) is truncated according to the trigger strength of the antecedent. For example, when the current density change rate is 0.25 A / dm 2 ·s (belongs to the "medium" membership degree 0.5) and the flow velocity fluctuation amplitude is 12% (belongs to the "medium" membership degree 0.8), the trigger strength of rule 5 is 0.5, and the corresponding "medium risk" membership function is truncated to the trapezoidal region of 0-0.5.

[0110] The defuzzification process adopts the barycenter method to convert the fuzzy output into an accurate quantitative value. The specific steps are as follows: first, divide the domain of the output variable (risk level) into continuous intervals of 01, corresponding to "low risk" (00.3), "medium risk" (0.30.6), and "high risk" (0.61) three levels; then calculate the barycenter coordinates of the membership function of all activated rules and the figure surrounded by the domain interval, the formula is:

[0111]

[0112] wherein μ i is the triggering strength of the ith rule, z i is the midpoint value of the corresponding output level (e.g. 0.15 for “low risk”, 0.45 for “medium risk”, and 0.85 for “high risk”). The calculated Z value is the quantified value of the potential risk level, which is compared with the preset threshold to achieve the level division: when Z < 0.3, it is determined as low risk, 0.3≤Z<0.6 as medium risk, and Z≥0.6 as high risk.

[0113] In hardware implementation, the fuzzy control model is realized by an embedded controller (e.g. TITMS320F28379D DSP) for real-time operation. The controller integrates a fuzzy logic processing unit internally, adopts a lookup table method to optimize the rule matching efficiency, and pre-stores the fuzzification results of input variables, rule base and defuzzification parameters in the on-chip flash memory. Data interaction is connected with the host computer through a parallel bus, and the input variables are transmitted to the controller after data preprocessing by the host computer, and the output results are returned to the host computer for strategy decision-making.

[0114] In software design, the fuzzy control model adopts a modular programming architecture, including an input preprocessing module, a fuzzification module, a rule reasoning module and a defuzzification module. The input preprocessing module normalizes the current density change rate and flow velocity fluctuation amplitude, converting physical quantities into numerical values within the domain range (e.g. current density change rate 00.5A / dm 2 s is normalized to 01); the fuzzification module calculates the membership values of each fuzzy subset according to the membership function; the rule reasoning module realizes parallel matching of multiple rules through matrix operation to generate a fuzzy set of output variables; and the defuzzification module performs center of gravity calculation and outputs the quantized value.

[0115] The initialization of model parameters is completed through offline debugging. First, the current density and flow velocity data under different working conditions are generated by using electroplating process simulation software (e.g. CoventorWare), and a training data set is established in combination with the plating layer thickness uniformity detection results (e.g. thickness standard deviation measured by a scanning electron microscope). The shape parameters (e.g. coordinates of inflection points of trapezoidal function, peak position of triangular function) of the membership function are optimized by a particle swarm optimization algorithm (PSO) to minimize the error between the model output risk level and the actual detection result.

[0116] During real-time operation, the input data of the fuzzy control model is updated at the same frequency as the sensor sampling frequency (20 Hz), ensuring the timeliness of risk assessment. When the dynamic deviation index exceeds the threshold, the model automatically switches to a high-frequency operation mode, with an operation period reduced from 50 ms to 20 ms to respond to rapidly changing process conditions. At the same time, the system activates the model output monitoring mechanism, which triggers the parameter self-calibration program when the fluctuation amplitude of the risk level output value exceeds 20% for five consecutive sampling periods. The offset of the membership function is fine-tuned (adjustment step 0.05) to adapt to the slow changes in the composition of the electroplating solution over time.

[0117] It should be noted that the design of the fuzzy control model fully considers the uncertainty of the electroplating process. For example, the coupling effect of current density change rate and flow rate fluctuation amplitude may result in a single variable being unable to fully characterize the risk state, while the fuzzy rule base can more comprehensively reflect complex physical phenomena through multi-variable logical combination. In addition, the trapezoidal and triangular design of the membership function allows for some degree of overlapping intervals, avoiding abrupt changes in assessment results due to small fluctuations in data and enhancing the robustness of the model.

