CN-121983169-A - Electroplating additive performance prediction and dynamic optimization method based on big data analysis

CN121983169ACN 121983169 ACN121983169 ACN 121983169ACN-121983169-A

Abstract

The invention discloses a method for predicting and dynamically optimizing performance of an electroplating additive based on big data analysis, in particular relates to the field of modeling of an electroplating process, and is used for solving the problem that the performance of the electroplating additive is difficult to accurately predict and stably control under detection lag and working condition fluctuation; the method comprises the steps of dividing a time axis of an electroplating production line by detection or dosing time points, constructing periodic profile features, extracting key process operation information, training an additive performance prediction model by using historical periodic profile features, additive performance characterization values and coating quality characterization values, inputting real-time periodic profile features into the model to obtain a basic performance prediction result and a process adjustment prediction result in a production operation stage, generating a control scheme by combining a prediction reliability index, a performance target interval and a control gear, and performing incremental update on the model based on subsequent detection feedback, so that online prediction and dynamic optimization control of the performance of the electroplating additive are realized.

Inventors

Nie Yiyi
LV MINGWEI
JIANG CHENG
YANG XIONG
ZHAO ZHISONG
Yuan Ciming

Assignees

武汉奥邦表面技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260318

Claims (9)

1. The electroplating additive performance prediction and dynamic optimization method based on big data analysis is characterized by comprising the following steps: S1, dividing a time axis of an electroplating production line into a plurality of control periods by using a detection or dosing time point, and carrying out normalization and difference processing on process operation data and detection results in time sequence in each control period to obtain a period profile characteristic sequence; S2, establishing association between a periodic profile characteristic sequence of each control period and an additive performance characteristic value and a coating quality characteristic value at the beginning of an adjacent control period, and constructing a training sample set which takes the periodic profile evolution as input and takes the additive performance change trend as output according to time sequence; S3, training an additive performance prediction model based on a training sample set, so that the model outputs a basic additive performance predicted value at a preset control moment and a process adjustment predicted value for compensating for the influence of detection hysteresis when receiving a cycle profile characteristic sequence of a current control cycle and a history control cycle; S4, inputting the cycle profile characteristic sequence of the current control cycle into a model to obtain a prediction result during production operation, determining the correction amplitude of the prediction result according to the time relation between the prediction result of the latest control cycle and the process response, generating a control scheme according to the corrected prediction result, executing the control scheme, and updating the model according to the execution result.
2. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 1, wherein the step S1 comprises the following steps: Dividing the time axis of the electroplating production line into a plurality of control periods according to the detection time point and the dosing time point, combining the control period with the time length lower than a preset lower limit threshold with the adjacent control period, and obtaining interpolation sampling values from the production load data sequence and the detection data sequence based on uniform relative time sampling points in each control period.
3. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 2, wherein step S1 further comprises the following steps: And setting a production load change range threshold and a detection change range threshold according to the historical noise range and the effective change range in each control period, performing amplitude normalization and adjacent difference on the production load sampling sequence and the detection sampling sequence, and combining four results at each relative time sampling point into a period profile characteristic sequence.
4. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 3, wherein the step S2 comprises the following steps: According to the obtained periodic profile characteristic sequence of each control period, calculating an additive performance characteristic value in a sequencing intermediate value mode in a detection time subinterval near the starting time of the corresponding control period, calculating a coating quality characteristic value in a quality detection time window corresponding to the starting time in a sequencing intermediate value mode, and establishing a one-to-one correspondence between the additive performance characteristic value and the coating quality characteristic value and the control period index.
5. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 4, wherein step S2 further comprises the following steps: and determining a value interval of the span period number according to the historical additive performance characterization value sequence, selecting the span period number according to the length of the performance monotone change interval, calculating a cross-period performance change trend value of each initial control period on the control period sequence according to the span period number, and forming a training sample set by the period profile feature sequence of a plurality of continuous control periods, the corresponding performance change trend value, and the starting point and end point plating quality characterization values.
6. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 5, wherein step S3 comprises the following steps: And constructing an additive performance prediction model comprising a convolutional coding map and a middle section aggregation structure by using a training sample set, coding a periodic profile feature sequence set of a continuous control period into a fixed-dimension intermediate representation vector, generating a cross-period additive performance variation trend prediction value by a performance trend output branch according to the intermediate representation vector, and generating a process adjustment prediction value by a process adjustment output branch according to the intermediate representation vector.
7. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 6, wherein step S3 further comprises the following steps: And respectively calculating the error between the performance change trend predicted value and the cross-period additive performance change trend value and the error between the process adjustment predicted value and the process adjustment reference quantity on each training sample, constructing two partial losses through a hyperbolic tangent function and an absolute value, taking one of the two partial losses with larger values as comprehensive loss in a single training sample, summing the comprehensive losses as integral loss on all the training samples, and updating all parameters of the additive performance prediction model through a gradient-based iterative optimization method.
8. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis according to claim 7, wherein step S4 comprises the following steps: And constructing a prediction window comprising a plurality of control periods with a span period number when each control period is finished, inputting a period profile characteristic sequence set in the prediction window into an additive performance prediction model, and calculating a reference performance value of the control period of a starting point of the prediction window and a predicted value of the basic additive performance of a target control period according to the additive performance characterization value obtained by last real detection and the predicted value of the historical single-period average performance change.
9. The method for predicting and dynamically optimizing performance of a plating additive based on big data analysis as recited in claim 8, wherein the step S4 further comprises the following steps: And constructing a prediction reliability index and a prediction reliability threshold based on an error ratio between the observed performance variation and a cross-period additive performance variation trend predicted value, combining a performance target interval determined by quality stable operation data and a control gear mapped by a process adjustment predicted value to generate a dosing and process parameter adjustment scheme of a target control period, and updating an error record and an additive performance prediction model parameter after acquiring a detection result of a coverage span period.