[0118] During system integration, the fuzzy control model and the state monitoring module are seamlessly connected through a shared data buffer. The current density change rate and flow rate fluctuation amplitude output by the state monitoring module are directly written to the buffer, and the fuzzy control model reads the data and performs operations at a fixed period, with the operation results stored in an independent output buffer for the strategy decision module to call. This design pattern reduces direct coupling between modules, improving the maintainability and scalability of the system.

[0119] Example 3:

[0120] The temperature effect analysis module quantitatively analyzes the anodic passivation phenomenon by constructing a multi-variable regression model. The system collects real-time electrolyte temperature data through a thermocouple array distributed in the electroplating tank. The thermocouple array adopts a grid-like layout, with temperature measurement points arranged at 10 cm intervals horizontally and 8 cm intervals vertically, forming a three-dimensional temperature monitoring network. Each temperature measurement point is equipped with a high-precision T-type thermocouple with a measurement range of 0-100°C and an accuracy of ±0.5°C, and the sampling frequency is set to 10 Hz to ensure that dynamic changes in the temperature field can be captured. The collected temperature data is transmitted to the data acquisition module through the RS-485 bus, and after analog-to-digital conversion, it is stored in a ring buffer with a capacity of 1 hour of continuous data, facilitating subsequent time series analysis.

[0121] In constructing the multivariate regression model, the anode passivation rate is taken as the dependent variable, and the electrolyte temperature, flow rate, and metal ion concentration are taken as the independent variables. Among them, the anode passivation rate is indirectly obtained by monitoring the potential change rate on the anode surface, measured by linear sweep voltammetry (LSV) at a scan rate of 5 mV / s in the potential range of -0.2 V to 0.8 V (relative to the reference electrode), with a scan period of 5 minutes. The electrolyte flow rate data come from an ultrasonic flow rate meter array distributed in the tank, and the metal ion concentration data are obtained by real-time monitoring with an ion-selective electrode.

[0122] The initial structure of the model is set to linear regression, and the initial coefficient matrix is obtained by least squares fitting. In the model training stage, the system collects 100 sets of sample data under different working conditions, each set of data containing the time series values of temperature, flow rate, ion concentration, and anode passivation rate within 30 consecutive minutes. The collected data are preprocessed, including outlier rejection (using the 3σ principle), moving average filtering (window size of 5 sampling points), and data standardization (mapping each variable to the 0-1 interval).

[0123] To dynamically adapt to the time-varying characteristics of the electroplating process, the system introduces a dynamic feedback correction mechanism. This mechanism calculates the temperature gradient change rate and ion concentration difference in adjacent time periods to generate a correction factor to adjust the coefficient matrix of the regression model in real time. Specifically, the temperature gradient change rate is defined as the difference between the temperature gradients of two adjacent sampling times divided by the time interval, and the ion concentration difference is the concentration difference between adjacent temperature measurement points. The calculation formula of the correction factor is:

[0124]

[0125] where λi is the correction factor corresponding to the ith temperature measurement point, τi is the temperature gradient change rate of the point (unit: °C / min), ΔC is the ion concentration difference between adjacent temperature measurement points (unit: ppm), and ε is a constant to prevent the denominator from being zero (value 0.1 ppm). i i i

[0126] In the model iteration optimization process, the system updates the coefficient matrix using the gradient descent method. After obtaining the new correction factor each time, the gradient of the objective function (the mean square error of the actual passivation rate and the model prediction value) with respect to the coefficient matrix is calculated, and then the coefficients are adjusted along the negative gradient direction. The adjustment step is dynamically determined according to the size of the correction factor (the larger the correction factor, the larger the step, but not more than 20% of the initial step). The iteration termination condition is set to a change rate of the objective function less than 0.1% or the number of iterations reaching 100 times.

[0127] ​​​The system realizes the analysis of the spatial and temporal distribution characteristics of the temperature gradient of the electrolyte by constructing a three-dimensional temperature field model. Using the collected discrete temperature data, the kriging interpolation method is used to generate a continuous temperature field distribution, with an interpolation accuracy of ±1°C. Based on the temperature field distribution, the temperature gradient vector of each point is calculated, and the modulus of the gradient vector represents the intensity of temperature change, and the direction represents the direction of temperature rise. By analyzing the spatiotemporal evolution law of the temperature gradient, the high-temperature area prone to anode passivation and its moving track are identified.