Description

Electroplating additive performance prediction and dynamic optimization method based on big data analysis Technical Field The invention relates to the field of electroplating process modeling, in particular to a method for predicting and dynamically optimizing performance of an electroplating additive based on big data analysis. Background In modern electroplating production, various organic additives are widely used for adjusting the brightness, leveling property and pore-filling capability of a plating layer, in order to reduce manual test boards and empirical dosing, an analysis device based on cyclic voltammetry analysis, electrochemical impedance analysis and microfluidic electrochemical detection is generally arranged on a production line, the content of different additives in the plating solution is calculated through steps of sampling, testing, fitting curves and the like, and then an automatic dosing system supplements the plating tank according to analysis results, so that an automatic operation mode integrating detection and control is formed. Particularly, in the occasions of continuous production, compact rhythm and complex product structure, electroplating enterprises highly depend on the automatic analysis and dosing system to maintain the relative stability of the plating solution state, and meanwhile, data such as process parameters, production beats, detection records, quality results and the like are continuously accumulated on site, so that the plating solution is used for daily traceability and process improvement, and a certain data base is laid for subsequent introduction of big data and machine learning means. In the operation mode combining the automatic detection and the automatic dosing, there is a key technical problem that the electrochemical analysis needs to go through a complete test and calculation process from sampling to giving a result, while the additives in the electroplating bath are continuously consumed along with the production in the process, the automatic dosing control is based on detecting the state of the plating bath at the moment, but the actual state in the plating bath has changed when the dosing is actually performed, and the detection result and the control object form a time dislocation. With the changes of production load, plate shape and process conditions, the dislocation can be expressed as the fluctuation of the concentration of the additive back and forth near the allowable range, the automatic dosing action and the coating performance change are difficult to synchronize in time, and the conditions that the brightness, pore filling capability or stress performance deviate obviously and the detection record still shows basically normal easily occur at certain stages. In the prior art, when improving such systems, most of the systems still stay on the level of improving single detection precision, shortening single detection time or adjusting the dosing threshold value, the state and the performance of the electroplating additive are predicted for future time without truly utilizing the process data and the quality data accumulated on a production line, and the dosing and the process parameters are optimized according to the state and the performance, so that a method for predicting and dynamically optimizing the performance of the electroplating additive based on big data analysis is not formed, and the stability and consistency problems caused by detection lag and control dislocation are difficult to be relieved fundamentally. In order to solve the above problems, a technical solution is now provided. Disclosure of Invention In order to overcome the defects in the prior art, the embodiment of the invention provides a method for predicting and dynamically optimizing the performance of an electroplating additive based on big data analysis, which is characterized in that a cycle profile is constructed by dividing an electroplating production line time axis by detecting or dosing time points, key process operation information is extracted, a model for predicting the performance of the additive is trained by utilizing historical cycle profile characteristics, additive performance characterization values and coating quality characterization values, a model for predicting the performance of the additive is input into the real-time cycle profile in a production operation stage to obtain a basic performance prediction result and a process adjustment prediction result, a control scheme is generated by combining a prediction credibility index, a performance target interval and a control gear, and the model is updated in an increment mode based on subsequent detection feedback, so that the problems in the background technology are solved. In order to achieve the above purpose, the present invention provides the following technical solutions: S1, dividing a time axis of an electroplating production line into a plurality of control periods by using a detection o