[0128] In terms of hardware implementation, the temperature data acquisition module uses a low-noise operational amplifier to condition the weak voltage signal output by the thermocouple, with a gain setting of 1000 times and a bandwidth limit of 0-5 Hz to suppress high-frequency noise. The analog-to-digital converter selects a 24-bit high-precision chip with a sampling rate of 100 Hz and an oversampling rate of 10 times to further improve the measurement accuracy through digital filtering. The data processing unit uses an ARM Cortex-A9 processor with 2GB DDR3 memory to ensure real-time processing of large amounts of temperature data.

[0129] In terms of software design, the temperature effect analysis module uses a multi-thread architecture. The data acquisition thread is responsible for communicating with the hardware device and storing the raw temperature data in the shared buffer; the data processing thread reads the data from the buffer, performs preprocessing, interpolation calculation, and gradient analysis; the model optimization thread updates the coefficient matrix of the regression model based on the processing results; the result output thread writes the analysis results to the database and sends them to the human-machine interface for display. The synchronization between threads is achieved through the semaphore mechanism to ensure the orderly processing of data.

[0130] The system sets up a model validity verification mechanism. After each update of the coefficient matrix, the prediction error of the model on the validation dataset is calculated. The validation dataset selects historical data from the last 24 hours, divided into 10 non-overlapping subsets in chronological order. Using the cross-validation method, each time 9 subsets are used for training and 1 subset is used for testing, and the average prediction error is calculated. When the average error exceeds the preset threshold (initially set to 5%), the model reset program is triggered, and the initial coefficient matrix is learned from the historical data.

[0131] During actual operation, the system quantitatively analyzes the cumulative effect intensity of temperature fluctuations on anode passivation based on the spatiotemporal distribution characteristics of the temperature gradient and the corrected regression model. The cumulative effect intensity is defined as the weighted product of the temperature gradient change rate and the anode passivation rate within a certain time period, and the weight coefficient is determined by the coefficient matrix of the regression model. Through real-time monitoring of the cumulative effect intensity, the system can predict the development trend of anode passivation in advance and provide a basis for adjusting the feeding strategy.

[0132] When the cumulative effect intensity of a certain region is detected to be continuously increasing, the system automatically adjusts the feeding strategy of the region. For example, the amount of additive near the region is increased to inhibit the development of anode passivation; or the electrolyte flow rate is adjusted to improve the mass transfer conditions of the region and reduce the temperature gradient. These adjustment measures are realized through a distributed control system, which can independently control the feeding devices and circulating pumps in each region of the electroplating tank, with a response time of less than 1 second.

[0133] Example 4:

[0134] The strategy decision module realizes dynamic adjustment of the feeding strategy based on the combined conditions of the risk level and the cumulative effect intensity.

[0135] The strategy decision module receives the risk level quantization value from the risk assessment module and the cumulative effect intensity value from the temperature effect analysis module, and generates corresponding control instructions through preset decision logic. The risk level quantization value is a continuous value between 0 and 1, corresponding to three risk level intervals of low, medium, and high; the cumulative effect intensity value represents the degree of influence of temperature fluctuations on anode passivation, with the unit of min -1 .

[0136] The system divides the combined space of risk level and cumulative effect intensity into four decision regions. When the risk level quantization value is <0.3 and the cumulative effect intensity is <0.05min -1 , it is determined to be a stable region, at which time the system maintains the current feeding strategy unchanged; when the risk level quantization value is between 0.3 and 0.6 and the cumulative effect intensity is between 0.05 and 0.1min *1 , it is determined to be a warning region, at which time the system triggers local feeding parameter adjustment; when the risk level quantization value is ≥0.6 or the cumulative effect intensity is ≥0.1min -1 , it is determined to be a high-risk region, at which time the system triggers global feeding strategy reorganization; when the risk level quantization value is <0.3 but the cumulative effect intensity is between 0.05 and 0.1min -1 , it is determined to be a special region, at which time the system executes preventive adjustment measures.

[0137] In the warning region, the system executes local feeding parameter adjustment. Specifically, for a specific anode section and its adjacent sections whose risk level quantization value and cumulative effect intensity both exceed the threshold, the system adjusts the feeding rate and additive concentration. The adjustment amplitude of the feeding rate is determined linearly according to the risk level quantization value, and the adjustment formula is:

[0138] ΔR=k·(R q -0.3)·R0

[0139] where ΔR is the feeding rate adjustment amount, R qFor risk level quantification, R0 is the current feeding rate, k is the proportional coefficient (value range 0.1-0.3, determined according to the type of electroplating process). The adjustment of additive concentration is related to the strength of cumulative effect, and the adjustment value is determined by table lookup method. The greater the cumulative effect strength, the more the additive concentration increases.

[0140] When entering the high-risk area, the system triggers a global feeding strategy reorganization. First, the historical data of all anode segments are analyzed retrospectively, extracting features such as current density distribution and electrolyte parameter change trend. Based on these features, the system calls genetic algorithm to optimize the feeding strategy. The optimization objective function is set as the comprehensive index of plating layer thickness uniformity and material consumption, where the plating layer thickness uniformity is evaluated by calculating the standard deviation of the thickness of each measurement point, and the material consumption is calculated by accumulating the feeding amount. The population size of genetic algorithm is set to 50, the iteration number is 20, the crossover probability is 0.8, and the mutation probability is 0.05.

[0141] In the special area, the system performs preventive adjustment measures. Although the risk level quantification value is low, the increase in cumulative effect strength indicates that potential risks are accumulating. At this time, the system fine-tunes the feeding parameters of the relevant area and increases the monitoring frequency. The fine-tuning amplitude of the feeding parameters is 30% of the normal adjustment amount, and the monitoring frequency is increased from the regular 20Hz to 50Hz to capture parameter change trends more timely.

[0142] The hardware implementation of the strategy decision module is based on an industrial programmable logic controller (PLC). The PLC adopts a modular design, including CPU module, digital input and output module, analog input and output module, and communication module. The CPU module selects a high-performance processor with a main frequency ≥1GHz and a memory ≥1GB to ensure fast processing of complex decision logic. The digital input and output module is used to control the start and stop of the feeding equipment, and the analog input and output module is used to accurately adjust the feeding rate and additive concentration. The communication module supports multiple industrial bus protocols, including Profinet, ModbusTCP, etc., to realize data interaction with other modules.

[0143] In software design, the strategy decision module adopts a state machine architecture. The system switches between different states according to the combination of risk level and cumulative effect strength, and each state corresponds to a different control strategy. The state switching logic is implemented through the ladder diagram programming of the PLC, ensuring clear logic and reliable execution. At the same time, the system sets a timeout protection mechanism, which automatically triggers a recovery program when a certain state duration exceeds a preset threshold (such as global strategy reorganization for more than 30 minutes), preventing the system from falling into a long-term unstable state.

[0144] The system establishes a decision log recording mechanism. Each time the strategy is adjusted, the adjustment time, adjustment reason, adjustment parameters and process state parameters before and after adjustment are recorded in detail. The log data is stored in SQLite database, and the storage period is 1 year. Through the analysis of the historical decision log, the system can summarize the experience and continuously optimize the decision logic. For example, the success rate of various decisions under different working conditions is counted, and the boundary values of the decision region are adjusted according to the statistical results.

[0145] During the execution of the strategy, the system adopts a closed-loop control mechanism. After the adjustment instruction is issued, the system monitors the changes of the relevant process parameters in real time, compares the actual changes with the expected changes, and if the actual changes deviate from the expected changes by more than 15%, the system automatically starts the compensation mechanism to make a second correction to the adjustment parameters. The amplitude of the second correction is dynamically determined according to the deviation, and a proportional-integral (PI) control algorithm is used to ensure that the system can quickly and stably reach the expected state.

[0146] The strategy decision module and other modules interact with each other through shared memory. The risk level quantization value, cumulative effect intensity value and other input data are written into the specified area of the shared memory by the corresponding module, and the strategy decision module reads these data regularly for decision calculation. The decision result is also written into the shared memory for the execution module to read and execute. This design reduces the direct dependence between modules and improves the scalability of the system.

[0147] To deal with sudden abnormal situations, the system sets up an emergency treatment plan. When the risk level quantization value suddenly rises above 0.5 or the cumulative effect intensity increases by more than 50% within 5 minutes, the system immediately triggers an emergency shutdown program, closes all feeding devices and sends out sound and light alarm signals. After the emergency shutdown, the system enters a safe mode and waits for manual intervention.

[0148] Example 5:

[0149] In the implementation process of the vertical electroplating feeding control method, the parameter decoupling and compensation calculation module realizes precise regulation and control of the feeding parameters by eliminating the coupling interference between variables. The following is the detailed implementation method of Example 5:

[0150] The system first analyzes the correlation of the historical data of current density and electrolyte flow rate. By arranging a high-density sensor array in the electroplating tank, the current density and flow rate data of a specific section are collected, with a sampling frequency of 100 Hz and a continuous collection time of more than 24 hours, forming a time series data set containing not less than 1000 data points. The collected data is preprocessed, including outlier rejection (using a method based on inter-quartile range), sliding average filtering (window size of 5 sampling points) and normalization.

[0151] In the covariance matrix calculation link, the system aligns the pre-processed current density data and flow rate data by time sequence, and constructs a two-dimensional data set. The covariance matrix of the data set is calculated, and the matrix element C ij represents the covariance of the i-th variable and the j-th variable. If the proportion of the sum of the absolute values of the non-diagonal elements of the matrix to the total sum of the matrix elements exceeds 30%, it is determined that there is a strong coupling relationship between the current density and the flow rate.

[0152] When it is determined that there is a strong coupling relationship, the system processes the covariance matrix using the QR orthogonal decomposition method. This method decomposes the covariance matrix into the product of an orthogonal matrix Q and an upper triangular matrix R, i.e. C = QR. The column vectors of the orthogonal matrix Q form a set of orthogonal bases in the data space, which can eliminate the correlation between the original variables. By extracting the column vectors of the Q matrix as independent influence factors, decoupling analysis of the current density and the flow rate is realized.

[0153] Based on the independent influence factors, the system constructs a material consumption compensation coefficient calculation model. This model considers the current density distribution characteristics, electrolyte flow state and temperature gradient correction amount, and generates a compensation coefficient for adjusting the feeding rate. The calculation formula of the compensation coefficient is:

[0154]

[0155] where K i is the compensation coefficient of the i-th section, μ I is the mean of the current density, σ v is the standard deviation of the electrolyte flow rate, and θ i is the temperature gradient correction factor.

[0156] The temperature gradient correction factor θ is obtained by a three-dimensional temperature field model. This model uses a thermocouple array distributed in the electroplating tank to collect temperature data in real time, and uses the finite difference method to solve the heat conduction equation to obtain the spatial distribution and gradient change of the temperature field. The temperature gradient correction factor is defined as the ratio of the temperature gradient amplitude of a specific region to the reference gradient amplitude, which reflects the influence of temperature distribution non-uniformity on material consumption.

[0157] The system generates feeding rate adjustment instructions according to the calculated compensation coefficient. The adjustment instructions are transmitted to the distributed control system through industrial Ethernet, which includes multiple programmable logic controllers (PLCs), each of which is responsible for controlling the feeding device in a specific region. The resolution of the adjustment instructions is 0.1 L / h, and the response time is less than 500 ms, ensuring that it can quickly respond to changes in process state.

[0158] After the adjustment of the feeding rate, the system monitors the current distribution balance in real time. Through the flexible current sensor array arranged on the anode surface, the current density values of each point are collected, and the standard deviation is calculated as the balance index. If the balance index changes after adjustment do not reach the expected threshold, the system will start the rule weight adjustment mechanism of the fuzzy control model. This mechanism optimizes the output results of the risk assessment model by adjusting the premise condition weight of the fuzzy control rule, thereby indirectly affecting the feeding strategy.

[0159] In terms of hardware implementation, the parameter decoupling and compensation calculation module uses a high-performance industrial computer as the core processing unit. This computer is equipped with an Intel Xeon E5-2620v4 processor, 32GB DDR4 memory, and 1TB SSD storage, ensuring fast processing of large amounts of data. The data acquisition card uses a 16-bit high-precision ADC with a sampling rate of up to 1MSPS, meeting the high-frequency data acquisition requirements. The communication interface supports multiple industrial protocols, including Profinet, Ethernet / IP, etc., enabling reliable connection with sensors and actuators.

[0160] In terms of software design, the module uses a multi-thread parallel processing architecture. The data acquisition thread is responsible for obtaining real-time data from the sensor, the data processing thread performs covariance matrix calculation, QR decomposition, and other mathematical operations, the control decision thread generates compensation coefficients and issues adjustment instructions, and the monitoring thread evaluates the adjustment effect in real time. The threads exchange data through shared memory to ensure processing efficiency.

[0161] The system has an adaptive learning mechanism that continuously optimizes the compensation coefficient calculation model using historical data. It regularly reviews and analyzes the process data from the previous 24 hours, compares the actual material consumption with the model prediction value, and adjusts the model parameters using the gradient descent method to gradually reduce the prediction error. The parameter adjustment step size is dynamically determined based on the error size to ensure that the model can quickly adapt to process changes and is not overly sensitive to produce oscillations.

[0162] In extreme conditions, such as a sudden increase in current density in a certain section exceeding 50%, the system will trigger an emergency compensation mode. At this time, the model automatically increases the weight of the temperature gradient correction factor and shortens the data acquisition period to 10ms to improve the control response speed. The adjustment instructions in the emergency compensation mode have higher priority than the regular control instructions to ensure safe and stable operation of the system.

[0163] To verify the decoupling effect, the system regularly performs disturbance tests. During normal production intervals, small step disturbances are applied to the electrolyte flow rate in a specific section, and the current density changes are monitored. By comparing the changes in correlation coefficients before and after the disturbance, the effectiveness of the decoupling algorithm is evaluated. If the correlation coefficient decreases by less than 20%, the model parameter self-tuning program is triggered to optimize the orthogonal decomposition process again.

[0164] The parameter decoupling and compensation calculation module keeps real-time data interaction with other control subsystems. Through the industrial field bus, the module sends the calculated compensation coefficients to the feeding control system, and receives real-time temperature field data from the temperature monitoring system. This closed-loop control structure ensures the coordinated optimization of the entire electroplating process.

[0165] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0166] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.

Claims

1. A vertical electroplating feed control method, characterized by, The method comprises the following steps: Real-time monitoring of anode current distribution state and electrolyte parameters in the electroplating tank, judging whether the matching degree of the current feeding rate and electrolysis reaction is lower than the set threshold value; When the matching degree of the feeding rate and electrolysis reaction is lower than the set threshold value, the interactive relationship between the anode current density and electrolyte flow rate is analyzed based on a fuzzy control model to evaluate the potential risk level of the thickness uniformity of the electroplated layer; The spatiotemporal distribution characteristics of the electrolyte temperature gradient are analyzed by combining multivariate regression analysis and a dynamic feedback mechanism to quantify the cumulative effect intensity of temperature fluctuation on anode passivation phenomenon; According to the potential risk level of the thickness uniformity of the electroplated layer and the cumulative effect intensity of temperature fluctuation on anode passivation phenomenon, it is determined whether to trigger dynamic restructuring of the multi-stage feeding strategy; When it is determined to trigger the restructuring of the multi-stage feeding strategy, the correlation decoupling of the current density and electrolyte flow rate of each anode section is performed, and the material consumption compensation coefficients of each section are calculated.

2. The vertical plating feed control method of claim 1, wherein, Real-time monitoring of anode current distribution state and electrolyte parameters in the electroplating tank, judging whether the matching degree of the current feeding rate and electrolysis reaction is lower than the set threshold value, specifically: Collecting real-time current values of each anode section and current gradient difference values between adjacent sections; Synchronously acquiring spatial distribution data of electrolyte flow rate, temperature and metal ion concentration; According to the matching deviation of the current gradient difference and the metal ion concentration distribution data, the dynamic deviation index of the current feeding rate is calculated; When the dynamic deviation index exceeds the preset threshold value, it is determined that the matching degree of the feeding rate and electrolysis reaction is lower than the set threshold value.

3. The vertical plating feed control method of claim 2, wherein, Based on the fuzzy control model, the interactive relationship between the anode current density and electrolyte flow rate is analyzed to evaluate the potential risk level of the thickness uniformity of the electroplated layer, specifically: The input variables of the fuzzy control model are defined as the current density change rate and the electrolyte flow rate fluctuation amplitude, and the output variable is the risk level membership degree; The membership function of the input variable is set and the fuzzy rule base is established, and the rule form is: if the current density change rate is higher than the threshold value and the flow rate fluctuation amplitude is in a specific interval, then the high risk membership degree is outputted; The potential risk level quantization value is obtained by defuzzification calculation, and the high risk, medium risk and low risk levels are divided according to the preset threshold value.

4. The vertical plating feed control method of claim 3, wherein, The spatiotemporal distribution characteristics of the electrolyte temperature gradient are analyzed by combining multivariate regression analysis and a dynamic feedback mechanism, specifically: A multivariate regression equation containing temperature, flow rate and ion concentration is constructed to calculate the contribution weight of temperature gradient of each section to the anode passivation rate; A dynamic feedback correction factor is established, and its expression is: where λ i is a correction factor, τ i is the rate of change of the temperature gradient of the i-th section, ΔC i is the difference in ion concentration between adjacent sections, and ε is a constant; The coefficient matrix of the regression equation is updated iteratively according to the correction factor to obtain the cumulative effect intensity of temperature fluctuation.

5. The vertical plating feed control method of claim 4, wherein, It is determined whether to trigger dynamic restructuring of the multi-stage feeding strategy, specifically: When the potential risk level is high risk or the cumulative effect intensity exceeds the critical value, the global feeding strategy restructuring is triggered; When the potential risk level is medium risk and the cumulative effect intensity is in the warning interval, local feeding parameter adjustment is triggered; In other cases, the current feeding strategy is maintained.

6. The vertical plating feed control method of claim 5, wherein, The correlation decoupling of the current density and electrolyte flow rate of each anode section is performed, specifically: The covariance matrix of the current density and flow rate is calculated to identify the nonlinear coupling strength between them; The orthogonal decomposition method is used to eliminate the cross interference terms in the covariance matrix, and independent influence factors and are obtained. A material consumption compensation coefficient calculation model for each section is constructed according to the independent influence factors.

7. The vertical plating feed control method of claim 6, wherein, The material consumption compensation coefficients of each section are calculated, specifically as follows: The compensation coefficient is defined, and its expression is as follows: wherein K i is the compensation coefficient of the i-th section, μ I is the mean of the current density, σ v is the standard deviation of the electrolyte flow rate, θ i is the temperature gradient correction amount; The feeding rate adjustment instruction of each anode section is generated according to the compensation coefficient.

8. The vertical plating feed control method of claim 7, wherein, After the feeding rate adjustment instruction is generated, the following steps are further included: The current distribution uniformity after the compensation coefficient is executed is monitored in real time. When the uniformity improvement rate does not meet the expectation, the rule weight of the fuzzy control model is recalculated, and the compensation coefficient is iteratively optimized.

9. A vertical electroplating dosing control system for implementing a vertical electroplating dosing control method according to any one of claims 1 to 8, characterized in that The state monitoring module, the risk assessment module, the effect analysis module, the strategy decision module, the parameter decoupling module and the compensation execution module are included. The state monitoring module: the anode current distribution state and the electrolyte parameters in the electroplating tank are monitored in real time. The risk assessment module: the potential risk level of the electroplating layer thickness uniformity is evaluated based on the fuzzy control model. The effect analysis module: the cumulative effect intensity of temperature fluctuation is quantified through multivariate regression analysis and dynamic feedback mechanism. The strategy decision module: whether to trigger the dynamic reorganization of the multi-level feeding strategy is determined. The parameter decoupling module: the correlation between the current density and the electrolyte flow rate is decoupled, and the material consumption compensation coefficient is calculated. The compensation execution module: the feeding rate adjustment instruction is generated according to the compensation coefficient, and the execution effect is monitored.

10. The vertical electroplating feed control system of claim 9, wherein, The compensation execution module further includes: The adaptive optimization unit is configured to dynamically adjust the rule base parameters of the fuzzy control model and the compensation coefficient calculation threshold according to the current distribution uniformity data.

Images